[PATCH v2] fstests: btrfs: add a regression test for incorrect inode incompressible flag

[PATCH v2] fstests: btrfs: add a regression test for incorrect inode incompressible flag

[Qu Wenruo]

[BUG]
Since kernel commit 59615e2c1f63 ("btrfs: reject single block sized
compression early"), a single block write at file offset 0, which can
not be inlined due the inode size, will mark the inode incompressible.

[REGRESSION TEST]
The new regression test will do:

- Create and mount the fs with compression,max_inline=2k

- Do the following operations:
  * Truncate the inode to 2 * blocksize
    This will rule out any future inlined writes.

  * Buffered write [0, 2K)
    Which will not be inlined.

  * Sync
    For affected kernels, this will set the inode with NOCOMPRESS
    and reject all future compression on that inode.

  * Buffered write [1M, 2M)
    For affected kernels, the range will not be compressed due
    to the NOCOMPRESS inode flag.

- Unmount the fs

- Make sure that:
  * The inode has no NOCOMPRESS flag
  * File extent at file offset 1M is being compressed

Reviewed-by: Filipe Manana <fdmanana@suse.com>
Signed-off-by: Qu Wenruo <wqu@suse.com>
---
Changelog:
v2:
- Remove an unnessary sentence
  Which is confusing because I missed the "and" to connect the two
  sentences, and it's not needed after the first paragraph anyway.

- Use full "btrfs inspect-internal" group name instead

- Add missing punctures
---
 tests/btrfs/343     | 47 +++++++++++++++++++++++++++++++++++++++++++++
 tests/btrfs/343.out |  2 ++
 2 files changed, 49 insertions(+)
 create mode 100755 tests/btrfs/343
 create mode 100644 tests/btrfs/343.out

diff --git a/tests/btrfs/343 b/tests/btrfs/343
new file mode 100755
index 00000000..78079eff
--- /dev/null
+++ b/tests/btrfs/343
@@ -0,0 +1,47 @@
+#! /bin/bash
+# SPDX-License-Identifier: GPL-2.0
+# Copyright (c) 2026 SUSE S.A.  All Rights Reserved.
+#
+# FS QA Test 343
+#
+# A regression test to make sure a single-block write at file offset 0 won't
+# incorrectly mark the inode incompressible.
+#
+. ./common/preamble
+_begin_fstest auto quick compress
+
+_require_scratch
+
+_fixed_by_kernel_commit xxxxxxxxxxxx \
+	"btrfs: fix the inline compressed extent check in inode_need_compress()"
+
+_scratch_mkfs >>$seqres.full 2>&1
+_scratch_mount "-o compress,max_inline=2048"
+
+blocksize=$(_get_block_size $SCRATCH_MNT)
+
+# Create a sparse file which is 2 * blocksize, then try a small write at
+# file offset 0 which should not be inlined.
+# Sync so that [0, 2K) range is written, then write a larger range which
+# should be able to be compressed.
+$XFS_IO_PROG -f -c "truncate $((2 * $blocksize))" -c "pwrite 0 2k" -c sync \
+		-c "pwrite 1m 1m" $SCRATCH_MNT/foobar >> $seqres.full
+ino=$(stat -c "%i" $SCRATCH_MNT/foobar)
+_scratch_unmount
+
+# Dump the fstree into seqres.full for debug
+$BTRFS_UTIL_PROG inspect dump-tree -t 5 $SCRATCH_DEV >> $seqres.full
+
+# Check the NOCOMPRESS flag of the inode.
+$BTRFS_UTIL_PROG inspect dump-tree -t 5 $SCRATCH_DEV |\
+grep -A 4 -e "item .* key ($ino INODE_ITEM 0)" | grep -q NOCOMPRESS
+[ $? -eq 0 ] && echo "inode $ino has NOCOMPRESS flag"
+
+# Check the file extent at fileoffset 1m
+$BTRFS_UTIL_PROG inspect dump-tree -t 5 $SCRATCH_DEV |\
+grep -A 4 -e "item .* key ($ino EXTENT_DATA 1048576)" | grep -q "compression 0"
+[ $? -eq 0 ] && echo "inode $ino file offset 1M is not compressed"
+
+echo "Silence is golden"
+# success, all done
+_exit 0
diff --git a/tests/btrfs/343.out b/tests/btrfs/343.out
new file mode 100644
index 00000000..2eb30e4f
--- /dev/null
+++ b/tests/btrfs/343.out
@@ -0,0 +1,2 @@
+QA output created by 343
+Silence is golden
-- 
2.51.2

