commit e548e9b93d makes the kclient
only re-send cap flush once during MDS failover. If the kclient sends
a cap flush after MDS enters reconnect stage but before MDS recovers.
The kclient will skip re-sending the same cap flush when MDS recovers.
This causes problem for newly created inode. The MDS handles cap
flushes before replaying unsafe requests, so it's possible that MDS
find corresponding inode is missing when handling cap flush. The fix
is reverting to old behaviour: always re-send when MDS recovers
Signed-off-by: Yan, Zheng <zyan@redhat.com>
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
Previously our dcache readdir code relies on that child dentries in
directory dentry's d_subdir list are sorted by dentry's offset in
descending order. When adding dentries to the dcache, if a dentry
already exists, our readdir code moves it to head of directory
dentry's d_subdir list. This design relies on dcache internals.
Al Viro suggests using ncpfs's approach: keeping array of pointers
to dentries in page cache of directory inode. the validity of those
pointers are presented by directory inode's complete and ordered
flags. When a dentry gets pruned, we clear directory inode's complete
flag in the d_prune() callback. Before moving a dentry to other
directory, we clear the ordered flag for both old and new directory.
Signed-off-by: Yan, Zheng <zyan@redhat.com>
if flushing caps were revoked, we should re-send the cap flush in
client reconnect stage. This guarantees that MDS processes the cap
flush message before issuing the flushing caps to other client.
Signed-off-by: Yan, Zheng <zyan@redhat.com>
So we know TID of the oldest pending caps flushing. Later patch will
send this information to MDS, so that MDS can trim its completed caps
flush list.
Tracking pending caps flushing globally also simplifies syncfs code.
Signed-off-by: Yan, Zheng <zyan@redhat.com>
Previously we do not trace accurate TID for flushing caps. when
MDS failovers, we have no choice but to re-send all flushing caps
with a new TID. This can cause problem because MDS can has already
flushed some caps and has issued the same caps to other client.
The re-sent cap flush has a new TID, which makes MDS unable to
detect if it has already processed the cap flush.
This patch adds code to track pending caps flushing accurately.
When re-sending cap flush is needed, we use its original flush
TID.
Signed-off-by: Yan, Zheng <zyan@redhat.com>
Previously we pre-allocate cap release messages for each caps. This
wastes lots of memory when there are large amount of caps. This patch
make the code not pre-allocate the cap release messages. Instead,
we add the corresponding ceph_cap struct to a list when releasing a
cap. Later when flush cap releases is needed, we allocate the cap
release messages dynamically.
Signed-off-by: Yan, Zheng <zyan@redhat.com>
when a snap notification contains no new snapshot, we can avoid
sending FLUSHSNAP message to MDS. But we still need to create
cap_snap in some case because it's required by write path and
page writeback path
Signed-off-by: Yan, Zheng <zyan@redhat.com>
Cached_context in ceph_snap_realm is directly accessed by
uninline_data() and get_pool_perm(). This is racy in theory.
both uninline_data() and get_pool_perm() do not modify existing
object, they only create new object. So we can pass the empty
snap context to them. Unlike cached_context in ceph_snap_realm,
we do not need to protect the empty snap context.
Signed-off-by: Yan, Zheng <zyan@redhat.com>
When snaprealm is created, its initial reference count is zero.
But in some rare cases, the newly created snaprealm is not referenced
by anyone. This causes snaprealm with zero reference count not freed.
The fix is set reference count of newly snaprealm to 1. The reference
is return the function who requests to create the snaprealm. When the
function finishes its job, it releases the reference.
Signed-off-by: Yan, Zheng <zyan@redhat.com>
After converting inline data to normal data, client need to flush
the new i_inline_version (CEPH_INLINE_NONE) to MDS. This commit makes
cap messages (sent to MDS) contain inline_version and inline_data.
Client always converts inline data to normal data before data write,
so the inline data length part is always zero.
