* [PATCHSET 0/1] fstests: reload entire iunlink lists
@ 2023-09-25 21:42 Darrick J. Wong
2023-09-25 21:43 ` [PATCH 1/1] xfs: test unlinked inode list repair on demand Darrick J. Wong
0 siblings, 1 reply; 8+ messages in thread
From: Darrick J. Wong @ 2023-09-25 21:42 UTC (permalink / raw)
To: djwong, zlang; +Cc: linux-xfs, fstests, guan, david
Hi all,
This is the second part of correcting XFS to reload the incore unlinked
inode list from the ondisk contents. Whereas part one tackled failures
from regular filesystem calls, this part takes on the problem of needing
to reload the entire incore unlinked inode list on account of somebody
loading an inode that's in the /middle/ of an unlinked list. This
happens during quotacheck, bulkstat, or even opening a file by handle.
In this case we don't know the length of the list that we're reloading,
so we don't want to create a new unbounded memory load while holding
resources locked. Instead, we'll target UNTRUSTED iget calls to reload
the entire bucket.
Note that this changes the definition of the incore unlinked inode list
slightly -- i_prev_unlinked == 0 now means "not on the incore list".
If you're going to start using this code, I strongly recommend pulling
from my git trees, which are linked below.
This has been running on the djcloud for months with no problems. Enjoy!
Comments and questions are, as always, welcome.
--D
kernel git tree:
https://git.kernel.org/cgit/linux/kernel/git/djwong/xfs-linux.git/log/?h=fix-iunlink-list
fstests git tree:
https://git.kernel.org/cgit/linux/kernel/git/djwong/xfstests-dev.git/log/?h=fix-iunlink-list
---
common/rc | 4 +
tests/xfs/1872 | 113 +++++++++++++++++++++++++++
tests/xfs/1872.out | 5 +
tests/xfs/1873 | 217 ++++++++++++++++++++++++++++++++++++++++++++++++++++
tests/xfs/1873.out | 6 +
5 files changed, 344 insertions(+), 1 deletion(-)
create mode 100755 tests/xfs/1872
create mode 100644 tests/xfs/1872.out
create mode 100755 tests/xfs/1873
create mode 100644 tests/xfs/1873.out
^ permalink raw reply [flat|nested] 8+ messages in thread
* [PATCH 1/1] xfs: test unlinked inode list repair on demand
2023-09-25 21:42 [PATCHSET 0/1] fstests: reload entire iunlink lists Darrick J. Wong
@ 2023-09-25 21:43 ` Darrick J. Wong
2023-09-25 22:02 ` Darrick J. Wong
0 siblings, 1 reply; 8+ messages in thread
From: Darrick J. Wong @ 2023-09-25 21:43 UTC (permalink / raw)
To: djwong, zlang; +Cc: linux-xfs, fstests, guan, david
From: Darrick J. Wong <djwong@kernel.org>
Create a test to exercise recovery of unlinked inodes on a clean
filesystem. This was definitely possible on old kernels that on an ro
mount would clean the log without processing the iunlink list.
Signed-off-by: Darrick J. Wong <djwong@kernel.org>
---
common/rc | 4 +
tests/xfs/1872 | 113 +++++++++++++++++++++++++++
tests/xfs/1872.out | 5 +
tests/xfs/1873 | 217 ++++++++++++++++++++++++++++++++++++++++++++++++++++
tests/xfs/1873.out | 6 +
5 files changed, 344 insertions(+), 1 deletion(-)
create mode 100755 tests/xfs/1872
create mode 100644 tests/xfs/1872.out
create mode 100755 tests/xfs/1873
create mode 100644 tests/xfs/1873.out
diff --git a/common/rc b/common/rc
index 76a7e77403..28d922382d 100644
--- a/common/rc
+++ b/common/rc
@@ -2668,9 +2668,11 @@ _require_xfs_io_command()
param_checked="$pwrite_opts $param"
;;
"scrub"|"repair")
- testio=`$XFS_IO_PROG -x -c "$command probe" $TEST_DIR 2>&1`
+ test -z "$param" && param="probe"
+ testio=`$XFS_IO_PROG -x -c "$command $param" $TEST_DIR 2>&1`
echo $testio | grep -q "Inappropriate ioctl" && \
_notrun "xfs_io $command support is missing"
+ param_checked="$param"
;;
"startupdate"|"commitupdate"|"cancelupdate")
$XFS_IO_PROG -f -c 'pwrite -S 0x58 0 128k -b 128k' $testfile > /dev/null
diff --git a/tests/xfs/1872 b/tests/xfs/1872
new file mode 100755
index 0000000000..3720a3d184
--- /dev/null
+++ b/tests/xfs/1872
@@ -0,0 +1,113 @@
+#! /bin/bash
+# SPDX-License-Identifier: GPL-2.0
+# Copyright (c) 2023 Oracle. All Rights Reserved.
+#
+# FS QA Test No. 1872
+#
+# Test using runtime code to fix unlinked inodes on a clean filesystem that
+# never got cleaned up.
+#
+. ./common/preamble
+_begin_fstest auto quick unlink
+
+# Import common functions.
+source ./common/filter
+source ./common/fuzzy
+source ./common/quota
+
+# real QA test starts here
+
+# Modify as appropriate.
+_supported_fs generic
+_require_xfs_db_command iunlink
+_require_scratch_nocheck # we'll run repair ourselves
+
+# From the AGI definition
+XFS_AGI_UNLINKED_BUCKETS=64
+
+# Try to make each iunlink bucket have this many inodes in it.
+IUNLINK_BUCKETLEN=5
+
+# XXX Forcibly disable quota since quotacheck will break this test
+orig_mount_options="$MOUNT_OPTIONS"
+_qmount_option 'noquota'
+
+format_scratch() {
+ _scratch_mkfs -d agcount=1 | _filter_mkfs 2> "${tmp}.mkfs" >> $seqres.full
+ source "${tmp}.mkfs"
+ test "${agcount}" -eq 1 || _notrun "test requires 1 AG for error injection"
+
+ local nr_iunlinks="$((IUNLINK_BUCKETLEN * XFS_AGI_UNLINKED_BUCKETS))"
+ readarray -t BADINODES < <(_scratch_xfs_db -x -c "iunlink -n $nr_iunlinks" | awk '{print $4}')
+}
+
+__repair_check_scratch() {
+ _scratch_xfs_repair -o force_geometry -n 2>&1 | \
+ tee -a $seqres.full | \
+ grep -E '(disconnected inode.*would move|next_unlinked in inode|unlinked bucket.*is.*in ag)'
+ return "${PIPESTATUS[0]}"
+}
+
+exercise_scratch() {
+ # Create a bunch of files...
+ declare -A inums
+ for ((i = 0; i < (XFS_AGI_UNLINKED_BUCKETS * 2); i++)); do
+ touch "${SCRATCH_MNT}/${i}" || break
+ inums["${i}"]="$(stat -c %i "${SCRATCH_MNT}/${i}")"
+ done
+
+ # ...then delete them to exercise the unlinked buckets
+ for ((i = 0; i < (XFS_AGI_UNLINKED_BUCKETS * 2); i++)); do
+ if ! rm -f "${SCRATCH_MNT}/${i}"; then
+ echo "rm failed on inum ${inums[$i]}"
+ break
+ fi
+ done
+}
+
+# Offline repair should not find anything
+final_check_scratch() {
+ __repair_check_scratch
+ res=$?
+ if [ $res -eq 2 ]; then
+ echo "scratch fs went offline?"
+ _scratch_mount
+ _scratch_unmount
+ __repair_check_scratch
+ fi
+ test "$res" -ne 0 && echo "repair returned $res?"
+}
+
+echo "+ Part 0: See if runtime can recover the unlinked list" | tee -a $seqres.full
+format_scratch
+_kernlog "part 0"
+_scratch_mount
+exercise_scratch
+_scratch_unmount
+final_check_scratch
+
+echo "+ Part 1: See if bulkstat can recover the unlinked list" | tee -a $seqres.full
+format_scratch
+_kernlog "part 1"
+_scratch_mount
+$XFS_IO_PROG -c 'bulkstat' $SCRATCH_MNT > /dev/null
+exercise_scratch
+_scratch_unmount
+final_check_scratch
+
+echo "+ Part 2: See if quotacheck can recover the unlinked list" | tee -a $seqres.full
+if [ -f /proc/fs/xfs/xqmstat ]; then
+ MOUNT_OPTIONS="$orig_mount_options"
+ _qmount_option 'quota'
+ format_scratch
+ _kernlog "part 2"
+ _scratch_mount
+ exercise_scratch
+ _scratch_unmount
+ final_check_scratch
+fi
+
+# success, all done
+echo Silence is golden
+status=0
+exit
diff --git a/tests/xfs/1872.out b/tests/xfs/1872.out
new file mode 100644
index 0000000000..248f0e2416
--- /dev/null
+++ b/tests/xfs/1872.out
@@ -0,0 +1,5 @@
+QA output created by 1872
++ Part 0: See if runtime can recover the unlinked list
++ Part 1: See if bulkstat can recover the unlinked list
++ Part 2: See if quotacheck can recover the unlinked list
+Silence is golden
diff --git a/tests/xfs/1873 b/tests/xfs/1873
new file mode 100755
index 0000000000..4712dee7ab
--- /dev/null
+++ b/tests/xfs/1873
@@ -0,0 +1,217 @@
+#! /bin/bash
+# SPDX-License-Identifier: GPL-2.0
+# Copyright (c) 2023 Oracle. All Rights Reserved.
+#
+# FS QA Test No. 1873
+#
+# Functional test of using online repair to fix unlinked inodes on a clean
+# filesystem that never got cleaned up.
+#
+. ./common/preamble
+_begin_fstest auto online_repair
+
+# Import common functions.
+source ./common/filter
+source ./common/fuzzy
+source ./common/quota
+
+# real QA test starts here
+
+# Modify as appropriate.
+_supported_fs generic
+_require_xfs_db_command iunlink
+# The iunlink bucket repair code wasn't added to the AGI repair code
+# until after the directory repair code was merged
+_require_xfs_io_command repair -R directory
+_require_scratch_nocheck # repair doesn't like single-AG fs
+
+# From the AGI definition
+XFS_AGI_UNLINKED_BUCKETS=64
+
+# Try to make each iunlink bucket have this many inodes in it.
+IUNLINK_BUCKETLEN=5
+
+# XXX Forcibly disable quota since quotacheck will break this test
+_qmount_option 'noquota'
+
+format_scratch() {
+ _scratch_mkfs -d agcount=1 | _filter_mkfs 2> "${tmp}.mkfs" >> $seqres.full
+ source "${tmp}.mkfs"
+ test "${agcount}" -eq 1 || _notrun "test requires 1 AG for error injection"
+
+ local nr_iunlinks="$((IUNLINK_BUCKETLEN * XFS_AGI_UNLINKED_BUCKETS))"
+ readarray -t BADINODES < <(_scratch_xfs_db -x -c "iunlink -n $nr_iunlinks" | awk '{print $4}')
+}
+
+__repair_check_scratch() {
+ _scratch_xfs_repair -o force_geometry -n 2>&1 | \
+ tee -a $seqres.full | \
+ grep -E '(disconnected inode.*would move|next_unlinked in inode|unlinked bucket.*is.*in ag)'
+ return "${PIPESTATUS[0]}"
+}
+
+corrupt_scratch() {
+ # How far into the iunlink bucket chain do we target inodes for corruption?
+ # 1 = target the inode pointed to by the AGI
+ # 3 = middle of bucket list
+ # 5 = last element in bucket
+ local corruption_bucket_depth="$1"
+ if ((corruption_bucket_depth < 1 || corruption_bucket_depth > IUNLINK_BUCKETLEN)); then
+ echo "${corruption_bucket_depth}: Value must be between 1 and ${IUNLINK_BUCKETLEN}."
+ return 1
+ fi
+
+ # Index of the inode numbers within BADINODES
+ local bad_ino1_idx=$(( (IUNLINK_BUCKETLEN - corruption_bucket_depth) * XFS_AGI_UNLINKED_BUCKETS))
+ local bad_ino2_idx=$((bad_ino1_idx + 1))
+
+ # Inode numbers to target
+ local bad_ino1="${BADINODES[bad_ino1_idx]}"
+ local bad_ino2="${BADINODES[bad_ino2_idx]}"
+ printf "bad: 0x%x 0x%x\n" "${bad_ino1}" "${bad_ino2}" | _tee_kernlog >> $seqres.full
+
+ # Bucket within AGI 0's iunlinked array.
+ local ino1_bucket="$((bad_ino1 % XFS_AGI_UNLINKED_BUCKETS))"
+ local ino2_bucket="$((bad_ino2 % XFS_AGI_UNLINKED_BUCKETS))"
+
+ # The first bad inode stays on the unlinked list but gets a nonzero
+ # nlink; the second bad inode is removed from the unlinked list but
+ # keeps its zero nlink
+ _scratch_xfs_db -x \
+ -c "inode ${bad_ino1}" -c "write -d core.nlinkv2 5555" \
+ -c "agi 0" -c "fuzz -d unlinked[${ino2_bucket}] ones" -c "print unlinked" >> $seqres.full
+
+ local iwatch=()
+ local idx
+
+ # Make a list of the adjacent iunlink bucket inodes for the first inode
+ # that we targeted.
+ if [ "${corruption_bucket_depth}" -gt 1 ]; then
+ # Previous ino in bucket
+ idx=$(( (IUNLINK_BUCKETLEN - corruption_bucket_depth + 1) * XFS_AGI_UNLINKED_BUCKETS))
+ iwatch+=("${BADINODES[idx]}")
+ fi
+ iwatch+=("${bad_ino1}")
+ if [ "$((corruption_bucket_depth + 1))" -lt "${IUNLINK_BUCKETLEN}" ]; then
+ # Next ino in bucket
+ idx=$(( (IUNLINK_BUCKETLEN - corruption_bucket_depth - 1) * XFS_AGI_UNLINKED_BUCKETS))
+ iwatch+=("${BADINODES[idx]}")
+ fi
+
+ # Make a list of the adjacent iunlink bucket inodes for the second
+ # inode that we targeted.
+ if [ "${corruption_bucket_depth}" -gt 1 ]; then
+ # Previous ino in bucket
+ idx=$(( (IUNLINK_BUCKETLEN - corruption_bucket_depth + 1) * XFS_AGI_UNLINKED_BUCKETS))
+ iwatch+=("${BADINODES[idx + 1]}")
+ fi
+ iwatch+=("${bad_ino2}")
+ if [ "$((corruption_bucket_depth + 1))" -lt "${IUNLINK_BUCKETLEN}" ]; then
+ # Next ino in bucket
+ idx=$(( (IUNLINK_BUCKETLEN - corruption_bucket_depth - 1) * XFS_AGI_UNLINKED_BUCKETS))
+ iwatch+=("${BADINODES[idx + 1]}")
+ fi
+
+ # Construct a grep string for tracepoints.
+ GREP_STR="(xrep_attempt|xrep_done|bucket ${ino1_bucket} |bucket ${ino2_bucket} |bucket ${fuzz_bucket} "
+ GREP_STR="(xrep_attempt|xrep_done|bucket ${ino1_bucket} |bucket ${ino2_bucket} "
+ for ino in "${iwatch[@]}"; do
+ f="$(printf "|ino 0x%x" "${ino}")"
+ GREP_STR="${GREP_STR}${f}"
+ done
+ GREP_STR="${GREP_STR})"
+ echo "grep -E \"${GREP_STR}\"" >> $seqres.full
+
+ # Dump everything we did to to the full file.
+ local db_dump=(-c 'agi 0' -c 'print unlinked')
+ db_dump+=(-c 'addr root' -c 'print')
+ test "${ino1_bucket}" -gt 0 && \
+ db_dump+=(-c "dump_iunlinked -a 0 -b $((ino1_bucket - 1))")
+ db_dump+=(-c "dump_iunlinked -a 0 -b ${ino1_bucket}")
+ db_dump+=(-c "dump_iunlinked -a 0 -b ${ino2_bucket}")
+ test "${ino2_bucket}" -lt 63 && \
+ db_dump+=(-c "dump_iunlinked -a 0 -b $((ino2_bucket + 1))")
+ db_dump+=(-c "inode $bad_ino1" -c 'print core.nlinkv2 v3.inumber next_unlinked')
+ db_dump+=(-c "inode $bad_ino2" -c 'print core.nlinkv2 v3.inumber next_unlinked')
+ _scratch_xfs_db "${db_dump[@]}" >> $seqres.full
+
+ # Test run of repair to make sure we find disconnected inodes
+ __repair_check_scratch | \
+ sed -e 's/disconnected inode \([0-9]*\)/disconnected inode XXXXXX/g' \
+ -e 's/next_unlinked in inode \([0-9]*\)/next_unlinked in inode XXXXXX/g' \
+ -e 's/unlinked bucket \([0-9]*\) is \([0-9]*\) in ag \([0-9]*\) .inode=\([0-9]*\)/unlinked bucket YY is XXXXXX in ag Z (inode=AAAAAA/g' | \
+ uniq -c >> $seqres.full
+ res=${PIPESTATUS[0]}
+ test "$res" -ne 0 || echo "repair returned $res after corruption?"
+}
+
+exercise_scratch() {
+ # Create a bunch of files...
+ declare -A inums
+ for ((i = 0; i < (XFS_AGI_UNLINKED_BUCKETS * 2); i++)); do
+ touch "${SCRATCH_MNT}/${i}" || break
+ inums["${i}"]="$(stat -c %i "${SCRATCH_MNT}/${i}")"
+ done
+
+ # ...then delete them to exercise the unlinked buckets
+ for ((i = 0; i < (XFS_AGI_UNLINKED_BUCKETS * 2); i++)); do
+ if ! rm -f "${SCRATCH_MNT}/${i}"; then
+ echo "rm failed on inum ${inums[$i]}"
+ break
+ fi
+ done
+}
+
+# Offline repair should not find anything
+final_check_scratch() {
+ __repair_check_scratch
+ res=$?
