[#LU-11904] jobid cache cleanup is not safe

[LU-11904] jobid cache cleanup is not safe Created: 30/Jan/19 Updated: 13/May/19
Status:	Open
Project:	Lustre
Component/s:	None
Affects Version/s:	None
Fix Version/s:	None

Type:

Bug

Priority:

Minor

Reporter:

Patrick Farrell (Inactive)

Assignee:

WC Triage

Resolution:

Unresolved

Votes:

Labels:

None

Severity:

Rank (Obsolete):

9223372036854775807

Description

jobid_cache_fini uses the jobid_hash_lock to protect the jobid_hash pointer:

spin_lock(&jobid_hash_lock);
tmp_hash = jobid_hash;
jobid_hash = NULL;
spin_unlock(&jobid_hash_lock);

But other places in the code use that pointer without holding the jobid_hash_lock (notably jobid_get_from_cache()).

This can cause unexpected behavior in those locations during unmount.

No plan to do a patch for this ATM.

Comments

Comment by Patrick Farrell (Inactive) [ 30/Jan/19 ]

One point:
I wonder if we should be allowed to get here while there's still activity on the mount. If not, then this isn't a bug - just some unnecessary locking.

Comment by Chris Hunter (Inactive) [ 13/May/19 ]

Related to ~~LU-12225~~ ?

Generated at Sat Feb 10 02:47:56 UTC 2024 using Jira 9.4.14#940014-sha1:734e6822bbf0d45eff9af51f82432957f73aa32c.

[LU-11904] jobid cache cleanup is not safe Created: 30/Jan/19 Updated: 13/May/19