Signed-off-by: Yan, Zheng <zyan@redhat.com>
Before any data write, convert inline data to normal data and set
i_inline_version to CEPH_INLINE_NONE. The OSD request that saves
inline data to object contains 3 operations (CMPXATTR, WRITE and
SETXATTR). It compares a xattr named 'inline_version' to prevent
old data overwrites newer data.
Signed-off-by: Yan, Zheng <zyan@redhat.com>
Add a new parameter 'locked_page' to ceph_do_getattr(). If inline data
in getattr reply will be copied to the page.
Signed-off-by: Yan, Zheng <zyan@redhat.com>
Request reply and cap message can contain inline data. add inline data
to the page cache if there is Fc cap.
Signed-off-by: Yan, Zheng <zyan@redhat.com>
Current snaphost code does not properly handle moving inode from one
empty snap realm to another empty snap realm. After changing inode's
snap realm, some dirty pages' snap context can be not equal to inode's
i_head_snap. This can trigger BUG() in ceph_put_wrbuffer_cap_refs()
The fix is introduce a global empty snap context for all empty snap
realm. This avoids triggering the BUG() for filesystem with no snapshot.
Fixes: http://tracker.ceph.com/issues/9928
Signed-off-by: Yan, Zheng <zyan@redhat.com>
Reviewed-by: Ilya Dryomov <idryomov@redhat.com>
After creating/deleting/renaming file, offsets of sibling dentries may
change. So we can not use cached dentries to satisfy readdir. But we can
still use the cached dentries to conclude -ENOENT for lookup.
This patch introduces a new inode flag indicating if child dentries are
ordered. The flag is set at the same time marking a directory complete.
After creating/deleting/renaming file, we clear the flag on directory
inode. This prevents ceph_readdir() from using cached dentries to satisfy
readdir syscall.
Signed-off-by: Yan, Zheng <zyan@redhat.com>
Current code set new file/directory's initial ACL in a non-atomic
manner.
Client first sends request to MDS to create new file/directory, then set
the initial ACL after the new file/directory is successfully created.
The fix is include the initial ACL in create/mkdir/mknod MDS requests.
So MDS can handle creating file/directory and setting the initial ACL in
one request.
Signed-off-by: Yan, Zheng <zyan@redhat.com>
Reviewed-by: Sage Weil <sage@redhat.com>
Following sequence of events can happen.
- Client releases an inode, queues cap release message.
- A 'lookup' reply brings the same inode back, but the reply
doesn't contain xattrs because MDS didn't receive the cap release
message and thought client already has up-to-data xattrs.
The fix is force sending a getattr request to MDS if xattrs_version
is 0. The getattr mask is set to CEPH_STAT_CAP_XATTR, so MDS knows client
does not have xattr.
Signed-off-by: Yan, Zheng <zyan@redhat.com>
So that ceph_add_cap() can be used while i_ceph_lock is locked.
This simplifies the code that handle cap import/export.
Signed-off-by: Yan, Zheng <zheng.z.yan@intel.com>
When creating a file, ceph_set_dentry_offset() puts the new dentry
at the end of directory's d_subdirs, then set the dentry's offset
based on directory's max offset. The offset does not reflect the
real postion of the dentry in directory. Later readdir reply from
MDS may change the dentry's position/offset. This inconsistency
can cause missing/duplicate entries in readdir result if readdir
is partly satisfied by dcache_readdir().
The fix is clear directory's completeness after creating/renaming
file. It prevents later readdir from using dcache_readdir().
Fixes: http://tracker.ceph.com/issues/8025
Signed-off-by: Yan, Zheng <zheng.z.yan@intel.com>
Reviewed-by: Sage Weil <sage@inktank.com>
flock and posix lock should use fl->fl_file instead of process ID
as owner identifier. (posix lock uses fl->fl_owner. fl->fl_owner
is usually equal to fl->fl_file, but it also can be a customized
value). The process ID of who holds the lock is just for F_GETLK
fcntl(2).
