cache objects. This is good, but still leaves a lot of kernel memory
available to be copied to/from userspace in the face of bugs. To further
restrict what memory is available for copying, this creates a way to
whitelist specific areas of a given slab cache object for copying to/from
userspace, allowing much finer granularity of access control. Slab caches
that are never exposed to userspace can declare no whitelist for their
objects, thereby keeping them unavailable to userspace via dynamic copy
operations. (Note, an implicit form of whitelisting is the use of constant
sizes in usercopy operations and get_user()/put_user(); these bypass all
hardened usercopy checks since these sizes cannot change at runtime.)
This new check is WARN-by-default, so any mistakes can be found over the
next several releases without breaking anyone's system.
The series has roughly the following sections:
- remove %p and improve reporting with offset
- prepare infrastructure and whitelist kmalloc
- update VFS subsystem with whitelists
- update SCSI subsystem with whitelists
- update network subsystem with whitelists
- update process memory with whitelists
- update per-architecture thread_struct with whitelists
- update KVM with whitelists and fix ioctl bug
- mark all other allocations as not whitelisted
- update lkdtm for more sensible test overage
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1
Comment: Kees Cook <kees@outflux.net>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=tzmJ
-----END PGP SIGNATURE-----
Merge tag 'usercopy-v4.16-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/kees/linux
Pull hardened usercopy whitelisting from Kees Cook:
"Currently, hardened usercopy performs dynamic bounds checking on slab
cache objects. This is good, but still leaves a lot of kernel memory
available to be copied to/from userspace in the face of bugs.
To further restrict what memory is available for copying, this creates
a way to whitelist specific areas of a given slab cache object for
copying to/from userspace, allowing much finer granularity of access
control.
Slab caches that are never exposed to userspace can declare no
whitelist for their objects, thereby keeping them unavailable to
userspace via dynamic copy operations. (Note, an implicit form of
whitelisting is the use of constant sizes in usercopy operations and
get_user()/put_user(); these bypass all hardened usercopy checks since
these sizes cannot change at runtime.)
This new check is WARN-by-default, so any mistakes can be found over
the next several releases without breaking anyone's system.
The series has roughly the following sections:
- remove %p and improve reporting with offset
- prepare infrastructure and whitelist kmalloc
- update VFS subsystem with whitelists
- update SCSI subsystem with whitelists
- update network subsystem with whitelists
- update process memory with whitelists
- update per-architecture thread_struct with whitelists
- update KVM with whitelists and fix ioctl bug
- mark all other allocations as not whitelisted
- update lkdtm for more sensible test overage"
* tag 'usercopy-v4.16-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/kees/linux: (38 commits)
lkdtm: Update usercopy tests for whitelisting
usercopy: Restrict non-usercopy caches to size 0
kvm: x86: fix KVM_XEN_HVM_CONFIG ioctl
kvm: whitelist struct kvm_vcpu_arch
arm: Implement thread_struct whitelist for hardened usercopy
arm64: Implement thread_struct whitelist for hardened usercopy
x86: Implement thread_struct whitelist for hardened usercopy
fork: Provide usercopy whitelisting for task_struct
fork: Define usercopy region in thread_stack slab caches
fork: Define usercopy region in mm_struct slab caches
net: Restrict unwhitelisted proto caches to size 0
sctp: Copy struct sctp_sock.autoclose to userspace using put_user()
sctp: Define usercopy region in SCTP proto slab cache
caif: Define usercopy region in caif proto slab cache
ip: Define usercopy region in IP proto slab cache
net: Define usercopy region in struct proto slab cache
scsi: Define usercopy region in scsi_sense_cache slab cache
cifs: Define usercopy region in cifs_request slab cache
vxfs: Define usercopy region in vxfs_inode slab cache
ufs: Define usercopy region in ufs_inode_cache slab cache
...
Function handle_IRQ_event was retired in v2.6.39 and replaced with
handle_irq_event but nobody changed it in lkdtm so INT_HW_IRQ_EN has
been broken for a while.
