UbiquitousOS
/
OpenCloudOS-Kernel
mirror of https://gitee.com/OpenCloudOS/OpenCloudOS-Kernel.git
11
0
Fork 0
Code Issues Projects Releases Wiki Activity
OpenCloudOS-Kernel/fs/crypto/keyring.c

1211 lines
36 KiB
C
Raw Normal View History

fscrypt: add FS_IOC_ADD_ENCRYPTION_KEY ioctl Add a new fscrypt ioctl, FS_IOC_ADD_ENCRYPTION_KEY. This ioctl adds an encryption key to the filesystem's fscrypt keyring ->s_master_keys, making any files encrypted with that key appear "unlocked". Why we need this ~~~~~~~~~~~~~~~~ The main problem is that the "locked/unlocked" (ciphertext/plaintext) status of encrypted files is global, but the fscrypt keys are not. fscrypt only looks for keys in the keyring(s) the process accessing the filesystem is subscribed to: the thread keyring, process keyring, and session keyring, where the session keyring may contain the user keyring. Therefore, userspace has to put fscrypt keys in the keyrings for individual users or sessions. But this means that when a process with a different keyring tries to access encrypted files, whether they appear "unlocked" or not is nondeterministic. This is because it depends on whether the files are currently present in the inode cache. Fixing this by consistently providing each process its own view of the filesystem depending on whether it has the key or not isn't feasible due to how the VFS caches work. Furthermore, while sometimes users expect this behavior, it is misguided for two reasons. First, it would be an OS-level access control mechanism largely redundant with existing access control mechanisms such as UNIX file permissions, ACLs, LSMs, etc. Encryption is actually for protecting the data at rest. Second, almost all users of fscrypt actually do need the keys to be global. The largest users of fscrypt, Android and Chromium OS, achieve this by having PID 1 create a "session keyring" that is inherited by every process. This works, but it isn't scalable because it prevents session keyrings from being used for any other purpose. On general-purpose Linux distros, the 'fscrypt' userspace tool [1] can't similarly abuse the session keyring, so to make 'sudo' work on all systems it has to link all the user keyrings into root's user keyring [2]. This is ugly and raises security concerns. Moreover it can't make the keys available to system services, such as sshd trying to access the user's '~/.ssh' directory (see [3], [4]) or NetworkManager trying to read certificates from the user's home directory (see [5]); or to Docker containers (see [6], [7]). By having an API to add a key to the *filesystem* we'll be able to fix the above bugs, remove userspace workarounds, and clearly express the intended semantics: the locked/unlocked status of an encrypted directory is global, and encryption is orthogonal to OS-level access control. Why not use the add_key() syscall ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ We use an ioctl for this API rather than the existing add_key() system call because the ioctl gives us the flexibility needed to implement fscrypt-specific semantics that will be introduced in later patches: - Supporting key removal with the semantics such that the secret is removed immediately and any unused inodes using the key are evicted; also, the eviction of any in-use inodes can be retried. - Calculating a key-dependent cryptographic identifier and returning it to userspace. - Allowing keys to be added and removed by non-root users, but only keys for v2 encryption policies; and to prevent denial-of-service attacks, users can only remove keys they themselves have added, and a key is only really removed after all users who added it have removed it. Trying to shoehorn these semantics into the keyrings syscalls would be very difficult, whereas the ioctls make things much easier. However, to reuse code the implementation still uses the keyrings service internally. Thus we get lockless RCU-mode key lookups without having to re-implement it, and the keys automatically show up in /proc/keys for debugging purposes. References: [1] https://github.com/google/fscrypt [2] https://goo.gl/55cCrI#heading=h.vf09isp98isb [3] https://github.com/google/fscrypt/issues/111#issuecomment-444347939 [4] https://github.com/google/fscrypt/issues/116 [5] https://bugs.launchpad.net/ubuntu/+source/fscrypt/+bug/1770715 [6] https://github.com/google/fscrypt/issues/128 [7] https://askubuntu.com/questions/1130306/cannot-run-docker-on-an-encrypted-filesystem Reviewed-by: Theodore Ts'o <tytso@mit.edu> Signed-off-by: Eric Biggers <ebiggers@google.com>
2019-08-05 10:35:46 +08:00
// SPDX-License-Identifier: GPL-2.0
/*
* Filesystem-level keyring for fscrypt
*
* Copyright 2019 Google LLC
*/
/*
* This file implements management of fscrypt master keys in the
* filesystem-level keyring, including the ioctls:
*
* - FS_IOC_ADD_ENCRYPTION_KEY
fscrypt: add FS_IOC_REMOVE_ENCRYPTION_KEY ioctl Add a new fscrypt ioctl, FS_IOC_REMOVE_ENCRYPTION_KEY. This ioctl removes an encryption key that was added by FS_IOC_ADD_ENCRYPTION_KEY. It wipes the secret key itself, then "locks" the encrypted files and directories that had been unlocked using that key -- implemented by evicting the relevant dentries and inodes from the VFS caches. The problem this solves is that many fscrypt users want the ability to remove encryption keys, causing the corresponding encrypted directories to appear "locked" (presented in ciphertext form) again. Moreover, users want removing an encryption key to *really* remove it, in the sense that the removed keys cannot be recovered even if kernel memory is compromised, e.g. by the exploit of a kernel security vulnerability or by a physical attack. This is desirable after a user logs out of the system, for example. In many cases users even already assume this to be the case and are surprised to hear when it's not. It is not sufficient to simply unlink the master key from the keyring (or to revoke or invalidate it), since the actual encryption transform objects are still pinned in memory by their inodes. Therefore, to really remove a key we must also evict the relevant inodes. Currently one workaround is to run 'sync && echo 2 > /proc/sys/vm/drop_caches'. But, that evicts all unused inodes in the system rather than just the inodes associated with the key being removed, causing severe performance problems. Moreover, it requires root privileges, so regular users can't "lock" their encrypted files. Another workaround, used in Chromium OS kernels, is to add a new VFS-level ioctl FS_IOC_DROP_CACHE which is a more restricted version of drop_caches that operates on a single super_block. It does: shrink_dcache_sb(sb); invalidate_inodes(sb, false); But it's still a hack. Yet, the major users of filesystem encryption want this feature badly enough that they are actually using these hacks. To properly solve the problem, start maintaining a list of the inodes which have been "unlocked" using each master key. Originally this wasn't possible because the kernel didn't keep track of in-use master keys at all. But, with the ->s_master_keys keyring it is now possible. Then, add an ioctl FS_IOC_REMOVE_ENCRYPTION_KEY. It finds the specified master key in ->s_master_keys, then wipes the secret key itself, which prevents any additional inodes from being unlocked with the key. Then, it syncs the filesystem and evicts the inodes in the key's list. The normal inode eviction code will free and wipe the per-file keys (in ->i_crypt_info). Note that freeing ->i_crypt_info without evicting the inodes was also considered, but would have been racy. Some inodes may still be in use when a master key is removed, and we can't simply revoke random file descriptors, mmap's, etc. Thus, the ioctl simply skips in-use inodes, and returns -EBUSY to indicate that some inodes weren't evicted. The master key *secret* is still removed, but the fscrypt_master_key struct remains to keep track of the remaining inodes. Userspace can then retry the ioctl to evict the remaining inodes. Alternatively, if userspace adds the key again, the refreshed secret will be associated with the existing list of inodes so they remain correctly tracked for future key removals. The ioctl doesn't wipe pagecache pages. Thus, we tolerate that after a kernel compromise some portions of plaintext file contents may still be recoverable from memory. This can be solved by enabling page poisoning system-wide, which security conscious users may choose to do. But it's very difficult to solve otherwise, e.g. note that plaintext file contents may have been read in other places than pagecache pages. Like FS_IOC_ADD_ENCRYPTION_KEY, FS_IOC_REMOVE_ENCRYPTION_KEY is initially restricted to privileged users only. This is sufficient for some use cases, but not all. A later patch will relax this restriction, but it will require introducing key hashes, among other changes. Reviewed-by: Theodore Ts'o <tytso@mit.edu> Signed-off-by: Eric Biggers <ebiggers@google.com>
2019-08-05 10:35:46 +08:00
* - FS_IOC_REMOVE_ENCRYPTION_KEY
fscrypt: add FS_IOC_REMOVE_ENCRYPTION_KEY_ALL_USERS ioctl Add a root-only variant of the FS_IOC_REMOVE_ENCRYPTION_KEY ioctl which removes all users' claims of the key, not just the current user's claim. I.e., it always removes the key itself, no matter how many users have added it. This is useful for forcing a directory to be locked, without having to figure out which user ID(s) the key was added under. This is planned to be used by a command like 'sudo fscrypt lock DIR --all-users' in the fscrypt userspace tool (http://github.com/google/fscrypt). Reviewed-by: Theodore Ts'o <tytso@mit.edu> Signed-off-by: Eric Biggers <ebiggers@google.com>
2019-08-05 10:35:47 +08:00
* - FS_IOC_REMOVE_ENCRYPTION_KEY_ALL_USERS
fscrypt: add FS_IOC_GET_ENCRYPTION_KEY_STATUS ioctl Add a new fscrypt ioctl, FS_IOC_GET_ENCRYPTION_KEY_STATUS. Given a key specified by 'struct fscrypt_key_specifier' (the same way a key is specified for the other fscrypt key management ioctls), it returns status information in a 'struct fscrypt_get_key_status_arg'. The main motivation for this is that applications need to be able to check whether an encrypted directory is "unlocked" or not, so that they can add the key if it is not, and avoid adding the key (which may involve prompting the user for a passphrase) if it already is. It's possible to use some workarounds such as checking whether opening a regular file fails with ENOKEY, or checking whether the filenames "look like gibberish" or not. However, no workaround is usable in all cases. Like the other key management ioctls, the keyrings syscalls may seem at first to be a good fit for this. Unfortunately, they are not. Even if we exposed the keyring ID of the ->s_master_keys keyring and gave everyone Search permission on it (note: currently the keyrings permission system would also allow everyone to "invalidate" the keyring too), the fscrypt keys have an additional state that doesn't map cleanly to the keyrings API: the secret can be removed, but we can be still tracking the files that were using the key, and the removal can be re-attempted or the secret added again. After later patches, some applications will also need a way to determine whether a key was added by the current user vs. by some other user. Reserved fields are included in fscrypt_get_key_status_arg for this and other future extensions. Reviewed-by: Theodore Ts'o <tytso@mit.edu> Signed-off-by: Eric Biggers <ebiggers@google.com>
2019-08-05 10:35:46 +08:00
* - FS_IOC_GET_ENCRYPTION_KEY_STATUS
fscrypt: add FS_IOC_ADD_ENCRYPTION_KEY ioctl Add a new fscrypt ioctl, FS_IOC_ADD_ENCRYPTION_KEY. This ioctl adds an encryption key to the filesystem's fscrypt keyring ->s_master_keys, making any files encrypted with that key appear "unlocked". Why we need this ~~~~~~~~~~~~~~~~ The main problem is that the "locked/unlocked" (ciphertext/plaintext) status of encrypted files is global, but the fscrypt keys are not. fscrypt only looks for keys in the keyring(s) the process accessing the filesystem is subscribed to: the thread keyring, process keyring, and session keyring, where the session keyring may contain the user keyring. Therefore, userspace has to put fscrypt keys in the keyrings for individual users or sessions. But this means that when a process with a different keyring tries to access encrypted files, whether they appear "unlocked" or not is nondeterministic. This is because it depends on whether the files are currently present in the inode cache. Fixing this by consistently providing each process its own view of the filesystem depending on whether it has the key or not isn't feasible due to how the VFS caches work. Furthermore, while sometimes users expect this behavior, it is misguided for two reasons. First, it would be an OS-level access control mechanism largely redundant with existing access control mechanisms such as UNIX file permissions, ACLs, LSMs, etc. Encryption is actually for protecting the data at rest. Second, almost all users of fscrypt actually do need the keys to be global. The largest users of fscrypt, Android and Chromium OS, achieve this by having PID 1 create a "session keyring" that is inherited by every process. This works, but it isn't scalable because it prevents session keyrings from being used for any other purpose. On general-purpose Linux distros, the 'fscrypt' userspace tool [1] can't similarly abuse the session keyring, so to make 'sudo' work on all systems it has to link all the user keyrings into root's user keyring [2]. This is ugly and raises security concerns. Moreover it can't make the keys available to system services, such as sshd trying to access the user's '~/.ssh' directory (see [3], [4]) or NetworkManager trying to read certificates from the user's home directory (see [5]); or to Docker containers (see [6], [7]). By having an API to add a key to the *filesystem* we'll be able to fix the above bugs, remove userspace workarounds, and clearly express the intended semantics: the locked/unlocked status of an encrypted directory is global, and encryption is orthogonal to OS-level access control. Why not use the add_key() syscall ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ We use an ioctl for this API rather than the existing add_key() system call because the ioctl gives us the flexibility needed to implement fscrypt-specific semantics that will be introduced in later patches: - Supporting key removal with the semantics such that the secret is removed immediately and any unused inodes using the key are evicted; also, the eviction of any in-use inodes can be retried. - Calculating a key-dependent cryptographic identifier and returning it to userspace. - Allowing keys to be added and removed by non-root users, but only keys for v2 encryption policies; and to prevent denial-of-service attacks, users can only remove keys they themselves have added, and a key is only really removed after all users who added it have removed it. Trying to shoehorn these semantics into the keyrings syscalls would be very difficult, whereas the ioctls make things much easier. However, to reuse code the implementation still uses the keyrings service internally. Thus we get lockless RCU-mode key lookups without having to re-implement it, and the keys automatically show up in /proc/keys for debugging purposes. References: [1] https://github.com/google/fscrypt [2] https://goo.gl/55cCrI#heading=h.vf09isp98isb [3] https://github.com/google/fscrypt/issues/111#issuecomment-444347939 [4] https://github.com/google/fscrypt/issues/116 [5] https://bugs.launchpad.net/ubuntu/+source/fscrypt/+bug/1770715 [6] https://github.com/google/fscrypt/issues/128 [7] https://askubuntu.com/questions/1130306/cannot-run-docker-on-an-encrypted-filesystem Reviewed-by: Theodore Ts'o <tytso@mit.edu> Signed-off-by: Eric Biggers <ebiggers@google.com>
2019-08-05 10:35:46 +08:00
*
* See the "User API" section of Documentation/filesystems/fscrypt.rst for more
* information about these ioctls.
*/
fscrypt: stop using keyrings subsystem for fscrypt_master_key The approach of fs/crypto/ internally managing the fscrypt_master_key structs as the payloads of "struct key" objects contained in a "struct key" keyring has outlived its usefulness. The original idea was to simplify the code by reusing code from the keyrings subsystem. However, several issues have arisen that can't easily be resolved: - When a master key struct is destroyed, blk_crypto_evict_key() must be called on any per-mode keys embedded in it. (This started being the case when inline encryption support was added.) Yet, the keyrings subsystem can arbitrarily delay the destruction of keys, even past the time the filesystem was unmounted. Therefore, currently there is no easy way to call blk_crypto_evict_key() when a master key is destroyed. Currently, this is worked around by holding an extra reference to the filesystem's request_queue(s). But it was overlooked that the request_queue reference is *not* guaranteed to pin the corresponding blk_crypto_profile too; for device-mapper devices that support inline crypto, it doesn't. This can cause a use-after-free. - When the last inode that was using an incompletely-removed master key is evicted, the master key removal is completed by removing the key struct from the keyring. Currently this is done via key_invalidate(). Yet, key_invalidate() takes the key semaphore. This can deadlock when called from the shrinker, since in fscrypt_ioctl_add_key(), memory is allocated with GFP_KERNEL under the same semaphore. - More generally, the fact that the keyrings subsystem can arbitrarily delay the destruction of keys (via garbage collection delay, or via random processes getting temporary key references) is undesirable, as it means we can't strictly guarantee that all secrets are ever wiped. - Doing the master key lookups via the keyrings subsystem results in the key_permission LSM hook being called. fscrypt doesn't want this, as all access control for encrypted files is designed to happen via the files themselves, like any other files. The workaround which SELinux users are using is to change their SELinux policy to grant key search access to all domains. This works, but it is an odd extra step that shouldn't really have to be done. The fix for all these issues is to change the implementation to what I should have done originally: don't use the keyrings subsystem to keep track of the filesystem's fscrypt_master_key structs. Instead, just store them in a regular kernel data structure, and rework the reference counting, locking, and lifetime accordingly. Retain support for RCU-mode key lookups by using a hash table. Replace fscrypt_sb_free() with fscrypt_sb_delete(), which releases the keys synchronously and runs a bit earlier during unmount, so that block devices are still available. A side effect of this patch is that neither the master keys themselves nor the filesystem keyrings will be listed in /proc/keys anymore. ("Master key users" and the master key users keyrings will still be listed.) However, this was mostly an implementation detail, and it was intended just for debugging purposes. I don't know of anyone using it. This patch does *not* change how "master key users" (->mk_users) works; that still uses the keyrings subsystem. That is still needed for key quotas, and changing that isn't necessary to solve the issues listed above. If we decide to change that too, it would be a separate patch. I've marked this as fixing the original commit that added the fscrypt keyring, but as noted above the most important issue that this patch fixes wasn't introduced until the addition of inline encryption support. Fixes: 22d94f493bfb ("fscrypt: add FS_IOC_ADD_ENCRYPTION_KEY ioctl") Signed-off-by: Eric Biggers <ebiggers@google.com> Link: https://lore.kernel.org/r/20220901193208.138056-2-ebiggers@kernel.org
2022-09-02 03:32:06 +08:00
#include <asm/unaligned.h>
fscrypt: v2 encryption policy support Add a new fscrypt policy version, "v2". It has the following changes from the original policy version, which we call "v1" (*): - Master keys (the user-provided encryption keys) are only ever used as input to HKDF-SHA512. This is more flexible and less error-prone, and it avoids the quirks and limitations of the AES-128-ECB based KDF. Three classes of cryptographically isolated subkeys are defined: - Per-file keys, like used in v1 policies except for the new KDF. - Per-mode keys. These implement the semantics of the DIRECT_KEY flag, which for v1 policies made the master key be used directly. These are also planned to be used for inline encryption when support for it is added. - Key identifiers (see below). - Each master key is identified by a 16-byte master_key_identifier, which is derived from the key itself using HKDF-SHA512. This prevents users from associating the wrong key with an encrypted file or directory. This was easily possible with v1 policies, which identified the key by an arbitrary 8-byte master_key_descriptor. - The key must be provided in the filesystem-level keyring, not in a process-subscribed keyring. The following UAPI additions are made: - The existing ioctl FS_IOC_SET_ENCRYPTION_POLICY can now be passed a fscrypt_policy_v2 to set a v2 encryption policy. It's disambiguated from fscrypt_policy/fscrypt_policy_v1 by the version code prefix. - A new ioctl FS_IOC_GET_ENCRYPTION_POLICY_EX is added. It allows getting the v1 or v2 encryption policy of an encrypted file or directory. The existing FS_IOC_GET_ENCRYPTION_POLICY ioctl could not be used because it did not have a way for userspace to indicate which policy structure is expected. The new ioctl includes a size field, so it is extensible to future fscrypt policy versions. - The ioctls FS_IOC_ADD_ENCRYPTION_KEY, FS_IOC_REMOVE_ENCRYPTION_KEY, and FS_IOC_GET_ENCRYPTION_KEY_STATUS now support managing keys for v2 encryption policies. Such keys are kept logically separate from keys for v1 encryption policies, and are identified by 'identifier' rather than by 'descriptor'. The 'identifier' need not be provided when adding a key, since the kernel will calculate it anyway. This patch temporarily keeps adding/removing v2 policy keys behind the same permission check done for adding/removing v1 policy keys: capable(CAP_SYS_ADMIN). However, the next patch will carefully take advantage of the cryptographically secure master_key_identifier to allow non-root users to add/remove v2 policy keys, thus providing a full replacement for v1 policies. (*) Actually, in the API fscrypt_policy::version is 0 while on-disk fscrypt_context::format is 1. But I believe it makes the most sense to advance both to '2' to have them be in sync, and to consider the numbering to start at 1 except for the API quirk. Reviewed-by: Paul Crowley <paulcrowley@google.com> Reviewed-by: Theodore Ts'o <tytso@mit.edu> Signed-off-by: Eric Biggers <ebiggers@google.com>
2019-08-05 10:35:47 +08:00
#include <crypto/skcipher.h>
fscrypt: add FS_IOC_ADD_ENCRYPTION_KEY ioctl Add a new fscrypt ioctl, FS_IOC_ADD_ENCRYPTION_KEY. This ioctl adds an encryption key to the filesystem's fscrypt keyring ->s_master_keys, making any files encrypted with that key appear "unlocked". Why we need this ~~~~~~~~~~~~~~~~ The main problem is that the "locked/unlocked" (ciphertext/plaintext) status of encrypted files is global, but the fscrypt keys are not. fscrypt only looks for keys in the keyring(s) the process accessing the filesystem is subscribed to: the thread keyring, process keyring, and session keyring, where the session keyring may contain the user keyring. Therefore, userspace has to put fscrypt keys in the keyrings for individual users or sessions. But this means that when a process with a different keyring tries to access encrypted files, whether they appear "unlocked" or not is nondeterministic. This is because it depends on whether the files are currently present in the inode cache. Fixing this by consistently providing each process its own view of the filesystem depending on whether it has the key or not isn't feasible due to how the VFS caches work. Furthermore, while sometimes users expect this behavior, it is misguided for two reasons. First, it would be an OS-level access control mechanism largely redundant with existing access control mechanisms such as UNIX file permissions, ACLs, LSMs, etc. Encryption is actually for protecting the data at rest. Second, almost all users of fscrypt actually do need the keys to be global. The largest users of fscrypt, Android and Chromium OS, achieve this by having PID 1 create a "session keyring" that is inherited by every process. This works, but it isn't scalable because it prevents session keyrings from being used for any other purpose. On general-purpose Linux distros, the 'fscrypt' userspace tool [1] can't similarly abuse the session keyring, so to make 'sudo' work on all systems it has to link all the user keyrings into root's user keyring [2]. This is ugly and raises security concerns. Moreover it can't make the keys available to system services, such as sshd trying to access the user's '~/.ssh' directory (see [3], [4]) or NetworkManager trying to read certificates from the user's home directory (see [5]); or to Docker containers (see [6], [7]). By having an API to add a key to the *filesystem* we'll be able to fix the above bugs, remove userspace workarounds, and clearly express the intended semantics: the locked/unlocked status of an encrypted directory is global, and encryption is orthogonal to OS-level access control. Why not use the add_key() syscall ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ We use an ioctl for this API rather than the existing add_key() system call because the ioctl gives us the flexibility needed to implement fscrypt-specific semantics that will be introduced in later patches: - Supporting key removal with the semantics such that the secret is removed immediately and any unused inodes using the key are evicted; also, the eviction of any in-use inodes can be retried. - Calculating a key-dependent cryptographic identifier and returning it to userspace. - Allowing keys to be added and removed by non-root users, but only keys for v2 encryption policies; and to prevent denial-of-service attacks, users can only remove keys they themselves have added, and a key is only really removed after all users who added it have removed it. Trying to shoehorn these semantics into the keyrings syscalls would be very difficult, whereas the ioctls make things much easier. However, to reuse code the implementation still uses the keyrings service internally. Thus we get lockless RCU-mode key lookups without having to re-implement it, and the keys automatically show up in /proc/keys for debugging purposes. References: [1] https://github.com/google/fscrypt [2] https://goo.gl/55cCrI#heading=h.vf09isp98isb [3] https://github.com/google/fscrypt/issues/111#issuecomment-444347939 [4] https://github.com/google/fscrypt/issues/116 [5] https://bugs.launchpad.net/ubuntu/+source/fscrypt/+bug/1770715 [6] https://github.com/google/fscrypt/issues/128 [7] https://askubuntu.com/questions/1130306/cannot-run-docker-on-an-encrypted-filesystem Reviewed-by: Theodore Ts'o <tytso@mit.edu> Signed-off-by: Eric Biggers <ebiggers@google.com>
2019-08-05 10:35:46 +08:00
#include <linux/key-type.h>
fscrypt: add fscrypt_add_test_dummy_key() Currently, the test_dummy_encryption mount option (which is used for encryption I/O testing with xfstests) uses v1 encryption policies, and it relies on userspace inserting a test key into the session keyring. We need test_dummy_encryption to support v2 encryption policies too. Requiring userspace to add the test key doesn't work well with v2 policies, since v2 policies only support the filesystem keyring (not the session keyring), and keys in the filesystem keyring are lost when the filesystem is unmounted. Hooking all test code that unmounts and re-mounts the filesystem would be difficult. Instead, let's make the filesystem automatically add the test key to its keyring when test_dummy_encryption is enabled. That puts the responsibility for choosing the test key on the kernel. We could just hard-code a key. But out of paranoia, let's first try using a per-boot random key, to prevent this code from being misused. A per-boot key will work as long as no one expects dummy-encrypted files to remain accessible after a reboot. (gce-xfstests doesn't.) Therefore, this patch adds a function fscrypt_add_test_dummy_key() which implements the above. The next patch will use it. Link: https://lore.kernel.org/r/20200512233251.118314-3-ebiggers@kernel.org Reviewed-by: Theodore Ts'o <tytso@mit.edu> Reviewed-by: Jaegeuk Kim <jaegeuk@kernel.org> Signed-off-by: Eric Biggers <ebiggers@google.com>
2020-05-13 07:32:49 +08:00
#include <linux/random.h>
fscrypt: add FS_IOC_ADD_ENCRYPTION_KEY ioctl Add a new fscrypt ioctl, FS_IOC_ADD_ENCRYPTION_KEY. This ioctl adds an encryption key to the filesystem's fscrypt keyring ->s_master_keys, making any files encrypted with that key appear "unlocked". Why we need this ~~~~~~~~~~~~~~~~ The main problem is that the "locked/unlocked" (ciphertext/plaintext) status of encrypted files is global, but the fscrypt keys are not. fscrypt only looks for keys in the keyring(s) the process accessing the filesystem is subscribed to: the thread keyring, process keyring, and session keyring, where the session keyring may contain the user keyring. Therefore, userspace has to put fscrypt keys in the keyrings for individual users or sessions. But this means that when a process with a different keyring tries to access encrypted files, whether they appear "unlocked" or not is nondeterministic. This is because it depends on whether the files are currently present in the inode cache. Fixing this by consistently providing each process its own view of the filesystem depending on whether it has the key or not isn't feasible due to how the VFS caches work. Furthermore, while sometimes users expect this behavior, it is misguided for two reasons. First, it would be an OS-level access control mechanism largely redundant with existing access control mechanisms such as UNIX file permissions, ACLs, LSMs, etc. Encryption is actually for protecting the data at rest. Second, almost all users of fscrypt actually do need the keys to be global. The largest users of fscrypt, Android and Chromium OS, achieve this by having PID 1 create a "session keyring" that is inherited by every process. This works, but it isn't scalable because it prevents session keyrings from being used for any other purpose. On general-purpose Linux distros, the 'fscrypt' userspace tool [1] can't similarly abuse the session keyring, so to make 'sudo' work on all systems it has to link all the user keyrings into root's user keyring [2]. This is ugly and raises security concerns. Moreover it can't make the keys available to system services, such as sshd trying to access the user's '~/.ssh' directory (see [3], [4]) or NetworkManager trying to read certificates from the user's home directory (see [5]); or to Docker containers (see [6], [7]). By having an API to add a key to the *filesystem* we'll be able to fix the above bugs, remove userspace workarounds, and clearly express the intended semantics: the locked/unlocked status of an encrypted directory is global, and encryption is orthogonal to OS-level access control. Why not use the add_key() syscall ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ We use an ioctl for this API rather than the existing add_key() system call because the ioctl gives us the flexibility needed to implement fscrypt-specific semantics that will be introduced in later patches: - Supporting key removal with the semantics such that the secret is removed immediately and any unused inodes using the key are evicted; also, the eviction of any in-use inodes can be retried. - Calculating a key-dependent cryptographic identifier and returning it to userspace. - Allowing keys to be added and removed by non-root users, but only keys for v2 encryption policies; and to prevent denial-of-service attacks, users can only remove keys they themselves have added, and a key is only really removed after all users who added it have removed it. Trying to shoehorn these semantics into the keyrings syscalls would be very difficult, whereas the ioctls make things much easier. However, to reuse code the implementation still uses the keyrings service internally. Thus we get lockless RCU-mode key lookups without having to re-implement it, and the keys automatically show up in /proc/keys for debugging purposes. References: [1] https://github.com/google/fscrypt [2] https://goo.gl/55cCrI#heading=h.vf09isp98isb [3] https://github.com/google/fscrypt/issues/111#issuecomment-444347939 [4] https://github.com/google/fscrypt/issues/116 [5] https://bugs.launchpad.net/ubuntu/+source/fscrypt/+bug/1770715 [6] https://github.com/google/fscrypt/issues/128 [7] https://askubuntu.com/questions/1130306/cannot-run-docker-on-an-encrypted-filesystem Reviewed-by: Theodore Ts'o <tytso@mit.edu> Signed-off-by: Eric Biggers <ebiggers@google.com>
2019-08-05 10:35:46 +08:00
#include <linux/seq_file.h>
#include "fscrypt_private.h"
fscrypt: stop using keyrings subsystem for fscrypt_master_key The approach of fs/crypto/ internally managing the fscrypt_master_key structs as the payloads of "struct key" objects contained in a "struct key" keyring has outlived its usefulness. The original idea was to simplify the code by reusing code from the keyrings subsystem. However, several issues have arisen that can't easily be resolved: - When a master key struct is destroyed, blk_crypto_evict_key() must be called on any per-mode keys embedded in it. (This started being the case when inline encryption support was added.) Yet, the keyrings subsystem can arbitrarily delay the destruction of keys, even past the time the filesystem was unmounted. Therefore, currently there is no easy way to call blk_crypto_evict_key() when a master key is destroyed. Currently, this is worked around by holding an extra reference to the filesystem's request_queue(s). But it was overlooked that the request_queue reference is *not* guaranteed to pin the corresponding blk_crypto_profile too; for device-mapper devices that support inline crypto, it doesn't. This can cause a use-after-free. - When the last inode that was using an incompletely-removed master key is evicted, the master key removal is completed by removing the key struct from the keyring. Currently this is done via key_invalidate(). Yet, key_invalidate() takes the key semaphore. This can deadlock when called from the shrinker, since in fscrypt_ioctl_add_key(), memory is allocated with GFP_KERNEL under the same semaphore. - More generally, the fact that the keyrings subsystem can arbitrarily delay the destruction of keys (via garbage collection delay, or via random processes getting temporary key references) is undesirable, as it means we can't strictly guarantee that all secrets are ever wiped. - Doing the master key lookups via the keyrings subsystem results in the key_permission LSM hook being called. fscrypt doesn't want this, as all access control for encrypted files is designed to happen via the files themselves, like any other files. The workaround which SELinux users are using is to change their SELinux policy to grant key search access to all domains. This works, but it is an odd extra step that shouldn't really have to be done. The fix for all these issues is to change the implementation to what I should have done originally: don't use the keyrings subsystem to keep track of the filesystem's fscrypt_master_key structs. Instead, just store them in a regular kernel data structure, and rework the reference counting, locking, and lifetime accordingly. Retain support for RCU-mode key lookups by using a hash table. Replace fscrypt_sb_free() with fscrypt_sb_delete(), which releases the keys synchronously and runs a bit earlier during unmount, so that block devices are still available. A side effect of this patch is that neither the master keys themselves nor the filesystem keyrings will be listed in /proc/keys anymore. ("Master key users" and the master key users keyrings will still be listed.) However, this was mostly an implementation detail, and it was intended just for debugging purposes. I don't know of anyone using it. This patch does *not* change how "master key users" (->mk_users) works; that still uses the keyrings subsystem. That is still needed for key quotas, and changing that isn't necessary to solve the issues listed above. If we decide to change that too, it would be a separate patch. I've marked this as fixing the original commit that added the fscrypt keyring, but as noted above the most important issue that this patch fixes wasn't introduced until the addition of inline encryption support. Fixes: 22d94f493bfb ("fscrypt: add FS_IOC_ADD_ENCRYPTION_KEY ioctl") Signed-off-by: Eric Biggers <ebiggers@google.com> Link: https://lore.kernel.org/r/20220901193208.138056-2-ebiggers@kernel.org
2022-09-02 03:32:06 +08:00
/* The master encryption keys for a filesystem (->s_master_keys) */
struct fscrypt_keyring {
/*
* Lock that protects ->key_hashtable. It does *not* protect the
* fscrypt_master_key structs themselves.
*/
spinlock_t lock;
/* Hash table that maps fscrypt_key_specifier to fscrypt_master_key */
struct hlist_head key_hashtable[128];
};
fscrypt: add FS_IOC_ADD_ENCRYPTION_KEY ioctl Add a new fscrypt ioctl, FS_IOC_ADD_ENCRYPTION_KEY. This ioctl adds an encryption key to the filesystem's fscrypt keyring ->s_master_keys, making any files encrypted with that key appear "unlocked". Why we need this ~~~~~~~~~~~~~~~~ The main problem is that the "locked/unlocked" (ciphertext/plaintext) status of encrypted files is global, but the fscrypt keys are not. fscrypt only looks for keys in the keyring(s) the process accessing the filesystem is subscribed to: the thread keyring, process keyring, and session keyring, where the session keyring may contain the user keyring. Therefore, userspace has to put fscrypt keys in the keyrings for individual users or sessions. But this means that when a process with a different keyring tries to access encrypted files, whether they appear "unlocked" or not is nondeterministic. This is because it depends on whether the files are currently present in the inode cache. Fixing this by consistently providing each process its own view of the filesystem depending on whether it has the key or not isn't feasible due to how the VFS caches work. Furthermore, while sometimes users expect this behavior, it is misguided for two reasons. First, it would be an OS-level access control mechanism largely redundant with existing access control mechanisms such as UNIX file permissions, ACLs, LSMs, etc. Encryption is actually for protecting the data at rest. Second, almost all users of fscrypt actually do need the keys to be global. The largest users of fscrypt, Android and Chromium OS, achieve this by having PID 1 create a "session keyring" that is inherited by every process. This works, but it isn't scalable because it prevents session keyrings from being used for any other purpose. On general-purpose Linux distros, the 'fscrypt' userspace tool [1] can't similarly abuse the session keyring, so to make 'sudo' work on all systems it has to link all the user keyrings into root's user keyring [2]. This is ugly and raises security concerns. Moreover it can't make the keys available to system services, such as sshd trying to access the user's '~/.ssh' directory (see [3], [4]) or NetworkManager trying to read certificates from the user's home directory (see [5]); or to Docker containers (see [6], [7]). By having an API to add a key to the *filesystem* we'll be able to fix the above bugs, remove userspace workarounds, and clearly express the intended semantics: the locked/unlocked status of an encrypted directory is global, and encryption is orthogonal to OS-level access control. Why not use the add_key() syscall ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ We use an ioctl for this API rather than the existing add_key() system call because the ioctl gives us the flexibility needed to implement fscrypt-specific semantics that will be introduced in later patches: - Supporting key removal with the semantics such that the secret is removed immediately and any unused inodes using the key are evicted; also, the eviction of any in-use inodes can be retried. - Calculating a key-dependent cryptographic identifier and returning it to userspace. - Allowing keys to be added and removed by non-root users, but only keys for v2 encryption policies; and to prevent denial-of-service attacks, users can only remove keys they themselves have added, and a key is only really removed after all users who added it have removed it. Trying to shoehorn these semantics into the keyrings syscalls would be very difficult, whereas the ioctls make things much easier. However, to reuse code the implementation still uses the keyrings service internally. Thus we get lockless RCU-mode key lookups without having to re-implement it, and the keys automatically show up in /proc/keys for debugging purposes. References: [1] https://github.com/google/fscrypt [2] https://goo.gl/55cCrI#heading=h.vf09isp98isb [3] https://github.com/google/fscrypt/issues/111#issuecomment-444347939 [4] https://github.com/google/fscrypt/issues/116 [5] https://bugs.launchpad.net/ubuntu/+source/fscrypt/+bug/1770715 [6] https://github.com/google/fscrypt/issues/128 [7] https://askubuntu.com/questions/1130306/cannot-run-docker-on-an-encrypted-filesystem Reviewed-by: Theodore Ts'o <tytso@mit.edu> Signed-off-by: Eric Biggers <ebiggers@google.com>
2019-08-05 10:35:46 +08:00
static void wipe_master_key_secret(struct fscrypt_master_key_secret *secret)
{
fscrypt: v2 encryption policy support Add a new fscrypt policy version, "v2". It has the following changes from the original policy version, which we call "v1" (*): - Master keys (the user-provided encryption keys) are only ever used as input to HKDF-SHA512. This is more flexible and less error-prone, and it avoids the quirks and limitations of the AES-128-ECB based KDF. Three classes of cryptographically isolated subkeys are defined: - Per-file keys, like used in v1 policies except for the new KDF. - Per-mode keys. These implement the semantics of the DIRECT_KEY flag, which for v1 policies made the master key be used directly. These are also planned to be used for inline encryption when support for it is added. - Key identifiers (see below). - Each master key is identified by a 16-byte master_key_identifier, which is derived from the key itself using HKDF-SHA512. This prevents users from associating the wrong key with an encrypted file or directory. This was easily possible with v1 policies, which identified the key by an arbitrary 8-byte master_key_descriptor. - The key must be provided in the filesystem-level keyring, not in a process-subscribed keyring. The following UAPI additions are made: - The existing ioctl FS_IOC_SET_ENCRYPTION_POLICY can now be passed a fscrypt_policy_v2 to set a v2 encryption policy. It's disambiguated from fscrypt_policy/fscrypt_policy_v1 by the version code prefix. - A new ioctl FS_IOC_GET_ENCRYPTION_POLICY_EX is added. It allows getting the v1 or v2 encryption policy of an encrypted file or directory. The existing FS_IOC_GET_ENCRYPTION_POLICY ioctl could not be used because it did not have a way for userspace to indicate which policy structure is expected. The new ioctl includes a size field, so it is extensible to future fscrypt policy versions. - The ioctls FS_IOC_ADD_ENCRYPTION_KEY, FS_IOC_REMOVE_ENCRYPTION_KEY, and FS_IOC_GET_ENCRYPTION_KEY_STATUS now support managing keys for v2 encryption policies. Such keys are kept logically separate from keys for v1 encryption policies, and are identified by 'identifier' rather than by 'descriptor'. The 'identifier' need not be provided when adding a key, since the kernel will calculate it anyway. This patch temporarily keeps adding/removing v2 policy keys behind the same permission check done for adding/removing v1 policy keys: capable(CAP_SYS_ADMIN). However, the next patch will carefully take advantage of the cryptographically secure master_key_identifier to allow non-root users to add/remove v2 policy keys, thus providing a full replacement for v1 policies. (*) Actually, in the API fscrypt_policy::version is 0 while on-disk fscrypt_context::format is 1. But I believe it makes the most sense to advance both to '2' to have them be in sync, and to consider the numbering to start at 1 except for the API quirk. Reviewed-by: Paul Crowley <paulcrowley@google.com> Reviewed-by: Theodore Ts'o <tytso@mit.edu> Signed-off-by: Eric Biggers <ebiggers@google.com>
2019-08-05 10:35:47 +08:00
fscrypt_destroy_hkdf(&secret->hkdf);
fscrypt: add FS_IOC_ADD_ENCRYPTION_KEY ioctl Add a new fscrypt ioctl, FS_IOC_ADD_ENCRYPTION_KEY. This ioctl adds an encryption key to the filesystem's fscrypt keyring ->s_master_keys, making any files encrypted with that key appear "unlocked". Why we need this ~~~~~~~~~~~~~~~~ The main problem is that the "locked/unlocked" (ciphertext/plaintext) status of encrypted files is global, but the fscrypt keys are not. fscrypt only looks for keys in the keyring(s) the process accessing the filesystem is subscribed to: the thread keyring, process keyring, and session keyring, where the session keyring may contain the user keyring. Therefore, userspace has to put fscrypt keys in the keyrings for individual users or sessions. But this means that when a process with a different keyring tries to access encrypted files, whether they appear "unlocked" or not is nondeterministic. This is because it depends on whether the files are currently present in the inode cache. Fixing this by consistently providing each process its own view of the filesystem depending on whether it has the key or not isn't feasible due to how the VFS caches work. Furthermore, while sometimes users expect this behavior, it is misguided for two reasons. First, it would be an OS-level access control mechanism largely redundant with existing access control mechanisms such as UNIX file permissions, ACLs, LSMs, etc. Encryption is actually for protecting the data at rest. Second, almost all users of fscrypt actually do need the keys to be global. The largest users of fscrypt, Android and Chromium OS, achieve this by having PID 1 create a "session keyring" that is inherited by every process. This works, but it isn't scalable because it prevents session keyrings from being used for any other purpose. On general-purpose Linux distros, the 'fscrypt' userspace tool [1] can't similarly abuse the session keyring, so to make 'sudo' work on all systems it has to link all the user keyrings into root's user keyring [2]. This is ugly and raises security concerns. Moreover it can't make the keys available to system services, such as sshd trying to access the user's '~/.ssh' directory (see [3], [4]) or NetworkManager trying to read certificates from the user's home directory (see [5]); or to Docker containers (see [6], [7]). By having an API to add a key to the *filesystem* we'll be able to fix the above bugs, remove userspace workarounds, and clearly express the intended semantics: the locked/unlocked status of an encrypted directory is global, and encryption is orthogonal to OS-level access control. Why not use the add_key() syscall ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ We use an ioctl for this API rather than the existing add_key() system call because the ioctl gives us the flexibility needed to implement fscrypt-specific semantics that will be introduced in later patches: - Supporting key removal with the semantics such that the secret is removed immediately and any unused inodes using the key are evicted; also, the eviction of any in-use inodes can be retried. - Calculating a key-dependent cryptographic identifier and returning it to userspace. - Allowing keys to be added and removed by non-root users, but only keys for v2 encryption policies; and to prevent denial-of-service attacks, users can only remove keys they themselves have added, and a key is only really removed after all users who added it have removed it. Trying to shoehorn these semantics into the keyrings syscalls would be very difficult, whereas the ioctls make things much easier. However, to reuse code the implementation still uses the keyrings service internally. Thus we get lockless RCU-mode key lookups without having to re-implement it, and the keys automatically show up in /proc/keys for debugging purposes. References: [1] https://github.com/google/fscrypt [2] https://goo.gl/55cCrI#heading=h.vf09isp98isb [3] https://github.com/google/fscrypt/issues/111#issuecomment-444347939 [4] https://github.com/google/fscrypt/issues/116 [5] https://bugs.launchpad.net/ubuntu/+source/fscrypt/+bug/1770715 [6] https://github.com/google/fscrypt/issues/128 [7] https://askubuntu.com/questions/1130306/cannot-run-docker-on-an-encrypted-filesystem Reviewed-by: Theodore Ts'o <tytso@mit.edu> Signed-off-by: Eric Biggers <ebiggers@google.com>
2019-08-05 10:35:46 +08:00
memzero_explicit(secret, sizeof(*secret));
}
static void move_master_key_secret(struct fscrypt_master_key_secret *dst,
struct fscrypt_master_key_secret *src)
{
memcpy(dst, src, sizeof(*dst));
memzero_explicit(src, sizeof(*src));
}
fscrypt: stop using keyrings subsystem for fscrypt_master_key The approach of fs/crypto/ internally managing the fscrypt_master_key structs as the payloads of "struct key" objects contained in a "struct key" keyring has outlived its usefulness. The original idea was to simplify the code by reusing code from the keyrings subsystem. However, several issues have arisen that can't easily be resolved: - When a master key struct is destroyed, blk_crypto_evict_key() must be called on any per-mode keys embedded in it. (This started being the case when inline encryption support was added.) Yet, the keyrings subsystem can arbitrarily delay the destruction of keys, even past the time the filesystem was unmounted. Therefore, currently there is no easy way to call blk_crypto_evict_key() when a master key is destroyed. Currently, this is worked around by holding an extra reference to the filesystem's request_queue(s). But it was overlooked that the request_queue reference is *not* guaranteed to pin the corresponding blk_crypto_profile too; for device-mapper devices that support inline crypto, it doesn't. This can cause a use-after-free. - When the last inode that was using an incompletely-removed master key is evicted, the master key removal is completed by removing the key struct from the keyring. Currently this is done via key_invalidate(). Yet, key_invalidate() takes the key semaphore. This can deadlock when called from the shrinker, since in fscrypt_ioctl_add_key(), memory is allocated with GFP_KERNEL under the same semaphore. - More generally, the fact that the keyrings subsystem can arbitrarily delay the destruction of keys (via garbage collection delay, or via random processes getting temporary key references) is undesirable, as it means we can't strictly guarantee that all secrets are ever wiped. - Doing the master key lookups via the keyrings subsystem results in the key_permission LSM hook being called. fscrypt doesn't want this, as all access control for encrypted files is designed to happen via the files themselves, like any other files. The workaround which SELinux users are using is to change their SELinux policy to grant key search access to all domains. This works, but it is an odd extra step that shouldn't really have to be done. The fix for all these issues is to change the implementation to what I should have done originally: don't use the keyrings subsystem to keep track of the filesystem's fscrypt_master_key structs. Instead, just store them in a regular kernel data structure, and rework the reference counting, locking, and lifetime accordingly. Retain support for RCU-mode key lookups by using a hash table. Replace fscrypt_sb_free() with fscrypt_sb_delete(), which releases the keys synchronously and runs a bit earlier during unmount, so that block devices are still available. A side effect of this patch is that neither the master keys themselves nor the filesystem keyrings will be listed in /proc/keys anymore. ("Master key users" and the master key users keyrings will still be listed.) However, this was mostly an implementation detail, and it was intended just for debugging purposes. I don't know of anyone using it. This patch does *not* change how "master key users" (->mk_users) works; that still uses the keyrings subsystem. That is still needed for key quotas, and changing that isn't necessary to solve the issues listed above. If we decide to change that too, it would be a separate patch. I've marked this as fixing the original commit that added the fscrypt keyring, but as noted above the most important issue that this patch fixes wasn't introduced until the addition of inline encryption support. Fixes: 22d94f493bfb ("fscrypt: add FS_IOC_ADD_ENCRYPTION_KEY ioctl") Signed-off-by: Eric Biggers <ebiggers@google.com> Link: https://lore.kernel.org/r/20220901193208.138056-2-ebiggers@kernel.org
2022-09-02 03:32:06 +08:00
static void fscrypt_free_master_key(struct rcu_head *head)
{
struct fscrypt_master_key *mk =
container_of(head, struct fscrypt_master_key, mk_rcu_head);
/*
* The master key secret and any embedded subkeys should have already
* been wiped when the last active reference to the fscrypt_master_key
* struct was dropped; doing it here would be unnecessarily late.
* Nevertheless, use kfree_sensitive() in case anything was missed.
*/
kfree_sensitive(mk);
}
void fscrypt_put_master_key(struct fscrypt_master_key *mk)
{
if (!refcount_dec_and_test(&mk->mk_struct_refs))
return;
/*
* No structural references left, so free ->mk_users, and also free the
* fscrypt_master_key struct itself after an RCU grace period ensures
* that concurrent keyring lookups can no longer find it.
*/
WARN_ON(refcount_read(&mk->mk_active_refs) != 0);
key_put(mk->mk_users);
mk->mk_users = NULL;
call_rcu(&mk->mk_rcu_head, fscrypt_free_master_key);
}
void fscrypt_put_master_key_activeref(struct fscrypt_master_key *mk)
fscrypt: add FS_IOC_ADD_ENCRYPTION_KEY ioctl Add a new fscrypt ioctl, FS_IOC_ADD_ENCRYPTION_KEY. This ioctl adds an encryption key to the filesystem's fscrypt keyring ->s_master_keys, making any files encrypted with that key appear "unlocked". Why we need this ~~~~~~~~~~~~~~~~ The main problem is that the "locked/unlocked" (ciphertext/plaintext) status of encrypted files is global, but the fscrypt keys are not. fscrypt only looks for keys in the keyring(s) the process accessing the filesystem is subscribed to: the thread keyring, process keyring, and session keyring, where the session keyring may contain the user keyring. Therefore, userspace has to put fscrypt keys in the keyrings for individual users or sessions. But this means that when a process with a different keyring tries to access encrypted files, whether they appear "unlocked" or not is nondeterministic. This is because it depends on whether the files are currently present in the inode cache. Fixing this by consistently providing each process its own view of the filesystem depending on whether it has the key or not isn't feasible due to how the VFS caches work. Furthermore, while sometimes users expect this behavior, it is misguided for two reasons. First, it would be an OS-level access control mechanism largely redundant with existing access control mechanisms such as UNIX file permissions, ACLs, LSMs, etc. Encryption is actually for protecting the data at rest. Second, almost all users of fscrypt actually do need the keys to be global. The largest users of fscrypt, Android and Chromium OS, achieve this by having PID 1 create a "session keyring" that is inherited by every process. This works, but it isn't scalable because it prevents session keyrings from being used for any other purpose. On general-purpose Linux distros, the 'fscrypt' userspace tool [1] can't similarly abuse the session keyring, so to make 'sudo' work on all systems it has to link all the user keyrings into root's user keyring [2]. This is ugly and raises security concerns. Moreover it can't make the keys available to system services, such as sshd trying to access the user's '~/.ssh' directory (see [3], [4]) or NetworkManager trying to read certificates from the user's home directory (see [5]); or to Docker containers (see [6], [7]). By having an API to add a key to the *filesystem* we'll be able to fix the above bugs, remove userspace workarounds, and clearly express the intended semantics: the locked/unlocked status of an encrypted directory is global, and encryption is orthogonal to OS-level access control. Why not use the add_key() syscall ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ We use an ioctl for this API rather than the existing add_key() system call because the ioctl gives us the flexibility needed to implement fscrypt-specific semantics that will be introduced in later patches: - Supporting key removal with the semantics such that the secret is removed immediately and any unused inodes using the key are evicted; also, the eviction of any in-use inodes can be retried. - Calculating a key-dependent cryptographic identifier and returning it to userspace. - Allowing keys to be added and removed by non-root users, but only keys for v2 encryption policies; and to prevent denial-of-service attacks, users can only remove keys they themselves have added, and a key is only really removed after all users who added it have removed it. Trying to shoehorn these semantics into the keyrings syscalls would be very difficult, whereas the ioctls make things much easier. However, to reuse code the implementation still uses the keyrings service internally. Thus we get lockless RCU-mode key lookups without having to re-implement it, and the keys automatically show up in /proc/keys for debugging purposes. References: [1] https://github.com/google/fscrypt [2] https://goo.gl/55cCrI#heading=h.vf09isp98isb [3] https://github.com/google/fscrypt/issues/111#issuecomment-444347939 [4] https://github.com/google/fscrypt/issues/116 [5] https://bugs.launchpad.net/ubuntu/+source/fscrypt/+bug/1770715 [6] https://github.com/google/fscrypt/issues/128 [7] https://askubuntu.com/questions/1130306/cannot-run-docker-on-an-encrypted-filesystem Reviewed-by: Theodore Ts'o <tytso@mit.edu> Signed-off-by: Eric Biggers <ebiggers@google.com>
2019-08-05 10:35:46 +08:00
{
fscrypt: stop using keyrings subsystem for fscrypt_master_key The approach of fs/crypto/ internally managing the fscrypt_master_key structs as the payloads of "struct key" objects contained in a "struct key" keyring has outlived its usefulness. The original idea was to simplify the code by reusing code from the keyrings subsystem. However, several issues have arisen that can't easily be resolved: - When a master key struct is destroyed, blk_crypto_evict_key() must be called on any per-mode keys embedded in it. (This started being the case when inline encryption support was added.) Yet, the keyrings subsystem can arbitrarily delay the destruction of keys, even past the time the filesystem was unmounted. Therefore, currently there is no easy way to call blk_crypto_evict_key() when a master key is destroyed. Currently, this is worked around by holding an extra reference to the filesystem's request_queue(s). But it was overlooked that the request_queue reference is *not* guaranteed to pin the corresponding blk_crypto_profile too; for device-mapper devices that support inline crypto, it doesn't. This can cause a use-after-free. - When the last inode that was using an incompletely-removed master key is evicted, the master key removal is completed by removing the key struct from the keyring. Currently this is done via key_invalidate(). Yet, key_invalidate() takes the key semaphore. This can deadlock when called from the shrinker, since in fscrypt_ioctl_add_key(), memory is allocated with GFP_KERNEL under the same semaphore. - More generally, the fact that the keyrings subsystem can arbitrarily delay the destruction of keys (via garbage collection delay, or via random processes getting temporary key references) is undesirable, as it means we can't strictly guarantee that all secrets are ever wiped. - Doing the master key lookups via the keyrings subsystem results in the key_permission LSM hook being called. fscrypt doesn't want this, as all access control for encrypted files is designed to happen via the files themselves, like any other files. The workaround which SELinux users are using is to change their SELinux policy to grant key search access to all domains. This works, but it is an odd extra step that shouldn't really have to be done. The fix for all these issues is to change the implementation to what I should have done originally: don't use the keyrings subsystem to keep track of the filesystem's fscrypt_master_key structs. Instead, just store them in a regular kernel data structure, and rework the reference counting, locking, and lifetime accordingly. Retain support for RCU-mode key lookups by using a hash table. Replace fscrypt_sb_free() with fscrypt_sb_delete(), which releases the keys synchronously and runs a bit earlier during unmount, so that block devices are still available. A side effect of this patch is that neither the master keys themselves nor the filesystem keyrings will be listed in /proc/keys anymore. ("Master key users" and the master key users keyrings will still be listed.) However, this was mostly an implementation detail, and it was intended just for debugging purposes. I don't know of anyone using it. This patch does *not* change how "master key users" (->mk_users) works; that still uses the keyrings subsystem. That is still needed for key quotas, and changing that isn't necessary to solve the issues listed above. If we decide to change that too, it would be a separate patch. I've marked this as fixing the original commit that added the fscrypt keyring, but as noted above the most important issue that this patch fixes wasn't introduced until the addition of inline encryption support. Fixes: 22d94f493bfb ("fscrypt: add FS_IOC_ADD_ENCRYPTION_KEY ioctl") Signed-off-by: Eric Biggers <ebiggers@google.com> Link: https://lore.kernel.org/r/20220901193208.138056-2-ebiggers@kernel.org
2022-09-02 03:32:06 +08:00
struct super_block *sb = mk->mk_sb;
struct fscrypt_keyring *keyring = sb->s_master_keys;
fscrypt: v2 encryption policy support Add a new fscrypt policy version, "v2". It has the following changes from the original policy version, which we call "v1" (*): - Master keys (the user-provided encryption keys) are only ever used as input to HKDF-SHA512. This is more flexible and less error-prone, and it avoids the quirks and limitations of the AES-128-ECB based KDF. Three classes of cryptographically isolated subkeys are defined: - Per-file keys, like used in v1 policies except for the new KDF. - Per-mode keys. These implement the semantics of the DIRECT_KEY flag, which for v1 policies made the master key be used directly. These are also planned to be used for inline encryption when support for it is added. - Key identifiers (see below). - Each master key is identified by a 16-byte master_key_identifier, which is derived from the key itself using HKDF-SHA512. This prevents users from associating the wrong key with an encrypted file or directory. This was easily possible with v1 policies, which identified the key by an arbitrary 8-byte master_key_descriptor. - The key must be provided in the filesystem-level keyring, not in a process-subscribed keyring. The following UAPI additions are made: - The existing ioctl FS_IOC_SET_ENCRYPTION_POLICY can now be passed a fscrypt_policy_v2 to set a v2 encryption policy. It's disambiguated from fscrypt_policy/fscrypt_policy_v1 by the version code prefix. - A new ioctl FS_IOC_GET_ENCRYPTION_POLICY_EX is added. It allows getting the v1 or v2 encryption policy of an encrypted file or directory. The existing FS_IOC_GET_ENCRYPTION_POLICY ioctl could not be used because it did not have a way for userspace to indicate which policy structure is expected. The new ioctl includes a size field, so it is extensible to future fscrypt policy versions. - The ioctls FS_IOC_ADD_ENCRYPTION_KEY, FS_IOC_REMOVE_ENCRYPTION_KEY, and FS_IOC_GET_ENCRYPTION_KEY_STATUS now support managing keys for v2 encryption policies. Such keys are kept logically separate from keys for v1 encryption policies, and are identified by 'identifier' rather than by 'descriptor'. The 'identifier' need not be provided when adding a key, since the kernel will calculate it anyway. This patch temporarily keeps adding/removing v2 policy keys behind the same permission check done for adding/removing v1 policy keys: capable(CAP_SYS_ADMIN). However, the next patch will carefully take advantage of the cryptographically secure master_key_identifier to allow non-root users to add/remove v2 policy keys, thus providing a full replacement for v1 policies. (*) Actually, in the API fscrypt_policy::version is 0 while on-disk fscrypt_context::format is 1. But I believe it makes the most sense to advance both to '2' to have them be in sync, and to consider the numbering to start at 1 except for the API quirk. Reviewed-by: Paul Crowley <paulcrowley@google.com> Reviewed-by: Theodore Ts'o <tytso@mit.edu> Signed-off-by: Eric Biggers <ebiggers@google.com>
2019-08-05 10:35:47 +08:00
size_t i;
fscrypt: stop using keyrings subsystem for fscrypt_master_key The approach of fs/crypto/ internally managing the fscrypt_master_key structs as the payloads of "struct key" objects contained in a "struct key" keyring has outlived its usefulness. The original idea was to simplify the code by reusing code from the keyrings subsystem. However, several issues have arisen that can't easily be resolved: - When a master key struct is destroyed, blk_crypto_evict_key() must be called on any per-mode keys embedded in it. (This started being the case when inline encryption support was added.) Yet, the keyrings subsystem can arbitrarily delay the destruction of keys, even past the time the filesystem was unmounted. Therefore, currently there is no easy way to call blk_crypto_evict_key() when a master key is destroyed. Currently, this is worked around by holding an extra reference to the filesystem's request_queue(s). But it was overlooked that the request_queue reference is *not* guaranteed to pin the corresponding blk_crypto_profile too; for device-mapper devices that support inline crypto, it doesn't. This can cause a use-after-free. - When the last inode that was using an incompletely-removed master key is evicted, the master key removal is completed by removing the key struct from the keyring. Currently this is done via key_invalidate(). Yet, key_invalidate() takes the key semaphore. This can deadlock when called from the shrinker, since in fscrypt_ioctl_add_key(), memory is allocated with GFP_KERNEL under the same semaphore. - More generally, the fact that the keyrings subsystem can arbitrarily delay the destruction of keys (via garbage collection delay, or via random processes getting temporary key references) is undesirable, as it means we can't strictly guarantee that all secrets are ever wiped. - Doing the master key lookups via the keyrings subsystem results in the key_permission LSM hook being called. fscrypt doesn't want this, as all access control for encrypted files is designed to happen via the files themselves, like any other files. The workaround which SELinux users are using is to change their SELinux policy to grant key search access to all domains. This works, but it is an odd extra step that shouldn't really have to be done. The fix for all these issues is to change the implementation to what I should have done originally: don't use the keyrings subsystem to keep track of the filesystem's fscrypt_master_key structs. Instead, just store them in a regular kernel data structure, and rework the reference counting, locking, and lifetime accordingly. Retain support for RCU-mode key lookups by using a hash table. Replace fscrypt_sb_free() with fscrypt_sb_delete(), which releases the keys synchronously and runs a bit earlier during unmount, so that block devices are still available. A side effect of this patch is that neither the master keys themselves nor the filesystem keyrings will be listed in /proc/keys anymore. ("Master key users" and the master key users keyrings will still be listed.) However, this was mostly an implementation detail, and it was intended just for debugging purposes. I don't know of anyone using it. This patch does *not* change how "master key users" (->mk_users) works; that still uses the keyrings subsystem. That is still needed for key quotas, and changing that isn't necessary to solve the issues listed above. If we decide to change that too, it would be a separate patch. I've marked this as fixing the original commit that added the fscrypt keyring, but as noted above the most important issue that this patch fixes wasn't introduced until the addition of inline encryption support. Fixes: 22d94f493bfb ("fscrypt: add FS_IOC_ADD_ENCRYPTION_KEY ioctl") Signed-off-by: Eric Biggers <ebiggers@google.com> Link: https://lore.kernel.org/r/20220901193208.138056-2-ebiggers@kernel.org
2022-09-02 03:32:06 +08:00
if (!refcount_dec_and_test(&mk->mk_active_refs))
return;
/*
* No active references left, so complete the full removal of this
* fscrypt_master_key struct by removing it from the keyring and
* destroying any subkeys embedded in it.
*/
spin_lock(&keyring->lock);
hlist_del_rcu(&mk->mk_node);
spin_unlock(&keyring->lock);
/*
* ->mk_active_refs == 0 implies that ->mk_secret is not present and
* that ->mk_decrypted_inodes is empty.
*/
WARN_ON(is_master_key_secret_present(&mk->mk_secret));
WARN_ON(!list_empty(&mk->mk_decrypted_inodes));
fscrypt: v2 encryption policy support Add a new fscrypt policy version, "v2". It has the following changes from the original policy version, which we call "v1" (*): - Master keys (the user-provided encryption keys) are only ever used as input to HKDF-SHA512. This is more flexible and less error-prone, and it avoids the quirks and limitations of the AES-128-ECB based KDF. Three classes of cryptographically isolated subkeys are defined: - Per-file keys, like used in v1 policies except for the new KDF. - Per-mode keys. These implement the semantics of the DIRECT_KEY flag, which for v1 policies made the master key be used directly. These are also planned to be used for inline encryption when support for it is added. - Key identifiers (see below). - Each master key is identified by a 16-byte master_key_identifier, which is derived from the key itself using HKDF-SHA512. This prevents users from associating the wrong key with an encrypted file or directory. This was easily possible with v1 policies, which identified the key by an arbitrary 8-byte master_key_descriptor. - The key must be provided in the filesystem-level keyring, not in a process-subscribed keyring. The following UAPI additions are made: - The existing ioctl FS_IOC_SET_ENCRYPTION_POLICY can now be passed a fscrypt_policy_v2 to set a v2 encryption policy. It's disambiguated from fscrypt_policy/fscrypt_policy_v1 by the version code prefix. - A new ioctl FS_IOC_GET_ENCRYPTION_POLICY_EX is added. It allows getting the v1 or v2 encryption policy of an encrypted file or directory. The existing FS_IOC_GET_ENCRYPTION_POLICY ioctl could not be used because it did not have a way for userspace to indicate which policy structure is expected. The new ioctl includes a size field, so it is extensible to future fscrypt policy versions. - The ioctls FS_IOC_ADD_ENCRYPTION_KEY, FS_IOC_REMOVE_ENCRYPTION_KEY, and FS_IOC_GET_ENCRYPTION_KEY_STATUS now support managing keys for v2 encryption policies. Such keys are kept logically separate from keys for v1 encryption policies, and are identified by 'identifier' rather than by 'descriptor'. The 'identifier' need not be provided when adding a key, since the kernel will calculate it anyway. This patch temporarily keeps adding/removing v2 policy keys behind the same permission check done for adding/removing v1 policy keys: capable(CAP_SYS_ADMIN). However, the next patch will carefully take advantage of the cryptographically secure master_key_identifier to allow non-root users to add/remove v2 policy keys, thus providing a full replacement for v1 policies. (*) Actually, in the API fscrypt_policy::version is 0 while on-disk fscrypt_context::format is 1. But I believe it makes the most sense to advance both to '2' to have them be in sync, and to consider the numbering to start at 1 except for the API quirk. Reviewed-by: Paul Crowley <paulcrowley@google.com> Reviewed-by: Theodore Ts'o <tytso@mit.edu> Signed-off-by: Eric Biggers <ebiggers@google.com>
2019-08-05 10:35:47 +08:00
fscrypt: remove kernel-internal constants from UAPI header There isn't really any valid reason to use __FSCRYPT_MODE_MAX or FSCRYPT_POLICY_FLAGS_VALID in a userspace program. These constants are only meant to be used by the kernel internally, and they are defined in the UAPI header next to the mode numbers and flags only so that kernel developers don't forget to update them when adding new modes or flags. In https://lkml.kernel.org/r/20201005074133.1958633-2-satyat@google.com there was an example of someone wanting to use __FSCRYPT_MODE_MAX in a user program, and it was wrong because the program would have broken if __FSCRYPT_MODE_MAX were ever increased. So having this definition available is harmful. FSCRYPT_POLICY_FLAGS_VALID has the same problem. So, remove these definitions from the UAPI header. Replace FSCRYPT_POLICY_FLAGS_VALID with just listing the valid flags explicitly in the one kernel function that needs it. Move __FSCRYPT_MODE_MAX to fscrypt_private.h, remove the double underscores (which were only present to discourage use by userspace), and add a BUILD_BUG_ON() and comments to (hopefully) ensure it is kept in sync. Keep the old name FS_POLICY_FLAGS_VALID, since it's been around for longer and there's a greater chance that removing it would break source compatibility with some program. Indeed, mtd-utils is using it in an #ifdef, and removing it would introduce compiler warnings (about FS_POLICY_FLAGS_PAD_* being redefined) into the mtd-utils build. However, reduce its value to 0x07 so that it only includes the flags with old names (the ones present before Linux 5.4), and try to make it clear that it's now "frozen" and no new flags should be added to it. Fixes: 2336d0deb2d4 ("fscrypt: use FSCRYPT_ prefix for uapi constants") Cc: <stable@vger.kernel.org> # v5.4+ Link: https://lore.kernel.org/r/20201024005132.495952-1-ebiggers@kernel.org Signed-off-by: Eric Biggers <ebiggers@google.com>
2020-10-24 08:51:31 +08:00
for (i = 0; i <= FSCRYPT_MODE_MAX; i++) {
fscrypt: stop holding extra request_queue references Now that the fscrypt_master_key lifetime has been reworked to not be subject to the quirks of the keyrings subsystem, blk_crypto_evict_key() no longer gets called after the filesystem has already been unmounted. Therefore, there is no longer any need to hold extra references to the filesystem's request_queue(s). (And these references didn't always do their intended job anyway, as pinning a request_queue doesn't necessarily pin the corresponding blk_crypto_profile.) Stop taking these extra references. Instead, just pass the super_block to fscrypt_destroy_inline_crypt_key(), and use it to get the list of block devices the key needs to be evicted from. Signed-off-by: Eric Biggers <ebiggers@google.com> Link: https://lore.kernel.org/r/20220901193208.138056-3-ebiggers@kernel.org
2022-09-02 03:32:07 +08:00
fscrypt_destroy_prepared_key(
sb, &mk->mk_direct_keys[i]);
fscrypt_destroy_prepared_key(
sb, &mk->mk_iv_ino_lblk_64_keys[i]);
fscrypt_destroy_prepared_key(
sb, &mk->mk_iv_ino_lblk_32_keys[i]);
fscrypt: add support for IV_INO_LBLK_64 policies Inline encryption hardware compliant with the UFS v2.1 standard or with the upcoming version of the eMMC standard has the following properties: (1) Per I/O request, the encryption key is specified by a previously loaded keyslot. There might be only a small number of keyslots. (2) Per I/O request, the starting IV is specified by a 64-bit "data unit number" (DUN). IV bits 64-127 are assumed to be 0. The hardware automatically increments the DUN for each "data unit" of configurable size in the request, e.g. for each filesystem block. Property (1) makes it inefficient to use the traditional fscrypt per-file keys. Property (2) precludes the use of the existing DIRECT_KEY fscrypt policy flag, which needs at least 192 IV bits. Therefore, add a new fscrypt policy flag IV_INO_LBLK_64 which causes the encryption to modified as follows: - The encryption keys are derived from the master key, encryption mode number, and filesystem UUID. - The IVs are chosen as (inode_number << 32) | file_logical_block_num. For filenames encryption, file_logical_block_num is 0. Since the file nonces aren't used in the key derivation, many files may share the same encryption key. This is much more efficient on the target hardware. Including the inode number in the IVs and mixing the filesystem UUID into the keys ensures that data in different files is nevertheless still encrypted differently. Additionally, limiting the inode and block numbers to 32 bits and placing the block number in the low bits maintains compatibility with the 64-bit DUN convention (property (2) above). Since this scheme assumes that inode numbers are stable (which may preclude filesystem shrinking) and that inode and file logical block numbers are at most 32-bit, IV_INO_LBLK_64 will only be allowed on filesystems that meet these constraints. These are acceptable limitations for the cases where this format would actually be used. Note that IV_INO_LBLK_64 is an on-disk format, not an implementation. This patch just adds support for it using the existing filesystem layer encryption. A later patch will add support for inline encryption. Reviewed-by: Paul Crowley <paulcrowley@google.com> Co-developed-by: Satya Tangirala <satyat@google.com> Signed-off-by: Satya Tangirala <satyat@google.com> Signed-off-by: Eric Biggers <ebiggers@google.com>
2019-10-25 05:54:36 +08:00
}
fscrypt: stop using keyrings subsystem for fscrypt_master_key The approach of fs/crypto/ internally managing the fscrypt_master_key structs as the payloads of "struct key" objects contained in a "struct key" keyring has outlived its usefulness. The original idea was to simplify the code by reusing code from the keyrings subsystem. However, several issues have arisen that can't easily be resolved: - When a master key struct is destroyed, blk_crypto_evict_key() must be called on any per-mode keys embedded in it. (This started being the case when inline encryption support was added.) Yet, the keyrings subsystem can arbitrarily delay the destruction of keys, even past the time the filesystem was unmounted. Therefore, currently there is no easy way to call blk_crypto_evict_key() when a master key is destroyed. Currently, this is worked around by holding an extra reference to the filesystem's request_queue(s). But it was overlooked that the request_queue reference is *not* guaranteed to pin the corresponding blk_crypto_profile too; for device-mapper devices that support inline crypto, it doesn't. This can cause a use-after-free. - When the last inode that was using an incompletely-removed master key is evicted, the master key removal is completed by removing the key struct from the keyring. Currently this is done via key_invalidate(). Yet, key_invalidate() takes the key semaphore. This can deadlock when called from the shrinker, since in fscrypt_ioctl_add_key(), memory is allocated with GFP_KERNEL under the same semaphore. - More generally, the fact that the keyrings subsystem can arbitrarily delay the destruction of keys (via garbage collection delay, or via random processes getting temporary key references) is undesirable, as it means we can't strictly guarantee that all secrets are ever wiped. - Doing the master key lookups via the keyrings subsystem results in the key_permission LSM hook being called. fscrypt doesn't want this, as all access control for encrypted files is designed to happen via the files themselves, like any other files. The workaround which SELinux users are using is to change their SELinux policy to grant key search access to all domains. This works, but it is an odd extra step that shouldn't really have to be done. The fix for all these issues is to change the implementation to what I should have done originally: don't use the keyrings subsystem to keep track of the filesystem's fscrypt_master_key structs. Instead, just store them in a regular kernel data structure, and rework the reference counting, locking, and lifetime accordingly. Retain support for RCU-mode key lookups by using a hash table. Replace fscrypt_sb_free() with fscrypt_sb_delete(), which releases the keys synchronously and runs a bit earlier during unmount, so that block devices are still available. A side effect of this patch is that neither the master keys themselves nor the filesystem keyrings will be listed in /proc/keys anymore. ("Master key users" and the master key users keyrings will still be listed.) However, this was mostly an implementation detail, and it was intended just for debugging purposes. I don't know of anyone using it. This patch does *not* change how "master key users" (->mk_users) works; that still uses the keyrings subsystem. That is still needed for key quotas, and changing that isn't necessary to solve the issues listed above. If we decide to change that too, it would be a separate patch. I've marked this as fixing the original commit that added the fscrypt keyring, but as noted above the most important issue that this patch fixes wasn't introduced until the addition of inline encryption support. Fixes: 22d94f493bfb ("fscrypt: add FS_IOC_ADD_ENCRYPTION_KEY ioctl") Signed-off-by: Eric Biggers <ebiggers@google.com> Link: https://lore.kernel.org/r/20220901193208.138056-2-ebiggers@kernel.org
2022-09-02 03:32:06 +08:00
memzero_explicit(&mk->mk_ino_hash_key,
sizeof(mk->mk_ino_hash_key));
mk->mk_ino_hash_key_initialized = false;
fscrypt: v2 encryption policy support Add a new fscrypt policy version, "v2". It has the following changes from the original policy version, which we call "v1" (*): - Master keys (the user-provided encryption keys) are only ever used as input to HKDF-SHA512. This is more flexible and less error-prone, and it avoids the quirks and limitations of the AES-128-ECB based KDF. Three classes of cryptographically isolated subkeys are defined: - Per-file keys, like used in v1 policies except for the new KDF. - Per-mode keys. These implement the semantics of the DIRECT_KEY flag, which for v1 policies made the master key be used directly. These are also planned to be used for inline encryption when support for it is added. - Key identifiers (see below). - Each master key is identified by a 16-byte master_key_identifier, which is derived from the key itself using HKDF-SHA512. This prevents users from associating the wrong key with an encrypted file or directory. This was easily possible with v1 policies, which identified the key by an arbitrary 8-byte master_key_descriptor. - The key must be provided in the filesystem-level keyring, not in a process-subscribed keyring. The following UAPI additions are made: - The existing ioctl FS_IOC_SET_ENCRYPTION_POLICY can now be passed a fscrypt_policy_v2 to set a v2 encryption policy. It's disambiguated from fscrypt_policy/fscrypt_policy_v1 by the version code prefix. - A new ioctl FS_IOC_GET_ENCRYPTION_POLICY_EX is added. It allows getting the v1 or v2 encryption policy of an encrypted file or directory. The existing FS_IOC_GET_ENCRYPTION_POLICY ioctl could not be used because it did not have a way for userspace to indicate which policy structure is expected. The new ioctl includes a size field, so it is extensible to future fscrypt policy versions. - The ioctls FS_IOC_ADD_ENCRYPTION_KEY, FS_IOC_REMOVE_ENCRYPTION_KEY, and FS_IOC_GET_ENCRYPTION_KEY_STATUS now support managing keys for v2 encryption policies. Such keys are kept logically separate from keys for v1 encryption policies, and are identified by 'identifier' rather than by 'descriptor'. The 'identifier' need not be provided when adding a key, since the kernel will calculate it anyway. This patch temporarily keeps adding/removing v2 policy keys behind the same permission check done for adding/removing v1 policy keys: capable(CAP_SYS_ADMIN). However, the next patch will carefully take advantage of the cryptographically secure master_key_identifier to allow non-root users to add/remove v2 policy keys, thus providing a full replacement for v1 policies. (*) Actually, in the API fscrypt_policy::version is 0 while on-disk fscrypt_context::format is 1. But I believe it makes the most sense to advance both to '2' to have them be in sync, and to consider the numbering to start at 1 except for the API quirk. Reviewed-by: Paul Crowley <paulcrowley@google.com> Reviewed-by: Theodore Ts'o <tytso@mit.edu> Signed-off-by: Eric Biggers <ebiggers@google.com>
2019-08-05 10:35:47 +08:00
fscrypt: stop using keyrings subsystem for fscrypt_master_key The approach of fs/crypto/ internally managing the fscrypt_master_key structs as the payloads of "struct key" objects contained in a "struct key" keyring has outlived its usefulness. The original idea was to simplify the code by reusing code from the keyrings subsystem. However, several issues have arisen that can't easily be resolved: - When a master key struct is destroyed, blk_crypto_evict_key() must be called on any per-mode keys embedded in it. (This started being the case when inline encryption support was added.) Yet, the keyrings subsystem can arbitrarily delay the destruction of keys, even past the time the filesystem was unmounted. Therefore, currently there is no easy way to call blk_crypto_evict_key() when a master key is destroyed. Currently, this is worked around by holding an extra reference to the filesystem's request_queue(s). But it was overlooked that the request_queue reference is *not* guaranteed to pin the corresponding blk_crypto_profile too; for device-mapper devices that support inline crypto, it doesn't. This can cause a use-after-free. - When the last inode that was using an incompletely-removed master key is evicted, the master key removal is completed by removing the key struct from the keyring. Currently this is done via key_invalidate(). Yet, key_invalidate() takes the key semaphore. This can deadlock when called from the shrinker, since in fscrypt_ioctl_add_key(), memory is allocated with GFP_KERNEL under the same semaphore. - More generally, the fact that the keyrings subsystem can arbitrarily delay the destruction of keys (via garbage collection delay, or via random processes getting temporary key references) is undesirable, as it means we can't strictly guarantee that all secrets are ever wiped. - Doing the master key lookups via the keyrings subsystem results in the key_permission LSM hook being called. fscrypt doesn't want this, as all access control for encrypted files is designed to happen via the files themselves, like any other files. The workaround which SELinux users are using is to change their SELinux policy to grant key search access to all domains. This works, but it is an odd extra step that shouldn't really have to be done. The fix for all these issues is to change the implementation to what I should have done originally: don't use the keyrings subsystem to keep track of the filesystem's fscrypt_master_key structs. Instead, just store them in a regular kernel data structure, and rework the reference counting, locking, and lifetime accordingly. Retain support for RCU-mode key lookups by using a hash table. Replace fscrypt_sb_free() with fscrypt_sb_delete(), which releases the keys synchronously and runs a bit earlier during unmount, so that block devices are still available. A side effect of this patch is that neither the master keys themselves nor the filesystem keyrings will be listed in /proc/keys anymore. ("Master key users" and the master key users keyrings will still be listed.) However, this was mostly an implementation detail, and it was intended just for debugging purposes. I don't know of anyone using it. This patch does *not* change how "master key users" (->mk_users) works; that still uses the keyrings subsystem. That is still needed for key quotas, and changing that isn't necessary to solve the issues listed above. If we decide to change that too, it would be a separate patch. I've marked this as fixing the original commit that added the fscrypt keyring, but as noted above the most important issue that this patch fixes wasn't introduced until the addition of inline encryption support. Fixes: 22d94f493bfb ("fscrypt: add FS_IOC_ADD_ENCRYPTION_KEY ioctl") Signed-off-by: Eric Biggers <ebiggers@google.com> Link: https://lore.kernel.org/r/20220901193208.138056-2-ebiggers@kernel.org
2022-09-02 03:32:06 +08:00
/* Drop the structural ref associated with the active refs. */
fscrypt_put_master_key(mk);
fscrypt: add FS_IOC_ADD_ENCRYPTION_KEY ioctl Add a new fscrypt ioctl, FS_IOC_ADD_ENCRYPTION_KEY. This ioctl adds an encryption key to the filesystem's fscrypt keyring ->s_master_keys, making any files encrypted with that key appear "unlocked". Why we need this ~~~~~~~~~~~~~~~~ The main problem is that the "locked/unlocked" (ciphertext/plaintext) status of encrypted files is global, but the fscrypt keys are not. fscrypt only looks for keys in the keyring(s) the process accessing the filesystem is subscribed to: the thread keyring, process keyring, and session keyring, where the session keyring may contain the user keyring. Therefore, userspace has to put fscrypt keys in the keyrings for individual users or sessions. But this means that when a process with a different keyring tries to access encrypted files, whether they appear "unlocked" or not is nondeterministic. This is because it depends on whether the files are currently present in the inode cache. Fixing this by consistently providing each process its own view of the filesystem depending on whether it has the key or not isn't feasible due to how the VFS caches work. Furthermore, while sometimes users expect this behavior, it is misguided for two reasons. First, it would be an OS-level access control mechanism largely redundant with existing access control mechanisms such as UNIX file permissions, ACLs, LSMs, etc. Encryption is actually for protecting the data at rest. Second, almost all users of fscrypt actually do need the keys to be global. The largest users of fscrypt, Android and Chromium OS, achieve this by having PID 1 create a "session keyring" that is inherited by every process. This works, but it isn't scalable because it prevents session keyrings from being used for any other purpose. On general-purpose Linux distros, the 'fscrypt' userspace tool [1] can't similarly abuse the session keyring, so to make 'sudo' work on all systems it has to link all the user keyrings into root's user keyring [2]. This is ugly and raises security concerns. Moreover it can't make the keys available to system services, such as sshd trying to access the user's '~/.ssh' directory (see [3], [4]) or NetworkManager trying to read certificates from the user's home directory (see [5]); or to Docker containers (see [6], [7]). By having an API to add a key to the *filesystem* we'll be able to fix the above bugs, remove userspace workarounds, and clearly express the intended semantics: the locked/unlocked status of an encrypted directory is global, and encryption is orthogonal to OS-level access control. Why not use the add_key() syscall ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ We use an ioctl for this API rather than the existing add_key() system call because the ioctl gives us the flexibility needed to implement fscrypt-specific semantics that will be introduced in later patches: - Supporting key removal with the semantics such that the secret is removed immediately and any unused inodes using the key are evicted; also, the eviction of any in-use inodes can be retried. - Calculating a key-dependent cryptographic identifier and returning it to userspace. - Allowing keys to be added and removed by non-root users, but only keys for v2 encryption policies; and to prevent denial-of-service attacks, users can only remove keys they themselves have added, and a key is only really removed after all users who added it have removed it. Trying to shoehorn these semantics into the keyrings syscalls would be very difficult, whereas the ioctls make things much easier. However, to reuse code the implementation still uses the keyrings service internally. Thus we get lockless RCU-mode key lookups without having to re-implement it, and the keys automatically show up in /proc/keys for debugging purposes. References: [1] https://github.com/google/fscrypt [2] https://goo.gl/55cCrI#heading=h.vf09isp98isb [3] https://github.com/google/fscrypt/issues/111#issuecomment-444347939 [4] https://github.com/google/fscrypt/issues/116 [5] https://bugs.launchpad.net/ubuntu/+source/fscrypt/+bug/1770715 [6] https://github.com/google/fscrypt/issues/128 [7] https://askubuntu.com/questions/1130306/cannot-run-docker-on-an-encrypted-filesystem Reviewed-by: Theodore Ts'o <tytso@mit.edu> Signed-off-by: Eric Biggers <ebiggers@google.com>
2019-08-05 10:35:46 +08:00
}
static inline bool valid_key_spec(const struct fscrypt_key_specifier *spec)
{
if (spec->__reserved)
return false;
return master_key_spec_len(spec) != 0;
}
fscrypt: allow unprivileged users to add/remove keys for v2 policies Allow the FS_IOC_ADD_ENCRYPTION_KEY and FS_IOC_REMOVE_ENCRYPTION_KEY ioctls to be used by non-root users to add and remove encryption keys from the filesystem-level crypto keyrings, subject to limitations. Motivation: while privileged fscrypt key management is sufficient for some users (e.g. Android and Chromium OS, where a privileged process manages all keys), the old API by design also allows non-root users to set up and use encrypted directories, and we don't want to regress on that. Especially, we don't want to force users to continue using the old API, running into the visibility mismatch between files and keyrings and being unable to "lock" encrypted directories. Intuitively, the ioctls have to be privileged since they manipulate filesystem-level state. However, it's actually safe to make them unprivileged if we very carefully enforce some specific limitations. First, each key must be identified by a cryptographic hash so that a user can't add the wrong key for another user's files. For v2 encryption policies, we use the key_identifier for this. v1 policies don't have this, so managing keys for them remains privileged. Second, each key a user adds is charged to their quota for the keyrings service. Thus, a user can't exhaust memory by adding a huge number of keys. By default each non-root user is allowed up to 200 keys; this can be changed using the existing sysctl 'kernel.keys.maxkeys'. Third, if multiple users add the same key, we keep track of those users of the key (of which there remains a single copy), and won't really remove the key, i.e. "lock" the encrypted files, until all those users have removed it. This prevents denial of service attacks that would be possible under simpler schemes, such allowing the first user who added a key to remove it -- since that could be a malicious user who has compromised the key. Of course, encryption keys should be kept secret, but the idea is that using encryption should never be *less* secure than not using encryption, even if your key was compromised. We tolerate that a user will be unable to really remove a key, i.e. unable to "lock" their encrypted files, if another user has added the same key. But in a sense, this is actually a good thing because it will avoid providing a false notion of security where a key appears to have been removed when actually it's still in memory, available to any attacker who compromises the operating system kernel. Reviewed-by: Theodore Ts'o <tytso@mit.edu> Signed-off-by: Eric Biggers <ebiggers@google.com>
2019-08-05 10:35:47 +08:00
static int fscrypt_user_key_instantiate(struct key *key,
struct key_preparsed_payload *prep)
{
/*
* We just charge FSCRYPT_MAX_KEY_SIZE bytes to the user's key quota for
* each key, regardless of the exact key size. The amount of memory
* actually used is greater than the size of the raw key anyway.
