On Wed, Oct 02, 2024 at 11:56:40AM -0400, Brian Foster wrote: > On Wed, Oct 02, 2024 at 07:57:15AM -0700, Darrick J. Wong wrote: > > On Wed, Oct 02, 2024 at 10:35:55AM -0400, Brian Foster wrote: > > > On Wed, Oct 02, 2024 at 09:38:00PM +0800, Zorro Lang wrote: > > > > On Tue, Oct 01, 2024 at 07:59:44AM -0700, Darrick J. Wong wrote: > > > > > On Tue, Sep 24, 2024 at 10:45:48AM +0200, Christoph Hellwig wrote: > > > > > > From: Dave Chinner <dchinner@xxxxxxxxxx> > > > > > > > > > > > > These tests create substantial file fragmentation as a result of > > > > > > application actions that defeat post-EOF preallocation > > > > > > optimisations. They are intended to replicate known vectors for > > > > > > these problems, and provide a check that the fragmentation levels > > > > > > have been controlled. The mitigations we make may not completely > > > > > > remove fragmentation (e.g. they may demonstrate speculative delalloc > > > > > > related extent size growth) so the checks don't assume we'll end up > > > > > > with perfect layouts and hence check for an exceptable level of > > > > > > fragmentation rather than none. > > > > > > > > > > > > Signed-off-by: Dave Chinner <dchinner@xxxxxxxxxx> > > > > > > [move to different test number, update to current xfstest APIs] > > > > > > Signed-off-by: Christoph Hellwig <hch@xxxxxx> > > > > > > --- > > > > > > tests/xfs/1500 | 66 +++++++++++++++++++++++++++++++++++++++ > > > > > > tests/xfs/1500.out | 9 ++++++ > > > > > > tests/xfs/1501 | 68 ++++++++++++++++++++++++++++++++++++++++ > > > > > > tests/xfs/1501.out | 9 ++++++ > > > > > > tests/xfs/1502 | 68 ++++++++++++++++++++++++++++++++++++++++ > > > > > > tests/xfs/1502.out | 9 ++++++ > > > > > > tests/xfs/1503 | 77 ++++++++++++++++++++++++++++++++++++++++++++++ > > > > > > tests/xfs/1503.out | 33 ++++++++++++++++++++ > > > > > > 8 files changed, 339 insertions(+) > > > > > > create mode 100755 tests/xfs/1500 > > > > > > create mode 100644 tests/xfs/1500.out > > > > > > create mode 100755 tests/xfs/1501 > > > > > > create mode 100644 tests/xfs/1501.out > > > > > > create mode 100755 tests/xfs/1502 > > > > > > create mode 100644 tests/xfs/1502.out > > > > > > create mode 100755 tests/xfs/1503 > > > > > > create mode 100644 tests/xfs/1503.out > > > > > > > > > > > > diff --git a/tests/xfs/1500 b/tests/xfs/1500 > > > > > > new file mode 100755 > > > > > > index 000000000..de0e1df62 > > > > > > --- /dev/null > > > > > > +++ b/tests/xfs/1500 > > > > > > @@ -0,0 +1,66 @@ > > > > > > +#! /bin/bash > > > > > > +# SPDX-License-Identifier: GPL-2.0 > > > > > > +# Copyright (c) 2019 Red Hat, Inc. All Rights Reserved. > > > > > > +# > > > > > > +# FS QA Test xfs/500 > > > > > > +# > > > > > > +# Post-EOF preallocation defeat test for O_SYNC buffered I/O. > > > > > > +# > > > > > > + > > > > > > +. ./common/preamble > > > > > > +_begin_fstest auto quick prealloc rw > > > > > > + > > > > > > +. ./common/rc > > > > > > +. ./common/filter > > > > > > + > > > > > > +_require_scratch > > > > > > + > > > > > > +_cleanup() > > > > > > +{ > > > > > > + # try to kill all background processes > > > > > > + wait > > > > > > + cd / > > > > > > + rm -r -f $tmp.