The fix is rename the 'pid' fields of struct ceph_mds_request_args
and struct ceph_filelock to 'owner', rename 'pid_namespace' fields
to 'pid'. Assign fl->fl_file to the 'owner' field of lock messages.
We also set the most significant bit of the 'owner' field. MDS can
use that bit to distinguish between old and new clients.
The MDS counterpart of this patch modifies the flock code to not
take the 'pid_namespace' into consideration when checking conflict
locks.
Signed-off-by: Yan, Zheng <zheng.z.yan@intel.com>
Reviewed-by: Sage Weil <sage@inktank.com>
Comparing offset with inode->i_sb->s_maxbytes doesn't make sense for
directory. For a fragmented directory, offset (frag_t, off) can be
larger than inode->i_sb->s_maxbytes.
At the very beginning of ceph_dir_llseek(), local variable old_offset
is initialized to parameter offset. This doesn't make sense neither.
Old_offset should be ceph_make_fpos(fi->frag, fi->next_offset).
Signed-off-by: Yan, Zheng <zheng.z.yan@intel.com>
Reviewed-by: Alex Elder <elder@linaro.org>
The merge of commit 7221fe4c2e ("ceph: add acl for cephfs") raced with
upstream changes in the generic POSIX ACL code (eg commit 2aeccbe957
"fs: add generic xattr_acl handlers" and others).
Some of the fallout was fixed in commit 4db658ea0c ("ceph: Fix up after
semantic merge conflict"), but it was incomplete: the set_acl
inode_operation wasn't getting set, and the prototype needed to be
adjusted a bit (it doesn't take a dentry anymore).
Signed-off-by: Sage Weil <sage@inktank.com>
Signed-off-by: Ilya Dryomov <ilya.dryomov@inktank.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
The previous ceph-client merge resulted in ceph not even building,
because there was a merge conflict that wasn't visible as an actual data
conflict: commit 7221fe4c2e ("ceph: add acl for cephfs") added support
for POSIX ACL's into Ceph, but unluckily we also had the VFS tree change
a lot of the POSIX ACL helper functions to be much more helpful to
filesystems (see for example commits 2aeccbe957 "fs: add generic
xattr_acl handlers", 5bf3258fd2 "fs: make posix_acl_chmod more useful"
and 37bc15392a "fs: make posix_acl_create more useful")
The reason this conflict wasn't obvious was many-fold: because it was a
semantic conflict rather than a data conflict, it wasn't visible in the
git merge as a conflict. And because the VFS tree hadn't been in
linux-next, people hadn't become aware of it that way. And because I
was at jury duty this morning, I was using my laptop and as a result not
doing constant "allmodconfig" builds.
Anyway, this fixes the build and generally removes a fair chunk of the
Ceph POSIX ACL support code, since the improved helpers seem to match
really well for Ceph too. But I don't actually have any way to *test*
the end result, and I was really hoping for some ACK's for this. Oh,
well.
Not compiling certainly doesn't make things easier to test, so I'm
committing this without the acks after having waited for four hours...
Plus it's what I would have done for the merge had I noticed the
semantic conflict..
Reported-by: Dave Jones <davej@redhat.com>
Cc: Sage Weil <sage@inktank.com>
Cc: Guangliang Zhao <lucienchao@gmail.com>
Cc: Li Wang <li.wang@ubuntykylin.com>
Cc: Christoph Hellwig <hch@infradead.org>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Version 3 cap export message includes information about the imported
caps. It allows us to add the imported caps if the corresponding cap
import message still hasn't been received.
This allow us to handle situation that the importer MDS crashes and
the cap import message is missing.
Signed-off-by: Yan, Zheng <zheng.z.yan@intel.com>
Some inodes in readdir reply may have no caps. Getattr mds request
for these inodes can return -ESTALE. The fix is consider dentry that
links to inode with no caps as invalid. Invalid dentry causes a
lookup request to send to the mds, the MDS will send caps back.