+ if [ $res -eq 2 ]; then
+ echo "scratch fs went offline?"
+ _scratch_mount
+ _scratch_unmount
+ __repair_check_scratch
+ fi
+ test "$res" -ne 0 && echo "repair returned $res?"
+}
+
+echo "+ Part 1: See if scrub can recover the unlinked list" | tee -a $seqres.full
+format_scratch
+_kernlog "no bad inodes"
+_scratch_mount
+_scratch_scrub >> $seqres.full
+exercise_scratch
+_scratch_unmount
+final_check_scratch
+
+echo "+ Part 2: Corrupt the first inode in the bucket" | tee -a $seqres.full
+format_scratch
+corrupt_scratch 1
+_scratch_mount
+_scratch_scrub >> $seqres.full
+exercise_scratch
+_scratch_unmount
+final_check_scratch
+
+echo "+ Part 3: Corrupt the middle inode in the bucket" | tee -a $seqres.full
+format_scratch
+corrupt_scratch 3
+_scratch_mount
+_scratch_scrub >> $seqres.full
+exercise_scratch
+_scratch_unmount
+final_check_scratch
+
+echo "+ Part 4: Corrupt the last inode in the bucket" | tee -a $seqres.full
+format_scratch
+corrupt_scratch 5
+_scratch_mount
+_scratch_scrub >> $seqres.full
+exercise_scratch
+_scratch_unmount
+final_check_scratch
+
+# success, all done
+echo Silence is golden
+status=0
+exit
diff --git a/tests/xfs/1873.out b/tests/xfs/1873.out
new file mode 100644
index 0000000000..0e36bd2304
--- /dev/null
+++ b/tests/xfs/1873.out
@@ -0,0 +1,6 @@
+QA output created by 1873
++ Part 1: See if scrub can recover the unlinked list
++ Part 2: Corrupt the first inode in the bucket
++ Part 3: Corrupt the middle inode in the bucket
++ Part 4: Corrupt the last inode in the bucket
+Silence is golden
^ permalink raw reply related [flat|nested] 8+ messages in thread
* Re: [PATCH 1/1] xfs: test unlinked inode list repair on demand
2023-09-25 21:43 ` [PATCH 1/1] xfs: test unlinked inode list repair on demand Darrick J. Wong
@ 2023-09-25 22:02 ` Darrick J. Wong
0 siblings, 0 replies; 8+ messages in thread
From: Darrick J. Wong @ 2023-09-25 22:02 UTC (permalink / raw)
To: zlang; +Cc: linux-xfs, fstests, guan, david
On Mon, Sep 25, 2023 at 02:43:05PM -0700, Darrick J. Wong wrote:
> From: Darrick J. Wong <djwong@kernel.org>
>
> Create a test to exercise recovery of unlinked inodes on a clean
> filesystem. This was definitely possible on old kernels that on an ro
> mount would clean the log without processing the iunlink list.
Just a note: For these tests to do anything useful, you have to patch
xfsprogs to include the following:
https://lore.kernel.org/linux-xfs/169567919111.2320475.5853139361331067059.stgit@frogsfrogsfrogs/T/#u
(This is why I'm patchbombing xfsprogs even though 6.5 isn't out yet.)
--D
>
> Signed-off-by: Darrick J. Wong <djwong@kernel.org>
> ---
> common/rc | 4 +
> tests/xfs/1872 | 113 +++++++++++++++++++++++++++
> tests/xfs/1872.out | 5 +
> tests/xfs/1873 | 217 ++++++++++++++++++++++++++++++++++++++++++++++++++++
> tests/xfs/1873.out | 6 +
> 5 files changed, 344 insertions(+), 1 deletion(-)
> create mode 100755 tests/xfs/1872
> create mode 100644 tests/xfs/1872.out
> create mode 100755 tests/xfs/1873
> create mode 100644 tests/xfs/1873.out
>
>
> diff --git a/common/rc b/common/rc
> index 76a7e77403..28d922382d 100644
> --- a/common/rc
> +++ b/common/rc
> @@ -2668,9 +2668,11 @@ _require_xfs_io_command()
> param_checked="$pwrite_opts $param"
> ;;
> "scrub"|"repair")
> - testio=`$XFS_IO_PROG -x -c "$command probe" $TEST_DIR 2>&1`
> + test -z "$param" && param="probe"
> + testio=`$XFS_IO_PROG -x -c "$command $param" $TEST_DIR 2>&1`
> echo $testio | grep -q "Inappropriate ioctl" && \
> _notrun "xfs_io $command support is missing"
> + param_checked="$param"
> ;;
> "startupdate"|"commitupdate"|"cancelupdate")
> $XFS_IO_PROG -f -c 'pwrite -S 0x58 0 128k -b 128k' $testfile > /dev/null
> diff --git a/tests/xfs/1872 b/tests/xfs/1872
> new file mode 100755
> index 0000000000..3720a3d184
> --- /dev/null
> +++ b/tests/xfs/1872
> @@ -0,0 +1,113 @@
> +#! /bin/bash
> +# SPDX-License-Identifier: GPL-2.0
> +# Copyright (c) 2023 Oracle. All Rights Reserved.
> +#
> +# FS QA Test No. 1872
> +#
> +# Test using runtime code to fix unlinked inodes on a clean filesystem that
> +# never got cleaned up.
> +#
> +. ./common/preamble
> +_begin_fstest auto quick unlink
> +
> +# Import common functions.
> +source ./common/filter
> +source ./common/fuzzy
> +source ./common/quota
> +
> +# real QA test starts here
> +
> +# Modify as appropriate.
> +_supported_fs generic
> +_require_xfs_db_command iunlink
> +_require_scratch_nocheck # we'll run repair ourselves
> +
> +# From the AGI definition
> +XFS_AGI_UNLINKED_BUCKETS=64
> +
> +# Try to make each iunlink bucket have this many inodes in it.
> +IUNLINK_BUCKETLEN=5
> +
> +# XXX Forcibly disable quota since quotacheck will break this test
> +orig_mount_options="$MOUNT_OPTIONS"
> +_qmount_option 'noquota'
> +
> +format_scratch() {
> + _scratch_mkfs -d agcount=1 | _filter_mkfs 2> "${tmp}.mkfs" >> $seqres.full
> + source "${tmp}.mkfs"
> + test "${agcount}" -eq 1 || _notrun "test requires 1 AG for error injection"
> +
> + local nr_iunlinks="$((IUNLINK_BUCKETLEN * XFS_AGI_UNLINKED_BUCKETS))"
> + readarray -t BADINODES < <(_scratch_xfs_db -x -c "iunlink -n $nr_iunlinks" | awk '{print $4}')
> +}
> +
> +__repair_check_scratch() {
> + _scratch_xfs_repair -o force_geometry -n 2>&1 | \
> + tee -a $seqres.full | \
> + grep -E '(disconnected inode.*would move|next_unlinked in inode|unlinked bucket.*is.*in ag)'
> + return "${PIPESTATUS[0]}"
> +}
> +
> +exercise_scratch() {
> + # Create a bunch of files...
> + declare -A inums
> + for ((i = 0; i < (XFS_AGI_UNLINKED_BUCKETS * 2); i++)); do
> + touch "${SCRATCH_MNT}/${i}" || break
> + inums["${i}"]="$(stat -c %i "${SCRATCH_MNT}/${i}")"
> + done
> +
> + # ...then delete them to exercise the unlinked buckets
> + for ((i = 0; i < (XFS_AGI_UNLINKED_BUCKETS * 2); i++)); do
> + if ! rm -f "${SCRATCH_MNT}/${i}"; then
> + echo "rm failed on inum ${inums[$i]}"
> + break
> + fi
> + done
> +}
> +
> +# Offline repair should not find anything
> +final_check_scratch() {
> + __repair_check_scratch
> + res=$?
> + if [ $res -eq 2 ]; then
> + echo "scratch fs went offline?"
> + _scratch_mount
> + _scratch_unmount
> + __repair_check_scratch
> + fi
> + test "$res" -ne 0 && echo "repair returned $res?"
> +}
> +
> +echo "+ Part 0: See if runtime can recover the unlinked list" | tee -a $seqres.full
> +format_scratch
> +_kernlog "part 0"
> +_scratch_mount
> +exercise_scratch
> +_scratch_unmount
> +final_check_scratch
> +
> +echo "+ Part 1: See if bulkstat can recover the unlinked list" | tee -a $seqres.full
> +format_scratch
> +_kernlog "part 1"
> +_scratch_mount
> +$XFS_IO_PROG -c 'bulkstat' $SCRATCH_MNT > /dev/null
> +exercise_scratch
> +_scratch_unmount
> +final_check_scratch
> +
> +echo "+ Part 2: See if quotacheck can recover the unlinked list" | tee -a $seqres.full
> +if [ -f /proc/fs/xfs/xqmstat ]; then
> + MOUNT_OPTIONS="$orig_mount_options"
> + _qmount_option 'quota'
> + format_scratch
> + _kernlog "part 2"
> + _scratch_mount
> + exercise_scratch
> + _scratch_unmount
> + final_check_scratch
> +fi
> +
> +# success, all done
> +echo Silence is golden
> +status=0
> +exit
> diff --git a/tests/xfs/1872.out b/tests/xfs/1872.out
> new file mode 100644
> index 0000000000..248f0e2416
> --- /dev/null
> +++ b/tests/xfs/1872.out
> @@ -0,0 +1,5 @@
> +QA output created by 1872
> ++ Part 0: See if runtime can recover the unlinked list
> ++ Part 1: See if bulkstat can recover the unlinked list
> ++ Part 2: See if quotacheck can recover the unlinked list
> +Silence is golden
> diff --git a/tests/xfs/1873 b/tests/xfs/1873
> new file mode 100755
> index 0000000000..4712dee7ab
> --- /dev/null
> +++ b/tests/xfs/1873
> @@ -0,0 +1,217 @@
> +#! /bin/bash
> +# SPDX-License-Identifier: GPL-2.0
> +# Copyright (c) 2023 Oracle. All Rights Reserved.
> +#
> +# FS QA Test No. 1873
> +#
> +# Functional test of using online repair to fix unlinked inodes on a clean
> +# filesystem that never got cleaned up.
> +#
> +. ./common/preamble
> +_begin_fstest auto online_repair
> +
> +# Import common functions.
> +source ./common/filter
> +source ./common/fuzzy
> +source ./common/quota
> +
> +# real QA test starts here
> +
> +# Modify as appropriate.
> +_supported_fs generic
> +_require_xfs_db_command iunlink
> +# The iunlink bucket repair code wasn't added to the AGI repair code
> +# until after the directory repair code was merged
> +_require_xfs_io_command repair -R directory
> +_require_scratch_nocheck # repair doesn't like single-AG fs
> +
> +# From the AGI definition
> +XFS_AGI_UNLINKED_BUCKETS=64
> +
> +# Try to make each iunlink bucket have this many inodes in it.
> +IUNLINK_BUCKETLEN=5
> +
> +# XXX Forcibly disable quota since quotacheck will break this test
> +_qmount_option 'noquota'
> +
> +format_scratch() {
> + _scratch_mkfs -d agcount=1 | _filter_mkfs 2> "${tmp}.mkfs" >> $seqres.full
> + source "${tmp}.mkfs"
> + test "${agcount}" -eq 1 || _notrun "test requires 1 AG for error injection"
> +
> + local nr_iunlinks="$((IUNLINK_BUCKETLEN * XFS_AGI_UNLINKED_BUCKETS))"
> + readarray -t BADINODES < <(_scratch_xfs_db -x -c "iunlink -n $nr_iunlinks" | awk '{print $4}')
> +}
> +
> +__repair_check_scratch() {
> + _scratch_xfs_repair -o force_geometry -n 2>&1 | \
> + tee -a $seqres.full | \
> + grep -E '(disconnected inode.*would move|next_unlinked in inode|unlinked bucket.*is.*in ag)'
> + return "${PIPESTATUS[0]}"
> +}
> +
> +corrupt_scratch() {
> + # How far into the iunlink bucket chain do we target inodes for corruption?
> + # 1 = target the inode pointed to by the AGI
> + # 3 = middle of bucket list
> + # 5 = last element in bucket
> + local corruption_bucket_depth="$1"
> + if ((corruption_bucket_depth < 1 || corruption_bucket_depth > IUNLINK_BUCKETLEN)); then
> + echo "${corruption_bucket_depth}: Value must be between 1 and ${IUNLINK_BUCKETLEN}."
> + return 1
> + fi
> +
> + # Index of the inode numbers within BADINODES
> + local bad_ino1_idx=$(( (IUNLINK_BUCKETLEN - corruption_bucket_depth) * XFS_AGI_UNLINKED_BUCKETS))
> + local bad_ino2_idx=$((bad_ino1_idx + 1))
> +
> + # Inode numbers to target
> + local bad_ino1="${BADINODES[bad_ino1_idx]}"
> + local bad_ino2="${BADINODES[bad_ino2_idx]}"
> + printf "bad: 0x%x 0x%x\n" "${bad_ino1}" "${bad_ino2}" | _tee_kernlog >> $seqres.full
> +
> + # Bucket within AGI 0's iunlinked array.
> + local ino1_bucket="$((bad_ino1 % XFS_AGI_UNLINKED_BUCKETS))"
> + local ino2_bucket="$((bad_ino2 % XFS_AGI_UNLINKED_BUCKETS))"
> +
> + # The first bad inode stays on the unlinked list but gets a nonzero
> + # nlink; the second bad inode is removed from the unlinked list but
> + # keeps its zero nlink
> + _scratch_xfs_db -x \
> + -c "inode ${bad_ino1}" -c "write -d core.nlinkv2 5555" \
> + -c "agi 0" -c "fuzz -d unlinked[${ino2_bucket}] ones" -c "print unlinked" >> $seqres.full
> +
> + local iwatch=()
> + local idx
> +
> + # Make a list of the adjacent iunlink bucket inodes for the first inode
> + # that we targeted.
> + if [ "${corruption_bucket_depth}" -gt 1 ]; then
> + # Previous ino in bucket
> + idx=$(( (IUNLINK_BUCKETLEN - corruption_bucket_depth + 1) * XFS_AGI_UNLINKED_BUCKETS))
> + iwatch+=("${BADINODES[idx]}")
> + fi
> + iwatch+=("${bad_ino1}")
> + if [ "$((corruption_bucket_depth + 1))" -lt "${IUNLINK_BUCKETLEN}" ]; then
> + # Next ino in bucket
> + idx=$(( (IUNLINK_BUCKETLEN - corruption_bucket_depth - 1) * XFS_AGI_UNLINKED_BUCKETS))
> + iwatch+=("${BADINODES[idx]}")
> + fi
> +
> + # Make a list of the adjacent iunlink bucket inodes for the second
> + # inode that we targeted.
> + if [ "${corruption_bucket_depth}" -gt 1 ]; then
> + # Previous ino in bucket
> + idx=$(( (IUNLINK_BUCKETLEN - corruption_bucket_depth + 1) * XFS_AGI_UNLINKED_BUCKETS))
> + iwatch+=("${BADINODES[idx + 1]}")
> + fi
> + iwatch+=("${bad_ino2}")
> + if [ "$((corruption_bucket_depth + 1))" -lt "${IUNLINK_BUCKETLEN}" ]; then
> + # Next ino in bucket
> + idx=$(( (IUNLINK_BUCKETLEN - corruption_bucket_depth - 1) * XFS_AGI_UNLINKED_BUCKETS))
> + iwatch+=("${BADINODES[idx + 1]}")
> + fi
> +
> + # Construct a grep string for tracepoints.
> + GREP_STR="(xrep_attempt|xrep_done|bucket ${ino1_bucket} |bucket ${ino2_bucket} |bucket ${fuzz_bucket} "
> + GREP_STR="(xrep_attempt|xrep_done|bucket ${ino1_bucket} |bucket ${ino2_bucket} "
> + for ino in "${iwatch[@]}"; do
> + f="$(printf "|ino 0x%x" "${ino}")"
> + GREP_STR="${GREP_STR}${f}"
> + done
> + GREP_STR="${GREP_STR})"
> + echo "grep -E \"${GREP_STR}\"" >> $seqres.full
> +
> + # Dump everything we did to to the full file.
> + local db_dump=(-c 'agi 0' -c 'print unlinked')
> + db_dump+=(-c 'addr root' -c 'print')
> + test "${ino1_bucket}" -gt 0 && \
> + db_dump+=(-c "dump_iunlinked -a 0 -b $((ino1_bucket - 1))")
> + db_dump+=(-c "dump_iunlinked -a 0 -b ${ino1_bucket}")
> + db_dump+=(-c "dump_iunlinked -a 0 -b ${ino2_bucket}")
> + test "${ino2_bucket}" -lt 63 && \
> + db_dump+=(-c "dump_iunlinked -a 0 -b $((ino2_bucket + 1))")
> + db_dump+=(-c "inode $bad_ino1" -c 'print core.nlinkv2 v3.inumber next_unlinked')
> + db_dump+=(-c "inode $bad_ino2" -c 'print core.nlinkv2 v3.inumber next_unlinked')
> + _scratch_xfs_db "${db_dump[@]}" >> $seqres.full
> +
> + # Test run of repair to make sure we find disconnected inodes
> + __repair_check_scratch | \
> + sed -e 's/disconnected inode \([0-9]*\)/disconnected inode XXXXXX/g' \
> + -e 's/next_unlinked in inode \([0-9]*\)/next_unlinked in inode XXXXXX/g' \
> + -e 's/unlinked bucket \([0-9]*\) is \([0-9]*\) in ag \([0-9]*\) .inode=\([0-9]*\)/unlinked bucket YY is XXXXXX in ag Z (inode=AAAAAA/g' | \
> + uniq -c >> $seqres.full
> + res=${PIPESTATUS[0]}
> + test "$res" -ne 0 || echo "repair returned $res after corruption?"
