xfstests/xfs: xfs_repair secondary sb verification regression test
PATCH 1/1 xfstests/xfs: xfs_repair secondary sb verification regression test
The secondary superblock verification in xfs_repair was subject to a bug that unnecessarily leads to a brute force superblock scan if the last superblock in the fs happens to be corrupt. Normally, xfs_repair handles one-off superblock corruption gracefully using a heuristic that finds the most consistent superblock content across the set of secondary superblocks.
Create a regression test for xfs_repair that corrupts the last superblock in the fs. Verify the superblock is updated from the previously verified sb content and a brute force scan is not initiated. In the event of failure, detect that a brute force scan has started and abort the repair in order to fail the test quickly.
To support the test, extend the xfs_repair filter to handle corrupted superblock repair output and provide generic test output for arbitrary AG counts.
---
Hi all,
This is an xfs_repair regression test to trigger the problem fixed by the following previously posted fix:
http://oss.sgi.com/archives/xfs/2015-01/msg00244.html
Thoughts appreciated, thanks.
Brian
common/repair | 4 ++ tests/xfs/069 | 110 ++++++++++++++++++++++++++++++++++++++++++++++++++++++ tests/xfs/069.out | 27 ++++++++++++++ tests/xfs/group | 1 + 4 files changed, 142 insertions(+) create mode 100755 tests/xfs/069 create mode 100644 tests/xfs/069.out
common/repair +4 -0
@@ -88,6 +88,10 @@ s/(inode chunk) (\d+)\/(\d+)/AGNO\/INO/; # sunit/swidth reset messages s/^(Note - .*) were copied.*/\1 fields have been reset./; s/^(Please) reset (with .*) if necessary/\1 set \2/; +# corrupt sb messages +s/(superblock) (\d+)/\1 AGNO/; +s/(AG \#)(\d+)/\1AGNO/; +s/(reset bad sb for ag) (\d+)/\1 AGNO/; print;' }
tests/xfs/069 new file +110 -0
110 lines not on this page · raw message
tests/xfs/069.out new file +27 -0
@@ -0,0 +1,27 @@ +QA output created by 069 +Phase 1 - find and verify superblock... +Phase 2 - using <TYPEOF> log + - zero log... + - scan filesystem freespace and inode maps... +bad magic number +bad on-disk superblock AGNO - bad magic number +primary/secondary superblock AGNO conflict - AG superblock geometry info conflicts with filesystem geometry +zeroing unused portion of secondary superblock (AG #AGNO) +reset bad sb for ag AGNO + - found root inode chunk +Phase 3 - for each AG... + - scan and clear agi unlinked lists... + - process known inodes and perform inode discovery... + - process newly discovered inodes... +Phase 4 - check for duplicate blocks... + - setting up duplicate extent list... + - check for inodes claiming duplicate blocks... +Phase 5 - rebuild AG headers and trees... + - reset superblock... +Phase 6 - check inode connectivity... + - resetting contents of realtime bitmap and summary inodes + - traversing filesystem ... + - traversal finished ... + - moving disconnected inodes to lost+found ... +Phase 7 - verify and correct link counts... +done
tests/xfs/group +1 -0
@@ -66,6 +66,7 @@ 066 dump ioctl auto quick 067 acl attr auto quick 068 auto stress dump +069 auto quick repair 071 rw auto 072 rw auto prealloc quick 073 copy auto
Signature · 1 line
1.8.3.1
Quoted 28 lines · Brian Foster
The secondary superblock verification in
xfs_repairwas subject to a bug that unnecessarily leads to a brute force superblock scan if the last superblock in the fs happens to be corrupt. Normally,xfs_repairhandles one-off superblock corruption gracefully using a heuristic that finds the most consistent superblock content across the set of secondary superblocks.Create a regression test for
xfs_repairthat corrupts the last superblock in the fs. Verify the superblock is updated from the previously verified sb content and a brute force scan is not initiated. In the event of failure, detect that a brute force scan has started and abort the repair in order to fail the test quickly.To support the test, extend the
xfs_repairfilter to handle corrupted superblock repair output and provide generic test output for arbitrary AG counts.Signed-off-by Brian Foster <[email protected]>---
Hi all,
This is an
xfs_repairregression test to trigger the problem fixed by the following previously posted fix:http://oss.sgi.com/archives/xfs/2015-01/msg00244.html
Thoughts appreciated, thanks.
...
tests/xfs/069 line 63
Quoted 11 lines
tests/xfs/069 new file +11
-0+# Start and monitor an xfs_repair of the scratch device. This test can induce a +# time consuming brute force superblock scan. Since a brute force scan means +# test failure, detect it and end the repair. +_xfs_repair_noscan() +{ + # invoke repair directly so we can kill the process if need be + $XFS_REPAIR_PROG $SCRATCH_DEV 2>&1 | tee -a $seqres.full > $tmp.repair & + repair_pid=$! + + # monitor progress for as long as it is running + while [ `ps -q $repair_pid > /dev/null; echo $?` == 0 ]; do
while [ pgrep xfs_repair -eq 0 ]; do
Quoted 27 lines
tests/xfs/069 new file +27
-0+ grep "couldn't verify primary superblock" $tmp.repair \ + > /dev/null 2>&1 + if [ $? == 0 ]; then + # we've started a brute force scan. kill repair and + # fail the test + kill -9 $repair_pid >> $seqres.full 2>&1 + wait >> $seqres.full 2>&1 + + _fail "xfs_repair resorted to brute force scan" + fi + + sleep 1 + done + + wait + + cat $tmp.repair | _filter_repair +} + +rm -f $seqres.full + +# get standard environment, filters and checks +. ./common/rc +. ./common/filter +. ./common/repair + +# real QA test starts here
tests/xfs/069 line 100
Quoted 10 lines
tests/xfs/069 new file +10
-0+ +# Modify as appropriate. +_supported_fs xfs +_supported_os Linux +_require_scratch_nocheck + +_scratch_mkfs >> $seqres.full 2>&1 || _fail "mkfs failed" + +# corrupt the last secondary sb in the fs +agcount=`$XFS_DB_PROG -c "sb" -c "p agcount" $SCRATCH_DEV | awk '{ print $3 }'`
scratch_mkfs | _filter_mkfs 2> $tmp.mkfs
. $tmp.mkfs
And now you have the agcount variable already set up (and most other fs geometry variables that mkfs outputs).
Quoted 2 lines
+last_secondary=$((agcount - 1))
+$XFS_DB_PROG -x -c "sb $last_secondary" -c "type data" \
you can just use "sb $((agcount - 1))" directly. The comment above tells us that it's the last secondary sb we are corrupting....
Otherwise look sok.
Cheers,
Dave.
Signature · 2 lines
Dave Chinner [email protected]