While xfs_buftarg_shrink() is freeing buffers from the dispose list (filled with
buffers from lru list), there is a possibility to have xfs_buf_stale() racing
with it, and removing buffers from dispose list before xfs_buftarg_shrink() does
it.

This happens because xfs_buftarg_shrink() handle the dispose list without
locking and the test condition in xfs_buf_stale() checks for the buffer being in
*any* list:

if (!list_empty(&bp->b_lru))

If the buffer happens to be on dispose list, this causes the buffer counter of
lru list (btp->bt_lru_nr) to be decremented twice (once in xfs_buftarg_shrink()
and another in xfs_buf_stale()) causing a wrong account usage of the lru list.

This may cause xfs_buftarg_shrink() to return a wrong value to the memory
shrinker shrink_slab(), and such account error may also cause an underflowed
value to be returned; since the counter is lower than the current number of
items in the lru list, a decrement may happen when the counter is 0, causing
an underflow on the counter.

The fix uses a new flag field (and a new buffer flag) to serialize buffer
handling during the shrink process. The new flag field has been designed to use
btp->bt_lru_lock/unlock instead of xfs_buf_lock/unlock mechanism.

dchinner, sandeen, aquini and aris also deserve credits for this.
Signed-off-by: NCarlos Maiolino <cmaiolino@redhat.com>
Reviewed-by: NBen Myers <bpm@sgi.com>
Reviewed-by: NDave Chinner <dchinner@redhat.com>
Signed-off-by: NBen Myers <bpm@sgi.com>

If range.start or range.minlen is bigger than filesystem size, return
invalid value error. This fixes possible overflow in BTOBB macro when
passed value was nearly ULLONG_MAX.
Signed-off-by: NTomas Racek <tracek@redhat.com>
Reviewed-by: NDave Chinner <dchinner@redhat.com>
Signed-off-by: NBen Myers <bpm@sgi.com>

Also update some commens in the area to make the code easier to read.
Signed-off-by: NChristoph Hellwig <hch@lst.de>
Reviewed-by: NMark Tinguely <tinguely@sgi.com>
Reviewed-by: NDave Chinner <dchinner@redhat.com>
Signed-off-by: NBen Myers <bpm@sgi.com>

I noticed that "struct xfs_mount_args" was still declared in
"fs/xfs/xfs_mount.h".  That struct doesn't even exist any more (and
is obviously not referenced elsewhere in that header file).  While
in there, delete four other unneeded struct declarations in that
file.

Doing so highlights that "fs/xfs/xfs_trace.h" was relying indirectly
on "xfs_mount.h" to be #included in order to declare "struct
xfs_bmbt_irec", so add that declaration to resolve that issue.
Signed-off-by: NAlex Elder <elder@inktank.com>
Signed-off-by: NBen Myers <bpm@sgi.com>

Results in this assert failure in generic/090:

XFS: Assertion failed: *nmap >= 1, file: fs/xfs/xfs_bmap.c, line: 4363
.....
Call Trace:
 [<ffffffff814680db>] xfs_bmapi_read+0x6b/0x370
 [<ffffffff814b64b2>] xfs_rtbuf_get+0x42/0x130
 [<ffffffff814b6f09>] xfs_rtget_summary+0x89/0x120
 [<ffffffff814b7bfe>] xfs_rtallocate_extent_size+0xce/0x340
 [<ffffffff814b89f0>] xfs_rtallocate_extent+0x240/0x290
 [<ffffffff81462c1a>] xfs_bmap_rtalloc+0x1ba/0x340
 [<ffffffff81463a65>] xfs_bmap_alloc+0x35/0x40
 [<ffffffff8146f111>] xfs_bmapi_allocate+0xf1/0x350
 [<ffffffff8146f9de>] xfs_bmapi_write+0x66e/0xa60
 [<ffffffff8144538a>] xfs_iomap_write_direct+0x22a/0x3f0
 [<ffffffff8143707b>] __xfs_get_blocks+0x38b/0x5d0
 [<ffffffff814372d4>] xfs_get_blocks_direct+0x14/0x20
 [<ffffffff811b0081>] do_blockdev_direct_IO+0xf71/0x1eb0
 [<ffffffff811b1015>] __blockdev_direct_IO+0x55/0x60
 [<ffffffff814355ca>] xfs_vm_direct_IO+0x11a/0x1e0
 [<ffffffff8112d617>] generic_file_direct_write+0xd7/0x1b0
 [<ffffffff8143e16c>] xfs_file_dio_aio_write+0x13c/0x320
 [<ffffffff8143e6f2>] xfs_file_aio_write+0x1c2/0x1d0
 [<ffffffff81174a07>] do_sync_write+0xa7/0xe0
 [<ffffffff81175288>] vfs_write+0xa8/0x160
 [<ffffffff81175702>] sys_pwrite64+0x92/0xb0
 [<ffffffff81b68f69>] system_call_fastpath+0x16/0x1b
Signed-off-by: NDave Chinner <dchinner@redhat.com>
Signed-off-by: NBen Myers <bpm@sgi.com>

