[LU-15979] writing client eviction due to mds failover Created: 28/Jun/22  Updated: 28/Jun/22

Status: Open
Project: Lustre
Component/s: None
Affects Version/s: None
Fix Version/s: None

Type: Bug Priority: Minor
Reporter: Vladimir Saveliev Assignee: Vladimir Saveliev
Resolution: Unresolved Votes: 0
Labels: None

Severity: 3
Rank (Obsolete): 9223372036854775807

 Description   

This is another scenario for LU-12347 not covered by https://review.whamcloud.com/44151

    cl_io_loop
      cl_io_lock                    <- ldlm lock is taken here
      cl_io_start
        vvp_io_write_start
        ...
          __generic_file_aio_write
            file_remove_privs
              security_inode_need_killpriv
              ...
                ll_xattr_get_common
                ...
                  mdc_intent_lock   <- enqueue rpc is sent here
      cl_io_unlock                  <- ldlm lock is released

    If enqueue rpc is delayed, the client may get evicted as not
    cancelling lock taken in cl_io_lock.


 Comments   
Comment by Gerrit Updater [ 28/Jun/22 ]

"Vladimir Saveliev <vladimir.saveliev@hpe.com>" uploaded a new patch: https://review.whamcloud.com/47820
Subject: LU-15979 llite: call file_remove_privs before taking ldlm
Project: fs/lustre-release
Branch: master
Current Patch Set: 1
Commit: 1d13ea878ebed3985b42d4858a5b7cd7c3eecea8

Generated at Sat Feb 10 03:22:56 UTC 2024 using Jira 9.4.14#940014-sha1:734e6822bbf0d45eff9af51f82432957f73aa32c.