> +}
> +
> +exercise_scratch() {
> + # Create a bunch of files...
> + declare -A inums
> + for ((i = 0; i < (XFS_AGI_UNLINKED_BUCKETS * 2); i++)); do
> + touch "${SCRATCH_MNT}/${i}" || break
> + inums["${i}"]="$(stat -c %i "${SCRATCH_MNT}/${i}")"
> + done
> +
> + # ...then delete them to exercise the unlinked buckets
> + for ((i = 0; i < (XFS_AGI_UNLINKED_BUCKETS * 2); i++)); do
> + if ! rm -f "${SCRATCH_MNT}/${i}"; then
> + echo "rm failed on inum ${inums[$i]}"
> + break
> + fi
> + done
> +}
> +
> +# Offline repair should not find anything
> +final_check_scratch() {
> + __repair_check_scratch
> + res=$?
> + if [ $res -eq 2 ]; then
> + echo "scratch fs went offline?"
> + _scratch_mount
> + _scratch_unmount
> + __repair_check_scratch
> + fi
> + test "$res" -ne 0 && echo "repair returned $res?"
> +}
> +
> +echo "+ Part 1: See if scrub can recover the unlinked list" | tee -a $seqres.full
> +format_scratch
> +_kernlog "no bad inodes"
> +_scratch_mount
> +_scratch_scrub >> $seqres.full
> +exercise_scratch
> +_scratch_unmount
> +final_check_scratch
> +
> +echo "+ Part 2: Corrupt the first inode in the bucket" | tee -a $seqres.full
> +format_scratch
> +corrupt_scratch 1
> +_scratch_mount
> +_scratch_scrub >> $seqres.full
> +exercise_scratch
> +_scratch_unmount
> +final_check_scratch
> +
> +echo "+ Part 3: Corrupt the middle inode in the bucket" | tee -a $seqres.full
> +format_scratch
> +corrupt_scratch 3
> +_scratch_mount
> +_scratch_scrub >> $seqres.full
> +exercise_scratch
> +_scratch_unmount
> +final_check_scratch
> +
> +echo "+ Part 4: Corrupt the last inode in the bucket" | tee -a $seqres.full
> +format_scratch
> +corrupt_scratch 5
> +_scratch_mount
> +_scratch_scrub >> $seqres.full
> +exercise_scratch
> +_scratch_unmount
> +final_check_scratch
> +
> +# success, all done
> +echo Silence is golden
> +status=0
> +exit
> diff --git a/tests/xfs/1873.out b/tests/xfs/1873.out
> new file mode 100644
> index 0000000000..0e36bd2304
> --- /dev/null
> +++ b/tests/xfs/1873.out
> @@ -0,0 +1,6 @@
> +QA output created by 1873
> ++ Part 1: See if scrub can recover the unlinked list
> ++ Part 2: Corrupt the first inode in the bucket
> ++ Part 3: Corrupt the middle inode in the bucket
> ++ Part 4: Corrupt the last inode in the bucket
> +Silence is golden
>
^ permalink raw reply [flat|nested] 8+ messages in thread
* [PATCH 1/1] xfs: test unlinked inode list repair on demand
2023-10-09 18:18 [PATCHSET v2 0/1] fstests: reload entire iunlink lists Darrick J. Wong
@ 2023-10-09 18:18 ` Darrick J. Wong
2023-11-01 17:20 ` Darrick J. Wong
0 siblings, 1 reply; 8+ messages in thread
From: Darrick J. Wong @ 2023-10-09 18:18 UTC (permalink / raw)
To: djwong, zlang; +Cc: linux-xfs, fstests, guan, david
From: Darrick J. Wong <djwong@kernel.org>
Create a test to exercise recovery of unlinked inodes on a clean
filesystem. This was definitely possible on old kernels that on an ro
mount would clean the log without processing the iunlink list.
Signed-off-by: Darrick J. Wong <djwong@kernel.org>
---
common/rc | 4 +
tests/xfs/1872 | 113 +++++++++++++++++++++++++++
tests/xfs/1872.out | 5 +
tests/xfs/1873 | 217 ++++++++++++++++++++++++++++++++++++++++++++++++++++
tests/xfs/1873.out | 6 +
5 files changed, 344 insertions(+), 1 deletion(-)
create mode 100755 tests/xfs/1872
create mode 100644 tests/xfs/1872.out
create mode 100755 tests/xfs/1873
create mode 100644 tests/xfs/1873.out
diff --git a/common/rc b/common/rc
index 259a1ffb09..20d3e61c4e 100644
--- a/common/rc
+++ b/common/rc
@@ -2668,9 +2668,11 @@ _require_xfs_io_command()
param_checked="$pwrite_opts $param"
;;
"scrub"|"repair")
- testio=`$XFS_IO_PROG -x -c "$command probe" $TEST_DIR 2>&1`
+ test -z "$param" && param="probe"
+ testio=`$XFS_IO_PROG -x -c "$command $param" $TEST_DIR 2>&1`
echo $testio | grep -q "Inappropriate ioctl" && \
_notrun "xfs_io $command support is missing"
+ param_checked="$param"
;;
"startupdate"|"commitupdate"|"cancelupdate")
$XFS_IO_PROG -f -c 'pwrite -S 0x58 0 128k -b 128k' $testfile > /dev/null
diff --git a/tests/xfs/1872 b/tests/xfs/1872
new file mode 100755
index 0000000000..3720a3d184
--- /dev/null
+++ b/tests/xfs/1872
@@ -0,0 +1,113 @@
+#! /bin/bash
+# SPDX-License-Identifier: GPL-2.0
+# Copyright (c) 2023 Oracle. All Rights Reserved.
+#
+# FS QA Test No. 1872
+#
+# Test using runtime code to fix unlinked inodes on a clean filesystem that
+# never got cleaned up.
+#
+. ./common/preamble
+_begin_fstest auto quick unlink
+
+# Import common functions.
+source ./common/filter
+source ./common/fuzzy
+source ./common/quota
+
+# real QA test starts here
+
+# Modify as appropriate.
+_supported_fs generic
+_require_xfs_db_command iunlink
+_require_scratch_nocheck # we'll run repair ourselves
+
+# From the AGI definition
+XFS_AGI_UNLINKED_BUCKETS=64
+
+# Try to make each iunlink bucket have this many inodes in it.
+IUNLINK_BUCKETLEN=5
+
+# XXX Forcibly disable quota since quotacheck will break this test
+orig_mount_options="$MOUNT_OPTIONS"
+_qmount_option 'noquota'
+
+format_scratch() {
+ _scratch_mkfs -d agcount=1 | _filter_mkfs 2> "${tmp}.mkfs" >> $seqres.full
+ source "${tmp}.mkfs"
+ test "${agcount}" -eq 1 || _notrun "test requires 1 AG for error injection"
+
+ local nr_iunlinks="$((IUNLINK_BUCKETLEN * XFS_AGI_UNLINKED_BUCKETS))"
+ readarray -t BADINODES < <(_scratch_xfs_db -x -c "iunlink -n $nr_iunlinks" | awk '{print $4}')
+}
+
+__repair_check_scratch() {
+ _scratch_xfs_repair -o force_geometry -n 2>&1 | \
+ tee -a $seqres.full | \
+ grep -E '(disconnected inode.*would move|next_unlinked in inode|unlinked bucket.*is.*in ag)'
+ return "${PIPESTATUS[0]}"
+}
+
+exercise_scratch() {
+ # Create a bunch of files...
+ declare -A inums
+ for ((i = 0; i < (XFS_AGI_UNLINKED_BUCKETS * 2); i++)); do
+ touch "${SCRATCH_MNT}/${i}" || break
+ inums["${i}"]="$(stat -c %i "${SCRATCH_MNT}/${i}")"
+ done
+
+ # ...then delete them to exercise the unlinked buckets
+ for ((i = 0; i < (XFS_AGI_UNLINKED_BUCKETS * 2); i++)); do
+ if ! rm -f "${SCRATCH_MNT}/${i}"; then
+ echo "rm failed on inum ${inums[$i]}"
+ break
+ fi
+ done
+}
+
+# Offline repair should not find anything
+final_check_scratch() {
+ __repair_check_scratch
+ res=$?
+ if [ $res -eq 2 ]; then
+ echo "scratch fs went offline?"
+ _scratch_mount
+ _scratch_unmount
+ __repair_check_scratch
+ fi
+ test "$res" -ne 0 && echo "repair returned $res?"
+}
+
+echo "+ Part 0: See if runtime can recover the unlinked list" | tee -a $seqres.full
+format_scratch
+_kernlog "part 0"
+_scratch_mount
+exercise_scratch
+_scratch_unmount
+final_check_scratch
+
+echo "+ Part 1: See if bulkstat can recover the unlinked list" | tee -a $seqres.full
+format_scratch
+_kernlog "part 1"
+_scratch_mount
+$XFS_IO_PROG -c 'bulkstat' $SCRATCH_MNT > /dev/null
+exercise_scratch
+_scratch_unmount
+final_check_scratch
+
+echo "+ Part 2: See if quotacheck can recover the unlinked list" | tee -a $seqres.full
+if [ -f /proc/fs/xfs/xqmstat ]; then
+ MOUNT_OPTIONS="$orig_mount_options"
+ _qmount_option 'quota'
+ format_scratch
+ _kernlog "part 2"
+ _scratch_mount
+ exercise_scratch
+ _scratch_unmount
+ final_check_scratch
+fi
+
+# success, all done
+echo Silence is golden
+status=0
+exit
diff --git a/tests/xfs/1872.out b/tests/xfs/1872.out
new file mode 100644
index 0000000000..248f0e2416
--- /dev/null
+++ b/tests/xfs/1872.out
@@ -0,0 +1,5 @@
+QA output created by 1872
++ Part 0: See if runtime can recover the unlinked list
++ Part 1: See if bulkstat can recover the unlinked list
++ Part 2: See if quotacheck can recover the unlinked list
+Silence is golden
diff --git a/tests/xfs/1873 b/tests/xfs/1873
new file mode 100755
index 0000000000..4712dee7ab
--- /dev/null
+++ b/tests/xfs/1873
@@ -0,0 +1,217 @@
+#! /bin/bash
+# SPDX-License-Identifier: GPL-2.0
+# Copyright (c) 2023 Oracle. All Rights Reserved.
+#
+# FS QA Test No. 1873
+#
+# Functional test of using online repair to fix unlinked inodes on a clean
+# filesystem that never got cleaned up.
+#
+. ./common/preamble
+_begin_fstest auto online_repair
+
+# Import common functions.
+source ./common/filter
+source ./common/fuzzy
+source ./common/quota
+
+# real QA test starts here
+
+# Modify as appropriate.
+_supported_fs generic
+_require_xfs_db_command iunlink
+# The iunlink bucket repair code wasn't added to the AGI repair code
+# until after the directory repair code was merged
+_require_xfs_io_command repair -R directory
+_require_scratch_nocheck # repair doesn't like single-AG fs
+
+# From the AGI definition
+XFS_AGI_UNLINKED_BUCKETS=64
+
+# Try to make each iunlink bucket have this many inodes in it.
+IUNLINK_BUCKETLEN=5
+
+# XXX Forcibly disable quota since quotacheck will break this test
+_qmount_option 'noquota'
+
+format_scratch() {
+ _scratch_mkfs -d agcount=1 | _filter_mkfs 2> "${tmp}.mkfs" >> $seqres.full
+ source "${tmp}.mkfs"
+ test "${agcount}" -eq 1 || _notrun "test requires 1 AG for error injection"
+
+ local nr_iunlinks="$((IUNLINK_BUCKETLEN * XFS_AGI_UNLINKED_BUCKETS))"
+ readarray -t BADINODES < <(_scratch_xfs_db -x -c "iunlink -n $nr_iunlinks" | awk '{print $4}')
+}
+
+__repair_check_scratch() {
+ _scratch_xfs_repair -o force_geometry -n 2>&1 | \
+ tee -a $seqres.full | \
+ grep -E '(disconnected inode.*would move|next_unlinked in inode|unlinked bucket.*is.*in ag)'
+ return "${PIPESTATUS[0]}"
+}
+
+corrupt_scratch() {
+ # How far into the iunlink bucket chain do we target inodes for corruption?
+ # 1 = target the inode pointed to by the AGI
+ # 3 = middle of bucket list
+ # 5 = last element in bucket
+ local corruption_bucket_depth="$1"
+ if ((corruption_bucket_depth < 1 || corruption_bucket_depth > IUNLINK_BUCKETLEN)); then
+ echo "${corruption_bucket_depth}: Value must be between 1 and ${IUNLINK_BUCKETLEN}."
+ return 1
+ fi
+
+ # Index of the inode numbers within BADINODES
+ local bad_ino1_idx=$(( (IUNLINK_BUCKETLEN - corruption_bucket_depth) * XFS_AGI_UNLINKED_BUCKETS))
+ local bad_ino2_idx=$((bad_ino1_idx + 1))
+
+ # Inode numbers to target
+ local bad_ino1="${BADINODES[bad_ino1_idx]}"
+ local bad_ino2="${BADINODES[bad_ino2_idx]}"
+ printf "bad: 0x%x 0x%x\n" "${bad_ino1}" "${bad_ino2}" | _tee_kernlog >> $seqres.full
+
+ # Bucket within AGI 0's iunlinked array.
+ local ino1_bucket="$((bad_ino1 % XFS_AGI_UNLINKED_BUCKETS))"
+ local ino2_bucket="$((bad_ino2 % XFS_AGI_UNLINKED_BUCKETS))"
+
+ # The first bad inode stays on the unlinked list but gets a nonzero
+ # nlink; the second bad inode is removed from the unlinked list but
+ # keeps its zero nlink
+ _scratch_xfs_db -x \
+ -c "inode ${bad_ino1}" -c "write -d core.nlinkv2 5555" \
+ -c "agi 0" -c "fuzz -d unlinked[${ino2_bucket}] ones" -c "print unlinked" >> $seqres.full
+
+ local iwatch=()
+ local idx
+
+ # Make a list of the adjacent iunlink bucket inodes for the first inode
+ # that we targeted.
+ if [ "${corruption_bucket_depth}" -gt 1 ]; then
+ # Previous ino in bucket
+ idx=$(( (IUNLINK_BUCKETLEN - corruption_bucket_depth + 1) * XFS_AGI_UNLINKED_BUCKETS))
+ iwatch+=("${BADINODES[idx]}")
+ fi
+ iwatch+=("${bad_ino1}")
+ if [ "$((corruption_bucket_depth + 1))" -lt "${IUNLINK_BUCKETLEN}" ]; then
+ # Next ino in bucket
+ idx=$(( (IUNLINK_BUCKETLEN - corruption_bucket_depth - 1) * XFS_AGI_UNLINKED_BUCKETS))
+ iwatch+=("${BADINODES[idx]}")
+ fi
+
+ # Make a list of the adjacent iunlink bucket inodes for the second
+ # inode that we targeted.
+ if [ "${corruption_bucket_depth}" -gt 1 ]; then
+ # Previous ino in bucket
+ idx=$(( (IUNLINK_BUCKETLEN - corruption_bucket_depth + 1) * XFS_AGI_UNLINKED_BUCKETS))
+ iwatch+=("${BADINODES[idx + 1]}")
+ fi
+ iwatch+=("${bad_ino2}")
+ if [ "$((corruption_bucket_depth + 1))" -lt "${IUNLINK_BUCKETLEN}" ]; then
+ # Next ino in bucket
+ idx=$(( (IUNLINK_BUCKETLEN - corruption_bucket_depth - 1) * XFS_AGI_UNLINKED_BUCKETS))
+ iwatch+=("${BADINODES[idx + 1]}")
+ fi
+
+ # Construct a grep string for tracepoints.
+ GREP_STR="(xrep_attempt|xrep_done|bucket ${ino1_bucket} |bucket ${ino2_bucket} |bucket ${fuzz_bucket} "
+ GREP_STR="(xrep_attempt|xrep_done|bucket ${ino1_bucket} |bucket ${ino2_bucket} "
+ for ino in "${iwatch[@]}"; do
+ f="$(printf "|ino 0x%x" "${ino}")"
+ GREP_STR="${GREP_STR}${f}"
+ done
+ GREP_STR="${GREP_STR})"
+ echo "grep -E \"${GREP_STR}\"" >> $seqres.full
+
+ # Dump everything we did to to the full file.
+ local db_dump=(-c 'agi 0' -c 'print unlinked')
+ db_dump+=(-c 'addr root' -c 'print')
+ test "${ino1_bucket}" -gt 0 && \
+ db_dump+=(-c "dump_iunlinked -a 0 -b $((ino1_bucket - 1))")
+ db_dump+=(-c "dump_iunlinked -a 0 -b ${ino1_bucket}")
+ db_dump+=(-c "dump_iunlinked -a 0 -b ${ino2_bucket}")
+ test "${ino2_bucket}" -lt 63 && \
+ db_dump+=(-c "dump_iunlinked -a 0 -b $((ino2_bucket + 1))")
+ db_dump+=(-c "inode $bad_ino1" -c 'print core.nlinkv2 v3.inumber next_unlinked')
+ db_dump+=(-c "inode $bad_ino2" -c 'print core.nlinkv2 v3.inumber next_unlinked')
+ _scratch_xfs_db "${db_dump[@]}" >> $seqres.full
+
+ # Test run of repair to make sure we find disconnected inodes
+ __repair_check_scratch | \
+ sed -e 's/disconnected inode \([0-9]*\)/disconnected inode XXXXXX/g' \
+ -e 's/next_unlinked in inode \([0-9]*\)/next_unlinked in inode XXXXXX/g' \
+ -e 's/unlinked bucket \([0-9]*\) is \([0-9]*\) in ag \([0-9]*\) .inode=\([0-9]*\)/unlinked bucket YY is XXXXXX in ag Z (inode=AAAAAA/g' | \
+ uniq -c >> $seqres.full
+ res=${PIPESTATUS[0]}
+ test "$res" -ne 0 || echo "repair returned $res after corruption?"