Generic code now blocks all writers from standard write paths. So we add
blocking of all writers coming from ioctl (we get a protection of ioctl against
racing remount read-only as a bonus) and convert xfs_file_aio_write() to a
non-racy freeze protection. We also keep freeze protection on transaction
start to block internal filesystem writes such as removal of preallocated
blocks.

CC: Ben Myers <bpm@sgi.com>
CC: Alex Elder <elder@kernel.org>
CC: xfs@oss.sgi.com
Signed-off-by: NJan Kara <jack@suse.cz>
Signed-off-by: NAl Viro <viro@zeniv.linux.org.uk>

v2: Add the xfs_buf_lock to xfs_quiesce_attr().
    Add explaination why xfs_buf_lock() is used to wait for write.

xfs_wait_buftarg() does not wait for the completion of the write of the
uncached superblock. This write can race with the shutdown of the log
and causes a panic if the write does not win the race.

During the log write, xfsaild_push() will lock the buffer and set the
XBF_ASYNC flag. Because the XBF_FLAG is set, complete() is not performed
on the buffer's iowait entry, we cannot call xfs_buf_iowait() to wait
for the write to complete. The buffer's lock is held until the write is
complete, so we can block on a xfs_buf_lock() request to be notified
that the write is complete.
Signed-off-by: NMark Tinguely <tinguely@sgi.com>
Reviewed-by: NChristoph Hellwig <hch@lst.de>
Signed-off-by: NBen Myers <bpm@sgi.com>

xfsaild idle mode logic currently leads to a couple hangs:

1.) If xfsaild is rescheduled in during an incremental scan
    (i.e., tout != 0) and the target has been updated since
    the previous run, we can hit the new target and go into
    idle mode with a still populated ail.
2.) A wake up is only issued when the target is pushed forward.
    The wake up can race with xfsaild if it is currently in the
    process of entering idle mode, causing future wake up
    events to be lost.

These hangs have been reproduced and verified as fixed by
running xfstests 273 in a loop on a slightly modified upstream
kernel. The kernel is modified to re-enable idle mode as
previously implemented (when count == 0) and with a revert of
commit 670ce93f, which includes performance improvements that
make this harder to reproduce.

The solution, the algorithm for which has been outlined by
Dave Chinner, is to modify xfsaild to enter idle mode only when
the ail is empty and the push target has not been moved forward
since the last push.
Signed-off-by: NBrian Foster <bfoster@redhat.com>
Reviewed-by: NDave Chinner <dchinner@redhat.com>
Reviewed-by: NChristoph Hellwig <hch@lst.de>
Signed-off-by: NBen Myers <bpm@sgi.com>

Content-Disposition: inline; filename=xfs-remove-iolock-classes

Now that we never take the iolock during inode reclaim we don't need
to play games with lock classes.
Signed-off-by: NChristoph Hellwig <hch@lst.de>
Reviewed-by: NRich Johnston <rjohnston@sgi.com>
Signed-off-by: NBen Myers <bpm@sgi.com>

Same rational as the last patch - these inodes are not reachable, so
don't bother with locking.
Signed-off-by: NChristoph Hellwig <hch@lst.de>
Reviewed-by: NRich Johnston <rjohnston@sgi.com>
Signed-off-by: NBen Myers <bpm@sgi.com>

An inode that enters xfs_inactive has been removed from all global
lists but the inode hash, and can't be recycled in xfs_iget before
it has been marked reclaimable.  Thus taking the iolock in here
is not nessecary at all, and given the amount of lockdep false
positives it has triggered already I'd rather remove the locking.