* > > > > > > +} > > > > > > + > > > > > > +_scratch_mkfs > "$seqres.full" 2>&1 > > > > > > +_scratch_mount > > > > > > + > > > > > > +# Write multiple files in parallel using synchronous buffered writes. Aim is to > > > > > > +# interleave allocations to fragment the files. Synchronous writes defeat the > > > > > > +# open/write/close heuristics in xfs_file_release() that prevent EOF block > > > > > > +# removal, so this should fragment badly. Typical problematic behaviour shows > > > > > > +# per-file extent counts of >900 (almost worse case) whilst fixed behaviour > > > > > > +# typically shows extent counts in the low 20s. > > > > > > > > > > Now that these are in for-next, I've noticed that these new tests > > > > > consistently fail in the above-documented manner on various configs -- > > > > > fsdax, always_cow, rtextsize > 1fsb, and sometimes 1k fsblock size. > > > > > > > > > > I'm not sure why this happens, but it probably needs to be looked at > > > > > along with all the FALLOC_FL_UNSHARE_RANGE brokenness that's also been > > > > > exposed by fstests that /does/ need to be fixed. > > > > > > > > Yes, some fsx tests fail on xfs, after the FALLOC_FL_UNSHARE_RANGE supporting. > > > > e.g. g/091, g/127, g/263, g/363 and g/616. I thought they're known issues as > > > > you known. If they're not, better to check. Hi Brian, are these failures as you > > > > known? > > > > > > > > > > So I'm aware of two fundamental issues that fsx unshare range support > > > uncovers. First is the XFS data loss issue that is addressed here[1], > > > second is the iomap unshare range warning/error splat that Julian Sun > > > has been working on (last version posted here[2] I believe). > > > > > > My initial testing of the fsx unshare range patch was to run fsx > > > directly on the fs until I could run for some notable number of > > > operations without triggering a failure (probably at least 1m+, but I > > > don't recall exactly). I was initially able to do that with the patches > > > from [1] plus a local hack to trim to i_size in iomap_unshare_range(), > > > so based on that I _think_ these are the only two outstanding issues > > > with unshare range. > > > > > > [1] https://lore.kernel.org/linux-xfs/20240906114051.120743-1-bfoster@xxxxxxxxxx/ > > > [2] https://lore.kernel.org/linux-fsdevel/20240927065344.2628691-1-sunjunchao2870@xxxxxxxxx/ > > > > > > The patches at [1] have been reviewed, but I'm not really sure where > > > they stand in terms of the XFS pipeline. Carlos? > > > > > > It looks like the fix associated with [2] is still under > > > development/review. In any event, I just ran the set of tests noted by > > > Zorro above (w/ unshare range support) and they all fail on my distro > > > kernel, but all but generic/363 pass on current master (6.12.0-rc1+) > > > plus [1]. The generic/363 failure produces the iomap error associated > > > with [2], so I suspect that all of these test failures can be > > > categorized into one of those two known issues. > > > > Yep. [1] fixes a lot of the splats, and the rest of the splats can be > > fixed by a couple of other patches that I'll send out today. > > > > Those same fsx tests above are still broken on fsdax though, so > > something is still wrong. :( > > > > Ah, Ok.. I hadn't tested for fsdax. Is that fixed by your corresponding > i_size fix, or even with that we're still failing? Note that if it's > isolated to generic/363 and not explicitly the iomap warning it could be > a zeroing issue rather than unshare. That could probably be confirmed by > disabling unshare and the eof pollution bits one at a time in fsx.. Oh, no, it's anything that uses fsx on fsdax. Roughly speaking, the "is this really shared?" predicate didn't get updated when the iomap_unshare_iter