2 # SPDX-License-Identifier: GPL-2.0
3 # Copyright (C) 2015 SUSE Linux Products GmbH. All Rights Reserved.
7 # This test is motivated by an fsync issue discovered in btrfs.
8 # The issue in btrfs was that adding a new hard link to an inode that already
9 # had a large number of hardlinks and fsync the inode, would make the fsync
10 # log replay code update the inode with a wrong link count (smaller than the
11 # correct value). This resulted later in dangling directory index entries,
12 # after removing most of the hard links (correct_value - wrong_value), that
13 # were visible to user space but it was impossible to delete them or do
14 # any other operation on them (since they pointed to an inode that didn't
15 # exist anymore, resulting in -ESTALE errors).
17 # The btrfs issue was fixed by the following linux kernel patch:
19 # Btrfs: fix fsync when extend references are added to an inode
21 # This issue was present in btrfs since the extrefs (extend references)
22 # feature was added (2012).
25 seqres=$RESULT_DIR/$seq
26 echo "QA output created by $seq"
30 status=1 # failure is the default!
36 trap "_cleanup; exit \$status" 0 1 2 3 15
38 # get standard environment, filters and checks
43 # real QA test starts here
48 _require_dm_target flakey
52 # If the test filesystem is btrfs, make sure we create a filesystem with
53 # the extend references (extrefs) feature enabled (it's enabled by default
54 # in recent versions of btrfs-progs).
55 if [ "$FSTYP" = "btrfs" ]; then
56 _scratch_mkfs "-O extref" >> $seqres.full 2>&1
58 _scratch_mkfs >> $seqres.full 2>&1
61 _require_metadata_journaling $SCRATCH_DEV
65 # Create a test file with 3001 hard links. This number is large enough to
66 # make btrfs start using extrefs at some point even if the fs has the maximum
67 # possible leaf/node size (64Kb).
68 echo "hello world" > $SCRATCH_MNT/foo
69 for i in `seq 1 3000`; do
70 ln $SCRATCH_MNT/foo $SCRATCH_MNT/foo_link_`printf "%04d" $i`
73 # Make sure all metadata and data are durably persisted.
76 # Add one more link to the inode that ends up being a btrfs extref and fsync
78 ln $SCRATCH_MNT/foo $SCRATCH_MNT/foo_link_3001
79 $XFS_IO_PROG -c "fsync" $SCRATCH_MNT/foo
81 _flakey_drop_and_remount
83 # Now after the fsync log replay btrfs left our inode with a wrong link count N,
84 # which was smaller than the correct link count M (N < M).
85 # So after removing N hard links, the remaining M - N directory entries were
86 # still visible to user space but it was impossible to do anything with them
87 # because they pointed to an inode that didn't exist anymore. This resulted in
88 # stale file handle errors (-ESTALE) when accessing those dentries for example.
90 # So remove all hard links except the first one and then attempt to read the
91 # file, to verify we don't get an -ESTALE error when accessing the inode.
93 # The btrfs fsck tool also detected the incorrect inode link count and it
94 # reported an error message like the following:
96 # root 5 inode 257 errors 2001, no inode item, link count wrong
97 # unresolved ref dir 256 index 2978 namelen 13 name foo_link_2976 filetype 1 errors 4, no inode ref
99 # The fstests framework automatically calls fsck after a test is run, so we
100 # don't need to call fsck explicitly here.
102 echo "Link count before rm foo_link_*: $(stat -c %h $SCRATCH_MNT/foo)"
103 rm -f $SCRATCH_MNT/foo_link_*
104 echo "Link count after rm foo_link_*: $(stat -c %h $SCRATCH_MNT/foo)"