Re: xfs_repair hanging

[Justin Piszcz <jpiszcz@xxxxxxxxxxxxxxx>]

On Sun, 9 Nov 2008, Justin Piszcz wrote:

> p34:~# xfs_repair -V
> xfs_repair version 2.9.8
>
> Hangs with xfs_repair as shown below or with -L option.
>
> p34:~# xfs_repair /dev/md3
> Phase 1 - find and verify superblock...
> Phase 2 - using internal log
>        - zero log...
>        - scan filesystem freespace and inode maps...
> bad magic # 0x58443242 in btbno block 0/1022
> expected level 1 got 2672 in btbno block 0/1022
> bad magic # 0x3f052a50 in btbno block 0/12392
> bad magic # 0 in btbno block 0/36892
> bad magic # 0 in btbno block 0/1241755
> bad magic # 0x2f2a0a20 in btbno block 0/2800
> bad magic # 0x44726976 in btbno block 0/2672
> bad magic # 0x30302d49 in btbno block 0/46
> <hang>
>
> strace of event:
> write(2, "bad magic # 0 in btbno block 0/36"..., 37bad magic # 0 in btbno 
> block 0/36892
> ) = 37
> pread(4, 
> "\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0"..., 
> 4096, 5086228480) = 4096
> write(2, "bad magic # 0 in btbno block 0/12"..., 39bad magic # 0 in btbno 
> block 0/1241755
> ) = 39
> pread(4, "/*\n * arch/arm/mach-loki/include/"..., 4096, 11468800) = 4096
> write(2, "bad magic # 0x2f2a0a20 in btbno b"..., 45bad magic # 0x2f2a0a20 in 
> btbno block 0/2800
> ) = 45
> pread(4, "Driver\n------\n\nInformations about"..., 4096, 10944512) = 4096
> write(2, "bad magic # 0x44726976 in btbno b"..., 45bad magic # 0x44726976 in 
> btbno block 0/2672
> ) = 45
> pread(4, "00-INDEX\n\t- this file (info on so"..., 4096, 188416) = 4096
> write(2, "bad magic # 0x30302d49 in btbno b"..., 43bad magic # 0x30302d49 in 
> btbno block 0/46
> ) = 43
> futex(0x185d098, FUTEX_WAIT_PRIVATE, 2, NULL
>
> I can mount the filesystem but its a bit mangled:
>
> p34:~# cd /r1
> p34:/r1# ls
> ls: cannot access x3: Structure needs cleaning
> ls: cannot access x4: Structure needs cleaning
> ls: cannot access x5: Structure needs cleaning
>
> Should I re-format or is there any possibility to fix the FS?  This was due 
> to some disks going in/out of a RAID5, not XFS' fault but xfs_repair failing 
> to repair seems strange.
>
> Justin.

[  215.558117] Filesystem "md3": Disabling barriers, trial barrier write failed
[  215.558231] XFS mounting filesystem md3
[  215.645960] Ending clean XFS mount for filesystem: md3
[  218.449389] 00000000: ac b3 63 97 3c 65 20 d1 8a 7c dd a7 93 9c f1 9e  ..c.<e
 ..|......
[  218.449399] Filesystem "md3": XFS internal error xfs_da_do_buf(2) at line 210
7 of file fs/xfs/xfs_da_btree.c.  Caller 0xffffffff80397924
[  218.449404] Pid: 5107, comm: find Not tainted 2.6.27.5 #1
[  218.449406] 
[  218.449407] Call Trace:
[  218.449414]  [<ffffffff80397924>] xfs_da_read_buf+0x24/0x30
[  218.449417]  [<ffffffff80397879>] xfs_da_do_buf+0x6f9/0x730
[  218.449421]  [<ffffffff80397924>] xfs_da_read_buf+0x24/0x30
[  218.449425]  [<ffffffff803edfa0>] _atomic_dec_and_lock+0x20/0x90
[  218.449429]  [<ffffffff80397924>] xfs_da_read_buf+0x24/0x30
[  218.449433]  [<ffffffff8039c083>] xfs_dir2_block_getdents+0xa3/0x210
[  218.449436]  [<ffffffff8039c083>] xfs_dir2_block_getdents+0xa3/0x210
[  218.449441]  [<ffffffff80299ee2>] __dentry_open+0xc2/0x280
[  218.449444]  [<ffffffff803d1fe0>] xfs_hack_filldir+0x0/0x60
[  218.449448]  [<ffffffff803d1fe0>] xfs_hack_filldir+0x0/0x60
[  218.449451]  [<ffffffff8039aa18>] xfs_readdir+0x98/0xf0
[  218.449454]  [<ffffffff803d1f12>] xfs_file_readdir+0xe2/0x1b0
[  218.449457]  [<ffffffff802a99e0>] filldir+0x0/0xe0
[  218.449460]  [<ffffffff802a99e0>] filldir+0x0/0xe0
[  218.449462]  [<ffffffff802a9c60>] vfs_readdir+0xc0/0xe0
[  218.449465]  [<ffffffff802a9de4>] sys_getdents+0x84/0xe0
[  218.449468]  [<ffffffff8029a97a>] sys_fchdir+0x4a/0x90
[  218.449472]  [<ffffffff80226d2b>] system_call_fastpath+0x16/0x1b
[  218.449474] 
[  239.765314] Filesystem "md3": Disabling barriers, trial barrier write failed
[  239.765538] XFS mounting filesystem md3
[  239.957389] Ending clean XFS mount for filesystem: md3

Looks like its stuck on a few bad files?  Is there anyway to use xfs_* to
delete the files its having problems on or should I just wipe it?

Justin.