The only change outside of xfs_inactive is relaxing an assert in
xfs_itruncate_extents.
Signed-off-by: NChristoph Hellwig <hch@lst.de>
Reviewed-by: NRich Johnston <rjohnston@sgi.com>
Signed-off-by: NBen Myers <bpm@sgi.com>

Remove this helper as the code flow is a lot more obvious when it gets
merged into its only caller.
Signed-off-by: NChristoph Hellwig <hch@lst.de>
Reviewed-by: NRich Johnston <rjohnston@sgi.com>
Signed-off-by: NBen Myers <bpm@sgi.com>

The code to reserve log space and join the inode to the transaction is
common for all cases, so don't duplicate it.  Also remove the trivial
xfs_inactive_symlink_local helper which can simply be opencode now.
Signed-off-by: NChristoph Hellwig <hch@lst.de>
Reviewed-by: NRich Johnston <rjohnston@sgi.com>
Signed-off-by: NBen Myers <bpm@sgi.com>

Refactor the AG selection loop in xfs_dialloc to operate on the in-memory
perag data as much as possible.  We only read the AGI buffer once we have
selected an AG to allocate inodes now instead of for every AG considered.
Signed-off-by: NChristoph Hellwig <hch@lst.de>
Reviewed-by: NDave Chinner <dchinner@redhat.com>
Reviewed-by: NMark Tinguely <tinguely@sgi.com>
Signed-off-by: NBen Myers <bpm@sgi.com>

Loop over the in-core perag structures and prefer using pagi_freecount over
going out to the AGI buffer where possible.
Signed-off-by: NChristoph Hellwig <hch@lst.de>
Reviewed-by: NDave Chinner <dchinner@redhat.com>
Reviewed-by: NMark Tinguely <tinguely@sgi.com>
Signed-off-by: NBen Myers <bpm@sgi.com>

In this case we already have selected an AG and know it has free space
beause the buffer lock never got released.  Jump directly into xfs_dialloc_ag
and short cut the AG selection loop.
Signed-off-by: NChristoph Hellwig <hch@lst.de>
Reviewed-by: NDave Chinner <dchinner@redhat.com>
Reviewed-by: NMark Tinguely <tinguely@sgi.com>
Signed-off-by: NBen Myers <bpm@sgi.com>

We can simplify check the IO_agbp pointer for being non-NULL instead of
passing another argument through two layers of function calls.
Signed-off-by: NChristoph Hellwig <hch@lst.de>
Reviewed-by: NDave Chinner <dchinner@redhat.com>
Reviewed-by: NMark Tinguely <tinguely@sgi.com>
Signed-off-by: NBen Myers <bpm@sgi.com>

Move the actual allocation once we have selected an allocation group into a
separate helper, and make xfs_dialloc a wrapper around it.
Signed-off-by: NChristoph Hellwig <hch@lst.de>
Reviewed-by: NDave Chinner <dchinner@redhat.com>
Reviewed-by: NBen Myers <bpm@sgi.com>
Signed-off-by: NBen Myers <bpm@sgi.com>

Signed-off-by: NAl Viro <viro@zeniv.linux.org.uk>

This function is entirely trivial and only has one caller, so remove it to
simplify the code.
Signed-off-by: NChristoph Hellwig <hch@lst.de>
Reviewed-by: NDave Chinner <dchinner@redhat.com>
Signed-off-by: NBen Myers <bpm@sgi.com>

Add a XFS_ prefix to IO_DIRECT,XFS_IO_DELALLOC, XFS_IO_UNWRITTEN and
XFS_IO_OVERWRITE. This to avoid namespace conflict with other modules.
Signed-off-by: NAlain Renaud <arenaud@sgi.com>
Reviewed-by: NRich Johnston <rjohnston@sgi.com>
Signed-off-by: NBen Myers <bpm@sgi.com>

