On 12/01/21 07:40, Qu Wenruo wrote:
> There is a bug that, when btrfs is beyond qgroup limit, touching a file
> could crash btrfs.
>
> Such beyond limit situation needs to be intentionally created, e.g.
> writing 1GiB file, then limit the subvolume to 512 MiB.
> As current qgroup works pretty well at preventing us from reaching the
> limit.
>
> This makes existing qgroup test cases unable to detect it.
>
> The regression is introduced by commit c53e9653605d ("btrfs: qgroup: try
> to flush qgroup space when we get -EDQUOT"), and the fix is titled
> "btrfs: qgroup: don't commit transaction when we have already
> hold a transaction handler"
By the way, the patch subject changed to
"btrfs: qgroup: don't commit transaction when we already hold the handle"
and it's already in Linus' tree, so its commit hash should be referenced.
https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=6f23277a49e68f8a9355385c846939ad0b1261e7
Thanks for refreshing it.
>
> Link: https://bugzilla.suse.com/show_bug.cgi?id=1178634
> Reviewed-by: Filipe Manana <fdmanana@xxxxxxxx>
> Signed-off-by: Qu Wenruo <wqu@xxxxxxxx>
> ---
> Changelog:
> v2:
> - Use "0/5" to replace the double "$SCRATCH_MNT" in btrfs qgroup command
> To reduce confusion.
> ---
> tests/btrfs/228 | 59 +++++++++++++++++++++++++++++++++++++++++++++
> tests/btrfs/228.out | 2 ++
> tests/btrfs/group | 1 +
> 3 files changed, 62 insertions(+)
> create mode 100755 tests/btrfs/228
> create mode 100644 tests/btrfs/228.out
>
> diff --git a/tests/btrfs/228 b/tests/btrfs/228
> new file mode 100755
> index 00000000..ecca3181
> --- /dev/null
> +++ b/tests/btrfs/228
> @@ -0,0 +1,59 @@
> +#! /bin/bash
> +# SPDX-License-Identifier: GPL-2.0
> +# Copyright (c) 2020 SUSE Linux Products GmbH. All Rights Reserved.
> +#
> +# FS QA Test 228
> +#
> +# Test if btrfs qgroup would crash if we're modifying the fs
> +# after exceeding the limit
> +#
> +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()
> +{
> + cd /
> + rm -f $tmp.*
> +}
> +
> +# get standard environment, filters and checks
> +. ./common/rc
> +. ./common/filter
> +
> +# remove previous $seqres.full before test
> +rm -f $seqres.full
> +
> +# real QA test starts here
> +
> +_supported_fs btrfs
> +
> +# Need at least 2GiB
> +_require_scratch_size $((2 * 1024 * 1024))
> +_scratch_mkfs > /dev/null 2>&1
> +_scratch_mount
> +
> +_pwrite_byte 0xcd 0 1G $SCRATCH_MNT/file >> $seqres.full
> +# Make sure the data reach disk so later qgroup scan can see it
> +sync
> +
> +$BTRFS_UTIL_PROG quota enable $SCRATCH_MNT
> +$BTRFS_UTIL_PROG quota rescan -w $SCRATCH_MNT >> $seqres.full
> +
> +# Set the limit to just 512MiB, which is way below the existing usage
> +$BTRFS_UTIL_PROG qgroup limit 512M 0/5 $SCRATCH_MNT
> +
> +# Touch above file, if kernel not patched, it will trigger an ASSERT()
> +#
> +# Even for patched kernel, we will still get EDQUOT error, but that
> +# is expected behavior.
> +touch $SCRATCH_MNT/file 2>&1 | _filter_scratch
> +
> +# success, all done
> +status=0
> +exit
> diff --git a/tests/btrfs/228.out b/tests/btrfs/228.out
> new file mode 100644
> index 00000000..9c250148
> --- /dev/null
> +++ b/tests/btrfs/228.out
> @@ -0,0 +1,2 @@
> +QA output created by 228
> +touch: setting times of 'SCRATCH_MNT/file': Disk quota exceeded
> diff --git a/tests/btrfs/group b/tests/btrfs/group
> index 1868208e..9b0dc5ca 100644
> --- a/tests/btrfs/group
> +++ b/tests/btrfs/group
> @@ -230,3 +230,4 @@
> 225 auto quick volume seed
> 226 auto quick rw snapshot clone prealloc punch
> 227 auto quick send
> +228 auto quick qgroup limit
>
