[LU-7441] Memory leak in nrs_tbf_*_startup

• Type: Bug
• Resolution: Fixed
• Priority: Major
• Fix Version/s: Lustre 2.10.0
• Affects Version/s: Lustre 2.8.0
• Labels: patch
• Severity: 3
• Assignee: Emoly Liu
• Reporter: Li Xi (Inactive)

Description

nrs_tbf_*_startup should free hash table if futher process fails.

Issue Links

• is related to: LU-9750 misc code cleanups in nrs policy code

Comments

[Joseph Gmitter (Inactive)]

Issue here is resolved and landed for 2.10.0. The code cleanup patch is moved to a separate ticket: LU-9750.

[Gerrit Updater]

Oleg Drokin ([e-mail]) merged in patch https://review.whamcloud.com/17224/
Subject: LU-7441 nrs: Free hash table if failed to start a nrs policy
Project: fs/lustre-release
Branch: master
Current Patch Set:
Commit: cd362fa9186a3e4de34c7c68908e6d3d429bb087

[Gerrit Updater]

Emoly Liu ([e-mail]) uploaded a new patch: https://review.whamcloud.com/25319
Subject: LU-7441 nrs: some code cleanup in NRS policies
Project: fs/lustre-release
Branch: master
Current Patch Set: 1
Commit: 740df75fe6952bada2b505a7bd751aff09a07e94

[Li Xi (Inactive)]

I don't think it is NRS state setting problem. If all funcitons of op_policy_start() always free all memory they allocats when failure happends, no memory will be leaked. And I checked all nrs_*_start functions. I didn't find any problem.

[Emoly Liu]

Lixi, thanks for your reproducer!

As we discussed, this problem exists in all NRS policies code, not only TBF, so I think it's related to NRS state setting. Your patch does work, but we need to know the root cause and then fix it.

[Li Xi (Inactive)]

This problem is easy to produce:

With following patch:

memset(&start, 0, sizeof(start));
start.tc_jobids_str = "*";

start.tc_rpc_rate = tbf_rate;
start.tc_rule_flags = NTRS_DEFAULT;
start.tc_name = NRS_TBF_DEFAULT_RULE;
INIT_LIST_HEAD(&start.tc_jobids);
//rc = nrs_tbf_rule_start(policy, head, &start);
rc = -EINVAL;
return rc;

[root@QYJ tests]# lctl set_param ost.OSS.ost_io.nrs_policies="tbf jobid"
ost.OSS.ost_io.nrs_policies=tbf jobid
error: set_param: setting /proc/fs/lustre/ost/OSS/ost_io/nrs_policies=tbf jobid: Invalid argument
[root@QYJ tests]# sh llmountcleanup.sh
Stopping clients: QYJ /mnt/lustre (opts:-f)
Stopping client QYJ /mnt/lustre opts:-f
Stopping clients: QYJ /mnt/lustre2 (opts:-f)
Stopping /mnt/mds1 (opts:-f) on QYJ
Stopping /mnt/ost1 (opts:-f) on QYJ
Stopping /mnt/ost2 (opts:-f) on QYJ

LNetError: 19587:0:(module.c:412:exit_libcfs_module()) Portals memory leaked: 131600 bytes
Memory leaks detected

[Peter Jones]

Emoly

Could you please take care of this patch?

Thanks

Peter

[Gerrit Updater]

Li Xi ([e-mail]) uploaded a new patch: http://review.whamcloud.com/17224
Subject: LU-7441 nrs: fix memory leak in nrs_tbf_*_startup
Project: fs/lustre-release
Branch: master
Current Patch Set: 1
Commit: f1b173201d137a8acd924084d03a76d2bdcab0a1