There is no need to keep this helper around, opencoding it in the only
caller is just as clear.
Signed-off-by: NChristoph Hellwig <hch@lst.de>
Reviewed-by: NMark Tinguely <tinguely@sgi.com>
Signed-off-by: NBen Myers <bpm@sgi.com>

All callers of xfs_imap_to_bp want the dinode pointer, so let's calculate it
inside xfs_imap_to_bp.  Once that is done xfs_itobp becomes a fairly pointless
wrapper which can be replaced with direct calls to xfs_imap_to_bp.
Signed-off-by: NChristoph Hellwig <hch@lst.de>
Reviewed-by: NMark Tinguely <tinguely@sgi.com>
Signed-off-by: NBen Myers <bpm@sgi.com>

We need to zero out part of a page which beyond EOF before setting uptodate,
otherwise, mapread or write will see non-zero data beyond EOF.

Based on the code in fs/buffer.c and the following ext4 commit:

  ext4: handle EOF correctly in ext4_bio_write_page()

And yes, I wish we had a good test case for it.
Signed-off-by: NChristoph Hellwig <hch@lst.de>
Reviewed-by: NMark Tinguely <tinguely@sgi.com>
Signed-off-by: NBen Myers <bpm@sgi.com>

Use this new method to replace our hacky use of ->dirty_inode.  An additional
benefit is that we can now propagate errors up the stack.
Signed-off-by: NChristoph Hellwig <hch@lst.de>
Reviewed-by: NDave Chinner <dchinner@redhat.com>
Signed-off-by: NBen Myers <bpm@sgi.com>

Fix trivial typo error that has written "It" to "Is".
Signed-off-by: NChen Baozi <baozich@gmail.com>
Reviewed-by: NBen Myers <bpm@sgi.com>
Signed-off-by: NBen Myers <bpm@sgi.com>

boolean "does it have to be exclusive?" flag is passed instead;
Local filesystem should just ignore it - the object is guaranteed
not to be there yet.
Signed-off-by: NAl Viro <viro@zeniv.linux.org.uk>

Just the flags; only NFS cares even about that, but there are
legitimate uses for such argument.  And getting rid of that
completely would require splitting ->lookup() into a couple
of methods (at least), so let's leave that alone for now...
Signed-off-by: NAl Viro <viro@zeniv.linux.org.uk>

xfs_bdstrat_cb only adds a check for a shutdown filesystem over
xfs_buf_iorequest, but xfs_buf_iodone_callbacks just checked for a shut down
filesystem a little earlier.  In addition the shutdown handling in
xfs_bdstrat_cb is not very suitable for this caller.
Signed-off-by: NChristoph Hellwig <hch@lst.de>
Reviewed-by: NDave Chinner <dchinner@redhat.com>
Signed-off-by: NBen Myers <bpm@sgi.com>

If the b_iodone handler is run in calling context in xfs_buf_iorequest we
can run into a recursion where xfs_buf_iodone_callbacks keeps calling back
into xfs_buf_iorequest because an I/O error happened, which keeps calling
back into xfs_buf_iorequest.  This chain will usually not take long
because the filesystem gets shut down because of log I/O errors, but even
over a short time it can cause stack overflows if run on the same context.

As a short term workaround make sure we always call the iodone handler in
workqueue context.
Signed-off-by: NChristoph Hellwig <hch@lst.de>
Reviewed-by: NDave Chinner <dchinner@redhat.com>
Signed-off-by: NBen Myers <bpm@sgi.com>

Almost all metadata allocations come from shallow stack usage
situations. Avoid the overhead of switching the allocation to a
workqueue as we are not in danger of running out of stack when
making these allocations. Metadata allocations are already marked
through the args that are passed down, so this is trivial to do.
Signed-off-by: NDave Chinner <dchinner@redhat.com>
Reported-by: NMel Gorman <mgorman@suse.de>
Tested-by: NMel Gorman <mgorman@suse.de>
Signed-off-by: NBen Myers <bpm@sgi.com>

