[LU-5734] LNet dynamic control: lnet_dyn_add_ni() can't clean up failed NI in some cases

• Type: Bug
• Resolution: Fixed
• Priority: Blocker
• Fix Version/s: Lustre 2.7.0
• Affects Version/s: Lustre 2.7.0
• Severity: 3
• Assignee: Amir Shehata (Inactive)
• Reporter: Isaac Huang (Inactive)

Description

In lnet_dyn_add_ni(), if lnet_startup_lndnis() fails, then NI on the local &net_head list is freed. But lnet_startup_lndnis(&net_head) can fail after list_del(&ni->ni_list), i.e. after the NI has been removed already from the &net_head. In this case I failed to see where the NI gets cleaned up and freed.

Issue Links

• is related to: LU-5849 DLC: parse networks function needs to return the number of parsed networks
• is related to: LU-6002 DLC: startup acceptor dynamically.
• is related to: LU-5839 LNetNIInit() should not call lnet_destroy_routes() before lnet_parse_routes() has been called
• is related to: LU-5850 DLC: lnet_startup_lndnis should clean up after itself
• is related to: LU-5568 kernel crash when when network initialization failed
• is related to: LU-2456 Dynamic LNet Config Main Development Work

Comments

[Jodi Levi (Inactive)]

Patch landed to Master.

[Gerrit Updater]

Oleg Drokin ([e-mail]) merged in patch http://review.whamcloud.com/12658/
Subject: LU-5734 lnet: improve clean up code and API
Project: fs/lustre-release
Branch: master
Current Patch Set:
Commit: c9501b87d0e06c36b180b80c08ca79b672f20c72

[Wang Shilong (Inactive)]

Amir Shehata just send this patch to fix it:

http://review.whamcloud.com/#/c/12658/

Best Regards,
Wang Shilong

[Wang Shilong (Inactive)]

Hi Andreas Dilger,

Previously, the fix is together with my patch, Now Amir is giving a quick fix for LU-5568.
I am going to separate enhanced patch after Amir's patch which should include this fix!

Best Regards,
Wang Shilong

[Andreas Dilger]

Wang, are you working on a patch for this issue?

[Isaac Huang (Inactive)]

Another related issue LU-5839, that should probably be fixed together.

[Wang Shilong (Inactive)]

Hi lsaac Huang,

I will try to fix this issue together with LU-5568.

Thanks very much for your comments and help!
Best regards,
Wang Shilong

[Isaac Huang (Inactive)]

It should be fixed together with LU-5568 - they are closely related.