Fixes: 33b054b867 ("genirq: Remove handle_IRQ_event")
Signed-off-by: Travis Brown <travisb@arista.com>
Signed-off-by: Ivan Delalande <colona@arista.com>
Acked-by: Kees Cook <keescook@chromium.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
This updates the USERCOPY_HEAP_FLAG_* tests to USERCOPY_HEAP_WHITELIST_*,
since the final form of usercopy whitelisting ended up using an offset/size
window instead of the earlier proposed allocation flags.
Signed-off-by: Kees Cook <keescook@chromium.org>
Since nothing in the crashtypes table ever changes, mark it const. Adds
some missing "static" markings as well.
Signed-off-by: Kees Cook <keescook@chromium.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
The jprobe subsystem is being removed, so convert to using kprobe instead.
Cc: Masami Hiramatsu <mhiramat@kernel.org>
Signed-off-by: Kees Cook <keescook@chromium.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
There wasn't an LKDTM test to distinguish between -fstack-protector and
-fstack-protector-strong in use. This adds CORRUPT_STACK_STRONG to see
the difference. Also adjusts the stack-clobber value to 0xff so execution
won't potentially jump into userspace when the stack protector is missing.
Signed-off-by: Kees Cook <keescook@chromium.org>
Two new tests STACK_GUARD_PAGE_LEADING and STACK_GUARD_PAGE_TRAILING
attempt to read the byte before and after, respectively, of the current
stack frame, which should fault.
Signed-off-by: Kees Cook <keescook@chromium.org>
While not a crash test, this does provide two tight atomic_t and
refcount_t loops for performance comparisons:
cd /sys/kernel/debug/provoke-crash
perf stat -B -- cat <(echo ATOMIC_TIMING) > DIRECT
perf stat -B -- cat <(echo REFCOUNT_TIMING) > DIRECT
Looking a CPU cycles is the best way to example the fast-path (rather
than instruction counts, since conditional jumps will be executed but
will be negligible due to branch-prediction).
Signed-off-by: Kees Cook <keescook@chromium.org>
The existing REFCOUNT_* LKDTM tests were designed only for testing a narrow
portion of CONFIG_REFCOUNT_FULL. This moves the tests to their own file and
expands their testing to poke each boundary condition.
Since the protections (CONFIG_REFCOUNT_FULL and x86-fast) use different
saturation values and reach-zero behavior, those have to be build-time
set so the tests can actually validate things are happening at the
right places.
Notably, the x86-fast protection will fail REFCOUNT_INC_ZERO and
REFCOUNT_ADD_ZERO since those conditions are not checked (only overflow
is critical to protecting refcount_t). CONFIG_REFCOUNT_FULL will warn for
each REFCOUNT_*_NEGATIVE test since it provides zero-pinning behaviors
(which allows it to pass REFCOUNT_INC_ZERO and REFCOUNT_ADD_ZERO).
Signed-off-by: Kees Cook <keescook@chromium.org>
This adds CORRUPT_USER_DS to check that the get_fs() test on syscall
return (via __VERIFY_PRE_USERMODE_STATE) still sees USER_DS. Since
trying to deal with values other than USER_DS and KERNEL_DS across all
architectures in a safe way is not sensible, this sets KERNEL_DS, but
since that could be extremely dangerous if the protection is not present,
it also raises SIGKILL for current, so that no matter what, the process
will die. A successful test will be visible with a BUG(), like all the
other LKDTM tests.
Signed-off-by: Kees Cook <keescook@chromium.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Here is the big char/misc driver patchset for 4.11-rc1.
Lots of different driver subsystems updated here. Rework for the hyperv
subsystem to handle new platforms better, mei and w1 and extcon driver
updates, as well as a number of other "minor" driver updates. Full
details are in the shortlog below.
All of these have been in linux-next for a while with no reported
issues.