Signed-off-by: Yan, Zheng <zheng.z.yan@intel.com>
handle following sequence of events:
- non-auth MDS revokes Fc cap. queue invalidate work
- auth MDS issues Fc cap through request reply. i_rdcache_gen gets
increased.
- invalidate work runs. it finds i_rdcache_revoking != i_rdcache_gen,
so it does nothing.
Signed-off-by: Yan, Zheng <zheng.z.yan@intel.com>
Positve dentry and corresponding inode are always accompanied in MDS reply.
So no need to keep inode in the cache after dropping all its aliases.
Signed-off-by: Yan, Zheng <zheng.z.yan@intel.com>
Reviewed-by: Sage Weil <sage@inktank.com>
commit 6f60f889 (ceph: fix freeing inode vs removing session caps race)
introduced ceph_lookup_inode(). But there is already a ceph_find_inode()
which provides similar function. So remove ceph_lookup_inode(), use
ceph_find_inode() instead.
Signed-off-by: Yan, Zheng <zheng.z.yan@intel.com>
Reviewed-by: Alex Elder <alex.elder@linary.org>
Reviewed-by: Sage Weil <sage@inktank.com>
Adding support for fscache to the Ceph filesystem. This would bring it to on
par with some of the other network filesystems in Linux (like NFS, AFS, etc...)
In order to mount the filesystem with fscache the 'fsc' mount option must be
passed.
Signed-off-by: Milosz Tanski <milosz@adfin.com>
Signed-off-by: Sage Weil <sage@inktank.com>
I encountered below deadlock when running fsstress
wmtruncate work truncate MDS
--------------- ------------------ --------------------------
lock i_mutex
<- truncate file
lock i_mutex (blocked)
<- revoking Fcb (filelock to MIX)
send request ->
handle request (xlock filelock)
At the initial time, there are some dirty pages in the page cache.
When the kclient receives the truncate message, it reduces inode size
and creates some 'out of i_size' dirty pages. wmtruncate work can't
truncate these dirty pages because it's blocked by the i_mutex. Later
when the kclient receives the cap message that revokes Fcb caps, It
can't flush all dirty pages because writepages() only flushes dirty
pages within the inode size.
When the MDS handles the 'truncate' request from kclient, it waits
for the filelock to become stable. But the filelock is stuck in
unstable state because it can't finish revoking kclient's Fcb caps.
The truncate pagecache locking has already caused lots of trouble
for use. I think it's time simplify it by introducing a new mutex.
We use the new mutex to prevent concurrent truncate_inode_pages().
There is no need to worry about race between buffered write and
truncate_inode_pages(), because our "get caps" mechanism prevents
them from concurrent execution.
Reviewed-by: Sage Weil <sage@inktank.com>
Signed-off-by: Yan, Zheng <zheng.z.yan@intel.com>
remove_session_caps() uses iterate_session_caps() to remove caps,
but iterate_session_caps() skips inodes that are being deleted.
So session->s_nr_caps can be non-zero after iterate_session_caps()
return.
We can fix the issue by waiting until deletions are complete.
__wait_on_freeing_inode() is designed for the job, but it is not
exported, so we use lookup inode function to access it.
Signed-off-by: Yan, Zheng <zheng.z.yan@intel.com>
The locking order for pending vmtruncate is wrong, it can lead to
following race:
write wmtruncate work
------------------------ ----------------------
lock i_mutex
check i_truncate_pending check i_truncate_pending
truncate_inode_pages() lock i_mutex (blocked)
copy data to page cache
unlock i_mutex
truncate_inode_pages()
The fix is take i_mutex before calling __ceph_do_pending_vmtruncate()
Fixes: http://tracker.ceph.com/issues/5453
Signed-off-by: Yan, Zheng <zheng.z.yan@intel.com>
Reviewed-by: Sage Weil <sage@inktank.com>
Drop ignored return value. Fix allocation failure case to not leak.
