RHS is recommending "-n 8k" for larger dir block size at mkfs time.
However, there are 2 bugs in xfs_repair if those 2 blocks aren't allocated contiguously:
9fe1cc862159b50651baa68a6b60ae0819d4f871 xfs_repair: handle fragmented multiblock dir2 in process_leaf_node_dir2()
b220548bf2e51f4a573b19b4a8b4b15856bd4a06 xfs_repair: Fix fragmented multiblock dir2 handling in blkmap_getn()
We need to get these backported to RHEL6's xfsprogs to handle this case.
This is tested in xfstests:
2a4ed91a018054c2ae187f3cd4ee26141f56234f xfstest 291: userspace vs. fragmented multiblock dir2
thx, added perf dept to cc
Since the problem described in this bug report should be
resolved in a recent advisory, it has been closed with a
resolution of ERRATA.
For information on the advisory, and where to find the updated
files, follow the link below.
If the solution does not work for you, open a new bug report.