[LU-8801] ldiskfs quota accounting iterator is flawed

• Type: Bug
• Resolution: Fixed
• Priority: Minor
• Fix Version/s: Lustre 2.10.0
• Severity: 3
• Assignee: Niu Yawei (Inactive)
• Reporter: Niu Yawei (Inactive)

Description

The ldiskfs quota accounting iterator is flawed, that'll result in some accounting entries being missed in the output of proc files (acct_user & acct_group).

This can be reproduced by:

• Create a fresh filesystem;
• Run the following script to generate large amount of accounting entries:

  #!/bin/bash
  for i in `seq 10000`; do
          touch /mnt/lustre/file_$i
          chown $i:$i /mnt/lustre/file_$i
  done
  

• Verify proc file by: cat /proc/fs/lustre/osd-ldiskfs/lustre-MDT0000/quota_slave/acct_user; Some entries are missed in the output.

Comments

[Peter Jones]

Landed for 2.10

[Gerrit Updater]

Oleg Drokin ([e-mail]) merged in patch https://review.whamcloud.com/23583/
Subject: LU-8801 quota: fix ldiskfs accounting iterator
Project: fs/lustre-release
Branch: master
Current Patch Set:
Commit: 623b315f6ddf8aa6ef435aa405e371ba93d0454a

[Niu Yawei (Inactive)]

Another problem is that reading acct_user/group will become extremely slow when the amount of id entries is large, because in each lprocfs_quota_seq_start(), we use it->next() to move the cursor to current the last processed position like this:

       while (offset--) {
               rc = iops->next(&lqp->lqp_env, it);
               if (rc != 0) /* Error or reach the end */
                       goto not_found;
       }

Which is quite time consuming when the accounting file getting large. I'll optimize it with above fix together.

[Gerrit Updater]

Niu Yawei ([e-mail]) uploaded a new patch: http://review.whamcloud.com/23583
Subject: LU-8801 quota: fix ldiskfs accounting iterator
Project: fs/lustre-release
Branch: master
Current Patch Set: 1
Commit: bcb301aad574beed8022b1d34dc2bab02c64ca5b