Signed-off-by: Jianpeng Ma <majianpeng@gmail.com>
Reviewed-by: Sage Weil <sage@inktank.com>
Current ceph code tracks directory's completeness in two places.
ceph_readdir() checks i_release_count to decide if it can set the
I_COMPLETE flag in i_ceph_flags. All other places check the I_COMPLETE
flag. This indirection introduces locking complexity.
This patch adds a new variable i_complete_count to ceph_inode_info.
Set i_release_count's value to it when marking a directory complete.
By comparing the two variables, we know if a directory is complete
Signed-off-by: Yan, Zheng <zheng.z.yan@intel.com>
make __ceph_do_pending_vmtruncate() acquire the i_mutex if the caller
does not hold the i_mutex, so ceph_aio_read() can call safely.
Signed-off-by: Yan, Zheng <zheng.z.yan@intel.com>
Reviewed-by: Greg Farnum <greg@inktank.com>
commit c6ffe10015 moved the flag that tracks if the dcache contents
for a directory are complete to dentry. The problem is there are
lots of places that use ceph_dir_{set,clear,test}_complete() while
holding i_ceph_lock. but ceph_dir_{set,clear,test}_complete() may
sleep because they call dput().
This patch basically reverts that commit. For ceph_d_prune(), it's
called with both the dentry to prune and the parent dentry are
locked. So it's safe to access the parent dentry's d_inode and
clear I_COMPLETE flag.
Signed-off-by: Yan, Zheng <zheng.z.yan@intel.com>
Reviewed-by: Greg Farnum <greg@inktank.com>
Reviewed-by: Sage Weil <sage@inktank.com>
Pull Ceph updates from Sage Weil:
"A few groups of patches here. Alex has been hard at work improving
the RBD code, layout groundwork for understanding the new formats and
doing layering. Most of the infrastructure is now in place for the
final bits that will come with the next window.
There are a few changes to the data layout. Jim Schutt's patch fixes
some non-ideal CRUSH behavior, and a set of patches from me updates
the client to speak a newer version of the protocol and implement an
improved hashing strategy across storage nodes (when the server side
supports it too).
A pair of patches from Sam Lang fix the atomicity of open+create
operations. Several patches from Yan, Zheng fix various mds/client
issues that turned up during multi-mds torture tests.
A final set of patches expose file layouts via virtual xattrs, and
allow the policies to be set on directories via xattrs as well
(avoiding the awkward ioctl interface and providing a consistent
interface for both kernel mount and ceph-fuse users)."
* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/sage/ceph-client: (143 commits)
libceph: add support for HASHPSPOOL pool flag
libceph: update osd request/reply encoding
libceph: calculate placement based on the internal data types
ceph: update support for PGID64, PGPOOL3, OSDENC protocol features
ceph: update "ceph_features.h"
libceph: decode into cpu-native ceph_pg type
libceph: rename ceph_pg -> ceph_pg_v1
rbd: pass length, not op for osd completions
rbd: move rbd_osd_trivial_callback()
libceph: use a do..while loop in con_work()
libceph: use a flag to indicate a fault has occurred
libceph: separate non-locked fault handling
libceph: encapsulate connection backoff
libceph: eliminate sparse warnings
ceph: eliminate sparse warnings in fs code
rbd: eliminate sparse warnings
libceph: define connection flag helpers
rbd: normalize dout() calls
rbd: barriers are hard
rbd: ignore zero-length requests
...
Different versions of glibc are broken in different ways, but the short of
it is that for the time being, frsize should == bsize, and be used as the
multiple for the blocks, free, and available fields. This mirrors what is
done for NFS. The previous reporting of the page size for frsize meant
that newer glibc and df would report a very small value for the fs size.
Fixes http://tracker.ceph.com/issues/3793.
Signed-off-by: Sage Weil <sage@inktank.com>
Reviewed-by: Greg Farnum <greg@inktank.com>
Yan, Zheng