*/
return key_payload_reserve(key, FSCRYPT_MAX_KEY_SIZE);
}
static void fscrypt_user_key_describe(const struct key *key, struct seq_file *m)
{
seq_puts(m, key->description);
}
/*
* Type of key in ->mk_users. Each key of this type represents a particular
* user who has added a particular master key.
*
* Note that the name of this key type really should be something like
* ".fscrypt-user" instead of simply ".fscrypt". But the shorter name is chosen
* mainly for simplicity of presentation in /proc/keys when read by a non-root
* user. And it is expected to be rare that a key is actually added by multiple
* users, since users should keep their encryption keys confidential.
*/
static struct key_type key_type_fscrypt_user = {
.name = ".fscrypt",
.instantiate = fscrypt_user_key_instantiate,
.describe = fscrypt_user_key_describe,
};
#define FSCRYPT_MK_USERS_DESCRIPTION_SIZE \
(CONST_STRLEN("fscrypt-") + 2 * FSCRYPT_KEY_IDENTIFIER_SIZE + \
CONST_STRLEN("-users") + 1)
#define FSCRYPT_MK_USER_DESCRIPTION_SIZE \
(2 * FSCRYPT_KEY_IDENTIFIER_SIZE + CONST_STRLEN(".uid.") + 10 + 1)
static void format_mk_users_keyring_description(
char description[FSCRYPT_MK_USERS_DESCRIPTION_SIZE],
const u8 mk_identifier[FSCRYPT_KEY_IDENTIFIER_SIZE])
{
sprintf(description, "fscrypt-%*phN-users",
FSCRYPT_KEY_IDENTIFIER_SIZE, mk_identifier);
}
static void format_mk_user_description(
char description[FSCRYPT_MK_USER_DESCRIPTION_SIZE],
const u8 mk_identifier[FSCRYPT_KEY_IDENTIFIER_SIZE])
{
sprintf(description, "%*phN.uid.%u", FSCRYPT_KEY_IDENTIFIER_SIZE,
mk_identifier, __kuid_val(current_fsuid()));
}
fscrypt: add FS_IOC_ADD_ENCRYPTION_KEY ioctl Add a new fscrypt ioctl, FS_IOC_ADD_ENCRYPTION_KEY. This ioctl adds an encryption key to the filesystem's fscrypt keyring ->s_master_keys, making any files encrypted with that key appear "unlocked". Why we need this ~~~~~~~~~~~~~~~~ The main problem is that the "locked/unlocked" (ciphertext/plaintext) status of encrypted files is global, but the fscrypt keys are not. fscrypt only looks for keys in the keyring(s) the process accessing the filesystem is subscribed to: the thread keyring, process keyring, and session keyring, where the session keyring may contain the user keyring. Therefore, userspace has to put fscrypt keys in the keyrings for individual users or sessions. But this means that when a process with a different keyring tries to access encrypted files, whether they appear "unlocked" or not is nondeterministic. This is because it depends on whether the files are currently present in the inode cache. Fixing this by consistently providing each process its own view of the filesystem depending on whether it has the key or not isn't feasible due to how the VFS caches work. Furthermore, while sometimes users expect this behavior, it is misguided for two reasons. First, it would be an OS-level access control mechanism largely redundant with existing access control mechanisms such as UNIX file permissions, ACLs, LSMs, etc. Encryption is actually for protecting the data at rest. Second, almost all users of fscrypt actually do need the keys to be global. The largest users of fscrypt, Android and Chromium OS, achieve this by having PID 1 create a "session keyring" that is inherited by every process. This works, but it isn't scalable because it prevents session keyrings from being used for any other purpose. On general-purpose Linux distros, the 'fscrypt' userspace tool [1] can't similarly abuse the session keyring, so to make 'sudo' work on all systems it has to link all the user keyrings into root's user keyring [2]. This is ugly and raises security concerns. Moreover it can't make the keys available to system services, such as sshd trying to access the user's '~/.ssh' directory (see [3], [4]) or NetworkManager trying to read certificates from the user's home directory (see [5]); or to Docker containers (see [6], [7]). By having an API to add a key to the *filesystem* we'll be able to fix the above bugs, remove userspace workarounds, and clearly express the intended semantics: the locked/unlocked status of an encrypted directory is global, and encryption is orthogonal to OS-level access control. Why not use the add_key() syscall ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ We use an ioctl for this API rather than the existing add_key() system call because the ioctl gives us the flexibility needed to implement fscrypt-specific semantics that will be introduced in later patches: - Supporting key removal with the semantics such that the secret is removed immediately and any unused inodes using the key are evicted; also, the eviction of any in-use inodes can be retried. - Calculating a key-dependent cryptographic identifier and returning it to userspace. - Allowing keys to be added and removed by non-root users, but only keys for v2 encryption policies; and to prevent denial-of-service attacks, users can only remove keys they themselves have added, and a key is only really removed after all users who added it have removed it. Trying to shoehorn these semantics into the keyrings syscalls would be very difficult, whereas the ioctls make things much easier. However, to reuse code the implementation still uses the keyrings service internally. Thus we get lockless RCU-mode key lookups without having to re-implement it, and the keys automatically show up in /proc/keys for debugging purposes. References: [1] https://github.com/google/fscrypt [2] https://goo.gl/55cCrI#heading=h.vf09isp98isb [3] https://github.com/google/fscrypt/issues/111#issuecomment-444347939 [4] https://github.com/google/fscrypt/issues/116 [5] https://bugs.launchpad.net/ubuntu/+source/fscrypt/+bug/1770715 [6] https://github.com/google/fscrypt/issues/128 [7] https://askubuntu.com/questions/1130306/cannot-run-docker-on-an-encrypted-filesystem Reviewed-by: Theodore Ts'o <tytso@mit.edu> Signed-off-by: Eric Biggers <ebiggers@google.com>
2019-08-05 10:35:46 +08:00
/* Create ->s_master_keys if needed. Synchronized by fscrypt_add_key_mutex. */
static int allocate_filesystem_keyring(struct super_block *sb)
{
fscrypt: stop using keyrings subsystem for fscrypt_master_key The approach of fs/crypto/ internally managing the fscrypt_master_key structs as the payloads of "struct key" objects contained in a "struct key" keyring has outlived its usefulness. The original idea was to simplify the code by reusing code from the keyrings subsystem. However, several issues have arisen that can't easily be resolved: - When a master key struct is destroyed, blk_crypto_evict_key() must be called on any per-mode keys embedded in it. (This started being the case when inline encryption support was added.) Yet, the keyrings subsystem can arbitrarily delay the destruction of keys, even past the time the filesystem was unmounted. Therefore, currently there is no easy way to call blk_crypto_evict_key() when a master key is destroyed. Currently, this is worked around by holding an extra reference to the filesystem's request_queue(s). But it was overlooked that the request_queue reference is *not* guaranteed to pin the corresponding blk_crypto_profile too; for device-mapper devices that support inline crypto, it doesn't. This can cause a use-after-free. - When the last inode that was using an incompletely-removed master key is evicted, the master key removal is completed by removing the key struct from the keyring. Currently this is done via key_invalidate(). Yet, key_invalidate() takes the key semaphore. This can deadlock when called from the shrinker, since in fscrypt_ioctl_add_key(), memory is allocated with GFP_KERNEL under the same semaphore. - More generally, the fact that the keyrings subsystem can arbitrarily delay the destruction of keys (via garbage collection delay, or via random processes getting temporary key references) is undesirable, as it means we can't strictly guarantee that all secrets are ever wiped. - Doing the master key lookups via the keyrings subsystem results in the key_permission LSM hook being called. fscrypt doesn't want this, as all access control for encrypted files is designed to happen via the files themselves, like any other files. The workaround which SELinux users are using is to change their SELinux policy to grant key search access to all domains. This works, but it is an odd extra step that shouldn't really have to be done. The fix for all these issues is to change the implementation to what I should have done originally: don't use the keyrings subsystem to keep track of the filesystem's fscrypt_master_key structs. Instead, just store them in a regular kernel data structure, and rework the reference counting, locking, and lifetime accordingly. Retain support for RCU-mode key lookups by using a hash table. Replace fscrypt_sb_free() with fscrypt_sb_delete(), which releases the keys synchronously and runs a bit earlier during unmount, so that block devices are still available. A side effect of this patch is that neither the master keys themselves nor the filesystem keyrings will be listed in /proc/keys anymore. ("Master key users" and the master key users keyrings will still be listed.) However, this was mostly an implementation detail, and it was intended just for debugging purposes. I don't know of anyone using it. This patch does *not* change how "master key users" (->mk_users) works; that still uses the keyrings subsystem. That is still needed for key quotas, and changing that isn't necessary to solve the issues listed above. If we decide to change that too, it would be a separate patch. I've marked this as fixing the original commit that added the fscrypt keyring, but as noted above the most important issue that this patch fixes wasn't introduced until the addition of inline encryption support. Fixes: 22d94f493bfb ("fscrypt: add FS_IOC_ADD_ENCRYPTION_KEY ioctl") Signed-off-by: Eric Biggers <ebiggers@google.com> Link: https://lore.kernel.org/r/20220901193208.138056-2-ebiggers@kernel.org
2022-09-02 03:32:06 +08:00
struct fscrypt_keyring *keyring;
fscrypt: add FS_IOC_ADD_ENCRYPTION_KEY ioctl Add a new fscrypt ioctl, FS_IOC_ADD_ENCRYPTION_KEY. This ioctl adds an encryption key to the filesystem's fscrypt keyring ->s_master_keys, making any files encrypted with that key appear "unlocked". Why we need this ~~~~~~~~~~~~~~~~ The main problem is that the "locked/unlocked" (ciphertext/plaintext) status of encrypted files is global, but the fscrypt keys are not. fscrypt only looks for keys in the keyring(s) the process accessing the filesystem is subscribed to: the thread keyring, process keyring, and session keyring, where the session keyring may contain the user keyring. Therefore, userspace has to put fscrypt keys in the keyrings for individual users or sessions. But this means that when a process with a different keyring tries to access encrypted files, whether they appear "unlocked" or not is nondeterministic. This is because it depends on whether the files are currently present in the inode cache. Fixing this by consistently providing each process its own view of the filesystem depending on whether it has the key or not isn't feasible due to how the VFS caches work. Furthermore, while sometimes users expect this behavior, it is misguided for two reasons. First, it would be an OS-level access control mechanism largely redundant with existing access control mechanisms such as UNIX file permissions, ACLs, LSMs, etc. Encryption is actually for protecting the data at rest. Second, almost all users of fscrypt actually do need the keys to be global. The largest users of fscrypt, Android and Chromium OS, achieve this by having PID 1 create a "session keyring" that is inherited by every process. This works, but it isn't scalable because it prevents session keyrings from being used for any other purpose. On general-purpose Linux distros, the 'fscrypt' userspace tool [1] can't similarly abuse the session keyring, so to make 'sudo' work on all systems it has to link all the user keyrings into root's user keyring [2]. This is ugly and raises security concerns. Moreover it can't make the keys available to system services, such as sshd trying to access the user's '~/.ssh' directory (see [3], [4]) or NetworkManager trying to read certificates from the user's home directory (see [5]); or to Docker containers (see [6], [7]). By having an API to add a key to the *filesystem* we'll be able to fix the above bugs, remove userspace workarounds, and clearly express the intended semantics: the locked/unlocked status of an encrypted directory is global, and encryption is orthogonal to OS-level access control. Why not use the add_key() syscall ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ We use an ioctl for this API rather than the existing add_key() system call because the ioctl gives us the flexibility needed to implement fscrypt-specific semantics that will be introduced in later patches: - Supporting key removal with the semantics such that the secret is removed immediately and any unused inodes using the key are evicted; also, the eviction of any in-use inodes can be retried. - Calculating a key-dependent cryptographic identifier and returning it to userspace. - Allowing keys to be added and removed by non-root users, but only keys for v2 encryption policies; and to prevent denial-of-service attacks, users can only remove keys they themselves have added, and a key is only really removed after all users who added it have removed it. Trying to shoehorn these semantics into the keyrings syscalls would be very difficult, whereas the ioctls make things much easier. However, to reuse code the implementation still uses the keyrings service internally. Thus we get lockless RCU-mode key lookups without having to re-implement it, and the keys automatically show up in /proc/keys for debugging purposes. References: [1] https://github.com/google/fscrypt [2] https://goo.gl/55cCrI#heading=h.vf09isp98isb [3] https://github.com/google/fscrypt/issues/111#issuecomment-444347939 [4] https://github.com/google/fscrypt/issues/116 [5] https://bugs.launchpad.net/ubuntu/+source/fscrypt/+bug/1770715 [6] https://github.com/google/fscrypt/issues/128 [7] https://askubuntu.com/questions/1130306/cannot-run-docker-on-an-encrypted-filesystem Reviewed-by: Theodore Ts'o <tytso@mit.edu> Signed-off-by: Eric Biggers <ebiggers@google.com>
2019-08-05 10:35:46 +08:00
if (sb->s_master_keys)
return 0;
fscrypt: stop using keyrings subsystem for fscrypt_master_key The approach of fs/crypto/ internally managing the fscrypt_master_key structs as the payloads of "struct key" objects contained in a "struct key" keyring has outlived its usefulness. The original idea was to simplify the code by reusing code from the keyrings subsystem. However, several issues have arisen that can't easily be resolved: - When a master key struct is destroyed, blk_crypto_evict_key() must be called on any per-mode keys embedded in it. (This started being the case when inline encryption support was added.) Yet, the keyrings subsystem can arbitrarily delay the destruction of keys, even past the time the filesystem was unmounted. Therefore, currently there is no easy way to call blk_crypto_evict_key() when a master key is destroyed. Currently, this is worked around by holding an extra reference to the filesystem's request_queue(s). But it was overlooked that the request_queue reference is *not* guaranteed to pin the corresponding blk_crypto_profile too; for device-mapper devices that support inline crypto, it doesn't. This can cause a use-after-free. - When the last inode that was using an incompletely-removed master key is evicted, the master key removal is completed by removing the key struct from the keyring. Currently this is done via key_invalidate(). Yet, key_invalidate() takes the key semaphore. This can deadlock when called from the shrinker, since in fscrypt_ioctl_add_key(), memory is allocated with GFP_KERNEL under the same semaphore. - More generally, the fact that the keyrings subsystem can arbitrarily delay the destruction of keys (via garbage collection delay, or via random processes getting temporary key references) is undesirable, as it means we can't strictly guarantee that all secrets are ever wiped. - Doing the master key lookups via the keyrings subsystem results in the key_permission LSM hook being called. fscrypt doesn't want this, as all access control for encrypted files is designed to happen via the files themselves, like any other files. The workaround which SELinux users are using is to change their SELinux policy to grant key search access to all domains. This works, but it is an odd extra step that shouldn't really have to be done. The fix for all these issues is to change the implementation to what I should have done originally: don't use the keyrings subsystem to keep track of the filesystem's fscrypt_master_key structs. Instead, just store them in a regular kernel data structure, and rework the reference counting, locking, and lifetime accordingly. Retain support for RCU-mode key lookups by using a hash table. Replace fscrypt_sb_free() with fscrypt_sb_delete(), which releases the keys synchronously and runs a bit earlier during unmount, so that block devices are still available. A side effect of this patch is that neither the master keys themselves nor the filesystem keyrings will be listed in /proc/keys anymore. ("Master key users" and the master key users keyrings will still be listed.) However, this was mostly an implementation detail, and it was intended just for debugging purposes. I don't know of anyone using it. This patch does *not* change how "master key users" (->mk_users) works; that still uses the keyrings subsystem. That is still needed for key quotas, and changing that isn't necessary to solve the issues listed above. If we decide to change that too, it would be a separate patch. I've marked this as fixing the original commit that added the fscrypt keyring, but as noted above the most important issue that this patch fixes wasn't introduced until the addition of inline encryption support. Fixes: 22d94f493bfb ("fscrypt: add FS_IOC_ADD_ENCRYPTION_KEY ioctl") Signed-off-by: Eric Biggers <ebiggers@google.com> Link: https://lore.kernel.org/r/20220901193208.138056-2-ebiggers@kernel.org
2022-09-02 03:32:06 +08:00
keyring = kzalloc(sizeof(*keyring), GFP_KERNEL);
if (!keyring)
return -ENOMEM;
spin_lock_init(&keyring->lock);
fscrypt: use smp_load_acquire() for ->s_master_keys Normally smp_store_release() or cmpxchg_release() is paired with smp_load_acquire(). Sometimes smp_load_acquire() can be replaced with the more lightweight READ_ONCE(). However, for this to be safe, all the published memory must only be accessed in a way that involves the pointer itself. This may not be the case if allocating the object also involves initializing a static or global variable, for example. super_block::s_master_keys is a keyring, which is internal to and is allocated by the keyrings subsystem. By using READ_ONCE() for it, we're relying on internal implementation details of the keyrings subsystem. Remove this fragile assumption by using smp_load_acquire() instead. (Note: I haven't seen any real-world problems here. This change is just fixing the code to be guaranteed correct and less fragile.) Fixes: 22d94f493bfb ("fscrypt: add FS_IOC_ADD_ENCRYPTION_KEY ioctl") Link: https://lore.kernel.org/r/20200721225920.114347-4-ebiggers@kernel.org Signed-off-by: Eric Biggers <ebiggers@google.com>
2020-07-22 06:59:18 +08:00
/*
* Pairs with the smp_load_acquire() in fscrypt_find_master_key().
* I.e., here we publish ->s_master_keys with a RELEASE barrier so that
* concurrent tasks can ACQUIRE it.
*/
fscrypt: add FS_IOC_ADD_ENCRYPTION_KEY ioctl Add a new fscrypt ioctl, FS_IOC_ADD_ENCRYPTION_KEY. This ioctl adds an encryption key to the filesystem's fscrypt keyring ->s_master_keys, making any files encrypted with that key appear "unlocked". Why we need this ~~~~~~~~~~~~~~~~ The main problem is that the "locked/unlocked" (ciphertext/plaintext) status of encrypted files is global, but the fscrypt keys are not. fscrypt only looks for keys in the keyring(s) the process accessing the filesystem is subscribed to: the thread keyring, process keyring, and session keyring, where the session keyring may contain the user keyring. Therefore, userspace has to put fscrypt keys in the keyrings for individual users or sessions. But this means that when a process with a different keyring tries to access encrypted files, whether they appear "unlocked" or not is nondeterministic. This is because it depends on whether the files are currently present in the inode cache. Fixing this by consistently providing each process its own view of the filesystem depending on whether it has the key or not isn't feasible due to how the VFS caches work. Furthermore, while sometimes users expect this behavior, it is misguided for two reasons. First, it would be an OS-level access control mechanism largely redundant with existing access control mechanisms such as UNIX file permissions, ACLs, LSMs, etc. Encryption is actually for protecting the data at rest. Second, almost all users of fscrypt actually do need the keys to be global. The largest users of fscrypt, Android and Chromium OS, achieve this by having PID 1 create a "session keyring" that is inherited by every process. This works, but it isn't scalable because it prevents session keyrings from being used for any other purpose. On general-purpose Linux distros, the 'fscrypt' userspace tool [1] can't similarly abuse the session keyring, so to make 'sudo' work on all systems it has to link all the user keyrings into root's user keyring [2]. This is ugly and raises security concerns. Moreover it can't make the keys available to system services, such as sshd trying to access the user's '~/.ssh' directory (see [3], [4]) or NetworkManager trying to read certificates from the user's home directory (see [5]); or to Docker containers (see [6], [7]). By having an API to add a key to the *filesystem* we'll be able to fix the above bugs, remove userspace workarounds, and clearly express the intended semantics: the locked/unlocked status of an encrypted directory is global, and encryption is orthogonal to OS-level access control. Why not use the add_key() syscall ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ We use an ioctl for this API rather than the existing add_key() system call because the ioctl gives us the flexibility needed to implement fscrypt-specific semantics that will be introduced in later patches: - Supporting key removal with the semantics such that the secret is removed immediately and any unused inodes using the key are evicted; also, the eviction of any in-use inodes can be retried. - Calculating a key-dependent cryptographic identifier and returning it to userspace. - Allowing keys to be added and removed by non-root users, but only keys for v2 encryption policies; and to prevent denial-of-service attacks, users can only remove keys they themselves have added, and a key is only really removed after all users who added it have removed it. Trying to shoehorn these semantics into the keyrings syscalls would be very difficult, whereas the ioctls make things much easier. However, to reuse code the implementation still uses the keyrings service internally. Thus we get lockless RCU-mode key lookups without having to re-implement it, and the keys automatically show up in /proc/keys for debugging purposes. References: [1] https://github.com/google/fscrypt [2] https://goo.gl/55cCrI#heading=h.vf09isp98isb [3] https://github.com/google/fscrypt/issues/111#issuecomment-444347939 [4] https://github.com/google/fscrypt/issues/116 [5] https://bugs.launchpad.net/ubuntu/+source/fscrypt/+bug/1770715 [6] https://github.com/google/fscrypt/issues/128 [7] https://askubuntu.com/questions/1130306/cannot-run-docker-on-an-encrypted-filesystem Reviewed-by: Theodore Ts'o <tytso@mit.edu> Signed-off-by: Eric Biggers <ebiggers@google.com>
2019-08-05 10:35:46 +08:00
smp_store_release(&sb->s_master_keys, keyring);
return 0;
}
fscrypt: stop using keyrings subsystem for fscrypt_master_key The approach of fs/crypto/ internally managing the fscrypt_master_key structs as the payloads of "struct key" objects contained in a "struct key" keyring has outlived its usefulness. The original idea was to simplify the code by reusing code from the keyrings subsystem. However, several issues have arisen that can't easily be resolved: - When a master key struct is destroyed, blk_crypto_evict_key() must be called on any per-mode keys embedded in it. (This started being the case when inline encryption support was added.) Yet, the keyrings subsystem can arbitrarily delay the destruction of keys, even past the time the filesystem was unmounted. Therefore, currently there is no easy way to call blk_crypto_evict_key() when a master key is destroyed. Currently, this is worked around by holding an extra reference to the filesystem's request_queue(s). But it was overlooked that the request_queue reference is *not* guaranteed to pin the corresponding blk_crypto_profile too; for device-mapper devices that support inline crypto, it doesn't. This can cause a use-after-free. - When the last inode that was using an incompletely-removed master key is evicted, the master key removal is completed by removing the key struct from the keyring. Currently this is done via key_invalidate(). Yet, key_invalidate() takes the key semaphore. This can deadlock when called from the shrinker, since in fscrypt_ioctl_add_key(), memory is allocated with GFP_KERNEL under the same semaphore. - More generally, the fact that the keyrings subsystem can arbitrarily delay the destruction of keys (via garbage collection delay, or via random processes getting temporary key references) is undesirable, as it means we can't strictly guarantee that all secrets are ever wiped. - Doing the master key lookups via the keyrings subsystem results in the key_permission LSM hook being called. fscrypt doesn't want this, as all access control for encrypted files is designed to happen via the files themselves, like any other files. The workaround which SELinux users are using is to change their SELinux policy to grant key search access to all domains. This works, but it is an odd extra step that shouldn't really have to be done. The fix for all these issues is to change the implementation to what I should have done originally: don't use the keyrings subsystem to keep track of the filesystem's fscrypt_master_key structs. Instead, just store them in a regular kernel data structure, and rework the reference counting, locking, and lifetime accordingly. Retain support for RCU-mode key lookups by using a hash table. Replace fscrypt_sb_free() with fscrypt_sb_delete(), which releases the keys synchronously and runs a bit earlier during unmount, so that block devices are still available. A side effect of this patch is that neither the master keys themselves nor the filesystem keyrings will be listed in /proc/keys anymore. ("Master key users" and the master key users keyrings will still be listed.) However, this was mostly an implementation detail, and it was intended just for debugging purposes. I don't know of anyone using it. This patch does *not* change how "master key users" (->mk_users) works; that still uses the keyrings subsystem. That is still needed for key quotas, and changing that isn't necessary to solve the issues listed above. If we decide to change that too, it would be a separate patch. I've marked this as fixing the original commit that added the fscrypt keyring, but as noted above the most important issue that this patch fixes wasn't introduced until the addition of inline encryption support. Fixes: 22d94f493bfb ("fscrypt: add FS_IOC_ADD_ENCRYPTION_KEY ioctl") Signed-off-by: Eric Biggers <ebiggers@google.com> Link: https://lore.kernel.org/r/20220901193208.138056-2-ebiggers@kernel.org
2022-09-02 03:32:06 +08:00
/*
* This is called at unmount time to release all encryption keys that have been
* added to the filesystem, along with the keyring that contains them.
*
* Note that besides clearing and freeing memory, this might need to evict keys
* from the keyslots of an inline crypto engine. Therefore, this must be called
* while the filesystem's underlying block device(s) are still available.
*/
void fscrypt_sb_delete(struct super_block *sb)
fscrypt: add FS_IOC_ADD_ENCRYPTION_KEY ioctl Add a new fscrypt ioctl, FS_IOC_ADD_ENCRYPTION_KEY. This ioctl adds an encryption key to the filesystem's fscrypt keyring ->s_master_keys, making any files encrypted with that key appear "unlocked". Why we need this ~~~~~~~~~~~~~~~~ The main problem is that the "locked/unlocked" (ciphertext/plaintext) status of encrypted files is global, but the fscrypt keys are not. fscrypt only looks for keys in the keyring(s) the process accessing the filesystem is subscribed to: the thread keyring, process keyring, and session keyring, where the session keyring may contain the user keyring. Therefore, userspace has to put fscrypt keys in the keyrings for individual users or sessions. But this means that when a process with a different keyring tries to access encrypted files, whether they appear "unlocked" or not is nondeterministic. This is because it depends on whether the files are currently present in the inode cache. Fixing this by consistently providing each process its own view of the filesystem depending on whether it has the key or not isn't feasible due to how the VFS caches work. Furthermore, while sometimes users expect this behavior, it is misguided for two reasons. First, it would be an OS-level access control mechanism largely redundant with existing access control mechanisms such as UNIX file permissions, ACLs, LSMs, etc. Encryption is actually for protecting the data at rest. Second, almost all users of fscrypt actually do need the keys to be global. The largest users of fscrypt, Android and Chromium OS, achieve this by having PID 1 create a "session keyring" that is inherited by every process. This works, but it isn't scalable because it prevents session keyrings from being used for any other purpose. On general-purpose Linux distros, the 'fscrypt' userspace tool [1] can't similarly abuse the session keyring, so to make 'sudo' work on all systems it has to link all the user keyrings into root's user keyring [2]. This is ugly and raises security concerns. Moreover it can't make the keys available to system services, such as sshd trying to access the user's '~/.ssh' directory (see [3], [4]) or NetworkManager trying to read certificates from the user's home directory (see [5]); or to Docker containers (see [6], [7]). By having an API to add a key to the *filesystem* we'll be able to fix the above bugs, remove userspace workarounds, and clearly express the intended semantics: the locked/unlocked status of an encrypted directory is global, and encryption is orthogonal to OS-level access control. Why not use the add_key() syscall ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ We use an ioctl for this API rather than the existing add_key() system call because the ioctl gives us the flexibility needed to implement fscrypt-specific semantics that will be introduced in later patches: - Supporting key removal with the semantics such that the secret is removed immediately and any unused inodes using the key are evicted; also, the eviction of any in-use inodes can be retried. - Calculating a key-dependent cryptographic identifier and returning it to userspace. - Allowing keys to be added and removed by non-root users, but only keys for v2 encryption policies; and to prevent denial-of-service attacks, users can only remove keys they themselves have added, and a key is only really removed after all users who added it have removed it. Trying to shoehorn these semantics into the keyrings syscalls would be very difficult, whereas the ioctls make things much easier. However, to reuse code the implementation still uses the keyrings service internally. Thus we get lockless RCU-mode key lookups without having to re-implement it, and the keys automatically show up in /proc/keys for debugging purposes. References: [1] https://github.com/google/fscrypt [2] https://goo.gl/55cCrI#heading=h.vf09isp98isb [3] https://github.com/google/fscrypt/issues/111#issuecomment-444347939 [4] https://github.com/google/fscrypt/issues/116 [5] https://bugs.launchpad.net/ubuntu/+source/fscrypt/+bug/1770715 [6] https://github.com/google/fscrypt/issues/128 [7] https://askubuntu.com/questions/1130306/cannot-run-docker-on-an-encrypted-filesystem Reviewed-by: Theodore Ts'o <tytso@mit.edu> Signed-off-by: Eric Biggers <ebiggers@google.com>
2019-08-05 10:35:46 +08:00
{
fscrypt: stop using keyrings subsystem for fscrypt_master_key The approach of fs/crypto/ internally managing the fscrypt_master_key structs as the payloads of "struct key" objects contained in a "struct key" keyring has outlived its usefulness. The original idea was to simplify the code by reusing code from the keyrings subsystem. However, several issues have arisen that can't easily be resolved: - When a master key struct is destroyed, blk_crypto_evict_key() must be called on any per-mode keys embedded in it. (This started being the case when inline encryption support was added.) Yet, the keyrings subsystem can arbitrarily delay the destruction of keys, even past the time the filesystem was unmounted. Therefore, currently there is no easy way to call blk_crypto_evict_key() when a master key is destroyed. Currently, this is worked around by holding an extra reference to the filesystem's request_queue(s). But it was overlooked that the request_queue reference is *not* guaranteed to pin the corresponding blk_crypto_profile too; for device-mapper devices that support inline crypto, it doesn't. This can cause a use-after-free. - When the last inode that was using an incompletely-removed master key is evicted, the master key removal is completed by removing the key struct from the keyring. Currently this is done via key_invalidate(). Yet, key_invalidate() takes the key semaphore. This can deadlock when called from the shrinker, since in fscrypt_ioctl_add_key(), memory is allocated with GFP_KERNEL under the same semaphore. - More generally, the fact that the keyrings subsystem can arbitrarily delay the destruction of keys (via garbage collection delay, or via random processes getting temporary key references) is undesirable, as it means we can't strictly guarantee that all secrets are ever wiped. - Doing the master key lookups via the keyrings subsystem results in the key_permission LSM hook being called. fscrypt doesn't want this, as all access control for encrypted files is designed to happen via the files themselves, like any other files. The workaround which SELinux users are using is to change their SELinux policy to grant key search access to all domains. This works, but it is an odd extra step that shouldn't really have to be done. The fix for all these issues is to change the implementation to what I should have done originally: don't use the keyrings subsystem to keep track of the filesystem's fscrypt_master_key structs. Instead, just store them in a regular kernel data structure, and rework the reference counting, locking, and lifetime accordingly. Retain support for RCU-mode key lookups by using a hash table. Replace fscrypt_sb_free() with fscrypt_sb_delete(), which releases the keys synchronously and runs a bit earlier during unmount, so that block devices are still available. A side effect of this patch is that neither the master keys themselves nor the filesystem keyrings will be listed in /proc/keys anymore. ("Master key users" and the master key users keyrings will still be listed.) However, this was mostly an implementation detail, and it was intended just for debugging purposes. I don't know of anyone using it. This patch does *not* change how "master key users" (->mk_users) works; that still uses the keyrings subsystem. That is still needed for key quotas, and changing that isn't necessary to solve the issues listed above. If we decide to change that too, it would be a separate patch. I've marked this as fixing the original commit that added the fscrypt keyring, but as noted above the most important issue that this patch fixes wasn't introduced until the addition of inline encryption support. Fixes: 22d94f493bfb ("fscrypt: add FS_IOC_ADD_ENCRYPTION_KEY ioctl") Signed-off-by: Eric Biggers <ebiggers@google.com> Link: https://lore.kernel.org/r/20220901193208.138056-2-ebiggers@kernel.org
2022-09-02 03:32:06 +08:00
struct fscrypt_keyring *keyring = sb->s_master_keys;
size_t i;
if (!keyring)
return;
for (i = 0; i < ARRAY_SIZE(keyring->key_hashtable); i++) {
struct hlist_head *bucket = &keyring->key_hashtable[i];
struct fscrypt_master_key *mk;
struct hlist_node *tmp;
hlist_for_each_entry_safe(mk, tmp, bucket, mk_node) {
/*
* Since all inodes were already evicted, every key
* remaining in the keyring should have an empty inode
* list, and should only still be in the keyring due to
* the single active ref associated with ->mk_secret.
* There should be no structural refs beyond the one
* associated with the active ref.
*/
WARN_ON(refcount_read(&mk->mk_active_refs) != 1);
WARN_ON(refcount_read(&mk->mk_struct_refs) != 1);
WARN_ON(!is_master_key_secret_present(&mk->mk_secret));
wipe_master_key_secret(&mk->mk_secret);
fscrypt_put_master_key_activeref(mk);
}
}
kfree_sensitive(keyring);
fscrypt: add FS_IOC_ADD_ENCRYPTION_KEY ioctl Add a new fscrypt ioctl, FS_IOC_ADD_ENCRYPTION_KEY. This ioctl adds an encryption key to the filesystem's fscrypt keyring ->s_master_keys, making any files encrypted with that key appear "unlocked". Why we need this ~~~~~~~~~~~~~~~~ The main problem is that the "locked/unlocked" (ciphertext/plaintext) status of encrypted files is global, but the fscrypt keys are not. fscrypt only looks for keys in the keyring(s) the process accessing the filesystem is subscribed to: the thread keyring, process keyring, and session keyring, where the session keyring may contain the user keyring. Therefore, userspace has to put fscrypt keys in the keyrings for individual users or sessions. But this means that when a process with a different keyring tries to access encrypted files, whether they appear "unlocked" or not is nondeterministic. This is because it depends on whether the files are currently present in the inode cache. Fixing this by consistently providing each process its own view of the filesystem depending on whether it has the key or not isn't feasible due to how the VFS caches work. Furthermore, while sometimes users expect this behavior, it is misguided for two reasons. First, it would be an OS-level access control mechanism largely redundant with existing access control mechanisms such as UNIX file permissions, ACLs, LSMs, etc. Encryption is actually for protecting the data at rest. Second, almost all users of fscrypt actually do need the keys to be global. The largest users of fscrypt, Android and Chromium OS, achieve this by having PID 1 create a "session keyring" that is inherited by every process. This works, but it isn't scalable because it prevents session keyrings from being used for any other purpose. On general-purpose Linux distros, the 'fscrypt' userspace tool [1] can't similarly abuse the session keyring, so to make 'sudo' work on all systems it has to link all the user keyrings into root's user keyring [2]. This is ugly and raises security concerns. Moreover it can't make the keys available to system services, such as sshd trying to access the user's '~/.ssh' directory (see [3], [4]) or NetworkManager trying to read certificates from the user's home directory (see [5]); or to Docker containers (see [6], [7]). By having an API to add a key to the *filesystem* we'll be able to fix the above bugs, remove userspace workarounds, and clearly express the intended semantics: the locked/unlocked status of an encrypted directory is global, and encryption is orthogonal to OS-level access control. Why not use the add_key() syscall ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ We use an ioctl for this API rather than the existing add_key() system call because the ioctl gives us the flexibility needed to implement fscrypt-specific semantics that will be introduced in later patches: - Supporting key removal with the semantics such that the secret is removed immediately and any unused inodes using the key are evicted; also, the eviction of any in-use inodes can be retried. - Calculating a key-dependent cryptographic identifier and returning it to userspace. - Allowing keys to be added and removed by non-root users, but only keys for v2 encryption policies; and to prevent denial-of-service attacks, users can only remove keys they themselves have added, and a key is only really removed after all users who added it have removed it. Trying to shoehorn these semantics into the keyrings syscalls would be very difficult, whereas the ioctls make things much easier. However, to reuse code the implementation still uses the keyrings service internally. Thus we get lockless RCU-mode key lookups without having to re-implement it, and the keys automatically show up in /proc/keys for debugging purposes. References: [1] https://github.com/google/fscrypt [2] https://goo.gl/55cCrI#heading=h.vf09isp98isb [3] https://github.com/google/fscrypt/issues/111#issuecomment-444347939 [4] https://github.com/google/fscrypt/issues/116 [5] https://bugs.launchpad.net/ubuntu/+source/fscrypt/+bug/1770715 [6] https://github.com/google/fscrypt/issues/128 [7] https://askubuntu.com/questions/1130306/cannot-run-docker-on-an-encrypted-filesystem Reviewed-by: Theodore Ts'o <tytso@mit.edu> Signed-off-by: Eric Biggers <ebiggers@google.com>
2019-08-05 10:35:46 +08:00
sb->s_master_keys = NULL;
}
fscrypt: stop using keyrings subsystem for fscrypt_master_key The approach of fs/crypto/ internally managing the fscrypt_master_key structs as the payloads of "struct key" objects contained in a "struct key" keyring has outlived its usefulness. The original idea was to simplify the code by reusing code from the keyrings subsystem. However, several issues have arisen that can't easily be resolved: - When a master key struct is destroyed, blk_crypto_evict_key() must be called on any per-mode keys embedded in it. (This started being the case when inline encryption support was added.) Yet, the keyrings subsystem can arbitrarily delay the destruction of keys, even past the time the filesystem was unmounted. Therefore, currently there is no easy way to call blk_crypto_evict_key() when a master key is destroyed. Currently, this is worked around by holding an extra reference to the filesystem's request_queue(s). But it was overlooked that the request_queue reference is *not* guaranteed to pin the corresponding blk_crypto_profile too; for device-mapper devices that support inline crypto, it doesn't. This can cause a use-after-free. - When the last inode that was using an incompletely-removed master key is evicted, the master key removal is completed by removing the key struct from the keyring. Currently this is done via key_invalidate(). Yet, key_invalidate() takes the key semaphore. This can deadlock when called from the shrinker, since in fscrypt_ioctl_add_key(), memory is allocated with GFP_KERNEL under the same semaphore. - More generally, the fact that the keyrings subsystem can arbitrarily delay the destruction of keys (via garbage collection delay, or via random processes getting temporary key references) is undesirable, as it means we can't strictly guarantee that all secrets are ever wiped. - Doing the master key lookups via the keyrings subsystem results in the key_permission LSM hook being called. fscrypt doesn't want this, as all access control for encrypted files is designed to happen via the files themselves, like any other files. The workaround which SELinux users are using is to change their SELinux policy to grant key search access to all domains. This works, but it is an odd extra step that shouldn't really have to be done. The fix for all these issues is to change the implementation to what I should have done originally: don't use the keyrings subsystem to keep track of the filesystem's fscrypt_master_key structs. Instead, just store them in a regular kernel data structure, and rework the reference counting, locking, and lifetime accordingly. Retain support for RCU-mode key lookups by using a hash table. Replace fscrypt_sb_free() with fscrypt_sb_delete(), which releases the keys synchronously and runs a bit earlier during unmount, so that block devices are still available. A side effect of this patch is that neither the master keys themselves nor the filesystem keyrings will be listed in /proc/keys anymore. ("Master key users" and the master key users keyrings will still be listed.) However, this was mostly an implementation detail, and it was intended just for debugging purposes. I don't know of anyone using it. This patch does *not* change how "master key users" (->mk_users) works; that still uses the keyrings subsystem. That is still needed for key quotas, and changing that isn't necessary to solve the issues listed above. If we decide to change that too, it would be a separate patch. I've marked this as fixing the original commit that added the fscrypt keyring, but as noted above the most important issue that this patch fixes wasn't introduced until the addition of inline encryption support. Fixes: 22d94f493bfb ("fscrypt: add FS_IOC_ADD_ENCRYPTION_KEY ioctl") Signed-off-by: Eric Biggers <ebiggers@google.com> Link: https://lore.kernel.org/r/20220901193208.138056-2-ebiggers@kernel.org
2022-09-02 03:32:06 +08:00
static struct hlist_head *
fscrypt_mk_hash_bucket(struct fscrypt_keyring *keyring,
const struct fscrypt_key_specifier *mk_spec)
{
/*
* Since key specifiers should be "random" values, it is sufficient to
* use a trivial hash function that just takes the first several bits of
* the key specifier.
*/
unsigned long i = get_unaligned((unsigned long *)&mk_spec->u);
return &keyring->key_hashtable[i % ARRAY_SIZE(keyring->key_hashtable)];
}
fscrypt: add FS_IOC_ADD_ENCRYPTION_KEY ioctl Add a new fscrypt ioctl, FS_IOC_ADD_ENCRYPTION_KEY. This ioctl adds an encryption key to the filesystem's fscrypt keyring ->s_master_keys, making any files encrypted with that key appear "unlocked". Why we need this ~~~~~~~~~~~~~~~~ The main problem is that the "locked/unlocked" (ciphertext/plaintext) status of encrypted files is global, but the fscrypt keys are not. fscrypt only looks for keys in the keyring(s) the process accessing the filesystem is subscribed to: the thread keyring, process keyring, and session keyring, where the session keyring may contain the user keyring. Therefore, userspace has to put fscrypt keys in the keyrings for individual users or sessions. But this means that when a process with a different keyring tries to access encrypted files, whether they appear "unlocked" or not is nondeterministic. This is because it depends on whether the files are currently present in the inode cache. Fixing this by consistently providing each process its own view of the filesystem depending on whether it has the key or not isn't feasible due to how the VFS caches work. Furthermore, while sometimes users expect this behavior, it is misguided for two reasons. First, it would be an OS-level access control mechanism largely redundant with existing access control mechanisms such as UNIX file permissions, ACLs, LSMs, etc. Encryption is actually for protecting the data at rest. Second, almost all users of fscrypt actually do need the keys to be global. The largest users of fscrypt, Android and Chromium OS, achieve this by having PID 1 create a "session keyring" that is inherited by every process. This works, but it isn't scalable because it prevents session keyrings from being used for any other purpose. On general-purpose Linux distros, the 'fscrypt' userspace tool [1] can't similarly abuse the session keyring, so to make 'sudo' work on all systems it has to link all the user keyrings into root's user keyring [2]. This is ugly and raises security concerns. Moreover it can't make the keys available to system services, such as sshd trying to access the user's '~/.ssh' directory (see [3], [4]) or NetworkManager trying to read certificates from the user's home directory (see [5]); or to Docker containers (see [6], [7]). By having an API to add a key to the *filesystem* we'll be able to fix the above bugs, remove userspace workarounds, and clearly express the intended semantics: the locked/unlocked status of an encrypted directory is global, and encryption is orthogonal to OS-level access control. Why not use the add_key() syscall ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ We use an ioctl for this API rather than the existing add_key() system call because the ioctl gives us the flexibility needed to implement fscrypt-specific semantics that will be introduced in later patches: - Supporting key removal with the semantics such that the secret is removed immediately and any unused inodes using the key are evicted; also, the eviction of any in-use inodes can be retried. - Calculating a key-dependent cryptographic identifier and returning it to userspace. - Allowing keys to be added and removed by non-root users, but only keys for v2 encryption policies; and to prevent denial-of-service attacks, users can only remove keys they themselves have added, and a key is only really removed after all users who added it have removed it. Trying to shoehorn these semantics into the keyrings syscalls would be very difficult, whereas the ioctls make things much easier. However, to reuse code the implementation still uses the keyrings service internally. Thus we get lockless RCU-mode key lookups without having to re-implement it, and the keys automatically show up in /proc/keys for debugging purposes. References: [1] https://github.com/google/fscrypt [2] https://goo.gl/55cCrI#heading=h.vf09isp98isb [3] https://github.com/google/fscrypt/issues/111#issuecomment-444347939 [4] https://github.com/google/fscrypt/issues/116 [5] https://bugs.launchpad.net/ubuntu/+source/fscrypt/+bug/1770715 [6] https://github.com/google/fscrypt/issues/128 [7] https://askubuntu.com/questions/1130306/cannot-run-docker-on-an-encrypted-filesystem Reviewed-by: Theodore Ts'o <tytso@mit.edu> Signed-off-by: Eric Biggers <ebiggers@google.com>
2019-08-05 10:35:46 +08:00
/*
fscrypt: stop using keyrings subsystem for fscrypt_master_key The approach of fs/crypto/ internally managing the fscrypt_master_key structs as the payloads of "struct key" objects contained in a "struct key" keyring has outlived its usefulness. The original idea was to simplify the code by reusing code from the keyrings subsystem. However, several issues have arisen that can't easily be resolved: - When a master key struct is destroyed, blk_crypto_evict_key() must be called on any per-mode keys embedded in it. (This started being the case when inline encryption support was added.) Yet, the keyrings subsystem can arbitrarily delay the destruction of keys, even past the time the filesystem was unmounted. Therefore, currently there is no easy way to call blk_crypto_evict_key() when a master key is destroyed. Currently, this is worked around by holding an extra reference to the filesystem's request_queue(s). But it was overlooked that the request_queue reference is *not* guaranteed to pin the corresponding blk_crypto_profile too; for device-mapper devices that support inline crypto, it doesn't. This can cause a use-after-free. - When the last inode that was using an incompletely-removed master key is evicted, the master key removal is completed by removing the key struct from the keyring. Currently this is done via key_invalidate(). Yet, key_invalidate() takes the key semaphore. This can deadlock when called from the shrinker, since in fscrypt_ioctl_add_key(), memory is allocated with GFP_KERNEL under the same semaphore. - More generally, the fact that the keyrings subsystem can arbitrarily delay the destruction of keys (via garbage collection delay, or via random processes getting temporary key references) is undesirable, as it means we can't strictly guarantee that all secrets are ever wiped. - Doing the master key lookups via the keyrings subsystem results in the key_permission LSM hook being called. fscrypt doesn't want this, as all access control for encrypted files is designed to happen via the files themselves, like any other files. The workaround which SELinux users are using is to change their SELinux policy to grant key search access to all domains. This works, but it is an odd extra step that shouldn't really have to be done. The fix for all these issues is to change the implementation to what I should have done originally: don't use the keyrings subsystem to keep track of the filesystem's fscrypt_master_key structs. Instead, just store them in a regular kernel data structure, and rework the reference counting, locking, and lifetime accordingly. Retain support for RCU-mode key lookups by using a hash table. Replace fscrypt_sb_free() with fscrypt_sb_delete(), which releases the keys synchronously and runs a bit earlier during unmount, so that block devices are still available. A side effect of this patch is that neither the master keys themselves nor the filesystem keyrings will be listed in /proc/keys anymore. ("Master key users" and the master key users keyrings will still be listed.) However, this was mostly an implementation detail, and it was intended just for debugging purposes. I don't know of anyone using it. This patch does *not* change how "master key users" (->mk_users) works; that still uses the keyrings subsystem. That is still needed for key quotas, and changing that isn't necessary to solve the issues listed above. If we decide to change that too, it would be a separate patch. I've marked this as fixing the original commit that added the fscrypt keyring, but as noted above the most important issue that this patch fixes wasn't introduced until the addition of inline encryption support. Fixes: 22d94f493bfb ("fscrypt: add FS_IOC_ADD_ENCRYPTION_KEY ioctl") Signed-off-by: Eric Biggers <ebiggers@google.com> Link: https://lore.kernel.org/r/20220901193208.138056-2-ebiggers@kernel.org
2022-09-02 03:32:06 +08:00
* Find the specified master key struct in ->s_master_keys and take a structural
* ref to it. The structural ref guarantees that the key struct continues to
* exist, but it does *not* guarantee that ->s_master_keys continues to contain
* the key struct. The structural ref needs to be dropped by
* fscrypt_put_master_key(). Returns NULL if the key struct is not found.
fscrypt: add FS_IOC_ADD_ENCRYPTION_KEY ioctl Add a new fscrypt ioctl, FS_IOC_ADD_ENCRYPTION_KEY. This ioctl adds an encryption key to the filesystem's fscrypt keyring ->s_master_keys, making any files encrypted with that key appear "unlocked". Why we need this ~~~~~~~~~~~~~~~~ The main problem is that the "locked/unlocked" (ciphertext/plaintext) status of encrypted files is global, but the fscrypt keys are not. fscrypt only looks for keys in the keyring(s) the process accessing the filesystem is subscribed to: the thread keyring, process keyring, and session keyring, where the session keyring may contain the user keyring. Therefore, userspace has to put fscrypt keys in the keyrings for individual users or sessions. But this means that when a process with a different keyring tries to access encrypted files, whether they appear "unlocked" or not is nondeterministic. This is because it depends on whether the files are currently present in the inode cache. Fixing this by consistently providing each process its own view of the filesystem depending on whether it has the key or not isn't feasible due to how the VFS caches work. Furthermore, while sometimes users expect this behavior, it is misguided for two reasons. First, it would be an OS-level access control mechanism largely redundant with existing access control mechanisms such as UNIX file permissions, ACLs, LSMs, etc. Encryption is actually for protecting the data at rest. Second, almost all users of fscrypt actually do need the keys to be global. The largest users of fscrypt, Android and Chromium OS, achieve this by having PID 1 create a "session keyring" that is inherited by every process. This works, but it isn't scalable because it prevents session keyrings from being used for any other purpose. On general-purpose Linux distros, the 'fscrypt' userspace tool [1] can't similarly abuse the session keyring, so to make 'sudo' work on all systems it has to link all the user keyrings into root's user keyring [2]. This is ugly and raises security concerns. Moreover it can't make the keys available to system services, such as sshd trying to access the user's '~/.ssh' directory (see [3], [4]) or NetworkManager trying to read certificates from the user's home directory (see [5]); or to Docker containers (see [6], [7]). By having an API to add a key to the *filesystem* we'll be able to fix the above bugs, remove userspace workarounds, and clearly express the intended semantics: the locked/unlocked status of an encrypted directory is global, and encryption is orthogonal to OS-level access control. Why not use the add_key() syscall ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ We use an ioctl for this API rather than the existing add_key() system call because the ioctl gives us the flexibility needed to implement fscrypt-specific semantics that will be introduced in later patches: - Supporting key removal with the semantics such that the secret is removed immediately and any unused inodes using the key are evicted; also, the eviction of any in-use inodes can be retried. - Calculating a key-dependent cryptographic identifier and returning it to userspace. - Allowing keys to be added and removed by non-root users, but only keys for v2 encryption policies; and to prevent denial-of-service attacks, users can only remove keys they themselves have added, and a key is only really removed after all users who added it have removed it. Trying to shoehorn these semantics into the keyrings syscalls would be very difficult, whereas the ioctls make things much easier. However, to reuse code the implementation still uses the keyrings service internally. Thus we get lockless RCU-mode key lookups without having to re-implement it, and the keys automatically show up in /proc/keys for debugging purposes. References: [1] https://github.com/google/fscrypt [2] https://goo.gl/55cCrI#heading=h.vf09isp98isb [3] https://github.com/google/fscrypt/issues/111#issuecomment-444347939 [4] https://github.com/google/fscrypt/issues/116 [5] https://bugs.launchpad.net/ubuntu/+source/fscrypt/+bug/1770715 [6] https://github.com/google/fscrypt/issues/128 [7] https://askubuntu.com/questions/1130306/cannot-run-docker-on-an-encrypted-filesystem Reviewed-by: Theodore Ts'o <tytso@mit.edu> Signed-off-by: Eric Biggers <ebiggers@google.com>
2019-08-05 10:35:46 +08:00
*/
fscrypt: stop using keyrings subsystem for fscrypt_master_key The approach of fs/crypto/ internally managing the fscrypt_master_key structs as the payloads of "struct key" objects contained in a "struct key" keyring has outlived its usefulness. The original idea was to simplify the code by reusing code from the keyrings subsystem. However, several issues have arisen that can't easily be resolved: - When a master key struct is destroyed, blk_crypto_evict_key() must be called on any per-mode keys embedded in it. (This started being the case when inline encryption support was added.) Yet, the keyrings subsystem can arbitrarily delay the destruction of keys, even past the time the filesystem was unmounted. Therefore, currently there is no easy way to call blk_crypto_evict_key() when a master key is destroyed. Currently, this is worked around by holding an extra reference to the filesystem's request_queue(s). But it was overlooked that the request_queue reference is *not* guaranteed to pin the corresponding blk_crypto_profile too; for device-mapper devices that support inline crypto, it doesn't. This can cause a use-after-free. - When the last inode that was using an incompletely-removed master key is evicted, the master key removal is completed by removing the key struct from the keyring. Currently this is done via key_invalidate(). Yet, key_invalidate() takes the key semaphore. This can deadlock when called from the shrinker, since in fscrypt_ioctl_add_key(), memory is allocated with GFP_KERNEL under the same semaphore. - More generally, the fact that the keyrings subsystem can arbitrarily delay the destruction of keys (via garbage collection delay, or via random processes getting temporary key references) is undesirable, as it means we can't strictly guarantee that all secrets are ever wiped. - Doing the master key lookups via the keyrings subsystem results in the key_permission LSM hook being called. fscrypt doesn't want this, as all access control for encrypted files is designed to happen via the files themselves, like any other files. The workaround which SELinux users are using is to change their SELinux policy to grant key search access to all domains. This works, but it is an odd extra step that shouldn't really have to be done. The fix for all these issues is to change the implementation to what I should have done originally: don't use the keyrings subsystem to keep track of the filesystem's fscrypt_master_key structs. Instead, just store them in a regular kernel data structure, and rework the reference counting, locking, and lifetime accordingly. Retain support for RCU-mode key lookups by using a hash table. Replace fscrypt_sb_free() with fscrypt_sb_delete(), which releases the keys synchronously and runs a bit earlier during unmount, so that block devices are still available. A side effect of this patch is that neither the master keys themselves nor the filesystem keyrings will be listed in /proc/keys anymore. ("Master key users" and the master key users keyrings will still be listed.) However, this was mostly an implementation detail, and it was intended just for debugging purposes. I don't know of anyone using it. This patch does *not* change how "master key users" (->mk_users) works; that still uses the keyrings subsystem. That is still needed for key quotas, and changing that isn't necessary to solve the issues listed above. If we decide to change that too, it would be a separate patch. I've marked this as fixing the original commit that added the fscrypt keyring, but as noted above the most important issue that this patch fixes wasn't introduced until the addition of inline encryption support. Fixes: 22d94f493bfb ("fscrypt: add FS_IOC_ADD_ENCRYPTION_KEY ioctl") Signed-off-by: Eric Biggers <ebiggers@google.com> Link: https://lore.kernel.org/r/20220901193208.138056-2-ebiggers@kernel.org
2022-09-02 03:32:06 +08:00
struct fscrypt_master_key *
fscrypt_find_master_key(struct super_block *sb,
const struct fscrypt_key_specifier *mk_spec)
fscrypt: add FS_IOC_ADD_ENCRYPTION_KEY ioctl Add a new fscrypt ioctl, FS_IOC_ADD_ENCRYPTION_KEY. This ioctl adds an encryption key to the filesystem's fscrypt keyring ->s_master_keys, making any files encrypted with that key appear "unlocked". Why we need this ~~~~~~~~~~~~~~~~ The main problem is that the "locked/unlocked" (ciphertext/plaintext) status of encrypted files is global, but the fscrypt keys are not. fscrypt only looks for keys in the keyring(s) the process accessing the filesystem is subscribed to: the thread keyring, process keyring, and session keyring, where the session keyring may contain the user keyring. Therefore, userspace has to put fscrypt keys in the keyrings for individual users or sessions. But this means that when a process with a different keyring tries to access encrypted files, whether they appear "unlocked" or not is nondeterministic. This is because it depends on whether the files are currently present in the inode cache. Fixing this by consistently providing each process its own view of the filesystem depending on whether it has the key or not isn't feasible due to how the VFS caches work. Furthermore, while sometimes users expect this behavior, it is misguided for two reasons. First, it would be an OS-level access control mechanism largely redundant with existing access control mechanisms such as UNIX file permissions, ACLs, LSMs, etc. Encryption is actually for protecting the data at rest. Second, almost all users of fscrypt actually do need the keys to be global. The largest users of fscrypt, Android and Chromium OS, achieve this by having PID 1 create a "session keyring" that is inherited by every process. This works, but it isn't scalable because it prevents session keyrings from being used for any other purpose. On general-purpose Linux distros, the 'fscrypt' userspace tool [1] can't similarly abuse the session keyring, so to make 'sudo' work on all systems it has to link all the user keyrings into root's user keyring [2]. This is ugly and raises security concerns. Moreover it can't make the keys available to system services, such as sshd trying to access the user's '~/.ssh' directory (see [3], [4]) or NetworkManager trying to read certificates from the user's home directory (see [5]); or to Docker containers (see [6], [7]). By having an API to add a key to the *filesystem* we'll be able to fix the above bugs, remove userspace workarounds, and clearly express the intended semantics: the locked/unlocked status of an encrypted directory is global, and encryption is orthogonal to OS-level access control. Why not use the add_key() syscall ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ We use an ioctl for this API rather than the existing add_key() system call because the ioctl gives us the flexibility needed to implement fscrypt-specific semantics that will be introduced in later patches: - Supporting key removal with the semantics such that the secret is removed immediately and any unused inodes using the key are evicted; also, the eviction of any in-use inodes can be retried. - Calculating a key-dependent cryptographic identifier and returning it to userspace. - Allowing keys to be added and removed by non-root users, but only keys for v2 encryption policies; and to prevent denial-of-service attacks, users can only remove keys they themselves have added, and a key is only really removed after all users who added it have removed it. Trying to shoehorn these semantics into the keyrings syscalls would be very difficult, whereas the ioctls make things much easier. However, to reuse code the implementation still uses the keyrings service internally. Thus we get lockless RCU-mode key lookups without having to re-implement it, and the keys automatically show up in /proc/keys for debugging purposes. References: [1] https://github.com/google/fscrypt [2] https://goo.gl/55cCrI#heading=h.vf09isp98isb [3] https://github.com/google/fscrypt/issues/111#issuecomment-444347939 [4] https://github.com/google/fscrypt/issues/116 [5] https://bugs.launchpad.net/ubuntu/+source/fscrypt/+bug/1770715 [6] https://github.com/google/fscrypt/issues/128 [7] https://askubuntu.com/questions/1130306/cannot-run-docker-on-an-encrypted-filesystem Reviewed-by: Theodore Ts'o <tytso@mit.edu> Signed-off-by: Eric Biggers <ebiggers@google.com>
2019-08-05 10:35:46 +08:00
{
fscrypt: stop using keyrings subsystem for fscrypt_master_key The approach of fs/crypto/ internally managing the fscrypt_master_key structs as the payloads of "struct key" objects contained in a "struct key" keyring has outlived its usefulness. The original idea was to simplify the code by reusing code from the keyrings subsystem. However, several issues have arisen that can't easily be resolved: - When a master key struct is destroyed, blk_crypto_evict_key() must be called on any per-mode keys embedded in it. (This started being the case when inline encryption support was added.) Yet, the keyrings subsystem can arbitrarily delay the destruction of keys, even past the time the filesystem was unmounted. Therefore, currently there is no easy way to call blk_crypto_evict_key() when a master key is destroyed. Currently, this is worked around by holding an extra reference to the filesystem's request_queue(s). But it was overlooked that the request_queue reference is *not* guaranteed to pin the corresponding blk_crypto_profile too; for device-mapper devices that support inline crypto, it doesn't. This can cause a use-after-free. - When the last inode that was using an incompletely-removed master key is evicted, the master key removal is completed by removing the key struct from the keyring. Currently this is done via key_invalidate(). Yet, key_invalidate() takes the key semaphore. This can deadlock when called from the shrinker, since in fscrypt_ioctl_add_key(), memory is allocated with GFP_KERNEL under the same semaphore. - More generally, the fact that the keyrings subsystem can arbitrarily delay the destruction of keys (via garbage collection delay, or via random processes getting temporary key references) is undesirable, as it means we can't strictly guarantee that all secrets are ever wiped. - Doing the master key lookups via the keyrings subsystem results in the key_permission LSM hook being called. fscrypt doesn't want this, as all access control for encrypted files is designed to happen via the files themselves, like any other files. The workaround which SELinux users are using is to change their SELinux policy to grant key search access to all domains. This works, but it is an odd extra step that shouldn't really have to be done. The fix for all these issues is to change the implementation to what I should have done originally: don't use the keyrings subsystem to keep track of the filesystem's fscrypt_master_key structs. Instead, just store them in a regular kernel data structure, and rework the reference counting, locking, and lifetime accordingly. Retain support for RCU-mode key lookups by using a hash table. Replace fscrypt_sb_free() with fscrypt_sb_delete(), which releases the keys synchronously and runs a bit earlier during unmount, so that block devices are still available. A side effect of this patch is that neither the master keys themselves nor the filesystem keyrings will be listed in /proc/keys anymore. ("Master key users" and the master key users keyrings will still be listed.) However, this was mostly an implementation detail, and it was intended just for debugging purposes. I don't know of anyone using it. This patch does *not* change how "master key users" (->mk_users) works; that still uses the keyrings subsystem. That is still needed for key quotas, and changing that isn't necessary to solve the issues listed above. If we decide to change that too, it would be a separate patch. I've marked this as fixing the original commit that added the fscrypt keyring, but as noted above the most important issue that this patch fixes wasn't introduced until the addition of inline encryption support. Fixes: 22d94f493bfb ("fscrypt: add FS_IOC_ADD_ENCRYPTION_KEY ioctl") Signed-off-by: Eric Biggers <ebiggers@google.com> Link: https://lore.kernel.org/r/20220901193208.138056-2-ebiggers@kernel.org
2022-09-02 03:32:06 +08:00
struct fscrypt_keyring *keyring;
struct hlist_head *bucket;
struct fscrypt_master_key *mk;
fscrypt: add FS_IOC_ADD_ENCRYPTION_KEY ioctl Add a new fscrypt ioctl, FS_IOC_ADD_ENCRYPTION_KEY. This ioctl adds an encryption key to the filesystem's fscrypt keyring ->s_master_keys, making any files encrypted with that key appear "unlocked". Why we need this ~~~~~~~~~~~~~~~~ The main problem is that the "locked/unlocked" (ciphertext/plaintext) status of encrypted files is global, but the fscrypt keys are not. fscrypt only looks for keys in the keyring(s) the process accessing the filesystem is subscribed to: the thread keyring, process keyring, and session keyring, where the session keyring may contain the user keyring. Therefore, userspace has to put fscrypt keys in the keyrings for individual users or sessions. But this means that when a process with a different keyring tries to access encrypted files, whether they appear "unlocked" or not is nondeterministic. This is because it depends on whether the files are currently present in the inode cache. Fixing this by consistently providing each process its own view of the filesystem depending on whether it has the key or not isn't feasible due to how the VFS caches work. Furthermore, while sometimes users expect this behavior, it is misguided for two reasons. First, it would be an OS-level access control mechanism largely redundant with existing access control mechanisms such as UNIX file permissions, ACLs, LSMs, etc. Encryption is actually for protecting the data at rest. Second, almost all users of fscrypt actually do need the keys to be global. The largest users of fscrypt, Android and Chromium OS, achieve this by having PID 1 create a "session keyring" that is inherited by every process. This works, but it isn't scalable because it prevents session keyrings from being used for any other purpose. On general-purpose Linux distros, the 'fscrypt' userspace tool [1] can't similarly abuse the session keyring, so to make 'sudo' work on all systems it has to link all the user keyrings into root's user keyring [2]. This is ugly and raises security concerns. Moreover it can't make the keys available to system services, such as sshd trying to access the user's '~/.ssh' directory (see [3], [4]) or NetworkManager trying to read certificates from the user's home directory (see [5]); or to Docker containers (see [6], [7]). By having an API to add a key to the *filesystem* we'll be able to fix the above bugs, remove userspace workarounds, and clearly express the intended semantics: the locked/unlocked status of an encrypted directory is global, and encryption is orthogonal to OS-level access control. Why not use the add_key() syscall ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ We use an ioctl for this API rather than the existing add_key() system call because the ioctl gives us the flexibility needed to implement fscrypt-specific semantics that will be introduced in later patches: - Supporting key removal with the semantics such that the secret is removed immediately and any unused inodes using the key are evicted; also, the eviction of any in-use inodes can be retried. - Calculating a key-dependent cryptographic identifier and returning it to userspace. - Allowing keys to be added and removed by non-root users, but only keys for v2 encryption policies; and to prevent denial-of-service attacks, users can only remove keys they themselves have added, and a key is only really removed after all users who added it have removed it. Trying to shoehorn these semantics into the keyrings syscalls would be very difficult, whereas the ioctls make things much easier. However, to reuse code the implementation still uses the keyrings service internally. Thus we get lockless RCU-mode key lookups without having to re-implement it, and the keys automatically show up in /proc/keys for debugging purposes. References: [1] https://github.com/google/fscrypt [2] https://goo.gl/55cCrI#heading=h.vf09isp98isb [3] https://github.com/google/fscrypt/issues/111#issuecomment-444347939 [4] https://github.com/google/fscrypt/issues/116 [5] https://bugs.launchpad.net/ubuntu/+source/fscrypt/+bug/1770715 [6] https://github.com/google/fscrypt/issues/128 [7] https://askubuntu.com/questions/1130306/cannot-run-docker-on-an-encrypted-filesystem Reviewed-by: Theodore Ts'o <tytso@mit.edu> Signed-off-by: Eric Biggers <ebiggers@google.com>
2019-08-05 10:35:46 +08:00
fscrypt: use smp_load_acquire() for ->s_master_keys Normally smp_store_release() or cmpxchg_release() is paired with smp_load_acquire(). Sometimes smp_load_acquire() can be replaced with the more lightweight READ_ONCE(). However, for this to be safe, all the published memory must only be accessed in a way that involves the pointer itself. This may not be the case if allocating the object also involves initializing a static or global variable, for example. super_block::s_master_keys is a keyring, which is internal to and is allocated by the keyrings subsystem. By using READ_ONCE() for it, we're relying on internal implementation details of the keyrings subsystem. Remove this fragile assumption by using smp_load_acquire() instead. (Note: I haven't seen any real-world problems here. This change is just fixing the code to be guaranteed correct and less fragile.) Fixes: 22d94f493bfb ("fscrypt: add FS_IOC_ADD_ENCRYPTION_KEY ioctl") Link: https://lore.kernel.org/r/20200721225920.114347-4-ebiggers@kernel.org Signed-off-by: Eric Biggers <ebiggers@google.com>
2020-07-22 06:59:18 +08:00
/*
* Pairs with the smp_store_release() in allocate_filesystem_keyring().
* I.e., another task can publish ->s_master_keys concurrently,
* executing a RELEASE barrier. We need to use smp_load_acquire() here
* to safely ACQUIRE the memory the other task published.