+}
+
+exercise_scratch() {
+ # Create a bunch of files...
+ declare -A inums
+ for ((i = 0; i < (XFS_AGI_UNLINKED_BUCKETS * 2); i++)); do
+ touch "${SCRATCH_MNT}/${i}" || break
+ inums["${i}"]="$(stat -c %i "${SCRATCH_MNT}/${i}")"
+ done
+
+ # ...then delete them to exercise the unlinked buckets
+ for ((i = 0; i < (XFS_AGI_UNLINKED_BUCKETS * 2); i++)); do
+ if ! rm -f "${SCRATCH_MNT}/${i}"; then
+ echo "rm failed on inum ${inums[$i]}"
+ break
+ fi
+ done
+}
+
+# Offline repair should not find anything
+final_check_scratch() {
+ __repair_check_scratch
+ res=$?
+ if [ $res -eq 2 ]; then
+ echo "scratch fs went offline?"
+ _scratch_mount
+ _scratch_unmount
+ __repair_check_scratch
+ fi
+ test "$res" -ne 0 && echo "repair returned $res?"
+}
+
+echo "+ Part 1: See if scrub can recover the unlinked list" | tee -a $seqres.full
+format_scratch
+_kernlog "no bad inodes"
+_scratch_mount
+_scratch_scrub >> $seqres.full
+exercise_scratch
+_scratch_unmount
+final_check_scratch
+
+echo "+ Part 2: Corrupt the first inode in the bucket" | tee -a $seqres.full
+format_scratch
+corrupt_scratch 1
+_scratch_mount
+_scratch_scrub >> $seqres.full
+exercise_scratch
+_scratch_unmount
+final_check_scratch
+
+echo "+ Part 3: Corrupt the middle inode in the bucket" | tee -a $seqres.full
+format_scratch
+corrupt_scratch 3
+_scratch_mount
+_scratch_scrub >> $seqres.full
+exercise_scratch
+_scratch_unmount
+final_check_scratch
+
+echo "+ Part 4: Corrupt the last inode in the bucket" | tee -a $seqres.full
+format_scratch
+corrupt_scratch 5
+_scratch_mount
+_scratch_scrub >> $seqres.full
+exercise_scratch
+_scratch_unmount
+final_check_scratch
+
+# success, all done
+echo Silence is golden
+status=0
+exit
diff --git a/tests/xfs/1873.out b/tests/xfs/1873.out
new file mode 100644
index 0000000000..0e36bd2304
--- /dev/null
+++ b/tests/xfs/1873.out
@@ -0,0 +1,6 @@
+QA output created by 1873
++ Part 1: See if scrub can recover the unlinked list
++ Part 2: Corrupt the first inode in the bucket
++ Part 3: Corrupt the middle inode in the bucket
++ Part 4: Corrupt the last inode in the bucket
+Silence is golden
^ permalink raw reply related [flat|nested] 8+ messages in thread
* Re: [PATCH 1/1] xfs: test unlinked inode list repair on demand
2023-10-09 18:18 ` [PATCH 1/1] xfs: test unlinked inode list repair on demand Darrick J. Wong
@ 2023-11-01 17:20 ` Darrick J. Wong
2023-11-02 20:12 ` Zorro Lang
0 siblings, 1 reply; 8+ messages in thread
From: Darrick J. Wong @ 2023-11-01 17:20 UTC (permalink / raw)
To: zlang; +Cc: linux-xfs, fstests, guan, david
On Mon, Oct 09, 2023 at 11:18:54AM -0700, Darrick J. Wong wrote:
> From: Darrick J. Wong <djwong@kernel.org>
>
> Create a test to exercise recovery of unlinked inodes on a clean
> filesystem. This was definitely possible on old kernels that on an ro
> mount would clean the log without processing the iunlink list.
>
> Signed-off-by: Darrick J. Wong <djwong@kernel.org>
Ping? The kernel fixes and the xfs_db commands have both been merged
upstream, could we please merge this functional test of the new unlinked
inode cleanup code in the kernel XFS driver?
--D
> ---
> common/rc | 4 +
> tests/xfs/1872 | 113 +++++++++++++++++++++++++++
> tests/xfs/1872.out | 5 +
> tests/xfs/1873 | 217 ++++++++++++++++++++++++++++++++++++++++++++++++++++
> tests/xfs/1873.out | 6 +
> 5 files changed, 344 insertions(+), 1 deletion(-)
> create mode 100755 tests/xfs/1872
> create mode 100644 tests/xfs/1872.out
> create mode 100755 tests/xfs/1873
> create mode 100644 tests/xfs/1873.out
>
>
> diff --git a/common/rc b/common/rc
> index 259a1ffb09..20d3e61c4e 100644
> --- a/common/rc
> +++ b/common/rc
> @@ -2668,9 +2668,11 @@ _require_xfs_io_command()
> param_checked="$pwrite_opts $param"
> ;;
> "scrub"|"repair")
> - testio=`$XFS_IO_PROG -x -c "$command probe" $TEST_DIR 2>&1`
> + test -z "$param" && param="probe"
> + testio=`$XFS_IO_PROG -x -c "$command $param" $TEST_DIR 2>&1`
> echo $testio | grep -q "Inappropriate ioctl" && \
> _notrun "xfs_io $command support is missing"
> + param_checked="$param"
> ;;
> "startupdate"|"commitupdate"|"cancelupdate")
> $XFS_IO_PROG -f -c 'pwrite -S 0x58 0 128k -b 128k' $testfile > /dev/null
> diff --git a/tests/xfs/1872 b/tests/xfs/1872
> new file mode 100755
> index 0000000000..3720a3d184
> --- /dev/null
> +++ b/tests/xfs/1872
> @@ -0,0 +1,113 @@
> +#! /bin/bash
> +# SPDX-License-Identifier: GPL-2.0
> +# Copyright (c) 2023 Oracle. All Rights Reserved.
> +#
> +# FS QA Test No. 1872
> +#
> +# Test using runtime code to fix unlinked inodes on a clean filesystem that
> +# never got cleaned up.
> +#
> +. ./common/preamble
> +_begin_fstest auto quick unlink
> +
> +# Import common functions.
> +source ./common/filter
> +source ./common/fuzzy
> +source ./common/quota
> +
> +# real QA test starts here
> +
> +# Modify as appropriate.
> +_supported_fs generic
> +_require_xfs_db_command iunlink
> +_require_scratch_nocheck # we'll run repair ourselves
> +
> +# From the AGI definition
> +XFS_AGI_UNLINKED_BUCKETS=64
> +
> +# Try to make each iunlink bucket have this many inodes in it.
> +IUNLINK_BUCKETLEN=5
> +
> +# XXX Forcibly disable quota since quotacheck will break this test
> +orig_mount_options="$MOUNT_OPTIONS"
> +_qmount_option 'noquota'
> +
> +format_scratch() {
> + _scratch_mkfs -d agcount=1 | _filter_mkfs 2> "${tmp}.mkfs" >> $seqres.full
> + source "${tmp}.mkfs"
> + test "${agcount}" -eq 1 || _notrun "test requires 1 AG for error injection"
> +
> + local nr_iunlinks="$((IUNLINK_BUCKETLEN * XFS_AGI_UNLINKED_BUCKETS))"
> + readarray -t BADINODES < <(_scratch_xfs_db -x -c "iunlink -n $nr_iunlinks" | awk '{print $4}')
> +}
> +
> +__repair_check_scratch() {
> + _scratch_xfs_repair -o force_geometry -n 2>&1 | \
> + tee -a $seqres.full | \
> + grep -E '(disconnected inode.*would move|next_unlinked in inode|unlinked bucket.*is.*in ag)'
> + return "${PIPESTATUS[0]}"
> +}
> +
> +exercise_scratch() {
> + # Create a bunch of files...
> + declare -A inums
> + for ((i = 0; i < (XFS_AGI_UNLINKED_BUCKETS * 2); i++)); do
> + touch "${SCRATCH_MNT}/${i}" || break
> + inums["${i}"]="$(stat -c %i "${SCRATCH_MNT}/${i}")"
> + done
> +
> + # ...then delete them to exercise the unlinked buckets
> + for ((i = 0; i < (XFS_AGI_UNLINKED_BUCKETS * 2); i++)); do
> + if ! rm -f "${SCRATCH_MNT}/${i}"; then
> + echo "rm failed on inum ${inums[$i]}"
> + break
> + fi
> + done
> +}
> +
> +# Offline repair should not find anything
> +final_check_scratch() {
> + __repair_check_scratch
> + res=$?
> + if [ $res -eq 2 ]; then
> + echo "scratch fs went offline?"
> + _scratch_mount
> + _scratch_unmount
> + __repair_check_scratch
> + fi
> + test "$res" -ne 0 && echo "repair returned $res?"
> +}
> +
> +echo "+ Part 0: See if runtime can recover the unlinked list" | tee -a $seqres.full
> +format_scratch
> +_kernlog "part 0"
> +_scratch_mount
> +exercise_scratch
> +_scratch_unmount
> +final_check_scratch
> +
> +echo "+ Part 1: See if bulkstat can recover the unlinked list" | tee -a $seqres.full
> +format_scratch
> +_kernlog "part 1"
> +_scratch_mount
> +$XFS_IO_PROG -c 'bulkstat' $SCRATCH_MNT > /dev/null
> +exercise_scratch
> +_scratch_unmount
> +final_check_scratch
> +
> +echo "+ Part 2: See if quotacheck can recover the unlinked list" | tee -a $seqres.full
> +if [ -f /proc/fs/xfs/xqmstat ]; then
> + MOUNT_OPTIONS="$orig_mount_options"
> + _qmount_option 'quota'
> + format_scratch
> + _kernlog "part 2"
> + _scratch_mount
> + exercise_scratch
> + _scratch_unmount
> + final_check_scratch
> +fi
> +
> +# success, all done
> +echo Silence is golden
> +status=0
> +exit
> diff --git a/tests/xfs/1872.out b/tests/xfs/1872.out
> new file mode 100644
> index 0000000000..248f0e2416
> --- /dev/null
> +++ b/tests/xfs/1872.out
> @@ -0,0 +1,5 @@
> +QA output created by 1872
> ++ Part 0: See if runtime can recover the unlinked list
> ++ Part 1: See if bulkstat can recover the unlinked list
> ++ Part 2: See if quotacheck can recover the unlinked list
> +Silence is golden
> diff --git a/tests/xfs/1873 b/tests/xfs/1873
> new file mode 100755
> index 0000000000..4712dee7ab
> --- /dev/null
> +++ b/tests/xfs/1873
> @@ -0,0 +1,217 @@
> +#! /bin/bash
> +# SPDX-License-Identifier: GPL-2.0
> +# Copyright (c) 2023 Oracle. All Rights Reserved.
> +#
> +# FS QA Test No. 1873
> +#
> +# Functional test of using online repair to fix unlinked inodes on a clean
> +# filesystem that never got cleaned up.
> +#
> +. ./common/preamble
> +_begin_fstest auto online_repair
> +
> +# Import common functions.
> +source ./common/filter
> +source ./common/fuzzy
> +source ./common/quota
> +
> +# real QA test starts here
> +
> +# Modify as appropriate.
> +_supported_fs generic
> +_require_xfs_db_command iunlink
> +# The iunlink bucket repair code wasn't added to the AGI repair code
> +# until after the directory repair code was merged
> +_require_xfs_io_command repair -R directory
> +_require_scratch_nocheck # repair doesn't like single-AG fs
> +
> +# From the AGI definition
> +XFS_AGI_UNLINKED_BUCKETS=64
> +
> +# Try to make each iunlink bucket have this many inodes in it.
> +IUNLINK_BUCKETLEN=5
> +
> +# XXX Forcibly disable quota since quotacheck will break this test
> +_qmount_option 'noquota'
> +
> +format_scratch() {
> + _scratch_mkfs -d agcount=1 | _filter_mkfs 2> "${tmp}.mkfs" >> $seqres.full
> + source "${tmp}.mkfs"
> + test "${agcount}" -eq 1 || _notrun "test requires 1 AG for error injection"
> +
> + local nr_iunlinks="$((IUNLINK_BUCKETLEN * XFS_AGI_UNLINKED_BUCKETS))"
> + readarray -t BADINODES < <(_scratch_xfs_db -x -c "iunlink -n $nr_iunlinks" | awk '{print $4}')
> +}
> +
> +__repair_check_scratch() {
> + _scratch_xfs_repair -o force_geometry -n 2>&1 | \
> + tee -a $seqres.full | \
> + grep -E '(disconnected inode.*would move|next_unlinked in inode|unlinked bucket.*is.*in ag)'
> + return "${PIPESTATUS[0]}"
> +}
> +
> +corrupt_scratch() {
> + # How far into the iunlink bucket chain do we target inodes for corruption?
> + # 1 = target the inode pointed to by the AGI
> + # 3 = middle of bucket list
> + # 5 = last element in bucket
> + local corruption_bucket_depth="$1"
> + if ((corruption_bucket_depth < 1 || corruption_bucket_depth > IUNLINK_BUCKETLEN)); then
> + echo "${corruption_bucket_depth}: Value must be between 1 and ${IUNLINK_BUCKETLEN}."
> + return 1
> + fi
> +
> + # Index of the inode numbers within BADINODES
> + local bad_ino1_idx=$(( (IUNLINK_BUCKETLEN - corruption_bucket_depth) * XFS_AGI_UNLINKED_BUCKETS))
> + local bad_ino2_idx=$((bad_ino1_idx + 1))
> +
> + # Inode numbers to target
> + local bad_ino1="${BADINODES[bad_ino1_idx]}"
> + local bad_ino2="${BADINODES[bad_ino2_idx]}"
> + printf "bad: 0x%x 0x%x\n" "${bad_ino1}" "${bad_ino2}" | _tee_kernlog >> $seqres.full
> +
> + # Bucket within AGI 0's iunlinked array.
> + local ino1_bucket="$((bad_ino1 % XFS_AGI_UNLINKED_BUCKETS))"
> + local ino2_bucket="$((bad_ino2 % XFS_AGI_UNLINKED_BUCKETS))"
> +
> + # The first bad inode stays on the unlinked list but gets a nonzero
> + # nlink; the second bad inode is removed from the unlinked list but
> + # keeps its zero nlink
> + _scratch_xfs_db -x \
> + -c "inode ${bad_ino1}" -c "write -d core.nlinkv2 5555" \
> + -c "agi 0" -c "fuzz -d unlinked[${ino2_bucket}] ones" -c "print unlinked" >> $seqres.full
> +
> + local iwatch=()
> + local idx
> +
> + # Make a list of the adjacent iunlink bucket inodes for the first inode
> + # that we targeted.
> + if [ "${corruption_bucket_depth}" -gt 1 ]; then
> + # Previous ino in bucket
> + idx=$(( (IUNLINK_BUCKETLEN - corruption_bucket_depth + 1) * XFS_AGI_UNLINKED_BUCKETS))
> + iwatch+=("${BADINODES[idx]}")
> + fi
> + iwatch+=("${bad_ino1}")
> + if [ "$((corruption_bucket_depth + 1))" -lt "${IUNLINK_BUCKETLEN}" ]; then
> + # Next ino in bucket
> + idx=$(( (IUNLINK_BUCKETLEN - corruption_bucket_depth - 1) * XFS_AGI_UNLINKED_BUCKETS))
> + iwatch+=("${BADINODES[idx]}")
> + fi
> +
> + # Make a list of the adjacent iunlink bucket inodes for the second
> + # inode that we targeted.
> + if [ "${corruption_bucket_depth}" -gt 1 ]; then
> + # Previous ino in bucket
> + idx=$(( (IUNLINK_BUCKETLEN - corruption_bucket_depth + 1) * XFS_AGI_UNLINKED_BUCKETS))
> + iwatch+=("${BADINODES[idx + 1]}")
> + fi
> + iwatch+=("${bad_ino2}")
> + if [ "$((corruption_bucket_depth + 1))" -lt "${IUNLINK_BUCKETLEN}" ]; then
> + # Next ino in bucket
> + idx=$(( (IUNLINK_BUCKETLEN - corruption_bucket_depth - 1) * XFS_AGI_UNLINKED_BUCKETS))
> + iwatch+=("${BADINODES[idx + 1]}")
> + fi
> +
> + # Construct a grep string for tracepoints.
> + GREP_STR="(xrep_attempt|xrep_done|bucket ${ino1_bucket} |bucket ${ino2_bucket} |bucket ${fuzz_bucket} "
> + GREP_STR="(xrep_attempt|xrep_done|bucket ${ino1_bucket} |bucket ${ino2_bucket} "
> + for ino in "${iwatch[@]}"; do
> + f="$(printf "|ino 0x%x" "${ino}")"
> + GREP_STR="${GREP_STR}${f}"
> + done
> + GREP_STR="${GREP_STR})"
> + echo "grep -E \"${GREP_STR}\"" >> $seqres.full
> +
> + # Dump everything we did to to the full file.