version did; the "if srcmap is hole or unwritten" code makes no sense because you can't share holes or unwritten extents; the range to memcpy should have been extended to align with the fsblocks because we cow whole blocks; the rounding down that the dax_iomap_direct_access method does to the returned pointer means it's been corrupting data; and it doesn't invalidate the mappings before doing the cow which means any real fsdax programs access the wrong memory afterwards. IOWs, we've been corrupting data since 2022. Oh and also xfs_direct_write_iomap_begin shouldn't be allocating cow extents for an unshare operation over a hole. So, uh, a big thank you to you for exposing this untested crap code (both iomap and fsdax versions). Patches coming. :) --D > Brian > > > --D > > > > > Brian > > > > > > > Thanks, > > > > Zorro > > > > > > > > > > > > > > --D > > > > > > > > > > > +# Failure is determined by golden output mismatch from _within_tolerance(). > > > > > > + > > > > > > +workfile=$SCRATCH_MNT/file > > > > > > +nfiles=8 > > > > > > +wsize=4096 > > > > > > +wcnt=1000 > > > > > > + > > > > > > +write_sync_file() > > > > > > +{ > > > > > > + idx=$1 > > > > > > + > > > > > > + for ((cnt=0; cnt<$wcnt; cnt++)); do > > > > > > + $XFS_IO_PROG -f -s -c "pwrite $((cnt * wsize)) $wsize" $workfile.$idx > > > > > > + done > > > > > > +} > > > > > > + > > > > > > +rm -f $workfile* > > > > > > +for ((n=0; n<$nfiles; n++)); do > > > > > > + write_sync_file $n > /dev/null 2>&1 & > > > > > > +done > > > > > > +wait > > > > > > +sync > > > > > > + > > > > > > +for ((n=0; n<$nfiles; n++)); do > > > > > > + count=$(_count_extents $workfile.$n) > > > > > > + # Acceptible extent count range is 1-40 > > > > > > + _within_tolerance "file.$n extent count" $count 21 19 -v > > > > > > +done > > > > > > + > > > > > > +status=0 > > > > > > +exit > > > > > > diff --git a/tests/xfs/1500.out b/tests/xfs/1500.out > > > > > > new file mode 100644 > > > > > > index 000000000..414df87ed > > > > > > --- /dev/null > > > > > > +++ b/tests/xfs/1500.out > > > > > > @@ -0,0 +1,9 @@ > > > > > > +QA output created by 1500 > > > > > > +file.0 extent count is in range > > > > > > +file.1 extent count is in range > > > > > > +file.2 extent count is in range > > > > > > +file.3 extent count is in range > > > > > > +file.4 extent count is in range > > > > > > +file.5 extent count is in range > > > > > > +file.6 extent count is in range > > > > > > +file.7 extent count is in range > > > > > > diff --git a/tests/xfs/1501 b/tests/xfs/1501 > > > > > > new file mode 100755 > > > > > > index 000000000..cf3cbf8b5 > > > > > > --- /dev/null > > > > > > +++ b/tests/xfs/1501 > > > > > > @@ -0,0 +1,68 @@ > > > > > > +#! /bin/bash > > > > > > +# SPDX-License-Identifier: GPL-2.0 > > > > > > +# Copyright (c) 2019 Red Hat, Inc. All Rights Reserved. > > > > > > +# > > > > > > +# FS QA Test xfs/501 > > > > > > +# > > > > > > +# Post-EOF preallocation defeat test for buffered I/O with extent size hints. > > > > > > +# > > > > > > + > > > > > > +. ./common/preamble > > > > > > +_begin_fstest auto quick prealloc rw > > > > > > + > > > > > > +. ./common/rc > > > > > > +. ./common/filter > > > > > > + > > > > > > +_require_scratch > > > > > > + > > > > > > +_cleanup() > > > > > > +{ > > > > > > + # try to kill all background processes > > > > > > + wait > > > > > > + cd / > > > > > > + rm -r -f $tmp.