When we log changes to the superblock, we first have to write them
to the on-disk buffer, and then log that. Right now we have a
complex bitfield based arrangement to only write the modified field
to the buffer before we log it.
This used to be necessary as a performance optimisation because we
logged the superblock buffer in every extent or inode allocation or
freeing, and so performance was extremely important. We haven't done
this for years, however, ever since the lazy superblock counters
pulled the superblock logging out of the transaction commit
fast path.
Hence we have a bunch of complexity that is not necessary that makes
writing the in-core superblock to disk much more complex than it
needs to be. We only need to log the superblock now during
management operations (e.g. during mount, unmount or quota control
operations) so it is not a performance critical path anymore.
As such, remove the complex field based logging mechanism and
replace it with a simple conversion function similar to what we use
for all other on-disk structures.
This means we always log the entirity of the superblock, but again
because we rarely modify the superblock this is not an issue for log
bandwidth or CPU time. Indeed, if we do log the superblock
frequently, delayed logging will minimise the impact of this
overhead.
[Fixed gquota/pquota inode sharing regression noticed by bfoster.]
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Brian Foster <bfoster@redhat.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>
This update contains:
o more on-disk format header consolidation
o move some structures shared with userspace to libxfs
o new per-mount workqueue to fix for deadlocks between nested loop
mounted filesystems
o various bug fixes for ENOSPC, stats, quota off and preallocation
o a bunch of compiler warning fixes for set-but-unused variables
o various code cleanups
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.12 (GNU/Linux)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=2SJP
-----END PGP SIGNATURE-----
Merge tag 'xfs-for-linus-3.19-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/dgc/linux-xfs
Pull xfs update from Dave Chinner:
"There's relatively little change in this update; it is mainly bug
fixes, cleanups and more of the on-going libxfs restructuring and
on-disk format header consolidation work.
Details:
- more on-disk format header consolidation
- move some structures shared with userspace to libxfs
- new per-mount workqueue to fix for deadlocks between nested loop
mounted filesystems
- various bug fixes for ENOSPC, stats, quota off and preallocation
- a bunch of compiler warning fixes for set-but-unused variables
- various code cleanups"
* tag 'xfs-for-linus-3.19-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/dgc/linux-xfs: (24 commits)
xfs: split metadata and log buffer completion to separate workqueues
xfs: fix set-but-unused warnings
xfs: move type conversion functions to xfs_dir.h
xfs: move ftype conversion functions to libxfs
xfs: lobotomise xfs_trans_read_buf_map()
xfs: active inodes stat is broken
xfs: cleanup xfs_bmse_merge returns
xfs: cleanup xfs_bmse_shift_one goto mess
xfs: fix premature enospc on inode allocation
xfs: overflow in xfs_iomap_eof_align_last_fsb
xfs: fix simple_return.cocci warning in xfs_bmse_shift_one
xfs: fix simple_return.cocci warning in xfs_file_readdir
libxfs: fix simple_return.cocci warnings
xfs: remove unnecessary null checks
xfs: merge xfs_inum.h into xfs_format.h
xfs: move most of xfs_sb.h to xfs_format.h
xfs: merge xfs_ag.h into xfs_format.h
xfs: move acl structures to xfs_format.h
xfs: merge xfs_dinode.h into xfs_format.h
xfs: catch invalid negative blknos in _xfs_buf_find()
...
Pull quota updates from Jan Kara:
"Quota improvements and some minor cleanups.
The main portion in the pull request are changes which move i_dquot
array from struct inode into fs-private part of an inode which saves
memory for filesystems which don't use VFS quotas"
* 'for_linus' of git://git.kernel.org/pub/scm/linux/kernel/git/jack/linux-fs:
udf: One function call less in udf_fill_super() after error detection
udf: Deletion of unnecessary checks before the function call "iput"
jbd: Deletion of an unnecessary check before the function call "iput"
vfs: Remove i_dquot field from inode
jfs: Convert to private i_dquot field
reiserfs: Convert to private i_dquot field
ocfs2: Convert to private i_dquot field
ext4: Convert to private i_dquot field
ext3: Convert to private i_dquot field
ext2: Convert to private i_dquot field
quota: Use function to provide i_dquot pointers
xfs: Set allowed quota types
gfs2: Set allowed quota types
quota: Allow each filesystem to specify which quota types it supports
quota: Remove const from function declarations
quota: Add log level to printk
XFS traditionally sends all buffer I/O completion work to a single
workqueue. This includes metadata buffer completion and log buffer
completion. The log buffer completion requires a high priority queue to
prevent stalls due to log forces getting stuck behind other queued work.