> + local db_dump=(-c 'agi 0' -c 'print unlinked')
> + db_dump+=(-c 'addr root' -c 'print')
> + test "${ino1_bucket}" -gt 0 && \
> + db_dump+=(-c "dump_iunlinked -a 0 -b $((ino1_bucket - 1))")
> + db_dump+=(-c "dump_iunlinked -a 0 -b ${ino1_bucket}")
> + db_dump+=(-c "dump_iunlinked -a 0 -b ${ino2_bucket}")
> + test "${ino2_bucket}" -lt 63 && \
> + db_dump+=(-c "dump_iunlinked -a 0 -b $((ino2_bucket + 1))")
> + db_dump+=(-c "inode $bad_ino1" -c 'print core.nlinkv2 v3.inumber next_unlinked')
> + db_dump+=(-c "inode $bad_ino2" -c 'print core.nlinkv2 v3.inumber next_unlinked')
> + _scratch_xfs_db "${db_dump[@]}" >> $seqres.full
> +
> + # Test run of repair to make sure we find disconnected inodes
> + __repair_check_scratch | \
> + sed -e 's/disconnected inode \([0-9]*\)/disconnected inode XXXXXX/g' \
> + -e 's/next_unlinked in inode \([0-9]*\)/next_unlinked in inode XXXXXX/g' \
> + -e 's/unlinked bucket \([0-9]*\) is \([0-9]*\) in ag \([0-9]*\) .inode=\([0-9]*\)/unlinked bucket YY is XXXXXX in ag Z (inode=AAAAAA/g' | \
> + uniq -c >> $seqres.full
> + res=${PIPESTATUS[0]}
> + test "$res" -ne 0 || echo "repair returned $res after corruption?"
> +}
> +
> +exercise_scratch() {
> + # Create a bunch of files...
> + declare -A inums
> + for ((i = 0; i < (XFS_AGI_UNLINKED_BUCKETS * 2); i++)); do
> + touch "${SCRATCH_MNT}/${i}" || break
> + inums["${i}"]="$(stat -c %i "${SCRATCH_MNT}/${i}")"
> + done
> +
> + # ...then delete them to exercise the unlinked buckets
> + for ((i = 0; i < (XFS_AGI_UNLINKED_BUCKETS * 2); i++)); do
> + if ! rm -f "${SCRATCH_MNT}/${i}"; then
> + echo "rm failed on inum ${inums[$i]}"
> + break
> + fi
> + done
> +}
> +
> +# Offline repair should not find anything
> +final_check_scratch() {
> + __repair_check_scratch
> + res=$?
> + if [ $res -eq 2 ]; then
> + echo "scratch fs went offline?"
> + _scratch_mount
> + _scratch_unmount
> + __repair_check_scratch
> + fi
> + test "$res" -ne 0 && echo "repair returned $res?"
> +}
> +
> +echo "+ Part 1: See if scrub can recover the unlinked list" | tee -a $seqres.full
> +format_scratch
> +_kernlog "no bad inodes"
> +_scratch_mount
> +_scratch_scrub >> $seqres.full
> +exercise_scratch
> +_scratch_unmount
> +final_check_scratch
> +
> +echo "+ Part 2: Corrupt the first inode in the bucket" | tee -a $seqres.full
> +format_scratch
> +corrupt_scratch 1
> +_scratch_mount
> +_scratch_scrub >> $seqres.full
> +exercise_scratch
> +_scratch_unmount
> +final_check_scratch
> +
> +echo "+ Part 3: Corrupt the middle inode in the bucket" | tee -a $seqres.full
> +format_scratch
> +corrupt_scratch 3
> +_scratch_mount
> +_scratch_scrub >> $seqres.full
> +exercise_scratch
> +_scratch_unmount
> +final_check_scratch
> +
> +echo "+ Part 4: Corrupt the last inode in the bucket" | tee -a $seqres.full
> +format_scratch
> +corrupt_scratch 5
> +_scratch_mount
> +_scratch_scrub >> $seqres.full
> +exercise_scratch
> +_scratch_unmount
> +final_check_scratch
> +
> +# success, all done
> +echo Silence is golden
> +status=0
> +exit
> diff --git a/tests/xfs/1873.out b/tests/xfs/1873.out
> new file mode 100644
> index 0000000000..0e36bd2304
> --- /dev/null
> +++ b/tests/xfs/1873.out
> @@ -0,0 +1,6 @@
> +QA output created by 1873
> ++ Part 1: See if scrub can recover the unlinked list
> ++ Part 2: Corrupt the first inode in the bucket
> ++ Part 3: Corrupt the middle inode in the bucket
> ++ Part 4: Corrupt the last inode in the bucket
> +Silence is golden
>
^ permalink raw reply [flat|nested] 8+ messages in thread
* Re: [PATCH 1/1] xfs: test unlinked inode list repair on demand
2023-11-01 17:20 ` Darrick J. Wong
@ 2023-11-02 20:12 ` Zorro Lang
2023-11-02 22:18 ` Darrick J. Wong
0 siblings, 1 reply; 8+ messages in thread
From: Zorro Lang @ 2023-11-02 20:12 UTC (permalink / raw)
To: Darrick J. Wong; +Cc: linux-xfs, fstests
On Wed, Nov 01, 2023 at 10:20:54AM -0700, Darrick J. Wong wrote:
> On Mon, Oct 09, 2023 at 11:18:54AM -0700, Darrick J. Wong wrote:
> > From: Darrick J. Wong <djwong@kernel.org>
> >
> > Create a test to exercise recovery of unlinked inodes on a clean
> > filesystem. This was definitely possible on old kernels that on an ro
> > mount would clean the log without processing the iunlink list.
> >
> > Signed-off-by: Darrick J. Wong <djwong@kernel.org>
>
> Ping? The kernel fixes and the xfs_db commands have both been merged
> upstream, could we please merge this functional test of the new unlinked
> inode cleanup code in the kernel XFS driver?
Sure Darrick. I'll try to merge it in next fstests release.
The x/1872 missed the _kernlog(), I think it's a Darrick specific common
helper :)
The x/1873 need the xfs_io repair -R, I'm using the lastest xfsprogs
for-next branch. Where can I get the -R support ?
Thanks,
Zorro
>
> --D
>
> > ---
> > common/rc | 4 +
> > tests/xfs/1872 | 113 +++++++++++++++++++++++++++
> > tests/xfs/1872.out | 5 +
> > tests/xfs/1873 | 217 ++++++++++++++++++++++++++++++++++++++++++++++++++++
> > tests/xfs/1873.out | 6 +
> > 5 files changed, 344 insertions(+), 1 deletion(-)
> > create mode 100755 tests/xfs/1872
> > create mode 100644 tests/xfs/1872.out
> > create mode 100755 tests/xfs/1873
> > create mode 100644 tests/xfs/1873.out
> >
> >
> > diff --git a/common/rc b/common/rc
> > index 259a1ffb09..20d3e61c4e 100644
> > --- a/common/rc
> > +++ b/common/rc
> > @@ -2668,9 +2668,11 @@ _require_xfs_io_command()
> > param_checked="$pwrite_opts $param"
> > ;;
> > "scrub"|"repair")
> > - testio=`$XFS_IO_PROG -x -c "$command probe" $TEST_DIR 2>&1`
> > + test -z "$param" && param="probe"
> > + testio=`$XFS_IO_PROG -x -c "$command $param" $TEST_DIR 2>&1`
> > echo $testio | grep -q "Inappropriate ioctl" && \
> > _notrun "xfs_io $command support is missing"
> > + param_checked="$param"
> > ;;
> > "startupdate"|"commitupdate"|"cancelupdate")
> > $XFS_IO_PROG -f -c 'pwrite -S 0x58 0 128k -b 128k' $testfile > /dev/null
> > diff --git a/tests/xfs/1872 b/tests/xfs/1872
> > new file mode 100755
> > index 0000000000..3720a3d184
> > --- /dev/null
> > +++ b/tests/xfs/1872
> > @@ -0,0 +1,113 @@
> > +#! /bin/bash
> > +# SPDX-License-Identifier: GPL-2.0
> > +# Copyright (c) 2023 Oracle. All Rights Reserved.
> > +#
> > +# FS QA Test No. 1872
> > +#
> > +# Test using runtime code to fix unlinked inodes on a clean filesystem that
> > +# never got cleaned up.
> > +#
> > +. ./common/preamble
> > +_begin_fstest auto quick unlink
> > +
> > +# Import common functions.
> > +source ./common/filter
> > +source ./common/fuzzy
> > +source ./common/quota
> > +
> > +# real QA test starts here
> > +
> > +# Modify as appropriate.
> > +_supported_fs generic
> > +_require_xfs_db_command iunlink
> > +_require_scratch_nocheck # we'll run repair ourselves
> > +
> > +# From the AGI definition
> > +XFS_AGI_UNLINKED_BUCKETS=64
> > +
> > +# Try to make each iunlink bucket have this many inodes in it.
> > +IUNLINK_BUCKETLEN=5
> > +
> > +# XXX Forcibly disable quota since quotacheck will break this test
> > +orig_mount_options="$MOUNT_OPTIONS"
> > +_qmount_option 'noquota'
> > +
> > +format_scratch() {
> > + _scratch_mkfs -d agcount=1 | _filter_mkfs 2> "${tmp}.mkfs" >> $seqres.full
> > + source "${tmp}.mkfs"
> > + test "${agcount}" -eq 1 || _notrun "test requires 1 AG for error injection"
> > +
> > + local nr_iunlinks="$((IUNLINK_BUCKETLEN * XFS_AGI_UNLINKED_BUCKETS))"
> > + readarray -t BADINODES < <(_scratch_xfs_db -x -c "iunlink -n $nr_iunlinks" | awk '{print $4}')
> > +}
> > +
> > +__repair_check_scratch() {
> > + _scratch_xfs_repair -o force_geometry -n 2>&1 | \
> > + tee -a $seqres.full | \
> > + grep -E '(disconnected inode.*would move|next_unlinked in inode|unlinked bucket.*is.*in ag)'
> > + return "${PIPESTATUS[0]}"
> > +}
> > +
> > +exercise_scratch() {
> > + # Create a bunch of files...
> > + declare -A inums
> > + for ((i = 0; i < (XFS_AGI_UNLINKED_BUCKETS * 2); i++)); do
> > + touch "${SCRATCH_MNT}/${i}" || break
> > + inums["${i}"]="$(stat -c %i "${SCRATCH_MNT}/${i}")"
> > + done
> > +
> > + # ...then delete them to exercise the unlinked buckets
> > + for ((i = 0; i < (XFS_AGI_UNLINKED_BUCKETS * 2); i++)); do
> > + if ! rm -f "${SCRATCH_MNT}/${i}"; then
> > + echo "rm failed on inum ${inums[$i]}"
> > + break
> > + fi
> > + done
> > +}
> > +
> > +# Offline repair should not find anything
> > +final_check_scratch() {
> > + __repair_check_scratch
> > + res=$?
> > + if [ $res -eq 2 ]; then
> > + echo "scratch fs went offline?"
> > + _scratch_mount
> > + _scratch_unmount
> > + __repair_check_scratch
> > + fi
> > + test "$res" -ne 0 && echo "repair returned $res?"
> > +}
> > +
> > +echo "+ Part 0: See if runtime can recover the unlinked list" | tee -a $seqres.full
> > +format_scratch
> > +_kernlog "part 0"
> > +_scratch_mount
> > +exercise_scratch
> > +_scratch_unmount
> > +final_check_scratch
> > +
> > +echo "+ Part 1: See if bulkstat can recover the unlinked list" | tee -a $seqres.full
> > +format_scratch
> > +_kernlog "part 1"
> > +_scratch_mount
> > +$XFS_IO_PROG -c 'bulkstat' $SCRATCH_MNT > /dev/null
> > +exercise_scratch
> > +_scratch_unmount
> > +final_check_scratch
> > +
> > +echo "+ Part 2: See if quotacheck can recover the unlinked list" | tee -a $seqres.full
> > +if [ -f /proc/fs/xfs/xqmstat ]; then
> > + MOUNT_OPTIONS="$orig_mount_options"
> > + _qmount_option 'quota'
> > + format_scratch
> > + _kernlog "part 2"
> > + _scratch_mount
> > + exercise_scratch
> > + _scratch_unmount
> > + final_check_scratch
> > +fi
> > +
> > +# success, all done
> > +echo Silence is golden
> > +status=0
> > +exit
> > diff --git a/tests/xfs/1872.out b/tests/xfs/1872.out
> > new file mode 100644
> > index 0000000000..248f0e2416
> > --- /dev/null
> > +++ b/tests/xfs/1872.out
> > @@ -0,0 +1,5 @@
> > +QA output created by 1872
> > ++ Part 0: See if runtime can recover the unlinked list
> > ++ Part 1: See if bulkstat can recover the unlinked list
> > ++ Part 2: See if quotacheck can recover the unlinked list
> > +Silence is golden
> > diff --git a/tests/xfs/1873 b/tests/xfs/1873
> > new file mode 100755
> > index 0000000000..4712dee7ab
> > --- /dev/null
> > +++ b/tests/xfs/1873
> > @@ -0,0 +1,217 @@
> > +#! /bin/bash
> > +# SPDX-License-Identifier: GPL-2.0
> > +# Copyright (c) 2023 Oracle. All Rights Reserved.
> > +#
> > +# FS QA Test No. 1873
> > +#
> > +# Functional test of using online repair to fix unlinked inodes on a clean
> > +# filesystem that never got cleaned up.
> > +#
> > +. ./common/preamble
> > +_begin_fstest auto online_repair
> > +
> > +# Import common functions.
> > +source ./common/filter
> > +source ./common/fuzzy
> > +source ./common/quota
> > +
> > +# real QA test starts here
> > +
> > +# Modify as appropriate.
> > +_supported_fs generic
> > +_require_xfs_db_command iunlink
> > +# The iunlink bucket repair code wasn't added to the AGI repair code
> > +# until after the directory repair code was merged
> > +_require_xfs_io_command repair -R directory
> > +_require_scratch_nocheck # repair doesn't like single-AG fs
> > +
> > +# From the AGI definition
> > +XFS_AGI_UNLINKED_BUCKETS=64
> > +
> > +# Try to make each iunlink bucket have this many inodes in it.
> > +IUNLINK_BUCKETLEN=5
> > +
> > +# XXX Forcibly disable quota since quotacheck will break this test
> > +_qmount_option 'noquota'
> > +
> > +format_scratch() {
> > + _scratch_mkfs -d agcount=1 | _filter_mkfs 2> "${tmp}.mkfs" >> $seqres.full
> > + source "${tmp}.mkfs"
> > + test "${agcount}" -eq 1 || _notrun "test requires 1 AG for error injection"
> > +
> > + local nr_iunlinks="$((IUNLINK_BUCKETLEN * XFS_AGI_UNLINKED_BUCKETS))"
> > + readarray -t BADINODES < <(_scratch_xfs_db -x -c "iunlink -n $nr_iunlinks" | awk '{print $4}')
> > +}
> > +
> > +__repair_check_scratch() {
> > + _scratch_xfs_repair -o force_geometry -n 2>&1 | \
> > + tee -a $seqres.full | \
> > + grep -E '(disconnected inode.*would move|next_unlinked in inode|unlinked bucket.*is.*in ag)'
> > + return "${PIPESTATUS[0]}"
> > +}
> > +
> > +corrupt_scratch() {
> > + # How far into the iunlink bucket chain do we target inodes for corruption?
> > + # 1 = target the inode pointed to by the AGI
> > + # 3 = middle of bucket list
> > + # 5 = last element in bucket
> > + local corruption_bucket_depth="$1"
> > + if ((corruption_bucket_depth < 1 || corruption_bucket_depth > IUNLINK_BUCKETLEN)); then
> > + echo "${corruption_bucket_depth}: Value must be between 1 and ${IUNLINK_BUCKETLEN}."
> > + return 1
> > + fi
> > +
> > + # Index of the inode numbers within BADINODES
> > + local bad_ino1_idx=$(( (IUNLINK_BUCKETLEN - corruption_bucket_depth) * XFS_AGI_UNLINKED_BUCKETS))
> > + local bad_ino2_idx=$((bad_ino1_idx + 1))
> > +
> > + # Inode numbers to target
> > + local bad_ino1="${BADINODES[bad_ino1_idx]}"
> > + local bad_ino2="${BADINODES[bad_ino2_idx]}"
> > + printf "bad: 0x%x 0x%x\n" "${bad_ino1}" "${bad_ino2}" | _tee_kernlog >> $seqres.full
> > +
> > + # Bucket within AGI 0's iunlinked array.
> > + local ino1_bucket="$((bad_ino1 % XFS_AGI_UNLINKED_BUCKETS))"
> > + local ino2_bucket="$((bad_ino2 % XFS_AGI_UNLINKED_BUCKETS))"
> > +
> > + # The first bad inode stays on the unlinked list but gets a nonzero
> > + # nlink; the second bad inode is removed from the unlinked list but
> > + # keeps its zero nlink
> > + _scratch_xfs_db -x \
> > + -c "inode ${bad_ino1}" -c "write -d core.nlinkv2 5555" \
> > + -c "agi 0" -c "fuzz -d unlinked[${ino2_bucket}] ones" -c "print unlinked" >> $seqres.full
> > +
> > + local iwatch=()
> > + local idx
> > +
> > + # Make a list of the adjacent iunlink bucket inodes for the first inode
> > + # that we targeted.
