Bug 105671 - CLONE_FS then CLONE_VFORK can deadlock distributed FS CEPH & OCFS2
Summary: CLONE_FS then CLONE_VFORK can deadlock distributed FS CEPH & OCFS2
Status: NEW
Alias: None
Product: Process Management
Classification: Unclassified
Component: Other (show other bugs)
Hardware: All Linux
: P1 normal
Assignee: process_other
URL:
Keywords:
Depends on:
Blocks:
 
Reported: 2015-10-08 10:19 UTC by Dzianis Kahanovich
Modified: 2015-10-08 10:20 UTC (History)
0 users

See Also:
Kernel Version: 4.2.3
Subsystem:
Regression: No
Bisected commit-id:

Attachments
Add an attachment (proposed patch, testcase, etc.)

Description Dzianis Kahanovich 2015-10-08 10:19:30 UTC 
Cannot confirm now, but forking (clone()) process with CLONE_FS then forking CLONE_VFORK deadlock distributed FS CEPH & OCFS2. Moving to PREEMPT & OCFS2 side - this is eventially deadlock (on short VM freezes as rbd snapshotting), simple PREEMPT_NONE heavy locked asap. Experienced by own changes to MPM-ITK, now simple remove both. Think single one (prefer CLONE_VFORK) must be good, but not now.

If somebody believe or confirm me, suggest to check not only CLONE_FS, but other data sharing flags and at least bit-compare previous & current clone mode & deny lockable combinations. I will look to code after relax.

Motivation to publish this exotic combination - IMHO (FIXME) can be server (or even cluster, as ceph mds daemons affected by long locks too) bomb from user side.

PS 4.1.7 probably too.
Comment 1 Dzianis Kahanovich 2015-10-08 10:20:36 UTC 
fix: *simple PREEMPT_NONE + CEPHFS from vm
Note You need to log in before you can comment on or make changes to this bug.