Project

General

Profile

Actions

Bug #23966

closed

Deleting a pool with active notify linger ops can result in seg fault

Added by Jason Dillaman about 6 years ago. Updated almost 6 years ago.

Status:
Resolved
Priority:
Normal
Assignee:
Jason Dillaman
Category:
-
Target version:
-
% Done:

0%

Source:
Tags:
Backport:
luminous,jewel
Regression:
No
Severity:
3 - minor
Reviewed:
Affected Versions:
ceph-qa-suite:
Component(RADOS):
Pull request ID:
Crash signature (v1):
Crash signature (v2):

Description

It's possible that if a notification is sent while a pool is being deleted, the Objecter will fail the Op w/ -ENOENT when it detects the pool deletion. However, the OSDs might not have seen the new map yet and might accept the op and issue the MWatchNotify op back to the client. Since the client doesn't expect to receive a MWatchNotify on a failed Op, it had already completed the 'on_notify_finish' callback but didn't reset the pointer to NULL.

http://qa-proxy.ceph.com/teuthology/trociny-2018-05-02_11:06:00-rbd-wip-mgolub-testing-distro-basic-smithi/2466798/teuthology.log


Related issues 2 (0 open2 closed)

Copied to RADOS - Backport #24058: jewel: Deleting a pool with active notify linger ops can result in seg faultResolvedKefu ChaiActions
Copied to RADOS - Backport #24059: luminous: Deleting a pool with active notify linger ops can result in seg faultResolvedPrashant DActions
Actions

Also available in: Atom PDF