> > + if [ "${corruption_bucket_depth}" -gt 1 ]; then
> > + # Previous ino in bucket
> > + idx=$(( (IUNLINK_BUCKETLEN - corruption_bucket_depth + 1) * XFS_AGI_UNLINKED_BUCKETS))
> > + iwatch+=("${BADINODES[idx]}")
> > + fi
> > + iwatch+=("${bad_ino1}")
> > + if [ "$((corruption_bucket_depth + 1))" -lt "${IUNLINK_BUCKETLEN}" ]; then
> > + # Next ino in bucket
> > + idx=$(( (IUNLINK_BUCKETLEN - corruption_bucket_depth - 1) * XFS_AGI_UNLINKED_BUCKETS))
> > + iwatch+=("${BADINODES[idx]}")
> > + fi
> > +
> > + # Make a list of the adjacent iunlink bucket inodes for the second
> > + # inode that we targeted.
> > + if [ "${corruption_bucket_depth}" -gt 1 ]; then
> > + # Previous ino in bucket
> > + idx=$(( (IUNLINK_BUCKETLEN - corruption_bucket_depth + 1) * XFS_AGI_UNLINKED_BUCKETS))
> > + iwatch+=("${BADINODES[idx + 1]}")
> > + fi
> > + iwatch+=("${bad_ino2}")
> > + if [ "$((corruption_bucket_depth + 1))" -lt "${IUNLINK_BUCKETLEN}" ]; then
> > + # Next ino in bucket
> > + idx=$(( (IUNLINK_BUCKETLEN - corruption_bucket_depth - 1) * XFS_AGI_UNLINKED_BUCKETS))
> > + iwatch+=("${BADINODES[idx + 1]}")
> > + fi
> > +
> > + # Construct a grep string for tracepoints.
> > + GREP_STR="(xrep_attempt|xrep_done|bucket ${ino1_bucket} |bucket ${ino2_bucket} |bucket ${fuzz_bucket} "
> > + GREP_STR="(xrep_attempt|xrep_done|bucket ${ino1_bucket} |bucket ${ino2_bucket} "
> > + for ino in "${iwatch[@]}"; do
> > + f="$(printf "|ino 0x%x" "${ino}")"
> > + GREP_STR="${GREP_STR}${f}"
> > + done
> > + GREP_STR="${GREP_STR})"
> > + echo "grep -E \"${GREP_STR}\"" >> $seqres.full
> > +
> > + # Dump everything we did to to the full file.
> > + local db_dump=(-c 'agi 0' -c 'print unlinked')
> > + db_dump+=(-c 'addr root' -c 'print')
> > + test "${ino1_bucket}" -gt 0 && \
> > + db_dump+=(-c "dump_iunlinked -a 0 -b $((ino1_bucket - 1))")
> > + db_dump+=(-c "dump_iunlinked -a 0 -b ${ino1_bucket}")
> > + db_dump+=(-c "dump_iunlinked -a 0 -b ${ino2_bucket}")
> > + test "${ino2_bucket}" -lt 63 && \
> > + db_dump+=(-c "dump_iunlinked -a 0 -b $((ino2_bucket + 1))")
> > + db_dump+=(-c "inode $bad_ino1" -c 'print core.nlinkv2 v3.inumber next_unlinked')
> > + db_dump+=(-c "inode $bad_ino2" -c 'print core.nlinkv2 v3.inumber next_unlinked')
> > + _scratch_xfs_db "${db_dump[@]}" >> $seqres.full
> > +
> > + # Test run of repair to make sure we find disconnected inodes
> > + __repair_check_scratch | \
> > + sed -e 's/disconnected inode \([0-9]*\)/disconnected inode XXXXXX/g' \
> > + -e 's/next_unlinked in inode \([0-9]*\)/next_unlinked in inode XXXXXX/g' \
> > + -e 's/unlinked bucket \([0-9]*\) is \([0-9]*\) in ag \([0-9]*\) .inode=\([0-9]*\)/unlinked bucket YY is XXXXXX in ag Z (inode=AAAAAA/g' | \
> > + uniq -c >> $seqres.full
> > + res=${PIPESTATUS[0]}
> > + test "$res" -ne 0 || echo "repair returned $res after corruption?"
> > +}
> > +
> > +exercise_scratch() {
> > + # Create a bunch of files...
> > + declare -A inums
> > + for ((i = 0; i < (XFS_AGI_UNLINKED_BUCKETS * 2); i++)); do
> > + touch "${SCRATCH_MNT}/${i}" || break
> > + inums["${i}"]="$(stat -c %i "${SCRATCH_MNT}/${i}")"
> > + done
> > +
> > + # ...then delete them to exercise the unlinked buckets
> > + for ((i = 0; i < (XFS_AGI_UNLINKED_BUCKETS * 2); i++)); do
> > + if ! rm -f "${SCRATCH_MNT}/${i}"; then
> > + echo "rm failed on inum ${inums[$i]}"
> > + break
> > + fi
> > + done
> > +}
> > +
> > +# Offline repair should not find anything
> > +final_check_scratch() {
> > + __repair_check_scratch
> > + res=$?
> > + if [ $res -eq 2 ]; then
> > + echo "scratch fs went offline?"
> > + _scratch_mount
> > + _scratch_unmount
> > + __repair_check_scratch
> > + fi
> > + test "$res" -ne 0 && echo "repair returned $res?"
> > +}
> > +
> > +echo "+ Part 1: See if scrub can recover the unlinked list" | tee -a $seqres.full
> > +format_scratch
> > +_kernlog "no bad inodes"
> > +_scratch_mount
> > +_scratch_scrub >> $seqres.full
> > +exercise_scratch
> > +_scratch_unmount
> > +final_check_scratch
> > +
> > +echo "+ Part 2: Corrupt the first inode in the bucket" | tee -a $seqres.full
> > +format_scratch
> > +corrupt_scratch 1
> > +_scratch_mount
> > +_scratch_scrub >> $seqres.full
> > +exercise_scratch
> > +_scratch_unmount
> > +final_check_scratch
> > +
> > +echo "+ Part 3: Corrupt the middle inode in the bucket" | tee -a $seqres.full
> > +format_scratch
> > +corrupt_scratch 3
> > +_scratch_mount
> > +_scratch_scrub >> $seqres.full
> > +exercise_scratch
> > +_scratch_unmount
> > +final_check_scratch
> > +
> > +echo "+ Part 4: Corrupt the last inode in the bucket" | tee -a $seqres.full
> > +format_scratch
> > +corrupt_scratch 5
> > +_scratch_mount
> > +_scratch_scrub >> $seqres.full
> > +exercise_scratch
> > +_scratch_unmount
> > +final_check_scratch
> > +
> > +# success, all done
> > +echo Silence is golden
> > +status=0
> > +exit
> > diff --git a/tests/xfs/1873.out b/tests/xfs/1873.out
> > new file mode 100644
> > index 0000000000..0e36bd2304
> > --- /dev/null
> > +++ b/tests/xfs/1873.out
> > @@ -0,0 +1,6 @@
> > +QA output created by 1873
> > ++ Part 1: See if scrub can recover the unlinked list
> > ++ Part 2: Corrupt the first inode in the bucket
> > ++ Part 3: Corrupt the middle inode in the bucket
> > ++ Part 4: Corrupt the last inode in the bucket
> > +Silence is golden
> >
>
^ permalink raw reply [flat|nested] 8+ messages in thread
* Re: [PATCH 1/1] xfs: test unlinked inode list repair on demand
2023-11-02 20:12 ` Zorro Lang
@ 2023-11-02 22:18 ` Darrick J. Wong
2023-11-03 6:33 ` Zorro Lang
0 siblings, 1 reply; 8+ messages in thread
From: Darrick J. Wong @ 2023-11-02 22:18 UTC (permalink / raw)
To: Zorro Lang; +Cc: linux-xfs, fstests
On Fri, Nov 03, 2023 at 04:12:31AM +0800, Zorro Lang wrote:
> On Wed, Nov 01, 2023 at 10:20:54AM -0700, Darrick J. Wong wrote:
> > On Mon, Oct 09, 2023 at 11:18:54AM -0700, Darrick J. Wong wrote:
> > > From: Darrick J. Wong <djwong@kernel.org>
> > >
> > > Create a test to exercise recovery of unlinked inodes on a clean
> > > filesystem. This was definitely possible on old kernels that on an ro
> > > mount would clean the log without processing the iunlink list.
> > >
> > > Signed-off-by: Darrick J. Wong <djwong@kernel.org>
> >
> > Ping? The kernel fixes and the xfs_db commands have both been merged
> > upstream, could we please merge this functional test of the new unlinked
> > inode cleanup code in the kernel XFS driver?
>
> Sure Darrick. I'll try to merge it in next fstests release.
>
> The x/1872 missed the _kernlog(), I think it's a Darrick specific common
> helper :)
D'oh! Yeah, that's part of ... oh, I guess I never sent /that/ series.
The actual commit is here:
https://git.kernel.org/pub/scm/linux/kernel/git/djwong/xfstests-dev.git/commit/?h=djwong-wtf&id=ca8940e822d83574772f6c953dcadb8c12b9b2b1
> The x/1873 need the xfs_io repair -R, I'm using the lastest xfsprogs
> for-next branch. Where can I get the -R support ?
D'oh! repair -R is in the series that Carlos said he'd pick up for
xfsprogs 6.6:
https://lore.kernel.org/linux-xfs/20231003113334.zi7p7c77pqvesmfb@andromeda/
Ok, I'll put this one back to sleep until xfsprogs catches up. Or
maybe next week.
(FWIW I pushed djwong-wtf of all three branches last night.)
--D
> Thanks,
> Zorro
>
> >
> > --D
> >
> > > ---
> > > common/rc | 4 +
> > > tests/xfs/1872 | 113 +++++++++++++++++++++++++++
> > > tests/xfs/1872.out | 5 +
> > > tests/xfs/1873 | 217 ++++++++++++++++++++++++++++++++++++++++++++++++++++
> > > tests/xfs/1873.out | 6 +
> > > 5 files changed, 344 insertions(+), 1 deletion(-)
> > > create mode 100755 tests/xfs/1872
> > > create mode 100644 tests/xfs/1872.out
> > > create mode 100755 tests/xfs/1873
> > > create mode 100644 tests/xfs/1873.out
> > >
> > >
> > > diff --git a/common/rc b/common/rc
> > > index 259a1ffb09..20d3e61c4e 100644
> > > --- a/common/rc
> > > +++ b/common/rc
> > > @@ -2668,9 +2668,11 @@ _require_xfs_io_command()
> > > param_checked="$pwrite_opts $param"
> > > ;;
> > > "scrub"|"repair")
> > > - testio=`$XFS_IO_PROG -x -c "$command probe" $TEST_DIR 2>&1`
> > > + test -z "$param" && param="probe"
> > > + testio=`$XFS_IO_PROG -x -c "$command $param" $TEST_DIR 2>&1`
> > > echo $testio | grep -q "Inappropriate ioctl" && \
> > > _notrun "xfs_io $command support is missing"
> > > + param_checked="$param"
> > > ;;
> > > "startupdate"|"commitupdate"|"cancelupdate")
> > > $XFS_IO_PROG -f -c 'pwrite -S 0x58 0 128k -b 128k' $testfile > /dev/null
> > > diff --git a/tests/xfs/1872 b/tests/xfs/1872
> > > new file mode 100755
> > > index 0000000000..3720a3d184
> > > --- /dev/null
> > > +++ b/tests/xfs/1872
> > > @@ -0,0 +1,113 @@
> > > +#! /bin/bash
> > > +# SPDX-License-Identifier: GPL-2.0
> > > +# Copyright (c) 2023 Oracle. All Rights Reserved.
> > > +#
> > > +# FS QA Test No. 1872
> > > +#
> > > +# Test using runtime code to fix unlinked inodes on a clean filesystem that
> > > +# never got cleaned up.
> > > +#
> > > +. ./common/preamble
> > > +_begin_fstest auto quick unlink
> > > +
> > > +# Import common functions.
> > > +source ./common/filter
> > > +source ./common/fuzzy
> > > +source ./common/quota
> > > +
> > > +# real QA test starts here
> > > +
> > > +# Modify as appropriate.
> > > +_supported_fs generic
> > > +_require_xfs_db_command iunlink
> > > +_require_scratch_nocheck # we'll run repair ourselves
> > > +
> > > +# From the AGI definition
> > > +XFS_AGI_UNLINKED_BUCKETS=64
> > > +
> > > +# Try to make each iunlink bucket have this many inodes in it.
> > > +IUNLINK_BUCKETLEN=5
> > > +
> > > +# XXX Forcibly disable quota since quotacheck will break this test
> > > +orig_mount_options="$MOUNT_OPTIONS"
> > > +_qmount_option 'noquota'
> > > +
> > > +format_scratch() {
> > > + _scratch_mkfs -d agcount=1 | _filter_mkfs 2> "${tmp}.mkfs" >> $seqres.full
> > > + source "${tmp}.mkfs"
> > > + test "${agcount}" -eq 1 || _notrun "test requires 1 AG for error injection"
> > > +
> > > + local nr_iunlinks="$((IUNLINK_BUCKETLEN * XFS_AGI_UNLINKED_BUCKETS))"
> > > + readarray -t BADINODES < <(_scratch_xfs_db -x -c "iunlink -n $nr_iunlinks" | awk '{print $4}')
> > > +}
> > > +
> > > +__repair_check_scratch() {
> > > + _scratch_xfs_repair -o force_geometry -n 2>&1 | \
> > > + tee -a $seqres.full | \
> > > + grep -E '(disconnected inode.*would move|next_unlinked in inode|unlinked bucket.*is.*in ag)'
> > > + return "${PIPESTATUS[0]}"
> > > +}
> > > +
> > > +exercise_scratch() {
> > > + # Create a bunch of files...
> > > + declare -A inums
> > > + for ((i = 0; i < (XFS_AGI_UNLINKED_BUCKETS * 2); i++)); do
> > > + touch "${SCRATCH_MNT}/${i}" || break
> > > + inums["${i}"]="$(stat -c %i "${SCRATCH_MNT}/${i}")"
> > > + done
> > > +
> > > + # ...then delete them to exercise the unlinked buckets
> > > + for ((i = 0; i < (XFS_AGI_UNLINKED_BUCKETS * 2); i++)); do
> > > + if ! rm -f "${SCRATCH_MNT}/${i}"; then
> > > + echo "rm failed on inum ${inums[$i]}"
> > > + break
> > > + fi
> > > + done
> > > +}
> > > +
> > > +# Offline repair should not find anything
> > > +final_check_scratch() {
> > > + __repair_check_scratch
> > > + res=$?
> > > + if [ $res -eq 2 ]; then
> > > + echo "scratch fs went offline?"
> > > + _scratch_mount
> > > + _scratch_unmount
> > > + __repair_check_scratch
> > > + fi
> > > + test "$res" -ne 0 && echo "repair returned $res?"
> > > +}
> > > +
> > > +echo "+ Part 0: See if runtime can recover the unlinked list" | tee -a $seqres.full
> > > +format_scratch
> > > +_kernlog "part 0"
> > > +_scratch_mount
> > > +exercise_scratch
> > > +_scratch_unmount
> > > +final_check_scratch
> > > +
> > > +echo "+ Part 1: See if bulkstat can recover the unlinked list" | tee -a $seqres.full
> > > +format_scratch
> > > +_kernlog "part 1"
> > > +_scratch_mount
> > > +$XFS_IO_PROG -c 'bulkstat' $SCRATCH_MNT > /dev/null
> > > +exercise_scratch
> > > +_scratch_unmount
> > > +final_check_scratch
> > > +
> > > +echo "+ Part 2: See if quotacheck can recover the unlinked list" | tee -a $seqres.full
> > > +if [ -f /proc/fs/xfs/xqmstat ]; then
> > > + MOUNT_OPTIONS="$orig_mount_options"
> > > + _qmount_option 'quota'
> > > + format_scratch
> > > + _kernlog "part 2"
> > > + _scratch_mount
> > > + exercise_scratch
> > > + _scratch_unmount
> > > + final_check_scratch
> > > +fi
> > > +
> > > +# success, all done
> > > +echo Silence is golden
> > > +status=0
> > > +exit
> > > diff --git a/tests/xfs/1872.out b/tests/xfs/1872.out
> > > new file mode 100644
> > > index 0000000000..248f0e2416
> > > --- /dev/null
> > > +++ b/tests/xfs/1872.out
> > > @@ -0,0 +1,5 @@
> > > +QA output created by 1872
> > > ++ Part 0: See if runtime can recover the unlinked list
> > > ++ Part 1: See if bulkstat can recover the unlinked list
> > > ++ Part 2: See if quotacheck can recover the unlinked list
> > > +Silence is golden
> > > diff --git a/tests/xfs/1873 b/tests/xfs/1873
> > > new file mode 100755
> > > index 0000000000..4712dee7ab
> > > --- /dev/null
> > > +++ b/tests/xfs/1873
> > > @@ -0,0 +1,217 @@
> > > +#! /bin/bash
> > > +# SPDX-License-Identifier: GPL-2.0
> > > +# Copyright (c) 2023 Oracle. All Rights Reserved.
> > > +#
> > > +# FS QA Test No. 1873
> > > +#
> > > +# Functional test of using online repair to fix unlinked inodes on a clean
> > > +# filesystem that never got cleaned up.
> > > +#
> > > +. ./common/preamble
> > > +_begin_fstest auto online_repair
> > > +
> > > +# Import common functions.
> > > +source ./common/filter
> > > +source ./common/fuzzy
> > > +source ./common/quota
> > > +
> > > +# real QA test starts here
> > > +
> > > +# Modify as appropriate.