* > > > > > > +} > > > > > > + > > > > > > +_scratch_mkfs > "$seqres.full" 2>&1 > > > > > > +_scratch_mount > > > > > > + > > > > > > +# Write multiple files in parallel using buffered writes with extent size hints. > > > > > > +# Aim is to interleave allocations to fragment the files. Writes w/ extent size > > > > > > +# hints set defeat the open/write/close heuristics in xfs_file_release() that > > > > > > +# prevent EOF block removal, so this should fragment badly. Typical problematic > > > > > > +# behaviour shows per-file extent counts of 1000 (worst case!) whilst > > > > > > +# fixed behaviour should show very few extents (almost best case). > > > > > > +# > > > > > > +# Failure is determined by golden output mismatch from _within_tolerance(). > > > > > > + > > > > > > +workfile=$SCRATCH_MNT/file > > > > > > +nfiles=8 > > > > > > +wsize=4096 > > > > > > +wcnt=1000 > > > > > > +extent_size=16m > > > > > > + > > > > > > +write_extsz_file() > > > > > > +{ > > > > > > + idx=$1 > > > > > > + > > > > > > + $XFS_IO_PROG -f -c "extsize $extent_size" $workfile.$idx > > > > > > + for ((cnt=0; cnt<$wcnt; cnt++)); do > > > > > > + $XFS_IO_PROG -f -c "pwrite $((cnt * wsize)) $wsize" $workfile.$idx > > > > > > + done > > > > > > +} > > > > > > + > > > > > > +rm -f $workfile* > > > > > > +for ((n=0; n<$nfiles; n++)); do > > > > > > + write_extsz_file $n > /dev/null 2>&1 & > > > > > > +done > > > > > > +wait > > > > > > +sync > > > > > > + > > > > > > +for ((n=0; n<$nfiles; n++)); do > > > > > > + count=$(_count_extents $workfile.$n) > > > > > > + # Acceptible extent count range is 1-10 > > > > > > + _within_tolerance "file.$n extent count" $count 2 1 8 -v > > > > > > +done > > > > > > + > > > > > > +status=0 > > > > > > +exit > > > > > > diff --git a/tests/xfs/1501.out b/tests/xfs/1501.out > > > > > > new file mode 100644 > > > > > > index 000000000..a266ef74b > > > > > > --- /dev/null > > > > > > +++ b/tests/xfs/1501.out > > > > > > @@ -0,0 +1,9 @@ > > > > > > +QA output created by 1501 > > > > > > +file.0 extent count is in range > > > > > > +file.1 extent count is in range > > > > > > +file.2 extent count is in range > > > > > > +file.3 extent count is in range > > > > > > +file.4 extent count is in range > > > > > > +file.5 extent count is in range > > > > > > +file.6 extent count is in range > > > > > > +file.7 extent count is in range > > > > > > diff --git a/tests/xfs/1502 b/tests/xfs/1502 > > > > > > new file mode 100755 > > > > > > index 000000000..f4228667a > > > > > > --- /dev/null > > > > > > +++ b/tests/xfs/1502 > > > > > > @@ -0,0 +1,68 @@ > > > > > > +#! /bin/bash > > > > > > +# SPDX-License-Identifier: GPL-2.0 > > > > > > +# Copyright (c) 2019 Red Hat, Inc. All Rights Reserved. > > > > > > +# > > > > > > +# FS QA Test xfs/502 > > > > > > +# > > > > > > +# Post-EOF preallocation defeat test for direct I/O with extent size hints. > > > > > > +# > > > > > > + > > > > > > +. ./common/preamble > > > > > > +_begin_fstest auto quick prealloc rw > > > > > > + > > > > > > +. ./common/rc > > > > > > +. ./common/filter > > > > > > + > > > > > > +_require_scratch > > > > > > + > > > > > > +_cleanup() > > > > > > +{ > > > > > > + # try to kill all background processes > > > > > > + wait > > > > > > + cd / > > > > > > + rm -r -f $tmp.