fstests: btrfs: add a test case to verify fsverity on compressed files
PATCH 1/1 fstests: btrfs: add a test case to verify fsverity on compressed files
There is a bug report that with the recently enabled large folios support, compressed files on btrfs can give random but reproducible warnings about file corruption.
It turns out to be a bug in the verification range, where blocks not included in the read range are passed to be verified.
Add a new regression test case for it, the workload is pretty simple:
- Create a 64M file with compressed file extents
- Enable fsverity digest on that file
- Drop page cache
- Read that file
- Verify there is no "FILE CORRUPTED!" warning
--- tests/btrfs/359 | 45 +++++++++++++++++++++++++++++++++++++++++++++ tests/btrfs/359.out | 6 ++++++ 2 files changed, 51 insertions(+) create mode 100755 tests/btrfs/359 create mode 100644 tests/btrfs/359.out
tests/btrfs/359 new file +45 -0
@@ -0,0 +1,45 @@ +#! /bin/bash +# SPDX-License-Identifier: GPL-2.0 +# Copyright (c) 2026 SUSE S.A. All Rights Reserved. +# +# FS QA Test 359 +# +# Test if btrfs handles fsverity read correctly without corruption warning +# +. ./common/preamble +_begin_fstest auto quick verity compress + +_fixed_by_kernel_commit xxxxxxxxxxxx \ + "btrfs: fix the fsverity callback where unexpected range is verified" + +. ./common/filter +. ./common/verity + +_require_scratch_verity + +_scratch_mkfs_verity >> $seqres.full + +# We need a read range that doesn't cover the full folio, the easiest way +# to create such read is compression, which limits the extent size to 128K, +# much smaller than the supported large folio size (256K or 2M for 4K page size). +_scratch_mount -o compress + +# Create a large enough file that can contain multiple compressed file extents. +_pwrite_byte "0xcd" 0 64m "$SCRATCH_MNT/foobar" | _filter_xfs_io + +# Enable verity on that file +_fsv_enable "$SCRATCH_MNT/foobar" + +# Cycle mount to drop page cache +_scratch_cycle_mount + +# Read the file, unpatched kernel will race and trigger fs-verity warnings. +# But the read itself should still be fine, as the failed part will be retried. +_hexdump "$SCRATCH_MNT/foobar" + +# Check for any kernel warnings from fs-verity on file corruption. +if _check_dmesg_for "FILE CORRUPTED!"; then + echo "fs-verity detected corruption" +fi + +_exit 0
tests/btrfs/359.out new file +6 -0
@@ -0,0 +1,6 @@ +QA output created by 359 +wrote 67108864/67108864 bytes at offset 0 +XXX Bytes, X ops; XX:XX:XX.X (XXX YYY/sec and XXX ops/sec) +000000 cd cd cd cd cd cd cd cd cd cd cd cd cd cd cd cd >................< +* +4000000
Signature · 1 line
2.56.0
Quoted 37 lines · Qu Wenruo
There is a bug report that with the recently enabled large folios support, compressed files on btrfs can give random but reproducible warnings about file corruption.
It turns out to be a bug in the verification range, where blocks not included in the read range are passed to be verified.
Add a new regression test case for it, the workload is pretty simple:
- Create a 64M file with compressed file extents
- Enable fsverity digest on that file
- Drop page cache
- Read that file
- Verify there is no "FILE CORRUPTED!" warningSigned-off-by Qu Wenruo <[email protected]>--- tests/btrfs/359 | 45 +++++++++++++++++++++++++++++++++++++++++++++ tests/btrfs/359.out | 6 ++++++ 2 files changed, 51 insertions(+) create mode 100755 tests/btrfs/359 create mode 100644 tests/btrfs/359.outtests/btrfs/359 new file +8
-0@@ -0,0 +1,45 @@ +#! /bin/bash +# SPDX-License-Identifier: GPL-2.0 +# Copyright (c) 2026 SUSE S.A. All Rights Reserved. +# +# FS QA Test 359 +# +# Test if btrfs handles fsverity read correctly without corruption warning +#
tests/btrfs/359 line 18
Quoted 10 lines
tests/btrfs/359 new file +10
-0+. ./common/preamble +_begin_fstest auto quick verity compress + +_fixed_by_kernel_commit xxxxxxxxxxxx \ + "btrfs: fix the fsverity callback where unexpected range is verified" + +. ./common/filter +. ./common/verity + +_require_scratch_verity
Missing _require_check_dmesg?
Also, sourcing common/verity adds nodatasum to the mount options. Doesn't that disable compression too?
- Eric
Quoted 42 lines · Eric Biggers
Quoted 41 lines · Qu Wenruo
There is a bug report that with the recently enabled large folios support, compressed files on btrfs can give random but reproducible warnings about file corruption.
It turns out to be a bug in the verification range, where blocks not included in the read range are passed to be verified.
Add a new regression test case for it, the workload is pretty simple:
- Create a 64M file with compressed file extents
- Enable fsverity digest on that file
- Drop page cache
- Read that file
- Verify there is no "FILE CORRUPTED!" warningSigned-off-by Qu Wenruo <[email protected]>--- tests/btrfs/359 | 45 +++++++++++++++++++++++++++++++++++++++++++++ tests/btrfs/359.out | 6 ++++++ 2 files changed, 51 insertions(+)create mode 100755 tests/btrfs/359
create mode 100644 tests/btrfs/359.outtests/btrfs/359 new file +12
-0@@ -0,0 +1,45 @@ +#! /bin/bash +# SPDX-License-Identifier: GPL-2.0 +# Copyright (c) 2026 SUSE S.A. All Rights Reserved. +# +# FS QA Test 359 +# +# Test if btrfs handles fsverity read correctly without corruption warning +# +. ./common/preamble +_begin_fstest auto quick verity compress + +_fixed_by_kernel_commit xxxxxxxxxxxx \
Quoted 8 lines
tests/btrfs/359 line 18
Quoted 6 lines
tests/btrfs/359 new file +6
-0+ "btrfs: fix the fsverity callback where unexpected range is verified" + +. ./common/filter +. ./common/verity + +_require_scratch_verityMissing
_require_check_dmesg?
Right, thanks for pointing out this.
Quoted 3 lines
Also, sourcing common/verity adds nodatasum to the mount options. Doesn't that disable compression too?
It can still be override by the extra "-o compress" option at later _scratch_mount() call.
In fact after the run, the fs still has compressed extents, e.g:
item 201 key (257 EXTENT_DATA 66977792) itemoff 5577 itemsize 53 generation 9 type 1 (regular) extent data disk byte 15724544 nr 4096 extent data offset 0 nr 131072 ram 131072 extent compression 1 (zlib) extent encryption 0
On the hand, I believe it's possible to craft a generic test case that doesn't require any btrfs specific mount option.
Using holes can still force btrfs to split the reads and trigger the warning.
I'll send out a generic test instead.
Thanks,
Qu
Quoted 2 lines
- Eric