*/
keyring = smp_load_acquire(&sb->s_master_keys);
fscrypt: add FS_IOC_ADD_ENCRYPTION_KEY ioctl Add a new fscrypt ioctl, FS_IOC_ADD_ENCRYPTION_KEY. This ioctl adds an encryption key to the filesystem's fscrypt keyring ->s_master_keys, making any files encrypted with that key appear "unlocked". Why we need this ~~~~~~~~~~~~~~~~ The main problem is that the "locked/unlocked" (ciphertext/plaintext) status of encrypted files is global, but the fscrypt keys are not. fscrypt only looks for keys in the keyring(s) the process accessing the filesystem is subscribed to: the thread keyring, process keyring, and session keyring, where the session keyring may contain the user keyring. Therefore, userspace has to put fscrypt keys in the keyrings for individual users or sessions. But this means that when a process with a different keyring tries to access encrypted files, whether they appear "unlocked" or not is nondeterministic. This is because it depends on whether the files are currently present in the inode cache. Fixing this by consistently providing each process its own view of the filesystem depending on whether it has the key or not isn't feasible due to how the VFS caches work. Furthermore, while sometimes users expect this behavior, it is misguided for two reasons. First, it would be an OS-level access control mechanism largely redundant with existing access control mechanisms such as UNIX file permissions, ACLs, LSMs, etc. Encryption is actually for protecting the data at rest. Second, almost all users of fscrypt actually do need the keys to be global. The largest users of fscrypt, Android and Chromium OS, achieve this by having PID 1 create a "session keyring" that is inherited by every process. This works, but it isn't scalable because it prevents session keyrings from being used for any other purpose. On general-purpose Linux distros, the 'fscrypt' userspace tool [1] can't similarly abuse the session keyring, so to make 'sudo' work on all systems it has to link all the user keyrings into root's user keyring [2]. This is ugly and raises security concerns. Moreover it can't make the keys available to system services, such as sshd trying to access the user's '~/.ssh' directory (see [3], [4]) or NetworkManager trying to read certificates from the user's home directory (see [5]); or to Docker containers (see [6], [7]). By having an API to add a key to the *filesystem* we'll be able to fix the above bugs, remove userspace workarounds, and clearly express the intended semantics: the locked/unlocked status of an encrypted directory is global, and encryption is orthogonal to OS-level access control. Why not use the add_key() syscall ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ We use an ioctl for this API rather than the existing add_key() system call because the ioctl gives us the flexibility needed to implement fscrypt-specific semantics that will be introduced in later patches: - Supporting key removal with the semantics such that the secret is removed immediately and any unused inodes using the key are evicted; also, the eviction of any in-use inodes can be retried. - Calculating a key-dependent cryptographic identifier and returning it to userspace. - Allowing keys to be added and removed by non-root users, but only keys for v2 encryption policies; and to prevent denial-of-service attacks, users can only remove keys they themselves have added, and a key is only really removed after all users who added it have removed it. Trying to shoehorn these semantics into the keyrings syscalls would be very difficult, whereas the ioctls make things much easier. However, to reuse code the implementation still uses the keyrings service internally. Thus we get lockless RCU-mode key lookups without having to re-implement it, and the keys automatically show up in /proc/keys for debugging purposes. References: [1] https://github.com/google/fscrypt [2] https://goo.gl/55cCrI#heading=h.vf09isp98isb [3] https://github.com/google/fscrypt/issues/111#issuecomment-444347939 [4] https://github.com/google/fscrypt/issues/116 [5] https://bugs.launchpad.net/ubuntu/+source/fscrypt/+bug/1770715 [6] https://github.com/google/fscrypt/issues/128 [7] https://askubuntu.com/questions/1130306/cannot-run-docker-on-an-encrypted-filesystem Reviewed-by: Theodore Ts'o <tytso@mit.edu> Signed-off-by: Eric Biggers <ebiggers@google.com>
2019-08-05 10:35:46 +08:00
if (keyring == NULL)
fscrypt: stop using keyrings subsystem for fscrypt_master_key The approach of fs/crypto/ internally managing the fscrypt_master_key structs as the payloads of "struct key" objects contained in a "struct key" keyring has outlived its usefulness. The original idea was to simplify the code by reusing code from the keyrings subsystem. However, several issues have arisen that can't easily be resolved: - When a master key struct is destroyed, blk_crypto_evict_key() must be called on any per-mode keys embedded in it. (This started being the case when inline encryption support was added.) Yet, the keyrings subsystem can arbitrarily delay the destruction of keys, even past the time the filesystem was unmounted. Therefore, currently there is no easy way to call blk_crypto_evict_key() when a master key is destroyed. Currently, this is worked around by holding an extra reference to the filesystem's request_queue(s). But it was overlooked that the request_queue reference is *not* guaranteed to pin the corresponding blk_crypto_profile too; for device-mapper devices that support inline crypto, it doesn't. This can cause a use-after-free. - When the last inode that was using an incompletely-removed master key is evicted, the master key removal is completed by removing the key struct from the keyring. Currently this is done via key_invalidate(). Yet, key_invalidate() takes the key semaphore. This can deadlock when called from the shrinker, since in fscrypt_ioctl_add_key(), memory is allocated with GFP_KERNEL under the same semaphore. - More generally, the fact that the keyrings subsystem can arbitrarily delay the destruction of keys (via garbage collection delay, or via random processes getting temporary key references) is undesirable, as it means we can't strictly guarantee that all secrets are ever wiped. - Doing the master key lookups via the keyrings subsystem results in the key_permission LSM hook being called. fscrypt doesn't want this, as all access control for encrypted files is designed to happen via the files themselves, like any other files. The workaround which SELinux users are using is to change their SELinux policy to grant key search access to all domains. This works, but it is an odd extra step that shouldn't really have to be done. The fix for all these issues is to change the implementation to what I should have done originally: don't use the keyrings subsystem to keep track of the filesystem's fscrypt_master_key structs. Instead, just store them in a regular kernel data structure, and rework the reference counting, locking, and lifetime accordingly. Retain support for RCU-mode key lookups by using a hash table. Replace fscrypt_sb_free() with fscrypt_sb_delete(), which releases the keys synchronously and runs a bit earlier during unmount, so that block devices are still available. A side effect of this patch is that neither the master keys themselves nor the filesystem keyrings will be listed in /proc/keys anymore. ("Master key users" and the master key users keyrings will still be listed.) However, this was mostly an implementation detail, and it was intended just for debugging purposes. I don't know of anyone using it. This patch does *not* change how "master key users" (->mk_users) works; that still uses the keyrings subsystem. That is still needed for key quotas, and changing that isn't necessary to solve the issues listed above. If we decide to change that too, it would be a separate patch. I've marked this as fixing the original commit that added the fscrypt keyring, but as noted above the most important issue that this patch fixes wasn't introduced until the addition of inline encryption support. Fixes: 22d94f493bfb ("fscrypt: add FS_IOC_ADD_ENCRYPTION_KEY ioctl") Signed-off-by: Eric Biggers <ebiggers@google.com> Link: https://lore.kernel.org/r/20220901193208.138056-2-ebiggers@kernel.org
2022-09-02 03:32:06 +08:00
return NULL; /* No keyring yet, so no keys yet. */
bucket = fscrypt_mk_hash_bucket(keyring, mk_spec);
rcu_read_lock();
switch (mk_spec->type) {
case FSCRYPT_KEY_SPEC_TYPE_DESCRIPTOR:
hlist_for_each_entry_rcu(mk, bucket, mk_node) {
if (mk->mk_spec.type ==
FSCRYPT_KEY_SPEC_TYPE_DESCRIPTOR &&
memcmp(mk->mk_spec.u.descriptor,
mk_spec->u.descriptor,
FSCRYPT_KEY_DESCRIPTOR_SIZE) == 0 &&
refcount_inc_not_zero(&mk->mk_struct_refs))
goto out;
}
break;
case FSCRYPT_KEY_SPEC_TYPE_IDENTIFIER:
hlist_for_each_entry_rcu(mk, bucket, mk_node) {
if (mk->mk_spec.type ==
FSCRYPT_KEY_SPEC_TYPE_IDENTIFIER &&
memcmp(mk->mk_spec.u.identifier,
mk_spec->u.identifier,
FSCRYPT_KEY_IDENTIFIER_SIZE) == 0 &&
refcount_inc_not_zero(&mk->mk_struct_refs))
goto out;
}
break;
}
mk = NULL;
out:
rcu_read_unlock();
return mk;
fscrypt: add FS_IOC_ADD_ENCRYPTION_KEY ioctl Add a new fscrypt ioctl, FS_IOC_ADD_ENCRYPTION_KEY. This ioctl adds an encryption key to the filesystem's fscrypt keyring ->s_master_keys, making any files encrypted with that key appear "unlocked". Why we need this ~~~~~~~~~~~~~~~~ The main problem is that the "locked/unlocked" (ciphertext/plaintext) status of encrypted files is global, but the fscrypt keys are not. fscrypt only looks for keys in the keyring(s) the process accessing the filesystem is subscribed to: the thread keyring, process keyring, and session keyring, where the session keyring may contain the user keyring. Therefore, userspace has to put fscrypt keys in the keyrings for individual users or sessions. But this means that when a process with a different keyring tries to access encrypted files, whether they appear "unlocked" or not is nondeterministic. This is because it depends on whether the files are currently present in the inode cache. Fixing this by consistently providing each process its own view of the filesystem depending on whether it has the key or not isn't feasible due to how the VFS caches work. Furthermore, while sometimes users expect this behavior, it is misguided for two reasons. First, it would be an OS-level access control mechanism largely redundant with existing access control mechanisms such as UNIX file permissions, ACLs, LSMs, etc. Encryption is actually for protecting the data at rest. Second, almost all users of fscrypt actually do need the keys to be global. The largest users of fscrypt, Android and Chromium OS, achieve this by having PID 1 create a "session keyring" that is inherited by every process. This works, but it isn't scalable because it prevents session keyrings from being used for any other purpose. On general-purpose Linux distros, the 'fscrypt' userspace tool [1] can't similarly abuse the session keyring, so to make 'sudo' work on all systems it has to link all the user keyrings into root's user keyring [2]. This is ugly and raises security concerns. Moreover it can't make the keys available to system services, such as sshd trying to access the user's '~/.ssh' directory (see [3], [4]) or NetworkManager trying to read certificates from the user's home directory (see [5]); or to Docker containers (see [6], [7]). By having an API to add a key to the *filesystem* we'll be able to fix the above bugs, remove userspace workarounds, and clearly express the intended semantics: the locked/unlocked status of an encrypted directory is global, and encryption is orthogonal to OS-level access control. Why not use the add_key() syscall ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ We use an ioctl for this API rather than the existing add_key() system call because the ioctl gives us the flexibility needed to implement fscrypt-specific semantics that will be introduced in later patches: - Supporting key removal with the semantics such that the secret is removed immediately and any unused inodes using the key are evicted; also, the eviction of any in-use inodes can be retried. - Calculating a key-dependent cryptographic identifier and returning it to userspace. - Allowing keys to be added and removed by non-root users, but only keys for v2 encryption policies; and to prevent denial-of-service attacks, users can only remove keys they themselves have added, and a key is only really removed after all users who added it have removed it. Trying to shoehorn these semantics into the keyrings syscalls would be very difficult, whereas the ioctls make things much easier. However, to reuse code the implementation still uses the keyrings service internally. Thus we get lockless RCU-mode key lookups without having to re-implement it, and the keys automatically show up in /proc/keys for debugging purposes. References: [1] https://github.com/google/fscrypt [2] https://goo.gl/55cCrI#heading=h.vf09isp98isb [3] https://github.com/google/fscrypt/issues/111#issuecomment-444347939 [4] https://github.com/google/fscrypt/issues/116 [5] https://bugs.launchpad.net/ubuntu/+source/fscrypt/+bug/1770715 [6] https://github.com/google/fscrypt/issues/128 [7] https://askubuntu.com/questions/1130306/cannot-run-docker-on-an-encrypted-filesystem Reviewed-by: Theodore Ts'o <tytso@mit.edu> Signed-off-by: Eric Biggers <ebiggers@google.com>
2019-08-05 10:35:46 +08:00
}
fscrypt: allow unprivileged users to add/remove keys for v2 policies Allow the FS_IOC_ADD_ENCRYPTION_KEY and FS_IOC_REMOVE_ENCRYPTION_KEY ioctls to be used by non-root users to add and remove encryption keys from the filesystem-level crypto keyrings, subject to limitations. Motivation: while privileged fscrypt key management is sufficient for some users (e.g. Android and Chromium OS, where a privileged process manages all keys), the old API by design also allows non-root users to set up and use encrypted directories, and we don't want to regress on that. Especially, we don't want to force users to continue using the old API, running into the visibility mismatch between files and keyrings and being unable to "lock" encrypted directories. Intuitively, the ioctls have to be privileged since they manipulate filesystem-level state. However, it's actually safe to make them unprivileged if we very carefully enforce some specific limitations. First, each key must be identified by a cryptographic hash so that a user can't add the wrong key for another user's files. For v2 encryption policies, we use the key_identifier for this. v1 policies don't have this, so managing keys for them remains privileged. Second, each key a user adds is charged to their quota for the keyrings service. Thus, a user can't exhaust memory by adding a huge number of keys. By default each non-root user is allowed up to 200 keys; this can be changed using the existing sysctl 'kernel.keys.maxkeys'. Third, if multiple users add the same key, we keep track of those users of the key (of which there remains a single copy), and won't really remove the key, i.e. "lock" the encrypted files, until all those users have removed it. This prevents denial of service attacks that would be possible under simpler schemes, such allowing the first user who added a key to remove it -- since that could be a malicious user who has compromised the key. Of course, encryption keys should be kept secret, but the idea is that using encryption should never be *less* secure than not using encryption, even if your key was compromised. We tolerate that a user will be unable to really remove a key, i.e. unable to "lock" their encrypted files, if another user has added the same key. But in a sense, this is actually a good thing because it will avoid providing a false notion of security where a key appears to have been removed when actually it's still in memory, available to any attacker who compromises the operating system kernel. Reviewed-by: Theodore Ts'o <tytso@mit.edu> Signed-off-by: Eric Biggers <ebiggers@google.com>
2019-08-05 10:35:47 +08:00
static int allocate_master_key_users_keyring(struct fscrypt_master_key *mk)
{
char description[FSCRYPT_MK_USERS_DESCRIPTION_SIZE];
struct key *keyring;
format_mk_users_keyring_description(description,
mk->mk_spec.u.identifier);
keyring = keyring_alloc(description, GLOBAL_ROOT_UID, GLOBAL_ROOT_GID,
current_cred(), KEY_POS_SEARCH |
KEY_USR_SEARCH | KEY_USR_READ | KEY_USR_VIEW,
KEY_ALLOC_NOT_IN_QUOTA, NULL, NULL);
if (IS_ERR(keyring))
return PTR_ERR(keyring);
mk->mk_users = keyring;
return 0;
}
/*
* Find the current user's "key" in the master key's ->mk_users.
* Returns ERR_PTR(-ENOKEY) if not found.
*/
static struct key *find_master_key_user(struct fscrypt_master_key *mk)
{
char description[FSCRYPT_MK_USER_DESCRIPTION_SIZE];
fscrypt: stop using keyrings subsystem for fscrypt_master_key The approach of fs/crypto/ internally managing the fscrypt_master_key structs as the payloads of "struct key" objects contained in a "struct key" keyring has outlived its usefulness. The original idea was to simplify the code by reusing code from the keyrings subsystem. However, several issues have arisen that can't easily be resolved: - When a master key struct is destroyed, blk_crypto_evict_key() must be called on any per-mode keys embedded in it. (This started being the case when inline encryption support was added.) Yet, the keyrings subsystem can arbitrarily delay the destruction of keys, even past the time the filesystem was unmounted. Therefore, currently there is no easy way to call blk_crypto_evict_key() when a master key is destroyed. Currently, this is worked around by holding an extra reference to the filesystem's request_queue(s). But it was overlooked that the request_queue reference is *not* guaranteed to pin the corresponding blk_crypto_profile too; for device-mapper devices that support inline crypto, it doesn't. This can cause a use-after-free. - When the last inode that was using an incompletely-removed master key is evicted, the master key removal is completed by removing the key struct from the keyring. Currently this is done via key_invalidate(). Yet, key_invalidate() takes the key semaphore. This can deadlock when called from the shrinker, since in fscrypt_ioctl_add_key(), memory is allocated with GFP_KERNEL under the same semaphore. - More generally, the fact that the keyrings subsystem can arbitrarily delay the destruction of keys (via garbage collection delay, or via random processes getting temporary key references) is undesirable, as it means we can't strictly guarantee that all secrets are ever wiped. - Doing the master key lookups via the keyrings subsystem results in the key_permission LSM hook being called. fscrypt doesn't want this, as all access control for encrypted files is designed to happen via the files themselves, like any other files. The workaround which SELinux users are using is to change their SELinux policy to grant key search access to all domains. This works, but it is an odd extra step that shouldn't really have to be done. The fix for all these issues is to change the implementation to what I should have done originally: don't use the keyrings subsystem to keep track of the filesystem's fscrypt_master_key structs. Instead, just store them in a regular kernel data structure, and rework the reference counting, locking, and lifetime accordingly. Retain support for RCU-mode key lookups by using a hash table. Replace fscrypt_sb_free() with fscrypt_sb_delete(), which releases the keys synchronously and runs a bit earlier during unmount, so that block devices are still available. A side effect of this patch is that neither the master keys themselves nor the filesystem keyrings will be listed in /proc/keys anymore. ("Master key users" and the master key users keyrings will still be listed.) However, this was mostly an implementation detail, and it was intended just for debugging purposes. I don't know of anyone using it. This patch does *not* change how "master key users" (->mk_users) works; that still uses the keyrings subsystem. That is still needed for key quotas, and changing that isn't necessary to solve the issues listed above. If we decide to change that too, it would be a separate patch. I've marked this as fixing the original commit that added the fscrypt keyring, but as noted above the most important issue that this patch fixes wasn't introduced until the addition of inline encryption support. Fixes: 22d94f493bfb ("fscrypt: add FS_IOC_ADD_ENCRYPTION_KEY ioctl") Signed-off-by: Eric Biggers <ebiggers@google.com> Link: https://lore.kernel.org/r/20220901193208.138056-2-ebiggers@kernel.org
2022-09-02 03:32:06 +08:00
key_ref_t keyref;
fscrypt: allow unprivileged users to add/remove keys for v2 policies Allow the FS_IOC_ADD_ENCRYPTION_KEY and FS_IOC_REMOVE_ENCRYPTION_KEY ioctls to be used by non-root users to add and remove encryption keys from the filesystem-level crypto keyrings, subject to limitations. Motivation: while privileged fscrypt key management is sufficient for some users (e.g. Android and Chromium OS, where a privileged process manages all keys), the old API by design also allows non-root users to set up and use encrypted directories, and we don't want to regress on that. Especially, we don't want to force users to continue using the old API, running into the visibility mismatch between files and keyrings and being unable to "lock" encrypted directories. Intuitively, the ioctls have to be privileged since they manipulate filesystem-level state. However, it's actually safe to make them unprivileged if we very carefully enforce some specific limitations. First, each key must be identified by a cryptographic hash so that a user can't add the wrong key for another user's files. For v2 encryption policies, we use the key_identifier for this. v1 policies don't have this, so managing keys for them remains privileged. Second, each key a user adds is charged to their quota for the keyrings service. Thus, a user can't exhaust memory by adding a huge number of keys. By default each non-root user is allowed up to 200 keys; this can be changed using the existing sysctl 'kernel.keys.maxkeys'. Third, if multiple users add the same key, we keep track of those users of the key (of which there remains a single copy), and won't really remove the key, i.e. "lock" the encrypted files, until all those users have removed it. This prevents denial of service attacks that would be possible under simpler schemes, such allowing the first user who added a key to remove it -- since that could be a malicious user who has compromised the key. Of course, encryption keys should be kept secret, but the idea is that using encryption should never be *less* secure than not using encryption, even if your key was compromised. We tolerate that a user will be unable to really remove a key, i.e. unable to "lock" their encrypted files, if another user has added the same key. But in a sense, this is actually a good thing because it will avoid providing a false notion of security where a key appears to have been removed when actually it's still in memory, available to any attacker who compromises the operating system kernel. Reviewed-by: Theodore Ts'o <tytso@mit.edu> Signed-off-by: Eric Biggers <ebiggers@google.com>
2019-08-05 10:35:47 +08:00
format_mk_user_description(description, mk->mk_spec.u.identifier);
fscrypt: stop using keyrings subsystem for fscrypt_master_key The approach of fs/crypto/ internally managing the fscrypt_master_key structs as the payloads of "struct key" objects contained in a "struct key" keyring has outlived its usefulness. The original idea was to simplify the code by reusing code from the keyrings subsystem. However, several issues have arisen that can't easily be resolved: - When a master key struct is destroyed, blk_crypto_evict_key() must be called on any per-mode keys embedded in it. (This started being the case when inline encryption support was added.) Yet, the keyrings subsystem can arbitrarily delay the destruction of keys, even past the time the filesystem was unmounted. Therefore, currently there is no easy way to call blk_crypto_evict_key() when a master key is destroyed. Currently, this is worked around by holding an extra reference to the filesystem's request_queue(s). But it was overlooked that the request_queue reference is *not* guaranteed to pin the corresponding blk_crypto_profile too; for device-mapper devices that support inline crypto, it doesn't. This can cause a use-after-free. - When the last inode that was using an incompletely-removed master key is evicted, the master key removal is completed by removing the key struct from the keyring. Currently this is done via key_invalidate(). Yet, key_invalidate() takes the key semaphore. This can deadlock when called from the shrinker, since in fscrypt_ioctl_add_key(), memory is allocated with GFP_KERNEL under the same semaphore. - More generally, the fact that the keyrings subsystem can arbitrarily delay the destruction of keys (via garbage collection delay, or via random processes getting temporary key references) is undesirable, as it means we can't strictly guarantee that all secrets are ever wiped. - Doing the master key lookups via the keyrings subsystem results in the key_permission LSM hook being called. fscrypt doesn't want this, as all access control for encrypted files is designed to happen via the files themselves, like any other files. The workaround which SELinux users are using is to change their SELinux policy to grant key search access to all domains. This works, but it is an odd extra step that shouldn't really have to be done. The fix for all these issues is to change the implementation to what I should have done originally: don't use the keyrings subsystem to keep track of the filesystem's fscrypt_master_key structs. Instead, just store them in a regular kernel data structure, and rework the reference counting, locking, and lifetime accordingly. Retain support for RCU-mode key lookups by using a hash table. Replace fscrypt_sb_free() with fscrypt_sb_delete(), which releases the keys synchronously and runs a bit earlier during unmount, so that block devices are still available. A side effect of this patch is that neither the master keys themselves nor the filesystem keyrings will be listed in /proc/keys anymore. ("Master key users" and the master key users keyrings will still be listed.) However, this was mostly an implementation detail, and it was intended just for debugging purposes. I don't know of anyone using it. This patch does *not* change how "master key users" (->mk_users) works; that still uses the keyrings subsystem. That is still needed for key quotas, and changing that isn't necessary to solve the issues listed above. If we decide to change that too, it would be a separate patch. I've marked this as fixing the original commit that added the fscrypt keyring, but as noted above the most important issue that this patch fixes wasn't introduced until the addition of inline encryption support. Fixes: 22d94f493bfb ("fscrypt: add FS_IOC_ADD_ENCRYPTION_KEY ioctl") Signed-off-by: Eric Biggers <ebiggers@google.com> Link: https://lore.kernel.org/r/20220901193208.138056-2-ebiggers@kernel.org
2022-09-02 03:32:06 +08:00
/*
* We need to mark the keyring reference as "possessed" so that we
* acquire permission to search it, via the KEY_POS_SEARCH permission.
*/
keyref = keyring_search(make_key_ref(mk->mk_users, true /*possessed*/),
&key_type_fscrypt_user, description, false);
if (IS_ERR(keyref)) {
if (PTR_ERR(keyref) == -EAGAIN || /* not found */
PTR_ERR(keyref) == -EKEYREVOKED) /* recently invalidated */
keyref = ERR_PTR(-ENOKEY);
return ERR_CAST(keyref);
}
return key_ref_to_ptr(keyref);
fscrypt: allow unprivileged users to add/remove keys for v2 policies Allow the FS_IOC_ADD_ENCRYPTION_KEY and FS_IOC_REMOVE_ENCRYPTION_KEY ioctls to be used by non-root users to add and remove encryption keys from the filesystem-level crypto keyrings, subject to limitations. Motivation: while privileged fscrypt key management is sufficient for some users (e.g. Android and Chromium OS, where a privileged process manages all keys), the old API by design also allows non-root users to set up and use encrypted directories, and we don't want to regress on that. Especially, we don't want to force users to continue using the old API, running into the visibility mismatch between files and keyrings and being unable to "lock" encrypted directories. Intuitively, the ioctls have to be privileged since they manipulate filesystem-level state. However, it's actually safe to make them unprivileged if we very carefully enforce some specific limitations. First, each key must be identified by a cryptographic hash so that a user can't add the wrong key for another user's files. For v2 encryption policies, we use the key_identifier for this. v1 policies don't have this, so managing keys for them remains privileged. Second, each key a user adds is charged to their quota for the keyrings service. Thus, a user can't exhaust memory by adding a huge number of keys. By default each non-root user is allowed up to 200 keys; this can be changed using the existing sysctl 'kernel.keys.maxkeys'. Third, if multiple users add the same key, we keep track of those users of the key (of which there remains a single copy), and won't really remove the key, i.e. "lock" the encrypted files, until all those users have removed it. This prevents denial of service attacks that would be possible under simpler schemes, such allowing the first user who added a key to remove it -- since that could be a malicious user who has compromised the key. Of course, encryption keys should be kept secret, but the idea is that using encryption should never be *less* secure than not using encryption, even if your key was compromised. We tolerate that a user will be unable to really remove a key, i.e. unable to "lock" their encrypted files, if another user has added the same key. But in a sense, this is actually a good thing because it will avoid providing a false notion of security where a key appears to have been removed when actually it's still in memory, available to any attacker who compromises the operating system kernel. Reviewed-by: Theodore Ts'o <tytso@mit.edu> Signed-off-by: Eric Biggers <ebiggers@google.com>
2019-08-05 10:35:47 +08:00
}
/*
* Give the current user a "key" in ->mk_users. This charges the user's quota
* and marks the master key as added by the current user, so that it cannot be
fscrypt: stop using keyrings subsystem for fscrypt_master_key The approach of fs/crypto/ internally managing the fscrypt_master_key structs as the payloads of "struct key" objects contained in a "struct key" keyring has outlived its usefulness. The original idea was to simplify the code by reusing code from the keyrings subsystem. However, several issues have arisen that can't easily be resolved: - When a master key struct is destroyed, blk_crypto_evict_key() must be called on any per-mode keys embedded in it. (This started being the case when inline encryption support was added.) Yet, the keyrings subsystem can arbitrarily delay the destruction of keys, even past the time the filesystem was unmounted. Therefore, currently there is no easy way to call blk_crypto_evict_key() when a master key is destroyed. Currently, this is worked around by holding an extra reference to the filesystem's request_queue(s). But it was overlooked that the request_queue reference is *not* guaranteed to pin the corresponding blk_crypto_profile too; for device-mapper devices that support inline crypto, it doesn't. This can cause a use-after-free. - When the last inode that was using an incompletely-removed master key is evicted, the master key removal is completed by removing the key struct from the keyring. Currently this is done via key_invalidate(). Yet, key_invalidate() takes the key semaphore. This can deadlock when called from the shrinker, since in fscrypt_ioctl_add_key(), memory is allocated with GFP_KERNEL under the same semaphore. - More generally, the fact that the keyrings subsystem can arbitrarily delay the destruction of keys (via garbage collection delay, or via random processes getting temporary key references) is undesirable, as it means we can't strictly guarantee that all secrets are ever wiped. - Doing the master key lookups via the keyrings subsystem results in the key_permission LSM hook being called. fscrypt doesn't want this, as all access control for encrypted files is designed to happen via the files themselves, like any other files. The workaround which SELinux users are using is to change their SELinux policy to grant key search access to all domains. This works, but it is an odd extra step that shouldn't really have to be done. The fix for all these issues is to change the implementation to what I should have done originally: don't use the keyrings subsystem to keep track of the filesystem's fscrypt_master_key structs. Instead, just store them in a regular kernel data structure, and rework the reference counting, locking, and lifetime accordingly. Retain support for RCU-mode key lookups by using a hash table. Replace fscrypt_sb_free() with fscrypt_sb_delete(), which releases the keys synchronously and runs a bit earlier during unmount, so that block devices are still available. A side effect of this patch is that neither the master keys themselves nor the filesystem keyrings will be listed in /proc/keys anymore. ("Master key users" and the master key users keyrings will still be listed.) However, this was mostly an implementation detail, and it was intended just for debugging purposes. I don't know of anyone using it. This patch does *not* change how "master key users" (->mk_users) works; that still uses the keyrings subsystem. That is still needed for key quotas, and changing that isn't necessary to solve the issues listed above. If we decide to change that too, it would be a separate patch. I've marked this as fixing the original commit that added the fscrypt keyring, but as noted above the most important issue that this patch fixes wasn't introduced until the addition of inline encryption support. Fixes: 22d94f493bfb ("fscrypt: add FS_IOC_ADD_ENCRYPTION_KEY ioctl") Signed-off-by: Eric Biggers <ebiggers@google.com> Link: https://lore.kernel.org/r/20220901193208.138056-2-ebiggers@kernel.org
2022-09-02 03:32:06 +08:00
* removed by another user with the key. Either ->mk_sem must be held for
* write, or the master key must be still undergoing initialization.
fscrypt: allow unprivileged users to add/remove keys for v2 policies Allow the FS_IOC_ADD_ENCRYPTION_KEY and FS_IOC_REMOVE_ENCRYPTION_KEY ioctls to be used by non-root users to add and remove encryption keys from the filesystem-level crypto keyrings, subject to limitations. Motivation: while privileged fscrypt key management is sufficient for some users (e.g. Android and Chromium OS, where a privileged process manages all keys), the old API by design also allows non-root users to set up and use encrypted directories, and we don't want to regress on that. Especially, we don't want to force users to continue using the old API, running into the visibility mismatch between files and keyrings and being unable to "lock" encrypted directories. Intuitively, the ioctls have to be privileged since they manipulate filesystem-level state. However, it's actually safe to make them unprivileged if we very carefully enforce some specific limitations. First, each key must be identified by a cryptographic hash so that a user can't add the wrong key for another user's files. For v2 encryption policies, we use the key_identifier for this. v1 policies don't have this, so managing keys for them remains privileged. Second, each key a user adds is charged to their quota for the keyrings service. Thus, a user can't exhaust memory by adding a huge number of keys. By default each non-root user is allowed up to 200 keys; this can be changed using the existing sysctl 'kernel.keys.maxkeys'. Third, if multiple users add the same key, we keep track of those users of the key (of which there remains a single copy), and won't really remove the key, i.e. "lock" the encrypted files, until all those users have removed it. This prevents denial of service attacks that would be possible under simpler schemes, such allowing the first user who added a key to remove it -- since that could be a malicious user who has compromised the key. Of course, encryption keys should be kept secret, but the idea is that using encryption should never be *less* secure than not using encryption, even if your key was compromised. We tolerate that a user will be unable to really remove a key, i.e. unable to "lock" their encrypted files, if another user has added the same key. But in a sense, this is actually a good thing because it will avoid providing a false notion of security where a key appears to have been removed when actually it's still in memory, available to any attacker who compromises the operating system kernel. Reviewed-by: Theodore Ts'o <tytso@mit.edu> Signed-off-by: Eric Biggers <ebiggers@google.com>
2019-08-05 10:35:47 +08:00
*/
static int add_master_key_user(struct fscrypt_master_key *mk)
{
char description[FSCRYPT_MK_USER_DESCRIPTION_SIZE];
struct key *mk_user;
int err;
format_mk_user_description(description, mk->mk_spec.u.identifier);
mk_user = key_alloc(&key_type_fscrypt_user, description,
current_fsuid(), current_gid(), current_cred(),
KEY_POS_SEARCH | KEY_USR_VIEW, 0, NULL);
if (IS_ERR(mk_user))
return PTR_ERR(mk_user);
err = key_instantiate_and_link(mk_user, NULL, 0, mk->mk_users, NULL);
key_put(mk_user);
return err;
}
/*
* Remove the current user's "key" from ->mk_users.
fscrypt: stop using keyrings subsystem for fscrypt_master_key The approach of fs/crypto/ internally managing the fscrypt_master_key structs as the payloads of "struct key" objects contained in a "struct key" keyring has outlived its usefulness. The original idea was to simplify the code by reusing code from the keyrings subsystem. However, several issues have arisen that can't easily be resolved: - When a master key struct is destroyed, blk_crypto_evict_key() must be called on any per-mode keys embedded in it. (This started being the case when inline encryption support was added.) Yet, the keyrings subsystem can arbitrarily delay the destruction of keys, even past the time the filesystem was unmounted. Therefore, currently there is no easy way to call blk_crypto_evict_key() when a master key is destroyed. Currently, this is worked around by holding an extra reference to the filesystem's request_queue(s). But it was overlooked that the request_queue reference is *not* guaranteed to pin the corresponding blk_crypto_profile too; for device-mapper devices that support inline crypto, it doesn't. This can cause a use-after-free. - When the last inode that was using an incompletely-removed master key is evicted, the master key removal is completed by removing the key struct from the keyring. Currently this is done via key_invalidate(). Yet, key_invalidate() takes the key semaphore. This can deadlock when called from the shrinker, since in fscrypt_ioctl_add_key(), memory is allocated with GFP_KERNEL under the same semaphore. - More generally, the fact that the keyrings subsystem can arbitrarily delay the destruction of keys (via garbage collection delay, or via random processes getting temporary key references) is undesirable, as it means we can't strictly guarantee that all secrets are ever wiped. - Doing the master key lookups via the keyrings subsystem results in the key_permission LSM hook being called. fscrypt doesn't want this, as all access control for encrypted files is designed to happen via the files themselves, like any other files. The workaround which SELinux users are using is to change their SELinux policy to grant key search access to all domains. This works, but it is an odd extra step that shouldn't really have to be done. The fix for all these issues is to change the implementation to what I should have done originally: don't use the keyrings subsystem to keep track of the filesystem's fscrypt_master_key structs. Instead, just store them in a regular kernel data structure, and rework the reference counting, locking, and lifetime accordingly. Retain support for RCU-mode key lookups by using a hash table. Replace fscrypt_sb_free() with fscrypt_sb_delete(), which releases the keys synchronously and runs a bit earlier during unmount, so that block devices are still available. A side effect of this patch is that neither the master keys themselves nor the filesystem keyrings will be listed in /proc/keys anymore. ("Master key users" and the master key users keyrings will still be listed.) However, this was mostly an implementation detail, and it was intended just for debugging purposes. I don't know of anyone using it. This patch does *not* change how "master key users" (->mk_users) works; that still uses the keyrings subsystem. That is still needed for key quotas, and changing that isn't necessary to solve the issues listed above. If we decide to change that too, it would be a separate patch. I've marked this as fixing the original commit that added the fscrypt keyring, but as noted above the most important issue that this patch fixes wasn't introduced until the addition of inline encryption support. Fixes: 22d94f493bfb ("fscrypt: add FS_IOC_ADD_ENCRYPTION_KEY ioctl") Signed-off-by: Eric Biggers <ebiggers@google.com> Link: https://lore.kernel.org/r/20220901193208.138056-2-ebiggers@kernel.org
2022-09-02 03:32:06 +08:00
* ->mk_sem must be held for write.
fscrypt: allow unprivileged users to add/remove keys for v2 policies Allow the FS_IOC_ADD_ENCRYPTION_KEY and FS_IOC_REMOVE_ENCRYPTION_KEY ioctls to be used by non-root users to add and remove encryption keys from the filesystem-level crypto keyrings, subject to limitations. Motivation: while privileged fscrypt key management is sufficient for some users (e.g. Android and Chromium OS, where a privileged process manages all keys), the old API by design also allows non-root users to set up and use encrypted directories, and we don't want to regress on that. Especially, we don't want to force users to continue using the old API, running into the visibility mismatch between files and keyrings and being unable to "lock" encrypted directories. Intuitively, the ioctls have to be privileged since they manipulate filesystem-level state. However, it's actually safe to make them unprivileged if we very carefully enforce some specific limitations. First, each key must be identified by a cryptographic hash so that a user can't add the wrong key for another user's files. For v2 encryption policies, we use the key_identifier for this. v1 policies don't have this, so managing keys for them remains privileged. Second, each key a user adds is charged to their quota for the keyrings service. Thus, a user can't exhaust memory by adding a huge number of keys. By default each non-root user is allowed up to 200 keys; this can be changed using the existing sysctl 'kernel.keys.maxkeys'. Third, if multiple users add the same key, we keep track of those users of the key (of which there remains a single copy), and won't really remove the key, i.e. "lock" the encrypted files, until all those users have removed it. This prevents denial of service attacks that would be possible under simpler schemes, such allowing the first user who added a key to remove it -- since that could be a malicious user who has compromised the key. Of course, encryption keys should be kept secret, but the idea is that using encryption should never be *less* secure than not using encryption, even if your key was compromised. We tolerate that a user will be unable to really remove a key, i.e. unable to "lock" their encrypted files, if another user has added the same key. But in a sense, this is actually a good thing because it will avoid providing a false notion of security where a key appears to have been removed when actually it's still in memory, available to any attacker who compromises the operating system kernel. Reviewed-by: Theodore Ts'o <tytso@mit.edu> Signed-off-by: Eric Biggers <ebiggers@google.com>
2019-08-05 10:35:47 +08:00
*
* Returns 0 if removed, -ENOKEY if not found, or another -errno code.
*/
static int remove_master_key_user(struct fscrypt_master_key *mk)
{
struct key *mk_user;
int err;
mk_user = find_master_key_user(mk);
if (IS_ERR(mk_user))
return PTR_ERR(mk_user);
err = key_unlink(mk->mk_users, mk_user);
key_put(mk_user);
return err;
}
fscrypt: add FS_IOC_ADD_ENCRYPTION_KEY ioctl Add a new fscrypt ioctl, FS_IOC_ADD_ENCRYPTION_KEY. This ioctl adds an encryption key to the filesystem's fscrypt keyring ->s_master_keys, making any files encrypted with that key appear "unlocked". Why we need this ~~~~~~~~~~~~~~~~ The main problem is that the "locked/unlocked" (ciphertext/plaintext) status of encrypted files is global, but the fscrypt keys are not. fscrypt only looks for keys in the keyring(s) the process accessing the filesystem is subscribed to: the thread keyring, process keyring, and session keyring, where the session keyring may contain the user keyring. Therefore, userspace has to put fscrypt keys in the keyrings for individual users or sessions. But this means that when a process with a different keyring tries to access encrypted files, whether they appear "unlocked" or not is nondeterministic. This is because it depends on whether the files are currently present in the inode cache. Fixing this by consistently providing each process its own view of the filesystem depending on whether it has the key or not isn't feasible due to how the VFS caches work. Furthermore, while sometimes users expect this behavior, it is misguided for two reasons. First, it would be an OS-level access control mechanism largely redundant with existing access control mechanisms such as UNIX file permissions, ACLs, LSMs, etc. Encryption is actually for protecting the data at rest. Second, almost all users of fscrypt actually do need the keys to be global. The largest users of fscrypt, Android and Chromium OS, achieve this by having PID 1 create a "session keyring" that is inherited by every process. This works, but it isn't scalable because it prevents session keyrings from being used for any other purpose. On general-purpose Linux distros, the 'fscrypt' userspace tool [1] can't similarly abuse the session keyring, so to make 'sudo' work on all systems it has to link all the user keyrings into root's user keyring [2]. This is ugly and raises security concerns. Moreover it can't make the keys available to system services, such as sshd trying to access the user's '~/.ssh' directory (see [3], [4]) or NetworkManager trying to read certificates from the user's home directory (see [5]); or to Docker containers (see [6], [7]). By having an API to add a key to the *filesystem* we'll be able to fix the above bugs, remove userspace workarounds, and clearly express the intended semantics: the locked/unlocked status of an encrypted directory is global, and encryption is orthogonal to OS-level access control. Why not use the add_key() syscall ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ We use an ioctl for this API rather than the existing add_key() system call because the ioctl gives us the flexibility needed to implement fscrypt-specific semantics that will be introduced in later patches: - Supporting key removal with the semantics such that the secret is removed immediately and any unused inodes using the key are evicted; also, the eviction of any in-use inodes can be retried. - Calculating a key-dependent cryptographic identifier and returning it to userspace. - Allowing keys to be added and removed by non-root users, but only keys for v2 encryption policies; and to prevent denial-of-service attacks, users can only remove keys they themselves have added, and a key is only really removed after all users who added it have removed it. Trying to shoehorn these semantics into the keyrings syscalls would be very difficult, whereas the ioctls make things much easier. However, to reuse code the implementation still uses the keyrings service internally. Thus we get lockless RCU-mode key lookups without having to re-implement it, and the keys automatically show up in /proc/keys for debugging purposes. References: [1] https://github.com/google/fscrypt [2] https://goo.gl/55cCrI#heading=h.vf09isp98isb [3] https://github.com/google/fscrypt/issues/111#issuecomment-444347939 [4] https://github.com/google/fscrypt/issues/116 [5] https://bugs.launchpad.net/ubuntu/+source/fscrypt/+bug/1770715 [6] https://github.com/google/fscrypt/issues/128 [7] https://askubuntu.com/questions/1130306/cannot-run-docker-on-an-encrypted-filesystem Reviewed-by: Theodore Ts'o <tytso@mit.edu> Signed-off-by: Eric Biggers <ebiggers@google.com>
2019-08-05 10:35:46 +08:00
/*
fscrypt: stop using keyrings subsystem for fscrypt_master_key The approach of fs/crypto/ internally managing the fscrypt_master_key structs as the payloads of "struct key" objects contained in a "struct key" keyring has outlived its usefulness. The original idea was to simplify the code by reusing code from the keyrings subsystem. However, several issues have arisen that can't easily be resolved: - When a master key struct is destroyed, blk_crypto_evict_key() must be called on any per-mode keys embedded in it. (This started being the case when inline encryption support was added.) Yet, the keyrings subsystem can arbitrarily delay the destruction of keys, even past the time the filesystem was unmounted. Therefore, currently there is no easy way to call blk_crypto_evict_key() when a master key is destroyed. Currently, this is worked around by holding an extra reference to the filesystem's request_queue(s). But it was overlooked that the request_queue reference is *not* guaranteed to pin the corresponding blk_crypto_profile too; for device-mapper devices that support inline crypto, it doesn't. This can cause a use-after-free. - When the last inode that was using an incompletely-removed master key is evicted, the master key removal is completed by removing the key struct from the keyring. Currently this is done via key_invalidate(). Yet, key_invalidate() takes the key semaphore. This can deadlock when called from the shrinker, since in fscrypt_ioctl_add_key(), memory is allocated with GFP_KERNEL under the same semaphore. - More generally, the fact that the keyrings subsystem can arbitrarily delay the destruction of keys (via garbage collection delay, or via random processes getting temporary key references) is undesirable, as it means we can't strictly guarantee that all secrets are ever wiped. - Doing the master key lookups via the keyrings subsystem results in the key_permission LSM hook being called. fscrypt doesn't want this, as all access control for encrypted files is designed to happen via the files themselves, like any other files. The workaround which SELinux users are using is to change their SELinux policy to grant key search access to all domains. This works, but it is an odd extra step that shouldn't really have to be done. The fix for all these issues is to change the implementation to what I should have done originally: don't use the keyrings subsystem to keep track of the filesystem's fscrypt_master_key structs. Instead, just store them in a regular kernel data structure, and rework the reference counting, locking, and lifetime accordingly. Retain support for RCU-mode key lookups by using a hash table. Replace fscrypt_sb_free() with fscrypt_sb_delete(), which releases the keys synchronously and runs a bit earlier during unmount, so that block devices are still available. A side effect of this patch is that neither the master keys themselves nor the filesystem keyrings will be listed in /proc/keys anymore. ("Master key users" and the master key users keyrings will still be listed.) However, this was mostly an implementation detail, and it was intended just for debugging purposes. I don't know of anyone using it. This patch does *not* change how "master key users" (->mk_users) works; that still uses the keyrings subsystem. That is still needed for key quotas, and changing that isn't necessary to solve the issues listed above. If we decide to change that too, it would be a separate patch. I've marked this as fixing the original commit that added the fscrypt keyring, but as noted above the most important issue that this patch fixes wasn't introduced until the addition of inline encryption support. Fixes: 22d94f493bfb ("fscrypt: add FS_IOC_ADD_ENCRYPTION_KEY ioctl") Signed-off-by: Eric Biggers <ebiggers@google.com> Link: https://lore.kernel.org/r/20220901193208.138056-2-ebiggers@kernel.org
2022-09-02 03:32:06 +08:00
* Allocate a new fscrypt_master_key, transfer the given secret over to it, and
* insert it into sb->s_master_keys.
fscrypt: add FS_IOC_ADD_ENCRYPTION_KEY ioctl Add a new fscrypt ioctl, FS_IOC_ADD_ENCRYPTION_KEY. This ioctl adds an encryption key to the filesystem's fscrypt keyring ->s_master_keys, making any files encrypted with that key appear "unlocked". Why we need this ~~~~~~~~~~~~~~~~ The main problem is that the "locked/unlocked" (ciphertext/plaintext) status of encrypted files is global, but the fscrypt keys are not. fscrypt only looks for keys in the keyring(s) the process accessing the filesystem is subscribed to: the thread keyring, process keyring, and session keyring, where the session keyring may contain the user keyring. Therefore, userspace has to put fscrypt keys in the keyrings for individual users or sessions. But this means that when a process with a different keyring tries to access encrypted files, whether they appear "unlocked" or not is nondeterministic. This is because it depends on whether the files are currently present in the inode cache. Fixing this by consistently providing each process its own view of the filesystem depending on whether it has the key or not isn't feasible due to how the VFS caches work. Furthermore, while sometimes users expect this behavior, it is misguided for two reasons. First, it would be an OS-level access control mechanism largely redundant with existing access control mechanisms such as UNIX file permissions, ACLs, LSMs, etc. Encryption is actually for protecting the data at rest. Second, almost all users of fscrypt actually do need the keys to be global. The largest users of fscrypt, Android and Chromium OS, achieve this by having PID 1 create a "session keyring" that is inherited by every process. This works, but it isn't scalable because it prevents session keyrings from being used for any other purpose. On general-purpose Linux distros, the 'fscrypt' userspace tool [1] can't similarly abuse the session keyring, so to make 'sudo' work on all systems it has to link all the user keyrings into root's user keyring [2]. This is ugly and raises security concerns. Moreover it can't make the keys available to system services, such as sshd trying to access the user's '~/.ssh' directory (see [3], [4]) or NetworkManager trying to read certificates from the user's home directory (see [5]); or to Docker containers (see [6], [7]). By having an API to add a key to the *filesystem* we'll be able to fix the above bugs, remove userspace workarounds, and clearly express the intended semantics: the locked/unlocked status of an encrypted directory is global, and encryption is orthogonal to OS-level access control. Why not use the add_key() syscall ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ We use an ioctl for this API rather than the existing add_key() system call because the ioctl gives us the flexibility needed to implement fscrypt-specific semantics that will be introduced in later patches: - Supporting key removal with the semantics such that the secret is removed immediately and any unused inodes using the key are evicted; also, the eviction of any in-use inodes can be retried. - Calculating a key-dependent cryptographic identifier and returning it to userspace. - Allowing keys to be added and removed by non-root users, but only keys for v2 encryption policies; and to prevent denial-of-service attacks, users can only remove keys they themselves have added, and a key is only really removed after all users who added it have removed it. Trying to shoehorn these semantics into the keyrings syscalls would be very difficult, whereas the ioctls make things much easier. However, to reuse code the implementation still uses the keyrings service internally. Thus we get lockless RCU-mode key lookups without having to re-implement it, and the keys automatically show up in /proc/keys for debugging purposes. References: [1] https://github.com/google/fscrypt [2] https://goo.gl/55cCrI#heading=h.vf09isp98isb [3] https://github.com/google/fscrypt/issues/111#issuecomment-444347939 [4] https://github.com/google/fscrypt/issues/116 [5] https://bugs.launchpad.net/ubuntu/+source/fscrypt/+bug/1770715 [6] https://github.com/google/fscrypt/issues/128 [7] https://askubuntu.com/questions/1130306/cannot-run-docker-on-an-encrypted-filesystem Reviewed-by: Theodore Ts'o <tytso@mit.edu> Signed-off-by: Eric Biggers <ebiggers@google.com>
2019-08-05 10:35:46 +08:00
*/
fscrypt: stop using keyrings subsystem for fscrypt_master_key The approach of fs/crypto/ internally managing the fscrypt_master_key structs as the payloads of "struct key" objects contained in a "struct key" keyring has outlived its usefulness. The original idea was to simplify the code by reusing code from the keyrings subsystem. However, several issues have arisen that can't easily be resolved: - When a master key struct is destroyed, blk_crypto_evict_key() must be called on any per-mode keys embedded in it. (This started being the case when inline encryption support was added.) Yet, the keyrings subsystem can arbitrarily delay the destruction of keys, even past the time the filesystem was unmounted. Therefore, currently there is no easy way to call blk_crypto_evict_key() when a master key is destroyed. Currently, this is worked around by holding an extra reference to the filesystem's request_queue(s). But it was overlooked that the request_queue reference is *not* guaranteed to pin the corresponding blk_crypto_profile too; for device-mapper devices that support inline crypto, it doesn't. This can cause a use-after-free. - When the last inode that was using an incompletely-removed master key is evicted, the master key removal is completed by removing the key struct from the keyring. Currently this is done via key_invalidate(). Yet, key_invalidate() takes the key semaphore. This can deadlock when called from the shrinker, since in fscrypt_ioctl_add_key(), memory is allocated with GFP_KERNEL under the same semaphore. - More generally, the fact that the keyrings subsystem can arbitrarily delay the destruction of keys (via garbage collection delay, or via random processes getting temporary key references) is undesirable, as it means we can't strictly guarantee that all secrets are ever wiped. - Doing the master key lookups via the keyrings subsystem results in the key_permission LSM hook being called. fscrypt doesn't want this, as all access control for encrypted files is designed to happen via the files themselves, like any other files. The workaround which SELinux users are using is to change their SELinux policy to grant key search access to all domains. This works, but it is an odd extra step that shouldn't really have to be done. The fix for all these issues is to change the implementation to what I should have done originally: don't use the keyrings subsystem to keep track of the filesystem's fscrypt_master_key structs. Instead, just store them in a regular kernel data structure, and rework the reference counting, locking, and lifetime accordingly. Retain support for RCU-mode key lookups by using a hash table. Replace fscrypt_sb_free() with fscrypt_sb_delete(), which releases the keys synchronously and runs a bit earlier during unmount, so that block devices are still available. A side effect of this patch is that neither the master keys themselves nor the filesystem keyrings will be listed in /proc/keys anymore. ("Master key users" and the master key users keyrings will still be listed.) However, this was mostly an implementation detail, and it was intended just for debugging purposes. I don't know of anyone using it. This patch does *not* change how "master key users" (->mk_users) works; that still uses the keyrings subsystem. That is still needed for key quotas, and changing that isn't necessary to solve the issues listed above. If we decide to change that too, it would be a separate patch. I've marked this as fixing the original commit that added the fscrypt keyring, but as noted above the most important issue that this patch fixes wasn't introduced until the addition of inline encryption support. Fixes: 22d94f493bfb ("fscrypt: add FS_IOC_ADD_ENCRYPTION_KEY ioctl") Signed-off-by: Eric Biggers <ebiggers@google.com> Link: https://lore.kernel.org/r/20220901193208.138056-2-ebiggers@kernel.org
2022-09-02 03:32:06 +08:00
static int add_new_master_key(struct super_block *sb,
struct fscrypt_master_key_secret *secret,
const struct fscrypt_key_specifier *mk_spec)
fscrypt: add FS_IOC_ADD_ENCRYPTION_KEY ioctl Add a new fscrypt ioctl, FS_IOC_ADD_ENCRYPTION_KEY. This ioctl adds an encryption key to the filesystem's fscrypt keyring ->s_master_keys, making any files encrypted with that key appear "unlocked". Why we need this ~~~~~~~~~~~~~~~~ The main problem is that the "locked/unlocked" (ciphertext/plaintext) status of encrypted files is global, but the fscrypt keys are not. fscrypt only looks for keys in the keyring(s) the process accessing the filesystem is subscribed to: the thread keyring, process keyring, and session keyring, where the session keyring may contain the user keyring. Therefore, userspace has to put fscrypt keys in the keyrings for individual users or sessions. But this means that when a process with a different keyring tries to access encrypted files, whether they appear "unlocked" or not is nondeterministic. This is because it depends on whether the files are currently present in the inode cache. Fixing this by consistently providing each process its own view of the filesystem depending on whether it has the key or not isn't feasible due to how the VFS caches work. Furthermore, while sometimes users expect this behavior, it is misguided for two reasons. First, it would be an OS-level access control mechanism largely redundant with existing access control mechanisms such as UNIX file permissions, ACLs, LSMs, etc. Encryption is actually for protecting the data at rest. Second, almost all users of fscrypt actually do need the keys to be global. The largest users of fscrypt, Android and Chromium OS, achieve this by having PID 1 create a "session keyring" that is inherited by every process. This works, but it isn't scalable because it prevents session keyrings from being used for any other purpose. On general-purpose Linux distros, the 'fscrypt' userspace tool [1] can't similarly abuse the session keyring, so to make 'sudo' work on all systems it has to link all the user keyrings into root's user keyring [2]. This is ugly and raises security concerns. Moreover it can't make the keys available to system services, such as sshd trying to access the user's '~/.ssh' directory (see [3], [4]) or NetworkManager trying to read certificates from the user's home directory (see [5]); or to Docker containers (see [6], [7]). By having an API to add a key to the *filesystem* we'll be able to fix the above bugs, remove userspace workarounds, and clearly express the intended semantics: the locked/unlocked status of an encrypted directory is global, and encryption is orthogonal to OS-level access control. Why not use the add_key() syscall ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ We use an ioctl for this API rather than the existing add_key() system call because the ioctl gives us the flexibility needed to implement fscrypt-specific semantics that will be introduced in later patches: - Supporting key removal with the semantics such that the secret is removed immediately and any unused inodes using the key are evicted; also, the eviction of any in-use inodes can be retried. - Calculating a key-dependent cryptographic identifier and returning it to userspace. - Allowing keys to be added and removed by non-root users, but only keys for v2 encryption policies; and to prevent denial-of-service attacks, users can only remove keys they themselves have added, and a key is only really removed after all users who added it have removed it. Trying to shoehorn these semantics into the keyrings syscalls would be very difficult, whereas the ioctls make things much easier. However, to reuse code the implementation still uses the keyrings service internally. Thus we get lockless RCU-mode key lookups without having to re-implement it, and the keys automatically show up in /proc/keys for debugging purposes. References: [1] https://github.com/google/fscrypt [2] https://goo.gl/55cCrI#heading=h.vf09isp98isb [3] https://github.com/google/fscrypt/issues/111#issuecomment-444347939 [4] https://github.com/google/fscrypt/issues/116 [5] https://bugs.launchpad.net/ubuntu/+source/fscrypt/+bug/1770715 [6] https://github.com/google/fscrypt/issues/128 [7] https://askubuntu.com/questions/1130306/cannot-run-docker-on-an-encrypted-filesystem Reviewed-by: Theodore Ts'o <tytso@mit.edu> Signed-off-by: Eric Biggers <ebiggers@google.com>
2019-08-05 10:35:46 +08:00
{
fscrypt: stop using keyrings subsystem for fscrypt_master_key The approach of fs/crypto/ internally managing the fscrypt_master_key structs as the payloads of "struct key" objects contained in a "struct key" keyring has outlived its usefulness. The original idea was to simplify the code by reusing code from the keyrings subsystem. However, several issues have arisen that can't easily be resolved: - When a master key struct is destroyed, blk_crypto_evict_key() must be called on any per-mode keys embedded in it. (This started being the case when inline encryption support was added.) Yet, the keyrings subsystem can arbitrarily delay the destruction of keys, even past the time the filesystem was unmounted. Therefore, currently there is no easy way to call blk_crypto_evict_key() when a master key is destroyed. Currently, this is worked around by holding an extra reference to the filesystem's request_queue(s). But it was overlooked that the request_queue reference is *not* guaranteed to pin the corresponding blk_crypto_profile too; for device-mapper devices that support inline crypto, it doesn't. This can cause a use-after-free. - When the last inode that was using an incompletely-removed master key is evicted, the master key removal is completed by removing the key struct from the keyring. Currently this is done via key_invalidate(). Yet, key_invalidate() takes the key semaphore. This can deadlock when called from the shrinker, since in fscrypt_ioctl_add_key(), memory is allocated with GFP_KERNEL under the same semaphore. - More generally, the fact that the keyrings subsystem can arbitrarily delay the destruction of keys (via garbage collection delay, or via random processes getting temporary key references) is undesirable, as it means we can't strictly guarantee that all secrets are ever wiped. - Doing the master key lookups via the keyrings subsystem results in the key_permission LSM hook being called. fscrypt doesn't want this, as all access control for encrypted files is designed to happen via the files themselves, like any other files. The workaround which SELinux users are using is to change their SELinux policy to grant key search access to all domains. This works, but it is an odd extra step that shouldn't really have to be done. The fix for all these issues is to change the implementation to what I should have done originally: don't use the keyrings subsystem to keep track of the filesystem's fscrypt_master_key structs. Instead, just store them in a regular kernel data structure, and rework the reference counting, locking, and lifetime accordingly. Retain support for RCU-mode key lookups by using a hash table. Replace fscrypt_sb_free() with fscrypt_sb_delete(), which releases the keys synchronously and runs a bit earlier during unmount, so that block devices are still available. A side effect of this patch is that neither the master keys themselves nor the filesystem keyrings will be listed in /proc/keys anymore. ("Master key users" and the master key users keyrings will still be listed.) However, this was mostly an implementation detail, and it was intended just for debugging purposes. I don't know of anyone using it. This patch does *not* change how "master key users" (->mk_users) works; that still uses the keyrings subsystem. That is still needed for key quotas, and changing that isn't necessary to solve the issues listed above. If we decide to change that too, it would be a separate patch. I've marked this as fixing the original commit that added the fscrypt keyring, but as noted above the most important issue that this patch fixes wasn't introduced until the addition of inline encryption support. Fixes: 22d94f493bfb ("fscrypt: add FS_IOC_ADD_ENCRYPTION_KEY ioctl") Signed-off-by: Eric Biggers <ebiggers@google.com> Link: https://lore.kernel.org/r/20220901193208.138056-2-ebiggers@kernel.org
2022-09-02 03:32:06 +08:00
struct fscrypt_keyring *keyring = sb->s_master_keys;
fscrypt: add FS_IOC_ADD_ENCRYPTION_KEY ioctl Add a new fscrypt ioctl, FS_IOC_ADD_ENCRYPTION_KEY. This ioctl adds an encryption key to the filesystem's fscrypt keyring ->s_master_keys, making any files encrypted with that key appear "unlocked". Why we need this ~~~~~~~~~~~~~~~~ The main problem is that the "locked/unlocked" (ciphertext/plaintext) status of encrypted files is global, but the fscrypt keys are not. fscrypt only looks for keys in the keyring(s) the process accessing the filesystem is subscribed to: the thread keyring, process keyring, and session keyring, where the session keyring may contain the user keyring. Therefore, userspace has to put fscrypt keys in the keyrings for individual users or sessions. But this means that when a process with a different keyring tries to access encrypted files, whether they appear "unlocked" or not is nondeterministic. This is because it depends on whether the files are currently present in the inode cache. Fixing this by consistently providing each process its own view of the filesystem depending on whether it has the key or not isn't feasible due to how the VFS caches work. Furthermore, while sometimes users expect this behavior, it is misguided for two reasons. First, it would be an OS-level access control mechanism largely redundant with existing access control mechanisms such as UNIX file permissions, ACLs, LSMs, etc. Encryption is actually for protecting the data at rest. Second, almost all users of fscrypt actually do need the keys to be global. The largest users of fscrypt, Android and Chromium OS, achieve this by having PID 1 create a "session keyring" that is inherited by every process. This works, but it isn't scalable because it prevents session keyrings from being used for any other purpose. On general-purpose Linux distros, the 'fscrypt' userspace tool [1] can't similarly abuse the session keyring, so to make 'sudo' work on all systems it has to link all the user keyrings into root's user keyring [2]. This is ugly and raises security concerns. Moreover it can't make the keys available to system services, such as sshd trying to access the user's '~/.ssh' directory (see [3], [4]) or NetworkManager trying to read certificates from the user's home directory (see [5]); or to Docker containers (see [6], [7]). By having an API to add a key to the *filesystem* we'll be able to fix the above bugs, remove userspace workarounds, and clearly express the intended semantics: the locked/unlocked status of an encrypted directory is global, and encryption is orthogonal to OS-level access control. Why not use the add_key() syscall ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ We use an ioctl for this API rather than the existing add_key() system call because the ioctl gives us the flexibility needed to implement fscrypt-specific semantics that will be introduced in later patches: - Supporting key removal with the semantics such that the secret is removed immediately and any unused inodes using the key are evicted; also, the eviction of any in-use inodes can be retried. - Calculating a key-dependent cryptographic identifier and returning it to userspace. - Allowing keys to be added and removed by non-root users, but only keys for v2 encryption policies; and to prevent denial-of-service attacks, users can only remove keys they themselves have added, and a key is only really removed after all users who added it have removed it. Trying to shoehorn these semantics into the keyrings syscalls would be very difficult, whereas the ioctls make things much easier. However, to reuse code the implementation still uses the keyrings service internally. Thus we get lockless RCU-mode key lookups without having to re-implement it, and the keys automatically show up in /proc/keys for debugging purposes. References: [1] https://github.com/google/fscrypt [2] https://goo.gl/55cCrI#heading=h.vf09isp98isb [3] https://github.com/google/fscrypt/issues/111#issuecomment-444347939 [4] https://github.com/google/fscrypt/issues/116 [5] https://bugs.launchpad.net/ubuntu/+source/fscrypt/+bug/1770715 [6] https://github.com/google/fscrypt/issues/128 [7] https://askubuntu.com/questions/1130306/cannot-run-docker-on-an-encrypted-filesystem Reviewed-by: Theodore Ts'o <tytso@mit.edu> Signed-off-by: Eric Biggers <ebiggers@google.com>
2019-08-05 10:35:46 +08:00
struct fscrypt_master_key *mk;
int err;
mk = kzalloc(sizeof(*mk), GFP_KERNEL);
if (!mk)
return -ENOMEM;
fscrypt: stop using keyrings subsystem for fscrypt_master_key The approach of fs/crypto/ internally managing the fscrypt_master_key structs as the payloads of "struct key" objects contained in a "struct key" keyring has outlived its usefulness. The original idea was to simplify the code by reusing code from the keyrings subsystem. However, several issues have arisen that can't easily be resolved: - When a master key struct is destroyed, blk_crypto_evict_key() must be called on any per-mode keys embedded in it. (This started being the case when inline encryption support was added.) Yet, the keyrings subsystem can arbitrarily delay the destruction of keys, even past the time the filesystem was unmounted. Therefore, currently there is no easy way to call blk_crypto_evict_key() when a master key is destroyed. Currently, this is worked around by holding an extra reference to the filesystem's request_queue(s). But it was overlooked that the request_queue reference is *not* guaranteed to pin the corresponding blk_crypto_profile too; for device-mapper devices that support inline crypto, it doesn't. This can cause a use-after-free. - When the last inode that was using an incompletely-removed master key is evicted, the master key removal is completed by removing the key struct from the keyring. Currently this is done via key_invalidate(). Yet, key_invalidate() takes the key semaphore. This can deadlock when called from the shrinker, since in fscrypt_ioctl_add_key(), memory is allocated with GFP_KERNEL under the same semaphore. - More generally, the fact that the keyrings subsystem can arbitrarily delay the destruction of keys (via garbage collection delay, or via random processes getting temporary key references) is undesirable, as it means we can't strictly guarantee that all secrets are ever wiped. - Doing the master key lookups via the keyrings subsystem results in the key_permission LSM hook being called. fscrypt doesn't want this, as all access control for encrypted files is designed to happen via the files themselves, like any other files. The workaround which SELinux users are using is to change their SELinux policy to grant key search access to all domains. This works, but it is an odd extra step that shouldn't really have to be done. The fix for all these issues is to change the implementation to what I should have done originally: don't use the keyrings subsystem to keep track of the filesystem's fscrypt_master_key structs. Instead, just store them in a regular kernel data structure, and rework the reference counting, locking, and lifetime accordingly. Retain support for RCU-mode key lookups by using a hash table. Replace fscrypt_sb_free() with fscrypt_sb_delete(), which releases the keys synchronously and runs a bit earlier during unmount, so that block devices are still available. A side effect of this patch is that neither the master keys themselves nor the filesystem keyrings will be listed in /proc/keys anymore. ("Master key users" and the master key users keyrings will still be listed.) However, this was mostly an implementation detail, and it was intended just for debugging purposes. I don't know of anyone using it. This patch does *not* change how "master key users" (->mk_users) works; that still uses the keyrings subsystem. That is still needed for key quotas, and changing that isn't necessary to solve the issues listed above. If we decide to change that too, it would be a separate patch. I've marked this as fixing the original commit that added the fscrypt keyring, but as noted above the most important issue that this patch fixes wasn't introduced until the addition of inline encryption support. Fixes: 22d94f493bfb ("fscrypt: add FS_IOC_ADD_ENCRYPTION_KEY ioctl") Signed-off-by: Eric Biggers <ebiggers@google.com> Link: https://lore.kernel.org/r/20220901193208.138056-2-ebiggers@kernel.org
2022-09-02 03:32:06 +08:00
mk->mk_sb = sb;
init_rwsem(&mk->mk_sem);
refcount_set(&mk->mk_struct_refs, 1);
fscrypt: add FS_IOC_ADD_ENCRYPTION_KEY ioctl Add a new fscrypt ioctl, FS_IOC_ADD_ENCRYPTION_KEY. This ioctl adds an encryption key to the filesystem's fscrypt keyring ->s_master_keys, making any files encrypted with that key appear "unlocked". Why we need this ~~~~~~~~~~~~~~~~ The main problem is that the "locked/unlocked" (ciphertext/plaintext) status of encrypted files is global, but the fscrypt keys are not. fscrypt only looks for keys in the keyring(s) the process accessing the filesystem is subscribed to: the thread keyring, process keyring, and session keyring, where the session keyring may contain the user keyring. Therefore, userspace has to put fscrypt keys in the keyrings for individual users or sessions. But this means that when a process with a different keyring tries to access encrypted files, whether they appear "unlocked" or not is nondeterministic. This is because it depends on whether the files are currently present in the inode cache. Fixing this by consistently providing each process its own view of the filesystem depending on whether it has the key or not isn't feasible due to how the VFS caches work. Furthermore, while sometimes users expect this behavior, it is misguided for two reasons. First, it would be an OS-level access control mechanism largely redundant with existing access control mechanisms such as UNIX file permissions, ACLs, LSMs, etc. Encryption is actually for protecting the data at rest. Second, almost all users of fscrypt actually do need the keys to be global. The largest users of fscrypt, Android and Chromium OS, achieve this by having PID 1 create a "session keyring" that is inherited by every process. This works, but it isn't scalable because it prevents session keyrings from being used for any other purpose. On general-purpose Linux distros, the 'fscrypt' userspace tool [1] can't similarly abuse the session keyring, so to make 'sudo' work on all systems it has to link all the user keyrings into root's user keyring [2]. This is ugly and raises security concerns. Moreover it can't make the keys available to system services, such as sshd trying to access the user's '~/.ssh' directory (see [3], [4]) or NetworkManager trying to read certificates from the user's home directory (see [5]); or to Docker containers (see [6], [7]). By having an API to add a key to the *filesystem* we'll be able to fix the above bugs, remove userspace workarounds, and clearly express the intended semantics: the locked/unlocked status of an encrypted directory is global, and encryption is orthogonal to OS-level access control. Why not use the add_key() syscall ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ We use an ioctl for this API rather than the existing add_key() system call because the ioctl gives us the flexibility needed to implement fscrypt-specific semantics that will be introduced in later patches: - Supporting key removal with the semantics such that the secret is removed immediately and any unused inodes using the key are evicted; also, the eviction of any in-use inodes can be retried. - Calculating a key-dependent cryptographic identifier and returning it to userspace. - Allowing keys to be added and removed by non-root users, but only keys for v2 encryption policies; and to prevent denial-of-service attacks, users can only remove keys they themselves have added, and a key is only really removed after all users who added it have removed it. Trying to shoehorn these semantics into the keyrings syscalls would be very difficult, whereas the ioctls make things much easier. However, to reuse code the implementation still uses the keyrings service internally. Thus we get lockless RCU-mode key lookups without having to re-implement it, and the keys automatically show up in /proc/keys for debugging purposes. References: [1] https://github.com/google/fscrypt [2] https://goo.gl/55cCrI#heading=h.vf09isp98isb [3] https://github.com/google/fscrypt/issues/111#issuecomment-444347939 [4] https://github.com/google/fscrypt/issues/116 [5] https://bugs.launchpad.net/ubuntu/+source/fscrypt/+bug/1770715 [6] https://github.com/google/fscrypt/issues/128 [7] https://askubuntu.com/questions/1130306/cannot-run-docker-on-an-encrypted-filesystem Reviewed-by: Theodore Ts'o <tytso@mit.edu> Signed-off-by: Eric Biggers <ebiggers@google.com>
2019-08-05 10:35:46 +08:00
mk->mk_spec = *mk_spec;
fscrypt: add FS_IOC_REMOVE_ENCRYPTION_KEY ioctl Add a new fscrypt ioctl, FS_IOC_REMOVE_ENCRYPTION_KEY. This ioctl removes an encryption key that was added by FS_IOC_ADD_ENCRYPTION_KEY. It wipes the secret key itself, then "locks" the encrypted files and directories that had been unlocked using that key -- implemented by evicting the relevant dentries and inodes from the VFS caches. The problem this solves is that many fscrypt users want the ability to remove encryption keys, causing the corresponding encrypted directories to appear "locked" (presented in ciphertext form) again. Moreover, users want removing an encryption key to *really* remove it, in the sense that the removed keys cannot be recovered even if kernel memory is compromised, e.g. by the exploit of a kernel security vulnerability or by a physical attack. This is desirable after a user logs out of the system, for example. In many cases users even already assume this to be the case and are surprised to hear when it's not. It is not sufficient to simply unlink the master key from the keyring (or to revoke or invalidate it), since the actual encryption transform objects are still pinned in memory by their inodes. Therefore, to really remove a key we must also evict the relevant inodes. Currently one workaround is to run 'sync && echo 2 > /proc/sys/vm/drop_caches'. But, that evicts all unused inodes in the system rather than just the inodes associated with the key being removed, causing severe performance problems. Moreover, it requires root privileges, so regular users can't "lock" their encrypted files. Another workaround, used in Chromium OS kernels, is to add a new VFS-level ioctl FS_IOC_DROP_CACHE which is a more restricted version of drop_caches that operates on a single super_block. It does: shrink_dcache_sb(sb); invalidate_inodes(sb, false); But it's still a hack. Yet, the major users of filesystem encryption want this feature badly enough that they are actually using these hacks. To properly solve the problem, start maintaining a list of the inodes which have been "unlocked" using each master key. Originally this wasn't possible because the kernel didn't keep track of in-use master keys at all. But, with the ->s_master_keys keyring it is now possible. Then, add an ioctl FS_IOC_REMOVE_ENCRYPTION_KEY. It finds the specified master key in ->s_master_keys, then wipes the secret key itself, which prevents any additional inodes from being unlocked with the key. Then, it syncs the filesystem and evicts the inodes in the key's list. The normal inode eviction code will free and wipe the per-file keys (in ->i_crypt_info). Note that freeing ->i_crypt_info without evicting the inodes was also considered, but would have been racy. Some inodes may still be in use when a master key is removed, and we can't simply revoke random file descriptors, mmap's, etc. Thus, the ioctl simply skips in-use inodes, and returns -EBUSY to indicate that some inodes weren't evicted. The master key *secret* is still removed, but the fscrypt_master_key struct remains to keep track of the remaining inodes. Userspace can then retry the ioctl to evict the remaining inodes. Alternatively, if userspace adds the key again, the refreshed secret will be associated with the existing list of inodes so they remain correctly tracked for future key removals. The ioctl doesn't wipe pagecache pages. Thus, we tolerate that after a kernel compromise some portions of plaintext file contents may still be recoverable from memory. This can be solved by enabling page poisoning system-wide, which security conscious users may choose to do. But it's very difficult to solve otherwise, e.g. note that plaintext file contents may have been read in other places than pagecache pages. Like FS_IOC_ADD_ENCRYPTION_KEY, FS_IOC_REMOVE_ENCRYPTION_KEY is initially restricted to privileged users only. This is sufficient for some use cases, but not all. A later patch will relax this restriction, but it will require introducing key hashes, among other changes. Reviewed-by: Theodore Ts'o <tytso@mit.edu> Signed-off-by: Eric Biggers <ebiggers@google.com>
2019-08-05 10:35:46 +08:00
INIT_LIST_HEAD(&mk->mk_decrypted_inodes);
spin_lock_init(&mk->mk_decrypted_inodes_lock);
fscrypt: allow unprivileged users to add/remove keys for v2 policies Allow the FS_IOC_ADD_ENCRYPTION_KEY and FS_IOC_REMOVE_ENCRYPTION_KEY ioctls to be used by non-root users to add and remove encryption keys from the filesystem-level crypto keyrings, subject to limitations. Motivation: while privileged fscrypt key management is sufficient for some users (e.g. Android and Chromium OS, where a privileged process manages all keys), the old API by design also allows non-root users to set up and use encrypted directories, and we don't want to regress on that. Especially, we don't want to force users to continue using the old API, running into the visibility mismatch between files and keyrings and being unable to "lock" encrypted directories. Intuitively, the ioctls have to be privileged since they manipulate filesystem-level state. However, it's actually safe to make them unprivileged if we very carefully enforce some specific limitations. First, each key must be identified by a cryptographic hash so that a user can't add the wrong key for another user's files. For v2 encryption policies, we use the key_identifier for this. v1 policies don't have this, so managing keys for them remains privileged. Second, each key a user adds is charged to their quota for the keyrings service. Thus, a user can't exhaust memory by adding a huge number of keys. By default each non-root user is allowed up to 200 keys; this can be changed using the existing sysctl 'kernel.keys.maxkeys'. Third, if multiple users add the same key, we keep track of those users of the key (of which there remains a single copy), and won't really remove the key, i.e. "lock" the encrypted files, until all those users have removed it. This prevents denial of service attacks that would be possible under simpler schemes, such allowing the first user who added a key to remove it -- since that could be a malicious user who has compromised the key. Of course, encryption keys should be kept secret, but the idea is that using encryption should never be *less* secure than not using encryption, even if your key was compromised. We tolerate that a user will be unable to really remove a key, i.e. unable to "lock" their encrypted files, if another user has added the same key. But in a sense, this is actually a good thing because it will avoid providing a false notion of security where a key appears to have been removed when actually it's still in memory, available to any attacker who compromises the operating system kernel. Reviewed-by: Theodore Ts'o <tytso@mit.edu> Signed-off-by: Eric Biggers <ebiggers@google.com>
2019-08-05 10:35:47 +08:00
if (mk_spec->type == FSCRYPT_KEY_SPEC_TYPE_IDENTIFIER) {
err = allocate_master_key_users_keyring(mk);
if (err)