Rather than continue to prioritize all buffer I/O completion due to the
needs of log completion, split log buffer completion off to
m_log_workqueue and move the high priority flag from m_buf_workqueue to
m_log_workqueue.
Add a b_ioend_wq wq pointer to xfs_buf to allow completion workqueue
customization on a per-buffer basis. Initialize b_ioend_wq to
m_buf_workqueue by default in the generic buffer I/O submission path.
Finally, override the default wq with the high priority m_log_workqueue
in the log buffer I/O submission path.
Signed-off-by: Brian Foster <bfoster@redhat.com>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>
The kernel compile doesn't turn on these checks by default, so it's
only when I do a kernel-user sync that I find that there are lots of
compiler warnings waiting to be fixed. Fix up these set-but-unused
warnings.
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Eric Sandeen <sandeen@redhat.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Dave Chinner <david@fromorbit.com>
These are currently considered private to libxfs, but they are
widely used by the userspace code to decode, walk and check
directory structures. Hence they really form part of the external
API and as such need to bemoved to xfs_dir2.h.
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Dave Chinner <david@fromorbit.com>
These functions are needed in userspace for repair and mkfs to
do the right thing. Move them to libxfs so they can be easily
shared.
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Dave Chinner <david@fromorbit.com>
There's a case in that code where it checks for a buffer match in a
transaction where the buffer is not marked done. i.e. trying to
catch a buffer we have locked in the transaction but have not
completed IO on.
The only way we can find a buffer that has not had IO completed on
it is if it had readahead issued on it, but we never do readahead on
buffers that we have already joined into a transaction. Hence this
condition cannot occur, and buffers locked and joined into a
transaction should always be marked done and not under IO.
Remove this code and re-order xfs_trans_read_buf_map() to remove
duplicated IO dispatch and error handling code.
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Dave Chinner <david@fromorbit.com>
vn_active only ever gets decremented, so it has a very large
negative number. Make it track the inode count we currently have
allocated properly so we can easily track the size of the inode
cache via tools like PCP.
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Mark Tinguely <tinguely@sgi.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>
Signed-off-by: Dave Chinner <dchinner@redhat.com>
xfs_bmse_merge() has a jump label for return that just returns the
error value. Convert all the code to just return the error directly
and use XFS_WANT_CORRUPTED_RETURN. This also allows the final call
to xfs_bmbt_update() to return directly.
Noticed while reviewing coccinelle return cleanup patches and
wondering why the same return pattern as in xfs_bmse_shift_one()
wasn't picked up by the checker pattern...
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Brian Foster <bfoster@redhat.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>
xfs_bmse_shift_one() jumps around determining whether to shift or
merge, making the code flow difficult to follow. Clean it up and
use direct error returns (including XFS_WANT_CORRUPTED_RETURN) to
make the code flow better and be easier to read.
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Brian Foster <bfoster@redhat.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>
After growing a filesystem, XFS can fail to allocate inodes even
though there is a large amount of space available in the filesystem
for inodes. The issue is caused by a nearly full allocation group
having enough free space in it to be considered for inode
allocation, but not enough contiguous free space to actually
allocation inodes. This situation results in successful selection
of the AG for allocation, then failure of the allocation resulting
in ENOSPC being reported to the caller.
It is caused by two possible issues. Firstly, we only consider the
lognest free extent and whether it would fit an inode chunk. If the
extent is not correctly aligned, then we can't allocate an inode
chunk in it regardless of the fact that it is large enough. This
tends to be a permanent error until space in the AG is freed.
The second issue is that we don't actually lock the AGI or AGF when
we are doing these checks, and so by the time we get to actually
allocating the inode chunk the space we thought we had in the AG may
have been allocated. This tends to be a spurious error as it
requires a race to trigger. Hence this case is ignored in this patch
as the reported problem is for permanent errors.
The first issue could be addressed by simply taking into account the
alignment when checking the longest extent. This, however, would
prevent allocation in AGs that have aligned, exact sized extents
free. However, this case should be fairly rare compared to the
number of allocations that occur near ENOSPC that would trigger this
condition.
Hence, when selecting the inode AG, take into account the inode
cluster alignment when checking the lognest free extent in the AG.
