btrfs: add a test case to verify the scrub error reports
PATCH 1/1 btrfs: add a test case to verify the scrub error reports
There is a regression in recent v6.4 cycle where a scrub rewrite changed how we report errors, especially repairable errors.
Before the rewrite, we report the initial errors hit, and the amount of repairable errors.
While after the rewrite, we no longer report the initial errors, but only the number of repairable errors.
This behavior change is a regression, thus needs a test case to prevent such problem from happening again.
The test case itself would:
- Create a btrfs using DUP data profile and 4K sector size
- Create a file with one 128K extent
- Corrupt the first mirror of that 128K extent
- Scrub and checks the detailed report
Both corrected errors and csum errors should be 32.
--- tests/btrfs/289 | 67 +++++++++++++++++++++++++++++++++++++++++++++ tests/btrfs/289.out | 2 ++ 2 files changed, 69 insertions(+) create mode 100755 tests/btrfs/289 create mode 100644 tests/btrfs/289.out
tests/btrfs/289 new file +67 -0
@@ -0,0 +1,67 @@ +#! /bin/bash +# SPDX-License-Identifier: GPL-2.0 +# Copyright (C) 2023 SUSE Linux Products GmbH. All Rights Reserved. +# +# FS QA Test 289 +# +# Make sure btrfs-scrub reports errors correctly for repaired sectors. +# +. ./common/preamble +_begin_fstest auto quick scrub repair + +# For filedefrag and all the filters +. ./common/filter + +# real QA test starts here + +# Modify as appropriate. +_supported_fs btrfs +_require_scratch + +_require_odirect +# Overwriting data is forbidden on a zoned block device +_require_non_zoned_device "${SCRATCH_DEV}" + +# The errors reported would be in the unit of sector, thus the number +# is dependent on the sectorsize. +_require_btrfs_support_sectorsize 4096 + +# Create a single btrfs with DUP data profile, and create one 128K file. +_scratch_mkfs -s 4k -d dup -b 1G >> $seqres.full 2>&1 +_scratch_mount +$XFS_IO_PROG -f -d -c "pwrite -S 0xaa -b 128K 0 128K" "$SCRATCH_MNT/foobar" \ + > /dev/null +sync + +logical=$(_btrfs_get_first_logical "$SCRATCH_MNT/foobar") + +physical1=$(_btrfs_get_physical ${logical} 1) +devpath1=$(_btrfs_get_device_path ${logical} 1) +_scratch_unmount + +echo " corrupt stripe #1, devpath $devpath1 physical $physical1" \ + >> $seqres.full +$XFS_IO_PROG -d -c "pwrite -S 0xf1 -b 64K $physical1 128K" $devpath1 \ + > /dev/null + +# Mount and do a scrub and compare the ouput +_scratch_mount +$BTRFS_UTIL_PROG scrub start -BR $SCRATCH_MNT >> $tmp.scrub_report 2>&1 +cat $tmp.scrub_report >> $seqres.full + +# Csum errors should be 128K/4K = 32 +csum_errors=$(grep "csum_errors" $tmp.scrub_report | awk '{print $2}') +if [ $csum_errors -ne 32 ]; then + echo "csum_errors incorrect, expect 32 has $csum_errors" +fi + +# And all errors should be repaired, thus corrected errors should also be 32. +corrected_errors=$(grep "corrected_errors" $tmp.scrub_report | awk '{print $2}') +if [ $corrected_errors -ne 32 ]; then + echo "csum_errors incorrect, expect 32 has $corrected_errors" +fi + +echo "Silence is golden" + +status=0 +exit
tests/btrfs/289.out new file +2 -0
@@ -0,0 +1,2 @@ +QA output created by 289 +Silence is golden
Signature · 1 line
2.39.0
Quoted 39 lines · Qu Wenruo
There is a regression in recent v6.4 cycle where a scrub rewrite changed how we report errors, especially repairable errors.
Before the rewrite, we report the initial errors hit, and the amount of repairable errors.
While after the rewrite, we no longer report the initial errors, but only the number of repairable errors.This behavior change is a regression, thus needs a test case to prevent such problem from happening again.