fscrypt: stop using keyrings subsystem for fscrypt_master_key The approach of fs/crypto/ internally managing the fscrypt_master_key structs as the payloads of "struct key" objects contained in a "struct key" keyring has outlived its usefulness. The original idea was to simplify the code by reusing code from the keyrings subsystem. However, several issues have arisen that can't easily be resolved: - When a master key struct is destroyed, blk_crypto_evict_key() must be called on any per-mode keys embedded in it. (This started being the case when inline encryption support was added.) Yet, the keyrings subsystem can arbitrarily delay the destruction of keys, even past the time the filesystem was unmounted. Therefore, currently there is no easy way to call blk_crypto_evict_key() when a master key is destroyed. Currently, this is worked around by holding an extra reference to the filesystem's request_queue(s). But it was overlooked that the request_queue reference is *not* guaranteed to pin the corresponding blk_crypto_profile too; for device-mapper devices that support inline crypto, it doesn't. This can cause a use-after-free. - When the last inode that was using an incompletely-removed master key is evicted, the master key removal is completed by removing the key struct from the keyring. Currently this is done via key_invalidate(). Yet, key_invalidate() takes the key semaphore. This can deadlock when called from the shrinker, since in fscrypt_ioctl_add_key(), memory is allocated with GFP_KERNEL under the same semaphore. - More generally, the fact that the keyrings subsystem can arbitrarily delay the destruction of keys (via garbage collection delay, or via random processes getting temporary key references) is undesirable, as it means we can't strictly guarantee that all secrets are ever wiped. - Doing the master key lookups via the keyrings subsystem results in the key_permission LSM hook being called. fscrypt doesn't want this, as all access control for encrypted files is designed to happen via the files themselves, like any other files. The workaround which SELinux users are using is to change their SELinux policy to grant key search access to all domains. This works, but it is an odd extra step that shouldn't really have to be done. The fix for all these issues is to change the implementation to what I should have done originally: don't use the keyrings subsystem to keep track of the filesystem's fscrypt_master_key structs. Instead, just store them in a regular kernel data structure, and rework the reference counting, locking, and lifetime accordingly. Retain support for RCU-mode key lookups by using a hash table. Replace fscrypt_sb_free() with fscrypt_sb_delete(), which releases the keys synchronously and runs a bit earlier during unmount, so that block devices are still available. A side effect of this patch is that neither the master keys themselves nor the filesystem keyrings will be listed in /proc/keys anymore. ("Master key users" and the master key users keyrings will still be listed.) However, this was mostly an implementation detail, and it was intended just for debugging purposes. I don't know of anyone using it. This patch does *not* change how "master key users" (->mk_users) works; that still uses the keyrings subsystem. That is still needed for key quotas, and changing that isn't necessary to solve the issues listed above. If we decide to change that too, it would be a separate patch. I've marked this as fixing the original commit that added the fscrypt keyring, but as noted above the most important issue that this patch fixes wasn't introduced until the addition of inline encryption support. Fixes: 22d94f493bfb ("fscrypt: add FS_IOC_ADD_ENCRYPTION_KEY ioctl") Signed-off-by: Eric Biggers <ebiggers@google.com> Link: https://lore.kernel.org/r/20220901193208.138056-2-ebiggers@kernel.org
2022-09-02 03:32:06 +08:00
goto out_put;
fscrypt: allow unprivileged users to add/remove keys for v2 policies Allow the FS_IOC_ADD_ENCRYPTION_KEY and FS_IOC_REMOVE_ENCRYPTION_KEY ioctls to be used by non-root users to add and remove encryption keys from the filesystem-level crypto keyrings, subject to limitations. Motivation: while privileged fscrypt key management is sufficient for some users (e.g. Android and Chromium OS, where a privileged process manages all keys), the old API by design also allows non-root users to set up and use encrypted directories, and we don't want to regress on that. Especially, we don't want to force users to continue using the old API, running into the visibility mismatch between files and keyrings and being unable to "lock" encrypted directories. Intuitively, the ioctls have to be privileged since they manipulate filesystem-level state. However, it's actually safe to make them unprivileged if we very carefully enforce some specific limitations. First, each key must be identified by a cryptographic hash so that a user can't add the wrong key for another user's files. For v2 encryption policies, we use the key_identifier for this. v1 policies don't have this, so managing keys for them remains privileged. Second, each key a user adds is charged to their quota for the keyrings service. Thus, a user can't exhaust memory by adding a huge number of keys. By default each non-root user is allowed up to 200 keys; this can be changed using the existing sysctl 'kernel.keys.maxkeys'. Third, if multiple users add the same key, we keep track of those users of the key (of which there remains a single copy), and won't really remove the key, i.e. "lock" the encrypted files, until all those users have removed it. This prevents denial of service attacks that would be possible under simpler schemes, such allowing the first user who added a key to remove it -- since that could be a malicious user who has compromised the key. Of course, encryption keys should be kept secret, but the idea is that using encryption should never be *less* secure than not using encryption, even if your key was compromised. We tolerate that a user will be unable to really remove a key, i.e. unable to "lock" their encrypted files, if another user has added the same key. But in a sense, this is actually a good thing because it will avoid providing a false notion of security where a key appears to have been removed when actually it's still in memory, available to any attacker who compromises the operating system kernel. Reviewed-by: Theodore Ts'o <tytso@mit.edu> Signed-off-by: Eric Biggers <ebiggers@google.com>
2019-08-05 10:35:47 +08:00
err = add_master_key_user(mk);
if (err)
fscrypt: stop using keyrings subsystem for fscrypt_master_key The approach of fs/crypto/ internally managing the fscrypt_master_key structs as the payloads of "struct key" objects contained in a "struct key" keyring has outlived its usefulness. The original idea was to simplify the code by reusing code from the keyrings subsystem. However, several issues have arisen that can't easily be resolved: - When a master key struct is destroyed, blk_crypto_evict_key() must be called on any per-mode keys embedded in it. (This started being the case when inline encryption support was added.) Yet, the keyrings subsystem can arbitrarily delay the destruction of keys, even past the time the filesystem was unmounted. Therefore, currently there is no easy way to call blk_crypto_evict_key() when a master key is destroyed. Currently, this is worked around by holding an extra reference to the filesystem's request_queue(s). But it was overlooked that the request_queue reference is *not* guaranteed to pin the corresponding blk_crypto_profile too; for device-mapper devices that support inline crypto, it doesn't. This can cause a use-after-free. - When the last inode that was using an incompletely-removed master key is evicted, the master key removal is completed by removing the key struct from the keyring. Currently this is done via key_invalidate(). Yet, key_invalidate() takes the key semaphore. This can deadlock when called from the shrinker, since in fscrypt_ioctl_add_key(), memory is allocated with GFP_KERNEL under the same semaphore. - More generally, the fact that the keyrings subsystem can arbitrarily delay the destruction of keys (via garbage collection delay, or via random processes getting temporary key references) is undesirable, as it means we can't strictly guarantee that all secrets are ever wiped. - Doing the master key lookups via the keyrings subsystem results in the key_permission LSM hook being called. fscrypt doesn't want this, as all access control for encrypted files is designed to happen via the files themselves, like any other files. The workaround which SELinux users are using is to change their SELinux policy to grant key search access to all domains. This works, but it is an odd extra step that shouldn't really have to be done. The fix for all these issues is to change the implementation to what I should have done originally: don't use the keyrings subsystem to keep track of the filesystem's fscrypt_master_key structs. Instead, just store them in a regular kernel data structure, and rework the reference counting, locking, and lifetime accordingly. Retain support for RCU-mode key lookups by using a hash table. Replace fscrypt_sb_free() with fscrypt_sb_delete(), which releases the keys synchronously and runs a bit earlier during unmount, so that block devices are still available. A side effect of this patch is that neither the master keys themselves nor the filesystem keyrings will be listed in /proc/keys anymore. ("Master key users" and the master key users keyrings will still be listed.) However, this was mostly an implementation detail, and it was intended just for debugging purposes. I don't know of anyone using it. This patch does *not* change how "master key users" (->mk_users) works; that still uses the keyrings subsystem. That is still needed for key quotas, and changing that isn't necessary to solve the issues listed above. If we decide to change that too, it would be a separate patch. I've marked this as fixing the original commit that added the fscrypt keyring, but as noted above the most important issue that this patch fixes wasn't introduced until the addition of inline encryption support. Fixes: 22d94f493bfb ("fscrypt: add FS_IOC_ADD_ENCRYPTION_KEY ioctl") Signed-off-by: Eric Biggers <ebiggers@google.com> Link: https://lore.kernel.org/r/20220901193208.138056-2-ebiggers@kernel.org
2022-09-02 03:32:06 +08:00
goto out_put;
fscrypt: allow unprivileged users to add/remove keys for v2 policies Allow the FS_IOC_ADD_ENCRYPTION_KEY and FS_IOC_REMOVE_ENCRYPTION_KEY ioctls to be used by non-root users to add and remove encryption keys from the filesystem-level crypto keyrings, subject to limitations. Motivation: while privileged fscrypt key management is sufficient for some users (e.g. Android and Chromium OS, where a privileged process manages all keys), the old API by design also allows non-root users to set up and use encrypted directories, and we don't want to regress on that. Especially, we don't want to force users to continue using the old API, running into the visibility mismatch between files and keyrings and being unable to "lock" encrypted directories. Intuitively, the ioctls have to be privileged since they manipulate filesystem-level state. However, it's actually safe to make them unprivileged if we very carefully enforce some specific limitations. First, each key must be identified by a cryptographic hash so that a user can't add the wrong key for another user's files. For v2 encryption policies, we use the key_identifier for this. v1 policies don't have this, so managing keys for them remains privileged. Second, each key a user adds is charged to their quota for the keyrings service. Thus, a user can't exhaust memory by adding a huge number of keys. By default each non-root user is allowed up to 200 keys; this can be changed using the existing sysctl 'kernel.keys.maxkeys'. Third, if multiple users add the same key, we keep track of those users of the key (of which there remains a single copy), and won't really remove the key, i.e. "lock" the encrypted files, until all those users have removed it. This prevents denial of service attacks that would be possible under simpler schemes, such allowing the first user who added a key to remove it -- since that could be a malicious user who has compromised the key. Of course, encryption keys should be kept secret, but the idea is that using encryption should never be *less* secure than not using encryption, even if your key was compromised. We tolerate that a user will be unable to really remove a key, i.e. unable to "lock" their encrypted files, if another user has added the same key. But in a sense, this is actually a good thing because it will avoid providing a false notion of security where a key appears to have been removed when actually it's still in memory, available to any attacker who compromises the operating system kernel. Reviewed-by: Theodore Ts'o <tytso@mit.edu> Signed-off-by: Eric Biggers <ebiggers@google.com>
2019-08-05 10:35:47 +08:00
}
fscrypt: stop using keyrings subsystem for fscrypt_master_key The approach of fs/crypto/ internally managing the fscrypt_master_key structs as the payloads of "struct key" objects contained in a "struct key" keyring has outlived its usefulness. The original idea was to simplify the code by reusing code from the keyrings subsystem. However, several issues have arisen that can't easily be resolved: - When a master key struct is destroyed, blk_crypto_evict_key() must be called on any per-mode keys embedded in it. (This started being the case when inline encryption support was added.) Yet, the keyrings subsystem can arbitrarily delay the destruction of keys, even past the time the filesystem was unmounted. Therefore, currently there is no easy way to call blk_crypto_evict_key() when a master key is destroyed. Currently, this is worked around by holding an extra reference to the filesystem's request_queue(s). But it was overlooked that the request_queue reference is *not* guaranteed to pin the corresponding blk_crypto_profile too; for device-mapper devices that support inline crypto, it doesn't. This can cause a use-after-free. - When the last inode that was using an incompletely-removed master key is evicted, the master key removal is completed by removing the key struct from the keyring. Currently this is done via key_invalidate(). Yet, key_invalidate() takes the key semaphore. This can deadlock when called from the shrinker, since in fscrypt_ioctl_add_key(), memory is allocated with GFP_KERNEL under the same semaphore. - More generally, the fact that the keyrings subsystem can arbitrarily delay the destruction of keys (via garbage collection delay, or via random processes getting temporary key references) is undesirable, as it means we can't strictly guarantee that all secrets are ever wiped. - Doing the master key lookups via the keyrings subsystem results in the key_permission LSM hook being called. fscrypt doesn't want this, as all access control for encrypted files is designed to happen via the files themselves, like any other files. The workaround which SELinux users are using is to change their SELinux policy to grant key search access to all domains. This works, but it is an odd extra step that shouldn't really have to be done. The fix for all these issues is to change the implementation to what I should have done originally: don't use the keyrings subsystem to keep track of the filesystem's fscrypt_master_key structs. Instead, just store them in a regular kernel data structure, and rework the reference counting, locking, and lifetime accordingly. Retain support for RCU-mode key lookups by using a hash table. Replace fscrypt_sb_free() with fscrypt_sb_delete(), which releases the keys synchronously and runs a bit earlier during unmount, so that block devices are still available. A side effect of this patch is that neither the master keys themselves nor the filesystem keyrings will be listed in /proc/keys anymore. ("Master key users" and the master key users keyrings will still be listed.) However, this was mostly an implementation detail, and it was intended just for debugging purposes. I don't know of anyone using it. This patch does *not* change how "master key users" (->mk_users) works; that still uses the keyrings subsystem. That is still needed for key quotas, and changing that isn't necessary to solve the issues listed above. If we decide to change that too, it would be a separate patch. I've marked this as fixing the original commit that added the fscrypt keyring, but as noted above the most important issue that this patch fixes wasn't introduced until the addition of inline encryption support. Fixes: 22d94f493bfb ("fscrypt: add FS_IOC_ADD_ENCRYPTION_KEY ioctl") Signed-off-by: Eric Biggers <ebiggers@google.com> Link: https://lore.kernel.org/r/20220901193208.138056-2-ebiggers@kernel.org
2022-09-02 03:32:06 +08:00
move_master_key_secret(&mk->mk_secret, secret);
refcount_set(&mk->mk_active_refs, 1); /* ->mk_secret is present */
fscrypt: add FS_IOC_ADD_ENCRYPTION_KEY ioctl Add a new fscrypt ioctl, FS_IOC_ADD_ENCRYPTION_KEY. This ioctl adds an encryption key to the filesystem's fscrypt keyring ->s_master_keys, making any files encrypted with that key appear "unlocked". Why we need this ~~~~~~~~~~~~~~~~ The main problem is that the "locked/unlocked" (ciphertext/plaintext) status of encrypted files is global, but the fscrypt keys are not. fscrypt only looks for keys in the keyring(s) the process accessing the filesystem is subscribed to: the thread keyring, process keyring, and session keyring, where the session keyring may contain the user keyring. Therefore, userspace has to put fscrypt keys in the keyrings for individual users or sessions. But this means that when a process with a different keyring tries to access encrypted files, whether they appear "unlocked" or not is nondeterministic. This is because it depends on whether the files are currently present in the inode cache. Fixing this by consistently providing each process its own view of the filesystem depending on whether it has the key or not isn't feasible due to how the VFS caches work. Furthermore, while sometimes users expect this behavior, it is misguided for two reasons. First, it would be an OS-level access control mechanism largely redundant with existing access control mechanisms such as UNIX file permissions, ACLs, LSMs, etc. Encryption is actually for protecting the data at rest. Second, almost all users of fscrypt actually do need the keys to be global. The largest users of fscrypt, Android and Chromium OS, achieve this by having PID 1 create a "session keyring" that is inherited by every process. This works, but it isn't scalable because it prevents session keyrings from being used for any other purpose. On general-purpose Linux distros, the 'fscrypt' userspace tool [1] can't similarly abuse the session keyring, so to make 'sudo' work on all systems it has to link all the user keyrings into root's user keyring [2]. This is ugly and raises security concerns. Moreover it can't make the keys available to system services, such as sshd trying to access the user's '~/.ssh' directory (see [3], [4]) or NetworkManager trying to read certificates from the user's home directory (see [5]); or to Docker containers (see [6], [7]). By having an API to add a key to the *filesystem* we'll be able to fix the above bugs, remove userspace workarounds, and clearly express the intended semantics: the locked/unlocked status of an encrypted directory is global, and encryption is orthogonal to OS-level access control. Why not use the add_key() syscall ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ We use an ioctl for this API rather than the existing add_key() system call because the ioctl gives us the flexibility needed to implement fscrypt-specific semantics that will be introduced in later patches: - Supporting key removal with the semantics such that the secret is removed immediately and any unused inodes using the key are evicted; also, the eviction of any in-use inodes can be retried. - Calculating a key-dependent cryptographic identifier and returning it to userspace. - Allowing keys to be added and removed by non-root users, but only keys for v2 encryption policies; and to prevent denial-of-service attacks, users can only remove keys they themselves have added, and a key is only really removed after all users who added it have removed it. Trying to shoehorn these semantics into the keyrings syscalls would be very difficult, whereas the ioctls make things much easier. However, to reuse code the implementation still uses the keyrings service internally. Thus we get lockless RCU-mode key lookups without having to re-implement it, and the keys automatically show up in /proc/keys for debugging purposes. References: [1] https://github.com/google/fscrypt [2] https://goo.gl/55cCrI#heading=h.vf09isp98isb [3] https://github.com/google/fscrypt/issues/111#issuecomment-444347939 [4] https://github.com/google/fscrypt/issues/116 [5] https://bugs.launchpad.net/ubuntu/+source/fscrypt/+bug/1770715 [6] https://github.com/google/fscrypt/issues/128 [7] https://askubuntu.com/questions/1130306/cannot-run-docker-on-an-encrypted-filesystem Reviewed-by: Theodore Ts'o <tytso@mit.edu> Signed-off-by: Eric Biggers <ebiggers@google.com>
2019-08-05 10:35:46 +08:00
fscrypt: stop using keyrings subsystem for fscrypt_master_key The approach of fs/crypto/ internally managing the fscrypt_master_key structs as the payloads of "struct key" objects contained in a "struct key" keyring has outlived its usefulness. The original idea was to simplify the code by reusing code from the keyrings subsystem. However, several issues have arisen that can't easily be resolved: - When a master key struct is destroyed, blk_crypto_evict_key() must be called on any per-mode keys embedded in it. (This started being the case when inline encryption support was added.) Yet, the keyrings subsystem can arbitrarily delay the destruction of keys, even past the time the filesystem was unmounted. Therefore, currently there is no easy way to call blk_crypto_evict_key() when a master key is destroyed. Currently, this is worked around by holding an extra reference to the filesystem's request_queue(s). But it was overlooked that the request_queue reference is *not* guaranteed to pin the corresponding blk_crypto_profile too; for device-mapper devices that support inline crypto, it doesn't. This can cause a use-after-free. - When the last inode that was using an incompletely-removed master key is evicted, the master key removal is completed by removing the key struct from the keyring. Currently this is done via key_invalidate(). Yet, key_invalidate() takes the key semaphore. This can deadlock when called from the shrinker, since in fscrypt_ioctl_add_key(), memory is allocated with GFP_KERNEL under the same semaphore. - More generally, the fact that the keyrings subsystem can arbitrarily delay the destruction of keys (via garbage collection delay, or via random processes getting temporary key references) is undesirable, as it means we can't strictly guarantee that all secrets are ever wiped. - Doing the master key lookups via the keyrings subsystem results in the key_permission LSM hook being called. fscrypt doesn't want this, as all access control for encrypted files is designed to happen via the files themselves, like any other files. The workaround which SELinux users are using is to change their SELinux policy to grant key search access to all domains. This works, but it is an odd extra step that shouldn't really have to be done. The fix for all these issues is to change the implementation to what I should have done originally: don't use the keyrings subsystem to keep track of the filesystem's fscrypt_master_key structs. Instead, just store them in a regular kernel data structure, and rework the reference counting, locking, and lifetime accordingly. Retain support for RCU-mode key lookups by using a hash table. Replace fscrypt_sb_free() with fscrypt_sb_delete(), which releases the keys synchronously and runs a bit earlier during unmount, so that block devices are still available. A side effect of this patch is that neither the master keys themselves nor the filesystem keyrings will be listed in /proc/keys anymore. ("Master key users" and the master key users keyrings will still be listed.) However, this was mostly an implementation detail, and it was intended just for debugging purposes. I don't know of anyone using it. This patch does *not* change how "master key users" (->mk_users) works; that still uses the keyrings subsystem. That is still needed for key quotas, and changing that isn't necessary to solve the issues listed above. If we decide to change that too, it would be a separate patch. I've marked this as fixing the original commit that added the fscrypt keyring, but as noted above the most important issue that this patch fixes wasn't introduced until the addition of inline encryption support. Fixes: 22d94f493bfb ("fscrypt: add FS_IOC_ADD_ENCRYPTION_KEY ioctl") Signed-off-by: Eric Biggers <ebiggers@google.com> Link: https://lore.kernel.org/r/20220901193208.138056-2-ebiggers@kernel.org
2022-09-02 03:32:06 +08:00
spin_lock(&keyring->lock);
hlist_add_head_rcu(&mk->mk_node,
fscrypt_mk_hash_bucket(keyring, mk_spec));
spin_unlock(&keyring->lock);
fscrypt: add FS_IOC_ADD_ENCRYPTION_KEY ioctl Add a new fscrypt ioctl, FS_IOC_ADD_ENCRYPTION_KEY. This ioctl adds an encryption key to the filesystem's fscrypt keyring ->s_master_keys, making any files encrypted with that key appear "unlocked". Why we need this ~~~~~~~~~~~~~~~~ The main problem is that the "locked/unlocked" (ciphertext/plaintext) status of encrypted files is global, but the fscrypt keys are not. fscrypt only looks for keys in the keyring(s) the process accessing the filesystem is subscribed to: the thread keyring, process keyring, and session keyring, where the session keyring may contain the user keyring. Therefore, userspace has to put fscrypt keys in the keyrings for individual users or sessions. But this means that when a process with a different keyring tries to access encrypted files, whether they appear "unlocked" or not is nondeterministic. This is because it depends on whether the files are currently present in the inode cache. Fixing this by consistently providing each process its own view of the filesystem depending on whether it has the key or not isn't feasible due to how the VFS caches work. Furthermore, while sometimes users expect this behavior, it is misguided for two reasons. First, it would be an OS-level access control mechanism largely redundant with existing access control mechanisms such as UNIX file permissions, ACLs, LSMs, etc. Encryption is actually for protecting the data at rest. Second, almost all users of fscrypt actually do need the keys to be global. The largest users of fscrypt, Android and Chromium OS, achieve this by having PID 1 create a "session keyring" that is inherited by every process. This works, but it isn't scalable because it prevents session keyrings from being used for any other purpose. On general-purpose Linux distros, the 'fscrypt' userspace tool [1] can't similarly abuse the session keyring, so to make 'sudo' work on all systems it has to link all the user keyrings into root's user keyring [2]. This is ugly and raises security concerns. Moreover it can't make the keys available to system services, such as sshd trying to access the user's '~/.ssh' directory (see [3], [4]) or NetworkManager trying to read certificates from the user's home directory (see [5]); or to Docker containers (see [6], [7]). By having an API to add a key to the *filesystem* we'll be able to fix the above bugs, remove userspace workarounds, and clearly express the intended semantics: the locked/unlocked status of an encrypted directory is global, and encryption is orthogonal to OS-level access control. Why not use the add_key() syscall ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ We use an ioctl for this API rather than the existing add_key() system call because the ioctl gives us the flexibility needed to implement fscrypt-specific semantics that will be introduced in later patches: - Supporting key removal with the semantics such that the secret is removed immediately and any unused inodes using the key are evicted; also, the eviction of any in-use inodes can be retried. - Calculating a key-dependent cryptographic identifier and returning it to userspace. - Allowing keys to be added and removed by non-root users, but only keys for v2 encryption policies; and to prevent denial-of-service attacks, users can only remove keys they themselves have added, and a key is only really removed after all users who added it have removed it. Trying to shoehorn these semantics into the keyrings syscalls would be very difficult, whereas the ioctls make things much easier. However, to reuse code the implementation still uses the keyrings service internally. Thus we get lockless RCU-mode key lookups without having to re-implement it, and the keys automatically show up in /proc/keys for debugging purposes. References: [1] https://github.com/google/fscrypt [2] https://goo.gl/55cCrI#heading=h.vf09isp98isb [3] https://github.com/google/fscrypt/issues/111#issuecomment-444347939 [4] https://github.com/google/fscrypt/issues/116 [5] https://bugs.launchpad.net/ubuntu/+source/fscrypt/+bug/1770715 [6] https://github.com/google/fscrypt/issues/128 [7] https://askubuntu.com/questions/1130306/cannot-run-docker-on-an-encrypted-filesystem Reviewed-by: Theodore Ts'o <tytso@mit.edu> Signed-off-by: Eric Biggers <ebiggers@google.com>
2019-08-05 10:35:46 +08:00
return 0;
fscrypt: stop using keyrings subsystem for fscrypt_master_key The approach of fs/crypto/ internally managing the fscrypt_master_key structs as the payloads of "struct key" objects contained in a "struct key" keyring has outlived its usefulness. The original idea was to simplify the code by reusing code from the keyrings subsystem. However, several issues have arisen that can't easily be resolved: - When a master key struct is destroyed, blk_crypto_evict_key() must be called on any per-mode keys embedded in it. (This started being the case when inline encryption support was added.) Yet, the keyrings subsystem can arbitrarily delay the destruction of keys, even past the time the filesystem was unmounted. Therefore, currently there is no easy way to call blk_crypto_evict_key() when a master key is destroyed. Currently, this is worked around by holding an extra reference to the filesystem's request_queue(s). But it was overlooked that the request_queue reference is *not* guaranteed to pin the corresponding blk_crypto_profile too; for device-mapper devices that support inline crypto, it doesn't. This can cause a use-after-free. - When the last inode that was using an incompletely-removed master key is evicted, the master key removal is completed by removing the key struct from the keyring. Currently this is done via key_invalidate(). Yet, key_invalidate() takes the key semaphore. This can deadlock when called from the shrinker, since in fscrypt_ioctl_add_key(), memory is allocated with GFP_KERNEL under the same semaphore. - More generally, the fact that the keyrings subsystem can arbitrarily delay the destruction of keys (via garbage collection delay, or via random processes getting temporary key references) is undesirable, as it means we can't strictly guarantee that all secrets are ever wiped. - Doing the master key lookups via the keyrings subsystem results in the key_permission LSM hook being called. fscrypt doesn't want this, as all access control for encrypted files is designed to happen via the files themselves, like any other files. The workaround which SELinux users are using is to change their SELinux policy to grant key search access to all domains. This works, but it is an odd extra step that shouldn't really have to be done. The fix for all these issues is to change the implementation to what I should have done originally: don't use the keyrings subsystem to keep track of the filesystem's fscrypt_master_key structs. Instead, just store them in a regular kernel data structure, and rework the reference counting, locking, and lifetime accordingly. Retain support for RCU-mode key lookups by using a hash table. Replace fscrypt_sb_free() with fscrypt_sb_delete(), which releases the keys synchronously and runs a bit earlier during unmount, so that block devices are still available. A side effect of this patch is that neither the master keys themselves nor the filesystem keyrings will be listed in /proc/keys anymore. ("Master key users" and the master key users keyrings will still be listed.) However, this was mostly an implementation detail, and it was intended just for debugging purposes. I don't know of anyone using it. This patch does *not* change how "master key users" (->mk_users) works; that still uses the keyrings subsystem. That is still needed for key quotas, and changing that isn't necessary to solve the issues listed above. If we decide to change that too, it would be a separate patch. I've marked this as fixing the original commit that added the fscrypt keyring, but as noted above the most important issue that this patch fixes wasn't introduced until the addition of inline encryption support. Fixes: 22d94f493bfb ("fscrypt: add FS_IOC_ADD_ENCRYPTION_KEY ioctl") Signed-off-by: Eric Biggers <ebiggers@google.com> Link: https://lore.kernel.org/r/20220901193208.138056-2-ebiggers@kernel.org
2022-09-02 03:32:06 +08:00
out_put:
fscrypt_put_master_key(mk);
fscrypt: add FS_IOC_ADD_ENCRYPTION_KEY ioctl Add a new fscrypt ioctl, FS_IOC_ADD_ENCRYPTION_KEY. This ioctl adds an encryption key to the filesystem's fscrypt keyring ->s_master_keys, making any files encrypted with that key appear "unlocked". Why we need this ~~~~~~~~~~~~~~~~ The main problem is that the "locked/unlocked" (ciphertext/plaintext) status of encrypted files is global, but the fscrypt keys are not. fscrypt only looks for keys in the keyring(s) the process accessing the filesystem is subscribed to: the thread keyring, process keyring, and session keyring, where the session keyring may contain the user keyring. Therefore, userspace has to put fscrypt keys in the keyrings for individual users or sessions. But this means that when a process with a different keyring tries to access encrypted files, whether they appear "unlocked" or not is nondeterministic. This is because it depends on whether the files are currently present in the inode cache. Fixing this by consistently providing each process its own view of the filesystem depending on whether it has the key or not isn't feasible due to how the VFS caches work. Furthermore, while sometimes users expect this behavior, it is misguided for two reasons. First, it would be an OS-level access control mechanism largely redundant with existing access control mechanisms such as UNIX file permissions, ACLs, LSMs, etc. Encryption is actually for protecting the data at rest. Second, almost all users of fscrypt actually do need the keys to be global. The largest users of fscrypt, Android and Chromium OS, achieve this by having PID 1 create a "session keyring" that is inherited by every process. This works, but it isn't scalable because it prevents session keyrings from being used for any other purpose. On general-purpose Linux distros, the 'fscrypt' userspace tool [1] can't similarly abuse the session keyring, so to make 'sudo' work on all systems it has to link all the user keyrings into root's user keyring [2]. This is ugly and raises security concerns. Moreover it can't make the keys available to system services, such as sshd trying to access the user's '~/.ssh' directory (see [3], [4]) or NetworkManager trying to read certificates from the user's home directory (see [5]); or to Docker containers (see [6], [7]). By having an API to add a key to the *filesystem* we'll be able to fix the above bugs, remove userspace workarounds, and clearly express the intended semantics: the locked/unlocked status of an encrypted directory is global, and encryption is orthogonal to OS-level access control. Why not use the add_key() syscall ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ We use an ioctl for this API rather than the existing add_key() system call because the ioctl gives us the flexibility needed to implement fscrypt-specific semantics that will be introduced in later patches: - Supporting key removal with the semantics such that the secret is removed immediately and any unused inodes using the key are evicted; also, the eviction of any in-use inodes can be retried. - Calculating a key-dependent cryptographic identifier and returning it to userspace. - Allowing keys to be added and removed by non-root users, but only keys for v2 encryption policies; and to prevent denial-of-service attacks, users can only remove keys they themselves have added, and a key is only really removed after all users who added it have removed it. Trying to shoehorn these semantics into the keyrings syscalls would be very difficult, whereas the ioctls make things much easier. However, to reuse code the implementation still uses the keyrings service internally. Thus we get lockless RCU-mode key lookups without having to re-implement it, and the keys automatically show up in /proc/keys for debugging purposes. References: [1] https://github.com/google/fscrypt [2] https://goo.gl/55cCrI#heading=h.vf09isp98isb [3] https://github.com/google/fscrypt/issues/111#issuecomment-444347939 [4] https://github.com/google/fscrypt/issues/116 [5] https://bugs.launchpad.net/ubuntu/+source/fscrypt/+bug/1770715 [6] https://github.com/google/fscrypt/issues/128 [7] https://askubuntu.com/questions/1130306/cannot-run-docker-on-an-encrypted-filesystem Reviewed-by: Theodore Ts'o <tytso@mit.edu> Signed-off-by: Eric Biggers <ebiggers@google.com>
2019-08-05 10:35:46 +08:00
return err;
}
fscrypt: add FS_IOC_REMOVE_ENCRYPTION_KEY ioctl Add a new fscrypt ioctl, FS_IOC_REMOVE_ENCRYPTION_KEY. This ioctl removes an encryption key that was added by FS_IOC_ADD_ENCRYPTION_KEY. It wipes the secret key itself, then "locks" the encrypted files and directories that had been unlocked using that key -- implemented by evicting the relevant dentries and inodes from the VFS caches. The problem this solves is that many fscrypt users want the ability to remove encryption keys, causing the corresponding encrypted directories to appear "locked" (presented in ciphertext form) again. Moreover, users want removing an encryption key to *really* remove it, in the sense that the removed keys cannot be recovered even if kernel memory is compromised, e.g. by the exploit of a kernel security vulnerability or by a physical attack. This is desirable after a user logs out of the system, for example. In many cases users even already assume this to be the case and are surprised to hear when it's not. It is not sufficient to simply unlink the master key from the keyring (or to revoke or invalidate it), since the actual encryption transform objects are still pinned in memory by their inodes. Therefore, to really remove a key we must also evict the relevant inodes. Currently one workaround is to run 'sync && echo 2 > /proc/sys/vm/drop_caches'. But, that evicts all unused inodes in the system rather than just the inodes associated with the key being removed, causing severe performance problems. Moreover, it requires root privileges, so regular users can't "lock" their encrypted files. Another workaround, used in Chromium OS kernels, is to add a new VFS-level ioctl FS_IOC_DROP_CACHE which is a more restricted version of drop_caches that operates on a single super_block. It does: shrink_dcache_sb(sb); invalidate_inodes(sb, false); But it's still a hack. Yet, the major users of filesystem encryption want this feature badly enough that they are actually using these hacks. To properly solve the problem, start maintaining a list of the inodes which have been "unlocked" using each master key. Originally this wasn't possible because the kernel didn't keep track of in-use master keys at all. But, with the ->s_master_keys keyring it is now possible. Then, add an ioctl FS_IOC_REMOVE_ENCRYPTION_KEY. It finds the specified master key in ->s_master_keys, then wipes the secret key itself, which prevents any additional inodes from being unlocked with the key. Then, it syncs the filesystem and evicts the inodes in the key's list. The normal inode eviction code will free and wipe the per-file keys (in ->i_crypt_info). Note that freeing ->i_crypt_info without evicting the inodes was also considered, but would have been racy. Some inodes may still be in use when a master key is removed, and we can't simply revoke random file descriptors, mmap's, etc. Thus, the ioctl simply skips in-use inodes, and returns -EBUSY to indicate that some inodes weren't evicted. The master key *secret* is still removed, but the fscrypt_master_key struct remains to keep track of the remaining inodes. Userspace can then retry the ioctl to evict the remaining inodes. Alternatively, if userspace adds the key again, the refreshed secret will be associated with the existing list of inodes so they remain correctly tracked for future key removals. The ioctl doesn't wipe pagecache pages. Thus, we tolerate that after a kernel compromise some portions of plaintext file contents may still be recoverable from memory. This can be solved by enabling page poisoning system-wide, which security conscious users may choose to do. But it's very difficult to solve otherwise, e.g. note that plaintext file contents may have been read in other places than pagecache pages. Like FS_IOC_ADD_ENCRYPTION_KEY, FS_IOC_REMOVE_ENCRYPTION_KEY is initially restricted to privileged users only. This is sufficient for some use cases, but not all. A later patch will relax this restriction, but it will require introducing key hashes, among other changes. Reviewed-by: Theodore Ts'o <tytso@mit.edu> Signed-off-by: Eric Biggers <ebiggers@google.com>
2019-08-05 10:35:46 +08:00
#define KEY_DEAD 1
static int add_existing_master_key(struct fscrypt_master_key *mk,
struct fscrypt_master_key_secret *secret)
{
fscrypt: allow unprivileged users to add/remove keys for v2 policies Allow the FS_IOC_ADD_ENCRYPTION_KEY and FS_IOC_REMOVE_ENCRYPTION_KEY ioctls to be used by non-root users to add and remove encryption keys from the filesystem-level crypto keyrings, subject to limitations. Motivation: while privileged fscrypt key management is sufficient for some users (e.g. Android and Chromium OS, where a privileged process manages all keys), the old API by design also allows non-root users to set up and use encrypted directories, and we don't want to regress on that. Especially, we don't want to force users to continue using the old API, running into the visibility mismatch between files and keyrings and being unable to "lock" encrypted directories. Intuitively, the ioctls have to be privileged since they manipulate filesystem-level state. However, it's actually safe to make them unprivileged if we very carefully enforce some specific limitations. First, each key must be identified by a cryptographic hash so that a user can't add the wrong key for another user's files. For v2 encryption policies, we use the key_identifier for this. v1 policies don't have this, so managing keys for them remains privileged. Second, each key a user adds is charged to their quota for the keyrings service. Thus, a user can't exhaust memory by adding a huge number of keys. By default each non-root user is allowed up to 200 keys; this can be changed using the existing sysctl 'kernel.keys.maxkeys'. Third, if multiple users add the same key, we keep track of those users of the key (of which there remains a single copy), and won't really remove the key, i.e. "lock" the encrypted files, until all those users have removed it. This prevents denial of service attacks that would be possible under simpler schemes, such allowing the first user who added a key to remove it -- since that could be a malicious user who has compromised the key. Of course, encryption keys should be kept secret, but the idea is that using encryption should never be *less* secure than not using encryption, even if your key was compromised. We tolerate that a user will be unable to really remove a key, i.e. unable to "lock" their encrypted files, if another user has added the same key. But in a sense, this is actually a good thing because it will avoid providing a false notion of security where a key appears to have been removed when actually it's still in memory, available to any attacker who compromises the operating system kernel. Reviewed-by: Theodore Ts'o <tytso@mit.edu> Signed-off-by: Eric Biggers <ebiggers@google.com>
2019-08-05 10:35:47 +08:00
int err;
fscrypt: add FS_IOC_REMOVE_ENCRYPTION_KEY ioctl Add a new fscrypt ioctl, FS_IOC_REMOVE_ENCRYPTION_KEY. This ioctl removes an encryption key that was added by FS_IOC_ADD_ENCRYPTION_KEY. It wipes the secret key itself, then "locks" the encrypted files and directories that had been unlocked using that key -- implemented by evicting the relevant dentries and inodes from the VFS caches. The problem this solves is that many fscrypt users want the ability to remove encryption keys, causing the corresponding encrypted directories to appear "locked" (presented in ciphertext form) again. Moreover, users want removing an encryption key to *really* remove it, in the sense that the removed keys cannot be recovered even if kernel memory is compromised, e.g. by the exploit of a kernel security vulnerability or by a physical attack. This is desirable after a user logs out of the system, for example. In many cases users even already assume this to be the case and are surprised to hear when it's not. It is not sufficient to simply unlink the master key from the keyring (or to revoke or invalidate it), since the actual encryption transform objects are still pinned in memory by their inodes. Therefore, to really remove a key we must also evict the relevant inodes. Currently one workaround is to run 'sync && echo 2 > /proc/sys/vm/drop_caches'. But, that evicts all unused inodes in the system rather than just the inodes associated with the key being removed, causing severe performance problems. Moreover, it requires root privileges, so regular users can't "lock" their encrypted files. Another workaround, used in Chromium OS kernels, is to add a new VFS-level ioctl FS_IOC_DROP_CACHE which is a more restricted version of drop_caches that operates on a single super_block. It does: shrink_dcache_sb(sb); invalidate_inodes(sb, false); But it's still a hack. Yet, the major users of filesystem encryption want this feature badly enough that they are actually using these hacks. To properly solve the problem, start maintaining a list of the inodes which have been "unlocked" using each master key. Originally this wasn't possible because the kernel didn't keep track of in-use master keys at all. But, with the ->s_master_keys keyring it is now possible. Then, add an ioctl FS_IOC_REMOVE_ENCRYPTION_KEY. It finds the specified master key in ->s_master_keys, then wipes the secret key itself, which prevents any additional inodes from being unlocked with the key. Then, it syncs the filesystem and evicts the inodes in the key's list. The normal inode eviction code will free and wipe the per-file keys (in ->i_crypt_info). Note that freeing ->i_crypt_info without evicting the inodes was also considered, but would have been racy. Some inodes may still be in use when a master key is removed, and we can't simply revoke random file descriptors, mmap's, etc. Thus, the ioctl simply skips in-use inodes, and returns -EBUSY to indicate that some inodes weren't evicted. The master key *secret* is still removed, but the fscrypt_master_key struct remains to keep track of the remaining inodes. Userspace can then retry the ioctl to evict the remaining inodes. Alternatively, if userspace adds the key again, the refreshed secret will be associated with the existing list of inodes so they remain correctly tracked for future key removals. The ioctl doesn't wipe pagecache pages. Thus, we tolerate that after a kernel compromise some portions of plaintext file contents may still be recoverable from memory. This can be solved by enabling page poisoning system-wide, which security conscious users may choose to do. But it's very difficult to solve otherwise, e.g. note that plaintext file contents may have been read in other places than pagecache pages. Like FS_IOC_ADD_ENCRYPTION_KEY, FS_IOC_REMOVE_ENCRYPTION_KEY is initially restricted to privileged users only. This is sufficient for some use cases, but not all. A later patch will relax this restriction, but it will require introducing key hashes, among other changes. Reviewed-by: Theodore Ts'o <tytso@mit.edu> Signed-off-by: Eric Biggers <ebiggers@google.com>
2019-08-05 10:35:46 +08:00
fscrypt: allow unprivileged users to add/remove keys for v2 policies Allow the FS_IOC_ADD_ENCRYPTION_KEY and FS_IOC_REMOVE_ENCRYPTION_KEY ioctls to be used by non-root users to add and remove encryption keys from the filesystem-level crypto keyrings, subject to limitations. Motivation: while privileged fscrypt key management is sufficient for some users (e.g. Android and Chromium OS, where a privileged process manages all keys), the old API by design also allows non-root users to set up and use encrypted directories, and we don't want to regress on that. Especially, we don't want to force users to continue using the old API, running into the visibility mismatch between files and keyrings and being unable to "lock" encrypted directories. Intuitively, the ioctls have to be privileged since they manipulate filesystem-level state. However, it's actually safe to make them unprivileged if we very carefully enforce some specific limitations. First, each key must be identified by a cryptographic hash so that a user can't add the wrong key for another user's files. For v2 encryption policies, we use the key_identifier for this. v1 policies don't have this, so managing keys for them remains privileged. Second, each key a user adds is charged to their quota for the keyrings service. Thus, a user can't exhaust memory by adding a huge number of keys. By default each non-root user is allowed up to 200 keys; this can be changed using the existing sysctl 'kernel.keys.maxkeys'. Third, if multiple users add the same key, we keep track of those users of the key (of which there remains a single copy), and won't really remove the key, i.e. "lock" the encrypted files, until all those users have removed it. This prevents denial of service attacks that would be possible under simpler schemes, such allowing the first user who added a key to remove it -- since that could be a malicious user who has compromised the key. Of course, encryption keys should be kept secret, but the idea is that using encryption should never be *less* secure than not using encryption, even if your key was compromised. We tolerate that a user will be unable to really remove a key, i.e. unable to "lock" their encrypted files, if another user has added the same key. But in a sense, this is actually a good thing because it will avoid providing a false notion of security where a key appears to have been removed when actually it's still in memory, available to any attacker who compromises the operating system kernel. Reviewed-by: Theodore Ts'o <tytso@mit.edu> Signed-off-by: Eric Biggers <ebiggers@google.com>
2019-08-05 10:35:47 +08:00
/*
* If the current user is already in ->mk_users, then there's nothing to
fscrypt: stop using keyrings subsystem for fscrypt_master_key The approach of fs/crypto/ internally managing the fscrypt_master_key structs as the payloads of "struct key" objects contained in a "struct key" keyring has outlived its usefulness. The original idea was to simplify the code by reusing code from the keyrings subsystem. However, several issues have arisen that can't easily be resolved: - When a master key struct is destroyed, blk_crypto_evict_key() must be called on any per-mode keys embedded in it. (This started being the case when inline encryption support was added.) Yet, the keyrings subsystem can arbitrarily delay the destruction of keys, even past the time the filesystem was unmounted. Therefore, currently there is no easy way to call blk_crypto_evict_key() when a master key is destroyed. Currently, this is worked around by holding an extra reference to the filesystem's request_queue(s). But it was overlooked that the request_queue reference is *not* guaranteed to pin the corresponding blk_crypto_profile too; for device-mapper devices that support inline crypto, it doesn't. This can cause a use-after-free. - When the last inode that was using an incompletely-removed master key is evicted, the master key removal is completed by removing the key struct from the keyring. Currently this is done via key_invalidate(). Yet, key_invalidate() takes the key semaphore. This can deadlock when called from the shrinker, since in fscrypt_ioctl_add_key(), memory is allocated with GFP_KERNEL under the same semaphore. - More generally, the fact that the keyrings subsystem can arbitrarily delay the destruction of keys (via garbage collection delay, or via random processes getting temporary key references) is undesirable, as it means we can't strictly guarantee that all secrets are ever wiped. - Doing the master key lookups via the keyrings subsystem results in the key_permission LSM hook being called. fscrypt doesn't want this, as all access control for encrypted files is designed to happen via the files themselves, like any other files. The workaround which SELinux users are using is to change their SELinux policy to grant key search access to all domains. This works, but it is an odd extra step that shouldn't really have to be done. The fix for all these issues is to change the implementation to what I should have done originally: don't use the keyrings subsystem to keep track of the filesystem's fscrypt_master_key structs. Instead, just store them in a regular kernel data structure, and rework the reference counting, locking, and lifetime accordingly. Retain support for RCU-mode key lookups by using a hash table. Replace fscrypt_sb_free() with fscrypt_sb_delete(), which releases the keys synchronously and runs a bit earlier during unmount, so that block devices are still available. A side effect of this patch is that neither the master keys themselves nor the filesystem keyrings will be listed in /proc/keys anymore. ("Master key users" and the master key users keyrings will still be listed.) However, this was mostly an implementation detail, and it was intended just for debugging purposes. I don't know of anyone using it. This patch does *not* change how "master key users" (->mk_users) works; that still uses the keyrings subsystem. That is still needed for key quotas, and changing that isn't necessary to solve the issues listed above. If we decide to change that too, it would be a separate patch. I've marked this as fixing the original commit that added the fscrypt keyring, but as noted above the most important issue that this patch fixes wasn't introduced until the addition of inline encryption support. Fixes: 22d94f493bfb ("fscrypt: add FS_IOC_ADD_ENCRYPTION_KEY ioctl") Signed-off-by: Eric Biggers <ebiggers@google.com> Link: https://lore.kernel.org/r/20220901193208.138056-2-ebiggers@kernel.org
2022-09-02 03:32:06 +08:00
* do. Otherwise, we need to add the user to ->mk_users. (Neither is
* applicable for v1 policy keys, which have NULL ->mk_users.)
fscrypt: allow unprivileged users to add/remove keys for v2 policies Allow the FS_IOC_ADD_ENCRYPTION_KEY and FS_IOC_REMOVE_ENCRYPTION_KEY ioctls to be used by non-root users to add and remove encryption keys from the filesystem-level crypto keyrings, subject to limitations. Motivation: while privileged fscrypt key management is sufficient for some users (e.g. Android and Chromium OS, where a privileged process manages all keys), the old API by design also allows non-root users to set up and use encrypted directories, and we don't want to regress on that. Especially, we don't want to force users to continue using the old API, running into the visibility mismatch between files and keyrings and being unable to "lock" encrypted directories. Intuitively, the ioctls have to be privileged since they manipulate filesystem-level state. However, it's actually safe to make them unprivileged if we very carefully enforce some specific limitations. First, each key must be identified by a cryptographic hash so that a user can't add the wrong key for another user's files. For v2 encryption policies, we use the key_identifier for this. v1 policies don't have this, so managing keys for them remains privileged. Second, each key a user adds is charged to their quota for the keyrings service. Thus, a user can't exhaust memory by adding a huge number of keys. By default each non-root user is allowed up to 200 keys; this can be changed using the existing sysctl 'kernel.keys.maxkeys'. Third, if multiple users add the same key, we keep track of those users of the key (of which there remains a single copy), and won't really remove the key, i.e. "lock" the encrypted files, until all those users have removed it. This prevents denial of service attacks that would be possible under simpler schemes, such allowing the first user who added a key to remove it -- since that could be a malicious user who has compromised the key. Of course, encryption keys should be kept secret, but the idea is that using encryption should never be *less* secure than not using encryption, even if your key was compromised. We tolerate that a user will be unable to really remove a key, i.e. unable to "lock" their encrypted files, if another user has added the same key. But in a sense, this is actually a good thing because it will avoid providing a false notion of security where a key appears to have been removed when actually it's still in memory, available to any attacker who compromises the operating system kernel. Reviewed-by: Theodore Ts'o <tytso@mit.edu> Signed-off-by: Eric Biggers <ebiggers@google.com>
2019-08-05 10:35:47 +08:00
*/
if (mk->mk_users) {
fscrypt: stop using keyrings subsystem for fscrypt_master_key The approach of fs/crypto/ internally managing the fscrypt_master_key structs as the payloads of "struct key" objects contained in a "struct key" keyring has outlived its usefulness. The original idea was to simplify the code by reusing code from the keyrings subsystem. However, several issues have arisen that can't easily be resolved: - When a master key struct is destroyed, blk_crypto_evict_key() must be called on any per-mode keys embedded in it. (This started being the case when inline encryption support was added.) Yet, the keyrings subsystem can arbitrarily delay the destruction of keys, even past the time the filesystem was unmounted. Therefore, currently there is no easy way to call blk_crypto_evict_key() when a master key is destroyed. Currently, this is worked around by holding an extra reference to the filesystem's request_queue(s). But it was overlooked that the request_queue reference is *not* guaranteed to pin the corresponding blk_crypto_profile too; for device-mapper devices that support inline crypto, it doesn't. This can cause a use-after-free. - When the last inode that was using an incompletely-removed master key is evicted, the master key removal is completed by removing the key struct from the keyring. Currently this is done via key_invalidate(). Yet, key_invalidate() takes the key semaphore. This can deadlock when called from the shrinker, since in fscrypt_ioctl_add_key(), memory is allocated with GFP_KERNEL under the same semaphore. - More generally, the fact that the keyrings subsystem can arbitrarily delay the destruction of keys (via garbage collection delay, or via random processes getting temporary key references) is undesirable, as it means we can't strictly guarantee that all secrets are ever wiped. - Doing the master key lookups via the keyrings subsystem results in the key_permission LSM hook being called. fscrypt doesn't want this, as all access control for encrypted files is designed to happen via the files themselves, like any other files. The workaround which SELinux users are using is to change their SELinux policy to grant key search access to all domains. This works, but it is an odd extra step that shouldn't really have to be done. The fix for all these issues is to change the implementation to what I should have done originally: don't use the keyrings subsystem to keep track of the filesystem's fscrypt_master_key structs. Instead, just store them in a regular kernel data structure, and rework the reference counting, locking, and lifetime accordingly. Retain support for RCU-mode key lookups by using a hash table. Replace fscrypt_sb_free() with fscrypt_sb_delete(), which releases the keys synchronously and runs a bit earlier during unmount, so that block devices are still available. A side effect of this patch is that neither the master keys themselves nor the filesystem keyrings will be listed in /proc/keys anymore. ("Master key users" and the master key users keyrings will still be listed.) However, this was mostly an implementation detail, and it was intended just for debugging purposes. I don't know of anyone using it. This patch does *not* change how "master key users" (->mk_users) works; that still uses the keyrings subsystem. That is still needed for key quotas, and changing that isn't necessary to solve the issues listed above. If we decide to change that too, it would be a separate patch. I've marked this as fixing the original commit that added the fscrypt keyring, but as noted above the most important issue that this patch fixes wasn't introduced until the addition of inline encryption support. Fixes: 22d94f493bfb ("fscrypt: add FS_IOC_ADD_ENCRYPTION_KEY ioctl") Signed-off-by: Eric Biggers <ebiggers@google.com> Link: https://lore.kernel.org/r/20220901193208.138056-2-ebiggers@kernel.org
2022-09-02 03:32:06 +08:00
struct key *mk_user = find_master_key_user(mk);
fscrypt: allow unprivileged users to add/remove keys for v2 policies Allow the FS_IOC_ADD_ENCRYPTION_KEY and FS_IOC_REMOVE_ENCRYPTION_KEY ioctls to be used by non-root users to add and remove encryption keys from the filesystem-level crypto keyrings, subject to limitations. Motivation: while privileged fscrypt key management is sufficient for some users (e.g. Android and Chromium OS, where a privileged process manages all keys), the old API by design also allows non-root users to set up and use encrypted directories, and we don't want to regress on that. Especially, we don't want to force users to continue using the old API, running into the visibility mismatch between files and keyrings and being unable to "lock" encrypted directories. Intuitively, the ioctls have to be privileged since they manipulate filesystem-level state. However, it's actually safe to make them unprivileged if we very carefully enforce some specific limitations. First, each key must be identified by a cryptographic hash so that a user can't add the wrong key for another user's files. For v2 encryption policies, we use the key_identifier for this. v1 policies don't have this, so managing keys for them remains privileged. Second, each key a user adds is charged to their quota for the keyrings service. Thus, a user can't exhaust memory by adding a huge number of keys. By default each non-root user is allowed up to 200 keys; this can be changed using the existing sysctl 'kernel.keys.maxkeys'. Third, if multiple users add the same key, we keep track of those users of the key (of which there remains a single copy), and won't really remove the key, i.e. "lock" the encrypted files, until all those users have removed it. This prevents denial of service attacks that would be possible under simpler schemes, such allowing the first user who added a key to remove it -- since that could be a malicious user who has compromised the key. Of course, encryption keys should be kept secret, but the idea is that using encryption should never be *less* secure than not using encryption, even if your key was compromised. We tolerate that a user will be unable to really remove a key, i.e. unable to "lock" their encrypted files, if another user has added the same key. But in a sense, this is actually a good thing because it will avoid providing a false notion of security where a key appears to have been removed when actually it's still in memory, available to any attacker who compromises the operating system kernel. Reviewed-by: Theodore Ts'o <tytso@mit.edu> Signed-off-by: Eric Biggers <ebiggers@google.com>
2019-08-05 10:35:47 +08:00
if (mk_user != ERR_PTR(-ENOKEY)) {
if (IS_ERR(mk_user))
return PTR_ERR(mk_user);
key_put(mk_user);
return 0;
}
err = add_master_key_user(mk);
fscrypt: stop using keyrings subsystem for fscrypt_master_key The approach of fs/crypto/ internally managing the fscrypt_master_key structs as the payloads of "struct key" objects contained in a "struct key" keyring has outlived its usefulness. The original idea was to simplify the code by reusing code from the keyrings subsystem. However, several issues have arisen that can't easily be resolved: - When a master key struct is destroyed, blk_crypto_evict_key() must be called on any per-mode keys embedded in it. (This started being the case when inline encryption support was added.) Yet, the keyrings subsystem can arbitrarily delay the destruction of keys, even past the time the filesystem was unmounted. Therefore, currently there is no easy way to call blk_crypto_evict_key() when a master key is destroyed. Currently, this is worked around by holding an extra reference to the filesystem's request_queue(s). But it was overlooked that the request_queue reference is *not* guaranteed to pin the corresponding blk_crypto_profile too; for device-mapper devices that support inline crypto, it doesn't. This can cause a use-after-free. - When the last inode that was using an incompletely-removed master key is evicted, the master key removal is completed by removing the key struct from the keyring. Currently this is done via key_invalidate(). Yet, key_invalidate() takes the key semaphore. This can deadlock when called from the shrinker, since in fscrypt_ioctl_add_key(), memory is allocated with GFP_KERNEL under the same semaphore. - More generally, the fact that the keyrings subsystem can arbitrarily delay the destruction of keys (via garbage collection delay, or via random processes getting temporary key references) is undesirable, as it means we can't strictly guarantee that all secrets are ever wiped. - Doing the master key lookups via the keyrings subsystem results in the key_permission LSM hook being called. fscrypt doesn't want this, as all access control for encrypted files is designed to happen via the files themselves, like any other files. The workaround which SELinux users are using is to change their SELinux policy to grant key search access to all domains. This works, but it is an odd extra step that shouldn't really have to be done. The fix for all these issues is to change the implementation to what I should have done originally: don't use the keyrings subsystem to keep track of the filesystem's fscrypt_master_key structs. Instead, just store them in a regular kernel data structure, and rework the reference counting, locking, and lifetime accordingly. Retain support for RCU-mode key lookups by using a hash table. Replace fscrypt_sb_free() with fscrypt_sb_delete(), which releases the keys synchronously and runs a bit earlier during unmount, so that block devices are still available. A side effect of this patch is that neither the master keys themselves nor the filesystem keyrings will be listed in /proc/keys anymore. ("Master key users" and the master key users keyrings will still be listed.) However, this was mostly an implementation detail, and it was intended just for debugging purposes. I don't know of anyone using it. This patch does *not* change how "master key users" (->mk_users) works; that still uses the keyrings subsystem. That is still needed for key quotas, and changing that isn't necessary to solve the issues listed above. If we decide to change that too, it would be a separate patch. I've marked this as fixing the original commit that added the fscrypt keyring, but as noted above the most important issue that this patch fixes wasn't introduced until the addition of inline encryption support. Fixes: 22d94f493bfb ("fscrypt: add FS_IOC_ADD_ENCRYPTION_KEY ioctl") Signed-off-by: Eric Biggers <ebiggers@google.com> Link: https://lore.kernel.org/r/20220901193208.138056-2-ebiggers@kernel.org
2022-09-02 03:32:06 +08:00
if (err)
fscrypt: allow unprivileged users to add/remove keys for v2 policies Allow the FS_IOC_ADD_ENCRYPTION_KEY and FS_IOC_REMOVE_ENCRYPTION_KEY ioctls to be used by non-root users to add and remove encryption keys from the filesystem-level crypto keyrings, subject to limitations. Motivation: while privileged fscrypt key management is sufficient for some users (e.g. Android and Chromium OS, where a privileged process manages all keys), the old API by design also allows non-root users to set up and use encrypted directories, and we don't want to regress on that. Especially, we don't want to force users to continue using the old API, running into the visibility mismatch between files and keyrings and being unable to "lock" encrypted directories. Intuitively, the ioctls have to be privileged since they manipulate filesystem-level state. However, it's actually safe to make them unprivileged if we very carefully enforce some specific limitations. First, each key must be identified by a cryptographic hash so that a user can't add the wrong key for another user's files. For v2 encryption policies, we use the key_identifier for this. v1 policies don't have this, so managing keys for them remains privileged. Second, each key a user adds is charged to their quota for the keyrings service. Thus, a user can't exhaust memory by adding a huge number of keys. By default each non-root user is allowed up to 200 keys; this can be changed using the existing sysctl 'kernel.keys.maxkeys'. Third, if multiple users add the same key, we keep track of those users of the key (of which there remains a single copy), and won't really remove the key, i.e. "lock" the encrypted files, until all those users have removed it. This prevents denial of service attacks that would be possible under simpler schemes, such allowing the first user who added a key to remove it -- since that could be a malicious user who has compromised the key. Of course, encryption keys should be kept secret, but the idea is that using encryption should never be *less* secure than not using encryption, even if your key was compromised. We tolerate that a user will be unable to really remove a key, i.e. unable to "lock" their encrypted files, if another user has added the same key. But in a sense, this is actually a good thing because it will avoid providing a false notion of security where a key appears to have been removed when actually it's still in memory, available to any attacker who compromises the operating system kernel. Reviewed-by: Theodore Ts'o <tytso@mit.edu> Signed-off-by: Eric Biggers <ebiggers@google.com>
2019-08-05 10:35:47 +08:00
return err;
}
/* Re-add the secret if needed. */
fscrypt: stop using keyrings subsystem for fscrypt_master_key The approach of fs/crypto/ internally managing the fscrypt_master_key structs as the payloads of "struct key" objects contained in a "struct key" keyring has outlived its usefulness. The original idea was to simplify the code by reusing code from the keyrings subsystem. However, several issues have arisen that can't easily be resolved: - When a master key struct is destroyed, blk_crypto_evict_key() must be called on any per-mode keys embedded in it. (This started being the case when inline encryption support was added.) Yet, the keyrings subsystem can arbitrarily delay the destruction of keys, even past the time the filesystem was unmounted. Therefore, currently there is no easy way to call blk_crypto_evict_key() when a master key is destroyed. Currently, this is worked around by holding an extra reference to the filesystem's request_queue(s). But it was overlooked that the request_queue reference is *not* guaranteed to pin the corresponding blk_crypto_profile too; for device-mapper devices that support inline crypto, it doesn't. This can cause a use-after-free. - When the last inode that was using an incompletely-removed master key is evicted, the master key removal is completed by removing the key struct from the keyring. Currently this is done via key_invalidate(). Yet, key_invalidate() takes the key semaphore. This can deadlock when called from the shrinker, since in fscrypt_ioctl_add_key(), memory is allocated with GFP_KERNEL under the same semaphore. - More generally, the fact that the keyrings subsystem can arbitrarily delay the destruction of keys (via garbage collection delay, or via random processes getting temporary key references) is undesirable, as it means we can't strictly guarantee that all secrets are ever wiped. - Doing the master key lookups via the keyrings subsystem results in the key_permission LSM hook being called. fscrypt doesn't want this, as all access control for encrypted files is designed to happen via the files themselves, like any other files. The workaround which SELinux users are using is to change their SELinux policy to grant key search access to all domains. This works, but it is an odd extra step that shouldn't really have to be done. The fix for all these issues is to change the implementation to what I should have done originally: don't use the keyrings subsystem to keep track of the filesystem's fscrypt_master_key structs. Instead, just store them in a regular kernel data structure, and rework the reference counting, locking, and lifetime accordingly. Retain support for RCU-mode key lookups by using a hash table. Replace fscrypt_sb_free() with fscrypt_sb_delete(), which releases the keys synchronously and runs a bit earlier during unmount, so that block devices are still available. A side effect of this patch is that neither the master keys themselves nor the filesystem keyrings will be listed in /proc/keys anymore. ("Master key users" and the master key users keyrings will still be listed.) However, this was mostly an implementation detail, and it was intended just for debugging purposes. I don't know of anyone using it. This patch does *not* change how "master key users" (->mk_users) works; that still uses the keyrings subsystem. That is still needed for key quotas, and changing that isn't necessary to solve the issues listed above. If we decide to change that too, it would be a separate patch. I've marked this as fixing the original commit that added the fscrypt keyring, but as noted above the most important issue that this patch fixes wasn't introduced until the addition of inline encryption support. Fixes: 22d94f493bfb ("fscrypt: add FS_IOC_ADD_ENCRYPTION_KEY ioctl") Signed-off-by: Eric Biggers <ebiggers@google.com> Link: https://lore.kernel.org/r/20220901193208.138056-2-ebiggers@kernel.org
2022-09-02 03:32:06 +08:00
if (!is_master_key_secret_present(&mk->mk_secret)) {
if (!refcount_inc_not_zero(&mk->mk_active_refs))
return KEY_DEAD;
fscrypt: allow unprivileged users to add/remove keys for v2 policies Allow the FS_IOC_ADD_ENCRYPTION_KEY and FS_IOC_REMOVE_ENCRYPTION_KEY ioctls to be used by non-root users to add and remove encryption keys from the filesystem-level crypto keyrings, subject to limitations. Motivation: while privileged fscrypt key management is sufficient for some users (e.g. Android and Chromium OS, where a privileged process manages all keys), the old API by design also allows non-root users to set up and use encrypted directories, and we don't want to regress on that. Especially, we don't want to force users to continue using the old API, running into the visibility mismatch between files and keyrings and being unable to "lock" encrypted directories. Intuitively, the ioctls have to be privileged since they manipulate filesystem-level state. However, it's actually safe to make them unprivileged if we very carefully enforce some specific limitations. First, each key must be identified by a cryptographic hash so that a user can't add the wrong key for another user's files. For v2 encryption policies, we use the key_identifier for this. v1 policies don't have this, so managing keys for them remains privileged. Second, each key a user adds is charged to their quota for the keyrings service. Thus, a user can't exhaust memory by adding a huge number of keys. By default each non-root user is allowed up to 200 keys; this can be changed using the existing sysctl 'kernel.keys.maxkeys'. Third, if multiple users add the same key, we keep track of those users of the key (of which there remains a single copy), and won't really remove the key, i.e. "lock" the encrypted files, until all those users have removed it. This prevents denial of service attacks that would be possible under simpler schemes, such allowing the first user who added a key to remove it -- since that could be a malicious user who has compromised the key. Of course, encryption keys should be kept secret, but the idea is that using encryption should never be *less* secure than not using encryption, even if your key was compromised. We tolerate that a user will be unable to really remove a key, i.e. unable to "lock" their encrypted files, if another user has added the same key. But in a sense, this is actually a good thing because it will avoid providing a false notion of security where a key appears to have been removed when actually it's still in memory, available to any attacker who compromises the operating system kernel. Reviewed-by: Theodore Ts'o <tytso@mit.edu> Signed-off-by: Eric Biggers <ebiggers@google.com>
2019-08-05 10:35:47 +08:00
move_master_key_secret(&mk->mk_secret, secret);
fscrypt: stop using keyrings subsystem for fscrypt_master_key The approach of fs/crypto/ internally managing the fscrypt_master_key structs as the payloads of "struct key" objects contained in a "struct key" keyring has outlived its usefulness. The original idea was to simplify the code by reusing code from the keyrings subsystem. However, several issues have arisen that can't easily be resolved: - When a master key struct is destroyed, blk_crypto_evict_key() must be called on any per-mode keys embedded in it. (This started being the case when inline encryption support was added.) Yet, the keyrings subsystem can arbitrarily delay the destruction of keys, even past the time the filesystem was unmounted. Therefore, currently there is no easy way to call blk_crypto_evict_key() when a master key is destroyed. Currently, this is worked around by holding an extra reference to the filesystem's request_queue(s). But it was overlooked that the request_queue reference is *not* guaranteed to pin the corresponding blk_crypto_profile too; for device-mapper devices that support inline crypto, it doesn't. This can cause a use-after-free. - When the last inode that was using an incompletely-removed master key is evicted, the master key removal is completed by removing the key struct from the keyring. Currently this is done via key_invalidate(). Yet, key_invalidate() takes the key semaphore. This can deadlock when called from the shrinker, since in fscrypt_ioctl_add_key(), memory is allocated with GFP_KERNEL under the same semaphore. - More generally, the fact that the keyrings subsystem can arbitrarily delay the destruction of keys (via garbage collection delay, or via random processes getting temporary key references) is undesirable, as it means we can't strictly guarantee that all secrets are ever wiped. - Doing the master key lookups via the keyrings subsystem results in the key_permission LSM hook being called. fscrypt doesn't want this, as all access control for encrypted files is designed to happen via the files themselves, like any other files. The workaround which SELinux users are using is to change their SELinux policy to grant key search access to all domains. This works, but it is an odd extra step that shouldn't really have to be done. The fix for all these issues is to change the implementation to what I should have done originally: don't use the keyrings subsystem to keep track of the filesystem's fscrypt_master_key structs. Instead, just store them in a regular kernel data structure, and rework the reference counting, locking, and lifetime accordingly. Retain support for RCU-mode key lookups by using a hash table. Replace fscrypt_sb_free() with fscrypt_sb_delete(), which releases the keys synchronously and runs a bit earlier during unmount, so that block devices are still available. A side effect of this patch is that neither the master keys themselves nor the filesystem keyrings will be listed in /proc/keys anymore. ("Master key users" and the master key users keyrings will still be listed.) However, this was mostly an implementation detail, and it was intended just for debugging purposes. I don't know of anyone using it. This patch does *not* change how "master key users" (->mk_users) works; that still uses the keyrings subsystem. That is still needed for key quotas, and changing that isn't necessary to solve the issues listed above. If we decide to change that too, it would be a separate patch. I've marked this as fixing the original commit that added the fscrypt keyring, but as noted above the most important issue that this patch fixes wasn't introduced until the addition of inline encryption support. Fixes: 22d94f493bfb ("fscrypt: add FS_IOC_ADD_ENCRYPTION_KEY ioctl") Signed-off-by: Eric Biggers <ebiggers@google.com> Link: https://lore.kernel.org/r/20220901193208.138056-2-ebiggers@kernel.org
2022-09-02 03:32:06 +08:00
}
fscrypt: add FS_IOC_REMOVE_ENCRYPTION_KEY ioctl Add a new fscrypt ioctl, FS_IOC_REMOVE_ENCRYPTION_KEY. This ioctl removes an encryption key that was added by FS_IOC_ADD_ENCRYPTION_KEY. It wipes the secret key itself, then "locks" the encrypted files and directories that had been unlocked using that key -- implemented by evicting the relevant dentries and inodes from the VFS caches. The problem this solves is that many fscrypt users want the ability to remove encryption keys, causing the corresponding encrypted directories to appear "locked" (presented in ciphertext form) again. Moreover, users want removing an encryption key to *really* remove it, in the sense that the removed keys cannot be recovered even if kernel memory is compromised, e.g. by the exploit of a kernel security vulnerability or by a physical attack. This is desirable after a user logs out of the system, for example. In many cases users even already assume this to be the case and are surprised to hear when it's not. It is not sufficient to simply unlink the master key from the keyring (or to revoke or invalidate it), since the actual encryption transform objects are still pinned in memory by their inodes. Therefore, to really remove a key we must also evict the relevant inodes. Currently one workaround is to run 'sync && echo 2 > /proc/sys/vm/drop_caches'. But, that evicts all unused inodes in the system rather than just the inodes associated with the key being removed, causing severe performance problems. Moreover, it requires root privileges, so regular users can't "lock" their encrypted files. Another workaround, used in Chromium OS kernels, is to add a new VFS-level ioctl FS_IOC_DROP_CACHE which is a more restricted version of drop_caches that operates on a single super_block. It does: shrink_dcache_sb(sb); invalidate_inodes(sb, false); But it's still a hack. Yet, the major users of filesystem encryption want this feature badly enough that they are actually using these hacks. To properly solve the problem, start maintaining a list of the inodes which have been "unlocked" using each master key. Originally this wasn't possible because the kernel didn't keep track of in-use master keys at all. But, with the ->s_master_keys keyring it is now possible. Then, add an ioctl FS_IOC_REMOVE_ENCRYPTION_KEY. It finds the specified master key in ->s_master_keys, then wipes the secret key itself, which prevents any additional inodes from being unlocked with the key. Then, it syncs the filesystem and evicts the inodes in the key's list. The normal inode eviction code will free and wipe the per-file keys (in ->i_crypt_info). Note that freeing ->i_crypt_info without evicting the inodes was also considered, but would have been racy. Some inodes may still be in use when a master key is removed, and we can't simply revoke random file descriptors, mmap's, etc. Thus, the ioctl simply skips in-use inodes, and returns -EBUSY to indicate that some inodes weren't evicted. The master key *secret* is still removed, but the fscrypt_master_key struct remains to keep track of the remaining inodes. Userspace can then retry the ioctl to evict the remaining inodes. Alternatively, if userspace adds the key again, the refreshed secret will be associated with the existing list of inodes so they remain correctly tracked for future key removals. The ioctl doesn't wipe pagecache pages. Thus, we tolerate that after a kernel compromise some portions of plaintext file contents may still be recoverable from memory. This can be solved by enabling page poisoning system-wide, which security conscious users may choose to do. But it's very difficult to solve otherwise, e.g. note that plaintext file contents may have been read in other places than pagecache pages. Like FS_IOC_ADD_ENCRYPTION_KEY, FS_IOC_REMOVE_ENCRYPTION_KEY is initially restricted to privileged users only. This is sufficient for some use cases, but not all. A later patch will relax this restriction, but it will require introducing key hashes, among other changes. Reviewed-by: Theodore Ts'o <tytso@mit.edu> Signed-off-by: Eric Biggers <ebiggers@google.com>
2019-08-05 10:35:46 +08:00
return 0;
}
fscrypt: add fscrypt_add_test_dummy_key() Currently, the test_dummy_encryption mount option (which is used for encryption I/O testing with xfstests) uses v1 encryption policies, and it relies on userspace inserting a test key into the session keyring. We need test_dummy_encryption to support v2 encryption policies too. Requiring userspace to add the test key doesn't work well with v2 policies, since v2 policies only support the filesystem keyring (not the session keyring), and keys in the filesystem keyring are lost when the filesystem is unmounted. Hooking all test code that unmounts and re-mounts the filesystem would be difficult. Instead, let's make the filesystem automatically add the test key to its keyring when test_dummy_encryption is enabled. That puts the responsibility for choosing the test key on the kernel. We could just hard-code a key. But out of paranoia, let's first try using a per-boot random key, to prevent this code from being misused. A per-boot key will work as long as no one expects dummy-encrypted files to remain accessible after a reboot. (gce-xfstests doesn't.) Therefore, this patch adds a function fscrypt_add_test_dummy_key() which implements the above. The next patch will use it. Link: https://lore.kernel.org/r/20200512233251.118314-3-ebiggers@kernel.org Reviewed-by: Theodore Ts'o <tytso@mit.edu> Reviewed-by: Jaegeuk Kim <jaegeuk@kernel.org> Signed-off-by: Eric Biggers <ebiggers@google.com>
2020-05-13 07:32:49 +08:00
static int do_add_master_key(struct super_block *sb,
struct fscrypt_master_key_secret *secret,
const struct fscrypt_key_specifier *mk_spec)
fscrypt: add FS_IOC_ADD_ENCRYPTION_KEY ioctl Add a new fscrypt ioctl, FS_IOC_ADD_ENCRYPTION_KEY. This ioctl adds an encryption key to the filesystem's fscrypt keyring ->s_master_keys, making any files encrypted with that key appear "unlocked". Why we need this ~~~~~~~~~~~~~~~~ The main problem is that the "locked/unlocked" (ciphertext/plaintext) status of encrypted files is global, but the fscrypt keys are not. fscrypt only looks for keys in the keyring(s) the process accessing the filesystem is subscribed to: the thread keyring, process keyring, and session keyring, where the session keyring may contain the user keyring. Therefore, userspace has to put fscrypt keys in the keyrings for individual users or sessions. But this means that when a process with a different keyring tries to access encrypted files, whether they appear "unlocked" or not is nondeterministic. This is because it depends on whether the files are currently present in the inode cache. Fixing this by consistently providing each process its own view of the filesystem depending on whether it has the key or not isn't feasible due to how the VFS caches work. Furthermore, while sometimes users expect this behavior, it is misguided for two reasons. First, it would be an OS-level access control mechanism largely redundant with existing access control mechanisms such as UNIX file permissions, ACLs, LSMs, etc. Encryption is actually for protecting the data at rest. Second, almost all users of fscrypt actually do need the keys to be global. The largest users of fscrypt, Android and Chromium OS, achieve this by having PID 1 create a "session keyring" that is inherited by every process. This works, but it isn't scalable because it prevents session keyrings from being used for any other purpose. On general-purpose Linux distros, the 'fscrypt' userspace tool [1] can't similarly abuse the session keyring, so to make 'sudo' work on all systems it has to link all the user keyrings into root's user keyring [2]. This is ugly and raises security concerns. Moreover it can't make the keys available to system services, such as sshd trying to access the user's '~/.ssh' directory (see [3], [4]) or NetworkManager trying to read certificates from the user's home directory (see [5]); or to Docker containers (see [6], [7]). By having an API to add a key to the *filesystem* we'll be able to fix the above bugs, remove userspace workarounds, and clearly express the intended semantics: the locked/unlocked status of an encrypted directory is global, and encryption is orthogonal to OS-level access control. Why not use the add_key() syscall ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ We use an ioctl for this API rather than the existing add_key() system call because the ioctl gives us the flexibility needed to implement fscrypt-specific semantics that will be introduced in later patches: - Supporting key removal with the semantics such that the secret is removed immediately and any unused inodes using the key are evicted; also, the eviction of any in-use inodes can be retried. - Calculating a key-dependent cryptographic identifier and returning it to userspace. - Allowing keys to be added and removed by non-root users, but only keys for v2 encryption policies; and to prevent denial-of-service attacks, users can only remove keys they themselves have added, and a key is only really removed after all users who added it have removed it. Trying to shoehorn these semantics into the keyrings syscalls would be very difficult, whereas the ioctls make things much easier. However, to reuse code the implementation still uses the keyrings service internally. Thus we get lockless RCU-mode key lookups without having to re-implement it, and the keys automatically show up in /proc/keys for debugging purposes. References: [1] https://github.com/google/fscrypt [2] https://goo.gl/55cCrI#heading=h.vf09isp98isb [3] https://github.com/google/fscrypt/issues/111#issuecomment-444347939 [4] https://github.com/google/fscrypt/issues/116 [5] https://bugs.launchpad.net/ubuntu/+source/fscrypt/+bug/1770715 [6] https://github.com/google/fscrypt/issues/128 [7] https://askubuntu.com/questions/1130306/cannot-run-docker-on-an-encrypted-filesystem Reviewed-by: Theodore Ts'o <tytso@mit.edu> Signed-off-by: Eric Biggers <ebiggers@google.com>