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
-----BEGIN PGP SIGNATURE-----
iG0EABECAC0WIQT0tgzFv3jCIUoxPcsxR9QN2y37KQUCWK2iRQ8cZ3JlZ0Brcm9h
aC5jb20ACgkQMUfUDdst+ynhFACguVE+/ixj5u5bT5DXQaZNai/6zIAAmgMWwd/t
YTD2cwsJsGbTT1fY3SUe
=CiSI
-----END PGP SIGNATURE-----
Merge tag 'char-misc-4.11-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/char-misc
Pull char/misc driver updates from Greg KH:
"Here is the big char/misc driver patchset for 4.11-rc1.
Lots of different driver subsystems updated here: rework for the
hyperv subsystem to handle new platforms better, mei and w1 and extcon
driver updates, as well as a number of other "minor" driver updates.
All of these have been in linux-next for a while with no reported
issues"
* tag 'char-misc-4.11-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/char-misc: (169 commits)
goldfish: Sanitize the broken interrupt handler
x86/platform/goldfish: Prevent unconditional loading
vmbus: replace modulus operation with subtraction
vmbus: constify parameters where possible
vmbus: expose hv_begin/end_read
vmbus: remove conditional locking of vmbus_write
vmbus: add direct isr callback mode
vmbus: change to per channel tasklet
vmbus: put related per-cpu variable together
vmbus: callback is in softirq not workqueue
binder: Add support for file-descriptor arrays
binder: Add support for scatter-gather
binder: Add extra size to allocator
binder: Refactor binder_transact()
binder: Support multiple /dev instances
binder: Deal with contexts in debugfs
binder: Support multiple context managers
binder: Split flat_binder_object
auxdisplay: ht16k33: remove private workqueue
auxdisplay: ht16k33: rework input device initialization
...
No jprobe is registered when the module is loaded without specifying a
crashpoint that uses a jprobe. At the moment, we unconditionally try to
unregister the jprobe on module unload which results in an Oops. Add a
check to fix this.
Signed-off-by: Juerg Haefliger <juerg.haefliger@hpe.com>
Acked-by: Kees Cook <keescook@chromium.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
When building under CONFIG_DEBUG_LIST, list addition and removal will be
sanity-checked. This validates that the check is working as expected by
setting up classic corruption attacks against list manipulations, available
with the new lkdtm tests CORRUPT_LIST_ADD and CORRUPT_LIST_DEL.
Signed-off-by: Kees Cook <keescook@chromium.org>
Acked-by: Steven Rostedt <rostedt@goodmis.org>
Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Acked-by: Rik van Riel <riel@redhat.com>
When building under W=1, the lack of lkdtm.h in lkdtm_usercopy.c and
lkdtm_rodata.c was discovered. This fixes the issue and consolidates
the common header and the pr_fmt macro for simplicity and regularity
across each test source file.
Signed-off-by: Kees Cook <keescook@chromium.org>
A conversion of the lkdtm core module added an "#ifdef CONFIG_KPROBES" check,
but a number of functions then become unused:
drivers/misc/lkdtm_core.c:340:16: error: 'lkdtm_debugfs_entry' defined but not used [-Werror=unused-function]
drivers/misc/lkdtm_core.c:122:12: error: 'jp_generic_ide_ioctl' defined but not used [-Werror=unused-function]
drivers/misc/lkdtm_core.c:114:12: error: 'jp_scsi_dispatch_cmd' defined but not used [-Werror=unused-function]
drivers/misc/lkdtm_core.c:106:12: error: 'jp_hrtimer_start' defined but not used [-Werror=unused-function]
drivers/misc/lkdtm_core.c:97:22: error: 'jp_shrink_inactive_list' defined but not used [-Werror=unused-function]
drivers/misc/lkdtm_core.c:89:13: error: 'jp_ll_rw_block' defined but not used [-Werror=unused-function]
drivers/misc/lkdtm_core.c:83:13: error: 'jp_tasklet_action' defined but not used [-Werror=unused-function]
drivers/misc/lkdtm_core.c:75:20: error: 'jp_handle_irq_event' defined but not used [-Werror=unused-function]
drivers/misc/lkdtm_core.c:68:21: error: 'jp_do_irq' defined but not used [-Werror=unused-function]
This adds the same #ifdef everywhere. There is probably a better way to do the
same thing, but for now this avoids the new warnings.