If we can't find any AGs with a contiguous free space large
enough to be aligned, drop the alignment addition and just try for
an AG that has enough contiguous free space available for an inode
chunk. This won't prevent issues from occurring, but should avoid
situations where other AGs have lots of free space but the selected
AG can't allocate due to alignment constraints.
Reported-by: Arkadiusz Miskiewicz <arekm@maven.pl>
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Brian Foster <bfoster@redhat.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>
If extsize is set and new_last_fsb is larger than 32 bits, the
roundup to extsize will overflow the align variable. Instead,
combine alignments by rounding stripe size up to extsize.
Signed-off-by: Peter Watkins <treestem@gmail.com>
Reviewed-by: Nathaniel W. Turner <nate@houseofnate.net>
Reviewed-by: Brian Foster <bfoster@redhat.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>
fs/xfs/libxfs/xfs_bmap.c:5591:1-6: WARNING: end returns can be simpified
Simplify a trivial if-return sequence. Possibly combine with a
preceding function call.
Generated by: scripts/coccinelle/misc/simple_return.cocci
CC: Brian Foster <bfoster@redhat.com>
Signed-off-by: Fengguang Wu <fengguang.wu@intel.com>
fs/xfs/xfs_file.c:919:1-6: WARNING: end returns can be simpified and declaration on line 902 can be dropped
Simplify a trivial if-return sequence. Possibly combine with a
preceding function call.
Generated by: scripts/coccinelle/misc/simple_return.cocci
Signed-off-by: Fengguang Wu <fengguang.wu@intel.com>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>
fs/xfs/libxfs/xfs_ialloc.c:1141:1-6: WARNING: end returns can be simpified
Simplify a trivial if-return sequence. Possibly combine with a
preceding function call.
Generated by: scripts/coccinelle/misc/simple_return.cocci
Signed-off-by: Fengguang Wu <fengguang.wu@intel.com>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>
The functions xfs_blkdev_put() and xfs_qm_dqrele() test whether
their argument is NULL and then return immediately. Thus the test
around the call is not needed.
This issue was detected by using the Coccinelle software.
Signed-off-by: Markus Elfring <elfring@users.sourceforge.net>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>
More on-disk format consolidation.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>
More on-disk format consolidation. A few declarations that weren't on-disk
format related move into better suitable spots.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>
Move the on-disk ACL format to xfs_format.h, so that repair can
use the common defintion.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>
More consolidatation for the on-disk format defintions. Note that the
XFS_IS_REALTIME_INODE moves to xfs_linux.h instead as it is not related
to the on disk format, but depends on a CONFIG_ option.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>
Here blkno is a daddr_t, which is a __s64; it's possible to hold
a value which is negative, and thus pass the (blkno >= eofs)
test. Then we try to do a xfs_perag_get() for a ridiculous
agno via xfs_daddr_to_agno(), and bad things happen when that
fails, and returns a null pag which is dereferenced shortly
thereafter.
Found via a user-supplied fuzzed image...
Signed-off-by: Eric Sandeen <sandeen@redhat.com>
Reviewed-by: Mark Tinguely <tinguely@sgi.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>
The expectation since the introduction the lazy superblock counters is
that the counters are synced and superblock logged appropriately as part
of the filesystem freeze sequence. This does not occur, however, due to
the logic in xfs_fs_writable() that prevents progress when the fs is in
any state other than SB_UNFROZEN.
While this is a bug, it has not been exposed to date because the last
thing XFS does during freeze is dirty the log. The log recovery process
recalculates the counters from AGI/AGF metadata to ensure everything is
correct. Therefore should a crash occur while an fs is frozen, the
subsequent log recovery puts everything back in order. See the following
commit for reference:
92821e2b [XFS] Lazy Superblock Counters
We might not always want to rely on dirtying the log on a frozen fs.
Modify xfs_log_sbcount() to proceed when the filesystem is freezing but
not once the freeze process has completed. Modify xfs_fs_writable() to
accept the minimum freeze level for which modifications should be
blocked to support various codepaths.
Signed-off-by: Brian Foster <bfoster@redhat.com>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>
The error handling in xfs_qm_log_quotaoff() has a couple problems. If
xfs_trans_commit() fails, we fall through to the error block and call
xfs_trans_cancel(). This is incorrect on commit failure. If
xfs_trans_reserve() fails, we jump to the error block, cancel the tp and
restore the superblock qflags to oldsbqflag. However, oldsbqflag has
been initialized to zero and not yet updated from the original flags so
we set the flags to zero.