2019-08-05 10:35:46 +08:00
{
static DEFINE_MUTEX(fscrypt_add_key_mutex);
fscrypt: stop using keyrings subsystem for fscrypt_master_key The approach of fs/crypto/ internally managing the fscrypt_master_key structs as the payloads of "struct key" objects contained in a "struct key" keyring has outlived its usefulness. The original idea was to simplify the code by reusing code from the keyrings subsystem. However, several issues have arisen that can't easily be resolved: - When a master key struct is destroyed, blk_crypto_evict_key() must be called on any per-mode keys embedded in it. (This started being the case when inline encryption support was added.) Yet, the keyrings subsystem can arbitrarily delay the destruction of keys, even past the time the filesystem was unmounted. Therefore, currently there is no easy way to call blk_crypto_evict_key() when a master key is destroyed. Currently, this is worked around by holding an extra reference to the filesystem's request_queue(s). But it was overlooked that the request_queue reference is *not* guaranteed to pin the corresponding blk_crypto_profile too; for device-mapper devices that support inline crypto, it doesn't. This can cause a use-after-free. - When the last inode that was using an incompletely-removed master key is evicted, the master key removal is completed by removing the key struct from the keyring. Currently this is done via key_invalidate(). Yet, key_invalidate() takes the key semaphore. This can deadlock when called from the shrinker, since in fscrypt_ioctl_add_key(), memory is allocated with GFP_KERNEL under the same semaphore. - More generally, the fact that the keyrings subsystem can arbitrarily delay the destruction of keys (via garbage collection delay, or via random processes getting temporary key references) is undesirable, as it means we can't strictly guarantee that all secrets are ever wiped. - Doing the master key lookups via the keyrings subsystem results in the key_permission LSM hook being called. fscrypt doesn't want this, as all access control for encrypted files is designed to happen via the files themselves, like any other files. The workaround which SELinux users are using is to change their SELinux policy to grant key search access to all domains. This works, but it is an odd extra step that shouldn't really have to be done. The fix for all these issues is to change the implementation to what I should have done originally: don't use the keyrings subsystem to keep track of the filesystem's fscrypt_master_key structs. Instead, just store them in a regular kernel data structure, and rework the reference counting, locking, and lifetime accordingly. Retain support for RCU-mode key lookups by using a hash table. Replace fscrypt_sb_free() with fscrypt_sb_delete(), which releases the keys synchronously and runs a bit earlier during unmount, so that block devices are still available. A side effect of this patch is that neither the master keys themselves nor the filesystem keyrings will be listed in /proc/keys anymore. ("Master key users" and the master key users keyrings will still be listed.) However, this was mostly an implementation detail, and it was intended just for debugging purposes. I don't know of anyone using it. This patch does *not* change how "master key users" (->mk_users) works; that still uses the keyrings subsystem. That is still needed for key quotas, and changing that isn't necessary to solve the issues listed above. If we decide to change that too, it would be a separate patch. I've marked this as fixing the original commit that added the fscrypt keyring, but as noted above the most important issue that this patch fixes wasn't introduced until the addition of inline encryption support. Fixes: 22d94f493bfb ("fscrypt: add FS_IOC_ADD_ENCRYPTION_KEY ioctl") Signed-off-by: Eric Biggers <ebiggers@google.com> Link: https://lore.kernel.org/r/20220901193208.138056-2-ebiggers@kernel.org
2022-09-02 03:32:06 +08:00
struct fscrypt_master_key *mk;
fscrypt: add FS_IOC_ADD_ENCRYPTION_KEY ioctl Add a new fscrypt ioctl, FS_IOC_ADD_ENCRYPTION_KEY. This ioctl adds an encryption key to the filesystem's fscrypt keyring ->s_master_keys, making any files encrypted with that key appear "unlocked". Why we need this ~~~~~~~~~~~~~~~~ The main problem is that the "locked/unlocked" (ciphertext/plaintext) status of encrypted files is global, but the fscrypt keys are not. fscrypt only looks for keys in the keyring(s) the process accessing the filesystem is subscribed to: the thread keyring, process keyring, and session keyring, where the session keyring may contain the user keyring. Therefore, userspace has to put fscrypt keys in the keyrings for individual users or sessions. But this means that when a process with a different keyring tries to access encrypted files, whether they appear "unlocked" or not is nondeterministic. This is because it depends on whether the files are currently present in the inode cache. Fixing this by consistently providing each process its own view of the filesystem depending on whether it has the key or not isn't feasible due to how the VFS caches work. Furthermore, while sometimes users expect this behavior, it is misguided for two reasons. First, it would be an OS-level access control mechanism largely redundant with existing access control mechanisms such as UNIX file permissions, ACLs, LSMs, etc. Encryption is actually for protecting the data at rest. Second, almost all users of fscrypt actually do need the keys to be global. The largest users of fscrypt, Android and Chromium OS, achieve this by having PID 1 create a "session keyring" that is inherited by every process. This works, but it isn't scalable because it prevents session keyrings from being used for any other purpose. On general-purpose Linux distros, the 'fscrypt' userspace tool [1] can't similarly abuse the session keyring, so to make 'sudo' work on all systems it has to link all the user keyrings into root's user keyring [2]. This is ugly and raises security concerns. Moreover it can't make the keys available to system services, such as sshd trying to access the user's '~/.ssh' directory (see [3], [4]) or NetworkManager trying to read certificates from the user's home directory (see [5]); or to Docker containers (see [6], [7]). By having an API to add a key to the *filesystem* we'll be able to fix the above bugs, remove userspace workarounds, and clearly express the intended semantics: the locked/unlocked status of an encrypted directory is global, and encryption is orthogonal to OS-level access control. Why not use the add_key() syscall ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ We use an ioctl for this API rather than the existing add_key() system call because the ioctl gives us the flexibility needed to implement fscrypt-specific semantics that will be introduced in later patches: - Supporting key removal with the semantics such that the secret is removed immediately and any unused inodes using the key are evicted; also, the eviction of any in-use inodes can be retried. - Calculating a key-dependent cryptographic identifier and returning it to userspace. - Allowing keys to be added and removed by non-root users, but only keys for v2 encryption policies; and to prevent denial-of-service attacks, users can only remove keys they themselves have added, and a key is only really removed after all users who added it have removed it. Trying to shoehorn these semantics into the keyrings syscalls would be very difficult, whereas the ioctls make things much easier. However, to reuse code the implementation still uses the keyrings service internally. Thus we get lockless RCU-mode key lookups without having to re-implement it, and the keys automatically show up in /proc/keys for debugging purposes. References: [1] https://github.com/google/fscrypt [2] https://goo.gl/55cCrI#heading=h.vf09isp98isb [3] https://github.com/google/fscrypt/issues/111#issuecomment-444347939 [4] https://github.com/google/fscrypt/issues/116 [5] https://bugs.launchpad.net/ubuntu/+source/fscrypt/+bug/1770715 [6] https://github.com/google/fscrypt/issues/128 [7] https://askubuntu.com/questions/1130306/cannot-run-docker-on-an-encrypted-filesystem Reviewed-by: Theodore Ts'o <tytso@mit.edu> Signed-off-by: Eric Biggers <ebiggers@google.com>
2019-08-05 10:35:46 +08:00
int err;
mutex_lock(&fscrypt_add_key_mutex); /* serialize find + link */
fscrypt: stop using keyrings subsystem for fscrypt_master_key The approach of fs/crypto/ internally managing the fscrypt_master_key structs as the payloads of "struct key" objects contained in a "struct key" keyring has outlived its usefulness. The original idea was to simplify the code by reusing code from the keyrings subsystem. However, several issues have arisen that can't easily be resolved: - When a master key struct is destroyed, blk_crypto_evict_key() must be called on any per-mode keys embedded in it. (This started being the case when inline encryption support was added.) Yet, the keyrings subsystem can arbitrarily delay the destruction of keys, even past the time the filesystem was unmounted. Therefore, currently there is no easy way to call blk_crypto_evict_key() when a master key is destroyed. Currently, this is worked around by holding an extra reference to the filesystem's request_queue(s). But it was overlooked that the request_queue reference is *not* guaranteed to pin the corresponding blk_crypto_profile too; for device-mapper devices that support inline crypto, it doesn't. This can cause a use-after-free. - When the last inode that was using an incompletely-removed master key is evicted, the master key removal is completed by removing the key struct from the keyring. Currently this is done via key_invalidate(). Yet, key_invalidate() takes the key semaphore. This can deadlock when called from the shrinker, since in fscrypt_ioctl_add_key(), memory is allocated with GFP_KERNEL under the same semaphore. - More generally, the fact that the keyrings subsystem can arbitrarily delay the destruction of keys (via garbage collection delay, or via random processes getting temporary key references) is undesirable, as it means we can't strictly guarantee that all secrets are ever wiped. - Doing the master key lookups via the keyrings subsystem results in the key_permission LSM hook being called. fscrypt doesn't want this, as all access control for encrypted files is designed to happen via the files themselves, like any other files. The workaround which SELinux users are using is to change their SELinux policy to grant key search access to all domains. This works, but it is an odd extra step that shouldn't really have to be done. The fix for all these issues is to change the implementation to what I should have done originally: don't use the keyrings subsystem to keep track of the filesystem's fscrypt_master_key structs. Instead, just store them in a regular kernel data structure, and rework the reference counting, locking, and lifetime accordingly. Retain support for RCU-mode key lookups by using a hash table. Replace fscrypt_sb_free() with fscrypt_sb_delete(), which releases the keys synchronously and runs a bit earlier during unmount, so that block devices are still available. A side effect of this patch is that neither the master keys themselves nor the filesystem keyrings will be listed in /proc/keys anymore. ("Master key users" and the master key users keyrings will still be listed.) However, this was mostly an implementation detail, and it was intended just for debugging purposes. I don't know of anyone using it. This patch does *not* change how "master key users" (->mk_users) works; that still uses the keyrings subsystem. That is still needed for key quotas, and changing that isn't necessary to solve the issues listed above. If we decide to change that too, it would be a separate patch. I've marked this as fixing the original commit that added the fscrypt keyring, but as noted above the most important issue that this patch fixes wasn't introduced until the addition of inline encryption support. Fixes: 22d94f493bfb ("fscrypt: add FS_IOC_ADD_ENCRYPTION_KEY ioctl") Signed-off-by: Eric Biggers <ebiggers@google.com> Link: https://lore.kernel.org/r/20220901193208.138056-2-ebiggers@kernel.org
2022-09-02 03:32:06 +08:00
mk = fscrypt_find_master_key(sb, mk_spec);
if (!mk) {
fscrypt: add FS_IOC_ADD_ENCRYPTION_KEY ioctl Add a new fscrypt ioctl, FS_IOC_ADD_ENCRYPTION_KEY. This ioctl adds an encryption key to the filesystem's fscrypt keyring ->s_master_keys, making any files encrypted with that key appear "unlocked". Why we need this ~~~~~~~~~~~~~~~~ The main problem is that the "locked/unlocked" (ciphertext/plaintext) status of encrypted files is global, but the fscrypt keys are not. fscrypt only looks for keys in the keyring(s) the process accessing the filesystem is subscribed to: the thread keyring, process keyring, and session keyring, where the session keyring may contain the user keyring. Therefore, userspace has to put fscrypt keys in the keyrings for individual users or sessions. But this means that when a process with a different keyring tries to access encrypted files, whether they appear "unlocked" or not is nondeterministic. This is because it depends on whether the files are currently present in the inode cache. Fixing this by consistently providing each process its own view of the filesystem depending on whether it has the key or not isn't feasible due to how the VFS caches work. Furthermore, while sometimes users expect this behavior, it is misguided for two reasons. First, it would be an OS-level access control mechanism largely redundant with existing access control mechanisms such as UNIX file permissions, ACLs, LSMs, etc. Encryption is actually for protecting the data at rest. Second, almost all users of fscrypt actually do need the keys to be global. The largest users of fscrypt, Android and Chromium OS, achieve this by having PID 1 create a "session keyring" that is inherited by every process. This works, but it isn't scalable because it prevents session keyrings from being used for any other purpose. On general-purpose Linux distros, the 'fscrypt' userspace tool [1] can't similarly abuse the session keyring, so to make 'sudo' work on all systems it has to link all the user keyrings into root's user keyring [2]. This is ugly and raises security concerns. Moreover it can't make the keys available to system services, such as sshd trying to access the user's '~/.ssh' directory (see [3], [4]) or NetworkManager trying to read certificates from the user's home directory (see [5]); or to Docker containers (see [6], [7]). By having an API to add a key to the *filesystem* we'll be able to fix the above bugs, remove userspace workarounds, and clearly express the intended semantics: the locked/unlocked status of an encrypted directory is global, and encryption is orthogonal to OS-level access control. Why not use the add_key() syscall ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ We use an ioctl for this API rather than the existing add_key() system call because the ioctl gives us the flexibility needed to implement fscrypt-specific semantics that will be introduced in later patches: - Supporting key removal with the semantics such that the secret is removed immediately and any unused inodes using the key are evicted; also, the eviction of any in-use inodes can be retried. - Calculating a key-dependent cryptographic identifier and returning it to userspace. - Allowing keys to be added and removed by non-root users, but only keys for v2 encryption policies; and to prevent denial-of-service attacks, users can only remove keys they themselves have added, and a key is only really removed after all users who added it have removed it. Trying to shoehorn these semantics into the keyrings syscalls would be very difficult, whereas the ioctls make things much easier. However, to reuse code the implementation still uses the keyrings service internally. Thus we get lockless RCU-mode key lookups without having to re-implement it, and the keys automatically show up in /proc/keys for debugging purposes. References: [1] https://github.com/google/fscrypt [2] https://goo.gl/55cCrI#heading=h.vf09isp98isb [3] https://github.com/google/fscrypt/issues/111#issuecomment-444347939 [4] https://github.com/google/fscrypt/issues/116 [5] https://bugs.launchpad.net/ubuntu/+source/fscrypt/+bug/1770715 [6] https://github.com/google/fscrypt/issues/128 [7] https://askubuntu.com/questions/1130306/cannot-run-docker-on-an-encrypted-filesystem Reviewed-by: Theodore Ts'o <tytso@mit.edu> Signed-off-by: Eric Biggers <ebiggers@google.com>
2019-08-05 10:35:46 +08:00
/* Didn't find the key in ->s_master_keys. Add it. */
err = allocate_filesystem_keyring(sb);
fscrypt: stop using keyrings subsystem for fscrypt_master_key The approach of fs/crypto/ internally managing the fscrypt_master_key structs as the payloads of "struct key" objects contained in a "struct key" keyring has outlived its usefulness. The original idea was to simplify the code by reusing code from the keyrings subsystem. However, several issues have arisen that can't easily be resolved: - When a master key struct is destroyed, blk_crypto_evict_key() must be called on any per-mode keys embedded in it. (This started being the case when inline encryption support was added.) Yet, the keyrings subsystem can arbitrarily delay the destruction of keys, even past the time the filesystem was unmounted. Therefore, currently there is no easy way to call blk_crypto_evict_key() when a master key is destroyed. Currently, this is worked around by holding an extra reference to the filesystem's request_queue(s). But it was overlooked that the request_queue reference is *not* guaranteed to pin the corresponding blk_crypto_profile too; for device-mapper devices that support inline crypto, it doesn't. This can cause a use-after-free. - When the last inode that was using an incompletely-removed master key is evicted, the master key removal is completed by removing the key struct from the keyring. Currently this is done via key_invalidate(). Yet, key_invalidate() takes the key semaphore. This can deadlock when called from the shrinker, since in fscrypt_ioctl_add_key(), memory is allocated with GFP_KERNEL under the same semaphore. - More generally, the fact that the keyrings subsystem can arbitrarily delay the destruction of keys (via garbage collection delay, or via random processes getting temporary key references) is undesirable, as it means we can't strictly guarantee that all secrets are ever wiped. - Doing the master key lookups via the keyrings subsystem results in the key_permission LSM hook being called. fscrypt doesn't want this, as all access control for encrypted files is designed to happen via the files themselves, like any other files. The workaround which SELinux users are using is to change their SELinux policy to grant key search access to all domains. This works, but it is an odd extra step that shouldn't really have to be done. The fix for all these issues is to change the implementation to what I should have done originally: don't use the keyrings subsystem to keep track of the filesystem's fscrypt_master_key structs. Instead, just store them in a regular kernel data structure, and rework the reference counting, locking, and lifetime accordingly. Retain support for RCU-mode key lookups by using a hash table. Replace fscrypt_sb_free() with fscrypt_sb_delete(), which releases the keys synchronously and runs a bit earlier during unmount, so that block devices are still available. A side effect of this patch is that neither the master keys themselves nor the filesystem keyrings will be listed in /proc/keys anymore. ("Master key users" and the master key users keyrings will still be listed.) However, this was mostly an implementation detail, and it was intended just for debugging purposes. I don't know of anyone using it. This patch does *not* change how "master key users" (->mk_users) works; that still uses the keyrings subsystem. That is still needed for key quotas, and changing that isn't necessary to solve the issues listed above. If we decide to change that too, it would be a separate patch. I've marked this as fixing the original commit that added the fscrypt keyring, but as noted above the most important issue that this patch fixes wasn't introduced until the addition of inline encryption support. Fixes: 22d94f493bfb ("fscrypt: add FS_IOC_ADD_ENCRYPTION_KEY ioctl") Signed-off-by: Eric Biggers <ebiggers@google.com> Link: https://lore.kernel.org/r/20220901193208.138056-2-ebiggers@kernel.org
2022-09-02 03:32:06 +08:00
if (!err)
err = add_new_master_key(sb, secret, mk_spec);
fscrypt: add FS_IOC_ADD_ENCRYPTION_KEY ioctl Add a new fscrypt ioctl, FS_IOC_ADD_ENCRYPTION_KEY. This ioctl adds an encryption key to the filesystem's fscrypt keyring ->s_master_keys, making any files encrypted with that key appear "unlocked". Why we need this ~~~~~~~~~~~~~~~~ The main problem is that the "locked/unlocked" (ciphertext/plaintext) status of encrypted files is global, but the fscrypt keys are not. fscrypt only looks for keys in the keyring(s) the process accessing the filesystem is subscribed to: the thread keyring, process keyring, and session keyring, where the session keyring may contain the user keyring. Therefore, userspace has to put fscrypt keys in the keyrings for individual users or sessions. But this means that when a process with a different keyring tries to access encrypted files, whether they appear "unlocked" or not is nondeterministic. This is because it depends on whether the files are currently present in the inode cache. Fixing this by consistently providing each process its own view of the filesystem depending on whether it has the key or not isn't feasible due to how the VFS caches work. Furthermore, while sometimes users expect this behavior, it is misguided for two reasons. First, it would be an OS-level access control mechanism largely redundant with existing access control mechanisms such as UNIX file permissions, ACLs, LSMs, etc. Encryption is actually for protecting the data at rest. Second, almost all users of fscrypt actually do need the keys to be global. The largest users of fscrypt, Android and Chromium OS, achieve this by having PID 1 create a "session keyring" that is inherited by every process. This works, but it isn't scalable because it prevents session keyrings from being used for any other purpose. On general-purpose Linux distros, the 'fscrypt' userspace tool [1] can't similarly abuse the session keyring, so to make 'sudo' work on all systems it has to link all the user keyrings into root's user keyring [2]. This is ugly and raises security concerns. Moreover it can't make the keys available to system services, such as sshd trying to access the user's '~/.ssh' directory (see [3], [4]) or NetworkManager trying to read certificates from the user's home directory (see [5]); or to Docker containers (see [6], [7]). By having an API to add a key to the *filesystem* we'll be able to fix the above bugs, remove userspace workarounds, and clearly express the intended semantics: the locked/unlocked status of an encrypted directory is global, and encryption is orthogonal to OS-level access control. Why not use the add_key() syscall ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ We use an ioctl for this API rather than the existing add_key() system call because the ioctl gives us the flexibility needed to implement fscrypt-specific semantics that will be introduced in later patches: - Supporting key removal with the semantics such that the secret is removed immediately and any unused inodes using the key are evicted; also, the eviction of any in-use inodes can be retried. - Calculating a key-dependent cryptographic identifier and returning it to userspace. - Allowing keys to be added and removed by non-root users, but only keys for v2 encryption policies; and to prevent denial-of-service attacks, users can only remove keys they themselves have added, and a key is only really removed after all users who added it have removed it. Trying to shoehorn these semantics into the keyrings syscalls would be very difficult, whereas the ioctls make things much easier. However, to reuse code the implementation still uses the keyrings service internally. Thus we get lockless RCU-mode key lookups without having to re-implement it, and the keys automatically show up in /proc/keys for debugging purposes. References: [1] https://github.com/google/fscrypt [2] https://goo.gl/55cCrI#heading=h.vf09isp98isb [3] https://github.com/google/fscrypt/issues/111#issuecomment-444347939 [4] https://github.com/google/fscrypt/issues/116 [5] https://bugs.launchpad.net/ubuntu/+source/fscrypt/+bug/1770715 [6] https://github.com/google/fscrypt/issues/128 [7] https://askubuntu.com/questions/1130306/cannot-run-docker-on-an-encrypted-filesystem Reviewed-by: Theodore Ts'o <tytso@mit.edu> Signed-off-by: Eric Biggers <ebiggers@google.com>
2019-08-05 10:35:46 +08:00
} else {
fscrypt: add FS_IOC_REMOVE_ENCRYPTION_KEY ioctl Add a new fscrypt ioctl, FS_IOC_REMOVE_ENCRYPTION_KEY. This ioctl removes an encryption key that was added by FS_IOC_ADD_ENCRYPTION_KEY. It wipes the secret key itself, then "locks" the encrypted files and directories that had been unlocked using that key -- implemented by evicting the relevant dentries and inodes from the VFS caches. The problem this solves is that many fscrypt users want the ability to remove encryption keys, causing the corresponding encrypted directories to appear "locked" (presented in ciphertext form) again. Moreover, users want removing an encryption key to *really* remove it, in the sense that the removed keys cannot be recovered even if kernel memory is compromised, e.g. by the exploit of a kernel security vulnerability or by a physical attack. This is desirable after a user logs out of the system, for example. In many cases users even already assume this to be the case and are surprised to hear when it's not. It is not sufficient to simply unlink the master key from the keyring (or to revoke or invalidate it), since the actual encryption transform objects are still pinned in memory by their inodes. Therefore, to really remove a key we must also evict the relevant inodes. Currently one workaround is to run 'sync && echo 2 > /proc/sys/vm/drop_caches'. But, that evicts all unused inodes in the system rather than just the inodes associated with the key being removed, causing severe performance problems. Moreover, it requires root privileges, so regular users can't "lock" their encrypted files. Another workaround, used in Chromium OS kernels, is to add a new VFS-level ioctl FS_IOC_DROP_CACHE which is a more restricted version of drop_caches that operates on a single super_block. It does: shrink_dcache_sb(sb); invalidate_inodes(sb, false); But it's still a hack. Yet, the major users of filesystem encryption want this feature badly enough that they are actually using these hacks. To properly solve the problem, start maintaining a list of the inodes which have been "unlocked" using each master key. Originally this wasn't possible because the kernel didn't keep track of in-use master keys at all. But, with the ->s_master_keys keyring it is now possible. Then, add an ioctl FS_IOC_REMOVE_ENCRYPTION_KEY. It finds the specified master key in ->s_master_keys, then wipes the secret key itself, which prevents any additional inodes from being unlocked with the key. Then, it syncs the filesystem and evicts the inodes in the key's list. The normal inode eviction code will free and wipe the per-file keys (in ->i_crypt_info). Note that freeing ->i_crypt_info without evicting the inodes was also considered, but would have been racy. Some inodes may still be in use when a master key is removed, and we can't simply revoke random file descriptors, mmap's, etc. Thus, the ioctl simply skips in-use inodes, and returns -EBUSY to indicate that some inodes weren't evicted. The master key *secret* is still removed, but the fscrypt_master_key struct remains to keep track of the remaining inodes. Userspace can then retry the ioctl to evict the remaining inodes. Alternatively, if userspace adds the key again, the refreshed secret will be associated with the existing list of inodes so they remain correctly tracked for future key removals. The ioctl doesn't wipe pagecache pages. Thus, we tolerate that after a kernel compromise some portions of plaintext file contents may still be recoverable from memory. This can be solved by enabling page poisoning system-wide, which security conscious users may choose to do. But it's very difficult to solve otherwise, e.g. note that plaintext file contents may have been read in other places than pagecache pages. Like FS_IOC_ADD_ENCRYPTION_KEY, FS_IOC_REMOVE_ENCRYPTION_KEY is initially restricted to privileged users only. This is sufficient for some use cases, but not all. A later patch will relax this restriction, but it will require introducing key hashes, among other changes. Reviewed-by: Theodore Ts'o <tytso@mit.edu> Signed-off-by: Eric Biggers <ebiggers@google.com>
2019-08-05 10:35:46 +08:00
/*
* Found the key in ->s_master_keys. Re-add the secret if
fscrypt: allow unprivileged users to add/remove keys for v2 policies Allow the FS_IOC_ADD_ENCRYPTION_KEY and FS_IOC_REMOVE_ENCRYPTION_KEY ioctls to be used by non-root users to add and remove encryption keys from the filesystem-level crypto keyrings, subject to limitations. Motivation: while privileged fscrypt key management is sufficient for some users (e.g. Android and Chromium OS, where a privileged process manages all keys), the old API by design also allows non-root users to set up and use encrypted directories, and we don't want to regress on that. Especially, we don't want to force users to continue using the old API, running into the visibility mismatch between files and keyrings and being unable to "lock" encrypted directories. Intuitively, the ioctls have to be privileged since they manipulate filesystem-level state. However, it's actually safe to make them unprivileged if we very carefully enforce some specific limitations. First, each key must be identified by a cryptographic hash so that a user can't add the wrong key for another user's files. For v2 encryption policies, we use the key_identifier for this. v1 policies don't have this, so managing keys for them remains privileged. Second, each key a user adds is charged to their quota for the keyrings service. Thus, a user can't exhaust memory by adding a huge number of keys. By default each non-root user is allowed up to 200 keys; this can be changed using the existing sysctl 'kernel.keys.maxkeys'. Third, if multiple users add the same key, we keep track of those users of the key (of which there remains a single copy), and won't really remove the key, i.e. "lock" the encrypted files, until all those users have removed it. This prevents denial of service attacks that would be possible under simpler schemes, such allowing the first user who added a key to remove it -- since that could be a malicious user who has compromised the key. Of course, encryption keys should be kept secret, but the idea is that using encryption should never be *less* secure than not using encryption, even if your key was compromised. We tolerate that a user will be unable to really remove a key, i.e. unable to "lock" their encrypted files, if another user has added the same key. But in a sense, this is actually a good thing because it will avoid providing a false notion of security where a key appears to have been removed when actually it's still in memory, available to any attacker who compromises the operating system kernel. Reviewed-by: Theodore Ts'o <tytso@mit.edu> Signed-off-by: Eric Biggers <ebiggers@google.com>
2019-08-05 10:35:47 +08:00
* needed, and add the user to ->mk_users if needed.
fscrypt: add FS_IOC_REMOVE_ENCRYPTION_KEY ioctl Add a new fscrypt ioctl, FS_IOC_REMOVE_ENCRYPTION_KEY. This ioctl removes an encryption key that was added by FS_IOC_ADD_ENCRYPTION_KEY. It wipes the secret key itself, then "locks" the encrypted files and directories that had been unlocked using that key -- implemented by evicting the relevant dentries and inodes from the VFS caches. The problem this solves is that many fscrypt users want the ability to remove encryption keys, causing the corresponding encrypted directories to appear "locked" (presented in ciphertext form) again. Moreover, users want removing an encryption key to *really* remove it, in the sense that the removed keys cannot be recovered even if kernel memory is compromised, e.g. by the exploit of a kernel security vulnerability or by a physical attack. This is desirable after a user logs out of the system, for example. In many cases users even already assume this to be the case and are surprised to hear when it's not. It is not sufficient to simply unlink the master key from the keyring (or to revoke or invalidate it), since the actual encryption transform objects are still pinned in memory by their inodes. Therefore, to really remove a key we must also evict the relevant inodes. Currently one workaround is to run 'sync && echo 2 > /proc/sys/vm/drop_caches'. But, that evicts all unused inodes in the system rather than just the inodes associated with the key being removed, causing severe performance problems. Moreover, it requires root privileges, so regular users can't "lock" their encrypted files. Another workaround, used in Chromium OS kernels, is to add a new VFS-level ioctl FS_IOC_DROP_CACHE which is a more restricted version of drop_caches that operates on a single super_block. It does: shrink_dcache_sb(sb); invalidate_inodes(sb, false); But it's still a hack. Yet, the major users of filesystem encryption want this feature badly enough that they are actually using these hacks. To properly solve the problem, start maintaining a list of the inodes which have been "unlocked" using each master key. Originally this wasn't possible because the kernel didn't keep track of in-use master keys at all. But, with the ->s_master_keys keyring it is now possible. Then, add an ioctl FS_IOC_REMOVE_ENCRYPTION_KEY. It finds the specified master key in ->s_master_keys, then wipes the secret key itself, which prevents any additional inodes from being unlocked with the key. Then, it syncs the filesystem and evicts the inodes in the key's list. The normal inode eviction code will free and wipe the per-file keys (in ->i_crypt_info). Note that freeing ->i_crypt_info without evicting the inodes was also considered, but would have been racy. Some inodes may still be in use when a master key is removed, and we can't simply revoke random file descriptors, mmap's, etc. Thus, the ioctl simply skips in-use inodes, and returns -EBUSY to indicate that some inodes weren't evicted. The master key *secret* is still removed, but the fscrypt_master_key struct remains to keep track of the remaining inodes. Userspace can then retry the ioctl to evict the remaining inodes. Alternatively, if userspace adds the key again, the refreshed secret will be associated with the existing list of inodes so they remain correctly tracked for future key removals. The ioctl doesn't wipe pagecache pages. Thus, we tolerate that after a kernel compromise some portions of plaintext file contents may still be recoverable from memory. This can be solved by enabling page poisoning system-wide, which security conscious users may choose to do. But it's very difficult to solve otherwise, e.g. note that plaintext file contents may have been read in other places than pagecache pages. Like FS_IOC_ADD_ENCRYPTION_KEY, FS_IOC_REMOVE_ENCRYPTION_KEY is initially restricted to privileged users only. This is sufficient for some use cases, but not all. A later patch will relax this restriction, but it will require introducing key hashes, among other changes. Reviewed-by: Theodore Ts'o <tytso@mit.edu> Signed-off-by: Eric Biggers <ebiggers@google.com>
2019-08-05 10:35:46 +08:00
*/
fscrypt: stop using keyrings subsystem for fscrypt_master_key The approach of fs/crypto/ internally managing the fscrypt_master_key structs as the payloads of "struct key" objects contained in a "struct key" keyring has outlived its usefulness. The original idea was to simplify the code by reusing code from the keyrings subsystem. However, several issues have arisen that can't easily be resolved: - When a master key struct is destroyed, blk_crypto_evict_key() must be called on any per-mode keys embedded in it. (This started being the case when inline encryption support was added.) Yet, the keyrings subsystem can arbitrarily delay the destruction of keys, even past the time the filesystem was unmounted. Therefore, currently there is no easy way to call blk_crypto_evict_key() when a master key is destroyed. Currently, this is worked around by holding an extra reference to the filesystem's request_queue(s). But it was overlooked that the request_queue reference is *not* guaranteed to pin the corresponding blk_crypto_profile too; for device-mapper devices that support inline crypto, it doesn't. This can cause a use-after-free. - When the last inode that was using an incompletely-removed master key is evicted, the master key removal is completed by removing the key struct from the keyring. Currently this is done via key_invalidate(). Yet, key_invalidate() takes the key semaphore. This can deadlock when called from the shrinker, since in fscrypt_ioctl_add_key(), memory is allocated with GFP_KERNEL under the same semaphore. - More generally, the fact that the keyrings subsystem can arbitrarily delay the destruction of keys (via garbage collection delay, or via random processes getting temporary key references) is undesirable, as it means we can't strictly guarantee that all secrets are ever wiped. - Doing the master key lookups via the keyrings subsystem results in the key_permission LSM hook being called. fscrypt doesn't want this, as all access control for encrypted files is designed to happen via the files themselves, like any other files. The workaround which SELinux users are using is to change their SELinux policy to grant key search access to all domains. This works, but it is an odd extra step that shouldn't really have to be done. The fix for all these issues is to change the implementation to what I should have done originally: don't use the keyrings subsystem to keep track of the filesystem's fscrypt_master_key structs. Instead, just store them in a regular kernel data structure, and rework the reference counting, locking, and lifetime accordingly. Retain support for RCU-mode key lookups by using a hash table. Replace fscrypt_sb_free() with fscrypt_sb_delete(), which releases the keys synchronously and runs a bit earlier during unmount, so that block devices are still available. A side effect of this patch is that neither the master keys themselves nor the filesystem keyrings will be listed in /proc/keys anymore. ("Master key users" and the master key users keyrings will still be listed.) However, this was mostly an implementation detail, and it was intended just for debugging purposes. I don't know of anyone using it. This patch does *not* change how "master key users" (->mk_users) works; that still uses the keyrings subsystem. That is still needed for key quotas, and changing that isn't necessary to solve the issues listed above. If we decide to change that too, it would be a separate patch. I've marked this as fixing the original commit that added the fscrypt keyring, but as noted above the most important issue that this patch fixes wasn't introduced until the addition of inline encryption support. Fixes: 22d94f493bfb ("fscrypt: add FS_IOC_ADD_ENCRYPTION_KEY ioctl") Signed-off-by: Eric Biggers <ebiggers@google.com> Link: https://lore.kernel.org/r/20220901193208.138056-2-ebiggers@kernel.org
2022-09-02 03:32:06 +08:00
down_write(&mk->mk_sem);
err = add_existing_master_key(mk, secret);
up_write(&mk->mk_sem);
fscrypt: add FS_IOC_REMOVE_ENCRYPTION_KEY ioctl Add a new fscrypt ioctl, FS_IOC_REMOVE_ENCRYPTION_KEY. This ioctl removes an encryption key that was added by FS_IOC_ADD_ENCRYPTION_KEY. It wipes the secret key itself, then "locks" the encrypted files and directories that had been unlocked using that key -- implemented by evicting the relevant dentries and inodes from the VFS caches. The problem this solves is that many fscrypt users want the ability to remove encryption keys, causing the corresponding encrypted directories to appear "locked" (presented in ciphertext form) again. Moreover, users want removing an encryption key to *really* remove it, in the sense that the removed keys cannot be recovered even if kernel memory is compromised, e.g. by the exploit of a kernel security vulnerability or by a physical attack. This is desirable after a user logs out of the system, for example. In many cases users even already assume this to be the case and are surprised to hear when it's not. It is not sufficient to simply unlink the master key from the keyring (or to revoke or invalidate it), since the actual encryption transform objects are still pinned in memory by their inodes. Therefore, to really remove a key we must also evict the relevant inodes. Currently one workaround is to run 'sync && echo 2 > /proc/sys/vm/drop_caches'. But, that evicts all unused inodes in the system rather than just the inodes associated with the key being removed, causing severe performance problems. Moreover, it requires root privileges, so regular users can't "lock" their encrypted files. Another workaround, used in Chromium OS kernels, is to add a new VFS-level ioctl FS_IOC_DROP_CACHE which is a more restricted version of drop_caches that operates on a single super_block. It does: shrink_dcache_sb(sb); invalidate_inodes(sb, false); But it's still a hack. Yet, the major users of filesystem encryption want this feature badly enough that they are actually using these hacks. To properly solve the problem, start maintaining a list of the inodes which have been "unlocked" using each master key. Originally this wasn't possible because the kernel didn't keep track of in-use master keys at all. But, with the ->s_master_keys keyring it is now possible. Then, add an ioctl FS_IOC_REMOVE_ENCRYPTION_KEY. It finds the specified master key in ->s_master_keys, then wipes the secret key itself, which prevents any additional inodes from being unlocked with the key. Then, it syncs the filesystem and evicts the inodes in the key's list. The normal inode eviction code will free and wipe the per-file keys (in ->i_crypt_info). Note that freeing ->i_crypt_info without evicting the inodes was also considered, but would have been racy. Some inodes may still be in use when a master key is removed, and we can't simply revoke random file descriptors, mmap's, etc. Thus, the ioctl simply skips in-use inodes, and returns -EBUSY to indicate that some inodes weren't evicted. The master key *secret* is still removed, but the fscrypt_master_key struct remains to keep track of the remaining inodes. Userspace can then retry the ioctl to evict the remaining inodes. Alternatively, if userspace adds the key again, the refreshed secret will be associated with the existing list of inodes so they remain correctly tracked for future key removals. The ioctl doesn't wipe pagecache pages. Thus, we tolerate that after a kernel compromise some portions of plaintext file contents may still be recoverable from memory. This can be solved by enabling page poisoning system-wide, which security conscious users may choose to do. But it's very difficult to solve otherwise, e.g. note that plaintext file contents may have been read in other places than pagecache pages. Like FS_IOC_ADD_ENCRYPTION_KEY, FS_IOC_REMOVE_ENCRYPTION_KEY is initially restricted to privileged users only. This is sufficient for some use cases, but not all. A later patch will relax this restriction, but it will require introducing key hashes, among other changes. Reviewed-by: Theodore Ts'o <tytso@mit.edu> Signed-off-by: Eric Biggers <ebiggers@google.com>
2019-08-05 10:35:46 +08:00
if (err == KEY_DEAD) {
fscrypt: stop using keyrings subsystem for fscrypt_master_key The approach of fs/crypto/ internally managing the fscrypt_master_key structs as the payloads of "struct key" objects contained in a "struct key" keyring has outlived its usefulness. The original idea was to simplify the code by reusing code from the keyrings subsystem. However, several issues have arisen that can't easily be resolved: - When a master key struct is destroyed, blk_crypto_evict_key() must be called on any per-mode keys embedded in it. (This started being the case when inline encryption support was added.) Yet, the keyrings subsystem can arbitrarily delay the destruction of keys, even past the time the filesystem was unmounted. Therefore, currently there is no easy way to call blk_crypto_evict_key() when a master key is destroyed. Currently, this is worked around by holding an extra reference to the filesystem's request_queue(s). But it was overlooked that the request_queue reference is *not* guaranteed to pin the corresponding blk_crypto_profile too; for device-mapper devices that support inline crypto, it doesn't. This can cause a use-after-free. - When the last inode that was using an incompletely-removed master key is evicted, the master key removal is completed by removing the key struct from the keyring. Currently this is done via key_invalidate(). Yet, key_invalidate() takes the key semaphore. This can deadlock when called from the shrinker, since in fscrypt_ioctl_add_key(), memory is allocated with GFP_KERNEL under the same semaphore. - More generally, the fact that the keyrings subsystem can arbitrarily delay the destruction of keys (via garbage collection delay, or via random processes getting temporary key references) is undesirable, as it means we can't strictly guarantee that all secrets are ever wiped. - Doing the master key lookups via the keyrings subsystem results in the key_permission LSM hook being called. fscrypt doesn't want this, as all access control for encrypted files is designed to happen via the files themselves, like any other files. The workaround which SELinux users are using is to change their SELinux policy to grant key search access to all domains. This works, but it is an odd extra step that shouldn't really have to be done. The fix for all these issues is to change the implementation to what I should have done originally: don't use the keyrings subsystem to keep track of the filesystem's fscrypt_master_key structs. Instead, just store them in a regular kernel data structure, and rework the reference counting, locking, and lifetime accordingly. Retain support for RCU-mode key lookups by using a hash table. Replace fscrypt_sb_free() with fscrypt_sb_delete(), which releases the keys synchronously and runs a bit earlier during unmount, so that block devices are still available. A side effect of this patch is that neither the master keys themselves nor the filesystem keyrings will be listed in /proc/keys anymore. ("Master key users" and the master key users keyrings will still be listed.) However, this was mostly an implementation detail, and it was intended just for debugging purposes. I don't know of anyone using it. This patch does *not* change how "master key users" (->mk_users) works; that still uses the keyrings subsystem. That is still needed for key quotas, and changing that isn't necessary to solve the issues listed above. If we decide to change that too, it would be a separate patch. I've marked this as fixing the original commit that added the fscrypt keyring, but as noted above the most important issue that this patch fixes wasn't introduced until the addition of inline encryption support. Fixes: 22d94f493bfb ("fscrypt: add FS_IOC_ADD_ENCRYPTION_KEY ioctl") Signed-off-by: Eric Biggers <ebiggers@google.com> Link: https://lore.kernel.org/r/20220901193208.138056-2-ebiggers@kernel.org
2022-09-02 03:32:06 +08:00
/*
* We found a key struct, but it's already been fully
* removed. Ignore the old struct and add a new one.
* fscrypt_add_key_mutex means we don't need to worry
* about concurrent adds.
*/
err = add_new_master_key(sb, secret, mk_spec);
fscrypt: add FS_IOC_REMOVE_ENCRYPTION_KEY ioctl Add a new fscrypt ioctl, FS_IOC_REMOVE_ENCRYPTION_KEY. This ioctl removes an encryption key that was added by FS_IOC_ADD_ENCRYPTION_KEY. It wipes the secret key itself, then "locks" the encrypted files and directories that had been unlocked using that key -- implemented by evicting the relevant dentries and inodes from the VFS caches. The problem this solves is that many fscrypt users want the ability to remove encryption keys, causing the corresponding encrypted directories to appear "locked" (presented in ciphertext form) again. Moreover, users want removing an encryption key to *really* remove it, in the sense that the removed keys cannot be recovered even if kernel memory is compromised, e.g. by the exploit of a kernel security vulnerability or by a physical attack. This is desirable after a user logs out of the system, for example. In many cases users even already assume this to be the case and are surprised to hear when it's not. It is not sufficient to simply unlink the master key from the keyring (or to revoke or invalidate it), since the actual encryption transform objects are still pinned in memory by their inodes. Therefore, to really remove a key we must also evict the relevant inodes. Currently one workaround is to run 'sync && echo 2 > /proc/sys/vm/drop_caches'. But, that evicts all unused inodes in the system rather than just the inodes associated with the key being removed, causing severe performance problems. Moreover, it requires root privileges, so regular users can't "lock" their encrypted files. Another workaround, used in Chromium OS kernels, is to add a new VFS-level ioctl FS_IOC_DROP_CACHE which is a more restricted version of drop_caches that operates on a single super_block. It does: shrink_dcache_sb(sb); invalidate_inodes(sb, false); But it's still a hack. Yet, the major users of filesystem encryption want this feature badly enough that they are actually using these hacks. To properly solve the problem, start maintaining a list of the inodes which have been "unlocked" using each master key. Originally this wasn't possible because the kernel didn't keep track of in-use master keys at all. But, with the ->s_master_keys keyring it is now possible. Then, add an ioctl FS_IOC_REMOVE_ENCRYPTION_KEY. It finds the specified master key in ->s_master_keys, then wipes the secret key itself, which prevents any additional inodes from being unlocked with the key. Then, it syncs the filesystem and evicts the inodes in the key's list. The normal inode eviction code will free and wipe the per-file keys (in ->i_crypt_info). Note that freeing ->i_crypt_info without evicting the inodes was also considered, but would have been racy. Some inodes may still be in use when a master key is removed, and we can't simply revoke random file descriptors, mmap's, etc. Thus, the ioctl simply skips in-use inodes, and returns -EBUSY to indicate that some inodes weren't evicted. The master key *secret* is still removed, but the fscrypt_master_key struct remains to keep track of the remaining inodes. Userspace can then retry the ioctl to evict the remaining inodes. Alternatively, if userspace adds the key again, the refreshed secret will be associated with the existing list of inodes so they remain correctly tracked for future key removals. The ioctl doesn't wipe pagecache pages. Thus, we tolerate that after a kernel compromise some portions of plaintext file contents may still be recoverable from memory. This can be solved by enabling page poisoning system-wide, which security conscious users may choose to do. But it's very difficult to solve otherwise, e.g. note that plaintext file contents may have been read in other places than pagecache pages. Like FS_IOC_ADD_ENCRYPTION_KEY, FS_IOC_REMOVE_ENCRYPTION_KEY is initially restricted to privileged users only. This is sufficient for some use cases, but not all. A later patch will relax this restriction, but it will require introducing key hashes, among other changes. Reviewed-by: Theodore Ts'o <tytso@mit.edu> Signed-off-by: Eric Biggers <ebiggers@google.com>
2019-08-05 10:35:46 +08:00
}
fscrypt: stop using keyrings subsystem for fscrypt_master_key The approach of fs/crypto/ internally managing the fscrypt_master_key structs as the payloads of "struct key" objects contained in a "struct key" keyring has outlived its usefulness. The original idea was to simplify the code by reusing code from the keyrings subsystem. However, several issues have arisen that can't easily be resolved: - When a master key struct is destroyed, blk_crypto_evict_key() must be called on any per-mode keys embedded in it. (This started being the case when inline encryption support was added.) Yet, the keyrings subsystem can arbitrarily delay the destruction of keys, even past the time the filesystem was unmounted. Therefore, currently there is no easy way to call blk_crypto_evict_key() when a master key is destroyed. Currently, this is worked around by holding an extra reference to the filesystem's request_queue(s). But it was overlooked that the request_queue reference is *not* guaranteed to pin the corresponding blk_crypto_profile too; for device-mapper devices that support inline crypto, it doesn't. This can cause a use-after-free. - When the last inode that was using an incompletely-removed master key is evicted, the master key removal is completed by removing the key struct from the keyring. Currently this is done via key_invalidate(). Yet, key_invalidate() takes the key semaphore. This can deadlock when called from the shrinker, since in fscrypt_ioctl_add_key(), memory is allocated with GFP_KERNEL under the same semaphore. - More generally, the fact that the keyrings subsystem can arbitrarily delay the destruction of keys (via garbage collection delay, or via random processes getting temporary key references) is undesirable, as it means we can't strictly guarantee that all secrets are ever wiped. - Doing the master key lookups via the keyrings subsystem results in the key_permission LSM hook being called. fscrypt doesn't want this, as all access control for encrypted files is designed to happen via the files themselves, like any other files. The workaround which SELinux users are using is to change their SELinux policy to grant key search access to all domains. This works, but it is an odd extra step that shouldn't really have to be done. The fix for all these issues is to change the implementation to what I should have done originally: don't use the keyrings subsystem to keep track of the filesystem's fscrypt_master_key structs. Instead, just store them in a regular kernel data structure, and rework the reference counting, locking, and lifetime accordingly. Retain support for RCU-mode key lookups by using a hash table. Replace fscrypt_sb_free() with fscrypt_sb_delete(), which releases the keys synchronously and runs a bit earlier during unmount, so that block devices are still available. A side effect of this patch is that neither the master keys themselves nor the filesystem keyrings will be listed in /proc/keys anymore. ("Master key users" and the master key users keyrings will still be listed.) However, this was mostly an implementation detail, and it was intended just for debugging purposes. I don't know of anyone using it. This patch does *not* change how "master key users" (->mk_users) works; that still uses the keyrings subsystem. That is still needed for key quotas, and changing that isn't necessary to solve the issues listed above. If we decide to change that too, it would be a separate patch. I've marked this as fixing the original commit that added the fscrypt keyring, but as noted above the most important issue that this patch fixes wasn't introduced until the addition of inline encryption support. Fixes: 22d94f493bfb ("fscrypt: add FS_IOC_ADD_ENCRYPTION_KEY ioctl") Signed-off-by: Eric Biggers <ebiggers@google.com> Link: https://lore.kernel.org/r/20220901193208.138056-2-ebiggers@kernel.org
2022-09-02 03:32:06 +08:00
fscrypt_put_master_key(mk);
fscrypt: add FS_IOC_ADD_ENCRYPTION_KEY ioctl Add a new fscrypt ioctl, FS_IOC_ADD_ENCRYPTION_KEY. This ioctl adds an encryption key to the filesystem's fscrypt keyring ->s_master_keys, making any files encrypted with that key appear "unlocked". Why we need this ~~~~~~~~~~~~~~~~ The main problem is that the "locked/unlocked" (ciphertext/plaintext) status of encrypted files is global, but the fscrypt keys are not. fscrypt only looks for keys in the keyring(s) the process accessing the filesystem is subscribed to: the thread keyring, process keyring, and session keyring, where the session keyring may contain the user keyring. Therefore, userspace has to put fscrypt keys in the keyrings for individual users or sessions. But this means that when a process with a different keyring tries to access encrypted files, whether they appear "unlocked" or not is nondeterministic. This is because it depends on whether the files are currently present in the inode cache. Fixing this by consistently providing each process its own view of the filesystem depending on whether it has the key or not isn't feasible due to how the VFS caches work. Furthermore, while sometimes users expect this behavior, it is misguided for two reasons. First, it would be an OS-level access control mechanism largely redundant with existing access control mechanisms such as UNIX file permissions, ACLs, LSMs, etc. Encryption is actually for protecting the data at rest. Second, almost all users of fscrypt actually do need the keys to be global. The largest users of fscrypt, Android and Chromium OS, achieve this by having PID 1 create a "session keyring" that is inherited by every process. This works, but it isn't scalable because it prevents session keyrings from being used for any other purpose. On general-purpose Linux distros, the 'fscrypt' userspace tool [1] can't similarly abuse the session keyring, so to make 'sudo' work on all systems it has to link all the user keyrings into root's user keyring [2]. This is ugly and raises security concerns. Moreover it can't make the keys available to system services, such as sshd trying to access the user's '~/.ssh' directory (see [3], [4]) or NetworkManager trying to read certificates from the user's home directory (see [5]); or to Docker containers (see [6], [7]). By having an API to add a key to the *filesystem* we'll be able to fix the above bugs, remove userspace workarounds, and clearly express the intended semantics: the locked/unlocked status of an encrypted directory is global, and encryption is orthogonal to OS-level access control. Why not use the add_key() syscall ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ We use an ioctl for this API rather than the existing add_key() system call because the ioctl gives us the flexibility needed to implement fscrypt-specific semantics that will be introduced in later patches: - Supporting key removal with the semantics such that the secret is removed immediately and any unused inodes using the key are evicted; also, the eviction of any in-use inodes can be retried. - Calculating a key-dependent cryptographic identifier and returning it to userspace. - Allowing keys to be added and removed by non-root users, but only keys for v2 encryption policies; and to prevent denial-of-service attacks, users can only remove keys they themselves have added, and a key is only really removed after all users who added it have removed it. Trying to shoehorn these semantics into the keyrings syscalls would be very difficult, whereas the ioctls make things much easier. However, to reuse code the implementation still uses the keyrings service internally. Thus we get lockless RCU-mode key lookups without having to re-implement it, and the keys automatically show up in /proc/keys for debugging purposes. References: [1] https://github.com/google/fscrypt [2] https://goo.gl/55cCrI#heading=h.vf09isp98isb [3] https://github.com/google/fscrypt/issues/111#issuecomment-444347939 [4] https://github.com/google/fscrypt/issues/116 [5] https://bugs.launchpad.net/ubuntu/+source/fscrypt/+bug/1770715 [6] https://github.com/google/fscrypt/issues/128 [7] https://askubuntu.com/questions/1130306/cannot-run-docker-on-an-encrypted-filesystem Reviewed-by: Theodore Ts'o <tytso@mit.edu> Signed-off-by: Eric Biggers <ebiggers@google.com>
2019-08-05 10:35:46 +08:00
}
mutex_unlock(&fscrypt_add_key_mutex);
return err;
}
fscrypt: add fscrypt_add_test_dummy_key() Currently, the test_dummy_encryption mount option (which is used for encryption I/O testing with xfstests) uses v1 encryption policies, and it relies on userspace inserting a test key into the session keyring. We need test_dummy_encryption to support v2 encryption policies too. Requiring userspace to add the test key doesn't work well with v2 policies, since v2 policies only support the filesystem keyring (not the session keyring), and keys in the filesystem keyring are lost when the filesystem is unmounted. Hooking all test code that unmounts and re-mounts the filesystem would be difficult. Instead, let's make the filesystem automatically add the test key to its keyring when test_dummy_encryption is enabled. That puts the responsibility for choosing the test key on the kernel. We could just hard-code a key. But out of paranoia, let's first try using a per-boot random key, to prevent this code from being misused. A per-boot key will work as long as no one expects dummy-encrypted files to remain accessible after a reboot. (gce-xfstests doesn't.) Therefore, this patch adds a function fscrypt_add_test_dummy_key() which implements the above. The next patch will use it. Link: https://lore.kernel.org/r/20200512233251.118314-3-ebiggers@kernel.org Reviewed-by: Theodore Ts'o <tytso@mit.edu> Reviewed-by: Jaegeuk Kim <jaegeuk@kernel.org> Signed-off-by: Eric Biggers <ebiggers@google.com>
2020-05-13 07:32:49 +08:00
static int add_master_key(struct super_block *sb,
struct fscrypt_master_key_secret *secret,
struct fscrypt_key_specifier *key_spec)
{
int err;
if (key_spec->type == FSCRYPT_KEY_SPEC_TYPE_IDENTIFIER) {
err = fscrypt_init_hkdf(&secret->hkdf, secret->raw,
secret->size);
if (err)
return err;
/*
* Now that the HKDF context is initialized, the raw key is no
* longer needed.
*/
memzero_explicit(secret->raw, secret->size);
/* Calculate the key identifier */
err = fscrypt_hkdf_expand(&secret->hkdf,
HKDF_CONTEXT_KEY_IDENTIFIER, NULL, 0,
key_spec->u.identifier,
FSCRYPT_KEY_IDENTIFIER_SIZE);
if (err)
return err;
}
return do_add_master_key(sb, secret, key_spec);
}
fscrypt: support passing a keyring key to FS_IOC_ADD_ENCRYPTION_KEY Extend the FS_IOC_ADD_ENCRYPTION_KEY ioctl to allow the raw key to be specified by a Linux keyring key, rather than specified directly. This is useful because fscrypt keys belong to a particular filesystem instance, so they are destroyed when that filesystem is unmounted. Usually this is desired. But in some cases, userspace may need to unmount and re-mount the filesystem while keeping the keys, e.g. during a system update. This requires keeping the keys somewhere else too. The keys could be kept in memory in a userspace daemon. But depending on the security architecture and assumptions, it can be preferable to keep them only in kernel memory, where they are unreadable by userspace. We also can't solve this by going back to the original fscrypt API (where for each file, the master key was looked up in the process's keyring hierarchy) because that caused lots of problems of its own. Therefore, add the ability for FS_IOC_ADD_ENCRYPTION_KEY to accept a Linux keyring key. This solves the problem by allowing userspace to (if needed) save the keys securely in a Linux keyring for re-provisioning, while still using the new fscrypt key management ioctls. This is analogous to how dm-crypt accepts a Linux keyring key, but the key is then stored internally in the dm-crypt data structures rather than being looked up again each time the dm-crypt device is accessed. Use a custom key type "fscrypt-provisioning" rather than one of the existing key types such as "logon". This is strongly desired because it enforces that these keys are only usable for a particular purpose: for fscrypt as input to a particular KDF. Otherwise, the keys could also be passed to any kernel API that accepts a "logon" key with any service prefix, e.g. dm-crypt, UBIFS, or (recently proposed) AF_ALG. This would risk leaking information about the raw key despite it ostensibly being unreadable. Of course, this mistake has already been made for multiple kernel APIs; but since this is a new API, let's do it right. This patch has been tested using an xfstest which I wrote to test it. Link: https://lore.kernel.org/r/20191119222447.226853-1-ebiggers@kernel.org Signed-off-by: Eric Biggers <ebiggers@google.com>
2019-11-20 06:24:47 +08:00
static int fscrypt_provisioning_key_preparse(struct key_preparsed_payload *prep)
{
const struct fscrypt_provisioning_key_payload *payload = prep->data;
if (prep->datalen < sizeof(*payload) + FSCRYPT_MIN_KEY_SIZE ||
prep->datalen > sizeof(*payload) + FSCRYPT_MAX_KEY_SIZE)
return -EINVAL;
if (payload->type != FSCRYPT_KEY_SPEC_TYPE_DESCRIPTOR &&
payload->type != FSCRYPT_KEY_SPEC_TYPE_IDENTIFIER)
return -EINVAL;
if (payload->__reserved)
return -EINVAL;
prep->payload.data[0] = kmemdup(payload, prep->datalen, GFP_KERNEL);
if (!prep->payload.data[0])
return -ENOMEM;
prep->quotalen = prep->datalen;
return 0;
}
static void fscrypt_provisioning_key_free_preparse(
struct key_preparsed_payload *prep)
{
mm, treewide: rename kzfree() to kfree_sensitive() As said by Linus: A symmetric naming is only helpful if it implies symmetries in use. Otherwise it's actively misleading. In "kzalloc()", the z is meaningful and an important part of what the caller wants. In "kzfree()", the z is actively detrimental, because maybe in the future we really _might_ want to use that "memfill(0xdeadbeef)" or something. The "zero" part of the interface isn't even _relevant_. The main reason that kzfree() exists is to clear sensitive information that should not be leaked to other future users of the same memory objects. Rename kzfree() to kfree_sensitive() to follow the example of the recently added kvfree_sensitive() and make the intention of the API more explicit. In addition, memzero_explicit() is used to clear the memory to make sure that it won't get optimized away by the compiler. The renaming is done by using the command sequence: git grep -w --name-only kzfree |\ xargs sed -i 's/kzfree/kfree_sensitive/' followed by some editing of the kfree_sensitive() kerneldoc and adding a kzfree backward compatibility macro in slab.h. [akpm@linux-foundation.org: fs/crypto/inline_crypt.c needs linux/slab.h] [akpm@linux-foundation.org: fix fs/crypto/inline_crypt.c some more] Suggested-by: Joe Perches <joe@perches.com> Signed-off-by: Waiman Long <longman@redhat.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Acked-by: David Howells <dhowells@redhat.com> Acked-by: Michal Hocko <mhocko@suse.com> Acked-by: Johannes Weiner <hannes@cmpxchg.org> Cc: Jarkko Sakkinen <jarkko.sakkinen@linux.intel.com> Cc: James Morris <jmorris@namei.org> Cc: "Serge E. Hallyn" <serge@hallyn.com> Cc: Joe Perches <joe@perches.com> Cc: Matthew Wilcox <willy@infradead.org> Cc: David Rientjes <rientjes@google.com> Cc: Dan Carpenter <dan.carpenter@oracle.com> Cc: "Jason A . Donenfeld" <Jason@zx2c4.com> Link: http://lkml.kernel.org/r/20200616154311.12314-3-longman@redhat.com Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2020-08-07 14:18:13 +08:00
kfree_sensitive(prep->payload.data[0]);
fscrypt: support passing a keyring key to FS_IOC_ADD_ENCRYPTION_KEY Extend the FS_IOC_ADD_ENCRYPTION_KEY ioctl to allow the raw key to be specified by a Linux keyring key, rather than specified directly. This is useful because fscrypt keys belong to a particular filesystem instance, so they are destroyed when that filesystem is unmounted. Usually this is desired. But in some cases, userspace may need to unmount and re-mount the filesystem while keeping the keys, e.g. during a system update. This requires keeping the keys somewhere else too. The keys could be kept in memory in a userspace daemon. But depending on the security architecture and assumptions, it can be preferable to keep them only in kernel memory, where they are unreadable by userspace. We also can't solve this by going back to the original fscrypt API (where for each file, the master key was looked up in the process's keyring hierarchy) because that caused lots of problems of its own. Therefore, add the ability for FS_IOC_ADD_ENCRYPTION_KEY to accept a Linux keyring key. This solves the problem by allowing userspace to (if needed) save the keys securely in a Linux keyring for re-provisioning, while still using the new fscrypt key management ioctls. This is analogous to how dm-crypt accepts a Linux keyring key, but the key is then stored internally in the dm-crypt data structures rather than being looked up again each time the dm-crypt device is accessed. Use a custom key type "fscrypt-provisioning" rather than one of the existing key types such as "logon". This is strongly desired because it enforces that these keys are only usable for a particular purpose: for fscrypt as input to a particular KDF. Otherwise, the keys could also be passed to any kernel API that accepts a "logon" key with any service prefix, e.g. dm-crypt, UBIFS, or (recently proposed) AF_ALG. This would risk leaking information about the raw key despite it ostensibly being unreadable. Of course, this mistake has already been made for multiple kernel APIs; but since this is a new API, let's do it right. This patch has been tested using an xfstest which I wrote to test it. Link: https://lore.kernel.org/r/20191119222447.226853-1-ebiggers@kernel.org Signed-off-by: Eric Biggers <ebiggers@google.com>
2019-11-20 06:24:47 +08:00
}
static void fscrypt_provisioning_key_describe(const struct key *key,
struct seq_file *m)
{
seq_puts(m, key->description);
if (key_is_positive(key)) {
const struct fscrypt_provisioning_key_payload *payload =
key->payload.data[0];
seq_printf(m, ": %u [%u]", key->datalen, payload->type);
}
}
static void fscrypt_provisioning_key_destroy(struct key *key)
{
mm, treewide: rename kzfree() to kfree_sensitive() As said by Linus: A symmetric naming is only helpful if it implies symmetries in use. Otherwise it's actively misleading. In "kzalloc()", the z is meaningful and an important part of what the caller wants. In "kzfree()", the z is actively detrimental, because maybe in the future we really _might_ want to use that "memfill(0xdeadbeef)" or something. The "zero" part of the interface isn't even _relevant_. The main reason that kzfree() exists is to clear sensitive information that should not be leaked to other future users of the same memory objects. Rename kzfree() to kfree_sensitive() to follow the example of the recently added kvfree_sensitive() and make the intention of the API more explicit. In addition, memzero_explicit() is used to clear the memory to make sure that it won't get optimized away by the compiler. The renaming is done by using the command sequence: git grep -w --name-only kzfree |\ xargs sed -i 's/kzfree/kfree_sensitive/' followed by some editing of the kfree_sensitive() kerneldoc and adding a kzfree backward compatibility macro in slab.h. [akpm@linux-foundation.org: fs/crypto/inline_crypt.c needs linux/slab.h] [akpm@linux-foundation.org: fix fs/crypto/inline_crypt.c some more] Suggested-by: Joe Perches <joe@perches.com> Signed-off-by: Waiman Long <longman@redhat.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Acked-by: David Howells <dhowells@redhat.com> Acked-by: Michal Hocko <mhocko@suse.com> Acked-by: Johannes Weiner <hannes@cmpxchg.org> Cc: Jarkko Sakkinen <jarkko.sakkinen@linux.intel.com> Cc: James Morris <jmorris@namei.org> Cc: "Serge E. Hallyn" <serge@hallyn.com> Cc: Joe Perches <joe@perches.com> Cc: Matthew Wilcox <willy@infradead.org> Cc: David Rientjes <rientjes@google.com> Cc: Dan Carpenter <dan.carpenter@oracle.com> Cc: "Jason A . Donenfeld" <Jason@zx2c4.com> Link: http://lkml.kernel.org/r/20200616154311.12314-3-longman@redhat.com Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2020-08-07 14:18:13 +08:00
kfree_sensitive(key->payload.data[0]);
fscrypt: support passing a keyring key to FS_IOC_ADD_ENCRYPTION_KEY Extend the FS_IOC_ADD_ENCRYPTION_KEY ioctl to allow the raw key to be specified by a Linux keyring key, rather than specified directly. This is useful because fscrypt keys belong to a particular filesystem instance, so they are destroyed when that filesystem is unmounted. Usually this is desired. But in some cases, userspace may need to unmount and re-mount the filesystem while keeping the keys, e.g. during a system update. This requires keeping the keys somewhere else too. The keys could be kept in memory in a userspace daemon. But depending on the security architecture and assumptions, it can be preferable to keep them only in kernel memory, where they are unreadable by userspace. We also can't solve this by going back to the original fscrypt API (where for each file, the master key was looked up in the process's keyring hierarchy) because that caused lots of problems of its own. Therefore, add the ability for FS_IOC_ADD_ENCRYPTION_KEY to accept a Linux keyring key. This solves the problem by allowing userspace to (if needed) save the keys securely in a Linux keyring for re-provisioning, while still using the new fscrypt key management ioctls. This is analogous to how dm-crypt accepts a Linux keyring key, but the key is then stored internally in the dm-crypt data structures rather than being looked up again each time the dm-crypt device is accessed. Use a custom key type "fscrypt-provisioning" rather than one of the existing key types such as "logon". This is strongly desired because it enforces that these keys are only usable for a particular purpose: for fscrypt as input to a particular KDF. Otherwise, the keys could also be passed to any kernel API that accepts a "logon" key with any service prefix, e.g. dm-crypt, UBIFS, or (recently proposed) AF_ALG. This would risk leaking information about the raw key despite it ostensibly being unreadable. Of course, this mistake has already been made for multiple kernel APIs; but since this is a new API, let's do it right. This patch has been tested using an xfstest which I wrote to test it. Link: https://lore.kernel.org/r/20191119222447.226853-1-ebiggers@kernel.org Signed-off-by: Eric Biggers <ebiggers@google.com>
2019-11-20 06:24:47 +08:00
}
static struct key_type key_type_fscrypt_provisioning = {
.name = "fscrypt-provisioning",
.preparse = fscrypt_provisioning_key_preparse,
.free_preparse = fscrypt_provisioning_key_free_preparse,
.instantiate = generic_key_instantiate,
.describe = fscrypt_provisioning_key_describe,
.destroy = fscrypt_provisioning_key_destroy,
};
/*
* Retrieve the raw key from the Linux keyring key specified by 'key_id', and
* store it into 'secret'.
*
* The key must be of type "fscrypt-provisioning" and must have the field
* fscrypt_provisioning_key_payload::type set to 'type', indicating that it's
* only usable with fscrypt with the particular KDF version identified by
* 'type'. We don't use the "logon" key type because there's no way to
* completely restrict the use of such keys; they can be used by any kernel API
* that accepts "logon" keys and doesn't require a specific service prefix.
*
* The ability to specify the key via Linux keyring key is intended for cases
* where userspace needs to re-add keys after the filesystem is unmounted and
* re-mounted. Most users should just provide the raw key directly instead.
*/
static int get_keyring_key(u32 key_id, u32 type,
struct fscrypt_master_key_secret *secret)
{
key_ref_t ref;
struct key *key;
const struct fscrypt_provisioning_key_payload *payload;
int err;
ref = lookup_user_key(key_id, 0, KEY_NEED_SEARCH);
if (IS_ERR(ref))
return PTR_ERR(ref);
key = key_ref_to_ptr(ref);
if (key->type != &key_type_fscrypt_provisioning)
goto bad_key;
payload = key->payload.data[0];
/* Don't allow fscrypt v1 keys to be used as v2 keys and vice versa. */
if (payload->type != type)
goto bad_key;
secret->size = key->datalen - sizeof(*payload);
memcpy(secret->raw, payload->raw, secret->size);
err = 0;
goto out_put;
bad_key:
err = -EKEYREJECTED;
out_put:
key_ref_put(ref);
return err;
}
fscrypt: add FS_IOC_ADD_ENCRYPTION_KEY ioctl Add a new fscrypt ioctl, FS_IOC_ADD_ENCRYPTION_KEY. This ioctl adds an encryption key to the filesystem's fscrypt keyring ->s_master_keys, making any files encrypted with that key appear "unlocked". Why we need this ~~~~~~~~~~~~~~~~ The main problem is that the "locked/unlocked" (ciphertext/plaintext) status of encrypted files is global, but the fscrypt keys are not. fscrypt only looks for keys in the keyring(s) the process accessing the filesystem is subscribed to: the thread keyring, process keyring, and session keyring, where the session keyring may contain the user keyring. Therefore, userspace has to put fscrypt keys in the keyrings for individual users or sessions. But this means that when a process with a different keyring tries to access encrypted files, whether they appear "unlocked" or not is nondeterministic. This is because it depends on whether the files are currently present in the inode cache. Fixing this by consistently providing each process its own view of the filesystem depending on whether it has the key or not isn't feasible due to how the VFS caches work. Furthermore, while sometimes users expect this behavior, it is misguided for two reasons. First, it would be an OS-level access control mechanism largely redundant with existing access control mechanisms such as UNIX file permissions, ACLs, LSMs, etc. Encryption is actually for protecting the data at rest. Second, almost all users of fscrypt actually do need the keys to be global. The largest users of fscrypt, Android and Chromium OS, achieve this by having PID 1 create a "session keyring" that is inherited by every process. This works, but it isn't scalable because it prevents session keyrings from being used for any other purpose. On general-purpose Linux distros, the 'fscrypt' userspace tool [1] can't similarly abuse the session keyring, so to make 'sudo' work on all systems it has to link all the user keyrings into root's user keyring [2]. This is ugly and raises security concerns. Moreover it can't make the keys available to system services, such as sshd trying to access the user's '~/.ssh' directory (see [3], [4]) or NetworkManager trying to read certificates from the user's home directory (see [5]); or to Docker containers (see [6], [7]). By having an API to add a key to the *filesystem* we'll be able to fix the above bugs, remove userspace workarounds, and clearly express the intended semantics: the locked/unlocked status of an encrypted directory is global, and encryption is orthogonal to OS-level access control. Why not use the add_key() syscall ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ We use an ioctl for this API rather than the existing add_key() system call because the ioctl gives us the flexibility needed to implement fscrypt-specific semantics that will be introduced in later patches: - Supporting key removal with the semantics such that the secret is removed immediately and any unused inodes using the key are evicted; also, the eviction of any in-use inodes can be retried. - Calculating a key-dependent cryptographic identifier and returning it to userspace. - Allowing keys to be added and removed by non-root users, but only keys for v2 encryption policies; and to prevent denial-of-service attacks, users can only remove keys they themselves have added, and a key is only really removed after all users who added it have removed it. Trying to shoehorn these semantics into the keyrings syscalls would be very difficult, whereas the ioctls make things much easier. However, to reuse code the implementation still uses the keyrings service internally. Thus we get lockless RCU-mode key lookups without having to re-implement it, and the keys automatically show up in /proc/keys for debugging purposes. References: [1] https://github.com/google/fscrypt [2] https://goo.gl/55cCrI#heading=h.vf09isp98isb [3] https://github.com/google/fscrypt/issues/111#issuecomment-444347939 [4] https://github.com/google/fscrypt/issues/116 [5] https://bugs.launchpad.net/ubuntu/+source/fscrypt/+bug/1770715 [6] https://github.com/google/fscrypt/issues/128 [7] https://askubuntu.com/questions/1130306/cannot-run-docker-on-an-encrypted-filesystem Reviewed-by: Theodore Ts'o <tytso@mit.edu> Signed-off-by: Eric Biggers <ebiggers@google.com>
2019-08-05 10:35:46 +08:00
/*
* Add a master encryption key to the filesystem, causing all files which were
* encrypted with it to appear "unlocked" (decrypted) when accessed.
*
fscrypt: allow unprivileged users to add/remove keys for v2 policies Allow the FS_IOC_ADD_ENCRYPTION_KEY and FS_IOC_REMOVE_ENCRYPTION_KEY ioctls to be used by non-root users to add and remove encryption keys from the filesystem-level crypto keyrings, subject to limitations. Motivation: while privileged fscrypt key management is sufficient for some users (e.g. Android and Chromium OS, where a privileged process manages all keys), the old API by design also allows non-root users to set up and use encrypted directories, and we don't want to regress on that. Especially, we don't want to force users to continue using the old API, running into the visibility mismatch between files and keyrings and being unable to "lock" encrypted directories. Intuitively, the ioctls have to be privileged since they manipulate filesystem-level state. However, it's actually safe to make them unprivileged if we very carefully enforce some specific limitations. First, each key must be identified by a cryptographic hash so that a user can't add the wrong key for another user's files. For v2 encryption policies, we use the key_identifier for this. v1 policies don't have this, so managing keys for them remains privileged. Second, each key a user adds is charged to their quota for the keyrings service. Thus, a user can't exhaust memory by adding a huge number of keys. By default each non-root user is allowed up to 200 keys; this can be changed using the existing sysctl 'kernel.keys.maxkeys'. Third, if multiple users add the same key, we keep track of those users of the key (of which there remains a single copy), and won't really remove the key, i.e. "lock" the encrypted files, until all those users have removed it. This prevents denial of service attacks that would be possible under simpler schemes, such allowing the first user who added a key to remove it -- since that could be a malicious user who has compromised the key. Of course, encryption keys should be kept secret, but the idea is that using encryption should never be *less* secure than not using encryption, even if your key was compromised. We tolerate that a user will be unable to really remove a key, i.e. unable to "lock" their encrypted files, if another user has added the same key. But in a sense, this is actually a good thing because it will avoid providing a false notion of security where a key appears to have been removed when actually it's still in memory, available to any attacker who compromises the operating system kernel. Reviewed-by: Theodore Ts'o <tytso@mit.edu> Signed-off-by: Eric Biggers <ebiggers@google.com>
2019-08-05 10:35:47 +08:00
* When adding a key for use by v1 encryption policies, this ioctl is
* privileged, and userspace must provide the 'key_descriptor'.
*
* When adding a key for use by v2+ encryption policies, this ioctl is
* unprivileged. This is needed, in general, to allow non-root users to use
* encryption without encountering the visibility problems of process-subscribed
* keyrings and the inability to properly remove keys. This works by having
* each key identified by its cryptographically secure hash --- the
* 'key_identifier'. The cryptographic hash ensures that a malicious user
* cannot add the wrong key for a given identifier. Furthermore, each added key
* is charged to the appropriate user's quota for the keyrings service, which
* prevents a malicious user from adding too many keys. Finally, we forbid a
* user from removing a key while other users have added it too, which prevents
* a user who knows another user's key from causing a denial-of-service by
* removing it at an inopportune time. (We tolerate that a user who knows a key
* can prevent other users from removing it.)
*
fscrypt: add FS_IOC_ADD_ENCRYPTION_KEY ioctl Add a new fscrypt ioctl, FS_IOC_ADD_ENCRYPTION_KEY. This ioctl adds an encryption key to the filesystem's fscrypt keyring ->s_master_keys, making any files encrypted with that key appear "unlocked". Why we need this ~~~~~~~~~~~~~~~~ The main problem is that the "locked/unlocked" (ciphertext/plaintext) status of encrypted files is global, but the fscrypt keys are not. fscrypt only looks for keys in the keyring(s) the process accessing the filesystem is subscribed to: the thread keyring, process keyring, and session keyring, where the session keyring may contain the user keyring. Therefore, userspace has to put fscrypt keys in the keyrings for individual users or sessions. But this means that when a process with a different keyring tries to access encrypted files, whether they appear "unlocked" or not is nondeterministic. This is because it depends on whether the files are currently present in the inode cache. Fixing this by consistently providing each process its own view of the filesystem depending on whether it has the key or not isn't feasible due to how the VFS caches work. Furthermore, while sometimes users expect this behavior, it is misguided for two reasons. First, it would be an OS-level access control mechanism largely redundant with existing access control mechanisms such as UNIX file permissions, ACLs, LSMs, etc. Encryption is actually for protecting the data at rest. Second, almost all users of fscrypt actually do need the keys to be global. The largest users of fscrypt, Android and Chromium OS, achieve this by having PID 1 create a "session keyring" that is inherited by every process. This works, but it isn't scalable because it prevents session keyrings from being used for any other purpose. On general-purpose Linux distros, the 'fscrypt' userspace tool [1] can't similarly abuse the session keyring, so to make 'sudo' work on all systems it has to link all the user keyrings into root's user keyring [2]. This is ugly and raises security concerns. Moreover it can't make the keys available to system services, such as sshd trying to access the user's '~/.ssh' directory (see [3], [4]) or NetworkManager trying to read certificates from the user's home directory (see [5]); or to Docker containers (see [6], [7]). By having an API to add a key to the *filesystem* we'll be able to fix the above bugs, remove userspace workarounds, and clearly express the intended semantics: the locked/unlocked status of an encrypted directory is global, and encryption is orthogonal to OS-level access control. Why not use the add_key() syscall ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ We use an ioctl for this API rather than the existing add_key() system call because the ioctl gives us the flexibility needed to implement fscrypt-specific semantics that will be introduced in later patches: - Supporting key removal with the semantics such that the secret is removed immediately and any unused inodes using the key are evicted; also, the eviction of any in-use inodes can be retried. - Calculating a key-dependent cryptographic identifier and returning it to userspace. - Allowing keys to be added and removed by non-root users, but only keys for v2 encryption policies; and to prevent denial-of-service attacks, users can only remove keys they themselves have added, and a key is only really removed after all users who added it have removed it. Trying to shoehorn these semantics into the keyrings syscalls would be very difficult, whereas the ioctls make things much easier. However, to reuse code the implementation still uses the keyrings service internally. Thus we get lockless RCU-mode key lookups without having to re-implement it, and the keys automatically show up in /proc/keys for debugging purposes. References: [1] https://github.com/google/fscrypt [2] https://goo.gl/55cCrI#heading=h.vf09isp98isb [3] https://github.com/google/fscrypt/issues/111#issuecomment-444347939 [4] https://github.com/google/fscrypt/issues/116 [5] https://bugs.launchpad.net/ubuntu/+source/fscrypt/+bug/1770715 [6] https://github.com/google/fscrypt/issues/128 [7] https://askubuntu.com/questions/1130306/cannot-run-docker-on-an-encrypted-filesystem Reviewed-by: Theodore Ts'o <tytso@mit.edu> Signed-off-by: Eric Biggers <ebiggers@google.com>
2019-08-05 10:35:46 +08:00
* For more details, see the "FS_IOC_ADD_ENCRYPTION_KEY" section of
* Documentation/filesystems/fscrypt.rst.