The current cursor is reallocated when retrying the allocation, so
the existing cursor needs to be destroyed in both the restart and
the failure cases.
Signed-off-by: NDave Chinner <dchinner@redhat.com>
Tested-by: NMike Snitzer <snitzer@redhat.com>
Signed-off-by: NBen Myers <bpm@sgi.com>

xfs_bdstrat_cb only adds a check for a shutdown filesystem over
xfs_buf_iorequest, but xfs_buf_iodone_callbacks just checked for a shut down
filesystem a little earlier.  In addition the shutdown handling in
xfs_bdstrat_cb is not very suitable for this caller.
Signed-off-by: NChristoph Hellwig <hch@lst.de>
Reviewed-by: NDave Chinner <dchinner@redhat.com>
Signed-off-by: NBen Myers <bpm@sgi.com>

If the b_iodone handler is run in calling context in xfs_buf_iorequest we
can run into a recursion where xfs_buf_iodone_callbacks keeps calling back
into xfs_buf_iorequest because an I/O error happened, which keeps calling
back into xfs_buf_iorequest.  This chain will usually not take long
because the filesystem gets shut down because of log I/O errors, but even
over a short time it can cause stack overflows if run on the same context.

As a short term workaround make sure we always call the iodone handler in
workqueue context.
Signed-off-by: NChristoph Hellwig <hch@lst.de>
Reviewed-by: NDave Chinner <dchinner@redhat.com>
Signed-off-by: NBen Myers <bpm@sgi.com>

Almost all metadata allocations come from shallow stack usage
situations. Avoid the overhead of switching the allocation to a
workqueue as we are not in danger of running out of stack when
making these allocations. Metadata allocations are already marked
through the args that are passed down, so this is trivial to do.
Signed-off-by: NDave Chinner <dchinner@redhat.com>
Reported-by: NMel Gorman <mgorman@suse.de>
Tested-by: NMel Gorman <mgorman@suse.de>
Signed-off-by: NBen Myers <bpm@sgi.com>

The current cursor is reallocated when retrying the allocation, so
the existing cursor needs to be destroyed in both the restart and
the failure cases.
Signed-off-by: NDave Chinner <dchinner@redhat.com>
Tested-by: NMike Snitzer <snitzer@redhat.com>
Signed-off-by: NBen Myers <bpm@sgi.com>

The buffer reading code in xfs_dir2_leaf_getdents is complex and difficult to
follow due to the readahead and all the context is carries. it is also badly
indented and so difficult to read. Factor it out into a separate function to
make it easier to understand and optimise in future patches.
Signed-off-by: NDave Chinner <dchinner@redhat.com>
Reviewed-by: NChristoph Hellwig <hch@lst.de>
Signed-off-by: NBen Myers <bpm@sgi.com>

The struct xfs_dabuf now only tracks a single xfs_buf and all the
information it holds can be gained directly from the xfs_buf. Hence
we can remove the struct dabuf and pass the xfs_buf around
everywhere.

Kill the struct dabuf and the associated infrastructure.
Signed-off-by: NDave Chinner <dchinner@redhat.com>
Reviewed-by: NChristoph Hellwig <hch@lst.de>
Signed-off-by: NBen Myers <bpm@sgi.com>

First step in converting the directory code to use native
discontiguous buffers and replacing the dabuf construct.
Signed-off-by: NDave Chinner <dchinner@redhat.com>
Reviewed-by: NChristoph Hellwig <hch@lst.de>
Signed-off-by: NBen Myers <bpm@sgi.com>

discontigous buffer in separate buffer format structures. This means log
recovery will recover all the changes on a per segment basis without
requiring any knowledge of the fact that it was logged from a
compound buffer.

To do this, we need to be able to determine what buffer segment any
given offset into the compound buffer sits over. This enables us to
translate the dirty bitmap in the number of separate buffer format
structures required.

We also need to be able to determine the number of bitmap elements
that a given buffer segment has, as this determines the size of the
buffer format structure. Hence we need to be able to determine the
both the start offset into the buffer and the length of a given
segment to be able to calculate this.

With this information, we can preallocate, build and format the
correct log vector array for each segment in a compound buffer to
appear exactly the same as individually logged buffers in the log.
Signed-off-by: NDave Chinner <dchinner@redhat.com>
Reviewed-by: NChristoph Hellwig <hch@lst.de>
Signed-off-by: NBen Myers <bpm@sgi.com>