The test case itself would:
- Create a btrfs using DUP data profile and 4K sector size
- Create a file with one 128K extent
- Corrupt the first mirror of that 128K extent
- Scrub and checks the detailed report
Both corrected errors and csum errors should be 32.Signed-off-by Qu Wenruo <[email protected]>--- tests/btrfs/289 | 67 +++++++++++++++++++++++++++++++++++++++++++++ tests/btrfs/289.out | 2 ++ 2 files changed, 69 insertions(+) create mode 100755 tests/btrfs/289 create mode 100644 tests/btrfs/289.outtests/btrfs/289 new file +2
-0@@ -0,0 +1,67 @@ +#! /bin/bash +# SPDX-License-Identifier: GPL-2.0
tests/btrfs/289 line 12
Quoted 10 lines
tests/btrfs/289 new file +10
-0+# Copyright (C) 2023 SUSE Linux Products GmbH. All Rights Reserved. +# +# FS QA Test 289 +# +# Make sure btrfs-scrub reports errors correctly for repaired sectors. +# +. ./common/preamble +_begin_fstest auto quick scrub repair + +# For filedefrag and all the filters
So almost the same comment as in the other test:
This comment is a bit confusing. File defrag? The test doesn't exercise defrag. I'm not seeing the test using filters either, the test is redirecting xfs_io's stdout to /dev/null
Quoted 5 lines
tests/btrfs/289 new file +5
-0+. ./common/filter + +# real QA test starts here + +# Modify as appropriate.
tests/btrfs/289 line 27
Quoted 10 lines
tests/btrfs/289 new file +10
-0+_supported_fs btrfs +_require_scratch + +_require_odirect +# Overwriting data is forbidden on a zoned block device +_require_non_zoned_device "${SCRATCH_DEV}" + +# The errors reported would be in the unit of sector, thus the number +# is dependent on the sectorsize. +_require_btrfs_support_sectorsize 4096
So same as before, can we please get a _fixed_by_kernel_commit to identify the patch that fixes the regression?
Quoted 10 lines
tests/btrfs/289 new file +10
-0+ +# Create a single btrfs with DUP data profile, and create one 128K file. +_scratch_mkfs -s 4k -d dup -b 1G >> $seqres.full 2>&1 +_scratch_mount +$XFS_IO_PROG -f -d -c "pwrite -S 0xaa -b 128K 0 128K" "$SCRATCH_MNT/foobar" \ + > /dev/null +sync + +logical=$(_btrfs_get_first_logical "$SCRATCH_MNT/foobar") +
tests/btrfs/289 line 47
Quoted 10 lines
tests/btrfs/289 new file +10
-0+physical1=$(_btrfs_get_physical ${logical} 1) +devpath1=$(_btrfs_get_device_path ${logical} 1) +_scratch_unmount + +echo " corrupt stripe #1, devpath $devpath1 physical $physical1" \ + >> $seqres.full +$XFS_IO_PROG -d -c "pwrite -S 0xf1 -b 64K $physical1 128K" $devpath1 \ + > /dev/null + +# Mount and do a scrub and compare the ouput
ouput -> output
tests/btrfs/289 line 53
Quoted 6 lines
tests/btrfs/289 new file +6
-0+_scratch_mount +$BTRFS_UTIL_PROG scrub start -BR $SCRATCH_MNT >> $tmp.scrub_report 2>&1 +cat $tmp.scrub_report >> $seqres.full + +# Csum errors should be 128K/4K = 32 +csum_errors=$(grep "csum_errors" $tmp.scrub_report | awk '{print $2}')
Use $AWK_PROG instead.
tests/btrfs/289 line 59
Quoted 6 lines
tests/btrfs/289 new file +6
-0+if [ $csum_errors -ne 32 ]; then + echo "csum_errors incorrect, expect 32 has $csum_errors" +fi + +# And all errors should be repaired, thus corrected errors should also be 32. +corrected_errors=$(grep "corrected_errors" $tmp.scrub_report | awk '{print $2}')
Same here, $AWK_PROG instead.
Otherwise, it looks fine, thanks.
Quoted 19 lines
tests/btrfs/289 new file +8
-0+if [ $corrected_errors -ne 32 ]; then + echo "csum_errors incorrect, expect 32 has $corrected_errors" +fi + +echo "Silence is golden" + +status=0 +exittests/btrfs/289.out new file +2
-1@@ -0,0 +1,2 @@ +QA output created by 289 +Silence is golden --2.39.0