*/
int fscrypt_ioctl_add_key(struct file *filp, void __user *_uarg)
{
struct super_block *sb = file_inode(filp)->i_sb;
struct fscrypt_add_key_arg __user *uarg = _uarg;
struct fscrypt_add_key_arg arg;
struct fscrypt_master_key_secret secret;
int err;
if (copy_from_user(&arg, uarg, sizeof(arg)))
return -EFAULT;
if (!valid_key_spec(&arg.key_spec))
return -EINVAL;
if (memchr_inv(arg.__reserved, 0, sizeof(arg.__reserved)))
return -EINVAL;
fscrypt: add fscrypt_add_test_dummy_key() Currently, the test_dummy_encryption mount option (which is used for encryption I/O testing with xfstests) uses v1 encryption policies, and it relies on userspace inserting a test key into the session keyring. We need test_dummy_encryption to support v2 encryption policies too. Requiring userspace to add the test key doesn't work well with v2 policies, since v2 policies only support the filesystem keyring (not the session keyring), and keys in the filesystem keyring are lost when the filesystem is unmounted. Hooking all test code that unmounts and re-mounts the filesystem would be difficult. Instead, let's make the filesystem automatically add the test key to its keyring when test_dummy_encryption is enabled. That puts the responsibility for choosing the test key on the kernel. We could just hard-code a key. But out of paranoia, let's first try using a per-boot random key, to prevent this code from being misused. A per-boot key will work as long as no one expects dummy-encrypted files to remain accessible after a reboot. (gce-xfstests doesn't.) Therefore, this patch adds a function fscrypt_add_test_dummy_key() which implements the above. The next patch will use it. Link: https://lore.kernel.org/r/20200512233251.118314-3-ebiggers@kernel.org Reviewed-by: Theodore Ts'o <tytso@mit.edu> Reviewed-by: Jaegeuk Kim <jaegeuk@kernel.org> Signed-off-by: Eric Biggers <ebiggers@google.com>
2020-05-13 07:32:49 +08:00
/*
* Only root can add keys that are identified by an arbitrary descriptor
* rather than by a cryptographic hash --- since otherwise a malicious
* user could add the wrong key.
*/
if (arg.key_spec.type == FSCRYPT_KEY_SPEC_TYPE_DESCRIPTOR &&
!capable(CAP_SYS_ADMIN))
return -EACCES;
fscrypt: add FS_IOC_ADD_ENCRYPTION_KEY ioctl Add a new fscrypt ioctl, FS_IOC_ADD_ENCRYPTION_KEY. This ioctl adds an encryption key to the filesystem's fscrypt keyring ->s_master_keys, making any files encrypted with that key appear "unlocked". Why we need this ~~~~~~~~~~~~~~~~ The main problem is that the "locked/unlocked" (ciphertext/plaintext) status of encrypted files is global, but the fscrypt keys are not. fscrypt only looks for keys in the keyring(s) the process accessing the filesystem is subscribed to: the thread keyring, process keyring, and session keyring, where the session keyring may contain the user keyring. Therefore, userspace has to put fscrypt keys in the keyrings for individual users or sessions. But this means that when a process with a different keyring tries to access encrypted files, whether they appear "unlocked" or not is nondeterministic. This is because it depends on whether the files are currently present in the inode cache. Fixing this by consistently providing each process its own view of the filesystem depending on whether it has the key or not isn't feasible due to how the VFS caches work. Furthermore, while sometimes users expect this behavior, it is misguided for two reasons. First, it would be an OS-level access control mechanism largely redundant with existing access control mechanisms such as UNIX file permissions, ACLs, LSMs, etc. Encryption is actually for protecting the data at rest. Second, almost all users of fscrypt actually do need the keys to be global. The largest users of fscrypt, Android and Chromium OS, achieve this by having PID 1 create a "session keyring" that is inherited by every process. This works, but it isn't scalable because it prevents session keyrings from being used for any other purpose. On general-purpose Linux distros, the 'fscrypt' userspace tool [1] can't similarly abuse the session keyring, so to make 'sudo' work on all systems it has to link all the user keyrings into root's user keyring [2]. This is ugly and raises security concerns. Moreover it can't make the keys available to system services, such as sshd trying to access the user's '~/.ssh' directory (see [3], [4]) or NetworkManager trying to read certificates from the user's home directory (see [5]); or to Docker containers (see [6], [7]). By having an API to add a key to the *filesystem* we'll be able to fix the above bugs, remove userspace workarounds, and clearly express the intended semantics: the locked/unlocked status of an encrypted directory is global, and encryption is orthogonal to OS-level access control. Why not use the add_key() syscall ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ We use an ioctl for this API rather than the existing add_key() system call because the ioctl gives us the flexibility needed to implement fscrypt-specific semantics that will be introduced in later patches: - Supporting key removal with the semantics such that the secret is removed immediately and any unused inodes using the key are evicted; also, the eviction of any in-use inodes can be retried. - Calculating a key-dependent cryptographic identifier and returning it to userspace. - Allowing keys to be added and removed by non-root users, but only keys for v2 encryption policies; and to prevent denial-of-service attacks, users can only remove keys they themselves have added, and a key is only really removed after all users who added it have removed it. Trying to shoehorn these semantics into the keyrings syscalls would be very difficult, whereas the ioctls make things much easier. However, to reuse code the implementation still uses the keyrings service internally. Thus we get lockless RCU-mode key lookups without having to re-implement it, and the keys automatically show up in /proc/keys for debugging purposes. References: [1] https://github.com/google/fscrypt [2] https://goo.gl/55cCrI#heading=h.vf09isp98isb [3] https://github.com/google/fscrypt/issues/111#issuecomment-444347939 [4] https://github.com/google/fscrypt/issues/116 [5] https://bugs.launchpad.net/ubuntu/+source/fscrypt/+bug/1770715 [6] https://github.com/google/fscrypt/issues/128 [7] https://askubuntu.com/questions/1130306/cannot-run-docker-on-an-encrypted-filesystem Reviewed-by: Theodore Ts'o <tytso@mit.edu> Signed-off-by: Eric Biggers <ebiggers@google.com>
2019-08-05 10:35:46 +08:00
memset(&secret, 0, sizeof(secret));
fscrypt: support passing a keyring key to FS_IOC_ADD_ENCRYPTION_KEY Extend the FS_IOC_ADD_ENCRYPTION_KEY ioctl to allow the raw key to be specified by a Linux keyring key, rather than specified directly. This is useful because fscrypt keys belong to a particular filesystem instance, so they are destroyed when that filesystem is unmounted. Usually this is desired. But in some cases, userspace may need to unmount and re-mount the filesystem while keeping the keys, e.g. during a system update. This requires keeping the keys somewhere else too. The keys could be kept in memory in a userspace daemon. But depending on the security architecture and assumptions, it can be preferable to keep them only in kernel memory, where they are unreadable by userspace. We also can't solve this by going back to the original fscrypt API (where for each file, the master key was looked up in the process's keyring hierarchy) because that caused lots of problems of its own. Therefore, add the ability for FS_IOC_ADD_ENCRYPTION_KEY to accept a Linux keyring key. This solves the problem by allowing userspace to (if needed) save the keys securely in a Linux keyring for re-provisioning, while still using the new fscrypt key management ioctls. This is analogous to how dm-crypt accepts a Linux keyring key, but the key is then stored internally in the dm-crypt data structures rather than being looked up again each time the dm-crypt device is accessed. Use a custom key type "fscrypt-provisioning" rather than one of the existing key types such as "logon". This is strongly desired because it enforces that these keys are only usable for a particular purpose: for fscrypt as input to a particular KDF. Otherwise, the keys could also be passed to any kernel API that accepts a "logon" key with any service prefix, e.g. dm-crypt, UBIFS, or (recently proposed) AF_ALG. This would risk leaking information about the raw key despite it ostensibly being unreadable. Of course, this mistake has already been made for multiple kernel APIs; but since this is a new API, let's do it right. This patch has been tested using an xfstest which I wrote to test it. Link: https://lore.kernel.org/r/20191119222447.226853-1-ebiggers@kernel.org Signed-off-by: Eric Biggers <ebiggers@google.com>
2019-11-20 06:24:47 +08:00
if (arg.key_id) {
if (arg.raw_size != 0)
return -EINVAL;
err = get_keyring_key(arg.key_id, arg.key_spec.type, &secret);
if (err)
goto out_wipe_secret;
} else {
if (arg.raw_size < FSCRYPT_MIN_KEY_SIZE ||
arg.raw_size > FSCRYPT_MAX_KEY_SIZE)
return -EINVAL;
secret.size = arg.raw_size;
err = -EFAULT;
if (copy_from_user(secret.raw, uarg->raw, secret.size))
goto out_wipe_secret;
}
fscrypt: add FS_IOC_ADD_ENCRYPTION_KEY ioctl Add a new fscrypt ioctl, FS_IOC_ADD_ENCRYPTION_KEY. This ioctl adds an encryption key to the filesystem's fscrypt keyring ->s_master_keys, making any files encrypted with that key appear "unlocked". Why we need this ~~~~~~~~~~~~~~~~ The main problem is that the "locked/unlocked" (ciphertext/plaintext) status of encrypted files is global, but the fscrypt keys are not. fscrypt only looks for keys in the keyring(s) the process accessing the filesystem is subscribed to: the thread keyring, process keyring, and session keyring, where the session keyring may contain the user keyring. Therefore, userspace has to put fscrypt keys in the keyrings for individual users or sessions. But this means that when a process with a different keyring tries to access encrypted files, whether they appear "unlocked" or not is nondeterministic. This is because it depends on whether the files are currently present in the inode cache. Fixing this by consistently providing each process its own view of the filesystem depending on whether it has the key or not isn't feasible due to how the VFS caches work. Furthermore, while sometimes users expect this behavior, it is misguided for two reasons. First, it would be an OS-level access control mechanism largely redundant with existing access control mechanisms such as UNIX file permissions, ACLs, LSMs, etc. Encryption is actually for protecting the data at rest. Second, almost all users of fscrypt actually do need the keys to be global. The largest users of fscrypt, Android and Chromium OS, achieve this by having PID 1 create a "session keyring" that is inherited by every process. This works, but it isn't scalable because it prevents session keyrings from being used for any other purpose. On general-purpose Linux distros, the 'fscrypt' userspace tool [1] can't similarly abuse the session keyring, so to make 'sudo' work on all systems it has to link all the user keyrings into root's user keyring [2]. This is ugly and raises security concerns. Moreover it can't make the keys available to system services, such as sshd trying to access the user's '~/.ssh' directory (see [3], [4]) or NetworkManager trying to read certificates from the user's home directory (see [5]); or to Docker containers (see [6], [7]). By having an API to add a key to the *filesystem* we'll be able to fix the above bugs, remove userspace workarounds, and clearly express the intended semantics: the locked/unlocked status of an encrypted directory is global, and encryption is orthogonal to OS-level access control. Why not use the add_key() syscall ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ We use an ioctl for this API rather than the existing add_key() system call because the ioctl gives us the flexibility needed to implement fscrypt-specific semantics that will be introduced in later patches: - Supporting key removal with the semantics such that the secret is removed immediately and any unused inodes using the key are evicted; also, the eviction of any in-use inodes can be retried. - Calculating a key-dependent cryptographic identifier and returning it to userspace. - Allowing keys to be added and removed by non-root users, but only keys for v2 encryption policies; and to prevent denial-of-service attacks, users can only remove keys they themselves have added, and a key is only really removed after all users who added it have removed it. Trying to shoehorn these semantics into the keyrings syscalls would be very difficult, whereas the ioctls make things much easier. However, to reuse code the implementation still uses the keyrings service internally. Thus we get lockless RCU-mode key lookups without having to re-implement it, and the keys automatically show up in /proc/keys for debugging purposes. References: [1] https://github.com/google/fscrypt [2] https://goo.gl/55cCrI#heading=h.vf09isp98isb [3] https://github.com/google/fscrypt/issues/111#issuecomment-444347939 [4] https://github.com/google/fscrypt/issues/116 [5] https://bugs.launchpad.net/ubuntu/+source/fscrypt/+bug/1770715 [6] https://github.com/google/fscrypt/issues/128 [7] https://askubuntu.com/questions/1130306/cannot-run-docker-on-an-encrypted-filesystem Reviewed-by: Theodore Ts'o <tytso@mit.edu> Signed-off-by: Eric Biggers <ebiggers@google.com>
2019-08-05 10:35:46 +08:00
fscrypt: add fscrypt_add_test_dummy_key() Currently, the test_dummy_encryption mount option (which is used for encryption I/O testing with xfstests) uses v1 encryption policies, and it relies on userspace inserting a test key into the session keyring. We need test_dummy_encryption to support v2 encryption policies too. Requiring userspace to add the test key doesn't work well with v2 policies, since v2 policies only support the filesystem keyring (not the session keyring), and keys in the filesystem keyring are lost when the filesystem is unmounted. Hooking all test code that unmounts and re-mounts the filesystem would be difficult. Instead, let's make the filesystem automatically add the test key to its keyring when test_dummy_encryption is enabled. That puts the responsibility for choosing the test key on the kernel. We could just hard-code a key. But out of paranoia, let's first try using a per-boot random key, to prevent this code from being misused. A per-boot key will work as long as no one expects dummy-encrypted files to remain accessible after a reboot. (gce-xfstests doesn't.) Therefore, this patch adds a function fscrypt_add_test_dummy_key() which implements the above. The next patch will use it. Link: https://lore.kernel.org/r/20200512233251.118314-3-ebiggers@kernel.org Reviewed-by: Theodore Ts'o <tytso@mit.edu> Reviewed-by: Jaegeuk Kim <jaegeuk@kernel.org> Signed-off-by: Eric Biggers <ebiggers@google.com>
2020-05-13 07:32:49 +08:00
err = add_master_key(sb, &secret, &arg.key_spec);
if (err)
fscrypt: allow unprivileged users to add/remove keys for v2 policies Allow the FS_IOC_ADD_ENCRYPTION_KEY and FS_IOC_REMOVE_ENCRYPTION_KEY ioctls to be used by non-root users to add and remove encryption keys from the filesystem-level crypto keyrings, subject to limitations. Motivation: while privileged fscrypt key management is sufficient for some users (e.g. Android and Chromium OS, where a privileged process manages all keys), the old API by design also allows non-root users to set up and use encrypted directories, and we don't want to regress on that. Especially, we don't want to force users to continue using the old API, running into the visibility mismatch between files and keyrings and being unable to "lock" encrypted directories. Intuitively, the ioctls have to be privileged since they manipulate filesystem-level state. However, it's actually safe to make them unprivileged if we very carefully enforce some specific limitations. First, each key must be identified by a cryptographic hash so that a user can't add the wrong key for another user's files. For v2 encryption policies, we use the key_identifier for this. v1 policies don't have this, so managing keys for them remains privileged. Second, each key a user adds is charged to their quota for the keyrings service. Thus, a user can't exhaust memory by adding a huge number of keys. By default each non-root user is allowed up to 200 keys; this can be changed using the existing sysctl 'kernel.keys.maxkeys'. Third, if multiple users add the same key, we keep track of those users of the key (of which there remains a single copy), and won't really remove the key, i.e. "lock" the encrypted files, until all those users have removed it. This prevents denial of service attacks that would be possible under simpler schemes, such allowing the first user who added a key to remove it -- since that could be a malicious user who has compromised the key. Of course, encryption keys should be kept secret, but the idea is that using encryption should never be *less* secure than not using encryption, even if your key was compromised. We tolerate that a user will be unable to really remove a key, i.e. unable to "lock" their encrypted files, if another user has added the same key. But in a sense, this is actually a good thing because it will avoid providing a false notion of security where a key appears to have been removed when actually it's still in memory, available to any attacker who compromises the operating system kernel. Reviewed-by: Theodore Ts'o <tytso@mit.edu> Signed-off-by: Eric Biggers <ebiggers@google.com>
2019-08-05 10:35:47 +08:00
goto out_wipe_secret;
fscrypt: v2 encryption policy support Add a new fscrypt policy version, "v2". It has the following changes from the original policy version, which we call "v1" (*): - Master keys (the user-provided encryption keys) are only ever used as input to HKDF-SHA512. This is more flexible and less error-prone, and it avoids the quirks and limitations of the AES-128-ECB based KDF. Three classes of cryptographically isolated subkeys are defined: - Per-file keys, like used in v1 policies except for the new KDF. - Per-mode keys. These implement the semantics of the DIRECT_KEY flag, which for v1 policies made the master key be used directly. These are also planned to be used for inline encryption when support for it is added. - Key identifiers (see below). - Each master key is identified by a 16-byte master_key_identifier, which is derived from the key itself using HKDF-SHA512. This prevents users from associating the wrong key with an encrypted file or directory. This was easily possible with v1 policies, which identified the key by an arbitrary 8-byte master_key_descriptor. - The key must be provided in the filesystem-level keyring, not in a process-subscribed keyring. The following UAPI additions are made: - The existing ioctl FS_IOC_SET_ENCRYPTION_POLICY can now be passed a fscrypt_policy_v2 to set a v2 encryption policy. It's disambiguated from fscrypt_policy/fscrypt_policy_v1 by the version code prefix. - A new ioctl FS_IOC_GET_ENCRYPTION_POLICY_EX is added. It allows getting the v1 or v2 encryption policy of an encrypted file or directory. The existing FS_IOC_GET_ENCRYPTION_POLICY ioctl could not be used because it did not have a way for userspace to indicate which policy structure is expected. The new ioctl includes a size field, so it is extensible to future fscrypt policy versions. - The ioctls FS_IOC_ADD_ENCRYPTION_KEY, FS_IOC_REMOVE_ENCRYPTION_KEY, and FS_IOC_GET_ENCRYPTION_KEY_STATUS now support managing keys for v2 encryption policies. Such keys are kept logically separate from keys for v1 encryption policies, and are identified by 'identifier' rather than by 'descriptor'. The 'identifier' need not be provided when adding a key, since the kernel will calculate it anyway. This patch temporarily keeps adding/removing v2 policy keys behind the same permission check done for adding/removing v1 policy keys: capable(CAP_SYS_ADMIN). However, the next patch will carefully take advantage of the cryptographically secure master_key_identifier to allow non-root users to add/remove v2 policy keys, thus providing a full replacement for v1 policies. (*) Actually, in the API fscrypt_policy::version is 0 while on-disk fscrypt_context::format is 1. But I believe it makes the most sense to advance both to '2' to have them be in sync, and to consider the numbering to start at 1 except for the API quirk. Reviewed-by: Paul Crowley <paulcrowley@google.com> Reviewed-by: Theodore Ts'o <tytso@mit.edu> Signed-off-by: Eric Biggers <ebiggers@google.com>
2019-08-05 10:35:47 +08:00
fscrypt: add fscrypt_add_test_dummy_key() Currently, the test_dummy_encryption mount option (which is used for encryption I/O testing with xfstests) uses v1 encryption policies, and it relies on userspace inserting a test key into the session keyring. We need test_dummy_encryption to support v2 encryption policies too. Requiring userspace to add the test key doesn't work well with v2 policies, since v2 policies only support the filesystem keyring (not the session keyring), and keys in the filesystem keyring are lost when the filesystem is unmounted. Hooking all test code that unmounts and re-mounts the filesystem would be difficult. Instead, let's make the filesystem automatically add the test key to its keyring when test_dummy_encryption is enabled. That puts the responsibility for choosing the test key on the kernel. We could just hard-code a key. But out of paranoia, let's first try using a per-boot random key, to prevent this code from being misused. A per-boot key will work as long as no one expects dummy-encrypted files to remain accessible after a reboot. (gce-xfstests doesn't.) Therefore, this patch adds a function fscrypt_add_test_dummy_key() which implements the above. The next patch will use it. Link: https://lore.kernel.org/r/20200512233251.118314-3-ebiggers@kernel.org Reviewed-by: Theodore Ts'o <tytso@mit.edu> Reviewed-by: Jaegeuk Kim <jaegeuk@kernel.org> Signed-off-by: Eric Biggers <ebiggers@google.com>
2020-05-13 07:32:49 +08:00
/* Return the key identifier to userspace, if applicable */
err = -EFAULT;
if (arg.key_spec.type == FSCRYPT_KEY_SPEC_TYPE_IDENTIFIER &&
copy_to_user(uarg->key_spec.u.identifier, arg.key_spec.u.identifier,
FSCRYPT_KEY_IDENTIFIER_SIZE))
goto out_wipe_secret;
err = 0;
fscrypt: add FS_IOC_ADD_ENCRYPTION_KEY ioctl Add a new fscrypt ioctl, FS_IOC_ADD_ENCRYPTION_KEY. This ioctl adds an encryption key to the filesystem's fscrypt keyring ->s_master_keys, making any files encrypted with that key appear "unlocked". Why we need this ~~~~~~~~~~~~~~~~ The main problem is that the "locked/unlocked" (ciphertext/plaintext) status of encrypted files is global, but the fscrypt keys are not. fscrypt only looks for keys in the keyring(s) the process accessing the filesystem is subscribed to: the thread keyring, process keyring, and session keyring, where the session keyring may contain the user keyring. Therefore, userspace has to put fscrypt keys in the keyrings for individual users or sessions. But this means that when a process with a different keyring tries to access encrypted files, whether they appear "unlocked" or not is nondeterministic. This is because it depends on whether the files are currently present in the inode cache. Fixing this by consistently providing each process its own view of the filesystem depending on whether it has the key or not isn't feasible due to how the VFS caches work. Furthermore, while sometimes users expect this behavior, it is misguided for two reasons. First, it would be an OS-level access control mechanism largely redundant with existing access control mechanisms such as UNIX file permissions, ACLs, LSMs, etc. Encryption is actually for protecting the data at rest. Second, almost all users of fscrypt actually do need the keys to be global. The largest users of fscrypt, Android and Chromium OS, achieve this by having PID 1 create a "session keyring" that is inherited by every process. This works, but it isn't scalable because it prevents session keyrings from being used for any other purpose. On general-purpose Linux distros, the 'fscrypt' userspace tool [1] can't similarly abuse the session keyring, so to make 'sudo' work on all systems it has to link all the user keyrings into root's user keyring [2]. This is ugly and raises security concerns. Moreover it can't make the keys available to system services, such as sshd trying to access the user's '~/.ssh' directory (see [3], [4]) or NetworkManager trying to read certificates from the user's home directory (see [5]); or to Docker containers (see [6], [7]). By having an API to add a key to the *filesystem* we'll be able to fix the above bugs, remove userspace workarounds, and clearly express the intended semantics: the locked/unlocked status of an encrypted directory is global, and encryption is orthogonal to OS-level access control. Why not use the add_key() syscall ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ We use an ioctl for this API rather than the existing add_key() system call because the ioctl gives us the flexibility needed to implement fscrypt-specific semantics that will be introduced in later patches: - Supporting key removal with the semantics such that the secret is removed immediately and any unused inodes using the key are evicted; also, the eviction of any in-use inodes can be retried. - Calculating a key-dependent cryptographic identifier and returning it to userspace. - Allowing keys to be added and removed by non-root users, but only keys for v2 encryption policies; and to prevent denial-of-service attacks, users can only remove keys they themselves have added, and a key is only really removed after all users who added it have removed it. Trying to shoehorn these semantics into the keyrings syscalls would be very difficult, whereas the ioctls make things much easier. However, to reuse code the implementation still uses the keyrings service internally. Thus we get lockless RCU-mode key lookups without having to re-implement it, and the keys automatically show up in /proc/keys for debugging purposes. References: [1] https://github.com/google/fscrypt [2] https://goo.gl/55cCrI#heading=h.vf09isp98isb [3] https://github.com/google/fscrypt/issues/111#issuecomment-444347939 [4] https://github.com/google/fscrypt/issues/116 [5] https://bugs.launchpad.net/ubuntu/+source/fscrypt/+bug/1770715 [6] https://github.com/google/fscrypt/issues/128 [7] https://askubuntu.com/questions/1130306/cannot-run-docker-on-an-encrypted-filesystem Reviewed-by: Theodore Ts'o <tytso@mit.edu> Signed-off-by: Eric Biggers <ebiggers@google.com>
2019-08-05 10:35:46 +08:00
out_wipe_secret:
wipe_master_key_secret(&secret);
return err;
}
EXPORT_SYMBOL_GPL(fscrypt_ioctl_add_key);
fscrypt: add new helper functions for test_dummy_encryption Unfortunately the design of fscrypt_set_test_dummy_encryption() doesn't work properly for the new mount API, as it combines too many steps into one function: - Parse the argument to test_dummy_encryption - Check the setting against the filesystem instance - Apply the setting to the filesystem instance The new mount API has split these into separate steps. ext4 partially worked around this by duplicating some of the logic, but it still had some bugs. To address this, add some new helper functions that split up the steps of fscrypt_set_test_dummy_encryption(): - fscrypt_parse_test_dummy_encryption() - fscrypt_dummy_policies_equal() - fscrypt_add_test_dummy_key() While we're add it, also add a function fscrypt_is_dummy_policy_set() which will be useful to avoid some #ifdef's. Signed-off-by: Eric Biggers <ebiggers@google.com> Link: https://lore.kernel.org/r/20220501050857.538984-5-ebiggers@kernel.org
2022-05-01 13:08:54 +08:00
static void
fscrypt_get_test_dummy_secret(struct fscrypt_master_key_secret *secret)
fscrypt: add fscrypt_add_test_dummy_key() Currently, the test_dummy_encryption mount option (which is used for encryption I/O testing with xfstests) uses v1 encryption policies, and it relies on userspace inserting a test key into the session keyring. We need test_dummy_encryption to support v2 encryption policies too. Requiring userspace to add the test key doesn't work well with v2 policies, since v2 policies only support the filesystem keyring (not the session keyring), and keys in the filesystem keyring are lost when the filesystem is unmounted. Hooking all test code that unmounts and re-mounts the filesystem would be difficult. Instead, let's make the filesystem automatically add the test key to its keyring when test_dummy_encryption is enabled. That puts the responsibility for choosing the test key on the kernel. We could just hard-code a key. But out of paranoia, let's first try using a per-boot random key, to prevent this code from being misused. A per-boot key will work as long as no one expects dummy-encrypted files to remain accessible after a reboot. (gce-xfstests doesn't.) Therefore, this patch adds a function fscrypt_add_test_dummy_key() which implements the above. The next patch will use it. Link: https://lore.kernel.org/r/20200512233251.118314-3-ebiggers@kernel.org Reviewed-by: Theodore Ts'o <tytso@mit.edu> Reviewed-by: Jaegeuk Kim <jaegeuk@kernel.org> Signed-off-by: Eric Biggers <ebiggers@google.com>
2020-05-13 07:32:49 +08:00
{
static u8 test_key[FSCRYPT_MAX_KEY_SIZE];
fscrypt: add new helper functions for test_dummy_encryption Unfortunately the design of fscrypt_set_test_dummy_encryption() doesn't work properly for the new mount API, as it combines too many steps into one function: - Parse the argument to test_dummy_encryption - Check the setting against the filesystem instance - Apply the setting to the filesystem instance The new mount API has split these into separate steps. ext4 partially worked around this by duplicating some of the logic, but it still had some bugs. To address this, add some new helper functions that split up the steps of fscrypt_set_test_dummy_encryption(): - fscrypt_parse_test_dummy_encryption() - fscrypt_dummy_policies_equal() - fscrypt_add_test_dummy_key() While we're add it, also add a function fscrypt_is_dummy_policy_set() which will be useful to avoid some #ifdef's. Signed-off-by: Eric Biggers <ebiggers@google.com> Link: https://lore.kernel.org/r/20220501050857.538984-5-ebiggers@kernel.org
2022-05-01 13:08:54 +08:00
get_random_once(test_key, FSCRYPT_MAX_KEY_SIZE);
memset(secret, 0, sizeof(*secret));
secret->size = FSCRYPT_MAX_KEY_SIZE;
memcpy(secret->raw, test_key, FSCRYPT_MAX_KEY_SIZE);
}
int fscrypt_get_test_dummy_key_identifier(
u8 key_identifier[FSCRYPT_KEY_IDENTIFIER_SIZE])
{
fscrypt: add fscrypt_add_test_dummy_key() Currently, the test_dummy_encryption mount option (which is used for encryption I/O testing with xfstests) uses v1 encryption policies, and it relies on userspace inserting a test key into the session keyring. We need test_dummy_encryption to support v2 encryption policies too. Requiring userspace to add the test key doesn't work well with v2 policies, since v2 policies only support the filesystem keyring (not the session keyring), and keys in the filesystem keyring are lost when the filesystem is unmounted. Hooking all test code that unmounts and re-mounts the filesystem would be difficult. Instead, let's make the filesystem automatically add the test key to its keyring when test_dummy_encryption is enabled. That puts the responsibility for choosing the test key on the kernel. We could just hard-code a key. But out of paranoia, let's first try using a per-boot random key, to prevent this code from being misused. A per-boot key will work as long as no one expects dummy-encrypted files to remain accessible after a reboot. (gce-xfstests doesn't.) Therefore, this patch adds a function fscrypt_add_test_dummy_key() which implements the above. The next patch will use it. Link: https://lore.kernel.org/r/20200512233251.118314-3-ebiggers@kernel.org Reviewed-by: Theodore Ts'o <tytso@mit.edu> Reviewed-by: Jaegeuk Kim <jaegeuk@kernel.org> Signed-off-by: Eric Biggers <ebiggers@google.com>
2020-05-13 07:32:49 +08:00
struct fscrypt_master_key_secret secret;
int err;
fscrypt: add new helper functions for test_dummy_encryption Unfortunately the design of fscrypt_set_test_dummy_encryption() doesn't work properly for the new mount API, as it combines too many steps into one function: - Parse the argument to test_dummy_encryption - Check the setting against the filesystem instance - Apply the setting to the filesystem instance The new mount API has split these into separate steps. ext4 partially worked around this by duplicating some of the logic, but it still had some bugs. To address this, add some new helper functions that split up the steps of fscrypt_set_test_dummy_encryption(): - fscrypt_parse_test_dummy_encryption() - fscrypt_dummy_policies_equal() - fscrypt_add_test_dummy_key() While we're add it, also add a function fscrypt_is_dummy_policy_set() which will be useful to avoid some #ifdef's. Signed-off-by: Eric Biggers <ebiggers@google.com> Link: https://lore.kernel.org/r/20220501050857.538984-5-ebiggers@kernel.org
2022-05-01 13:08:54 +08:00
fscrypt_get_test_dummy_secret(&secret);
fscrypt: add fscrypt_add_test_dummy_key() Currently, the test_dummy_encryption mount option (which is used for encryption I/O testing with xfstests) uses v1 encryption policies, and it relies on userspace inserting a test key into the session keyring. We need test_dummy_encryption to support v2 encryption policies too. Requiring userspace to add the test key doesn't work well with v2 policies, since v2 policies only support the filesystem keyring (not the session keyring), and keys in the filesystem keyring are lost when the filesystem is unmounted. Hooking all test code that unmounts and re-mounts the filesystem would be difficult. Instead, let's make the filesystem automatically add the test key to its keyring when test_dummy_encryption is enabled. That puts the responsibility for choosing the test key on the kernel. We could just hard-code a key. But out of paranoia, let's first try using a per-boot random key, to prevent this code from being misused. A per-boot key will work as long as no one expects dummy-encrypted files to remain accessible after a reboot. (gce-xfstests doesn't.) Therefore, this patch adds a function fscrypt_add_test_dummy_key() which implements the above. The next patch will use it. Link: https://lore.kernel.org/r/20200512233251.118314-3-ebiggers@kernel.org Reviewed-by: Theodore Ts'o <tytso@mit.edu> Reviewed-by: Jaegeuk Kim <jaegeuk@kernel.org> Signed-off-by: Eric Biggers <ebiggers@google.com>
2020-05-13 07:32:49 +08:00
fscrypt: add new helper functions for test_dummy_encryption Unfortunately the design of fscrypt_set_test_dummy_encryption() doesn't work properly for the new mount API, as it combines too many steps into one function: - Parse the argument to test_dummy_encryption - Check the setting against the filesystem instance - Apply the setting to the filesystem instance The new mount API has split these into separate steps. ext4 partially worked around this by duplicating some of the logic, but it still had some bugs. To address this, add some new helper functions that split up the steps of fscrypt_set_test_dummy_encryption(): - fscrypt_parse_test_dummy_encryption() - fscrypt_dummy_policies_equal() - fscrypt_add_test_dummy_key() While we're add it, also add a function fscrypt_is_dummy_policy_set() which will be useful to avoid some #ifdef's. Signed-off-by: Eric Biggers <ebiggers@google.com> Link: https://lore.kernel.org/r/20220501050857.538984-5-ebiggers@kernel.org
2022-05-01 13:08:54 +08:00
err = fscrypt_init_hkdf(&secret.hkdf, secret.raw, secret.size);
if (err)
goto out;
err = fscrypt_hkdf_expand(&secret.hkdf, HKDF_CONTEXT_KEY_IDENTIFIER,
NULL, 0, key_identifier,
FSCRYPT_KEY_IDENTIFIER_SIZE);
out:
wipe_master_key_secret(&secret);
return err;
}
fscrypt: add fscrypt_add_test_dummy_key() Currently, the test_dummy_encryption mount option (which is used for encryption I/O testing with xfstests) uses v1 encryption policies, and it relies on userspace inserting a test key into the session keyring. We need test_dummy_encryption to support v2 encryption policies too. Requiring userspace to add the test key doesn't work well with v2 policies, since v2 policies only support the filesystem keyring (not the session keyring), and keys in the filesystem keyring are lost when the filesystem is unmounted. Hooking all test code that unmounts and re-mounts the filesystem would be difficult. Instead, let's make the filesystem automatically add the test key to its keyring when test_dummy_encryption is enabled. That puts the responsibility for choosing the test key on the kernel. We could just hard-code a key. But out of paranoia, let's first try using a per-boot random key, to prevent this code from being misused. A per-boot key will work as long as no one expects dummy-encrypted files to remain accessible after a reboot. (gce-xfstests doesn't.) Therefore, this patch adds a function fscrypt_add_test_dummy_key() which implements the above. The next patch will use it. Link: https://lore.kernel.org/r/20200512233251.118314-3-ebiggers@kernel.org Reviewed-by: Theodore Ts'o <tytso@mit.edu> Reviewed-by: Jaegeuk Kim <jaegeuk@kernel.org> Signed-off-by: Eric Biggers <ebiggers@google.com>
2020-05-13 07:32:49 +08:00
fscrypt: add new helper functions for test_dummy_encryption Unfortunately the design of fscrypt_set_test_dummy_encryption() doesn't work properly for the new mount API, as it combines too many steps into one function: - Parse the argument to test_dummy_encryption - Check the setting against the filesystem instance - Apply the setting to the filesystem instance The new mount API has split these into separate steps. ext4 partially worked around this by duplicating some of the logic, but it still had some bugs. To address this, add some new helper functions that split up the steps of fscrypt_set_test_dummy_encryption(): - fscrypt_parse_test_dummy_encryption() - fscrypt_dummy_policies_equal() - fscrypt_add_test_dummy_key() While we're add it, also add a function fscrypt_is_dummy_policy_set() which will be useful to avoid some #ifdef's. Signed-off-by: Eric Biggers <ebiggers@google.com> Link: https://lore.kernel.org/r/20220501050857.538984-5-ebiggers@kernel.org
2022-05-01 13:08:54 +08:00
/**
* fscrypt_add_test_dummy_key() - add the test dummy encryption key
* @sb: the filesystem instance to add the key to
* @dummy_policy: the encryption policy for test_dummy_encryption
*
* If needed, add the key for the test_dummy_encryption mount option to the
* filesystem. To prevent misuse of this mount option, a per-boot random key is
* used instead of a hardcoded one. This makes it so that any encrypted files
* created using this option won't be accessible after a reboot.
*
* Return: 0 on success, -errno on failure
*/
int fscrypt_add_test_dummy_key(struct super_block *sb,
const struct fscrypt_dummy_policy *dummy_policy)
{
const union fscrypt_policy *policy = dummy_policy->policy;
struct fscrypt_key_specifier key_spec;
struct fscrypt_master_key_secret secret;
int err;
if (!policy)
return 0;
err = fscrypt_policy_to_key_spec(policy, &key_spec);
if (err)
return err;
fscrypt_get_test_dummy_secret(&secret);
err = add_master_key(sb, &secret, &key_spec);
fscrypt: add fscrypt_add_test_dummy_key() Currently, the test_dummy_encryption mount option (which is used for encryption I/O testing with xfstests) uses v1 encryption policies, and it relies on userspace inserting a test key into the session keyring. We need test_dummy_encryption to support v2 encryption policies too. Requiring userspace to add the test key doesn't work well with v2 policies, since v2 policies only support the filesystem keyring (not the session keyring), and keys in the filesystem keyring are lost when the filesystem is unmounted. Hooking all test code that unmounts and re-mounts the filesystem would be difficult. Instead, let's make the filesystem automatically add the test key to its keyring when test_dummy_encryption is enabled. That puts the responsibility for choosing the test key on the kernel. We could just hard-code a key. But out of paranoia, let's first try using a per-boot random key, to prevent this code from being misused. A per-boot key will work as long as no one expects dummy-encrypted files to remain accessible after a reboot. (gce-xfstests doesn't.) Therefore, this patch adds a function fscrypt_add_test_dummy_key() which implements the above. The next patch will use it. Link: https://lore.kernel.org/r/20200512233251.118314-3-ebiggers@kernel.org Reviewed-by: Theodore Ts'o <tytso@mit.edu> Reviewed-by: Jaegeuk Kim <jaegeuk@kernel.org> Signed-off-by: Eric Biggers <ebiggers@google.com>
2020-05-13 07:32:49 +08:00
wipe_master_key_secret(&secret);
return err;
}
fscrypt: add new helper functions for test_dummy_encryption Unfortunately the design of fscrypt_set_test_dummy_encryption() doesn't work properly for the new mount API, as it combines too many steps into one function: - Parse the argument to test_dummy_encryption - Check the setting against the filesystem instance - Apply the setting to the filesystem instance The new mount API has split these into separate steps. ext4 partially worked around this by duplicating some of the logic, but it still had some bugs. To address this, add some new helper functions that split up the steps of fscrypt_set_test_dummy_encryption(): - fscrypt_parse_test_dummy_encryption() - fscrypt_dummy_policies_equal() - fscrypt_add_test_dummy_key() While we're add it, also add a function fscrypt_is_dummy_policy_set() which will be useful to avoid some #ifdef's. Signed-off-by: Eric Biggers <ebiggers@google.com> Link: https://lore.kernel.org/r/20220501050857.538984-5-ebiggers@kernel.org
2022-05-01 13:08:54 +08:00
EXPORT_SYMBOL_GPL(fscrypt_add_test_dummy_key);
fscrypt: add fscrypt_add_test_dummy_key() Currently, the test_dummy_encryption mount option (which is used for encryption I/O testing with xfstests) uses v1 encryption policies, and it relies on userspace inserting a test key into the session keyring. We need test_dummy_encryption to support v2 encryption policies too. Requiring userspace to add the test key doesn't work well with v2 policies, since v2 policies only support the filesystem keyring (not the session keyring), and keys in the filesystem keyring are lost when the filesystem is unmounted. Hooking all test code that unmounts and re-mounts the filesystem would be difficult. Instead, let's make the filesystem automatically add the test key to its keyring when test_dummy_encryption is enabled. That puts the responsibility for choosing the test key on the kernel. We could just hard-code a key. But out of paranoia, let's first try using a per-boot random key, to prevent this code from being misused. A per-boot key will work as long as no one expects dummy-encrypted files to remain accessible after a reboot. (gce-xfstests doesn't.) Therefore, this patch adds a function fscrypt_add_test_dummy_key() which implements the above. The next patch will use it. Link: https://lore.kernel.org/r/20200512233251.118314-3-ebiggers@kernel.org Reviewed-by: Theodore Ts'o <tytso@mit.edu> Reviewed-by: Jaegeuk Kim <jaegeuk@kernel.org> Signed-off-by: Eric Biggers <ebiggers@google.com>
2020-05-13 07:32:49 +08:00
fscrypt: require that key be added when setting a v2 encryption policy By looking up the master keys in a filesystem-level keyring rather than in the calling processes' key hierarchy, it becomes possible for a user to set an encryption policy which refers to some key they don't actually know, then encrypt their files using that key. Cryptographically this isn't much of a problem, but the semantics of this would be a bit weird. Thus, enforce that a v2 encryption policy can only be set if the user has previously added the key, or has capable(CAP_FOWNER). We tolerate that this problem will continue to exist for v1 encryption policies, however; there is no way around that. Reviewed-by: Theodore Ts'o <tytso@mit.edu> Signed-off-by: Eric Biggers <ebiggers@google.com>
2019-08-05 10:35:48 +08:00
/*
* Verify that the current user has added a master key with the given identifier
* (returns -ENOKEY if not). This is needed to prevent a user from encrypting
* their files using some other user's key which they don't actually know.
* Cryptographically this isn't much of a problem, but the semantics of this
* would be a bit weird, so it's best to just forbid it.
*
* The system administrator (CAP_FOWNER) can override this, which should be
* enough for any use cases where encryption policies are being set using keys
* that were chosen ahead of time but aren't available at the moment.
*
* Note that the key may have already removed by the time this returns, but
* that's okay; we just care whether the key was there at some point.
*
* Return: 0 if the key is added, -ENOKEY if it isn't, or another -errno code
*/
int fscrypt_verify_key_added(struct super_block *sb,
const u8 identifier[FSCRYPT_KEY_IDENTIFIER_SIZE])
{
struct fscrypt_key_specifier mk_spec;
struct fscrypt_master_key *mk;
fscrypt: stop using keyrings subsystem for fscrypt_master_key The approach of fs/crypto/ internally managing the fscrypt_master_key structs as the payloads of "struct key" objects contained in a "struct key" keyring has outlived its usefulness. The original idea was to simplify the code by reusing code from the keyrings subsystem. However, several issues have arisen that can't easily be resolved: - When a master key struct is destroyed, blk_crypto_evict_key() must be called on any per-mode keys embedded in it. (This started being the case when inline encryption support was added.) Yet, the keyrings subsystem can arbitrarily delay the destruction of keys, even past the time the filesystem was unmounted. Therefore, currently there is no easy way to call blk_crypto_evict_key() when a master key is destroyed. Currently, this is worked around by holding an extra reference to the filesystem's request_queue(s). But it was overlooked that the request_queue reference is *not* guaranteed to pin the corresponding blk_crypto_profile too; for device-mapper devices that support inline crypto, it doesn't. This can cause a use-after-free. - When the last inode that was using an incompletely-removed master key is evicted, the master key removal is completed by removing the key struct from the keyring. Currently this is done via key_invalidate(). Yet, key_invalidate() takes the key semaphore. This can deadlock when called from the shrinker, since in fscrypt_ioctl_add_key(), memory is allocated with GFP_KERNEL under the same semaphore. - More generally, the fact that the keyrings subsystem can arbitrarily delay the destruction of keys (via garbage collection delay, or via random processes getting temporary key references) is undesirable, as it means we can't strictly guarantee that all secrets are ever wiped. - Doing the master key lookups via the keyrings subsystem results in the key_permission LSM hook being called. fscrypt doesn't want this, as all access control for encrypted files is designed to happen via the files themselves, like any other files. The workaround which SELinux users are using is to change their SELinux policy to grant key search access to all domains. This works, but it is an odd extra step that shouldn't really have to be done. The fix for all these issues is to change the implementation to what I should have done originally: don't use the keyrings subsystem to keep track of the filesystem's fscrypt_master_key structs. Instead, just store them in a regular kernel data structure, and rework the reference counting, locking, and lifetime accordingly. Retain support for RCU-mode key lookups by using a hash table. Replace fscrypt_sb_free() with fscrypt_sb_delete(), which releases the keys synchronously and runs a bit earlier during unmount, so that block devices are still available. A side effect of this patch is that neither the master keys themselves nor the filesystem keyrings will be listed in /proc/keys anymore. ("Master key users" and the master key users keyrings will still be listed.) However, this was mostly an implementation detail, and it was intended just for debugging purposes. I don't know of anyone using it. This patch does *not* change how "master key users" (->mk_users) works; that still uses the keyrings subsystem. That is still needed for key quotas, and changing that isn't necessary to solve the issues listed above. If we decide to change that too, it would be a separate patch. I've marked this as fixing the original commit that added the fscrypt keyring, but as noted above the most important issue that this patch fixes wasn't introduced until the addition of inline encryption support. Fixes: 22d94f493bfb ("fscrypt: add FS_IOC_ADD_ENCRYPTION_KEY ioctl") Signed-off-by: Eric Biggers <ebiggers@google.com> Link: https://lore.kernel.org/r/20220901193208.138056-2-ebiggers@kernel.org
2022-09-02 03:32:06 +08:00
struct key *mk_user;
fscrypt: require that key be added when setting a v2 encryption policy By looking up the master keys in a filesystem-level keyring rather than in the calling processes' key hierarchy, it becomes possible for a user to set an encryption policy which refers to some key they don't actually know, then encrypt their files using that key. Cryptographically this isn't much of a problem, but the semantics of this would be a bit weird. Thus, enforce that a v2 encryption policy can only be set if the user has previously added the key, or has capable(CAP_FOWNER). We tolerate that this problem will continue to exist for v1 encryption policies, however; there is no way around that. Reviewed-by: Theodore Ts'o <tytso@mit.edu> Signed-off-by: Eric Biggers <ebiggers@google.com>
2019-08-05 10:35:48 +08:00
int err;
mk_spec.type = FSCRYPT_KEY_SPEC_TYPE_IDENTIFIER;
memcpy(mk_spec.u.identifier, identifier, FSCRYPT_KEY_IDENTIFIER_SIZE);
fscrypt: stop using keyrings subsystem for fscrypt_master_key The approach of fs/crypto/ internally managing the fscrypt_master_key structs as the payloads of "struct key" objects contained in a "struct key" keyring has outlived its usefulness. The original idea was to simplify the code by reusing code from the keyrings subsystem. However, several issues have arisen that can't easily be resolved: - When a master key struct is destroyed, blk_crypto_evict_key() must be called on any per-mode keys embedded in it. (This started being the case when inline encryption support was added.) Yet, the keyrings subsystem can arbitrarily delay the destruction of keys, even past the time the filesystem was unmounted. Therefore, currently there is no easy way to call blk_crypto_evict_key() when a master key is destroyed. Currently, this is worked around by holding an extra reference to the filesystem's request_queue(s). But it was overlooked that the request_queue reference is *not* guaranteed to pin the corresponding blk_crypto_profile too; for device-mapper devices that support inline crypto, it doesn't. This can cause a use-after-free. - When the last inode that was using an incompletely-removed master key is evicted, the master key removal is completed by removing the key struct from the keyring. Currently this is done via key_invalidate(). Yet, key_invalidate() takes the key semaphore. This can deadlock when called from the shrinker, since in fscrypt_ioctl_add_key(), memory is allocated with GFP_KERNEL under the same semaphore. - More generally, the fact that the keyrings subsystem can arbitrarily delay the destruction of keys (via garbage collection delay, or via random processes getting temporary key references) is undesirable, as it means we can't strictly guarantee that all secrets are ever wiped. - Doing the master key lookups via the keyrings subsystem results in the key_permission LSM hook being called. fscrypt doesn't want this, as all access control for encrypted files is designed to happen via the files themselves, like any other files. The workaround which SELinux users are using is to change their SELinux policy to grant key search access to all domains. This works, but it is an odd extra step that shouldn't really have to be done. The fix for all these issues is to change the implementation to what I should have done originally: don't use the keyrings subsystem to keep track of the filesystem's fscrypt_master_key structs. Instead, just store them in a regular kernel data structure, and rework the reference counting, locking, and lifetime accordingly. Retain support for RCU-mode key lookups by using a hash table. Replace fscrypt_sb_free() with fscrypt_sb_delete(), which releases the keys synchronously and runs a bit earlier during unmount, so that block devices are still available. A side effect of this patch is that neither the master keys themselves nor the filesystem keyrings will be listed in /proc/keys anymore. ("Master key users" and the master key users keyrings will still be listed.) However, this was mostly an implementation detail, and it was intended just for debugging purposes. I don't know of anyone using it. This patch does *not* change how "master key users" (->mk_users) works; that still uses the keyrings subsystem. That is still needed for key quotas, and changing that isn't necessary to solve the issues listed above. If we decide to change that too, it would be a separate patch. I've marked this as fixing the original commit that added the fscrypt keyring, but as noted above the most important issue that this patch fixes wasn't introduced until the addition of inline encryption support. Fixes: 22d94f493bfb ("fscrypt: add FS_IOC_ADD_ENCRYPTION_KEY ioctl") Signed-off-by: Eric Biggers <ebiggers@google.com> Link: https://lore.kernel.org/r/20220901193208.138056-2-ebiggers@kernel.org
2022-09-02 03:32:06 +08:00
mk = fscrypt_find_master_key(sb, &mk_spec);
if (!mk) {
err = -ENOKEY;
fscrypt: require that key be added when setting a v2 encryption policy By looking up the master keys in a filesystem-level keyring rather than in the calling processes' key hierarchy, it becomes possible for a user to set an encryption policy which refers to some key they don't actually know, then encrypt their files using that key. Cryptographically this isn't much of a problem, but the semantics of this would be a bit weird. Thus, enforce that a v2 encryption policy can only be set if the user has previously added the key, or has capable(CAP_FOWNER). We tolerate that this problem will continue to exist for v1 encryption policies, however; there is no way around that. Reviewed-by: Theodore Ts'o <tytso@mit.edu> Signed-off-by: Eric Biggers <ebiggers@google.com>
2019-08-05 10:35:48 +08:00
goto out;
}
fscrypt: stop using keyrings subsystem for fscrypt_master_key The approach of fs/crypto/ internally managing the fscrypt_master_key structs as the payloads of "struct key" objects contained in a "struct key" keyring has outlived its usefulness. The original idea was to simplify the code by reusing code from the keyrings subsystem. However, several issues have arisen that can't easily be resolved: - When a master key struct is destroyed, blk_crypto_evict_key() must be called on any per-mode keys embedded in it. (This started being the case when inline encryption support was added.) Yet, the keyrings subsystem can arbitrarily delay the destruction of keys, even past the time the filesystem was unmounted. Therefore, currently there is no easy way to call blk_crypto_evict_key() when a master key is destroyed. Currently, this is worked around by holding an extra reference to the filesystem's request_queue(s). But it was overlooked that the request_queue reference is *not* guaranteed to pin the corresponding blk_crypto_profile too; for device-mapper devices that support inline crypto, it doesn't. This can cause a use-after-free. - When the last inode that was using an incompletely-removed master key is evicted, the master key removal is completed by removing the key struct from the keyring. Currently this is done via key_invalidate(). Yet, key_invalidate() takes the key semaphore. This can deadlock when called from the shrinker, since in fscrypt_ioctl_add_key(), memory is allocated with GFP_KERNEL under the same semaphore. - More generally, the fact that the keyrings subsystem can arbitrarily delay the destruction of keys (via garbage collection delay, or via random processes getting temporary key references) is undesirable, as it means we can't strictly guarantee that all secrets are ever wiped. - Doing the master key lookups via the keyrings subsystem results in the key_permission LSM hook being called. fscrypt doesn't want this, as all access control for encrypted files is designed to happen via the files themselves, like any other files. The workaround which SELinux users are using is to change their SELinux policy to grant key search access to all domains. This works, but it is an odd extra step that shouldn't really have to be done. The fix for all these issues is to change the implementation to what I should have done originally: don't use the keyrings subsystem to keep track of the filesystem's fscrypt_master_key structs. Instead, just store them in a regular kernel data structure, and rework the reference counting, locking, and lifetime accordingly. Retain support for RCU-mode key lookups by using a hash table. Replace fscrypt_sb_free() with fscrypt_sb_delete(), which releases the keys synchronously and runs a bit earlier during unmount, so that block devices are still available. A side effect of this patch is that neither the master keys themselves nor the filesystem keyrings will be listed in /proc/keys anymore. ("Master key users" and the master key users keyrings will still be listed.) However, this was mostly an implementation detail, and it was intended just for debugging purposes. I don't know of anyone using it. This patch does *not* change how "master key users" (->mk_users) works; that still uses the keyrings subsystem. That is still needed for key quotas, and changing that isn't necessary to solve the issues listed above. If we decide to change that too, it would be a separate patch. I've marked this as fixing the original commit that added the fscrypt keyring, but as noted above the most important issue that this patch fixes wasn't introduced until the addition of inline encryption support. Fixes: 22d94f493bfb ("fscrypt: add FS_IOC_ADD_ENCRYPTION_KEY ioctl") Signed-off-by: Eric Biggers <ebiggers@google.com> Link: https://lore.kernel.org/r/20220901193208.138056-2-ebiggers@kernel.org
2022-09-02 03:32:06 +08:00
down_read(&mk->mk_sem);
fscrypt: require that key be added when setting a v2 encryption policy By looking up the master keys in a filesystem-level keyring rather than in the calling processes' key hierarchy, it becomes possible for a user to set an encryption policy which refers to some key they don't actually know, then encrypt their files using that key. Cryptographically this isn't much of a problem, but the semantics of this would be a bit weird. Thus, enforce that a v2 encryption policy can only be set if the user has previously added the key, or has capable(CAP_FOWNER). We tolerate that this problem will continue to exist for v1 encryption policies, however; there is no way around that. Reviewed-by: Theodore Ts'o <tytso@mit.edu> Signed-off-by: Eric Biggers <ebiggers@google.com>
2019-08-05 10:35:48 +08:00
mk_user = find_master_key_user(mk);
if (IS_ERR(mk_user)) {
err = PTR_ERR(mk_user);
} else {
key_put(mk_user);
err = 0;
}
fscrypt: stop using keyrings subsystem for fscrypt_master_key The approach of fs/crypto/ internally managing the fscrypt_master_key structs as the payloads of "struct key" objects contained in a "struct key" keyring has outlived its usefulness. The original idea was to simplify the code by reusing code from the keyrings subsystem. However, several issues have arisen that can't easily be resolved: - When a master key struct is destroyed, blk_crypto_evict_key() must be called on any per-mode keys embedded in it. (This started being the case when inline encryption support was added.) Yet, the keyrings subsystem can arbitrarily delay the destruction of keys, even past the time the filesystem was unmounted. Therefore, currently there is no easy way to call blk_crypto_evict_key() when a master key is destroyed. Currently, this is worked around by holding an extra reference to the filesystem's request_queue(s). But it was overlooked that the request_queue reference is *not* guaranteed to pin the corresponding blk_crypto_profile too; for device-mapper devices that support inline crypto, it doesn't. This can cause a use-after-free. - When the last inode that was using an incompletely-removed master key is evicted, the master key removal is completed by removing the key struct from the keyring. Currently this is done via key_invalidate(). Yet, key_invalidate() takes the key semaphore. This can deadlock when called from the shrinker, since in fscrypt_ioctl_add_key(), memory is allocated with GFP_KERNEL under the same semaphore. - More generally, the fact that the keyrings subsystem can arbitrarily delay the destruction of keys (via garbage collection delay, or via random processes getting temporary key references) is undesirable, as it means we can't strictly guarantee that all secrets are ever wiped. - Doing the master key lookups via the keyrings subsystem results in the key_permission LSM hook being called. fscrypt doesn't want this, as all access control for encrypted files is designed to happen via the files themselves, like any other files. The workaround which SELinux users are using is to change their SELinux policy to grant key search access to all domains. This works, but it is an odd extra step that shouldn't really have to be done. The fix for all these issues is to change the implementation to what I should have done originally: don't use the keyrings subsystem to keep track of the filesystem's fscrypt_master_key structs. Instead, just store them in a regular kernel data structure, and rework the reference counting, locking, and lifetime accordingly. Retain support for RCU-mode key lookups by using a hash table. Replace fscrypt_sb_free() with fscrypt_sb_delete(), which releases the keys synchronously and runs a bit earlier during unmount, so that block devices are still available. A side effect of this patch is that neither the master keys themselves nor the filesystem keyrings will be listed in /proc/keys anymore. ("Master key users" and the master key users keyrings will still be listed.) However, this was mostly an implementation detail, and it was intended just for debugging purposes. I don't know of anyone using it. This patch does *not* change how "master key users" (->mk_users) works; that still uses the keyrings subsystem. That is still needed for key quotas, and changing that isn't necessary to solve the issues listed above. If we decide to change that too, it would be a separate patch. I've marked this as fixing the original commit that added the fscrypt keyring, but as noted above the most important issue that this patch fixes wasn't introduced until the addition of inline encryption support. Fixes: 22d94f493bfb ("fscrypt: add FS_IOC_ADD_ENCRYPTION_KEY ioctl") Signed-off-by: Eric Biggers <ebiggers@google.com> Link: https://lore.kernel.org/r/20220901193208.138056-2-ebiggers@kernel.org
2022-09-02 03:32:06 +08:00
up_read(&mk->mk_sem);
fscrypt_put_master_key(mk);
fscrypt: require that key be added when setting a v2 encryption policy By looking up the master keys in a filesystem-level keyring rather than in the calling processes' key hierarchy, it becomes possible for a user to set an encryption policy which refers to some key they don't actually know, then encrypt their files using that key. Cryptographically this isn't much of a problem, but the semantics of this would be a bit weird. Thus, enforce that a v2 encryption policy can only be set if the user has previously added the key, or has capable(CAP_FOWNER). We tolerate that this problem will continue to exist for v1 encryption policies, however; there is no way around that. Reviewed-by: Theodore Ts'o <tytso@mit.edu> Signed-off-by: Eric Biggers <ebiggers@google.com>
2019-08-05 10:35:48 +08:00
out:
if (err == -ENOKEY && capable(CAP_FOWNER))
err = 0;
return err;
}
fscrypt: add FS_IOC_REMOVE_ENCRYPTION_KEY ioctl Add a new fscrypt ioctl, FS_IOC_REMOVE_ENCRYPTION_KEY. This ioctl removes an encryption key that was added by FS_IOC_ADD_ENCRYPTION_KEY. It wipes the secret key itself, then "locks" the encrypted files and directories that had been unlocked using that key -- implemented by evicting the relevant dentries and inodes from the VFS caches. The problem this solves is that many fscrypt users want the ability to remove encryption keys, causing the corresponding encrypted directories to appear "locked" (presented in ciphertext form) again. Moreover, users want removing an encryption key to *really* remove it, in the sense that the removed keys cannot be recovered even if kernel memory is compromised, e.g. by the exploit of a kernel security vulnerability or by a physical attack. This is desirable after a user logs out of the system, for example. In many cases users even already assume this to be the case and are surprised to hear when it's not. It is not sufficient to simply unlink the master key from the keyring (or to revoke or invalidate it), since the actual encryption transform objects are still pinned in memory by their inodes. Therefore, to really remove a key we must also evict the relevant inodes. Currently one workaround is to run 'sync && echo 2 > /proc/sys/vm/drop_caches'. But, that evicts all unused inodes in the system rather than just the inodes associated with the key being removed, causing severe performance problems. Moreover, it requires root privileges, so regular users can't "lock" their encrypted files. Another workaround, used in Chromium OS kernels, is to add a new VFS-level ioctl FS_IOC_DROP_CACHE which is a more restricted version of drop_caches that operates on a single super_block. It does: shrink_dcache_sb(sb); invalidate_inodes(sb, false); But it's still a hack. Yet, the major users of filesystem encryption want this feature badly enough that they are actually using these hacks. To properly solve the problem, start maintaining a list of the inodes which have been "unlocked" using each master key. Originally this wasn't possible because the kernel didn't keep track of in-use master keys at all. But, with the ->s_master_keys keyring it is now possible. Then, add an ioctl FS_IOC_REMOVE_ENCRYPTION_KEY. It finds the specified master key in ->s_master_keys, then wipes the secret key itself, which prevents any additional inodes from being unlocked with the key. Then, it syncs the filesystem and evicts the inodes in the key's list. The normal inode eviction code will free and wipe the per-file keys (in ->i_crypt_info). Note that freeing ->i_crypt_info without evicting the inodes was also considered, but would have been racy. Some inodes may still be in use when a master key is removed, and we can't simply revoke random file descriptors, mmap's, etc. Thus, the ioctl simply skips in-use inodes, and returns -EBUSY to indicate that some inodes weren't evicted. The master key *secret* is still removed, but the fscrypt_master_key struct remains to keep track of the remaining inodes. Userspace can then retry the ioctl to evict the remaining inodes. Alternatively, if userspace adds the key again, the refreshed secret will be associated with the existing list of inodes so they remain correctly tracked for future key removals. The ioctl doesn't wipe pagecache pages. Thus, we tolerate that after a kernel compromise some portions of plaintext file contents may still be recoverable from memory. This can be solved by enabling page poisoning system-wide, which security conscious users may choose to do. But it's very difficult to solve otherwise, e.g. note that plaintext file contents may have been read in other places than pagecache pages. Like FS_IOC_ADD_ENCRYPTION_KEY, FS_IOC_REMOVE_ENCRYPTION_KEY is initially restricted to privileged users only. This is sufficient for some use cases, but not all. A later patch will relax this restriction, but it will require introducing key hashes, among other changes. Reviewed-by: Theodore Ts'o <tytso@mit.edu> Signed-off-by: Eric Biggers <ebiggers@google.com>
2019-08-05 10:35:46 +08:00
/*
* Try to evict the inode's dentries from the dentry cache. If the inode is a
* directory, then it can have at most one dentry; however, that dentry may be
* pinned by child dentries, so first try to evict the children too.
*/
static void shrink_dcache_inode(struct inode *inode)
{
struct dentry *dentry;
if (S_ISDIR(inode->i_mode)) {
dentry = d_find_any_alias(inode);
if (dentry) {
shrink_dcache_parent(dentry);
dput(dentry);
}
}
d_prune_aliases(inode);
}
static void evict_dentries_for_decrypted_inodes(struct fscrypt_master_key *mk)
{
struct fscrypt_info *ci;
struct inode *inode;
struct inode *toput_inode = NULL;
spin_lock(&mk->mk_decrypted_inodes_lock);
list_for_each_entry(ci, &mk->mk_decrypted_inodes, ci_master_key_link) {
inode = ci->ci_inode;
spin_lock(&inode->i_lock);
if (inode->i_state & (I_FREEING | I_WILL_FREE | I_NEW)) {
spin_unlock(&inode->i_lock);
continue;
}
__iget(inode);
spin_unlock(&inode->i_lock);
spin_unlock(&mk->mk_decrypted_inodes_lock);
shrink_dcache_inode(inode);
iput(toput_inode);
toput_inode = inode;
spin_lock(&mk->mk_decrypted_inodes_lock);
}
spin_unlock(&mk->mk_decrypted_inodes_lock);
iput(toput_inode);
}
static int check_for_busy_inodes(struct super_block *sb,
struct fscrypt_master_key *mk)
{
struct list_head *pos;
size_t busy_count = 0;
unsigned long ino;
fscrypt: adjust logging for in-creation inodes Now that a fscrypt_info may be set up for inodes that are currently being created and haven't yet had an inode number assigned, avoid logging confusing messages about "inode 0". Acked-by: Jeff Layton <jlayton@kernel.org> Link: https://lore.kernel.org/r/20200917041136.178600-7-ebiggers@kernel.org Signed-off-by: Eric Biggers <ebiggers@google.com>
2020-09-17 12:11:29 +08:00
char ino_str[50] = "";
fscrypt: add FS_IOC_REMOVE_ENCRYPTION_KEY ioctl Add a new fscrypt ioctl, FS_IOC_REMOVE_ENCRYPTION_KEY. This ioctl removes an encryption key that was added by FS_IOC_ADD_ENCRYPTION_KEY. It wipes the secret key itself, then "locks" the encrypted files and directories that had been unlocked using that key -- implemented by evicting the relevant dentries and inodes from the VFS caches. The problem this solves is that many fscrypt users want the ability to remove encryption keys, causing the corresponding encrypted directories to appear "locked" (presented in ciphertext form) again. Moreover, users want removing an encryption key to *really* remove it, in the sense that the removed keys cannot be recovered even if kernel memory is compromised, e.g. by the exploit of a kernel security vulnerability or by a physical attack. This is desirable after a user logs out of the system, for example. In many cases users even already assume this to be the case and are surprised to hear when it's not. It is not sufficient to simply unlink the master key from the keyring (or to revoke or invalidate it), since the actual encryption transform objects are still pinned in memory by their inodes. Therefore, to really remove a key we must also evict the relevant inodes. Currently one workaround is to run 'sync && echo 2 > /proc/sys/vm/drop_caches'. But, that evicts all unused inodes in the system rather than just the inodes associated with the key being removed, causing severe performance problems. Moreover, it requires root privileges, so regular users can't "lock" their encrypted files. Another workaround, used in Chromium OS kernels, is to add a new VFS-level ioctl FS_IOC_DROP_CACHE which is a more restricted version of drop_caches that operates on a single super_block. It does: shrink_dcache_sb(sb); invalidate_inodes(sb, false); But it's still a hack. Yet, the major users of filesystem encryption want this feature badly enough that they are actually using these hacks. To properly solve the problem, start maintaining a list of the inodes which have been "unlocked" using each master key. Originally this wasn't possible because the kernel didn't keep track of in-use master keys at all. But, with the ->s_master_keys keyring it is now possible. Then, add an ioctl FS_IOC_REMOVE_ENCRYPTION_KEY. It finds the specified master key in ->s_master_keys, then wipes the secret key itself, which prevents any additional inodes from being unlocked with the key. Then, it syncs the filesystem and evicts the inodes in the key's list. The normal inode eviction code will free and wipe the per-file keys (in ->i_crypt_info). Note that freeing ->i_crypt_info without evicting the inodes was also considered, but would have been racy. Some inodes may still be in use when a master key is removed, and we can't simply revoke random file descriptors, mmap's, etc. Thus, the ioctl simply skips in-use inodes, and returns -EBUSY to indicate that some inodes weren't evicted. The master key *secret* is still removed, but the fscrypt_master_key struct remains to keep track of the remaining inodes. Userspace can then retry the ioctl to evict the remaining inodes. Alternatively, if userspace adds the key again, the refreshed secret will be associated with the existing list of inodes so they remain correctly tracked for future key removals. The ioctl doesn't wipe pagecache pages. Thus, we tolerate that after a kernel compromise some portions of plaintext file contents may still be recoverable from memory. This can be solved by enabling page poisoning system-wide, which security conscious users may choose to do. But it's very difficult to solve otherwise, e.g. note that plaintext file contents may have been read in other places than pagecache pages. Like FS_IOC_ADD_ENCRYPTION_KEY, FS_IOC_REMOVE_ENCRYPTION_KEY is initially restricted to privileged users only. This is sufficient for some use cases, but not all. A later patch will relax this restriction, but it will require introducing key hashes, among other changes. Reviewed-by: Theodore Ts'o <tytso@mit.edu> Signed-off-by: Eric Biggers <ebiggers@google.com>
2019-08-05 10:35:46 +08:00
spin_lock(&mk->mk_decrypted_inodes_lock);
list_for_each(pos, &mk->mk_decrypted_inodes)
busy_count++;
if (busy_count == 0) {
spin_unlock(&mk->mk_decrypted_inodes_lock);
return 0;
}
{
/* select an example file to show for debugging purposes */
struct inode *inode =
list_first_entry(&mk->mk_decrypted_inodes,
struct fscrypt_info,
ci_master_key_link)->ci_inode;
ino = inode->i_ino;
}
spin_unlock(&mk->mk_decrypted_inodes_lock);
fscrypt: adjust logging for in-creation inodes Now that a fscrypt_info may be set up for inodes that are currently being created and haven't yet had an inode number assigned, avoid logging confusing messages about "inode 0". Acked-by: Jeff Layton <jlayton@kernel.org> Link: https://lore.kernel.org/r/20200917041136.178600-7-ebiggers@kernel.org Signed-off-by: Eric Biggers <ebiggers@google.com>
2020-09-17 12:11:29 +08:00
/* If the inode is currently being created, ino may still be 0. */
if (ino)
snprintf(ino_str, sizeof(ino_str), ", including ino %lu", ino);
fscrypt: add FS_IOC_REMOVE_ENCRYPTION_KEY ioctl Add a new fscrypt ioctl, FS_IOC_REMOVE_ENCRYPTION_KEY. This ioctl removes an encryption key that was added by FS_IOC_ADD_ENCRYPTION_KEY. It wipes the secret key itself, then "locks" the encrypted files and directories that had been unlocked using that key -- implemented by evicting the relevant dentries and inodes from the VFS caches. The problem this solves is that many fscrypt users want the ability to remove encryption keys, causing the corresponding encrypted directories to appear "locked" (presented in ciphertext form) again. Moreover, users want removing an encryption key to *really* remove it, in the sense that the removed keys cannot be recovered even if kernel memory is compromised, e.g. by the exploit of a kernel security vulnerability or by a physical attack. This is desirable after a user logs out of the system, for example. In many cases users even already assume this to be the case and are surprised to hear when it's not. It is not sufficient to simply unlink the master key from the keyring (or to revoke or invalidate it), since the actual encryption transform objects are still pinned in memory by their inodes. Therefore, to really remove a key we must also evict the relevant inodes. Currently one workaround is to run 'sync && echo 2 > /proc/sys/vm/drop_caches'. But, that evicts all unused inodes in the system rather than just the inodes associated with the key being removed, causing severe performance problems. Moreover, it requires root privileges, so regular users can't "lock" their encrypted files. Another workaround, used in Chromium OS kernels, is to add a new VFS-level ioctl FS_IOC_DROP_CACHE which is a more restricted version of drop_caches that operates on a single super_block. It does: shrink_dcache_sb(sb); invalidate_inodes(sb, false); But it's still a hack. Yet, the major users of filesystem encryption want this feature badly enough that they are actually using these hacks. To properly solve the problem, start maintaining a list of the inodes which have been "unlocked" using each master key. Originally this wasn't possible because the kernel didn't keep track of in-use master keys at all. But, with the ->s_master_keys keyring it is now possible. Then, add an ioctl FS_IOC_REMOVE_ENCRYPTION_KEY. It finds the specified master key in ->s_master_keys, then wipes the secret key itself, which prevents any additional inodes from being unlocked with the key. Then, it syncs the filesystem and evicts the inodes in the key's list. The normal inode eviction code will free and wipe the per-file keys (in ->i_crypt_info). Note that freeing ->i_crypt_info without evicting the inodes was also considered, but would have been racy. Some inodes may still be in use when a master key is removed, and we can't simply revoke random file descriptors, mmap's, etc. Thus, the ioctl simply skips in-use inodes, and returns -EBUSY to indicate that some inodes weren't evicted. The master key *secret* is still removed, but the fscrypt_master_key struct remains to keep track of the remaining inodes. Userspace can then retry the ioctl to evict the remaining inodes. Alternatively, if userspace adds the key again, the refreshed secret will be associated with the existing list of inodes so they remain correctly tracked for future key removals. The ioctl doesn't wipe pagecache pages. Thus, we tolerate that after a kernel compromise some portions of plaintext file contents may still be recoverable from memory. This can be solved by enabling page poisoning system-wide, which security conscious users may choose to do. But it's very difficult to solve otherwise, e.g. note that plaintext file contents may have been read in other places than pagecache pages. Like FS_IOC_ADD_ENCRYPTION_KEY, FS_IOC_REMOVE_ENCRYPTION_KEY is initially restricted to privileged users only. This is sufficient for some use cases, but not all. A later patch will relax this restriction, but it will require introducing key hashes, among other changes. Reviewed-by: Theodore Ts'o <tytso@mit.edu> Signed-off-by: Eric Biggers <ebiggers@google.com>
2019-08-05 10:35:46 +08:00
fscrypt_warn(NULL,
fscrypt: adjust logging for in-creation inodes Now that a fscrypt_info may be set up for inodes that are currently being created and haven't yet had an inode number assigned, avoid logging confusing messages about "inode 0". Acked-by: Jeff Layton <jlayton@kernel.org> Link: https://lore.kernel.org/r/20200917041136.178600-7-ebiggers@kernel.org Signed-off-by: Eric Biggers <ebiggers@google.com>
2020-09-17 12:11:29 +08:00
"%s: %zu inode(s) still busy after removing key with %s %*phN%s",
fscrypt: add FS_IOC_REMOVE_ENCRYPTION_KEY ioctl Add a new fscrypt ioctl, FS_IOC_REMOVE_ENCRYPTION_KEY. This ioctl removes an encryption key that was added by FS_IOC_ADD_ENCRYPTION_KEY. It wipes the secret key itself, then "locks" the encrypted files and directories that had been unlocked using that key -- implemented by evicting the relevant dentries and inodes from the VFS caches. The problem this solves is that many fscrypt users want the ability to remove encryption keys, causing the corresponding encrypted directories to appear "locked" (presented in ciphertext form) again. Moreover, users want removing an encryption key to *really* remove it, in the sense that the removed keys cannot be recovered even if kernel memory is compromised, e.g. by the exploit of a kernel security vulnerability or by a physical attack. This is desirable after a user logs out of the system, for example. In many cases users even already assume this to be the case and are surprised to hear when it's not. It is not sufficient to simply unlink the master key from the keyring (or to revoke or invalidate it), since the actual encryption transform objects are still pinned in memory by their inodes. Therefore, to really remove a key we must also evict the relevant inodes. Currently one workaround is to run 'sync && echo 2 > /proc/sys/vm/drop_caches'. But, that evicts all unused inodes in the system rather than just the inodes associated with the key being removed, causing severe performance problems. Moreover, it requires root privileges, so regular users can't "lock" their encrypted files. Another workaround, used in Chromium OS kernels, is to add a new VFS-level ioctl FS_IOC_DROP_CACHE which is a more restricted version of drop_caches that operates on a single super_block. It does: shrink_dcache_sb(sb); invalidate_inodes(sb, false); But it's still a hack. Yet, the major users of filesystem encryption want this feature badly enough that they are actually using these hacks. To properly solve the problem, start maintaining a list of the inodes which have been "unlocked" using each master key. Originally this wasn't possible because the kernel didn't keep track of in-use master keys at all. But, with the ->s_master_keys keyring it is now possible. Then, add an ioctl FS_IOC_REMOVE_ENCRYPTION_KEY. It finds the specified master key in ->s_master_keys, then wipes the secret key itself, which prevents any additional inodes from being unlocked with the key. Then, it syncs the filesystem and evicts the inodes in the key's list. The normal inode eviction code will free and wipe the per-file keys (in ->i_crypt_info). Note that freeing ->i_crypt_info without evicting the inodes was also considered, but would have been racy. Some inodes may still be in use when a master key is removed, and we can't simply revoke random file descriptors, mmap's, etc. Thus, the ioctl simply skips in-use inodes, and returns -EBUSY to indicate that some inodes weren't evicted. The master key *secret* is still removed, but the fscrypt_master_key struct remains to keep track of the remaining inodes. Userspace can then retry the ioctl to evict the remaining inodes. Alternatively, if userspace adds the key again, the refreshed secret will be associated with the existing list of inodes so they remain correctly tracked for future key removals. The ioctl doesn't wipe pagecache pages. Thus, we tolerate that after a kernel compromise some portions of plaintext file contents may still be recoverable from memory. This can be solved by enabling page poisoning system-wide, which security conscious users may choose to do. But it's very difficult to solve otherwise, e.g. note that plaintext file contents may have been read in other places than pagecache pages. Like FS_IOC_ADD_ENCRYPTION_KEY, FS_IOC_REMOVE_ENCRYPTION_KEY is initially restricted to privileged users only. This is sufficient for some use cases, but not all. A later patch will relax this restriction, but it will require introducing key hashes, among other changes. Reviewed-by: Theodore Ts'o <tytso@mit.edu> Signed-off-by: Eric Biggers <ebiggers@google.com>
2019-08-05 10:35:46 +08:00
sb->s_id, busy_count, master_key_spec_type(&mk->mk_spec),
master_key_spec_len(&mk->mk_spec), (u8 *)&mk->mk_spec.u,
fscrypt: adjust logging for in-creation inodes Now that a fscrypt_info may be set up for inodes that are currently being created and haven't yet had an inode number assigned, avoid logging confusing messages about "inode 0". Acked-by: Jeff Layton <jlayton@kernel.org> Link: https://lore.kernel.org/r/20200917041136.178600-7-ebiggers@kernel.org Signed-off-by: Eric Biggers <ebiggers@google.com>
2020-09-17 12:11:29 +08:00
ino_str);
fscrypt: add FS_IOC_REMOVE_ENCRYPTION_KEY ioctl Add a new fscrypt ioctl, FS_IOC_REMOVE_ENCRYPTION_KEY. This ioctl removes an encryption key that was added by FS_IOC_ADD_ENCRYPTION_KEY. It wipes the secret key itself, then "locks" the encrypted files and directories that had been unlocked using that key -- implemented by evicting the relevant dentries and inodes from the VFS caches. The problem this solves is that many fscrypt users want the ability to remove encryption keys, causing the corresponding encrypted directories to appear "locked" (presented in ciphertext form) again. Moreover, users want removing an encryption key to *really* remove it, in the sense that the removed keys cannot be recovered even if kernel memory is compromised, e.g. by the exploit of a kernel security vulnerability or by a physical attack. This is desirable after a user logs out of the system, for example. In many cases users even already assume this to be the case and are surprised to hear when it's not. It is not sufficient to simply unlink the master key from the keyring (or to revoke or invalidate it), since the actual encryption transform objects are still pinned in memory by their inodes. Therefore, to really remove a key we must also evict the relevant inodes. Currently one workaround is to run 'sync && echo 2 > /proc/sys/vm/drop_caches'. But, that evicts all unused inodes in the system rather than just the inodes associated with the key being removed, causing severe performance problems. Moreover, it requires root privileges, so regular users can't "lock" their encrypted files. Another workaround, used in Chromium OS kernels, is to add a new VFS-level ioctl FS_IOC_DROP_CACHE which is a more restricted version of drop_caches that operates on a single super_block. It does: shrink_dcache_sb(sb); invalidate_inodes(sb, false); But it's still a hack. Yet, the major users of filesystem encryption want this feature badly enough that they are actually using these hacks. To properly solve the problem, start maintaining a list of the inodes which have been "unlocked" using each master key. Originally this wasn't possible because the kernel didn't keep track of in-use master keys at all. But, with the ->s_master_keys keyring it is now possible. Then, add an ioctl FS_IOC_REMOVE_ENCRYPTION_KEY. It finds the specified master key in ->s_master_keys, then wipes the secret key itself, which prevents any additional inodes from being unlocked with the key. Then, it syncs the filesystem and evicts the inodes in the key's list. The normal inode eviction code will free and wipe the per-file keys (in ->i_crypt_info). Note that freeing ->i_crypt_info without evicting the inodes was also considered, but would have been racy. Some inodes may still be in use when a master key is removed, and we can't simply revoke random file descriptors, mmap's, etc. Thus, the ioctl simply skips in-use inodes, and returns -EBUSY to indicate that some inodes weren't evicted. The master key *secret* is still removed, but the fscrypt_master_key struct remains to keep track of the remaining inodes. Userspace can then retry the ioctl to evict the remaining inodes. Alternatively, if userspace adds the key again, the refreshed secret will be associated with the existing list of inodes so they remain correctly tracked for future key removals. The ioctl doesn't wipe pagecache pages. Thus, we tolerate that after a kernel compromise some portions of plaintext file contents may still be recoverable from memory. This can be solved by enabling page poisoning system-wide, which security conscious users may choose to do. But it's very difficult to solve otherwise, e.g. note that plaintext file contents may have been read in other places than pagecache pages. Like FS_IOC_ADD_ENCRYPTION_KEY, FS_IOC_REMOVE_ENCRYPTION_KEY is initially restricted to privileged users only. This is sufficient for some use cases, but not all. A later patch will relax this restriction, but it will require introducing key hashes, among other changes. Reviewed-by: Theodore Ts'o <tytso@mit.edu> Signed-off-by: Eric Biggers <ebiggers@google.com>
2019-08-05 10:35:46 +08:00
return -EBUSY;
}
static int try_to_lock_encrypted_files(struct super_block *sb,
struct fscrypt_master_key *mk)
{
int err1;
int err2;
/*
* An inode can't be evicted while it is dirty or has dirty pages.
