i'm running 2.4.30 with linux-2.4.22-VFS-lock.patch (from EVMS sources)
and using lvm1 snapshots -- without doing xfs_freeze.  the snapshots are
mounted -o ro.

last night i got this while a rdiff-backup was running on a snapshot:

XFS internal error XFS_WANT_CORRUPTED_GOTO at line 4340 of file xfs_bmap.c. 
Caller 0xc01c5157
ea2e3b2c c019e03d c030a27e 00000001 00000000 c030a39d 000010f4 c01c5157
       dd0bc670 00000068 00004683 00000000 00000246 00000000 c03bd584 00000004
       c876c268 dffb5680 00000000 ea2e2000 cb7a1c00 00000000 c876c268 00000001
Call Trace:    [xfs_bmap_read_extents+941/1184] [xfs_iread_extents+119/272]
[xfs_iread_extents+119/272] [xfs_bmapi+516/5600]
[_pagebuf_iodone+115/160]
  [locate_hd_struct+56/160] [__scsi_end_request+424/560]
[scsi_io_completion+353/1040] [rw_intr+90/464]
[locate_hd_struct+56/160] [account_io_start+56/96]
  [req_new_io+103/144] [xfs_iomap+410/1344] [<f8dba2bf>]
[linvfs_get_block_core+135/672] [ip_local_deliver_finish+0/368]
[get_unused_buffer_head+93/192]
  [create_buffers+107/256] [linvfs_get_block+54/64]
[block_read_full_page+635/704] [do_generic_file_read+702/1168]
[linvfs_get_block+0/64] [xfs_iunlock+62/128]
  [file_read_actor+0/256] [file_read_actor+0/256] [generic_file_read+183/432]
[file_read_actor+0/256] [__vma_link+86/192]
[xfs_read+205/752]
  [linvfs_read+93/224] [sys_read+153/272] [system_call+51/56]

i can't tell if the above was the snapshot filesystem or not, is the
device node hidden in that debug output somewhere?  none of my filesystems
went down at this point -- the backup continued for 3 more minutes without
error, and this error is about 7 minutes into the backup.

this hardware has been rock solid for over a year now, i just switched to
xfs last weekend.  i have no reason to suspect hardware.

-dean