Bug 1303979

Summary: OOM with kernel 4.3.x (regression)
Product: [Fedora] Fedora Reporter: Stuart D Gathman <stuart>
Component: kernelAssignee: Kernel Maintainer List <kernel-maint>
Status: CLOSED DUPLICATE QA Contact: Fedora Extras Quality Assurance <extras-qa>
Severity: unspecified Docs Contact:
Priority: unspecified    
Version: 22CC: gansalmon, itamar, jonathan, kernel-maint, madhu.chinakonda, mchehab
Target Milestone: ---   
Target Release: ---   
Hardware: Unspecified   
OS: Unspecified   
Whiteboard:
Fixed In Version: Doc Type: Bug Fix
Doc Text:
Story Points: ---
Clone Of: Environment:
Last Closed: 2016-02-03 08:16:37 UTC Type: Bug
Regression: --- Mount Type: ---
Documentation: --- CRM:
Verified Versions: Category: ---
oVirt Team: --- RHEL 7.3 requirements from Atomic Host:
Cloudforms Team: --- Target Upstream Version:
Attachments:
Description Flags
oom log
none
systemd-jrournal gets the axe none

Description Stuart D Gathman 2016-02-02 15:34:16 UTC 
Description of problem:
When booted from 4.3 kernels, laptop starts killing process for OOM within 8 hours.

Version-Release number of selected component (if applicable):
kernel-4.3.4-200.fc22.i686

How reproducible:
Always, but no hard timeline.

Steps to Reproduce:
1. Boot with 4.3.4 (hang on to last working kernel-4.2.8)
2. Do stuff.  Network seems important
3.

Actual results:
OOM, with userspace memory only 50% used.

Expected results:
Same stable operation as 4.2.8 kernel.

Additional info:
Adding a USB wifi dongle (for a total of 2 connected wifi nets) seemed to cut time to OOM in half.  But this was subjective and squishy.  Haven't done it often enough to quantify.

This is different from https://bugzilla.redhat.com/show_bug.cgi?id=1303270 in that SUnreclaim is not increasing without bound.  However, it is similar in that user space processes don't seem to be using the memory. (And in 4.3 kernel being generally borked.)  I'll be adding logs and another regression for intel video.
Comment 1 Stuart D Gathman 2016-02-02 16:11:55 UTC 
Created attachment 1120490 [details]
oom log

This may be related to bug#1303990, as video was "jumpy" prior to the OOMs, and firefox was using more memory than for the same set of tabs with 4.2.8.  Thunderbird is always a pig, but the OOMs follow 4.3 kernels.
Comment 2 Stuart D Gathman 2016-02-02 16:18:01 UTC 
Note, while the OOM handler starts with killing thunderbird and firefox (the biggest pigs), it continues killing processes, including mate-panel, various applets, etc, until the desktop is empty!  It only seems to kill processes using X, however.  Which is why I think bug#1303990 might be related.
Comment 3 Stuart D Gathman 2016-02-02 16:27:08 UTC 
Created attachment 1120497 [details]
systemd-jrournal gets the axe
Comment 4 Josh Boyer 2016-02-03 08:16:37 UTC 

*** This bug has been marked as a duplicate of bug 1303270 ***