* Thus, we first have to clean the inodes in ->mk_decrypted_inodes.
*
* Just do it the easy way: call sync_filesystem(). It's overkill, but
* it works, and it's more important to minimize the amount of caches we
* drop than the amount of data we sync. Also, unprivileged users can
* already call sync_filesystem() via sys_syncfs() or sys_sync().
*/
down_read(&sb->s_umount);
err1 = sync_filesystem(sb);
up_read(&sb->s_umount);
/* If a sync error occurs, still try to evict as much as possible. */
/*
* Inodes are pinned by their dentries, so we have to evict their
* dentries. shrink_dcache_sb() would suffice, but would be overkill
* and inappropriate for use by unprivileged users. So instead go
* through the inodes' alias lists and try to evict each dentry.
*/
evict_dentries_for_decrypted_inodes(mk);
/*
* evict_dentries_for_decrypted_inodes() already iput() each inode in
* the list; any inodes for which that dropped the last reference will
* have been evicted due to fscrypt_drop_inode() detecting the key
* removal and telling the VFS to evict the inode. So to finish, we
* just need to check whether any inodes couldn't be evicted.
*/
err2 = check_for_busy_inodes(sb, mk);
return err1 ?: err2;
}
/*
* Try to remove an fscrypt master encryption key.
*
fscrypt: add FS_IOC_REMOVE_ENCRYPTION_KEY_ALL_USERS ioctl Add a root-only variant of the FS_IOC_REMOVE_ENCRYPTION_KEY ioctl which removes all users' claims of the key, not just the current user's claim. I.e., it always removes the key itself, no matter how many users have added it. This is useful for forcing a directory to be locked, without having to figure out which user ID(s) the key was added under. This is planned to be used by a command like 'sudo fscrypt lock DIR --all-users' in the fscrypt userspace tool (http://github.com/google/fscrypt). Reviewed-by: Theodore Ts'o <tytso@mit.edu> Signed-off-by: Eric Biggers <ebiggers@google.com>
2019-08-05 10:35:47 +08:00
* FS_IOC_REMOVE_ENCRYPTION_KEY (all_users=false) removes the current user's
* claim to the key, then removes the key itself if no other users have claims.
* FS_IOC_REMOVE_ENCRYPTION_KEY_ALL_USERS (all_users=true) always removes the
* key itself.
fscrypt: allow unprivileged users to add/remove keys for v2 policies Allow the FS_IOC_ADD_ENCRYPTION_KEY and FS_IOC_REMOVE_ENCRYPTION_KEY ioctls to be used by non-root users to add and remove encryption keys from the filesystem-level crypto keyrings, subject to limitations. Motivation: while privileged fscrypt key management is sufficient for some users (e.g. Android and Chromium OS, where a privileged process manages all keys), the old API by design also allows non-root users to set up and use encrypted directories, and we don't want to regress on that. Especially, we don't want to force users to continue using the old API, running into the visibility mismatch between files and keyrings and being unable to "lock" encrypted directories. Intuitively, the ioctls have to be privileged since they manipulate filesystem-level state. However, it's actually safe to make them unprivileged if we very carefully enforce some specific limitations. First, each key must be identified by a cryptographic hash so that a user can't add the wrong key for another user's files. For v2 encryption policies, we use the key_identifier for this. v1 policies don't have this, so managing keys for them remains privileged. Second, each key a user adds is charged to their quota for the keyrings service. Thus, a user can't exhaust memory by adding a huge number of keys. By default each non-root user is allowed up to 200 keys; this can be changed using the existing sysctl 'kernel.keys.maxkeys'. Third, if multiple users add the same key, we keep track of those users of the key (of which there remains a single copy), and won't really remove the key, i.e. "lock" the encrypted files, until all those users have removed it. This prevents denial of service attacks that would be possible under simpler schemes, such allowing the first user who added a key to remove it -- since that could be a malicious user who has compromised the key. Of course, encryption keys should be kept secret, but the idea is that using encryption should never be *less* secure than not using encryption, even if your key was compromised. We tolerate that a user will be unable to really remove a key, i.e. unable to "lock" their encrypted files, if another user has added the same key. But in a sense, this is actually a good thing because it will avoid providing a false notion of security where a key appears to have been removed when actually it's still in memory, available to any attacker who compromises the operating system kernel. Reviewed-by: Theodore Ts'o <tytso@mit.edu> Signed-off-by: Eric Biggers <ebiggers@google.com>
2019-08-05 10:35:47 +08:00
*
* To "remove the key itself", first we wipe the actual master key secret, so
* that no more inodes can be unlocked with it. Then we try to evict all cached
* inodes that had been unlocked with the key.
fscrypt: add FS_IOC_REMOVE_ENCRYPTION_KEY ioctl Add a new fscrypt ioctl, FS_IOC_REMOVE_ENCRYPTION_KEY. This ioctl removes an encryption key that was added by FS_IOC_ADD_ENCRYPTION_KEY. It wipes the secret key itself, then "locks" the encrypted files and directories that had been unlocked using that key -- implemented by evicting the relevant dentries and inodes from the VFS caches. The problem this solves is that many fscrypt users want the ability to remove encryption keys, causing the corresponding encrypted directories to appear "locked" (presented in ciphertext form) again. Moreover, users want removing an encryption key to *really* remove it, in the sense that the removed keys cannot be recovered even if kernel memory is compromised, e.g. by the exploit of a kernel security vulnerability or by a physical attack. This is desirable after a user logs out of the system, for example. In many cases users even already assume this to be the case and are surprised to hear when it's not. It is not sufficient to simply unlink the master key from the keyring (or to revoke or invalidate it), since the actual encryption transform objects are still pinned in memory by their inodes. Therefore, to really remove a key we must also evict the relevant inodes. Currently one workaround is to run 'sync && echo 2 > /proc/sys/vm/drop_caches'. But, that evicts all unused inodes in the system rather than just the inodes associated with the key being removed, causing severe performance problems. Moreover, it requires root privileges, so regular users can't "lock" their encrypted files. Another workaround, used in Chromium OS kernels, is to add a new VFS-level ioctl FS_IOC_DROP_CACHE which is a more restricted version of drop_caches that operates on a single super_block. It does: shrink_dcache_sb(sb); invalidate_inodes(sb, false); But it's still a hack. Yet, the major users of filesystem encryption want this feature badly enough that they are actually using these hacks. To properly solve the problem, start maintaining a list of the inodes which have been "unlocked" using each master key. Originally this wasn't possible because the kernel didn't keep track of in-use master keys at all. But, with the ->s_master_keys keyring it is now possible. Then, add an ioctl FS_IOC_REMOVE_ENCRYPTION_KEY. It finds the specified master key in ->s_master_keys, then wipes the secret key itself, which prevents any additional inodes from being unlocked with the key. Then, it syncs the filesystem and evicts the inodes in the key's list. The normal inode eviction code will free and wipe the per-file keys (in ->i_crypt_info). Note that freeing ->i_crypt_info without evicting the inodes was also considered, but would have been racy. Some inodes may still be in use when a master key is removed, and we can't simply revoke random file descriptors, mmap's, etc. Thus, the ioctl simply skips in-use inodes, and returns -EBUSY to indicate that some inodes weren't evicted. The master key *secret* is still removed, but the fscrypt_master_key struct remains to keep track of the remaining inodes. Userspace can then retry the ioctl to evict the remaining inodes. Alternatively, if userspace adds the key again, the refreshed secret will be associated with the existing list of inodes so they remain correctly tracked for future key removals. The ioctl doesn't wipe pagecache pages. Thus, we tolerate that after a kernel compromise some portions of plaintext file contents may still be recoverable from memory. This can be solved by enabling page poisoning system-wide, which security conscious users may choose to do. But it's very difficult to solve otherwise, e.g. note that plaintext file contents may have been read in other places than pagecache pages. Like FS_IOC_ADD_ENCRYPTION_KEY, FS_IOC_REMOVE_ENCRYPTION_KEY is initially restricted to privileged users only. This is sufficient for some use cases, but not all. A later patch will relax this restriction, but it will require introducing key hashes, among other changes. Reviewed-by: Theodore Ts'o <tytso@mit.edu> Signed-off-by: Eric Biggers <ebiggers@google.com>
2019-08-05 10:35:46 +08:00
*
* If all inodes were evicted, then we unlink the fscrypt_master_key from the
* keyring. Otherwise it remains in the keyring in the "incompletely removed"
* state (without the actual secret key) where it tracks the list of remaining
* inodes. Userspace can execute the ioctl again later to retry eviction, or
* alternatively can re-add the secret key again.
*
* For more details, see the "Removing keys" section of
* Documentation/filesystems/fscrypt.rst.
*/
fscrypt: add FS_IOC_REMOVE_ENCRYPTION_KEY_ALL_USERS ioctl Add a root-only variant of the FS_IOC_REMOVE_ENCRYPTION_KEY ioctl which removes all users' claims of the key, not just the current user's claim. I.e., it always removes the key itself, no matter how many users have added it. This is useful for forcing a directory to be locked, without having to figure out which user ID(s) the key was added under. This is planned to be used by a command like 'sudo fscrypt lock DIR --all-users' in the fscrypt userspace tool (http://github.com/google/fscrypt). Reviewed-by: Theodore Ts'o <tytso@mit.edu> Signed-off-by: Eric Biggers <ebiggers@google.com>
2019-08-05 10:35:47 +08:00
static int do_remove_key(struct file *filp, void __user *_uarg, bool all_users)
fscrypt: add FS_IOC_REMOVE_ENCRYPTION_KEY ioctl Add a new fscrypt ioctl, FS_IOC_REMOVE_ENCRYPTION_KEY. This ioctl removes an encryption key that was added by FS_IOC_ADD_ENCRYPTION_KEY. It wipes the secret key itself, then "locks" the encrypted files and directories that had been unlocked using that key -- implemented by evicting the relevant dentries and inodes from the VFS caches. The problem this solves is that many fscrypt users want the ability to remove encryption keys, causing the corresponding encrypted directories to appear "locked" (presented in ciphertext form) again. Moreover, users want removing an encryption key to *really* remove it, in the sense that the removed keys cannot be recovered even if kernel memory is compromised, e.g. by the exploit of a kernel security vulnerability or by a physical attack. This is desirable after a user logs out of the system, for example. In many cases users even already assume this to be the case and are surprised to hear when it's not. It is not sufficient to simply unlink the master key from the keyring (or to revoke or invalidate it), since the actual encryption transform objects are still pinned in memory by their inodes. Therefore, to really remove a key we must also evict the relevant inodes. Currently one workaround is to run 'sync && echo 2 > /proc/sys/vm/drop_caches'. But, that evicts all unused inodes in the system rather than just the inodes associated with the key being removed, causing severe performance problems. Moreover, it requires root privileges, so regular users can't "lock" their encrypted files. Another workaround, used in Chromium OS kernels, is to add a new VFS-level ioctl FS_IOC_DROP_CACHE which is a more restricted version of drop_caches that operates on a single super_block. It does: shrink_dcache_sb(sb); invalidate_inodes(sb, false); But it's still a hack. Yet, the major users of filesystem encryption want this feature badly enough that they are actually using these hacks. To properly solve the problem, start maintaining a list of the inodes which have been "unlocked" using each master key. Originally this wasn't possible because the kernel didn't keep track of in-use master keys at all. But, with the ->s_master_keys keyring it is now possible. Then, add an ioctl FS_IOC_REMOVE_ENCRYPTION_KEY. It finds the specified master key in ->s_master_keys, then wipes the secret key itself, which prevents any additional inodes from being unlocked with the key. Then, it syncs the filesystem and evicts the inodes in the key's list. The normal inode eviction code will free and wipe the per-file keys (in ->i_crypt_info). Note that freeing ->i_crypt_info without evicting the inodes was also considered, but would have been racy. Some inodes may still be in use when a master key is removed, and we can't simply revoke random file descriptors, mmap's, etc. Thus, the ioctl simply skips in-use inodes, and returns -EBUSY to indicate that some inodes weren't evicted. The master key *secret* is still removed, but the fscrypt_master_key struct remains to keep track of the remaining inodes. Userspace can then retry the ioctl to evict the remaining inodes. Alternatively, if userspace adds the key again, the refreshed secret will be associated with the existing list of inodes so they remain correctly tracked for future key removals. The ioctl doesn't wipe pagecache pages. Thus, we tolerate that after a kernel compromise some portions of plaintext file contents may still be recoverable from memory. This can be solved by enabling page poisoning system-wide, which security conscious users may choose to do. But it's very difficult to solve otherwise, e.g. note that plaintext file contents may have been read in other places than pagecache pages. Like FS_IOC_ADD_ENCRYPTION_KEY, FS_IOC_REMOVE_ENCRYPTION_KEY is initially restricted to privileged users only. This is sufficient for some use cases, but not all. A later patch will relax this restriction, but it will require introducing key hashes, among other changes. Reviewed-by: Theodore Ts'o <tytso@mit.edu> Signed-off-by: Eric Biggers <ebiggers@google.com>
2019-08-05 10:35:46 +08:00
{
struct super_block *sb = file_inode(filp)->i_sb;
struct fscrypt_remove_key_arg __user *uarg = _uarg;
struct fscrypt_remove_key_arg arg;
struct fscrypt_master_key *mk;
u32 status_flags = 0;
int err;
fscrypt: stop using keyrings subsystem for fscrypt_master_key The approach of fs/crypto/ internally managing the fscrypt_master_key structs as the payloads of "struct key" objects contained in a "struct key" keyring has outlived its usefulness. The original idea was to simplify the code by reusing code from the keyrings subsystem. However, several issues have arisen that can't easily be resolved: - When a master key struct is destroyed, blk_crypto_evict_key() must be called on any per-mode keys embedded in it. (This started being the case when inline encryption support was added.) Yet, the keyrings subsystem can arbitrarily delay the destruction of keys, even past the time the filesystem was unmounted. Therefore, currently there is no easy way to call blk_crypto_evict_key() when a master key is destroyed. Currently, this is worked around by holding an extra reference to the filesystem's request_queue(s). But it was overlooked that the request_queue reference is *not* guaranteed to pin the corresponding blk_crypto_profile too; for device-mapper devices that support inline crypto, it doesn't. This can cause a use-after-free. - When the last inode that was using an incompletely-removed master key is evicted, the master key removal is completed by removing the key struct from the keyring. Currently this is done via key_invalidate(). Yet, key_invalidate() takes the key semaphore. This can deadlock when called from the shrinker, since in fscrypt_ioctl_add_key(), memory is allocated with GFP_KERNEL under the same semaphore. - More generally, the fact that the keyrings subsystem can arbitrarily delay the destruction of keys (via garbage collection delay, or via random processes getting temporary key references) is undesirable, as it means we can't strictly guarantee that all secrets are ever wiped. - Doing the master key lookups via the keyrings subsystem results in the key_permission LSM hook being called. fscrypt doesn't want this, as all access control for encrypted files is designed to happen via the files themselves, like any other files. The workaround which SELinux users are using is to change their SELinux policy to grant key search access to all domains. This works, but it is an odd extra step that shouldn't really have to be done. The fix for all these issues is to change the implementation to what I should have done originally: don't use the keyrings subsystem to keep track of the filesystem's fscrypt_master_key structs. Instead, just store them in a regular kernel data structure, and rework the reference counting, locking, and lifetime accordingly. Retain support for RCU-mode key lookups by using a hash table. Replace fscrypt_sb_free() with fscrypt_sb_delete(), which releases the keys synchronously and runs a bit earlier during unmount, so that block devices are still available. A side effect of this patch is that neither the master keys themselves nor the filesystem keyrings will be listed in /proc/keys anymore. ("Master key users" and the master key users keyrings will still be listed.) However, this was mostly an implementation detail, and it was intended just for debugging purposes. I don't know of anyone using it. This patch does *not* change how "master key users" (->mk_users) works; that still uses the keyrings subsystem. That is still needed for key quotas, and changing that isn't necessary to solve the issues listed above. If we decide to change that too, it would be a separate patch. I've marked this as fixing the original commit that added the fscrypt keyring, but as noted above the most important issue that this patch fixes wasn't introduced until the addition of inline encryption support. Fixes: 22d94f493bfb ("fscrypt: add FS_IOC_ADD_ENCRYPTION_KEY ioctl") Signed-off-by: Eric Biggers <ebiggers@google.com> Link: https://lore.kernel.org/r/20220901193208.138056-2-ebiggers@kernel.org
2022-09-02 03:32:06 +08:00
bool inodes_remain;
fscrypt: add FS_IOC_REMOVE_ENCRYPTION_KEY ioctl Add a new fscrypt ioctl, FS_IOC_REMOVE_ENCRYPTION_KEY. This ioctl removes an encryption key that was added by FS_IOC_ADD_ENCRYPTION_KEY. It wipes the secret key itself, then "locks" the encrypted files and directories that had been unlocked using that key -- implemented by evicting the relevant dentries and inodes from the VFS caches. The problem this solves is that many fscrypt users want the ability to remove encryption keys, causing the corresponding encrypted directories to appear "locked" (presented in ciphertext form) again. Moreover, users want removing an encryption key to *really* remove it, in the sense that the removed keys cannot be recovered even if kernel memory is compromised, e.g. by the exploit of a kernel security vulnerability or by a physical attack. This is desirable after a user logs out of the system, for example. In many cases users even already assume this to be the case and are surprised to hear when it's not. It is not sufficient to simply unlink the master key from the keyring (or to revoke or invalidate it), since the actual encryption transform objects are still pinned in memory by their inodes. Therefore, to really remove a key we must also evict the relevant inodes. Currently one workaround is to run 'sync && echo 2 > /proc/sys/vm/drop_caches'. But, that evicts all unused inodes in the system rather than just the inodes associated with the key being removed, causing severe performance problems. Moreover, it requires root privileges, so regular users can't "lock" their encrypted files. Another workaround, used in Chromium OS kernels, is to add a new VFS-level ioctl FS_IOC_DROP_CACHE which is a more restricted version of drop_caches that operates on a single super_block. It does: shrink_dcache_sb(sb); invalidate_inodes(sb, false); But it's still a hack. Yet, the major users of filesystem encryption want this feature badly enough that they are actually using these hacks. To properly solve the problem, start maintaining a list of the inodes which have been "unlocked" using each master key. Originally this wasn't possible because the kernel didn't keep track of in-use master keys at all. But, with the ->s_master_keys keyring it is now possible. Then, add an ioctl FS_IOC_REMOVE_ENCRYPTION_KEY. It finds the specified master key in ->s_master_keys, then wipes the secret key itself, which prevents any additional inodes from being unlocked with the key. Then, it syncs the filesystem and evicts the inodes in the key's list. The normal inode eviction code will free and wipe the per-file keys (in ->i_crypt_info). Note that freeing ->i_crypt_info without evicting the inodes was also considered, but would have been racy. Some inodes may still be in use when a master key is removed, and we can't simply revoke random file descriptors, mmap's, etc. Thus, the ioctl simply skips in-use inodes, and returns -EBUSY to indicate that some inodes weren't evicted. The master key *secret* is still removed, but the fscrypt_master_key struct remains to keep track of the remaining inodes. Userspace can then retry the ioctl to evict the remaining inodes. Alternatively, if userspace adds the key again, the refreshed secret will be associated with the existing list of inodes so they remain correctly tracked for future key removals. The ioctl doesn't wipe pagecache pages. Thus, we tolerate that after a kernel compromise some portions of plaintext file contents may still be recoverable from memory. This can be solved by enabling page poisoning system-wide, which security conscious users may choose to do. But it's very difficult to solve otherwise, e.g. note that plaintext file contents may have been read in other places than pagecache pages. Like FS_IOC_ADD_ENCRYPTION_KEY, FS_IOC_REMOVE_ENCRYPTION_KEY is initially restricted to privileged users only. This is sufficient for some use cases, but not all. A later patch will relax this restriction, but it will require introducing key hashes, among other changes. Reviewed-by: Theodore Ts'o <tytso@mit.edu> Signed-off-by: Eric Biggers <ebiggers@google.com>
2019-08-05 10:35:46 +08:00
if (copy_from_user(&arg, uarg, sizeof(arg)))
return -EFAULT;
if (!valid_key_spec(&arg.key_spec))
return -EINVAL;
if (memchr_inv(arg.__reserved, 0, sizeof(arg.__reserved)))
return -EINVAL;
fscrypt: allow unprivileged users to add/remove keys for v2 policies Allow the FS_IOC_ADD_ENCRYPTION_KEY and FS_IOC_REMOVE_ENCRYPTION_KEY ioctls to be used by non-root users to add and remove encryption keys from the filesystem-level crypto keyrings, subject to limitations. Motivation: while privileged fscrypt key management is sufficient for some users (e.g. Android and Chromium OS, where a privileged process manages all keys), the old API by design also allows non-root users to set up and use encrypted directories, and we don't want to regress on that. Especially, we don't want to force users to continue using the old API, running into the visibility mismatch between files and keyrings and being unable to "lock" encrypted directories. Intuitively, the ioctls have to be privileged since they manipulate filesystem-level state. However, it's actually safe to make them unprivileged if we very carefully enforce some specific limitations. First, each key must be identified by a cryptographic hash so that a user can't add the wrong key for another user's files. For v2 encryption policies, we use the key_identifier for this. v1 policies don't have this, so managing keys for them remains privileged. Second, each key a user adds is charged to their quota for the keyrings service. Thus, a user can't exhaust memory by adding a huge number of keys. By default each non-root user is allowed up to 200 keys; this can be changed using the existing sysctl 'kernel.keys.maxkeys'. Third, if multiple users add the same key, we keep track of those users of the key (of which there remains a single copy), and won't really remove the key, i.e. "lock" the encrypted files, until all those users have removed it. This prevents denial of service attacks that would be possible under simpler schemes, such allowing the first user who added a key to remove it -- since that could be a malicious user who has compromised the key. Of course, encryption keys should be kept secret, but the idea is that using encryption should never be *less* secure than not using encryption, even if your key was compromised. We tolerate that a user will be unable to really remove a key, i.e. unable to "lock" their encrypted files, if another user has added the same key. But in a sense, this is actually a good thing because it will avoid providing a false notion of security where a key appears to have been removed when actually it's still in memory, available to any attacker who compromises the operating system kernel. Reviewed-by: Theodore Ts'o <tytso@mit.edu> Signed-off-by: Eric Biggers <ebiggers@google.com>
2019-08-05 10:35:47 +08:00
/*
* Only root can add and remove keys that are identified by an arbitrary
* descriptor rather than by a cryptographic hash.
*/
if (arg.key_spec.type == FSCRYPT_KEY_SPEC_TYPE_DESCRIPTOR &&
!capable(CAP_SYS_ADMIN))
fscrypt: add FS_IOC_REMOVE_ENCRYPTION_KEY ioctl Add a new fscrypt ioctl, FS_IOC_REMOVE_ENCRYPTION_KEY. This ioctl removes an encryption key that was added by FS_IOC_ADD_ENCRYPTION_KEY. It wipes the secret key itself, then "locks" the encrypted files and directories that had been unlocked using that key -- implemented by evicting the relevant dentries and inodes from the VFS caches. The problem this solves is that many fscrypt users want the ability to remove encryption keys, causing the corresponding encrypted directories to appear "locked" (presented in ciphertext form) again. Moreover, users want removing an encryption key to *really* remove it, in the sense that the removed keys cannot be recovered even if kernel memory is compromised, e.g. by the exploit of a kernel security vulnerability or by a physical attack. This is desirable after a user logs out of the system, for example. In many cases users even already assume this to be the case and are surprised to hear when it's not. It is not sufficient to simply unlink the master key from the keyring (or to revoke or invalidate it), since the actual encryption transform objects are still pinned in memory by their inodes. Therefore, to really remove a key we must also evict the relevant inodes. Currently one workaround is to run 'sync && echo 2 > /proc/sys/vm/drop_caches'. But, that evicts all unused inodes in the system rather than just the inodes associated with the key being removed, causing severe performance problems. Moreover, it requires root privileges, so regular users can't "lock" their encrypted files. Another workaround, used in Chromium OS kernels, is to add a new VFS-level ioctl FS_IOC_DROP_CACHE which is a more restricted version of drop_caches that operates on a single super_block. It does: shrink_dcache_sb(sb); invalidate_inodes(sb, false); But it's still a hack. Yet, the major users of filesystem encryption want this feature badly enough that they are actually using these hacks. To properly solve the problem, start maintaining a list of the inodes which have been "unlocked" using each master key. Originally this wasn't possible because the kernel didn't keep track of in-use master keys at all. But, with the ->s_master_keys keyring it is now possible. Then, add an ioctl FS_IOC_REMOVE_ENCRYPTION_KEY. It finds the specified master key in ->s_master_keys, then wipes the secret key itself, which prevents any additional inodes from being unlocked with the key. Then, it syncs the filesystem and evicts the inodes in the key's list. The normal inode eviction code will free and wipe the per-file keys (in ->i_crypt_info). Note that freeing ->i_crypt_info without evicting the inodes was also considered, but would have been racy. Some inodes may still be in use when a master key is removed, and we can't simply revoke random file descriptors, mmap's, etc. Thus, the ioctl simply skips in-use inodes, and returns -EBUSY to indicate that some inodes weren't evicted. The master key *secret* is still removed, but the fscrypt_master_key struct remains to keep track of the remaining inodes. Userspace can then retry the ioctl to evict the remaining inodes. Alternatively, if userspace adds the key again, the refreshed secret will be associated with the existing list of inodes so they remain correctly tracked for future key removals. The ioctl doesn't wipe pagecache pages. Thus, we tolerate that after a kernel compromise some portions of plaintext file contents may still be recoverable from memory. This can be solved by enabling page poisoning system-wide, which security conscious users may choose to do. But it's very difficult to solve otherwise, e.g. note that plaintext file contents may have been read in other places than pagecache pages. Like FS_IOC_ADD_ENCRYPTION_KEY, FS_IOC_REMOVE_ENCRYPTION_KEY is initially restricted to privileged users only. This is sufficient for some use cases, but not all. A later patch will relax this restriction, but it will require introducing key hashes, among other changes. Reviewed-by: Theodore Ts'o <tytso@mit.edu> Signed-off-by: Eric Biggers <ebiggers@google.com>
2019-08-05 10:35:46 +08:00
return -EACCES;
/* Find the key being removed. */
fscrypt: stop using keyrings subsystem for fscrypt_master_key The approach of fs/crypto/ internally managing the fscrypt_master_key structs as the payloads of "struct key" objects contained in a "struct key" keyring has outlived its usefulness. The original idea was to simplify the code by reusing code from the keyrings subsystem. However, several issues have arisen that can't easily be resolved: - When a master key struct is destroyed, blk_crypto_evict_key() must be called on any per-mode keys embedded in it. (This started being the case when inline encryption support was added.) Yet, the keyrings subsystem can arbitrarily delay the destruction of keys, even past the time the filesystem was unmounted. Therefore, currently there is no easy way to call blk_crypto_evict_key() when a master key is destroyed. Currently, this is worked around by holding an extra reference to the filesystem's request_queue(s). But it was overlooked that the request_queue reference is *not* guaranteed to pin the corresponding blk_crypto_profile too; for device-mapper devices that support inline crypto, it doesn't. This can cause a use-after-free. - When the last inode that was using an incompletely-removed master key is evicted, the master key removal is completed by removing the key struct from the keyring. Currently this is done via key_invalidate(). Yet, key_invalidate() takes the key semaphore. This can deadlock when called from the shrinker, since in fscrypt_ioctl_add_key(), memory is allocated with GFP_KERNEL under the same semaphore. - More generally, the fact that the keyrings subsystem can arbitrarily delay the destruction of keys (via garbage collection delay, or via random processes getting temporary key references) is undesirable, as it means we can't strictly guarantee that all secrets are ever wiped. - Doing the master key lookups via the keyrings subsystem results in the key_permission LSM hook being called. fscrypt doesn't want this, as all access control for encrypted files is designed to happen via the files themselves, like any other files. The workaround which SELinux users are using is to change their SELinux policy to grant key search access to all domains. This works, but it is an odd extra step that shouldn't really have to be done. The fix for all these issues is to change the implementation to what I should have done originally: don't use the keyrings subsystem to keep track of the filesystem's fscrypt_master_key structs. Instead, just store them in a regular kernel data structure, and rework the reference counting, locking, and lifetime accordingly. Retain support for RCU-mode key lookups by using a hash table. Replace fscrypt_sb_free() with fscrypt_sb_delete(), which releases the keys synchronously and runs a bit earlier during unmount, so that block devices are still available. A side effect of this patch is that neither the master keys themselves nor the filesystem keyrings will be listed in /proc/keys anymore. ("Master key users" and the master key users keyrings will still be listed.) However, this was mostly an implementation detail, and it was intended just for debugging purposes. I don't know of anyone using it. This patch does *not* change how "master key users" (->mk_users) works; that still uses the keyrings subsystem. That is still needed for key quotas, and changing that isn't necessary to solve the issues listed above. If we decide to change that too, it would be a separate patch. I've marked this as fixing the original commit that added the fscrypt keyring, but as noted above the most important issue that this patch fixes wasn't introduced until the addition of inline encryption support. Fixes: 22d94f493bfb ("fscrypt: add FS_IOC_ADD_ENCRYPTION_KEY ioctl") Signed-off-by: Eric Biggers <ebiggers@google.com> Link: https://lore.kernel.org/r/20220901193208.138056-2-ebiggers@kernel.org
2022-09-02 03:32:06 +08:00
mk = fscrypt_find_master_key(sb, &arg.key_spec);
if (!mk)
return -ENOKEY;
down_write(&mk->mk_sem);
fscrypt: add FS_IOC_REMOVE_ENCRYPTION_KEY ioctl Add a new fscrypt ioctl, FS_IOC_REMOVE_ENCRYPTION_KEY. This ioctl removes an encryption key that was added by FS_IOC_ADD_ENCRYPTION_KEY. It wipes the secret key itself, then "locks" the encrypted files and directories that had been unlocked using that key -- implemented by evicting the relevant dentries and inodes from the VFS caches. The problem this solves is that many fscrypt users want the ability to remove encryption keys, causing the corresponding encrypted directories to appear "locked" (presented in ciphertext form) again. Moreover, users want removing an encryption key to *really* remove it, in the sense that the removed keys cannot be recovered even if kernel memory is compromised, e.g. by the exploit of a kernel security vulnerability or by a physical attack. This is desirable after a user logs out of the system, for example. In many cases users even already assume this to be the case and are surprised to hear when it's not. It is not sufficient to simply unlink the master key from the keyring (or to revoke or invalidate it), since the actual encryption transform objects are still pinned in memory by their inodes. Therefore, to really remove a key we must also evict the relevant inodes. Currently one workaround is to run 'sync && echo 2 > /proc/sys/vm/drop_caches'. But, that evicts all unused inodes in the system rather than just the inodes associated with the key being removed, causing severe performance problems. Moreover, it requires root privileges, so regular users can't "lock" their encrypted files. Another workaround, used in Chromium OS kernels, is to add a new VFS-level ioctl FS_IOC_DROP_CACHE which is a more restricted version of drop_caches that operates on a single super_block. It does: shrink_dcache_sb(sb); invalidate_inodes(sb, false); But it's still a hack. Yet, the major users of filesystem encryption want this feature badly enough that they are actually using these hacks. To properly solve the problem, start maintaining a list of the inodes which have been "unlocked" using each master key. Originally this wasn't possible because the kernel didn't keep track of in-use master keys at all. But, with the ->s_master_keys keyring it is now possible. Then, add an ioctl FS_IOC_REMOVE_ENCRYPTION_KEY. It finds the specified master key in ->s_master_keys, then wipes the secret key itself, which prevents any additional inodes from being unlocked with the key. Then, it syncs the filesystem and evicts the inodes in the key's list. The normal inode eviction code will free and wipe the per-file keys (in ->i_crypt_info). Note that freeing ->i_crypt_info without evicting the inodes was also considered, but would have been racy. Some inodes may still be in use when a master key is removed, and we can't simply revoke random file descriptors, mmap's, etc. Thus, the ioctl simply skips in-use inodes, and returns -EBUSY to indicate that some inodes weren't evicted. The master key *secret* is still removed, but the fscrypt_master_key struct remains to keep track of the remaining inodes. Userspace can then retry the ioctl to evict the remaining inodes. Alternatively, if userspace adds the key again, the refreshed secret will be associated with the existing list of inodes so they remain correctly tracked for future key removals. The ioctl doesn't wipe pagecache pages. Thus, we tolerate that after a kernel compromise some portions of plaintext file contents may still be recoverable from memory. This can be solved by enabling page poisoning system-wide, which security conscious users may choose to do. But it's very difficult to solve otherwise, e.g. note that plaintext file contents may have been read in other places than pagecache pages. Like FS_IOC_ADD_ENCRYPTION_KEY, FS_IOC_REMOVE_ENCRYPTION_KEY is initially restricted to privileged users only. This is sufficient for some use cases, but not all. A later patch will relax this restriction, but it will require introducing key hashes, among other changes. Reviewed-by: Theodore Ts'o <tytso@mit.edu> Signed-off-by: Eric Biggers <ebiggers@google.com>
2019-08-05 10:35:46 +08:00
fscrypt: add FS_IOC_REMOVE_ENCRYPTION_KEY_ALL_USERS ioctl Add a root-only variant of the FS_IOC_REMOVE_ENCRYPTION_KEY ioctl which removes all users' claims of the key, not just the current user's claim. I.e., it always removes the key itself, no matter how many users have added it. This is useful for forcing a directory to be locked, without having to figure out which user ID(s) the key was added under. This is planned to be used by a command like 'sudo fscrypt lock DIR --all-users' in the fscrypt userspace tool (http://github.com/google/fscrypt). Reviewed-by: Theodore Ts'o <tytso@mit.edu> Signed-off-by: Eric Biggers <ebiggers@google.com>
2019-08-05 10:35:47 +08:00
/* If relevant, remove current user's (or all users) claim to the key */
fscrypt: allow unprivileged users to add/remove keys for v2 policies Allow the FS_IOC_ADD_ENCRYPTION_KEY and FS_IOC_REMOVE_ENCRYPTION_KEY ioctls to be used by non-root users to add and remove encryption keys from the filesystem-level crypto keyrings, subject to limitations. Motivation: while privileged fscrypt key management is sufficient for some users (e.g. Android and Chromium OS, where a privileged process manages all keys), the old API by design also allows non-root users to set up and use encrypted directories, and we don't want to regress on that. Especially, we don't want to force users to continue using the old API, running into the visibility mismatch between files and keyrings and being unable to "lock" encrypted directories. Intuitively, the ioctls have to be privileged since they manipulate filesystem-level state. However, it's actually safe to make them unprivileged if we very carefully enforce some specific limitations. First, each key must be identified by a cryptographic hash so that a user can't add the wrong key for another user's files. For v2 encryption policies, we use the key_identifier for this. v1 policies don't have this, so managing keys for them remains privileged. Second, each key a user adds is charged to their quota for the keyrings service. Thus, a user can't exhaust memory by adding a huge number of keys. By default each non-root user is allowed up to 200 keys; this can be changed using the existing sysctl 'kernel.keys.maxkeys'. Third, if multiple users add the same key, we keep track of those users of the key (of which there remains a single copy), and won't really remove the key, i.e. "lock" the encrypted files, until all those users have removed it. This prevents denial of service attacks that would be possible under simpler schemes, such allowing the first user who added a key to remove it -- since that could be a malicious user who has compromised the key. Of course, encryption keys should be kept secret, but the idea is that using encryption should never be *less* secure than not using encryption, even if your key was compromised. We tolerate that a user will be unable to really remove a key, i.e. unable to "lock" their encrypted files, if another user has added the same key. But in a sense, this is actually a good thing because it will avoid providing a false notion of security where a key appears to have been removed when actually it's still in memory, available to any attacker who compromises the operating system kernel. Reviewed-by: Theodore Ts'o <tytso@mit.edu> Signed-off-by: Eric Biggers <ebiggers@google.com>
2019-08-05 10:35:47 +08:00
if (mk->mk_users && mk->mk_users->keys.nr_leaves_on_tree != 0) {
fscrypt: add FS_IOC_REMOVE_ENCRYPTION_KEY_ALL_USERS ioctl Add a root-only variant of the FS_IOC_REMOVE_ENCRYPTION_KEY ioctl which removes all users' claims of the key, not just the current user's claim. I.e., it always removes the key itself, no matter how many users have added it. This is useful for forcing a directory to be locked, without having to figure out which user ID(s) the key was added under. This is planned to be used by a command like 'sudo fscrypt lock DIR --all-users' in the fscrypt userspace tool (http://github.com/google/fscrypt). Reviewed-by: Theodore Ts'o <tytso@mit.edu> Signed-off-by: Eric Biggers <ebiggers@google.com>
2019-08-05 10:35:47 +08:00
if (all_users)
err = keyring_clear(mk->mk_users);
else
err = remove_master_key_user(mk);
fscrypt: allow unprivileged users to add/remove keys for v2 policies Allow the FS_IOC_ADD_ENCRYPTION_KEY and FS_IOC_REMOVE_ENCRYPTION_KEY ioctls to be used by non-root users to add and remove encryption keys from the filesystem-level crypto keyrings, subject to limitations. Motivation: while privileged fscrypt key management is sufficient for some users (e.g. Android and Chromium OS, where a privileged process manages all keys), the old API by design also allows non-root users to set up and use encrypted directories, and we don't want to regress on that. Especially, we don't want to force users to continue using the old API, running into the visibility mismatch between files and keyrings and being unable to "lock" encrypted directories. Intuitively, the ioctls have to be privileged since they manipulate filesystem-level state. However, it's actually safe to make them unprivileged if we very carefully enforce some specific limitations. First, each key must be identified by a cryptographic hash so that a user can't add the wrong key for another user's files. For v2 encryption policies, we use the key_identifier for this. v1 policies don't have this, so managing keys for them remains privileged. Second, each key a user adds is charged to their quota for the keyrings service. Thus, a user can't exhaust memory by adding a huge number of keys. By default each non-root user is allowed up to 200 keys; this can be changed using the existing sysctl 'kernel.keys.maxkeys'. Third, if multiple users add the same key, we keep track of those users of the key (of which there remains a single copy), and won't really remove the key, i.e. "lock" the encrypted files, until all those users have removed it. This prevents denial of service attacks that would be possible under simpler schemes, such allowing the first user who added a key to remove it -- since that could be a malicious user who has compromised the key. Of course, encryption keys should be kept secret, but the idea is that using encryption should never be *less* secure than not using encryption, even if your key was compromised. We tolerate that a user will be unable to really remove a key, i.e. unable to "lock" their encrypted files, if another user has added the same key. But in a sense, this is actually a good thing because it will avoid providing a false notion of security where a key appears to have been removed when actually it's still in memory, available to any attacker who compromises the operating system kernel. Reviewed-by: Theodore Ts'o <tytso@mit.edu> Signed-off-by: Eric Biggers <ebiggers@google.com>
2019-08-05 10:35:47 +08:00
if (err) {
fscrypt: stop using keyrings subsystem for fscrypt_master_key The approach of fs/crypto/ internally managing the fscrypt_master_key structs as the payloads of "struct key" objects contained in a "struct key" keyring has outlived its usefulness. The original idea was to simplify the code by reusing code from the keyrings subsystem. However, several issues have arisen that can't easily be resolved: - When a master key struct is destroyed, blk_crypto_evict_key() must be called on any per-mode keys embedded in it. (This started being the case when inline encryption support was added.) Yet, the keyrings subsystem can arbitrarily delay the destruction of keys, even past the time the filesystem was unmounted. Therefore, currently there is no easy way to call blk_crypto_evict_key() when a master key is destroyed. Currently, this is worked around by holding an extra reference to the filesystem's request_queue(s). But it was overlooked that the request_queue reference is *not* guaranteed to pin the corresponding blk_crypto_profile too; for device-mapper devices that support inline crypto, it doesn't. This can cause a use-after-free. - When the last inode that was using an incompletely-removed master key is evicted, the master key removal is completed by removing the key struct from the keyring. Currently this is done via key_invalidate(). Yet, key_invalidate() takes the key semaphore. This can deadlock when called from the shrinker, since in fscrypt_ioctl_add_key(), memory is allocated with GFP_KERNEL under the same semaphore. - More generally, the fact that the keyrings subsystem can arbitrarily delay the destruction of keys (via garbage collection delay, or via random processes getting temporary key references) is undesirable, as it means we can't strictly guarantee that all secrets are ever wiped. - Doing the master key lookups via the keyrings subsystem results in the key_permission LSM hook being called. fscrypt doesn't want this, as all access control for encrypted files is designed to happen via the files themselves, like any other files. The workaround which SELinux users are using is to change their SELinux policy to grant key search access to all domains. This works, but it is an odd extra step that shouldn't really have to be done. The fix for all these issues is to change the implementation to what I should have done originally: don't use the keyrings subsystem to keep track of the filesystem's fscrypt_master_key structs. Instead, just store them in a regular kernel data structure, and rework the reference counting, locking, and lifetime accordingly. Retain support for RCU-mode key lookups by using a hash table. Replace fscrypt_sb_free() with fscrypt_sb_delete(), which releases the keys synchronously and runs a bit earlier during unmount, so that block devices are still available. A side effect of this patch is that neither the master keys themselves nor the filesystem keyrings will be listed in /proc/keys anymore. ("Master key users" and the master key users keyrings will still be listed.) However, this was mostly an implementation detail, and it was intended just for debugging purposes. I don't know of anyone using it. This patch does *not* change how "master key users" (->mk_users) works; that still uses the keyrings subsystem. That is still needed for key quotas, and changing that isn't necessary to solve the issues listed above. If we decide to change that too, it would be a separate patch. I've marked this as fixing the original commit that added the fscrypt keyring, but as noted above the most important issue that this patch fixes wasn't introduced until the addition of inline encryption support. Fixes: 22d94f493bfb ("fscrypt: add FS_IOC_ADD_ENCRYPTION_KEY ioctl") Signed-off-by: Eric Biggers <ebiggers@google.com> Link: https://lore.kernel.org/r/20220901193208.138056-2-ebiggers@kernel.org
2022-09-02 03:32:06 +08:00
up_write(&mk->mk_sem);
fscrypt: allow unprivileged users to add/remove keys for v2 policies Allow the FS_IOC_ADD_ENCRYPTION_KEY and FS_IOC_REMOVE_ENCRYPTION_KEY ioctls to be used by non-root users to add and remove encryption keys from the filesystem-level crypto keyrings, subject to limitations. Motivation: while privileged fscrypt key management is sufficient for some users (e.g. Android and Chromium OS, where a privileged process manages all keys), the old API by design also allows non-root users to set up and use encrypted directories, and we don't want to regress on that. Especially, we don't want to force users to continue using the old API, running into the visibility mismatch between files and keyrings and being unable to "lock" encrypted directories. Intuitively, the ioctls have to be privileged since they manipulate filesystem-level state. However, it's actually safe to make them unprivileged if we very carefully enforce some specific limitations. First, each key must be identified by a cryptographic hash so that a user can't add the wrong key for another user's files. For v2 encryption policies, we use the key_identifier for this. v1 policies don't have this, so managing keys for them remains privileged. Second, each key a user adds is charged to their quota for the keyrings service. Thus, a user can't exhaust memory by adding a huge number of keys. By default each non-root user is allowed up to 200 keys; this can be changed using the existing sysctl 'kernel.keys.maxkeys'. Third, if multiple users add the same key, we keep track of those users of the key (of which there remains a single copy), and won't really remove the key, i.e. "lock" the encrypted files, until all those users have removed it. This prevents denial of service attacks that would be possible under simpler schemes, such allowing the first user who added a key to remove it -- since that could be a malicious user who has compromised the key. Of course, encryption keys should be kept secret, but the idea is that using encryption should never be *less* secure than not using encryption, even if your key was compromised. We tolerate that a user will be unable to really remove a key, i.e. unable to "lock" their encrypted files, if another user has added the same key. But in a sense, this is actually a good thing because it will avoid providing a false notion of security where a key appears to have been removed when actually it's still in memory, available to any attacker who compromises the operating system kernel. Reviewed-by: Theodore Ts'o <tytso@mit.edu> Signed-off-by: Eric Biggers <ebiggers@google.com>
2019-08-05 10:35:47 +08:00
goto out_put_key;
}
if (mk->mk_users->keys.nr_leaves_on_tree != 0) {
/*
* Other users have still added the key too. We removed
* the current user's claim to the key, but we still
* can't remove the key itself.
*/
status_flags |=
FSCRYPT_KEY_REMOVAL_STATUS_FLAG_OTHER_USERS;
err = 0;
fscrypt: stop using keyrings subsystem for fscrypt_master_key The approach of fs/crypto/ internally managing the fscrypt_master_key structs as the payloads of "struct key" objects contained in a "struct key" keyring has outlived its usefulness. The original idea was to simplify the code by reusing code from the keyrings subsystem. However, several issues have arisen that can't easily be resolved: - When a master key struct is destroyed, blk_crypto_evict_key() must be called on any per-mode keys embedded in it. (This started being the case when inline encryption support was added.) Yet, the keyrings subsystem can arbitrarily delay the destruction of keys, even past the time the filesystem was unmounted. Therefore, currently there is no easy way to call blk_crypto_evict_key() when a master key is destroyed. Currently, this is worked around by holding an extra reference to the filesystem's request_queue(s). But it was overlooked that the request_queue reference is *not* guaranteed to pin the corresponding blk_crypto_profile too; for device-mapper devices that support inline crypto, it doesn't. This can cause a use-after-free. - When the last inode that was using an incompletely-removed master key is evicted, the master key removal is completed by removing the key struct from the keyring. Currently this is done via key_invalidate(). Yet, key_invalidate() takes the key semaphore. This can deadlock when called from the shrinker, since in fscrypt_ioctl_add_key(), memory is allocated with GFP_KERNEL under the same semaphore. - More generally, the fact that the keyrings subsystem can arbitrarily delay the destruction of keys (via garbage collection delay, or via random processes getting temporary key references) is undesirable, as it means we can't strictly guarantee that all secrets are ever wiped. - Doing the master key lookups via the keyrings subsystem results in the key_permission LSM hook being called. fscrypt doesn't want this, as all access control for encrypted files is designed to happen via the files themselves, like any other files. The workaround which SELinux users are using is to change their SELinux policy to grant key search access to all domains. This works, but it is an odd extra step that shouldn't really have to be done. The fix for all these issues is to change the implementation to what I should have done originally: don't use the keyrings subsystem to keep track of the filesystem's fscrypt_master_key structs. Instead, just store them in a regular kernel data structure, and rework the reference counting, locking, and lifetime accordingly. Retain support for RCU-mode key lookups by using a hash table. Replace fscrypt_sb_free() with fscrypt_sb_delete(), which releases the keys synchronously and runs a bit earlier during unmount, so that block devices are still available. A side effect of this patch is that neither the master keys themselves nor the filesystem keyrings will be listed in /proc/keys anymore. ("Master key users" and the master key users keyrings will still be listed.) However, this was mostly an implementation detail, and it was intended just for debugging purposes. I don't know of anyone using it. This patch does *not* change how "master key users" (->mk_users) works; that still uses the keyrings subsystem. That is still needed for key quotas, and changing that isn't necessary to solve the issues listed above. If we decide to change that too, it would be a separate patch. I've marked this as fixing the original commit that added the fscrypt keyring, but as noted above the most important issue that this patch fixes wasn't introduced until the addition of inline encryption support. Fixes: 22d94f493bfb ("fscrypt: add FS_IOC_ADD_ENCRYPTION_KEY ioctl") Signed-off-by: Eric Biggers <ebiggers@google.com> Link: https://lore.kernel.org/r/20220901193208.138056-2-ebiggers@kernel.org
2022-09-02 03:32:06 +08:00
up_write(&mk->mk_sem);
fscrypt: allow unprivileged users to add/remove keys for v2 policies Allow the FS_IOC_ADD_ENCRYPTION_KEY and FS_IOC_REMOVE_ENCRYPTION_KEY ioctls to be used by non-root users to add and remove encryption keys from the filesystem-level crypto keyrings, subject to limitations. Motivation: while privileged fscrypt key management is sufficient for some users (e.g. Android and Chromium OS, where a privileged process manages all keys), the old API by design also allows non-root users to set up and use encrypted directories, and we don't want to regress on that. Especially, we don't want to force users to continue using the old API, running into the visibility mismatch between files and keyrings and being unable to "lock" encrypted directories. Intuitively, the ioctls have to be privileged since they manipulate filesystem-level state. However, it's actually safe to make them unprivileged if we very carefully enforce some specific limitations. First, each key must be identified by a cryptographic hash so that a user can't add the wrong key for another user's files. For v2 encryption policies, we use the key_identifier for this. v1 policies don't have this, so managing keys for them remains privileged. Second, each key a user adds is charged to their quota for the keyrings service. Thus, a user can't exhaust memory by adding a huge number of keys. By default each non-root user is allowed up to 200 keys; this can be changed using the existing sysctl 'kernel.keys.maxkeys'. Third, if multiple users add the same key, we keep track of those users of the key (of which there remains a single copy), and won't really remove the key, i.e. "lock" the encrypted files, until all those users have removed it. This prevents denial of service attacks that would be possible under simpler schemes, such allowing the first user who added a key to remove it -- since that could be a malicious user who has compromised the key. Of course, encryption keys should be kept secret, but the idea is that using encryption should never be *less* secure than not using encryption, even if your key was compromised. We tolerate that a user will be unable to really remove a key, i.e. unable to "lock" their encrypted files, if another user has added the same key. But in a sense, this is actually a good thing because it will avoid providing a false notion of security where a key appears to have been removed when actually it's still in memory, available to any attacker who compromises the operating system kernel. Reviewed-by: Theodore Ts'o <tytso@mit.edu> Signed-off-by: Eric Biggers <ebiggers@google.com>
2019-08-05 10:35:47 +08:00
goto out_put_key;
}
}
/* No user claims remaining. Go ahead and wipe the secret. */
fscrypt: stop using keyrings subsystem for fscrypt_master_key The approach of fs/crypto/ internally managing the fscrypt_master_key structs as the payloads of "struct key" objects contained in a "struct key" keyring has outlived its usefulness. The original idea was to simplify the code by reusing code from the keyrings subsystem. However, several issues have arisen that can't easily be resolved: - When a master key struct is destroyed, blk_crypto_evict_key() must be called on any per-mode keys embedded in it. (This started being the case when inline encryption support was added.) Yet, the keyrings subsystem can arbitrarily delay the destruction of keys, even past the time the filesystem was unmounted. Therefore, currently there is no easy way to call blk_crypto_evict_key() when a master key is destroyed. Currently, this is worked around by holding an extra reference to the filesystem's request_queue(s). But it was overlooked that the request_queue reference is *not* guaranteed to pin the corresponding blk_crypto_profile too; for device-mapper devices that support inline crypto, it doesn't. This can cause a use-after-free. - When the last inode that was using an incompletely-removed master key is evicted, the master key removal is completed by removing the key struct from the keyring. Currently this is done via key_invalidate(). Yet, key_invalidate() takes the key semaphore. This can deadlock when called from the shrinker, since in fscrypt_ioctl_add_key(), memory is allocated with GFP_KERNEL under the same semaphore. - More generally, the fact that the keyrings subsystem can arbitrarily delay the destruction of keys (via garbage collection delay, or via random processes getting temporary key references) is undesirable, as it means we can't strictly guarantee that all secrets are ever wiped. - Doing the master key lookups via the keyrings subsystem results in the key_permission LSM hook being called. fscrypt doesn't want this, as all access control for encrypted files is designed to happen via the files themselves, like any other files. The workaround which SELinux users are using is to change their SELinux policy to grant key search access to all domains. This works, but it is an odd extra step that shouldn't really have to be done. The fix for all these issues is to change the implementation to what I should have done originally: don't use the keyrings subsystem to keep track of the filesystem's fscrypt_master_key structs. Instead, just store them in a regular kernel data structure, and rework the reference counting, locking, and lifetime accordingly. Retain support for RCU-mode key lookups by using a hash table. Replace fscrypt_sb_free() with fscrypt_sb_delete(), which releases the keys synchronously and runs a bit earlier during unmount, so that block devices are still available. A side effect of this patch is that neither the master keys themselves nor the filesystem keyrings will be listed in /proc/keys anymore. ("Master key users" and the master key users keyrings will still be listed.) However, this was mostly an implementation detail, and it was intended just for debugging purposes. I don't know of anyone using it. This patch does *not* change how "master key users" (->mk_users) works; that still uses the keyrings subsystem. That is still needed for key quotas, and changing that isn't necessary to solve the issues listed above. If we decide to change that too, it would be a separate patch. I've marked this as fixing the original commit that added the fscrypt keyring, but as noted above the most important issue that this patch fixes wasn't introduced until the addition of inline encryption support. Fixes: 22d94f493bfb ("fscrypt: add FS_IOC_ADD_ENCRYPTION_KEY ioctl") Signed-off-by: Eric Biggers <ebiggers@google.com> Link: https://lore.kernel.org/r/20220901193208.138056-2-ebiggers@kernel.org
2022-09-02 03:32:06 +08:00
err = -ENOKEY;
fscrypt: add FS_IOC_REMOVE_ENCRYPTION_KEY ioctl Add a new fscrypt ioctl, FS_IOC_REMOVE_ENCRYPTION_KEY. This ioctl removes an encryption key that was added by FS_IOC_ADD_ENCRYPTION_KEY. It wipes the secret key itself, then "locks" the encrypted files and directories that had been unlocked using that key -- implemented by evicting the relevant dentries and inodes from the VFS caches. The problem this solves is that many fscrypt users want the ability to remove encryption keys, causing the corresponding encrypted directories to appear "locked" (presented in ciphertext form) again. Moreover, users want removing an encryption key to *really* remove it, in the sense that the removed keys cannot be recovered even if kernel memory is compromised, e.g. by the exploit of a kernel security vulnerability or by a physical attack. This is desirable after a user logs out of the system, for example. In many cases users even already assume this to be the case and are surprised to hear when it's not. It is not sufficient to simply unlink the master key from the keyring (or to revoke or invalidate it), since the actual encryption transform objects are still pinned in memory by their inodes. Therefore, to really remove a key we must also evict the relevant inodes. Currently one workaround is to run 'sync && echo 2 > /proc/sys/vm/drop_caches'. But, that evicts all unused inodes in the system rather than just the inodes associated with the key being removed, causing severe performance problems. Moreover, it requires root privileges, so regular users can't "lock" their encrypted files. Another workaround, used in Chromium OS kernels, is to add a new VFS-level ioctl FS_IOC_DROP_CACHE which is a more restricted version of drop_caches that operates on a single super_block. It does: shrink_dcache_sb(sb); invalidate_inodes(sb, false); But it's still a hack. Yet, the major users of filesystem encryption want this feature badly enough that they are actually using these hacks. To properly solve the problem, start maintaining a list of the inodes which have been "unlocked" using each master key. Originally this wasn't possible because the kernel didn't keep track of in-use master keys at all. But, with the ->s_master_keys keyring it is now possible. Then, add an ioctl FS_IOC_REMOVE_ENCRYPTION_KEY. It finds the specified master key in ->s_master_keys, then wipes the secret key itself, which prevents any additional inodes from being unlocked with the key. Then, it syncs the filesystem and evicts the inodes in the key's list. The normal inode eviction code will free and wipe the per-file keys (in ->i_crypt_info). Note that freeing ->i_crypt_info without evicting the inodes was also considered, but would have been racy. Some inodes may still be in use when a master key is removed, and we can't simply revoke random file descriptors, mmap's, etc. Thus, the ioctl simply skips in-use inodes, and returns -EBUSY to indicate that some inodes weren't evicted. The master key *secret* is still removed, but the fscrypt_master_key struct remains to keep track of the remaining inodes. Userspace can then retry the ioctl to evict the remaining inodes. Alternatively, if userspace adds the key again, the refreshed secret will be associated with the existing list of inodes so they remain correctly tracked for future key removals. The ioctl doesn't wipe pagecache pages. Thus, we tolerate that after a kernel compromise some portions of plaintext file contents may still be recoverable from memory. This can be solved by enabling page poisoning system-wide, which security conscious users may choose to do. But it's very difficult to solve otherwise, e.g. note that plaintext file contents may have been read in other places than pagecache pages. Like FS_IOC_ADD_ENCRYPTION_KEY, FS_IOC_REMOVE_ENCRYPTION_KEY is initially restricted to privileged users only. This is sufficient for some use cases, but not all. A later patch will relax this restriction, but it will require introducing key hashes, among other changes. Reviewed-by: Theodore Ts'o <tytso@mit.edu> Signed-off-by: Eric Biggers <ebiggers@google.com>
2019-08-05 10:35:46 +08:00
if (is_master_key_secret_present(&mk->mk_secret)) {
wipe_master_key_secret(&mk->mk_secret);
fscrypt: stop using keyrings subsystem for fscrypt_master_key The approach of fs/crypto/ internally managing the fscrypt_master_key structs as the payloads of "struct key" objects contained in a "struct key" keyring has outlived its usefulness. The original idea was to simplify the code by reusing code from the keyrings subsystem. However, several issues have arisen that can't easily be resolved: - When a master key struct is destroyed, blk_crypto_evict_key() must be called on any per-mode keys embedded in it. (This started being the case when inline encryption support was added.) Yet, the keyrings subsystem can arbitrarily delay the destruction of keys, even past the time the filesystem was unmounted. Therefore, currently there is no easy way to call blk_crypto_evict_key() when a master key is destroyed. Currently, this is worked around by holding an extra reference to the filesystem's request_queue(s). But it was overlooked that the request_queue reference is *not* guaranteed to pin the corresponding blk_crypto_profile too; for device-mapper devices that support inline crypto, it doesn't. This can cause a use-after-free. - When the last inode that was using an incompletely-removed master key is evicted, the master key removal is completed by removing the key struct from the keyring. Currently this is done via key_invalidate(). Yet, key_invalidate() takes the key semaphore. This can deadlock when called from the shrinker, since in fscrypt_ioctl_add_key(), memory is allocated with GFP_KERNEL under the same semaphore. - More generally, the fact that the keyrings subsystem can arbitrarily delay the destruction of keys (via garbage collection delay, or via random processes getting temporary key references) is undesirable, as it means we can't strictly guarantee that all secrets are ever wiped. - Doing the master key lookups via the keyrings subsystem results in the key_permission LSM hook being called. fscrypt doesn't want this, as all access control for encrypted files is designed to happen via the files themselves, like any other files. The workaround which SELinux users are using is to change their SELinux policy to grant key search access to all domains. This works, but it is an odd extra step that shouldn't really have to be done. The fix for all these issues is to change the implementation to what I should have done originally: don't use the keyrings subsystem to keep track of the filesystem's fscrypt_master_key structs. Instead, just store them in a regular kernel data structure, and rework the reference counting, locking, and lifetime accordingly. Retain support for RCU-mode key lookups by using a hash table. Replace fscrypt_sb_free() with fscrypt_sb_delete(), which releases the keys synchronously and runs a bit earlier during unmount, so that block devices are still available. A side effect of this patch is that neither the master keys themselves nor the filesystem keyrings will be listed in /proc/keys anymore. ("Master key users" and the master key users keyrings will still be listed.) However, this was mostly an implementation detail, and it was intended just for debugging purposes. I don't know of anyone using it. This patch does *not* change how "master key users" (->mk_users) works; that still uses the keyrings subsystem. That is still needed for key quotas, and changing that isn't necessary to solve the issues listed above. If we decide to change that too, it would be a separate patch. I've marked this as fixing the original commit that added the fscrypt keyring, but as noted above the most important issue that this patch fixes wasn't introduced until the addition of inline encryption support. Fixes: 22d94f493bfb ("fscrypt: add FS_IOC_ADD_ENCRYPTION_KEY ioctl") Signed-off-by: Eric Biggers <ebiggers@google.com> Link: https://lore.kernel.org/r/20220901193208.138056-2-ebiggers@kernel.org
2022-09-02 03:32:06 +08:00
fscrypt_put_master_key_activeref(mk);
fscrypt: add FS_IOC_REMOVE_ENCRYPTION_KEY ioctl Add a new fscrypt ioctl, FS_IOC_REMOVE_ENCRYPTION_KEY. This ioctl removes an encryption key that was added by FS_IOC_ADD_ENCRYPTION_KEY. It wipes the secret key itself, then "locks" the encrypted files and directories that had been unlocked using that key -- implemented by evicting the relevant dentries and inodes from the VFS caches. The problem this solves is that many fscrypt users want the ability to remove encryption keys, causing the corresponding encrypted directories to appear "locked" (presented in ciphertext form) again. Moreover, users want removing an encryption key to *really* remove it, in the sense that the removed keys cannot be recovered even if kernel memory is compromised, e.g. by the exploit of a kernel security vulnerability or by a physical attack. This is desirable after a user logs out of the system, for example. In many cases users even already assume this to be the case and are surprised to hear when it's not. It is not sufficient to simply unlink the master key from the keyring (or to revoke or invalidate it), since the actual encryption transform objects are still pinned in memory by their inodes. Therefore, to really remove a key we must also evict the relevant inodes. Currently one workaround is to run 'sync && echo 2 > /proc/sys/vm/drop_caches'. But, that evicts all unused inodes in the system rather than just the inodes associated with the key being removed, causing severe performance problems. Moreover, it requires root privileges, so regular users can't "lock" their encrypted files. Another workaround, used in Chromium OS kernels, is to add a new VFS-level ioctl FS_IOC_DROP_CACHE which is a more restricted version of drop_caches that operates on a single super_block. It does: shrink_dcache_sb(sb); invalidate_inodes(sb, false); But it's still a hack. Yet, the major users of filesystem encryption want this feature badly enough that they are actually using these hacks. To properly solve the problem, start maintaining a list of the inodes which have been "unlocked" using each master key. Originally this wasn't possible because the kernel didn't keep track of in-use master keys at all. But, with the ->s_master_keys keyring it is now possible. Then, add an ioctl FS_IOC_REMOVE_ENCRYPTION_KEY. It finds the specified master key in ->s_master_keys, then wipes the secret key itself, which prevents any additional inodes from being unlocked with the key. Then, it syncs the filesystem and evicts the inodes in the key's list. The normal inode eviction code will free and wipe the per-file keys (in ->i_crypt_info). Note that freeing ->i_crypt_info without evicting the inodes was also considered, but would have been racy. Some inodes may still be in use when a master key is removed, and we can't simply revoke random file descriptors, mmap's, etc. Thus, the ioctl simply skips in-use inodes, and returns -EBUSY to indicate that some inodes weren't evicted. The master key *secret* is still removed, but the fscrypt_master_key struct remains to keep track of the remaining inodes. Userspace can then retry the ioctl to evict the remaining inodes. Alternatively, if userspace adds the key again, the refreshed secret will be associated with the existing list of inodes so they remain correctly tracked for future key removals. The ioctl doesn't wipe pagecache pages. Thus, we tolerate that after a kernel compromise some portions of plaintext file contents may still be recoverable from memory. This can be solved by enabling page poisoning system-wide, which security conscious users may choose to do. But it's very difficult to solve otherwise, e.g. note that plaintext file contents may have been read in other places than pagecache pages. Like FS_IOC_ADD_ENCRYPTION_KEY, FS_IOC_REMOVE_ENCRYPTION_KEY is initially restricted to privileged users only. This is sufficient for some use cases, but not all. A later patch will relax this restriction, but it will require introducing key hashes, among other changes. Reviewed-by: Theodore Ts'o <tytso@mit.edu> Signed-off-by: Eric Biggers <ebiggers@google.com>
2019-08-05 10:35:46 +08:00
err = 0;
fscrypt: stop using keyrings subsystem for fscrypt_master_key The approach of fs/crypto/ internally managing the fscrypt_master_key structs as the payloads of "struct key" objects contained in a "struct key" keyring has outlived its usefulness. The original idea was to simplify the code by reusing code from the keyrings subsystem. However, several issues have arisen that can't easily be resolved: - When a master key struct is destroyed, blk_crypto_evict_key() must be called on any per-mode keys embedded in it. (This started being the case when inline encryption support was added.) Yet, the keyrings subsystem can arbitrarily delay the destruction of keys, even past the time the filesystem was unmounted. Therefore, currently there is no easy way to call blk_crypto_evict_key() when a master key is destroyed. Currently, this is worked around by holding an extra reference to the filesystem's request_queue(s). But it was overlooked that the request_queue reference is *not* guaranteed to pin the corresponding blk_crypto_profile too; for device-mapper devices that support inline crypto, it doesn't. This can cause a use-after-free. - When the last inode that was using an incompletely-removed master key is evicted, the master key removal is completed by removing the key struct from the keyring. Currently this is done via key_invalidate(). Yet, key_invalidate() takes the key semaphore. This can deadlock when called from the shrinker, since in fscrypt_ioctl_add_key(), memory is allocated with GFP_KERNEL under the same semaphore. - More generally, the fact that the keyrings subsystem can arbitrarily delay the destruction of keys (via garbage collection delay, or via random processes getting temporary key references) is undesirable, as it means we can't strictly guarantee that all secrets are ever wiped. - Doing the master key lookups via the keyrings subsystem results in the key_permission LSM hook being called. fscrypt doesn't want this, as all access control for encrypted files is designed to happen via the files themselves, like any other files. The workaround which SELinux users are using is to change their SELinux policy to grant key search access to all domains. This works, but it is an odd extra step that shouldn't really have to be done. The fix for all these issues is to change the implementation to what I should have done originally: don't use the keyrings subsystem to keep track of the filesystem's fscrypt_master_key structs. Instead, just store them in a regular kernel data structure, and rework the reference counting, locking, and lifetime accordingly. Retain support for RCU-mode key lookups by using a hash table. Replace fscrypt_sb_free() with fscrypt_sb_delete(), which releases the keys synchronously and runs a bit earlier during unmount, so that block devices are still available. A side effect of this patch is that neither the master keys themselves nor the filesystem keyrings will be listed in /proc/keys anymore. ("Master key users" and the master key users keyrings will still be listed.) However, this was mostly an implementation detail, and it was intended just for debugging purposes. I don't know of anyone using it. This patch does *not* change how "master key users" (->mk_users) works; that still uses the keyrings subsystem. That is still needed for key quotas, and changing that isn't necessary to solve the issues listed above. If we decide to change that too, it would be a separate patch. I've marked this as fixing the original commit that added the fscrypt keyring, but as noted above the most important issue that this patch fixes wasn't introduced until the addition of inline encryption support. Fixes: 22d94f493bfb ("fscrypt: add FS_IOC_ADD_ENCRYPTION_KEY ioctl") Signed-off-by: Eric Biggers <ebiggers@google.com> Link: https://lore.kernel.org/r/20220901193208.138056-2-ebiggers@kernel.org
2022-09-02 03:32:06 +08:00
}
inodes_remain = refcount_read(&mk->mk_active_refs) > 0;
up_write(&mk->mk_sem);
if (inodes_remain) {
fscrypt: add FS_IOC_REMOVE_ENCRYPTION_KEY ioctl Add a new fscrypt ioctl, FS_IOC_REMOVE_ENCRYPTION_KEY. This ioctl removes an encryption key that was added by FS_IOC_ADD_ENCRYPTION_KEY. It wipes the secret key itself, then "locks" the encrypted files and directories that had been unlocked using that key -- implemented by evicting the relevant dentries and inodes from the VFS caches. The problem this solves is that many fscrypt users want the ability to remove encryption keys, causing the corresponding encrypted directories to appear "locked" (presented in ciphertext form) again. Moreover, users want removing an encryption key to *really* remove it, in the sense that the removed keys cannot be recovered even if kernel memory is compromised, e.g. by the exploit of a kernel security vulnerability or by a physical attack. This is desirable after a user logs out of the system, for example. In many cases users even already assume this to be the case and are surprised to hear when it's not. It is not sufficient to simply unlink the master key from the keyring (or to revoke or invalidate it), since the actual encryption transform objects are still pinned in memory by their inodes. Therefore, to really remove a key we must also evict the relevant inodes. Currently one workaround is to run 'sync && echo 2 > /proc/sys/vm/drop_caches'. But, that evicts all unused inodes in the system rather than just the inodes associated with the key being removed, causing severe performance problems. Moreover, it requires root privileges, so regular users can't "lock" their encrypted files. Another workaround, used in Chromium OS kernels, is to add a new VFS-level ioctl FS_IOC_DROP_CACHE which is a more restricted version of drop_caches that operates on a single super_block. It does: shrink_dcache_sb(sb); invalidate_inodes(sb, false); But it's still a hack. Yet, the major users of filesystem encryption want this feature badly enough that they are actually using these hacks. To properly solve the problem, start maintaining a list of the inodes which have been "unlocked" using each master key. Originally this wasn't possible because the kernel didn't keep track of in-use master keys at all. But, with the ->s_master_keys keyring it is now possible. Then, add an ioctl FS_IOC_REMOVE_ENCRYPTION_KEY. It finds the specified master key in ->s_master_keys, then wipes the secret key itself, which prevents any additional inodes from being unlocked with the key. Then, it syncs the filesystem and evicts the inodes in the key's list. The normal inode eviction code will free and wipe the per-file keys (in ->i_crypt_info). Note that freeing ->i_crypt_info without evicting the inodes was also considered, but would have been racy. Some inodes may still be in use when a master key is removed, and we can't simply revoke random file descriptors, mmap's, etc. Thus, the ioctl simply skips in-use inodes, and returns -EBUSY to indicate that some inodes weren't evicted. The master key *secret* is still removed, but the fscrypt_master_key struct remains to keep track of the remaining inodes. Userspace can then retry the ioctl to evict the remaining inodes. Alternatively, if userspace adds the key again, the refreshed secret will be associated with the existing list of inodes so they remain correctly tracked for future key removals. The ioctl doesn't wipe pagecache pages. Thus, we tolerate that after a kernel compromise some portions of plaintext file contents may still be recoverable from memory. This can be solved by enabling page poisoning system-wide, which security conscious users may choose to do. But it's very difficult to solve otherwise, e.g. note that plaintext file contents may have been read in other places than pagecache pages. Like FS_IOC_ADD_ENCRYPTION_KEY, FS_IOC_REMOVE_ENCRYPTION_KEY is initially restricted to privileged users only. This is sufficient for some use cases, but not all. A later patch will relax this restriction, but it will require introducing key hashes, among other changes. Reviewed-by: Theodore Ts'o <tytso@mit.edu> Signed-off-by: Eric Biggers <ebiggers@google.com>
2019-08-05 10:35:46 +08:00
/* Some inodes still reference this key; try to evict them. */
err = try_to_lock_encrypted_files(sb, mk);
if (err == -EBUSY) {
status_flags |=
FSCRYPT_KEY_REMOVAL_STATUS_FLAG_FILES_BUSY;
err = 0;
}
}
/*
fscrypt: allow unprivileged users to add/remove keys for v2 policies Allow the FS_IOC_ADD_ENCRYPTION_KEY and FS_IOC_REMOVE_ENCRYPTION_KEY ioctls to be used by non-root users to add and remove encryption keys from the filesystem-level crypto keyrings, subject to limitations. Motivation: while privileged fscrypt key management is sufficient for some users (e.g. Android and Chromium OS, where a privileged process manages all keys), the old API by design also allows non-root users to set up and use encrypted directories, and we don't want to regress on that. Especially, we don't want to force users to continue using the old API, running into the visibility mismatch between files and keyrings and being unable to "lock" encrypted directories. Intuitively, the ioctls have to be privileged since they manipulate filesystem-level state. However, it's actually safe to make them unprivileged if we very carefully enforce some specific limitations. First, each key must be identified by a cryptographic hash so that a user can't add the wrong key for another user's files. For v2 encryption policies, we use the key_identifier for this. v1 policies don't have this, so managing keys for them remains privileged. Second, each key a user adds is charged to their quota for the keyrings service. Thus, a user can't exhaust memory by adding a huge number of keys. By default each non-root user is allowed up to 200 keys; this can be changed using the existing sysctl 'kernel.keys.maxkeys'. Third, if multiple users add the same key, we keep track of those users of the key (of which there remains a single copy), and won't really remove the key, i.e. "lock" the encrypted files, until all those users have removed it. This prevents denial of service attacks that would be possible under simpler schemes, such allowing the first user who added a key to remove it -- since that could be a malicious user who has compromised the key. Of course, encryption keys should be kept secret, but the idea is that using encryption should never be *less* secure than not using encryption, even if your key was compromised. We tolerate that a user will be unable to really remove a key, i.e. unable to "lock" their encrypted files, if another user has added the same key. But in a sense, this is actually a good thing because it will avoid providing a false notion of security where a key appears to have been removed when actually it's still in memory, available to any attacker who compromises the operating system kernel. Reviewed-by: Theodore Ts'o <tytso@mit.edu> Signed-off-by: Eric Biggers <ebiggers@google.com>
2019-08-05 10:35:47 +08:00
* We return 0 if we successfully did something: removed a claim to the
* key, wiped the secret, or tried locking the files again. Users need
* to check the informational status flags if they care whether the key
* has been fully removed including all files locked.