* > > > > > > +} > > > > > > + > > > > > > +_scratch_mkfs > "$seqres.full" 2>&1 > > > > > > +_scratch_mount > > > > > > + > > > > > > +# Write multiple files in parallel using O_DIRECT writes w/ extent size hints. > > > > > > +# Aim is to interleave allocations to fragment the files. O_DIRECT writes defeat > > > > > > +# the open/write/close heuristics in xfs_file_release() that prevent EOF block > > > > > > +# removal, so this should fragment badly. Typical problematic behaviour shows > > > > > > +# per-file extent counts of ~1000 (worst case) whilst fixed behaviour typically > > > > > > +# shows extent counts in the low single digits (almost best case) > > > > > > +# > > > > > > +# Failure is determined by golden output mismatch from _within_tolerance(). > > > > > > + > > > > > > +workfile=$SCRATCH_MNT/file > > > > > > +nfiles=8 > > > > > > +wsize=4096 > > > > > > +wcnt=1000 > > > > > > +extent_size=16m > > > > > > + > > > > > > +write_direct_file() > > > > > > +{ > > > > > > + idx=$1 > > > > > > + > > > > > > + $XFS_IO_PROG -f -c "extsize $extent_size" $workfile.$idx > > > > > > + for ((cnt=0; cnt<$wcnt; cnt++)); do > > > > > > + $XFS_IO_PROG -f -d -c "pwrite $((cnt * wsize)) $wsize" $workfile.$idx > > > > > > + done > > > > > > +} > > > > > > + > > > > > > +rm -f $workfile* > > > > > > +for ((n=0; n<$nfiles; n++)); do > > > > > > + write_direct_file $n > /dev/null 2>&1 & > > > > > > +done > > > > > > +wait > > > > > > +sync > > > > > > + > > > > > > +for ((n=0; n<$nfiles; n++)); do > > > > > > + count=$(_count_extents $workfile.$n) > > > > > > + # Acceptible extent count range is 1-10 > > > > > > + _within_tolerance "file.$n extent count" $count 2 1 8 -v > > > > > > +done > > > > > > + > > > > > > +status=0 > > > > > > +exit > > > > > > diff --git a/tests/xfs/1502.out b/tests/xfs/1502.out > > > > > > new file mode 100644 > > > > > > index 000000000..82c8760a3 > > > > > > --- /dev/null > > > > > > +++ b/tests/xfs/1502.out > > > > > > @@ -0,0 +1,9 @@ > > > > > > +QA output created by 1502 > > > > > > +file.0 extent count is in range > > > > > > +file.1 extent count is in range > > > > > > +file.2 extent count is in range > > > > > > +file.3 extent count is in range > > > > > > +file.4 extent count is in range > > > > > > +file.5 extent count is in range > > > > > > +file.6 extent count is in range > > > > > > +file.7 extent count is in range > > > > > > diff --git a/tests/xfs/1503 b/tests/xfs/1503 > > > > > > new file mode 100755 > > > > > > index 000000000..9002f87e6 > > > > > > --- /dev/null > > > > > > +++ b/tests/xfs/1503 > > > > > > @@ -0,0 +1,77 @@ > > > > > > +#! /bin/bash > > > > > > +# SPDX-License-Identifier: GPL-2.0 > > > > > > +# Copyright (c) 2019 Red Hat, Inc. All Rights Reserved. > > > > > > +# > > > > > > +# FS QA Test xfs/503 > > > > > > +# > > > > > > +# Post-EOF preallocation defeat test with O_SYNC buffered I/O that repeatedly > > > > > > +# closes and reopens the files. > > > > > > +# > > > > > > + > > > > > > +. ./common/preamble > > > > > > +_begin_fstest auto prealloc rw > > > > > > + > > > > > > +. ./common/rc > > > > > > +. ./common/filter > > > > > > + > > > > > > +_require_scratch > > > > > > + > > > > > > +_cleanup() > > > > > > +{ > > > > > > + # try to kill all background processes > > > > > > + wait > > > > > > + cd / > > > > > > + rm -r -f $tmp.