Re: [PATCH v2] fstests: btrfs: add a regression test for incorrect inode incompressible flag

[Filipe Manana]

On Mon, Feb 9, 2026 at 9:50 AM Qu Wenruo <wqu@suse.com> wrote:
>
> [BUG]
> Since kernel commit 59615e2c1f63 ("btrfs: reject single block sized
> compression early"), a single block write at file offset 0, which can
> not be inlined due the inode size, will mark the inode incompressible.
>
> [REGRESSION TEST]
> The new regression test will do:
>
> - Create and mount the fs with compression,max_inline=2k
>
> - Do the following operations:
>   * Truncate the inode to 2 * blocksize
>     This will rule out any future inlined writes.
>
>   * Buffered write [0, 2K)
>     Which will not be inlined.
>
>   * Sync
>     For affected kernels, this will set the inode with NOCOMPRESS
>     and reject all future compression on that inode.
>
>   * Buffered write [1M, 2M)
>     For affected kernels, the range will not be compressed due
>     to the NOCOMPRESS inode flag.
>
> - Unmount the fs
>
> - Make sure that:
>   * The inode has no NOCOMPRESS flag
>   * File extent at file offset 1M is being compressed
>
> Reviewed-by: Filipe Manana <fdmanana@suse.com>
> Signed-off-by: Qu Wenruo <wqu@suse.com>
> ---
> Changelog:
> v2:
> - Remove an unnessary sentence
>   Which is confusing because I missed the "and" to connect the two
>   sentences, and it's not needed after the first paragraph anyway.
>
> - Use full "btrfs inspect-internal" group name instead
>
> - Add missing punctures
> ---
>  tests/btrfs/343     | 47 +++++++++++++++++++++++++++++++++++++++++++++
>  tests/btrfs/343.out |  2 ++
>  2 files changed, 49 insertions(+)
>  create mode 100755 tests/btrfs/343
>  create mode 100644 tests/btrfs/343.out
>
> diff --git a/tests/btrfs/343 b/tests/btrfs/343
> new file mode 100755
> index 00000000..78079eff
> --- /dev/null
> +++ b/tests/btrfs/343
> @@ -0,0 +1,47 @@
> +#! /bin/bash
> +# SPDX-License-Identifier: GPL-2.0
> +# Copyright (c) 2026 SUSE S.A.  All Rights Reserved.
> +#
> +# FS QA Test 343
> +#
> +# A regression test to make sure a single-block write at file offset 0 won't
> +# incorrectly mark the inode incompressible.
> +#
> +. ./common/preamble
> +_begin_fstest auto quick compress
> +
> +_require_scratch

Still missing a:

_require_btrfs_command inspect-internal dump-tree

As noted in the review of the first version.
Otherwise it looks good, thanks.