Fix up the error handling in xfs_qm_log_quotaoff() to not restore flags
if they haven't been modified and not cancel the tp on commit failure.
Remove the flag restore code altogether because commit error is the only
failure condition and we don't know whether the transaction made it to
disk.
Reported-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: Brian Foster <bfoster@redhat.com>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>
There's no need to store a full struct xfs_trans_res on the stack in
xfs_create() and copy the fields. Use a pointer to the appropriate
structures embedded in the xfs_mount.
Signed-off-by: Brian Foster <bfoster@redhat.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Dave Chinner <david@fromorbit.com>
The xfslogd workqueue is a global, single-job workqueue for buffer ioend
processing. This means we allow for a single work item at a time for all
possible XFS mounts on a system. fsstress testing in loopback XFS over
XFS configurations has reproduced xfslogd deadlocks due to the single
threaded nature of the queue and dependencies introduced between the
separate XFS instances by online discard (-o discard).
Discard over a loopback device converts the discard request to a hole
punch (fallocate) on the underlying file. Online discard requests are
issued synchronously and from xfslogd context in XFS, hence the xfslogd
workqueue is blocked in the upper fs waiting on a hole punch request to
be servied in the lower fs. If the lower fs issues I/O that depends on
xfslogd to complete, both filesystems end up hung indefinitely. This is
reproduced reliabily by generic/013 on XFS->loop->XFS test devices with
the '-o discard' mount option.
Further, docker implementations appear to use this kind of configuration
for container instance filesystems by default (container fs->dm->
loop->base fs) and therefore are subject to this deadlock when running
on XFS.
Replace the global xfslogd workqueue with a per-mount variant. This
guarantees each mount access to a single worker and prevents deadlocks
due to inter-fs dependencies introduced by discard. Since the queue is
only responsible for buffer iodone processing at this point in time,
rename xfslogd to xfs-buf.
Signed-off-by: Brian Foster <bfoster@redhat.com>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>
We support user, group, and project quotas. Tell VFS about it.
CC: xfs@oss.sgi.com
CC: Dave Chinner <david@fromorbit.com>
Acked-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Jan Kara <jack@suse.cz>
The bulkstat main loop progress is tracked by the "lastino"
variable, which is a full 64 bit inode. However, the loop actually
works on agno/agino pairs, and so there's a significant disconnect
between the rest of the loop and the main cursor. Convert this to
use the agino, and pass the agino into the chunk formatting function
and convert it too.
This gets rid of the inconsistency in the loop processing, and
finally makes it simple for us to skip inodes at any point in the
loop simply by incrementing the agino cursor.
cc: <stable@vger.kernel.org> # 3.17
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Brian Foster <bfoster@redhat.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>
The error propagation is a horror - xfs_bulkstat() returns
a rval variable which is only set if there are formatter errors. Any
sort of btree walk error or corruption will cause the bulkstat walk
to terminate but will not pass an error back to userspace. Worse
is the fact that formatter errors will also be ignored if any inodes
were correctly formatted into the user buffer.
Hence bulkstat can fail badly yet still report success to userspace.
This causes significant issues with xfsdump not dumping everything
in the filesystem yet reporting success. It's not until a restore
fails that there is any indication that the dump was bad and tha
bulkstat failed. This patch now triggers xfsdump to fail with
bulkstat errors rather than silently missing files in the dump.
This now causes bulkstat to fail when the lastino cookie does not
fall inside an existing inode chunk. The pre-3.17 code tolerated
that error by allowing the code to move to the next inode chunk
as the agino target is guaranteed to fall into the next btree
record.
With the fixes up to this point in the series, xfsdump now passes on
the troublesome filesystem image that exposes all these bugs.
cc: <stable@vger.kernel.org>
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Brian Foster <bfoster@redhat.com>
There are a bunch of variables tha tare more wildy scoped than they
need to be, obfuscated user buffer checks and tortured "next inode"
tracking. This all needs cleaning up to expose the real issues that
need fixing.
cc: <stable@vger.kernel.org> # 3.17
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Brian Foster <bfoster@redhat.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>
The loop construct has issues:
- clustidx is completely unused, so remove it.
- the loop tries to be smart by terminating when the
"freecount" tells it that all inodes are free. Just drop
it as in most cases we have to scan all inodes in the
chunk anyway.
