I witnessed xfs_force_shutdown doing its best yesterday ;-)
I was trying to untarbz2 little damaged .tbz2 archive
(xaa+xab+xad+xac .... wrongfully connected files after split -b).
Since it occured on rootfs I copied output manually.
xfs_force_shutdown (ide0(3,1), 0x8) called from line 1020 of file
Return address = 0xc01b25b9 (which is xfs_trans_cancel+69 on this kernel)
Fatal error on root filesystem
Corruption of in-memory data detected.
Shutting down filesystem ide0(3,1)
Please umount filesystem, and rectify the problems(s)
Then i did all things as usual - Alt+SysRQ+UUUB :)
After reboot f/s could NOT be remounted r/w - same things
So i booted from CD and repaired f/s. The only thing that may be
interesting (during f/s repair) was:
in_freecount/free mismatch, inode chunk 6/214624, freecount 1 nfree 0
After next reboot everything was absolutely correct. I tried to reproduce
it but i didn't manage to....
Machine was up not more than an hour, i think.
It was quite idle during that time.
Kernel came from cvs checkout on 12 September 2001.
No highmem, no lvm, md as module but totally unused.
I would exclude memory corruption - I've been using it for
quite a long time.. I'll memtest86 it out today, however.
UP Celeron 466 Mendocino / 96MB / VIA82C596 (southbridge) - nothing special
2GB WDC drive, one xfs partition (rootfs). about 130MB of swap.
That was the only occurence of xfs_force_shutdown during my journey with xfs.
I'll provide more info if needed.