> > > +_supported_fs generic
> > > +_require_xfs_db_command iunlink
> > > +# The iunlink bucket repair code wasn't added to the AGI repair code
> > > +# until after the directory repair code was merged
> > > +_require_xfs_io_command repair -R directory
> > > +_require_scratch_nocheck # repair doesn't like single-AG fs
> > > +
> > > +# From the AGI definition
> > > +XFS_AGI_UNLINKED_BUCKETS=64
> > > +
> > > +# Try to make each iunlink bucket have this many inodes in it.
> > > +IUNLINK_BUCKETLEN=5
> > > +
> > > +# XXX Forcibly disable quota since quotacheck will break this test
> > > +_qmount_option 'noquota'
> > > +
> > > +format_scratch() {
> > > + _scratch_mkfs -d agcount=1 | _filter_mkfs 2> "${tmp}.mkfs" >> $seqres.full
> > > + source "${tmp}.mkfs"
> > > + test "${agcount}" -eq 1 || _notrun "test requires 1 AG for error injection"
> > > +
> > > + local nr_iunlinks="$((IUNLINK_BUCKETLEN * XFS_AGI_UNLINKED_BUCKETS))"
> > > + readarray -t BADINODES < <(_scratch_xfs_db -x -c "iunlink -n $nr_iunlinks" | awk '{print $4}')
> > > +}
> > > +
> > > +__repair_check_scratch() {
> > > + _scratch_xfs_repair -o force_geometry -n 2>&1 | \
> > > + tee -a $seqres.full | \
> > > + grep -E '(disconnected inode.*would move|next_unlinked in inode|unlinked bucket.*is.*in ag)'
> > > + return "${PIPESTATUS[0]}"
> > > +}
> > > +
> > > +corrupt_scratch() {
> > > + # How far into the iunlink bucket chain do we target inodes for corruption?
> > > + # 1 = target the inode pointed to by the AGI
> > > + # 3 = middle of bucket list
> > > + # 5 = last element in bucket
> > > + local corruption_bucket_depth="$1"
> > > + if ((corruption_bucket_depth < 1 || corruption_bucket_depth > IUNLINK_BUCKETLEN)); then
> > > + echo "${corruption_bucket_depth}: Value must be between 1 and ${IUNLINK_BUCKETLEN}."
> > > + return 1
> > > + fi
> > > +
> > > + # Index of the inode numbers within BADINODES
> > > + local bad_ino1_idx=$(( (IUNLINK_BUCKETLEN - corruption_bucket_depth) * XFS_AGI_UNLINKED_BUCKETS))
> > > + local bad_ino2_idx=$((bad_ino1_idx + 1))
> > > +
> > > + # Inode numbers to target
> > > + local bad_ino1="${BADINODES[bad_ino1_idx]}"
> > > + local bad_ino2="${BADINODES[bad_ino2_idx]}"
> > > + printf "bad: 0x%x 0x%x\n" "${bad_ino1}" "${bad_ino2}" | _tee_kernlog >> $seqres.full
> > > +
> > > + # Bucket within AGI 0's iunlinked array.
> > > + local ino1_bucket="$((bad_ino1 % XFS_AGI_UNLINKED_BUCKETS))"
> > > + local ino2_bucket="$((bad_ino2 % XFS_AGI_UNLINKED_BUCKETS))"
> > > +
> > > + # The first bad inode stays on the unlinked list but gets a nonzero
> > > + # nlink; the second bad inode is removed from the unlinked list but
> > > + # keeps its zero nlink
> > > + _scratch_xfs_db -x \
> > > + -c "inode ${bad_ino1}" -c "write -d core.nlinkv2 5555" \
> > > + -c "agi 0" -c "fuzz -d unlinked[${ino2_bucket}] ones" -c "print unlinked" >> $seqres.full
> > > +
> > > + local iwatch=()
> > > + local idx
> > > +
> > > + # Make a list of the adjacent iunlink bucket inodes for the first inode
> > > + # that we targeted.
> > > + if [ "${corruption_bucket_depth}" -gt 1 ]; then
> > > + # Previous ino in bucket
> > > + idx=$(( (IUNLINK_BUCKETLEN - corruption_bucket_depth + 1) * XFS_AGI_UNLINKED_BUCKETS))
> > > + iwatch+=("${BADINODES[idx]}")
> > > + fi
> > > + iwatch+=("${bad_ino1}")
> > > + if [ "$((corruption_bucket_depth + 1))" -lt "${IUNLINK_BUCKETLEN}" ]; then
> > > + # Next ino in bucket
> > > + idx=$(( (IUNLINK_BUCKETLEN - corruption_bucket_depth - 1) * XFS_AGI_UNLINKED_BUCKETS))
> > > + iwatch+=("${BADINODES[idx]}")
> > > + fi
> > > +
> > > + # Make a list of the adjacent iunlink bucket inodes for the second
> > > + # inode that we targeted.
> > > + if [ "${corruption_bucket_depth}" -gt 1 ]; then
> > > + # Previous ino in bucket
> > > + idx=$(( (IUNLINK_BUCKETLEN - corruption_bucket_depth + 1) * XFS_AGI_UNLINKED_BUCKETS))
> > > + iwatch+=("${BADINODES[idx + 1]}")
> > > + fi
> > > + iwatch+=("${bad_ino2}")
> > > + if [ "$((corruption_bucket_depth + 1))" -lt "${IUNLINK_BUCKETLEN}" ]; then
> > > + # Next ino in bucket
> > > + idx=$(( (IUNLINK_BUCKETLEN - corruption_bucket_depth - 1) * XFS_AGI_UNLINKED_BUCKETS))
> > > + iwatch+=("${BADINODES[idx + 1]}")
> > > + fi
> > > +
> > > + # Construct a grep string for tracepoints.
> > > + GREP_STR="(xrep_attempt|xrep_done|bucket ${ino1_bucket} |bucket ${ino2_bucket} |bucket ${fuzz_bucket} "
> > > + GREP_STR="(xrep_attempt|xrep_done|bucket ${ino1_bucket} |bucket ${ino2_bucket} "
> > > + for ino in "${iwatch[@]}"; do
> > > + f="$(printf "|ino 0x%x" "${ino}")"
> > > + GREP_STR="${GREP_STR}${f}"
> > > + done
> > > + GREP_STR="${GREP_STR})"
> > > + echo "grep -E \"${GREP_STR}\"" >> $seqres.full
> > > +
> > > + # Dump everything we did to to the full file.
> > > + local db_dump=(-c 'agi 0' -c 'print unlinked')
> > > + db_dump+=(-c 'addr root' -c 'print')
> > > + test "${ino1_bucket}" -gt 0 && \
> > > + db_dump+=(-c "dump_iunlinked -a 0 -b $((ino1_bucket - 1))")
> > > + db_dump+=(-c "dump_iunlinked -a 0 -b ${ino1_bucket}")
> > > + db_dump+=(-c "dump_iunlinked -a 0 -b ${ino2_bucket}")
> > > + test "${ino2_bucket}" -lt 63 && \
> > > + db_dump+=(-c "dump_iunlinked -a 0 -b $((ino2_bucket + 1))")
> > > + db_dump+=(-c "inode $bad_ino1" -c 'print core.nlinkv2 v3.inumber next_unlinked')
> > > + db_dump+=(-c "inode $bad_ino2" -c 'print core.nlinkv2 v3.inumber next_unlinked')
> > > + _scratch_xfs_db "${db_dump[@]}" >> $seqres.full
> > > +
> > > + # Test run of repair to make sure we find disconnected inodes
> > > + __repair_check_scratch | \
> > > + sed -e 's/disconnected inode \([0-9]*\)/disconnected inode XXXXXX/g' \
> > > + -e 's/next_unlinked in inode \([0-9]*\)/next_unlinked in inode XXXXXX/g' \
> > > + -e 's/unlinked bucket \([0-9]*\) is \([0-9]*\) in ag \([0-9]*\) .inode=\([0-9]*\)/unlinked bucket YY is XXXXXX in ag Z (inode=AAAAAA/g' | \
> > > + uniq -c >> $seqres.full
> > > + res=${PIPESTATUS[0]}
> > > + test "$res" -ne 0 || echo "repair returned $res after corruption?"
> > > +}
> > > +
> > > +exercise_scratch() {
> > > + # Create a bunch of files...
> > > + declare -A inums
> > > + for ((i = 0; i < (XFS_AGI_UNLINKED_BUCKETS * 2); i++)); do
> > > + touch "${SCRATCH_MNT}/${i}" || break
> > > + inums["${i}"]="$(stat -c %i "${SCRATCH_MNT}/${i}")"
> > > + done
> > > +
> > > + # ...then delete them to exercise the unlinked buckets
> > > + for ((i = 0; i < (XFS_AGI_UNLINKED_BUCKETS * 2); i++)); do
> > > + if ! rm -f "${SCRATCH_MNT}/${i}"; then
> > > + echo "rm failed on inum ${inums[$i]}"
> > > + break
> > > + fi
> > > + done
> > > +}
> > > +
> > > +# Offline repair should not find anything
> > > +final_check_scratch() {
> > > + __repair_check_scratch
> > > + res=$?
> > > + if [ $res -eq 2 ]; then
> > > + echo "scratch fs went offline?"
> > > + _scratch_mount
> > > + _scratch_unmount
> > > + __repair_check_scratch
> > > + fi
> > > + test "$res" -ne 0 && echo "repair returned $res?"
> > > +}
> > > +
> > > +echo "+ Part 1: See if scrub can recover the unlinked list" | tee -a $seqres.full
> > > +format_scratch
> > > +_kernlog "no bad inodes"
> > > +_scratch_mount
> > > +_scratch_scrub >> $seqres.full
> > > +exercise_scratch
> > > +_scratch_unmount
> > > +final_check_scratch
> > > +
> > > +echo "+ Part 2: Corrupt the first inode in the bucket" | tee -a $seqres.full
> > > +format_scratch
> > > +corrupt_scratch 1
> > > +_scratch_mount
> > > +_scratch_scrub >> $seqres.full
> > > +exercise_scratch
> > > +_scratch_unmount
> > > +final_check_scratch
> > > +
> > > +echo "+ Part 3: Corrupt the middle inode in the bucket" | tee -a $seqres.full
> > > +format_scratch
> > > +corrupt_scratch 3
> > > +_scratch_mount
> > > +_scratch_scrub >> $seqres.full
> > > +exercise_scratch
> > > +_scratch_unmount
> > > +final_check_scratch
> > > +
> > > +echo "+ Part 4: Corrupt the last inode in the bucket" | tee -a $seqres.full
> > > +format_scratch
> > > +corrupt_scratch 5
> > > +_scratch_mount
> > > +_scratch_scrub >> $seqres.full
> > > +exercise_scratch
> > > +_scratch_unmount
> > > +final_check_scratch
> > > +
> > > +# success, all done
> > > +echo Silence is golden
> > > +status=0
> > > +exit
> > > diff --git a/tests/xfs/1873.out b/tests/xfs/1873.out
> > > new file mode 100644
> > > index 0000000000..0e36bd2304
> > > --- /dev/null
> > > +++ b/tests/xfs/1873.out
> > > @@ -0,0 +1,6 @@
> > > +QA output created by 1873
> > > ++ Part 1: See if scrub can recover the unlinked list
> > > ++ Part 2: Corrupt the first inode in the bucket
> > > ++ Part 3: Corrupt the middle inode in the bucket
> > > ++ Part 4: Corrupt the last inode in the bucket
> > > +Silence is golden
> > >
> >
>
^ permalink raw reply [flat|nested] 8+ messages in thread
* Re: [PATCH 1/1] xfs: test unlinked inode list repair on demand
2023-11-02 22:18 ` Darrick J. Wong
@ 2023-11-03 6:33 ` Zorro Lang
0 siblings, 0 replies; 8+ messages in thread
From: Zorro Lang @ 2023-11-03 6:33 UTC (permalink / raw)
To: Darrick J. Wong; +Cc: linux-xfs, fstests
On Thu, Nov 02, 2023 at 03:18:18PM -0700, Darrick J. Wong wrote:
> On Fri, Nov 03, 2023 at 04:12:31AM +0800, Zorro Lang wrote:
> > On Wed, Nov 01, 2023 at 10:20:54AM -0700, Darrick J. Wong wrote:
> > > On Mon, Oct 09, 2023 at 11:18:54AM -0700, Darrick J. Wong wrote:
> > > > From: Darrick J. Wong <djwong@kernel.org>
> > > >
> > > > Create a test to exercise recovery of unlinked inodes on a clean
> > > > filesystem. This was definitely possible on old kernels that on an ro
> > > > mount would clean the log without processing the iunlink list.
> > > >
> > > > Signed-off-by: Darrick J. Wong <djwong@kernel.org>
> > >
> > > Ping? The kernel fixes and the xfs_db commands have both been merged
> > > upstream, could we please merge this functional test of the new unlinked
> > > inode cleanup code in the kernel XFS driver?
> >
> > Sure Darrick. I'll try to merge it in next fstests release.
> >
> > The x/1872 missed the _kernlog(), I think it's a Darrick specific common
> > helper :)
>
> D'oh! Yeah, that's part of ... oh, I guess I never sent /that/ series.
> The actual commit is here:
> https://git.kernel.org/pub/scm/linux/kernel/git/djwong/xfstests-dev.git/commit/?h=djwong-wtf&id=ca8940e822d83574772f6c953dcadb8c12b9b2b1
>
> > The x/1873 need the xfs_io repair -R, I'm using the lastest xfsprogs
> > for-next branch. Where can I get the -R support ?
>
> D'oh! repair -R is in the series that Carlos said he'd pick up for
> xfsprogs 6.6:
> https://lore.kernel.org/linux-xfs/20231003113334.zi7p7c77pqvesmfb@andromeda/
>
> Ok, I'll put this one back to sleep until xfsprogs catches up. Or
> maybe next week.
I'm planning to have next fstests release next weekend. So if you hope, we can
catch that :)
>
> (FWIW I pushed djwong-wtf of all three branches last night.)
>
> --D
>
> > Thanks,
> > Zorro
> >
> > >
> > > --D
> > >
> > > > ---
> > > > common/rc | 4 +
> > > > tests/xfs/1872 | 113 +++++++++++++++++++++++++++
> > > > tests/xfs/1872.out | 5 +
> > > > tests/xfs/1873 | 217 ++++++++++++++++++++++++++++++++++++++++++++++++++++
> > > > tests/xfs/1873.out | 6 +
> > > > 5 files changed, 344 insertions(+), 1 deletion(-)
> > > > create mode 100755 tests/xfs/1872
> > > > create mode 100644 tests/xfs/1872.out
> > > > create mode 100755 tests/xfs/1873
> > > > create mode 100644 tests/xfs/1873.out
> > > >
> > > >
> > > > diff --git a/common/rc b/common/rc
> > > > index 259a1ffb09..20d3e61c4e 100644
> > > > --- a/common/rc
> > > > +++ b/common/rc
> > > > @@ -2668,9 +2668,11 @@ _require_xfs_io_command()
> > > > param_checked="$pwrite_opts $param"
> > > > ;;
> > > > "scrub"|"repair")
> > > > - testio=`$XFS_IO_PROG -x -c "$command probe" $TEST_DIR 2>&1`
> > > > + test -z "$param" && param="probe"
> > > > + testio=`$XFS_IO_PROG -x -c "$command $param" $TEST_DIR 2>&1`
> > > > echo $testio | grep -q "Inappropriate ioctl" && \
> > > > _notrun "xfs_io $command support is missing"
> > > > + param_checked="$param"
> > > > ;;
> > > > "startupdate"|"commitupdate"|"cancelupdate")
> > > > $XFS_IO_PROG -f -c 'pwrite -S 0x58 0 128k -b 128k' $testfile > /dev/null
> > > > diff --git a/tests/xfs/1872 b/tests/xfs/1872
> > > > new file mode 100755
> > > > index 0000000000..3720a3d184
> > > > --- /dev/null
> > > > +++ b/tests/xfs/1872
> > > > @@ -0,0 +1,113 @@
> > > > +#! /bin/bash
> > > > +# SPDX-License-Identifier: GPL-2.0
> > > > +# Copyright (c) 2023 Oracle. All Rights Reserved.
> > > > +#
> > > > +# FS QA Test No. 1872
> > > > +#
> > > > +# Test using runtime code to fix unlinked inodes on a clean filesystem that
> > > > +# never got cleaned up.
> > > > +#
> > > > +. ./common/preamble
> > > > +_begin_fstest auto quick unlink
> > > > +
> > > > +# Import common functions.
> > > > +source ./common/filter
> > > > +source ./common/fuzzy
> > > > +source ./common/quota
> > > > +
> > > > +# real QA test starts here
> > > > +
> > > > +# Modify as appropriate.
> > > > +_supported_fs generic
> > > > +_require_xfs_db_command iunlink
> > > > +_require_scratch_nocheck # we'll run repair ourselves
> > > > +
> > > > +# From the AGI definition
> > > > +XFS_AGI_UNLINKED_BUCKETS=64
> > > > +
> > > > +# Try to make each iunlink bucket have this many inodes in it.
> > > > +IUNLINK_BUCKETLEN=5
> > > > +
> > > > +# XXX Forcibly disable quota since quotacheck will break this test
> > > > +orig_mount_options="$MOUNT_OPTIONS"
> > > > +_qmount_option 'noquota'
> > > > +
> > > > +format_scratch() {
> > > > + _scratch_mkfs -d agcount=1 | _filter_mkfs 2> "${tmp}.mkfs" >> $seqres.full
> > > > + source "${tmp}.mkfs"
> > > > + test "${agcount}" -eq 1 || _notrun "test requires 1 AG for error injection"
> > > > +
> > > > + local nr_iunlinks="$((IUNLINK_BUCKETLEN * XFS_AGI_UNLINKED_BUCKETS))"
> > > > + readarray -t BADINODES < <(_scratch_xfs_db -x -c "iunlink -n $nr_iunlinks" | awk '{print $4}')
> > > > +}
> > > > +
> > > > +__repair_check_scratch() {
> > > > + _scratch_xfs_repair -o force_geometry -n 2>&1 | \
> > > > + tee -a $seqres.full | \
> > > > + grep -E '(disconnected inode.*would move|next_unlinked in inode|unlinked bucket.*is.*in ag)'
> > > > + return "${PIPESTATUS[0]}"
> > > > +}
> > > > +
> > > > +exercise_scratch() {
> > > > + # Create a bunch of files...
