This fix adds more checks to xfs_repair for corruption seen on a
filesystem at LANL. The filesystem had an inode whose extent B+Tree
was corrupted; the corruption was not detected by xfs_repair, however
it prevented the file from being accessed (a 'cp' would report the
error "Filesystem is corrupted") or removed (a dirty transaciton would
be cancelled causing the filesystem to be shut down).
Date: Wed Aug 16 12:19:23 PDT 2000
Workarea: jen.americas.sgi.com:/src/lord/xfs-linux.2.4.0-test5
The following file(s) were checked into:
bonnie.engr.sgi.com:/isms/slinx/2.4.0-test1-xfs
Modid: 2.4.0-test1-xfs:slinx:72201a
cmd/xfs/repair/dinode.c - 1.80
- Add a check to process_btinode to verifty that B+Tree forward sibling
block numbers are NULL in the rightmost B+Tree block. This check
catches only problems in the highest level B+Tree blocks.
cmd/xfs/repair/scan.c - 1.43
- Add a B+Tree left sibling check to scanfunc_bmap. The left sibling
block number must be NULL in the leftmost B+Tree block.
Also add to scanfunc_bmap the same check that was added to
process_dinode_int to verifty that B+Tree forward sibling block
numbers are NULL in the rightmost B+Tree block. This check catches
problems in all but the highest B+Tree level.
|