> +
> +_fixed_by_kernel_commit xxxxxxxxxxxx \
> +       "btrfs: fix the inline compressed extent check in inode_need_compress()"
> +
> +_scratch_mkfs >>$seqres.full 2>&1
> +_scratch_mount "-o compress,max_inline=2048"
> +
> +blocksize=$(_get_block_size $SCRATCH_MNT)
> +
> +# Create a sparse file which is 2 * blocksize, then try a small write at
> +# file offset 0 which should not be inlined.
> +# Sync so that [0, 2K) range is written, then write a larger range which
> +# should be able to be compressed.
> +$XFS_IO_PROG -f -c "truncate $((2 * $blocksize))" -c "pwrite 0 2k" -c sync \
> +               -c "pwrite 1m 1m" $SCRATCH_MNT/foobar >> $seqres.full
> +ino=$(stat -c "%i" $SCRATCH_MNT/foobar)
> +_scratch_unmount
> +
> +# Dump the fstree into seqres.full for debug
> +$BTRFS_UTIL_PROG inspect dump-tree -t 5 $SCRATCH_DEV >> $seqres.full
> +
> +# Check the NOCOMPRESS flag of the inode.
> +$BTRFS_UTIL_PROG inspect dump-tree -t 5 $SCRATCH_DEV |\
> +grep -A 4 -e "item .* key ($ino INODE_ITEM 0)" | grep -q NOCOMPRESS
> +[ $? -eq 0 ] && echo "inode $ino has NOCOMPRESS flag"
> +
> +# Check the file extent at fileoffset 1m
> +$BTRFS_UTIL_PROG inspect dump-tree -t 5 $SCRATCH_DEV |\
> +grep -A 4 -e "item .* key ($ino EXTENT_DATA 1048576)" | grep -q "compression 0"
> +[ $? -eq 0 ] && echo "inode $ino file offset 1M is not compressed"
> +
> +echo "Silence is golden"
> +# success, all done
> +_exit 0
> diff --git a/tests/btrfs/343.out b/tests/btrfs/343.out
> new file mode 100644
> index 00000000..2eb30e4f
> --- /dev/null
> +++ b/tests/btrfs/343.out
> @@ -0,0 +1,2 @@
> +QA output created by 343
> +Silence is golden
> --
> 2.51.2
>
>

Re: [PATCH v2] fstests: btrfs: add a regression test for incorrect inode incompressible flag

[Qu Wenruo]

在 2026/2/9 20:24, Filipe Manana 写道:
> On Mon, Feb 9, 2026 at 9:50 AM Qu Wenruo <wqu@suse.com> wrote:
>>
>> [BUG]
>> Since kernel commit 59615e2c1f63 ("btrfs: reject single block sized
>> compression early"), a single block write at file offset 0, which can
>> not be inlined due the inode size, will mark the inode incompressible.
>>
>> [REGRESSION TEST]
>> The new regression test will do:
>>
>> - Create and mount the fs with compression,max_inline=2k
>>
>> - Do the following operations:
>>    * Truncate the inode to 2 * blocksize
>>      This will rule out any future inlined writes.
>>
>>    * Buffered write [0, 2K)
>>      Which will not be inlined.
>>
>>    * Sync
>>      For affected kernels, this will set the inode with NOCOMPRESS
>>      and reject all future compression on that inode.
>>
>>    * Buffered write [1M, 2M)
>>      For affected kernels, the range will not be compressed due
>>      to the NOCOMPRESS inode flag.
>>
>> - Unmount the fs
>>
>> - Make sure that:
>>    * The inode has no NOCOMPRESS flag
>>    * File extent at file offset 1M is being compressed
>>
>> Reviewed-by: Filipe Manana <fdmanana@suse.com>
>> Signed-off-by: Qu Wenruo <wqu@suse.com>
>> ---
>> Changelog:
>> v2:
>> - Remove an unnessary sentence
>>    Which is confusing because I missed the "and" to connect the two
>>    sentences, and it's not needed after the first paragraph anyway.
>>
>> - Use full "btrfs inspect-internal" group name instead
>>
>> - Add missing punctures
>> ---
>>   tests/btrfs/343     | 47 +++++++++++++++++++++++++++++++++++++++++++++
>>   tests/btrfs/343.out |  2 ++
>>   2 files changed, 49 insertions(+)
>>   create mode 100755 tests/btrfs/343
>>   create mode 100644 tests/btrfs/343.out
>>
>> diff --git a/tests/btrfs/343 b/tests/btrfs/343
>> new file mode 100755
>> index 00000000..78079eff
>> --- /dev/null
>> +++ b/tests/btrfs/343
>> @@ -0,0 +1,47 @@
>> +#! /bin/bash
>> +# SPDX-License-Identifier: GPL-2.0
>> +# Copyright (c) 2026 SUSE S.A.  All Rights Reserved.
>> +#
>> +# FS QA Test 343
>> +#
>> +# A regression test to make sure a single-block write at file offset 0 won't
>> +# incorrectly mark the inode incompressible.
>> +#
>> +. ./common/preamble
>> +_begin_fstest auto quick compress
>> +
>> +_require_scratch
> 
> Still missing a:
> 
> _require_btrfs_command inspect-internal dump-tree
> 
> As noted in the review of the first version.
> Otherwise it looks good, thanks.

Facepalm, I forgot to commit...