> > > > + declare -A inums
> > > > + for ((i = 0; i < (XFS_AGI_UNLINKED_BUCKETS * 2); i++)); do
> > > > + touch "${SCRATCH_MNT}/${i}" || break
> > > > + inums["${i}"]="$(stat -c %i "${SCRATCH_MNT}/${i}")"
> > > > + done
> > > > +
> > > > + # ...then delete them to exercise the unlinked buckets
> > > > + for ((i = 0; i < (XFS_AGI_UNLINKED_BUCKETS * 2); i++)); do
> > > > + if ! rm -f "${SCRATCH_MNT}/${i}"; then
> > > > + echo "rm failed on inum ${inums[$i]}"
> > > > + break
> > > > + fi
> > > > + done
> > > > +}
> > > > +
> > > > +# Offline repair should not find anything
> > > > +final_check_scratch() {
> > > > + __repair_check_scratch
> > > > + res=$?
> > > > + if [ $res -eq 2 ]; then
> > > > + echo "scratch fs went offline?"
> > > > + _scratch_mount
> > > > + _scratch_unmount
> > > > + __repair_check_scratch
> > > > + fi
> > > > + test "$res" -ne 0 && echo "repair returned $res?"
> > > > +}
> > > > +
> > > > +echo "+ Part 0: See if runtime can recover the unlinked list" | tee -a $seqres.full
> > > > +format_scratch
> > > > +_kernlog "part 0"
> > > > +_scratch_mount
> > > > +exercise_scratch
> > > > +_scratch_unmount
> > > > +final_check_scratch
> > > > +
> > > > +echo "+ Part 1: See if bulkstat can recover the unlinked list" | tee -a $seqres.full
> > > > +format_scratch
> > > > +_kernlog "part 1"
> > > > +_scratch_mount
> > > > +$XFS_IO_PROG -c 'bulkstat' $SCRATCH_MNT > /dev/null
> > > > +exercise_scratch
> > > > +_scratch_unmount
> > > > +final_check_scratch
> > > > +
> > > > +echo "+ Part 2: See if quotacheck can recover the unlinked list" | tee -a $seqres.full
> > > > +if [ -f /proc/fs/xfs/xqmstat ]; then
> > > > + MOUNT_OPTIONS="$orig_mount_options"
> > > > + _qmount_option 'quota'
> > > > + format_scratch
> > > > + _kernlog "part 2"
> > > > + _scratch_mount
> > > > + exercise_scratch
> > > > + _scratch_unmount
> > > > + final_check_scratch
> > > > +fi
> > > > +
> > > > +# success, all done
> > > > +echo Silence is golden
> > > > +status=0
> > > > +exit
> > > > diff --git a/tests/xfs/1872.out b/tests/xfs/1872.out
> > > > new file mode 100644
> > > > index 0000000000..248f0e2416
> > > > --- /dev/null
> > > > +++ b/tests/xfs/1872.out
> > > > @@ -0,0 +1,5 @@
> > > > +QA output created by 1872
> > > > ++ Part 0: See if runtime can recover the unlinked list
> > > > ++ Part 1: See if bulkstat can recover the unlinked list
> > > > ++ Part 2: See if quotacheck can recover the unlinked list
> > > > +Silence is golden
> > > > diff --git a/tests/xfs/1873 b/tests/xfs/1873
> > > > new file mode 100755
> > > > index 0000000000..4712dee7ab
> > > > --- /dev/null
> > > > +++ b/tests/xfs/1873
> > > > @@ -0,0 +1,217 @@
> > > > +#! /bin/bash
> > > > +# SPDX-License-Identifier: GPL-2.0
> > > > +# Copyright (c) 2023 Oracle. All Rights Reserved.
> > > > +#
> > > > +# FS QA Test No. 1873
> > > > +#
> > > > +# Functional test of using online repair to fix unlinked inodes on a clean
> > > > +# filesystem that never got cleaned up.
> > > > +#
> > > > +. ./common/preamble
> > > > +_begin_fstest auto online_repair
> > > > +
> > > > +# Import common functions.
> > > > +source ./common/filter
> > > > +source ./common/fuzzy
> > > > +source ./common/quota
> > > > +
> > > > +# real QA test starts here
> > > > +
> > > > +# Modify as appropriate.
> > > > +_supported_fs generic
> > > > +_require_xfs_db_command iunlink
> > > > +# The iunlink bucket repair code wasn't added to the AGI repair code
> > > > +# until after the directory repair code was merged
> > > > +_require_xfs_io_command repair -R directory
> > > > +_require_scratch_nocheck # repair doesn't like single-AG fs
> > > > +
> > > > +# From the AGI definition
> > > > +XFS_AGI_UNLINKED_BUCKETS=64
> > > > +
> > > > +# Try to make each iunlink bucket have this many inodes in it.
> > > > +IUNLINK_BUCKETLEN=5
> > > > +
> > > > +# XXX Forcibly disable quota since quotacheck will break this test
> > > > +_qmount_option 'noquota'
> > > > +
> > > > +format_scratch() {
> > > > + _scratch_mkfs -d agcount=1 | _filter_mkfs 2> "${tmp}.mkfs" >> $seqres.full
> > > > + source "${tmp}.mkfs"
> > > > + test "${agcount}" -eq 1 || _notrun "test requires 1 AG for error injection"
> > > > +
> > > > + local nr_iunlinks="$((IUNLINK_BUCKETLEN * XFS_AGI_UNLINKED_BUCKETS))"
> > > > + readarray -t BADINODES < <(_scratch_xfs_db -x -c "iunlink -n $nr_iunlinks" | awk '{print $4}')
> > > > +}
> > > > +
> > > > +__repair_check_scratch() {
> > > > + _scratch_xfs_repair -o force_geometry -n 2>&1 | \
> > > > + tee -a $seqres.full | \
> > > > + grep -E '(disconnected inode.*would move|next_unlinked in inode|unlinked bucket.*is.*in ag)'
> > > > + return "${PIPESTATUS[0]}"
> > > > +}
> > > > +
> > > > +corrupt_scratch() {
> > > > + # How far into the iunlink bucket chain do we target inodes for corruption?
> > > > + # 1 = target the inode pointed to by the AGI
> > > > + # 3 = middle of bucket list
> > > > + # 5 = last element in bucket
> > > > + local corruption_bucket_depth="$1"
> > > > + if ((corruption_bucket_depth < 1 || corruption_bucket_depth > IUNLINK_BUCKETLEN)); then
> > > > + echo "${corruption_bucket_depth}: Value must be between 1 and ${IUNLINK_BUCKETLEN}."
> > > > + return 1
> > > > + fi
> > > > +
> > > > + # Index of the inode numbers within BADINODES
> > > > + local bad_ino1_idx=$(( (IUNLINK_BUCKETLEN - corruption_bucket_depth) * XFS_AGI_UNLINKED_BUCKETS))
> > > > + local bad_ino2_idx=$((bad_ino1_idx + 1))
> > > > +
> > > > + # Inode numbers to target
> > > > + local bad_ino1="${BADINODES[bad_ino1_idx]}"
> > > > + local bad_ino2="${BADINODES[bad_ino2_idx]}"
> > > > + printf "bad: 0x%x 0x%x\n" "${bad_ino1}" "${bad_ino2}" | _tee_kernlog >> $seqres.full
> > > > +
> > > > + # Bucket within AGI 0's iunlinked array.
> > > > + local ino1_bucket="$((bad_ino1 % XFS_AGI_UNLINKED_BUCKETS))"
> > > > + local ino2_bucket="$((bad_ino2 % XFS_AGI_UNLINKED_BUCKETS))"
> > > > +
> > > > + # The first bad inode stays on the unlinked list but gets a nonzero
> > > > + # nlink; the second bad inode is removed from the unlinked list but
> > > > + # keeps its zero nlink
> > > > + _scratch_xfs_db -x \
> > > > + -c "inode ${bad_ino1}" -c "write -d core.nlinkv2 5555" \
> > > > + -c "agi 0" -c "fuzz -d unlinked[${ino2_bucket}] ones" -c "print unlinked" >> $seqres.full
> > > > +
> > > > + local iwatch=()
> > > > + local idx
> > > > +
> > > > + # Make a list of the adjacent iunlink bucket inodes for the first inode
> > > > + # that we targeted.
> > > > + if [ "${corruption_bucket_depth}" -gt 1 ]; then
> > > > + # Previous ino in bucket
> > > > + idx=$(( (IUNLINK_BUCKETLEN - corruption_bucket_depth + 1) * XFS_AGI_UNLINKED_BUCKETS))
> > > > + iwatch+=("${BADINODES[idx]}")
> > > > + fi
> > > > + iwatch+=("${bad_ino1}")
> > > > + if [ "$((corruption_bucket_depth + 1))" -lt "${IUNLINK_BUCKETLEN}" ]; then
> > > > + # Next ino in bucket
> > > > + idx=$(( (IUNLINK_BUCKETLEN - corruption_bucket_depth - 1) * XFS_AGI_UNLINKED_BUCKETS))
> > > > + iwatch+=("${BADINODES[idx]}")
> > > > + fi
> > > > +
> > > > + # Make a list of the adjacent iunlink bucket inodes for the second
> > > > + # inode that we targeted.
> > > > + if [ "${corruption_bucket_depth}" -gt 1 ]; then
> > > > + # Previous ino in bucket
> > > > + idx=$(( (IUNLINK_BUCKETLEN - corruption_bucket_depth + 1) * XFS_AGI_UNLINKED_BUCKETS))
> > > > + iwatch+=("${BADINODES[idx + 1]}")
> > > > + fi
> > > > + iwatch+=("${bad_ino2}")
> > > > + if [ "$((corruption_bucket_depth + 1))" -lt "${IUNLINK_BUCKETLEN}" ]; then
> > > > + # Next ino in bucket
> > > > + idx=$(( (IUNLINK_BUCKETLEN - corruption_bucket_depth - 1) * XFS_AGI_UNLINKED_BUCKETS))
> > > > + iwatch+=("${BADINODES[idx + 1]}")
> > > > + fi
> > > > +
> > > > + # Construct a grep string for tracepoints.
> > > > + GREP_STR="(xrep_attempt|xrep_done|bucket ${ino1_bucket} |bucket ${ino2_bucket} |bucket ${fuzz_bucket} "
> > > > + GREP_STR="(xrep_attempt|xrep_done|bucket ${ino1_bucket} |bucket ${ino2_bucket} "
> > > > + for ino in "${iwatch[@]}"; do
> > > > + f="$(printf "|ino 0x%x" "${ino}")"
> > > > + GREP_STR="${GREP_STR}${f}"
> > > > + done
> > > > + GREP_STR="${GREP_STR})"
> > > > + echo "grep -E \"${GREP_STR}\"" >> $seqres.full
> > > > +
> > > > + # Dump everything we did to to the full file.
> > > > + local db_dump=(-c 'agi 0' -c 'print unlinked')
> > > > + db_dump+=(-c 'addr root' -c 'print')
> > > > + test "${ino1_bucket}" -gt 0 && \
> > > > + db_dump+=(-c "dump_iunlinked -a 0 -b $((ino1_bucket - 1))")
> > > > + db_dump+=(-c "dump_iunlinked -a 0 -b ${ino1_bucket}")
> > > > + db_dump+=(-c "dump_iunlinked -a 0 -b ${ino2_bucket}")
> > > > + test "${ino2_bucket}" -lt 63 && \
> > > > + db_dump+=(-c "dump_iunlinked -a 0 -b $((ino2_bucket + 1))")
> > > > + db_dump+=(-c "inode $bad_ino1" -c 'print core.nlinkv2 v3.inumber next_unlinked')
> > > > + db_dump+=(-c "inode $bad_ino2" -c 'print core.nlinkv2 v3.inumber next_unlinked')
> > > > + _scratch_xfs_db "${db_dump[@]}" >> $seqres.full
> > > > +
> > > > + # Test run of repair to make sure we find disconnected inodes
> > > > + __repair_check_scratch | \
> > > > + sed -e 's/disconnected inode \([0-9]*\)/disconnected inode XXXXXX/g' \
> > > > + -e 's/next_unlinked in inode \([0-9]*\)/next_unlinked in inode XXXXXX/g' \
> > > > + -e 's/unlinked bucket \([0-9]*\) is \([0-9]*\) in ag \([0-9]*\) .inode=\([0-9]*\)/unlinked bucket YY is XXXXXX in ag Z (inode=AAAAAA/g' | \
> > > > + uniq -c >> $seqres.full
> > > > + res=${PIPESTATUS[0]}
> > > > + test "$res" -ne 0 || echo "repair returned $res after corruption?"
> > > > +}
> > > > +
> > > > +exercise_scratch() {
> > > > + # Create a bunch of files...
> > > > + declare -A inums
> > > > + for ((i = 0; i < (XFS_AGI_UNLINKED_BUCKETS * 2); i++)); do
> > > > + touch "${SCRATCH_MNT}/${i}" || break
> > > > + inums["${i}"]="$(stat -c %i "${SCRATCH_MNT}/${i}")"
> > > > + done
> > > > +
> > > > + # ...then delete them to exercise the unlinked buckets
> > > > + for ((i = 0; i < (XFS_AGI_UNLINKED_BUCKETS * 2); i++)); do
> > > > + if ! rm -f "${SCRATCH_MNT}/${i}"; then
> > > > + echo "rm failed on inum ${inums[$i]}"
> > > > + break
> > > > + fi
> > > > + done
> > > > +}
> > > > +
> > > > +# Offline repair should not find anything
> > > > +final_check_scratch() {
> > > > + __repair_check_scratch
> > > > + res=$?
> > > > + if [ $res -eq 2 ]; then
> > > > + echo "scratch fs went offline?"
> > > > + _scratch_mount
> > > > + _scratch_unmount
> > > > + __repair_check_scratch
> > > > + fi
> > > > + test "$res" -ne 0 && echo "repair returned $res?"
> > > > +}
> > > > +
> > > > +echo "+ Part 1: See if scrub can recover the unlinked list" | tee -a $seqres.full
> > > > +format_scratch
> > > > +_kernlog "no bad inodes"
> > > > +_scratch_mount
> > > > +_scratch_scrub >> $seqres.full
> > > > +exercise_scratch
> > > > +_scratch_unmount
> > > > +final_check_scratch
> > > > +
> > > > +echo "+ Part 2: Corrupt the first inode in the bucket" | tee -a $seqres.full
> > > > +format_scratch
> > > > +corrupt_scratch 1
> > > > +_scratch_mount
> > > > +_scratch_scrub >> $seqres.full
> > > > +exercise_scratch
> > > > +_scratch_unmount
> > > > +final_check_scratch
> > > > +
> > > > +echo "+ Part 3: Corrupt the middle inode in the bucket" | tee -a $seqres.full
> > > > +format_scratch
> > > > +corrupt_scratch 3
> > > > +_scratch_mount
> > > > +_scratch_scrub >> $seqres.full
> > > > +exercise_scratch
> > > > +_scratch_unmount
> > > > +final_check_scratch
> > > > +
> > > > +echo "+ Part 4: Corrupt the last inode in the bucket" | tee -a $seqres.full
> > > > +format_scratch
> > > > +corrupt_scratch 5
> > > > +_scratch_mount
> > > > +_scratch_scrub >> $seqres.full
> > > > +exercise_scratch
> > > > +_scratch_unmount
> > > > +final_check_scratch
> > > > +
> > > > +# success, all done
> > > > +echo Silence is golden
> > > > +status=0
> > > > +exit
> > > > diff --git a/tests/xfs/1873.out b/tests/xfs/1873.out
> > > > new file mode 100644
> > > > index 0000000000..0e36bd2304
> > > > --- /dev/null
> > > > +++ b/tests/xfs/1873.out
> > > > @@ -0,0 +1,6 @@
> > > > +QA output created by 1873
> > > > ++ Part 1: See if scrub can recover the unlinked list
> > > > ++ Part 2: Corrupt the first inode in the bucket
> > > > ++ Part 3: Corrupt the middle inode in the bucket
> > > > ++ Part 4: Corrupt the last inode in the bucket
> > > > +Silence is golden
> > > >
> > >
> >
>
^ permalink raw reply [flat|nested] 8+ messages in thread
end of thread, other threads:[~2023-11-03 6:34 UTC | newest]
Thread overview: 8+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2023-09-25 21:42 [PATCHSET 0/1] fstests: reload entire iunlink lists Darrick J. Wong
2023-09-25 21:43 ` [PATCH 1/1] xfs: test unlinked inode list repair on demand Darrick J. Wong
2023-09-25 22:02 ` Darrick J. Wong
-- strict thread matches above, loose matches on Subject: below --
2023-10-09 18:18 [PATCHSET v2 0/1] fstests: reload entire iunlink lists Darrick J. Wong
2023-10-09 18:18 ` [PATCH 1/1] xfs: test unlinked inode list repair on demand Darrick J. Wong
2023-11-01 17:20 ` Darrick J. Wong
2023-11-02 20:12 ` Zorro Lang
2023-11-02 22:18 ` Darrick J. Wong
2023-11-03 6:33 ` Zorro Lang
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).