On Wed, May 8, 2019 at 2:27 PM Qu Wenruo <wqu@xxxxxxxx> wrote:
>
> This test will test if we can still do the following operations when a
> full is full:
> - buffered write into unpopulated preallocated extent
> - clone the untouched preallocated extent
> - fsync
> - no data loss if power loss happens after above fsync
> Above operations should not fail, as they takes no extra data space.
>
> Xfs passes the test, while btrfs fails at fsync and has data loss.
> The fix for btrfs is:
> "btrfs: Flush before reflinking any extent to prevent NOCOW write falling
> back to CoW without data reservation"
>
> Signed-off-by: Qu Wenruo <wqu@xxxxxxxx>
> ---
> changelog:
> v2:
> - Change the comment and commit message to make it describe the test
> itself, not the btrfs specific part.
> - Use $XFS_IO_PROG to replace xfs_io.
> v3:
> - Move the current test result and btrfs fix to commit message
> - Also test if data is consistent after power loss and log recovery
> ---
> tests/generic/545 | 92 +++++++++++++++++++++++++++++++++++++++++++
> tests/generic/545.out | 2 +
> tests/generic/group | 1 +
> 3 files changed, 95 insertions(+)
> create mode 100755 tests/generic/545
> create mode 100644 tests/generic/545.out
>
> diff --git a/tests/generic/545 b/tests/generic/545
> new file mode 100755
> index 00000000..e8419585
> --- /dev/null
> +++ b/tests/generic/545
> @@ -0,0 +1,92 @@
> +#! /bin/bash
> +# SPDX-License-Identifier: GPL-2.0
> +# Copyright (C) 2019 SUSE Linux Products GmbH. All Rights Reserved.
> +#
> +# FS QA Test 545
> +#
> +# Test when a fs is full we can still:
> +# - Do buffered write into a unpopulated preallocated extent
> +# - Clone the untouched part of that preallocated extent
> +# - Fsync
> +# - No data loss even power loss happens after fsync
> +# All operations above should not fail.
> +#
> +seq=`basename $0`
> +seqres=$RESULT_DIR/$seq
> +echo "QA output created by $seq"
> +
> +here=`pwd`
> +tmp=/tmp/$$
> +status=1 # failure is the default!
> +trap "_cleanup; exit \$status" 0 1 2 3 15
> +
> +_cleanup()
> +{
> + _cleanup_flakey
> + cd /
> + rm -f $tmp.*
> +}
> +
> +# get standard environment, filters and checks
> +. ./common/rc
> +. ./common/filter
> +. ./common/reflink
> +. ./common/dmflakey
> +
> +# remove previous $seqres.full before test
> +rm -f $seqres.full
> +
> +# real QA test starts here
> +
> +# Modify as appropriate.
> +_supported_fs generic
> +_supported_os Linux
> +_require_scratch_reflink
> +_require_dm_target flakey
> +
> +_scratch_mkfs_sized $((512 * 1024 * 1024)) >> $seqres.full 2>&1
Missing here: _require_metadata_journaling $SCRATCH_DEV
> +_init_flakey
> +_mount_flakey
> +
> +# Create preallocated extent where we can write into
> +$XFS_IO_PROG -f -c 'falloc 8k 64m' "$SCRATCH_MNT/foobar" >> $seqres.full
> +
> +# Use up all data space, to test later write-into-preallocate behavior
> +_pwrite_byte 0x00 0 512m "$SCRATCH_MNT/padding" >> $seqres.full 2>&1
> +
> +# Sync to ensure that padding file reach disk so that at log recovery we
> +# still have no data space
> +sync
> +
> +# This should not fail
> +_pwrite_byte 0xcd 1m 16m "$SCRATCH_MNT/foobar" >> $seqres.full
> +
> +# Do reflink here, we shouldn't use extra data space, thus it should not fail
> +$XFS_IO_PROG -c "reflink ${SCRATCH_MNT}/foobar 8k 0 4k" "$SCRATCH_MNT/foobar" \
> + >> $seqres.full
> +
> +# Checkpoint before power loss
checkpoint -> checksum
> +c1foobar=$(_md5_checksum "$SCRATCH_MNT/foobar")
> +echo "md5 foobar before: $c1foobar" >> $seqres.full
> +c1padding=$(_md5_checksum "$SCRATCH_MNT/padding")
> +echo "md5 padding before: $c1padding" >> $seqres.full
What's the point of verifying the checksum of the padding file?
We sync'ed before, and we aren't modifying it after the sync.
Just leave it... Doesn't add any value, and it falls out of the test
description.
> +
> +# Fsync to check if writeback is ok
> +$XFS_IO_PROG -c 'fsync' "$SCRATCH_MNT/foobar"
> +
> +# Now emulate power loss
> +_flakey_drop_and_remount
> +
> +# Checkpoint after power loss
checkpoint -> checksum
> +c2foobar=$(_md5_checksum "$SCRATCH_MNT/foobar")
> +echo "md5 foobar after: $c2foobar" >> $seqres.full
> +c2padding=$(_md5_checksum "$SCRATCH_MNT/padding")
> +echo "md5 padding after: $c2padding" >> $seqres.full
Same here.
> +
> +test $c1foobar = $c2foobar || echo "foobar doesn't match after log recovery"
> +test $c1padding = $c2padding || echo "padding doesn't match after log recovery"
We generally verify things by making output of commands match what we
have on the golden ouput.
That is, just echo the checksum (before and after the power loss) and
have the golden output have the correct checksums.
If there are no complaints, I'm fine with it.
Thanks.
> +
> +echo "Silence is golden"
> +# success, all done
> +status=0
> +exit
> diff --git a/tests/generic/545.out b/tests/generic/545.out
> new file mode 100644
> index 00000000..920d7244
> --- /dev/null
> +++ b/tests/generic/545.out
> @@ -0,0 +1,2 @@
> +QA output created by 545
> +Silence is golden
> diff --git a/tests/generic/group b/tests/generic/group
> index 40deb4d0..84892a60 100644
> --- a/tests/generic/group
> +++ b/tests/generic/group
> @@ -547,3 +547,4 @@
> 542 auto quick clone
> 543 auto quick clone
> 544 auto quick clone
> +545 auto quick clone enospc log
> --
> 2.21.0
>
--
Filipe David Manana,
“Whether you think you can, or you think you can't — you're right.”