- move the "user buffer left" condition check to the only
point where we consume space int eh user buffer.
- move the initialisation of agino out of the loop, leaving
just a simple loop control logic using the clusteridx.
Also, double handling of the user buffer variables leads to problems
tracking the current state - use the cursor variables directly
rather than keeping local copies and then having to update the
cursor before returning.
cc: <stable@vger.kernel.org> # 3.17
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Brian Foster <bfoster@redhat.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>
The xfs_bulkstat_agichunk formatting cursor takes buffer values from
the main loop and passes them via the structure to the chunk
formatter, and the writes the changed values back into the main loop
local variables. Unfortunately, this complex dance is full of corner
cases that aren't handled correctly.
The biggest problem is that it is double handling the information in
both the main loop and the chunk formatting function, leading to
inconsistent updates and endless loops where progress is not made.
To fix this, push the struct xfs_bulkstat_agichunk outwards to be
the primary holder of user buffer information. this removes the
double handling in the main loop.
Also, pass the last inode processed by the chunk formatter as a
separate parameter as it purely an output variable and is not
related to the user buffer consumption cursor.
Finally, the chunk formatting code is not shared by anyone, so make
it local to xfs_itable.c.
cc: <stable@vger.kernel.org> # 3.17
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Brian Foster <bfoster@redhat.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>
The bulkstat code has several different ways of detecting the end of
an AG when doing a walk. They are not consistently detected, and the
code that checks for the end of AG conditions is not consistently
coded. Hence the are conditions where the walk code can get stuck in
an endless loop making no progress and not triggering any
termination conditions.
Convert all the "tmp/i" status return codes from btree operations
to a common name (stat) and apply end-of-ag detection to these
operations consistently.
cc: <stable@vger.kernel.org> # 3.17
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Brian Foster <bfoster@redhat.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>
The zero range operation is analogous to fallocate with the exception of
converting the range to zeroes. E.g., it attempts to allocate zeroed
blocks over the range specified by the caller. The XFS implementation
kills all delalloc blocks currently over the aligned range, converts the
range to allocated zero blocks (unwritten extents) and handles the
partial pages at the ends of the range by sending writes through the
pagecache.
The current implementation suffers from several problems associated with
inode size. If the aligned range covers an extending I/O, said I/O is
discarded and an inode size update from a previous write never makes it
to disk. Further, if an unaligned zero range extends beyond eof, the
page write induced for the partial end page can itself increase the
inode size, even if the zero range request is not supposed to update
i_size (via KEEP_SIZE, similar to an fallocate beyond EOF).
The latter behavior not only incorrectly increases the inode size, but
can lead to stray delalloc blocks on the inode. Typically, post-eof
preallocation blocks are either truncated on release or inode eviction
or explicitly written to by xfs_zero_eof() on natural file size
extension. If the inode size increases due to zero range, however,
associated blocks leak into the address space having never been
converted or mapped to pagecache pages. A direct I/O to such an
uncovered range cannot convert the extent via writeback and will BUG().
For example:
$ xfs_io -fc "pwrite 0 128k" -c "fzero -k 1m 54321" <file>
...
$ xfs_io -d -c "pread 128k 128k" <file>
<BUG>
If the entire delalloc extent happens to not have page coverage
whatsoever (e.g., delalloc conversion couldn't find a large enough free
space extent), even a full file writeback won't convert what's left of
the extent and we'll assert on inode eviction.
Rework xfs_zero_file_space() to avoid buffered I/O for partial pages.
Use the existing hole punch and prealloc mechanisms as primitives for
zero range. This implementation is not efficient nor ideal as we
writeback dirty data over the range and remove existing extents rather
than convert to unwrittern. The former writeback, however, is currently
the only mechanism available to ensure consistency between pagecache and
extent state. Even a pagecache truncate/delalloc punch prior to hole
punch has lead to inconsistencies due to racing with writeback.
This provides a consistent, correct implementation of zero range that
survives fsstress/fsx testing without assert failures. The
implementation can be optimized from this point forward once the
fundamental issue of pagecache and delalloc extent state consistency is
addressed.
Signed-off-by: Brian Foster <bfoster@redhat.com>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>
xfs_bulkstat() doesn't check error return from xfs_btree_increment(). In
case of specific fs corruption that could result in xfs_bulkstat()
entering an infinite loop because we would be looping over the same
chunk over and over again. Fix the problem by checking the return value
and terminating the loop properly.