* > > > > > > +} > > > > > > + > > > > > > +_scratch_mkfs > "$seqres.full" 2>&1 > > > > > > +_scratch_mount > > > > > > + > > > > > > +# Write multiple files in parallel using synchronous buffered writes that > > > > > > +# repeatedly close and reopen the fails. Aim is to interleave allocations to > > > > > > +# fragment the files. Assuming we've fixed the synchronous write defeat, we can > > > > > > +# still trigger the same issue with a open/read/close on O_RDONLY files. We > > > > > > +# should not be triggering EOF preallocation removal on files we don't have > > > > > > +# permission to write, so until this is fixed it should fragment badly. Typical > > > > > > +# problematic behaviour shows per-file extent counts of 50-350 whilst fixed > > > > > > +# behaviour typically demonstrates post-eof speculative delalloc growth in > > > > > > +# extent size (~6 extents for 50MB file). > > > > > > +# > > > > > > +# Failure is determined by golden output mismatch from _within_tolerance(). > > > > > > + > > > > > > +workfile=$SCRATCH_MNT/file > > > > > > +nfiles=32 > > > > > > +wsize=4096 > > > > > > +wcnt=1000 > > > > > > + > > > > > > +write_file() > > > > > > +{ > > > > > > + idx=$1 > > > > > > + > > > > > > + $XFS_IO_PROG -f -s -c "pwrite -b 64k 0 50m" $workfile.$idx > > > > > > +} > > > > > > + > > > > > > +read_file() > > > > > > +{ > > > > > > + idx=$1 > > > > > > + > > > > > > + for ((cnt=0; cnt<$wcnt; cnt++)); do > > > > > > + $XFS_IO_PROG -f -r -c "pread 0 28" $workfile.$idx > > > > > > + done > > > > > > +} > > > > > > + > > > > > > +rm -f $workdir/file* > > > > > > +for ((n=0; n<$((nfiles)); n++)); do > > > > > > + write_file $n > /dev/null 2>&1 & > > > > > > + read_file $n > /dev/null 2>&1 & > > > > > > +done > > > > > > +wait > > > > > > + > > > > > > +for ((n=0; n<$nfiles; n++)); do > > > > > > + count=$(_count_extents $workfile.$n) > > > > > > + # Acceptible extent count range is 1-40 > > > > > > + _within_tolerance "file.$n extent count" $count 6 5 10 -v > > > > > > +done > > > > > > + > > > > > > +status=0 > > > > > > +exit > > > > > > diff --git a/tests/xfs/1503.out b/tests/xfs/1503.out > > > > > > new file mode 100644 > > > > > > index 000000000..1780b16df > > > > > > --- /dev/null > > > > > > +++ b/tests/xfs/1503.out > > > > > > @@ -0,0 +1,33 @@ > > > > > > +QA output created by 1503 > > > > > > +file.0 extent count is in range > > > > > > +file.1 extent count is in range > > > > > > +file.2 extent count is in range > > > > > > +file.3 extent count is in range > > > > > > +file.4 extent count is in range > > > > > > +file.5 extent count is in range > > > > > > +file.6 extent count is in range > > > > > > +file.7 extent count is in range > > > > > > +file.8 extent count is in range > > > > > > +file.9 extent count is in range > > > > > > +file.10 extent count is in range > > > > > > +file.11 extent count is in range > > > > > > +file.12 extent count is in range > > > > > > +file.13 extent count is in range > > > > > > +file.14 extent count is in range > > > > > > +file.15 extent count is in range > > > > > > +file.16 extent count is in range > > > > > > +file.17 extent count is in range > > > > > > +file.18 extent count is in range > > > > > > +file.19 extent count is in range > > > > > > +file.20 extent count is in range > > > > > > +file.21 extent count is in range > > > > > > +file.22 extent count is in range > > > > > > +file.23 extent count is in range > > > > > > +file.24 extent count is in range > > > > > > +file.25 extent count is in range > > > > > > +file.26 extent count is in range > > > > > > +file.27 extent count is in range > > > > > > +file.28 extent count is in range > > > > > > +file.29 extent count is in range > > > > > > +file.30 extent count is in range > > > > > > +file.31 extent count is in range > > > > > > -- > > > > > > 2.45.2 > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > >