[LU-4603] NFS reexport leads to problems of "ls"

• Type: Bug
• Resolution: Fixed
• Priority: Blocker
• Fix Version/s: Lustre 2.6.0
• Affects Version/s: Lustre 2.6.0
• Labels: HB patch
• Severity: 3
• Assignee: Di Wang (Inactive)
• Reporter: Li Xi (Inactive)

Description

When "ls" under a NFS directory, following error message showes up.

ls: reading directory .: Too many levels of symbolic links

And what is more, if the dirctory is large, (e.g. has more than 23 files), there will be some files missing when doing "ls".

[root@vm1 nfs]# ls * | wc -l
97
[root@vm1 nfs]# ls -l | wc -l
ls: reading directory .: Too many levels of symbolic links
98
[root@server1 113_nfs]# ls /mnt/lustre/nfs/* | wc -l
100

We've confirmed that these problems is introduced by 4e0c8ae(LU-3531 llite: move dir cache to MDC layer). A patch which fixes these problems will be pushed soon.

Issue Links

• is related to: LU-4869 rmdir allowed on nonempty striped directories
• is related to: LU-4945 req_capsule_get: Wrong buffer for field `name' (5 of 6) in format `LDLM_INTENT_GETATTR': 3 vs. 0 (client)
• is related to: LU-4601 Failure on test suite parallel-scale-nfsv3 test_compilebench
• is related to: LU-3531 DNE2: striped directory

Comments

[Jodi Levi (Inactive)]

Patch landed to Master and fix confirmed in last comment. Please reopen if more work is needed in this ticket.

[Li Xi (Inactive)]

Hi Oleg,

I've confirmed that the newest patch set (Patch Set 28) fixes my problem. Thanks for reminding.

[Oleg Drokin]

Li Xi: I wonder if you tried the revised patch to ensure it still fixes your original report? Can you give it a try please?