Coverity-id: 1231338
cc: <stable@vger.kernel.org>
Signed-off-by: Jan Kara <jack@suse.cz>
Reviewed-by: Jie Liu <jeff.u.liu@gmail.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>
The recent refactoring of the bulkstat code left a small landmine in
the code. If a inobt read fails, then the tree walk is aborted and
returns without releasing the AGI buffer or freeing the cursor. This
can lead to a subsequent bulkstat call hanging trying to grab the
AGI buffer again.
cc: <stable@vger.kernel.org>
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Brian Foster <bfoster@redhat.com>
Reviewed-by: Eric Sandeen <sandeen@redhat.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>
Pull core block layer changes from Jens Axboe:
"This is the core block IO pull request for 3.18. Apart from the new
and improved flush machinery for blk-mq, this is all mostly bug fixes
and cleanups.
- blk-mq timeout updates and fixes from Christoph.
- Removal of REQ_END, also from Christoph. We pass it through the
->queue_rq() hook for blk-mq instead, freeing up one of the request
bits. The space was overly tight on 32-bit, so Martin also killed
REQ_KERNEL since it's no longer used.
- blk integrity updates and fixes from Martin and Gu Zheng.
- Update to the flush machinery for blk-mq from Ming Lei. Now we
have a per hardware context flush request, which both cleans up the
code should scale better for flush intensive workloads on blk-mq.
- Improve the error printing, from Rob Elliott.
- Backing device improvements and cleanups from Tejun.
- Fixup of a misplaced rq_complete() tracepoint from Hannes.
- Make blk_get_request() return error pointers, fixing up issues
where we NULL deref when a device goes bad or missing. From Joe
Lawrence.
- Prep work for drastically reducing the memory consumption of dm
devices from Junichi Nomura. This allows creating clone bio sets
without preallocating a lot of memory.
- Fix a blk-mq hang on certain combinations of queue depths and
hardware queues from me.
- Limit memory consumption for blk-mq devices for crash dump
scenarios and drivers that use crazy high depths (certain SCSI
shared tag setups). We now just use a single queue and limited
depth for that"
* 'for-3.18/core' of git://git.kernel.dk/linux-block: (58 commits)
block: Remove REQ_KERNEL
blk-mq: allocate cpumask on the home node
bio-integrity: remove the needless fail handle of bip_slab creating
block: include func name in __get_request prints
block: make blk_update_request print prefix match ratelimited prefix
blk-merge: don't compute bi_phys_segments from bi_vcnt for cloned bio
block: fix alignment_offset math that assumes io_min is a power-of-2
blk-mq: Make bt_clear_tag() easier to read
blk-mq: fix potential hang if rolling wakeup depth is too high
block: add bioset_create_nobvec()
block: use bio_clone_fast() in blk_rq_prep_clone()
block: misplaced rq_complete tracepoint
sd: Honor block layer integrity handling flags
block: Replace strnicmp with strncasecmp
block: Add T10 Protection Information functions
block: Don't merge requests if integrity flags differ
block: Integrity checksum flag
block: Relocate bio integrity flags
block: Add a disk flag to block integrity profile
block: Add prefix to block integrity profile flags
...
caused a regression in xfs_inumbers, which in turn broke
xfsdump, causing incomplete dumps.
The loop in xfs_inumbers() needs to fill the user-supplied
buffers, and iterates via xfs_btree_increment, reading new
ags as needed.
But the first time through the loop, if xfs_btree_increment()
succeeds, we continue, which triggers the ++agno at the bottom
of the loop, and we skip to soon to the next ag - without
the proper setup under next_ag to read the next ag.
Fix this by removing the agno increment from the loop conditional,
and only increment agno if we have actually hit the code under
the next_ag: target.
Cc: stable@vger.kernel.org
Signed-off-by: Eric Sandeen <sandeen@redhat.com>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>
Commit 3013683 ("xfs: remove all the inodes on a buffer from the AIL
in bulk") made the xfs inode flush callback more efficient by
combining all the inode writes on the buffer and the deletions of
the inode log item from AIL.
The initial loop in this patch should be looping through all
the log items on the buffer to see which items have
xfs_iflush_done as their callback function. But currently,
only the log item passed to the function has its callback
compared to xfs_iflush_done. If the log item pointer passed to
the function does have the xfs_iflush_done callback function,
then all the log items on the buffer are removed from the
li_bio_list on the buffer b_fspriv and could be removed from
the AIL even though they may have not been written yet.
