Bugzilla – Bug 277
xfs_force_shutdown after kernel bug
Last modified: 2004-02-16 07:32:30 CST
You need to log in before you can comment on or make changes to this bug.
I have NFS serving XFS on top of softraid md1. Linux Distribution is Redhat 8.0, kernel is version 2.4.19, XFS is version 1.2. After running this combination for about half a year without problems, I couldn't access the directory any more. 'ls' tells me it's empty, 'mount' tells me it's still mounted. Can't umount (busy), and can't shut down NFS. Output of 'dmesg' follows: xfs_iget_core: ambiguous vns: vp/0xd6ffe080, invp/0xd6ffe480 kernel BUG at debug.c:96! invalid operand: 0000 CPU: 0 EIP: 0010:[<c0215b73>] Not tainted EFLAGS: 00010246 eax: 00000040 ebx: 00000000 ecx: dec22000 edx: 00000000 esi: 33fe2c1f edi: 00000000 ebp: db37cdf0 esp: dec23c70 ds: 0018 es: 0018 ss: 0018 Process nfsd (pid: 624, stackpage=dec23000) Stack: c030ba49 c030ba21 c03bbcc0 d6ffe49c c01e54ae 00000000 c0315a80 d6ffe080 d6ffe480 dff93688 df879c00 33fe2c1f c014d6cb df879c00 df592cf8 00000000 00000000 db37cdf0 d6ffe49c d6ffe480 33fe2c1f 00000000 c01e56f7 d6ffe480 Call Trace: [<c01e54ae>] [<c014d6cb>] [<c01e56f7>] [<c0200909>] [<c021535e>] [<c018da86>] [<c018de46>] [<c018e262>] [<c02cf1b1>] [<c018f930>] [<c018feff>] [<c02d61b2>] [<c029b875>] [<c01170d3>] [<c0195eba>] [<c018b92e>] [<c02efcdf>] [<c018b71b>] [<c010733e>] [<c018b560>] Code: 0f 0b 60 00 41 ba 30 c0 8b 5c 24 0c 83 c4 10 c3 90 8d b6 00 <5>xfs_force_shutdown(md(9,7),0x8) called from line 1042 of file xfs_trans.c. Return address = 0xc01fc208 Corruption of in-memory data detected. Shutting down filesystem: md(9,7) Please umount the filesystem, and rectify the problem(s)
One correction of my original post: The partition in question is build on top of software raid level 0 (striped), not 1.
umount the volume if you can, or reboot the machine. Then run xfs_check to see if the FS is ok, if not run xfs_repair. Note if forced to reboot the machine mount the file system first (that will run recovery) and then unmount the FS before running xfs_check
Looks like there were no follups to this. Is this reproduible?
No, happened just once in about a year of productive use. No idea what triggered it.
Marking REMIND as it's absolutely not reproducible