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Fixes: c479e3fd88 ("lkdtm: use struct arrays instead of enums")
[kees: moved some code around to better consolidate the #ifdefs]
Signed-off-by: Kees Cook <keescook@chromium.org>
This removes the use of enums in favor of much more readable and compact
structure arrays. This requires changing all the enum passing to pointers
instead, but the results are much cleaner.
Signed-off-by: Kees Cook <keescook@chromium.org>
In preparation of referencing the jprobe entry points in a structure,
this moves them to the start of the source since they operate mostly
separately from everything else.
Signed-off-by: Kees Cook <keescook@chromium.org>
This reorganizes module parameters and global variables in the source
so they're grouped together with comments. Also moves early function
declarations to the top of the file.
Signed-off-by: Kees Cook <keescook@chromium.org>
The global variables used to track the active crashpoint and crashtype
are hard to distinguish from local variable names, so add a "lkdtm_"
prefix to them (or in the case of "lkdtm", add a "_jprobe" suffix).
Signed-off-by: Kees Cook <keescook@chromium.org>
The "count" variable name was not easy to understand, since it was regularly
obscured by local variables of the same name, and it's purpose wasn't clear.
This renames it (and its lock) to "crash_count", which is more readable.
Signed-off-by: Kees Cook <keescook@chromium.org>
There wasn't a good reason for keeping the enum and the names out of sync
by 1 position just to avoid "NONE" and "INVALID" from being in the string
lists.
Signed-off-by: Kees Cook <keescook@chromium.org>
This splits all the remaining tests from lkdtm_core.c into the new
lkdtm_bugs.c file to help separate things better for readability.
Signed-off-by: Kees Cook <keescook@chromium.org>
This splits the *_AFTER_FREE and related tests into the new lkdtm_heap.c
file to help separate things better for readability.
Signed-off-by: Kees Cook <keescook@chromium.org>
This splits the EXEC_*, WRITE_* and related tests into the new lkdtm_perms.c
file to help separate things better for readability.
Signed-off-by: Kees Cook <keescook@chromium.org>
This splits the USERCOPY_* tests into the new lkdtm_usercopy.c file to
help separate things better for readability.
Signed-off-by: Kees Cook <keescook@chromium.org>
There is no good reason to have the alloc_size parameter currently. The
compiler-tricking value used to exercise the stack can just use a stack
address instead. Similarly hard-code cache_size.
Signed-off-by: Kees Cook <keescook@chromium.org>
The upcoming HARDENED_USERCOPY checks will also block access to the
kernel text, so provide a test for this as well.
Signed-off-by: Kees Cook <keescook@chromium.org>
This adds test to detect copy_to_user/copy_from_user problems that are
protected by PAX_USERCOPY (and will be protected by HARDENED_USERCOPY).
Explicitly tests both "to" and "from" directions of heap object size
problems, heap object markings and, stack frame misalignment.
Signed-off-by: Kees Cook <keescook@chromium.org>
This adds a function that lives in the .rodata section. The section
flags are corrected using objcopy since there is no way with gcc to
declare section flags in an architecture-agnostic way.
Signed-off-by: Kees Cook <keescook@chromium.org>
This cleans up comments a bit to improve readability, adjusts the
name of the module after the source file renaming, and corrects the
MAINTAINERS for the upcoming lkdtm files.
Signed-off-by: Kees Cook <keescook@chromium.org>
Kbuild lacks a way to do in-place objcopy or other modifications of
built targets, so in order to move functions into non-text sections
without renaming the kernel module, the build of lkdtm must be split
into separate source files. This renames lkdtm.c to lkdtm_core.c in
preparation for adding the source file for the .rodata section.
Signed-off-by: Kees Cook <keescook@chromium.org>
Linus Torvalds