This problem is masked by the fact that currently all inodes on a
buffer will have the same calback function - either xfs_iflush_done
or xfs_istale_done - and hence the bug cannot manifest in any way.
Still, we need to remove the landmine so that if we add new
callbacks in future this doesn't cause us problems.
Signed-off-by: Mark Tinguely <tinguely@sgi.com>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>
XFS currently discards delalloc blocks within the target range of a
zero range request. Unaligned start and end offsets are zeroed
through the page cache and the internal, aligned blocks are
converted to unwritten extents.
If EOF is page aligned and covered by a delayed allocation extent.
The inode size is not updated until I/O completion. If a zero range
request discards a delalloc range that covers page aligned EOF as
such, the inode size update never occurs. For example:
$ rm -f /mnt/file
$ xfs_io -fc "pwrite 0 64k" -c "zero 60k 4k" /mnt/file
$ stat -c "%s" /mnt/file
65536
$ umount /mnt
$ mount <dev> /mnt
$ stat -c "%s" /mnt/file
61440
Update xfs_zero_file_space() to flush the range rather than discard
delalloc blocks to ensure that inode size updates occur
appropriately.
[dchinner: Note that this is really a workaround to avoid the
underlying problems. More work is needed (and ongoing) to fix those
issues so this fix is being added as a temporary stop-gap measure. ]
Signed-off-by: Brian Foster <bfoster@redhat.com>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>
xfs_vm_writepage() walks each buffer_head on the page, maps to the block
on disk and attaches to a running ioend structure that represents the
I/O submission. A new ioend is created when the type of I/O (unwritten,
delayed allocation or overwrite) required for a particular buffer_head
differs from the previous. If a buffer_head is a delalloc or unwritten
buffer, the associated bits are cleared by xfs_map_at_offset() once the
buffer_head is added to the ioend.
The process of mapping each buffer_head occurs in xfs_map_blocks() and
acquires the ilock in blocking or non-blocking mode, depending on the
type of writeback in progress. If the lock cannot be acquired for
non-blocking writeback, we cancel the ioend, redirty the page and
return. Writeback will revisit the page at some later point.
Note that we acquire the ilock for each buffer on the page. Therefore
during non-blocking writeback, it is possible to add an unwritten buffer
to the ioend, clear the unwritten state, fail to acquire the ilock when
mapping a subsequent buffer and cancel the ioend. If this occurs, the
unwritten status of the buffer sitting in the ioend has been lost. The
page will eventually hit writeback again, but xfs_vm_writepage() submits
overwrite I/O instead of unwritten I/O and does not perform unwritten
extent conversion at I/O completion. This leads to data corruption
because unwritten extents are treated as holes on reads and zeroes are
returned instead of reading from disk.
Modify xfs_cancel_ioend() to restore the buffer unwritten bit for ioends
of type XFS_IO_UNWRITTEN. This ensures that unwritten extent conversion
occurs once the page is eventually written back.
Signed-off-by: Brian Foster <bfoster@redhat.com>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>
Coverity spotted this.
Granted, we *just* checked xfs_inod_dquot() in the caller (by
calling xfs_quota_need_throttle). However, this is the only place we
don't check the return value but the check is cheap and future-proof
so add it.
Signed-off-by: Eric Sandeen <sandeen@redhat.com>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>
I discovered this in userspace, but the same change applies
to the kernel.
If we xfs_mdrestore an image from a non-crc filesystem, lo
and behold the restored image has gained a CRC:
# db/xfs_metadump.sh -o /dev/sdc1 - | xfs_mdrestore - test.img
# xfs_db -c "sb 0" -c "p crc" /dev/sdc1
crc = 0 (correct)
# xfs_db -c "sb 0" -c "p crc" test.img
crc = 0xb6f8d6a0 (correct)
This is because xfs_sb_from_disk doesn't fill in sb_crc,
but xfs_sb_to_disk(XFS_SB_ALL_BITS) does write the in-memory
CRC to disk - so we get uninitialized memory on disk.
Fix this by always initializing sb_crc to 0 when we read
the superblock, and masking out the CRC bit from ALL_BITS
when we write it.
Signed-off-by: Eric Sandeen <sandeen@redhat.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Dave Chinner <david@fromorbit.com>