fscrypt: add FS_IOC_REMOVE_ENCRYPTION_KEY ioctl Add a new fscrypt ioctl, FS_IOC_REMOVE_ENCRYPTION_KEY. This ioctl removes an encryption key that was added by FS_IOC_ADD_ENCRYPTION_KEY. It wipes the secret key itself, then "locks" the encrypted files and directories that had been unlocked using that key -- implemented by evicting the relevant dentries and inodes from the VFS caches. The problem this solves is that many fscrypt users want the ability to remove encryption keys, causing the corresponding encrypted directories to appear "locked" (presented in ciphertext form) again. Moreover, users want removing an encryption key to *really* remove it, in the sense that the removed keys cannot be recovered even if kernel memory is compromised, e.g. by the exploit of a kernel security vulnerability or by a physical attack. This is desirable after a user logs out of the system, for example. In many cases users even already assume this to be the case and are surprised to hear when it's not. It is not sufficient to simply unlink the master key from the keyring (or to revoke or invalidate it), since the actual encryption transform objects are still pinned in memory by their inodes. Therefore, to really remove a key we must also evict the relevant inodes. Currently one workaround is to run 'sync && echo 2 > /proc/sys/vm/drop_caches'. But, that evicts all unused inodes in the system rather than just the inodes associated with the key being removed, causing severe performance problems. Moreover, it requires root privileges, so regular users can't "lock" their encrypted files. Another workaround, used in Chromium OS kernels, is to add a new VFS-level ioctl FS_IOC_DROP_CACHE which is a more restricted version of drop_caches that operates on a single super_block. It does: shrink_dcache_sb(sb); invalidate_inodes(sb, false); But it's still a hack. Yet, the major users of filesystem encryption want this feature badly enough that they are actually using these hacks. To properly solve the problem, start maintaining a list of the inodes which have been "unlocked" using each master key. Originally this wasn't possible because the kernel didn't keep track of in-use master keys at all. But, with the ->s_master_keys keyring it is now possible. Then, add an ioctl FS_IOC_REMOVE_ENCRYPTION_KEY. It finds the specified master key in ->s_master_keys, then wipes the secret key itself, which prevents any additional inodes from being unlocked with the key. Then, it syncs the filesystem and evicts the inodes in the key's list. The normal inode eviction code will free and wipe the per-file keys (in ->i_crypt_info). Note that freeing ->i_crypt_info without evicting the inodes was also considered, but would have been racy. Some inodes may still be in use when a master key is removed, and we can't simply revoke random file descriptors, mmap's, etc. Thus, the ioctl simply skips in-use inodes, and returns -EBUSY to indicate that some inodes weren't evicted. The master key *secret* is still removed, but the fscrypt_master_key struct remains to keep track of the remaining inodes. Userspace can then retry the ioctl to evict the remaining inodes. Alternatively, if userspace adds the key again, the refreshed secret will be associated with the existing list of inodes so they remain correctly tracked for future key removals. The ioctl doesn't wipe pagecache pages. Thus, we tolerate that after a kernel compromise some portions of plaintext file contents may still be recoverable from memory. This can be solved by enabling page poisoning system-wide, which security conscious users may choose to do. But it's very difficult to solve otherwise, e.g. note that plaintext file contents may have been read in other places than pagecache pages. Like FS_IOC_ADD_ENCRYPTION_KEY, FS_IOC_REMOVE_ENCRYPTION_KEY is initially restricted to privileged users only. This is sufficient for some use cases, but not all. A later patch will relax this restriction, but it will require introducing key hashes, among other changes. Reviewed-by: Theodore Ts'o <tytso@mit.edu> Signed-off-by: Eric Biggers <ebiggers@google.com>
2019-08-05 10:35:46 +08:00
*/
fscrypt: allow unprivileged users to add/remove keys for v2 policies Allow the FS_IOC_ADD_ENCRYPTION_KEY and FS_IOC_REMOVE_ENCRYPTION_KEY ioctls to be used by non-root users to add and remove encryption keys from the filesystem-level crypto keyrings, subject to limitations. Motivation: while privileged fscrypt key management is sufficient for some users (e.g. Android and Chromium OS, where a privileged process manages all keys), the old API by design also allows non-root users to set up and use encrypted directories, and we don't want to regress on that. Especially, we don't want to force users to continue using the old API, running into the visibility mismatch between files and keyrings and being unable to "lock" encrypted directories. Intuitively, the ioctls have to be privileged since they manipulate filesystem-level state. However, it's actually safe to make them unprivileged if we very carefully enforce some specific limitations. First, each key must be identified by a cryptographic hash so that a user can't add the wrong key for another user's files. For v2 encryption policies, we use the key_identifier for this. v1 policies don't have this, so managing keys for them remains privileged. Second, each key a user adds is charged to their quota for the keyrings service. Thus, a user can't exhaust memory by adding a huge number of keys. By default each non-root user is allowed up to 200 keys; this can be changed using the existing sysctl 'kernel.keys.maxkeys'. Third, if multiple users add the same key, we keep track of those users of the key (of which there remains a single copy), and won't really remove the key, i.e. "lock" the encrypted files, until all those users have removed it. This prevents denial of service attacks that would be possible under simpler schemes, such allowing the first user who added a key to remove it -- since that could be a malicious user who has compromised the key. Of course, encryption keys should be kept secret, but the idea is that using encryption should never be *less* secure than not using encryption, even if your key was compromised. We tolerate that a user will be unable to really remove a key, i.e. unable to "lock" their encrypted files, if another user has added the same key. But in a sense, this is actually a good thing because it will avoid providing a false notion of security where a key appears to have been removed when actually it's still in memory, available to any attacker who compromises the operating system kernel. Reviewed-by: Theodore Ts'o <tytso@mit.edu> Signed-off-by: Eric Biggers <ebiggers@google.com>
2019-08-05 10:35:47 +08:00
out_put_key:
fscrypt: stop using keyrings subsystem for fscrypt_master_key The approach of fs/crypto/ internally managing the fscrypt_master_key structs as the payloads of "struct key" objects contained in a "struct key" keyring has outlived its usefulness. The original idea was to simplify the code by reusing code from the keyrings subsystem. However, several issues have arisen that can't easily be resolved: - When a master key struct is destroyed, blk_crypto_evict_key() must be called on any per-mode keys embedded in it. (This started being the case when inline encryption support was added.) Yet, the keyrings subsystem can arbitrarily delay the destruction of keys, even past the time the filesystem was unmounted. Therefore, currently there is no easy way to call blk_crypto_evict_key() when a master key is destroyed. Currently, this is worked around by holding an extra reference to the filesystem's request_queue(s). But it was overlooked that the request_queue reference is *not* guaranteed to pin the corresponding blk_crypto_profile too; for device-mapper devices that support inline crypto, it doesn't. This can cause a use-after-free. - When the last inode that was using an incompletely-removed master key is evicted, the master key removal is completed by removing the key struct from the keyring. Currently this is done via key_invalidate(). Yet, key_invalidate() takes the key semaphore. This can deadlock when called from the shrinker, since in fscrypt_ioctl_add_key(), memory is allocated with GFP_KERNEL under the same semaphore. - More generally, the fact that the keyrings subsystem can arbitrarily delay the destruction of keys (via garbage collection delay, or via random processes getting temporary key references) is undesirable, as it means we can't strictly guarantee that all secrets are ever wiped. - Doing the master key lookups via the keyrings subsystem results in the key_permission LSM hook being called. fscrypt doesn't want this, as all access control for encrypted files is designed to happen via the files themselves, like any other files. The workaround which SELinux users are using is to change their SELinux policy to grant key search access to all domains. This works, but it is an odd extra step that shouldn't really have to be done. The fix for all these issues is to change the implementation to what I should have done originally: don't use the keyrings subsystem to keep track of the filesystem's fscrypt_master_key structs. Instead, just store them in a regular kernel data structure, and rework the reference counting, locking, and lifetime accordingly. Retain support for RCU-mode key lookups by using a hash table. Replace fscrypt_sb_free() with fscrypt_sb_delete(), which releases the keys synchronously and runs a bit earlier during unmount, so that block devices are still available. A side effect of this patch is that neither the master keys themselves nor the filesystem keyrings will be listed in /proc/keys anymore. ("Master key users" and the master key users keyrings will still be listed.) However, this was mostly an implementation detail, and it was intended just for debugging purposes. I don't know of anyone using it. This patch does *not* change how "master key users" (->mk_users) works; that still uses the keyrings subsystem. That is still needed for key quotas, and changing that isn't necessary to solve the issues listed above. If we decide to change that too, it would be a separate patch. I've marked this as fixing the original commit that added the fscrypt keyring, but as noted above the most important issue that this patch fixes wasn't introduced until the addition of inline encryption support. Fixes: 22d94f493bfb ("fscrypt: add FS_IOC_ADD_ENCRYPTION_KEY ioctl") Signed-off-by: Eric Biggers <ebiggers@google.com> Link: https://lore.kernel.org/r/20220901193208.138056-2-ebiggers@kernel.org
2022-09-02 03:32:06 +08:00
fscrypt_put_master_key(mk);
fscrypt: add FS_IOC_REMOVE_ENCRYPTION_KEY ioctl Add a new fscrypt ioctl, FS_IOC_REMOVE_ENCRYPTION_KEY. This ioctl removes an encryption key that was added by FS_IOC_ADD_ENCRYPTION_KEY. It wipes the secret key itself, then "locks" the encrypted files and directories that had been unlocked using that key -- implemented by evicting the relevant dentries and inodes from the VFS caches. The problem this solves is that many fscrypt users want the ability to remove encryption keys, causing the corresponding encrypted directories to appear "locked" (presented in ciphertext form) again. Moreover, users want removing an encryption key to *really* remove it, in the sense that the removed keys cannot be recovered even if kernel memory is compromised, e.g. by the exploit of a kernel security vulnerability or by a physical attack. This is desirable after a user logs out of the system, for example. In many cases users even already assume this to be the case and are surprised to hear when it's not. It is not sufficient to simply unlink the master key from the keyring (or to revoke or invalidate it), since the actual encryption transform objects are still pinned in memory by their inodes. Therefore, to really remove a key we must also evict the relevant inodes. Currently one workaround is to run 'sync && echo 2 > /proc/sys/vm/drop_caches'. But, that evicts all unused inodes in the system rather than just the inodes associated with the key being removed, causing severe performance problems. Moreover, it requires root privileges, so regular users can't "lock" their encrypted files. Another workaround, used in Chromium OS kernels, is to add a new VFS-level ioctl FS_IOC_DROP_CACHE which is a more restricted version of drop_caches that operates on a single super_block. It does: shrink_dcache_sb(sb); invalidate_inodes(sb, false); But it's still a hack. Yet, the major users of filesystem encryption want this feature badly enough that they are actually using these hacks. To properly solve the problem, start maintaining a list of the inodes which have been "unlocked" using each master key. Originally this wasn't possible because the kernel didn't keep track of in-use master keys at all. But, with the ->s_master_keys keyring it is now possible. Then, add an ioctl FS_IOC_REMOVE_ENCRYPTION_KEY. It finds the specified master key in ->s_master_keys, then wipes the secret key itself, which prevents any additional inodes from being unlocked with the key. Then, it syncs the filesystem and evicts the inodes in the key's list. The normal inode eviction code will free and wipe the per-file keys (in ->i_crypt_info). Note that freeing ->i_crypt_info without evicting the inodes was also considered, but would have been racy. Some inodes may still be in use when a master key is removed, and we can't simply revoke random file descriptors, mmap's, etc. Thus, the ioctl simply skips in-use inodes, and returns -EBUSY to indicate that some inodes weren't evicted. The master key *secret* is still removed, but the fscrypt_master_key struct remains to keep track of the remaining inodes. Userspace can then retry the ioctl to evict the remaining inodes. Alternatively, if userspace adds the key again, the refreshed secret will be associated with the existing list of inodes so they remain correctly tracked for future key removals. The ioctl doesn't wipe pagecache pages. Thus, we tolerate that after a kernel compromise some portions of plaintext file contents may still be recoverable from memory. This can be solved by enabling page poisoning system-wide, which security conscious users may choose to do. But it's very difficult to solve otherwise, e.g. note that plaintext file contents may have been read in other places than pagecache pages. Like FS_IOC_ADD_ENCRYPTION_KEY, FS_IOC_REMOVE_ENCRYPTION_KEY is initially restricted to privileged users only. This is sufficient for some use cases, but not all. A later patch will relax this restriction, but it will require introducing key hashes, among other changes. Reviewed-by: Theodore Ts'o <tytso@mit.edu> Signed-off-by: Eric Biggers <ebiggers@google.com>
2019-08-05 10:35:46 +08:00
if (err == 0)
err = put_user(status_flags, &uarg->removal_status_flags);
return err;
}
fscrypt: add FS_IOC_REMOVE_ENCRYPTION_KEY_ALL_USERS ioctl Add a root-only variant of the FS_IOC_REMOVE_ENCRYPTION_KEY ioctl which removes all users' claims of the key, not just the current user's claim. I.e., it always removes the key itself, no matter how many users have added it. This is useful for forcing a directory to be locked, without having to figure out which user ID(s) the key was added under. This is planned to be used by a command like 'sudo fscrypt lock DIR --all-users' in the fscrypt userspace tool (http://github.com/google/fscrypt). Reviewed-by: Theodore Ts'o <tytso@mit.edu> Signed-off-by: Eric Biggers <ebiggers@google.com>
2019-08-05 10:35:47 +08:00
int fscrypt_ioctl_remove_key(struct file *filp, void __user *uarg)
{
return do_remove_key(filp, uarg, false);
}
fscrypt: add FS_IOC_REMOVE_ENCRYPTION_KEY ioctl Add a new fscrypt ioctl, FS_IOC_REMOVE_ENCRYPTION_KEY. This ioctl removes an encryption key that was added by FS_IOC_ADD_ENCRYPTION_KEY. It wipes the secret key itself, then "locks" the encrypted files and directories that had been unlocked using that key -- implemented by evicting the relevant dentries and inodes from the VFS caches. The problem this solves is that many fscrypt users want the ability to remove encryption keys, causing the corresponding encrypted directories to appear "locked" (presented in ciphertext form) again. Moreover, users want removing an encryption key to *really* remove it, in the sense that the removed keys cannot be recovered even if kernel memory is compromised, e.g. by the exploit of a kernel security vulnerability or by a physical attack. This is desirable after a user logs out of the system, for example. In many cases users even already assume this to be the case and are surprised to hear when it's not. It is not sufficient to simply unlink the master key from the keyring (or to revoke or invalidate it), since the actual encryption transform objects are still pinned in memory by their inodes. Therefore, to really remove a key we must also evict the relevant inodes. Currently one workaround is to run 'sync && echo 2 > /proc/sys/vm/drop_caches'. But, that evicts all unused inodes in the system rather than just the inodes associated with the key being removed, causing severe performance problems. Moreover, it requires root privileges, so regular users can't "lock" their encrypted files. Another workaround, used in Chromium OS kernels, is to add a new VFS-level ioctl FS_IOC_DROP_CACHE which is a more restricted version of drop_caches that operates on a single super_block. It does: shrink_dcache_sb(sb); invalidate_inodes(sb, false); But it's still a hack. Yet, the major users of filesystem encryption want this feature badly enough that they are actually using these hacks. To properly solve the problem, start maintaining a list of the inodes which have been "unlocked" using each master key. Originally this wasn't possible because the kernel didn't keep track of in-use master keys at all. But, with the ->s_master_keys keyring it is now possible. Then, add an ioctl FS_IOC_REMOVE_ENCRYPTION_KEY. It finds the specified master key in ->s_master_keys, then wipes the secret key itself, which prevents any additional inodes from being unlocked with the key. Then, it syncs the filesystem and evicts the inodes in the key's list. The normal inode eviction code will free and wipe the per-file keys (in ->i_crypt_info). Note that freeing ->i_crypt_info without evicting the inodes was also considered, but would have been racy. Some inodes may still be in use when a master key is removed, and we can't simply revoke random file descriptors, mmap's, etc. Thus, the ioctl simply skips in-use inodes, and returns -EBUSY to indicate that some inodes weren't evicted. The master key *secret* is still removed, but the fscrypt_master_key struct remains to keep track of the remaining inodes. Userspace can then retry the ioctl to evict the remaining inodes. Alternatively, if userspace adds the key again, the refreshed secret will be associated with the existing list of inodes so they remain correctly tracked for future key removals. The ioctl doesn't wipe pagecache pages. Thus, we tolerate that after a kernel compromise some portions of plaintext file contents may still be recoverable from memory. This can be solved by enabling page poisoning system-wide, which security conscious users may choose to do. But it's very difficult to solve otherwise, e.g. note that plaintext file contents may have been read in other places than pagecache pages. Like FS_IOC_ADD_ENCRYPTION_KEY, FS_IOC_REMOVE_ENCRYPTION_KEY is initially restricted to privileged users only. This is sufficient for some use cases, but not all. A later patch will relax this restriction, but it will require introducing key hashes, among other changes. Reviewed-by: Theodore Ts'o <tytso@mit.edu> Signed-off-by: Eric Biggers <ebiggers@google.com>
2019-08-05 10:35:46 +08:00
EXPORT_SYMBOL_GPL(fscrypt_ioctl_remove_key);
fscrypt: add FS_IOC_REMOVE_ENCRYPTION_KEY_ALL_USERS ioctl Add a root-only variant of the FS_IOC_REMOVE_ENCRYPTION_KEY ioctl which removes all users' claims of the key, not just the current user's claim. I.e., it always removes the key itself, no matter how many users have added it. This is useful for forcing a directory to be locked, without having to figure out which user ID(s) the key was added under. This is planned to be used by a command like 'sudo fscrypt lock DIR --all-users' in the fscrypt userspace tool (http://github.com/google/fscrypt). Reviewed-by: Theodore Ts'o <tytso@mit.edu> Signed-off-by: Eric Biggers <ebiggers@google.com>
2019-08-05 10:35:47 +08:00
int fscrypt_ioctl_remove_key_all_users(struct file *filp, void __user *uarg)
{
if (!capable(CAP_SYS_ADMIN))
return -EACCES;
return do_remove_key(filp, uarg, true);
}
EXPORT_SYMBOL_GPL(fscrypt_ioctl_remove_key_all_users);
fscrypt: add FS_IOC_GET_ENCRYPTION_KEY_STATUS ioctl Add a new fscrypt ioctl, FS_IOC_GET_ENCRYPTION_KEY_STATUS. Given a key specified by 'struct fscrypt_key_specifier' (the same way a key is specified for the other fscrypt key management ioctls), it returns status information in a 'struct fscrypt_get_key_status_arg'. The main motivation for this is that applications need to be able to check whether an encrypted directory is "unlocked" or not, so that they can add the key if it is not, and avoid adding the key (which may involve prompting the user for a passphrase) if it already is. It's possible to use some workarounds such as checking whether opening a regular file fails with ENOKEY, or checking whether the filenames "look like gibberish" or not. However, no workaround is usable in all cases. Like the other key management ioctls, the keyrings syscalls may seem at first to be a good fit for this. Unfortunately, they are not. Even if we exposed the keyring ID of the ->s_master_keys keyring and gave everyone Search permission on it (note: currently the keyrings permission system would also allow everyone to "invalidate" the keyring too), the fscrypt keys have an additional state that doesn't map cleanly to the keyrings API: the secret can be removed, but we can be still tracking the files that were using the key, and the removal can be re-attempted or the secret added again. After later patches, some applications will also need a way to determine whether a key was added by the current user vs. by some other user. Reserved fields are included in fscrypt_get_key_status_arg for this and other future extensions. Reviewed-by: Theodore Ts'o <tytso@mit.edu> Signed-off-by: Eric Biggers <ebiggers@google.com>
2019-08-05 10:35:46 +08:00
/*
* Retrieve the status of an fscrypt master encryption key.
*
* We set ->status to indicate whether the key is absent, present, or
* incompletely removed. "Incompletely removed" means that the master key
* secret has been removed, but some files which had been unlocked with it are
* still in use. This field allows applications to easily determine the state
* of an encrypted directory without using a hack such as trying to open a
* regular file in it (which can confuse the "incompletely removed" state with
* absent or present).
*
fscrypt: allow unprivileged users to add/remove keys for v2 policies Allow the FS_IOC_ADD_ENCRYPTION_KEY and FS_IOC_REMOVE_ENCRYPTION_KEY ioctls to be used by non-root users to add and remove encryption keys from the filesystem-level crypto keyrings, subject to limitations. Motivation: while privileged fscrypt key management is sufficient for some users (e.g. Android and Chromium OS, where a privileged process manages all keys), the old API by design also allows non-root users to set up and use encrypted directories, and we don't want to regress on that. Especially, we don't want to force users to continue using the old API, running into the visibility mismatch between files and keyrings and being unable to "lock" encrypted directories. Intuitively, the ioctls have to be privileged since they manipulate filesystem-level state. However, it's actually safe to make them unprivileged if we very carefully enforce some specific limitations. First, each key must be identified by a cryptographic hash so that a user can't add the wrong key for another user's files. For v2 encryption policies, we use the key_identifier for this. v1 policies don't have this, so managing keys for them remains privileged. Second, each key a user adds is charged to their quota for the keyrings service. Thus, a user can't exhaust memory by adding a huge number of keys. By default each non-root user is allowed up to 200 keys; this can be changed using the existing sysctl 'kernel.keys.maxkeys'. Third, if multiple users add the same key, we keep track of those users of the key (of which there remains a single copy), and won't really remove the key, i.e. "lock" the encrypted files, until all those users have removed it. This prevents denial of service attacks that would be possible under simpler schemes, such allowing the first user who added a key to remove it -- since that could be a malicious user who has compromised the key. Of course, encryption keys should be kept secret, but the idea is that using encryption should never be *less* secure than not using encryption, even if your key was compromised. We tolerate that a user will be unable to really remove a key, i.e. unable to "lock" their encrypted files, if another user has added the same key. But in a sense, this is actually a good thing because it will avoid providing a false notion of security where a key appears to have been removed when actually it's still in memory, available to any attacker who compromises the operating system kernel. Reviewed-by: Theodore Ts'o <tytso@mit.edu> Signed-off-by: Eric Biggers <ebiggers@google.com>
2019-08-05 10:35:47 +08:00
* In addition, for v2 policy keys we allow applications to determine, via
* ->status_flags and ->user_count, whether the key has been added by the
* current user, by other users, or by both. Most applications should not need
* this, since ordinarily only one user should know a given key. However, if a
* secret key is shared by multiple users, applications may wish to add an
* already-present key to prevent other users from removing it. This ioctl can
* be used to check whether that really is the case before the work is done to
* add the key --- which might e.g. require prompting the user for a passphrase.
*
fscrypt: add FS_IOC_GET_ENCRYPTION_KEY_STATUS ioctl Add a new fscrypt ioctl, FS_IOC_GET_ENCRYPTION_KEY_STATUS. Given a key specified by 'struct fscrypt_key_specifier' (the same way a key is specified for the other fscrypt key management ioctls), it returns status information in a 'struct fscrypt_get_key_status_arg'. The main motivation for this is that applications need to be able to check whether an encrypted directory is "unlocked" or not, so that they can add the key if it is not, and avoid adding the key (which may involve prompting the user for a passphrase) if it already is. It's possible to use some workarounds such as checking whether opening a regular file fails with ENOKEY, or checking whether the filenames "look like gibberish" or not. However, no workaround is usable in all cases. Like the other key management ioctls, the keyrings syscalls may seem at first to be a good fit for this. Unfortunately, they are not. Even if we exposed the keyring ID of the ->s_master_keys keyring and gave everyone Search permission on it (note: currently the keyrings permission system would also allow everyone to "invalidate" the keyring too), the fscrypt keys have an additional state that doesn't map cleanly to the keyrings API: the secret can be removed, but we can be still tracking the files that were using the key, and the removal can be re-attempted or the secret added again. After later patches, some applications will also need a way to determine whether a key was added by the current user vs. by some other user. Reserved fields are included in fscrypt_get_key_status_arg for this and other future extensions. Reviewed-by: Theodore Ts'o <tytso@mit.edu> Signed-off-by: Eric Biggers <ebiggers@google.com>
2019-08-05 10:35:46 +08:00
* For more details, see the "FS_IOC_GET_ENCRYPTION_KEY_STATUS" section of
* Documentation/filesystems/fscrypt.rst.
*/
int fscrypt_ioctl_get_key_status(struct file *filp, void __user *uarg)
{
struct super_block *sb = file_inode(filp)->i_sb;
struct fscrypt_get_key_status_arg arg;
struct fscrypt_master_key *mk;
int err;
if (copy_from_user(&arg, uarg, sizeof(arg)))
return -EFAULT;
if (!valid_key_spec(&arg.key_spec))
return -EINVAL;
if (memchr_inv(arg.__reserved, 0, sizeof(arg.__reserved)))
return -EINVAL;
fscrypt: allow unprivileged users to add/remove keys for v2 policies Allow the FS_IOC_ADD_ENCRYPTION_KEY and FS_IOC_REMOVE_ENCRYPTION_KEY ioctls to be used by non-root users to add and remove encryption keys from the filesystem-level crypto keyrings, subject to limitations. Motivation: while privileged fscrypt key management is sufficient for some users (e.g. Android and Chromium OS, where a privileged process manages all keys), the old API by design also allows non-root users to set up and use encrypted directories, and we don't want to regress on that. Especially, we don't want to force users to continue using the old API, running into the visibility mismatch between files and keyrings and being unable to "lock" encrypted directories. Intuitively, the ioctls have to be privileged since they manipulate filesystem-level state. However, it's actually safe to make them unprivileged if we very carefully enforce some specific limitations. First, each key must be identified by a cryptographic hash so that a user can't add the wrong key for another user's files. For v2 encryption policies, we use the key_identifier for this. v1 policies don't have this, so managing keys for them remains privileged. Second, each key a user adds is charged to their quota for the keyrings service. Thus, a user can't exhaust memory by adding a huge number of keys. By default each non-root user is allowed up to 200 keys; this can be changed using the existing sysctl 'kernel.keys.maxkeys'. Third, if multiple users add the same key, we keep track of those users of the key (of which there remains a single copy), and won't really remove the key, i.e. "lock" the encrypted files, until all those users have removed it. This prevents denial of service attacks that would be possible under simpler schemes, such allowing the first user who added a key to remove it -- since that could be a malicious user who has compromised the key. Of course, encryption keys should be kept secret, but the idea is that using encryption should never be *less* secure than not using encryption, even if your key was compromised. We tolerate that a user will be unable to really remove a key, i.e. unable to "lock" their encrypted files, if another user has added the same key. But in a sense, this is actually a good thing because it will avoid providing a false notion of security where a key appears to have been removed when actually it's still in memory, available to any attacker who compromises the operating system kernel. Reviewed-by: Theodore Ts'o <tytso@mit.edu> Signed-off-by: Eric Biggers <ebiggers@google.com>
2019-08-05 10:35:47 +08:00
arg.status_flags = 0;
arg.user_count = 0;
fscrypt: add FS_IOC_GET_ENCRYPTION_KEY_STATUS ioctl Add a new fscrypt ioctl, FS_IOC_GET_ENCRYPTION_KEY_STATUS. Given a key specified by 'struct fscrypt_key_specifier' (the same way a key is specified for the other fscrypt key management ioctls), it returns status information in a 'struct fscrypt_get_key_status_arg'. The main motivation for this is that applications need to be able to check whether an encrypted directory is "unlocked" or not, so that they can add the key if it is not, and avoid adding the key (which may involve prompting the user for a passphrase) if it already is. It's possible to use some workarounds such as checking whether opening a regular file fails with ENOKEY, or checking whether the filenames "look like gibberish" or not. However, no workaround is usable in all cases. Like the other key management ioctls, the keyrings syscalls may seem at first to be a good fit for this. Unfortunately, they are not. Even if we exposed the keyring ID of the ->s_master_keys keyring and gave everyone Search permission on it (note: currently the keyrings permission system would also allow everyone to "invalidate" the keyring too), the fscrypt keys have an additional state that doesn't map cleanly to the keyrings API: the secret can be removed, but we can be still tracking the files that were using the key, and the removal can be re-attempted or the secret added again. After later patches, some applications will also need a way to determine whether a key was added by the current user vs. by some other user. Reserved fields are included in fscrypt_get_key_status_arg for this and other future extensions. Reviewed-by: Theodore Ts'o <tytso@mit.edu> Signed-off-by: Eric Biggers <ebiggers@google.com>
2019-08-05 10:35:46 +08:00
memset(arg.__out_reserved, 0, sizeof(arg.__out_reserved));
fscrypt: stop using keyrings subsystem for fscrypt_master_key The approach of fs/crypto/ internally managing the fscrypt_master_key structs as the payloads of "struct key" objects contained in a "struct key" keyring has outlived its usefulness. The original idea was to simplify the code by reusing code from the keyrings subsystem. However, several issues have arisen that can't easily be resolved: - When a master key struct is destroyed, blk_crypto_evict_key() must be called on any per-mode keys embedded in it. (This started being the case when inline encryption support was added.) Yet, the keyrings subsystem can arbitrarily delay the destruction of keys, even past the time the filesystem was unmounted. Therefore, currently there is no easy way to call blk_crypto_evict_key() when a master key is destroyed. Currently, this is worked around by holding an extra reference to the filesystem's request_queue(s). But it was overlooked that the request_queue reference is *not* guaranteed to pin the corresponding blk_crypto_profile too; for device-mapper devices that support inline crypto, it doesn't. This can cause a use-after-free. - When the last inode that was using an incompletely-removed master key is evicted, the master key removal is completed by removing the key struct from the keyring. Currently this is done via key_invalidate(). Yet, key_invalidate() takes the key semaphore. This can deadlock when called from the shrinker, since in fscrypt_ioctl_add_key(), memory is allocated with GFP_KERNEL under the same semaphore. - More generally, the fact that the keyrings subsystem can arbitrarily delay the destruction of keys (via garbage collection delay, or via random processes getting temporary key references) is undesirable, as it means we can't strictly guarantee that all secrets are ever wiped. - Doing the master key lookups via the keyrings subsystem results in the key_permission LSM hook being called. fscrypt doesn't want this, as all access control for encrypted files is designed to happen via the files themselves, like any other files. The workaround which SELinux users are using is to change their SELinux policy to grant key search access to all domains. This works, but it is an odd extra step that shouldn't really have to be done. The fix for all these issues is to change the implementation to what I should have done originally: don't use the keyrings subsystem to keep track of the filesystem's fscrypt_master_key structs. Instead, just store them in a regular kernel data structure, and rework the reference counting, locking, and lifetime accordingly. Retain support for RCU-mode key lookups by using a hash table. Replace fscrypt_sb_free() with fscrypt_sb_delete(), which releases the keys synchronously and runs a bit earlier during unmount, so that block devices are still available. A side effect of this patch is that neither the master keys themselves nor the filesystem keyrings will be listed in /proc/keys anymore. ("Master key users" and the master key users keyrings will still be listed.) However, this was mostly an implementation detail, and it was intended just for debugging purposes. I don't know of anyone using it. This patch does *not* change how "master key users" (->mk_users) works; that still uses the keyrings subsystem. That is still needed for key quotas, and changing that isn't necessary to solve the issues listed above. If we decide to change that too, it would be a separate patch. I've marked this as fixing the original commit that added the fscrypt keyring, but as noted above the most important issue that this patch fixes wasn't introduced until the addition of inline encryption support. Fixes: 22d94f493bfb ("fscrypt: add FS_IOC_ADD_ENCRYPTION_KEY ioctl") Signed-off-by: Eric Biggers <ebiggers@google.com> Link: https://lore.kernel.org/r/20220901193208.138056-2-ebiggers@kernel.org
2022-09-02 03:32:06 +08:00
mk = fscrypt_find_master_key(sb, &arg.key_spec);
if (!mk) {
fscrypt: add FS_IOC_GET_ENCRYPTION_KEY_STATUS ioctl Add a new fscrypt ioctl, FS_IOC_GET_ENCRYPTION_KEY_STATUS. Given a key specified by 'struct fscrypt_key_specifier' (the same way a key is specified for the other fscrypt key management ioctls), it returns status information in a 'struct fscrypt_get_key_status_arg'. The main motivation for this is that applications need to be able to check whether an encrypted directory is "unlocked" or not, so that they can add the key if it is not, and avoid adding the key (which may involve prompting the user for a passphrase) if it already is. It's possible to use some workarounds such as checking whether opening a regular file fails with ENOKEY, or checking whether the filenames "look like gibberish" or not. However, no workaround is usable in all cases. Like the other key management ioctls, the keyrings syscalls may seem at first to be a good fit for this. Unfortunately, they are not. Even if we exposed the keyring ID of the ->s_master_keys keyring and gave everyone Search permission on it (note: currently the keyrings permission system would also allow everyone to "invalidate" the keyring too), the fscrypt keys have an additional state that doesn't map cleanly to the keyrings API: the secret can be removed, but we can be still tracking the files that were using the key, and the removal can be re-attempted or the secret added again. After later patches, some applications will also need a way to determine whether a key was added by the current user vs. by some other user. Reserved fields are included in fscrypt_get_key_status_arg for this and other future extensions. Reviewed-by: Theodore Ts'o <tytso@mit.edu> Signed-off-by: Eric Biggers <ebiggers@google.com>
2019-08-05 10:35:46 +08:00
arg.status = FSCRYPT_KEY_STATUS_ABSENT;
err = 0;
goto out;
}
fscrypt: stop using keyrings subsystem for fscrypt_master_key The approach of fs/crypto/ internally managing the fscrypt_master_key structs as the payloads of "struct key" objects contained in a "struct key" keyring has outlived its usefulness. The original idea was to simplify the code by reusing code from the keyrings subsystem. However, several issues have arisen that can't easily be resolved: - When a master key struct is destroyed, blk_crypto_evict_key() must be called on any per-mode keys embedded in it. (This started being the case when inline encryption support was added.) Yet, the keyrings subsystem can arbitrarily delay the destruction of keys, even past the time the filesystem was unmounted. Therefore, currently there is no easy way to call blk_crypto_evict_key() when a master key is destroyed. Currently, this is worked around by holding an extra reference to the filesystem's request_queue(s). But it was overlooked that the request_queue reference is *not* guaranteed to pin the corresponding blk_crypto_profile too; for device-mapper devices that support inline crypto, it doesn't. This can cause a use-after-free. - When the last inode that was using an incompletely-removed master key is evicted, the master key removal is completed by removing the key struct from the keyring. Currently this is done via key_invalidate(). Yet, key_invalidate() takes the key semaphore. This can deadlock when called from the shrinker, since in fscrypt_ioctl_add_key(), memory is allocated with GFP_KERNEL under the same semaphore. - More generally, the fact that the keyrings subsystem can arbitrarily delay the destruction of keys (via garbage collection delay, or via random processes getting temporary key references) is undesirable, as it means we can't strictly guarantee that all secrets are ever wiped. - Doing the master key lookups via the keyrings subsystem results in the key_permission LSM hook being called. fscrypt doesn't want this, as all access control for encrypted files is designed to happen via the files themselves, like any other files. The workaround which SELinux users are using is to change their SELinux policy to grant key search access to all domains. This works, but it is an odd extra step that shouldn't really have to be done. The fix for all these issues is to change the implementation to what I should have done originally: don't use the keyrings subsystem to keep track of the filesystem's fscrypt_master_key structs. Instead, just store them in a regular kernel data structure, and rework the reference counting, locking, and lifetime accordingly. Retain support for RCU-mode key lookups by using a hash table. Replace fscrypt_sb_free() with fscrypt_sb_delete(), which releases the keys synchronously and runs a bit earlier during unmount, so that block devices are still available. A side effect of this patch is that neither the master keys themselves nor the filesystem keyrings will be listed in /proc/keys anymore. ("Master key users" and the master key users keyrings will still be listed.) However, this was mostly an implementation detail, and it was intended just for debugging purposes. I don't know of anyone using it. This patch does *not* change how "master key users" (->mk_users) works; that still uses the keyrings subsystem. That is still needed for key quotas, and changing that isn't necessary to solve the issues listed above. If we decide to change that too, it would be a separate patch. I've marked this as fixing the original commit that added the fscrypt keyring, but as noted above the most important issue that this patch fixes wasn't introduced until the addition of inline encryption support. Fixes: 22d94f493bfb ("fscrypt: add FS_IOC_ADD_ENCRYPTION_KEY ioctl") Signed-off-by: Eric Biggers <ebiggers@google.com> Link: https://lore.kernel.org/r/20220901193208.138056-2-ebiggers@kernel.org
2022-09-02 03:32:06 +08:00
down_read(&mk->mk_sem);
fscrypt: add FS_IOC_GET_ENCRYPTION_KEY_STATUS ioctl Add a new fscrypt ioctl, FS_IOC_GET_ENCRYPTION_KEY_STATUS. Given a key specified by 'struct fscrypt_key_specifier' (the same way a key is specified for the other fscrypt key management ioctls), it returns status information in a 'struct fscrypt_get_key_status_arg'. The main motivation for this is that applications need to be able to check whether an encrypted directory is "unlocked" or not, so that they can add the key if it is not, and avoid adding the key (which may involve prompting the user for a passphrase) if it already is. It's possible to use some workarounds such as checking whether opening a regular file fails with ENOKEY, or checking whether the filenames "look like gibberish" or not. However, no workaround is usable in all cases. Like the other key management ioctls, the keyrings syscalls may seem at first to be a good fit for this. Unfortunately, they are not. Even if we exposed the keyring ID of the ->s_master_keys keyring and gave everyone Search permission on it (note: currently the keyrings permission system would also allow everyone to "invalidate" the keyring too), the fscrypt keys have an additional state that doesn't map cleanly to the keyrings API: the secret can be removed, but we can be still tracking the files that were using the key, and the removal can be re-attempted or the secret added again. After later patches, some applications will also need a way to determine whether a key was added by the current user vs. by some other user. Reserved fields are included in fscrypt_get_key_status_arg for this and other future extensions. Reviewed-by: Theodore Ts'o <tytso@mit.edu> Signed-off-by: Eric Biggers <ebiggers@google.com>
2019-08-05 10:35:46 +08:00
if (!is_master_key_secret_present(&mk->mk_secret)) {
fscrypt: stop using keyrings subsystem for fscrypt_master_key The approach of fs/crypto/ internally managing the fscrypt_master_key structs as the payloads of "struct key" objects contained in a "struct key" keyring has outlived its usefulness. The original idea was to simplify the code by reusing code from the keyrings subsystem. However, several issues have arisen that can't easily be resolved: - When a master key struct is destroyed, blk_crypto_evict_key() must be called on any per-mode keys embedded in it. (This started being the case when inline encryption support was added.) Yet, the keyrings subsystem can arbitrarily delay the destruction of keys, even past the time the filesystem was unmounted. Therefore, currently there is no easy way to call blk_crypto_evict_key() when a master key is destroyed. Currently, this is worked around by holding an extra reference to the filesystem's request_queue(s). But it was overlooked that the request_queue reference is *not* guaranteed to pin the corresponding blk_crypto_profile too; for device-mapper devices that support inline crypto, it doesn't. This can cause a use-after-free. - When the last inode that was using an incompletely-removed master key is evicted, the master key removal is completed by removing the key struct from the keyring. Currently this is done via key_invalidate(). Yet, key_invalidate() takes the key semaphore. This can deadlock when called from the shrinker, since in fscrypt_ioctl_add_key(), memory is allocated with GFP_KERNEL under the same semaphore. - More generally, the fact that the keyrings subsystem can arbitrarily delay the destruction of keys (via garbage collection delay, or via random processes getting temporary key references) is undesirable, as it means we can't strictly guarantee that all secrets are ever wiped. - Doing the master key lookups via the keyrings subsystem results in the key_permission LSM hook being called. fscrypt doesn't want this, as all access control for encrypted files is designed to happen via the files themselves, like any other files. The workaround which SELinux users are using is to change their SELinux policy to grant key search access to all domains. This works, but it is an odd extra step that shouldn't really have to be done. The fix for all these issues is to change the implementation to what I should have done originally: don't use the keyrings subsystem to keep track of the filesystem's fscrypt_master_key structs. Instead, just store them in a regular kernel data structure, and rework the reference counting, locking, and lifetime accordingly. Retain support for RCU-mode key lookups by using a hash table. Replace fscrypt_sb_free() with fscrypt_sb_delete(), which releases the keys synchronously and runs a bit earlier during unmount, so that block devices are still available. A side effect of this patch is that neither the master keys themselves nor the filesystem keyrings will be listed in /proc/keys anymore. ("Master key users" and the master key users keyrings will still be listed.) However, this was mostly an implementation detail, and it was intended just for debugging purposes. I don't know of anyone using it. This patch does *not* change how "master key users" (->mk_users) works; that still uses the keyrings subsystem. That is still needed for key quotas, and changing that isn't necessary to solve the issues listed above. If we decide to change that too, it would be a separate patch. I've marked this as fixing the original commit that added the fscrypt keyring, but as noted above the most important issue that this patch fixes wasn't introduced until the addition of inline encryption support. Fixes: 22d94f493bfb ("fscrypt: add FS_IOC_ADD_ENCRYPTION_KEY ioctl") Signed-off-by: Eric Biggers <ebiggers@google.com> Link: https://lore.kernel.org/r/20220901193208.138056-2-ebiggers@kernel.org
2022-09-02 03:32:06 +08:00
arg.status = refcount_read(&mk->mk_active_refs) > 0 ?
FSCRYPT_KEY_STATUS_INCOMPLETELY_REMOVED :
FSCRYPT_KEY_STATUS_ABSENT /* raced with full removal */;
fscrypt: add FS_IOC_GET_ENCRYPTION_KEY_STATUS ioctl Add a new fscrypt ioctl, FS_IOC_GET_ENCRYPTION_KEY_STATUS. Given a key specified by 'struct fscrypt_key_specifier' (the same way a key is specified for the other fscrypt key management ioctls), it returns status information in a 'struct fscrypt_get_key_status_arg'. The main motivation for this is that applications need to be able to check whether an encrypted directory is "unlocked" or not, so that they can add the key if it is not, and avoid adding the key (which may involve prompting the user for a passphrase) if it already is. It's possible to use some workarounds such as checking whether opening a regular file fails with ENOKEY, or checking whether the filenames "look like gibberish" or not. However, no workaround is usable in all cases. Like the other key management ioctls, the keyrings syscalls may seem at first to be a good fit for this. Unfortunately, they are not. Even if we exposed the keyring ID of the ->s_master_keys keyring and gave everyone Search permission on it (note: currently the keyrings permission system would also allow everyone to "invalidate" the keyring too), the fscrypt keys have an additional state that doesn't map cleanly to the keyrings API: the secret can be removed, but we can be still tracking the files that were using the key, and the removal can be re-attempted or the secret added again. After later patches, some applications will also need a way to determine whether a key was added by the current user vs. by some other user. Reserved fields are included in fscrypt_get_key_status_arg for this and other future extensions. Reviewed-by: Theodore Ts'o <tytso@mit.edu> Signed-off-by: Eric Biggers <ebiggers@google.com>
2019-08-05 10:35:46 +08:00
err = 0;
goto out_release_key;
}
arg.status = FSCRYPT_KEY_STATUS_PRESENT;
fscrypt: allow unprivileged users to add/remove keys for v2 policies Allow the FS_IOC_ADD_ENCRYPTION_KEY and FS_IOC_REMOVE_ENCRYPTION_KEY ioctls to be used by non-root users to add and remove encryption keys from the filesystem-level crypto keyrings, subject to limitations. Motivation: while privileged fscrypt key management is sufficient for some users (e.g. Android and Chromium OS, where a privileged process manages all keys), the old API by design also allows non-root users to set up and use encrypted directories, and we don't want to regress on that. Especially, we don't want to force users to continue using the old API, running into the visibility mismatch between files and keyrings and being unable to "lock" encrypted directories. Intuitively, the ioctls have to be privileged since they manipulate filesystem-level state. However, it's actually safe to make them unprivileged if we very carefully enforce some specific limitations. First, each key must be identified by a cryptographic hash so that a user can't add the wrong key for another user's files. For v2 encryption policies, we use the key_identifier for this. v1 policies don't have this, so managing keys for them remains privileged. Second, each key a user adds is charged to their quota for the keyrings service. Thus, a user can't exhaust memory by adding a huge number of keys. By default each non-root user is allowed up to 200 keys; this can be changed using the existing sysctl 'kernel.keys.maxkeys'. Third, if multiple users add the same key, we keep track of those users of the key (of which there remains a single copy), and won't really remove the key, i.e. "lock" the encrypted files, until all those users have removed it. This prevents denial of service attacks that would be possible under simpler schemes, such allowing the first user who added a key to remove it -- since that could be a malicious user who has compromised the key. Of course, encryption keys should be kept secret, but the idea is that using encryption should never be *less* secure than not using encryption, even if your key was compromised. We tolerate that a user will be unable to really remove a key, i.e. unable to "lock" their encrypted files, if another user has added the same key. But in a sense, this is actually a good thing because it will avoid providing a false notion of security where a key appears to have been removed when actually it's still in memory, available to any attacker who compromises the operating system kernel. Reviewed-by: Theodore Ts'o <tytso@mit.edu> Signed-off-by: Eric Biggers <ebiggers@google.com>
2019-08-05 10:35:47 +08:00
if (mk->mk_users) {
struct key *mk_user;
arg.user_count = mk->mk_users->keys.nr_leaves_on_tree;
mk_user = find_master_key_user(mk);
if (!IS_ERR(mk_user)) {
arg.status_flags |=
FSCRYPT_KEY_STATUS_FLAG_ADDED_BY_SELF;
key_put(mk_user);
} else if (mk_user != ERR_PTR(-ENOKEY)) {
err = PTR_ERR(mk_user);
goto out_release_key;
}
}
fscrypt: add FS_IOC_GET_ENCRYPTION_KEY_STATUS ioctl Add a new fscrypt ioctl, FS_IOC_GET_ENCRYPTION_KEY_STATUS. Given a key specified by 'struct fscrypt_key_specifier' (the same way a key is specified for the other fscrypt key management ioctls), it returns status information in a 'struct fscrypt_get_key_status_arg'. The main motivation for this is that applications need to be able to check whether an encrypted directory is "unlocked" or not, so that they can add the key if it is not, and avoid adding the key (which may involve prompting the user for a passphrase) if it already is. It's possible to use some workarounds such as checking whether opening a regular file fails with ENOKEY, or checking whether the filenames "look like gibberish" or not. However, no workaround is usable in all cases. Like the other key management ioctls, the keyrings syscalls may seem at first to be a good fit for this. Unfortunately, they are not. Even if we exposed the keyring ID of the ->s_master_keys keyring and gave everyone Search permission on it (note: currently the keyrings permission system would also allow everyone to "invalidate" the keyring too), the fscrypt keys have an additional state that doesn't map cleanly to the keyrings API: the secret can be removed, but we can be still tracking the files that were using the key, and the removal can be re-attempted or the secret added again. After later patches, some applications will also need a way to determine whether a key was added by the current user vs. by some other user. Reserved fields are included in fscrypt_get_key_status_arg for this and other future extensions. Reviewed-by: Theodore Ts'o <tytso@mit.edu> Signed-off-by: Eric Biggers <ebiggers@google.com>
2019-08-05 10:35:46 +08:00
err = 0;
out_release_key:
fscrypt: stop using keyrings subsystem for fscrypt_master_key The approach of fs/crypto/ internally managing the fscrypt_master_key structs as the payloads of "struct key" objects contained in a "struct key" keyring has outlived its usefulness. The original idea was to simplify the code by reusing code from the keyrings subsystem. However, several issues have arisen that can't easily be resolved: - When a master key struct is destroyed, blk_crypto_evict_key() must be called on any per-mode keys embedded in it. (This started being the case when inline encryption support was added.) Yet, the keyrings subsystem can arbitrarily delay the destruction of keys, even past the time the filesystem was unmounted. Therefore, currently there is no easy way to call blk_crypto_evict_key() when a master key is destroyed. Currently, this is worked around by holding an extra reference to the filesystem's request_queue(s). But it was overlooked that the request_queue reference is *not* guaranteed to pin the corresponding blk_crypto_profile too; for device-mapper devices that support inline crypto, it doesn't. This can cause a use-after-free. - When the last inode that was using an incompletely-removed master key is evicted, the master key removal is completed by removing the key struct from the keyring. Currently this is done via key_invalidate(). Yet, key_invalidate() takes the key semaphore. This can deadlock when called from the shrinker, since in fscrypt_ioctl_add_key(), memory is allocated with GFP_KERNEL under the same semaphore. - More generally, the fact that the keyrings subsystem can arbitrarily delay the destruction of keys (via garbage collection delay, or via random processes getting temporary key references) is undesirable, as it means we can't strictly guarantee that all secrets are ever wiped. - Doing the master key lookups via the keyrings subsystem results in the key_permission LSM hook being called. fscrypt doesn't want this, as all access control for encrypted files is designed to happen via the files themselves, like any other files. The workaround which SELinux users are using is to change their SELinux policy to grant key search access to all domains. This works, but it is an odd extra step that shouldn't really have to be done. The fix for all these issues is to change the implementation to what I should have done originally: don't use the keyrings subsystem to keep track of the filesystem's fscrypt_master_key structs. Instead, just store them in a regular kernel data structure, and rework the reference counting, locking, and lifetime accordingly. Retain support for RCU-mode key lookups by using a hash table. Replace fscrypt_sb_free() with fscrypt_sb_delete(), which releases the keys synchronously and runs a bit earlier during unmount, so that block devices are still available. A side effect of this patch is that neither the master keys themselves nor the filesystem keyrings will be listed in /proc/keys anymore. ("Master key users" and the master key users keyrings will still be listed.) However, this was mostly an implementation detail, and it was intended just for debugging purposes. I don't know of anyone using it. This patch does *not* change how "master key users" (->mk_users) works; that still uses the keyrings subsystem. That is still needed for key quotas, and changing that isn't necessary to solve the issues listed above. If we decide to change that too, it would be a separate patch. I've marked this as fixing the original commit that added the fscrypt keyring, but as noted above the most important issue that this patch fixes wasn't introduced until the addition of inline encryption support. Fixes: 22d94f493bfb ("fscrypt: add FS_IOC_ADD_ENCRYPTION_KEY ioctl") Signed-off-by: Eric Biggers <ebiggers@google.com> Link: https://lore.kernel.org/r/20220901193208.138056-2-ebiggers@kernel.org
2022-09-02 03:32:06 +08:00
up_read(&mk->mk_sem);
fscrypt_put_master_key(mk);
fscrypt: add FS_IOC_GET_ENCRYPTION_KEY_STATUS ioctl Add a new fscrypt ioctl, FS_IOC_GET_ENCRYPTION_KEY_STATUS. Given a key specified by 'struct fscrypt_key_specifier' (the same way a key is specified for the other fscrypt key management ioctls), it returns status information in a 'struct fscrypt_get_key_status_arg'. The main motivation for this is that applications need to be able to check whether an encrypted directory is "unlocked" or not, so that they can add the key if it is not, and avoid adding the key (which may involve prompting the user for a passphrase) if it already is. It's possible to use some workarounds such as checking whether opening a regular file fails with ENOKEY, or checking whether the filenames "look like gibberish" or not. However, no workaround is usable in all cases. Like the other key management ioctls, the keyrings syscalls may seem at first to be a good fit for this. Unfortunately, they are not. Even if we exposed the keyring ID of the ->s_master_keys keyring and gave everyone Search permission on it (note: currently the keyrings permission system would also allow everyone to "invalidate" the keyring too), the fscrypt keys have an additional state that doesn't map cleanly to the keyrings API: the secret can be removed, but we can be still tracking the files that were using the key, and the removal can be re-attempted or the secret added again. After later patches, some applications will also need a way to determine whether a key was added by the current user vs. by some other user. Reserved fields are included in fscrypt_get_key_status_arg for this and other future extensions. Reviewed-by: Theodore Ts'o <tytso@mit.edu> Signed-off-by: Eric Biggers <ebiggers@google.com>
2019-08-05 10:35:46 +08:00
out:
if (!err && copy_to_user(uarg, &arg, sizeof(arg)))
err = -EFAULT;
return err;
}
EXPORT_SYMBOL_GPL(fscrypt_ioctl_get_key_status);
fscrypt: add FS_IOC_ADD_ENCRYPTION_KEY ioctl Add a new fscrypt ioctl, FS_IOC_ADD_ENCRYPTION_KEY. This ioctl adds an encryption key to the filesystem's fscrypt keyring ->s_master_keys, making any files encrypted with that key appear "unlocked". Why we need this ~~~~~~~~~~~~~~~~ The main problem is that the "locked/unlocked" (ciphertext/plaintext) status of encrypted files is global, but the fscrypt keys are not. fscrypt only looks for keys in the keyring(s) the process accessing the filesystem is subscribed to: the thread keyring, process keyring, and session keyring, where the session keyring may contain the user keyring. Therefore, userspace has to put fscrypt keys in the keyrings for individual users or sessions. But this means that when a process with a different keyring tries to access encrypted files, whether they appear "unlocked" or not is nondeterministic. This is because it depends on whether the files are currently present in the inode cache. Fixing this by consistently providing each process its own view of the filesystem depending on whether it has the key or not isn't feasible due to how the VFS caches work. Furthermore, while sometimes users expect this behavior, it is misguided for two reasons. First, it would be an OS-level access control mechanism largely redundant with existing access control mechanisms such as UNIX file permissions, ACLs, LSMs, etc. Encryption is actually for protecting the data at rest. Second, almost all users of fscrypt actually do need the keys to be global. The largest users of fscrypt, Android and Chromium OS, achieve this by having PID 1 create a "session keyring" that is inherited by every process. This works, but it isn't scalable because it prevents session keyrings from being used for any other purpose. On general-purpose Linux distros, the 'fscrypt' userspace tool [1] can't similarly abuse the session keyring, so to make 'sudo' work on all systems it has to link all the user keyrings into root's user keyring [2]. This is ugly and raises security concerns. Moreover it can't make the keys available to system services, such as sshd trying to access the user's '~/.ssh' directory (see [3], [4]) or NetworkManager trying to read certificates from the user's home directory (see [5]); or to Docker containers (see [6], [7]). By having an API to add a key to the *filesystem* we'll be able to fix the above bugs, remove userspace workarounds, and clearly express the intended semantics: the locked/unlocked status of an encrypted directory is global, and encryption is orthogonal to OS-level access control. Why not use the add_key() syscall ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ We use an ioctl for this API rather than the existing add_key() system call because the ioctl gives us the flexibility needed to implement fscrypt-specific semantics that will be introduced in later patches: - Supporting key removal with the semantics such that the secret is removed immediately and any unused inodes using the key are evicted; also, the eviction of any in-use inodes can be retried. - Calculating a key-dependent cryptographic identifier and returning it to userspace. - Allowing keys to be added and removed by non-root users, but only keys for v2 encryption policies; and to prevent denial-of-service attacks, users can only remove keys they themselves have added, and a key is only really removed after all users who added it have removed it. Trying to shoehorn these semantics into the keyrings syscalls would be very difficult, whereas the ioctls make things much easier. However, to reuse code the implementation still uses the keyrings service internally. Thus we get lockless RCU-mode key lookups without having to re-implement it, and the keys automatically show up in /proc/keys for debugging purposes. References: [1] https://github.com/google/fscrypt [2] https://goo.gl/55cCrI#heading=h.vf09isp98isb [3] https://github.com/google/fscrypt/issues/111#issuecomment-444347939 [4] https://github.com/google/fscrypt/issues/116 [5] https://bugs.launchpad.net/ubuntu/+source/fscrypt/+bug/1770715 [6] https://github.com/google/fscrypt/issues/128 [7] https://askubuntu.com/questions/1130306/cannot-run-docker-on-an-encrypted-filesystem Reviewed-by: Theodore Ts'o <tytso@mit.edu> Signed-off-by: Eric Biggers <ebiggers@google.com>
2019-08-05 10:35:46 +08:00
int __init fscrypt_init_keyring(void)
{
fscrypt: allow unprivileged users to add/remove keys for v2 policies Allow the FS_IOC_ADD_ENCRYPTION_KEY and FS_IOC_REMOVE_ENCRYPTION_KEY ioctls to be used by non-root users to add and remove encryption keys from the filesystem-level crypto keyrings, subject to limitations. Motivation: while privileged fscrypt key management is sufficient for some users (e.g. Android and Chromium OS, where a privileged process manages all keys), the old API by design also allows non-root users to set up and use encrypted directories, and we don't want to regress on that. Especially, we don't want to force users to continue using the old API, running into the visibility mismatch between files and keyrings and being unable to "lock" encrypted directories. Intuitively, the ioctls have to be privileged since they manipulate filesystem-level state. However, it's actually safe to make them unprivileged if we very carefully enforce some specific limitations. First, each key must be identified by a cryptographic hash so that a user can't add the wrong key for another user's files. For v2 encryption policies, we use the key_identifier for this. v1 policies don't have this, so managing keys for them remains privileged. Second, each key a user adds is charged to their quota for the keyrings service. Thus, a user can't exhaust memory by adding a huge number of keys. By default each non-root user is allowed up to 200 keys; this can be changed using the existing sysctl 'kernel.keys.maxkeys'. Third, if multiple users add the same key, we keep track of those users of the key (of which there remains a single copy), and won't really remove the key, i.e. "lock" the encrypted files, until all those users have removed it. This prevents denial of service attacks that would be possible under simpler schemes, such allowing the first user who added a key to remove it -- since that could be a malicious user who has compromised the key. Of course, encryption keys should be kept secret, but the idea is that using encryption should never be *less* secure than not using encryption, even if your key was compromised. We tolerate that a user will be unable to really remove a key, i.e. unable to "lock" their encrypted files, if another user has added the same key. But in a sense, this is actually a good thing because it will avoid providing a false notion of security where a key appears to have been removed when actually it's still in memory, available to any attacker who compromises the operating system kernel. Reviewed-by: Theodore Ts'o <tytso@mit.edu> Signed-off-by: Eric Biggers <ebiggers@google.com>
2019-08-05 10:35:47 +08:00
int err;
err = register_key_type(&key_type_fscrypt_user);
if (err)
fscrypt: stop using keyrings subsystem for fscrypt_master_key The approach of fs/crypto/ internally managing the fscrypt_master_key structs as the payloads of "struct key" objects contained in a "struct key" keyring has outlived its usefulness. The original idea was to simplify the code by reusing code from the keyrings subsystem. However, several issues have arisen that can't easily be resolved: - When a master key struct is destroyed, blk_crypto_evict_key() must be called on any per-mode keys embedded in it. (This started being the case when inline encryption support was added.) Yet, the keyrings subsystem can arbitrarily delay the destruction of keys, even past the time the filesystem was unmounted. Therefore, currently there is no easy way to call blk_crypto_evict_key() when a master key is destroyed. Currently, this is worked around by holding an extra reference to the filesystem's request_queue(s). But it was overlooked that the request_queue reference is *not* guaranteed to pin the corresponding blk_crypto_profile too; for device-mapper devices that support inline crypto, it doesn't. This can cause a use-after-free. - When the last inode that was using an incompletely-removed master key is evicted, the master key removal is completed by removing the key struct from the keyring. Currently this is done via key_invalidate(). Yet, key_invalidate() takes the key semaphore. This can deadlock when called from the shrinker, since in fscrypt_ioctl_add_key(), memory is allocated with GFP_KERNEL under the same semaphore. - More generally, the fact that the keyrings subsystem can arbitrarily delay the destruction of keys (via garbage collection delay, or via random processes getting temporary key references) is undesirable, as it means we can't strictly guarantee that all secrets are ever wiped. - Doing the master key lookups via the keyrings subsystem results in the key_permission LSM hook being called. fscrypt doesn't want this, as all access control for encrypted files is designed to happen via the files themselves, like any other files. The workaround which SELinux users are using is to change their SELinux policy to grant key search access to all domains. This works, but it is an odd extra step that shouldn't really have to be done. The fix for all these issues is to change the implementation to what I should have done originally: don't use the keyrings subsystem to keep track of the filesystem's fscrypt_master_key structs. Instead, just store them in a regular kernel data structure, and rework the reference counting, locking, and lifetime accordingly. Retain support for RCU-mode key lookups by using a hash table. Replace fscrypt_sb_free() with fscrypt_sb_delete(), which releases the keys synchronously and runs a bit earlier during unmount, so that block devices are still available. A side effect of this patch is that neither the master keys themselves nor the filesystem keyrings will be listed in /proc/keys anymore. ("Master key users" and the master key users keyrings will still be listed.) However, this was mostly an implementation detail, and it was intended just for debugging purposes. I don't know of anyone using it. This patch does *not* change how "master key users" (->mk_users) works; that still uses the keyrings subsystem. That is still needed for key quotas, and changing that isn't necessary to solve the issues listed above. If we decide to change that too, it would be a separate patch. I've marked this as fixing the original commit that added the fscrypt keyring, but as noted above the most important issue that this patch fixes wasn't introduced until the addition of inline encryption support. Fixes: 22d94f493bfb ("fscrypt: add FS_IOC_ADD_ENCRYPTION_KEY ioctl") Signed-off-by: Eric Biggers <ebiggers@google.com> Link: https://lore.kernel.org/r/20220901193208.138056-2-ebiggers@kernel.org
2022-09-02 03:32:06 +08:00
return err;
fscrypt: allow unprivileged users to add/remove keys for v2 policies Allow the FS_IOC_ADD_ENCRYPTION_KEY and FS_IOC_REMOVE_ENCRYPTION_KEY ioctls to be used by non-root users to add and remove encryption keys from the filesystem-level crypto keyrings, subject to limitations. Motivation: while privileged fscrypt key management is sufficient for some users (e.g. Android and Chromium OS, where a privileged process manages all keys), the old API by design also allows non-root users to set up and use encrypted directories, and we don't want to regress on that. Especially, we don't want to force users to continue using the old API, running into the visibility mismatch between files and keyrings and being unable to "lock" encrypted directories. Intuitively, the ioctls have to be privileged since they manipulate filesystem-level state. However, it's actually safe to make them unprivileged if we very carefully enforce some specific limitations. First, each key must be identified by a cryptographic hash so that a user can't add the wrong key for another user's files. For v2 encryption policies, we use the key_identifier for this. v1 policies don't have this, so managing keys for them remains privileged. Second, each key a user adds is charged to their quota for the keyrings service. Thus, a user can't exhaust memory by adding a huge number of keys. By default each non-root user is allowed up to 200 keys; this can be changed using the existing sysctl 'kernel.keys.maxkeys'. Third, if multiple users add the same key, we keep track of those users of the key (of which there remains a single copy), and won't really remove the key, i.e. "lock" the encrypted files, until all those users have removed it. This prevents denial of service attacks that would be possible under simpler schemes, such allowing the first user who added a key to remove it -- since that could be a malicious user who has compromised the key. Of course, encryption keys should be kept secret, but the idea is that using encryption should never be *less* secure than not using encryption, even if your key was compromised. We tolerate that a user will be unable to really remove a key, i.e. unable to "lock" their encrypted files, if another user has added the same key. But in a sense, this is actually a good thing because it will avoid providing a false notion of security where a key appears to have been removed when actually it's still in memory, available to any attacker who compromises the operating system kernel. Reviewed-by: Theodore Ts'o <tytso@mit.edu> Signed-off-by: Eric Biggers <ebiggers@google.com>
2019-08-05 10:35:47 +08:00
fscrypt: support passing a keyring key to FS_IOC_ADD_ENCRYPTION_KEY Extend the FS_IOC_ADD_ENCRYPTION_KEY ioctl to allow the raw key to be specified by a Linux keyring key, rather than specified directly. This is useful because fscrypt keys belong to a particular filesystem instance, so they are destroyed when that filesystem is unmounted. Usually this is desired. But in some cases, userspace may need to unmount and re-mount the filesystem while keeping the keys, e.g. during a system update. This requires keeping the keys somewhere else too. The keys could be kept in memory in a userspace daemon. But depending on the security architecture and assumptions, it can be preferable to keep them only in kernel memory, where they are unreadable by userspace. We also can't solve this by going back to the original fscrypt API (where for each file, the master key was looked up in the process's keyring hierarchy) because that caused lots of problems of its own. Therefore, add the ability for FS_IOC_ADD_ENCRYPTION_KEY to accept a Linux keyring key. This solves the problem by allowing userspace to (if needed) save the keys securely in a Linux keyring for re-provisioning, while still using the new fscrypt key management ioctls. This is analogous to how dm-crypt accepts a Linux keyring key, but the key is then stored internally in the dm-crypt data structures rather than being looked up again each time the dm-crypt device is accessed. Use a custom key type "fscrypt-provisioning" rather than one of the existing key types such as "logon". This is strongly desired because it enforces that these keys are only usable for a particular purpose: for fscrypt as input to a particular KDF. Otherwise, the keys could also be passed to any kernel API that accepts a "logon" key with any service prefix, e.g. dm-crypt, UBIFS, or (recently proposed) AF_ALG. This would risk leaking information about the raw key despite it ostensibly being unreadable. Of course, this mistake has already been made for multiple kernel APIs; but since this is a new API, let's do it right. This patch has been tested using an xfstest which I wrote to test it. Link: https://lore.kernel.org/r/20191119222447.226853-1-ebiggers@kernel.org Signed-off-by: Eric Biggers <ebiggers@google.com>
2019-11-20 06:24:47 +08:00
err = register_key_type(&key_type_fscrypt_provisioning);
if (err)
goto err_unregister_fscrypt_user;
fscrypt: allow unprivileged users to add/remove keys for v2 policies Allow the FS_IOC_ADD_ENCRYPTION_KEY and FS_IOC_REMOVE_ENCRYPTION_KEY ioctls to be used by non-root users to add and remove encryption keys from the filesystem-level crypto keyrings, subject to limitations. Motivation: while privileged fscrypt key management is sufficient for some users (e.g. Android and Chromium OS, where a privileged process manages all keys), the old API by design also allows non-root users to set up and use encrypted directories, and we don't want to regress on that. Especially, we don't want to force users to continue using the old API, running into the visibility mismatch between files and keyrings and being unable to "lock" encrypted directories. Intuitively, the ioctls have to be privileged since they manipulate filesystem-level state. However, it's actually safe to make them unprivileged if we very carefully enforce some specific limitations. First, each key must be identified by a cryptographic hash so that a user can't add the wrong key for another user's files. For v2 encryption policies, we use the key_identifier for this. v1 policies don't have this, so managing keys for them remains privileged. Second, each key a user adds is charged to their quota for the keyrings service. Thus, a user can't exhaust memory by adding a huge number of keys. By default each non-root user is allowed up to 200 keys; this can be changed using the existing sysctl 'kernel.keys.maxkeys'. Third, if multiple users add the same key, we keep track of those users of the key (of which there remains a single copy), and won't really remove the key, i.e. "lock" the encrypted files, until all those users have removed it. This prevents denial of service attacks that would be possible under simpler schemes, such allowing the first user who added a key to remove it -- since that could be a malicious user who has compromised the key. Of course, encryption keys should be kept secret, but the idea is that using encryption should never be *less* secure than not using encryption, even if your key was compromised. We tolerate that a user will be unable to really remove a key, i.e. unable to "lock" their encrypted files, if another user has added the same key. But in a sense, this is actually a good thing because it will avoid providing a false notion of security where a key appears to have been removed when actually it's still in memory, available to any attacker who compromises the operating system kernel. Reviewed-by: Theodore Ts'o <tytso@mit.edu> Signed-off-by: Eric Biggers <ebiggers@google.com>
2019-08-05 10:35:47 +08:00
return 0;
fscrypt: support passing a keyring key to FS_IOC_ADD_ENCRYPTION_KEY Extend the FS_IOC_ADD_ENCRYPTION_KEY ioctl to allow the raw key to be specified by a Linux keyring key, rather than specified directly. This is useful because fscrypt keys belong to a particular filesystem instance, so they are destroyed when that filesystem is unmounted. Usually this is desired. But in some cases, userspace may need to unmount and re-mount the filesystem while keeping the keys, e.g. during a system update. This requires keeping the keys somewhere else too. The keys could be kept in memory in a userspace daemon. But depending on the security architecture and assumptions, it can be preferable to keep them only in kernel memory, where they are unreadable by userspace. We also can't solve this by going back to the original fscrypt API (where for each file, the master key was looked up in the process's keyring hierarchy) because that caused lots of problems of its own. Therefore, add the ability for FS_IOC_ADD_ENCRYPTION_KEY to accept a Linux keyring key. This solves the problem by allowing userspace to (if needed) save the keys securely in a Linux keyring for re-provisioning, while still using the new fscrypt key management ioctls. This is analogous to how dm-crypt accepts a Linux keyring key, but the key is then stored internally in the dm-crypt data structures rather than being looked up again each time the dm-crypt device is accessed. Use a custom key type "fscrypt-provisioning" rather than one of the existing key types such as "logon". This is strongly desired because it enforces that these keys are only usable for a particular purpose: for fscrypt as input to a particular KDF. Otherwise, the keys could also be passed to any kernel API that accepts a "logon" key with any service prefix, e.g. dm-crypt, UBIFS, or (recently proposed) AF_ALG. This would risk leaking information about the raw key despite it ostensibly being unreadable. Of course, this mistake has already been made for multiple kernel APIs; but since this is a new API, let's do it right. This patch has been tested using an xfstest which I wrote to test it. Link: https://lore.kernel.org/r/20191119222447.226853-1-ebiggers@kernel.org Signed-off-by: Eric Biggers <ebiggers@google.com>
2019-11-20 06:24:47 +08:00
err_unregister_fscrypt_user:
unregister_key_type(&key_type_fscrypt_user);
fscrypt: allow unprivileged users to add/remove keys for v2 policies Allow the FS_IOC_ADD_ENCRYPTION_KEY and FS_IOC_REMOVE_ENCRYPTION_KEY ioctls to be used by non-root users to add and remove encryption keys from the filesystem-level crypto keyrings, subject to limitations. Motivation: while privileged fscrypt key management is sufficient for some users (e.g. Android and Chromium OS, where a privileged process manages all keys), the old API by design also allows non-root users to set up and use encrypted directories, and we don't want to regress on that. Especially, we don't want to force users to continue using the old API, running into the visibility mismatch between files and keyrings and being unable to "lock" encrypted directories. Intuitively, the ioctls have to be privileged since they manipulate filesystem-level state. However, it's actually safe to make them unprivileged if we very carefully enforce some specific limitations. First, each key must be identified by a cryptographic hash so that a user can't add the wrong key for another user's files. For v2 encryption policies, we use the key_identifier for this. v1 policies don't have this, so managing keys for them remains privileged. Second, each key a user adds is charged to their quota for the keyrings service. Thus, a user can't exhaust memory by adding a huge number of keys. By default each non-root user is allowed up to 200 keys; this can be changed using the existing sysctl 'kernel.keys.maxkeys'. Third, if multiple users add the same key, we keep track of those users of the key (of which there remains a single copy), and won't really remove the key, i.e. "lock" the encrypted files, until all those users have removed it. This prevents denial of service attacks that would be possible under simpler schemes, such allowing the first user who added a key to remove it -- since that could be a malicious user who has compromised the key. Of course, encryption keys should be kept secret, but the idea is that using encryption should never be *less* secure than not using encryption, even if your key was compromised. We tolerate that a user will be unable to really remove a key, i.e. unable to "lock" their encrypted files, if another user has added the same key. But in a sense, this is actually a good thing because it will avoid providing a false notion of security where a key appears to have been removed when actually it's still in memory, available to any attacker who compromises the operating system kernel. Reviewed-by: Theodore Ts'o <tytso@mit.edu> Signed-off-by: Eric Biggers <ebiggers@google.com>
2019-08-05 10:35:47 +08:00
return err;
fscrypt: add FS_IOC_ADD_ENCRYPTION_KEY ioctl Add a new fscrypt ioctl, FS_IOC_ADD_ENCRYPTION_KEY. This ioctl adds an encryption key to the filesystem's fscrypt keyring ->s_master_keys, making any files encrypted with that key appear "unlocked". Why we need this ~~~~~~~~~~~~~~~~ The main problem is that the "locked/unlocked" (ciphertext/plaintext) status of encrypted files is global, but the fscrypt keys are not. fscrypt only looks for keys in the keyring(s) the process accessing the filesystem is subscribed to: the thread keyring, process keyring, and session keyring, where the session keyring may contain the user keyring. Therefore, userspace has to put fscrypt keys in the keyrings for individual users or sessions. But this means that when a process with a different keyring tries to access encrypted files, whether they appear "unlocked" or not is nondeterministic. This is because it depends on whether the files are currently present in the inode cache. Fixing this by consistently providing each process its own view of the filesystem depending on whether it has the key or not isn't feasible due to how the VFS caches work. Furthermore, while sometimes users expect this behavior, it is misguided for two reasons. First, it would be an OS-level access control mechanism largely redundant with existing access control mechanisms such as UNIX file permissions, ACLs, LSMs, etc. Encryption is actually for protecting the data at rest. Second, almost all users of fscrypt actually do need the keys to be global. The largest users of fscrypt, Android and Chromium OS, achieve this by having PID 1 create a "session keyring" that is inherited by every process. This works, but it isn't scalable because it prevents session keyrings from being used for any other purpose. On general-purpose Linux distros, the 'fscrypt' userspace tool [1] can't similarly abuse the session keyring, so to make 'sudo' work on all systems it has to link all the user keyrings into root's user keyring [2]. This is ugly and raises security concerns. Moreover it can't make the keys available to system services, such as sshd trying to access the user's '~/.ssh' directory (see [3], [4]) or NetworkManager trying to read certificates from the user's home directory (see [5]); or to Docker containers (see [6], [7]). By having an API to add a key to the *filesystem* we'll be able to fix the above bugs, remove userspace workarounds, and clearly express the intended semantics: the locked/unlocked status of an encrypted directory is global, and encryption is orthogonal to OS-level access control. Why not use the add_key() syscall ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ We use an ioctl for this API rather than the existing add_key() system call because the ioctl gives us the flexibility needed to implement fscrypt-specific semantics that will be introduced in later patches: - Supporting key removal with the semantics such that the secret is removed immediately and any unused inodes using the key are evicted; also, the eviction of any in-use inodes can be retried. - Calculating a key-dependent cryptographic identifier and returning it to userspace. - Allowing keys to be added and removed by non-root users, but only keys for v2 encryption policies; and to prevent denial-of-service attacks, users can only remove keys they themselves have added, and a key is only really removed after all users who added it have removed it. Trying to shoehorn these semantics into the keyrings syscalls would be very difficult, whereas the ioctls make things much easier. However, to reuse code the implementation still uses the keyrings service internally. Thus we get lockless RCU-mode key lookups without having to re-implement it, and the keys automatically show up in /proc/keys for debugging purposes. References: [1] https://github.com/google/fscrypt [2] https://goo.gl/55cCrI#heading=h.vf09isp98isb [3] https://github.com/google/fscrypt/issues/111#issuecomment-444347939 [4] https://github.com/google/fscrypt/issues/116 [5] https://bugs.launchpad.net/ubuntu/+source/fscrypt/+bug/1770715 [6] https://github.com/google/fscrypt/issues/128 [7] https://askubuntu.com/questions/1130306/cannot-run-docker-on-an-encrypted-filesystem Reviewed-by: Theodore Ts'o <tytso@mit.edu> Signed-off-by: Eric Biggers <ebiggers@google.com>
2019-08-05 10:35:46 +08:00
}