linux-kernel - my experience with "unable to handle kernel paging request" and "nohz=off"

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [thread-next>] [day] [month] [year] [list]

Message-ID: <4EC3CF6E.70607@gmx.net>
Date:	Wed, 16 Nov 2011 15:57:50 +0100
From:	Dirk <noisyb@....net>
To:	LKML <linux-kernel@...r.kernel.org>
Subject: my experience with "unable to handle kernel paging request" and "nohz=off"

Hello,

i am one of those guys who get kernel panics "unable to handle kernel 
paging request" if i don't add nohz=off to the kernel commandline...
(i also added highres=off to it)

this happens mostly when that system (stock IBM Intellistation M PRO but 
with Geforce 7600GT... p4/3.2GHz and 1G RAM) is under heavy load.. 
watching (long) videos with flash online while doing other stuff... but 
not exclusively when watching videos.. everything using xulrunner and 3D 
hardware is guaranteed to cause memleaks and, sooner or later, a kernel 
panic on this machine if nohz=off and highres=off are not set...

i think it started 1 or 2 versions before i stopped using 2.6 kernels 
which was when 3.0 was released... so the actual 2.6 kernel during 3.0 
release minus 2 minor versions or something like that is when it started..

with debugfs i was able to see memleaks when using anything that would 
use the nvidia driver and thought that caused the kernel panic in the 
long run... i did some memtest without problems..

so i switched to nouveau.. the memleaks were gone... but the kernel 
panic still happened under load.. but unlike with the nvidia driver i 
was able to read it now.. because the nouveau driver exited to the 
cmdline before the whole thing died...

blablalba "unable to handle kernel paging request" blablabla...

so i found the tip to set nohz=off and highres=off which i did... no 
kernel panic has occured since then.. but this is the spare machine i 
use to play quake3 so i installed the nvidia driver again since nouveau 
doesn't give me 125fps.. yet..

so kernel panic gone... memleaks back.. every 2nd boot i also get a 
"swapper" memleak but that seems to be unrelated to everything else and 
happens only through boot time...

so.. is there something i should do/test with this old machine to help 
tagging what is going on?

Dirk

Attached files:

kmemleaks (with nohz=off and nvidia driver instead of nouveau... with 
nouveau only the swapper memleaks would be there.. or not..)

config for the current 3.2rc2 kernel running on this machine...

View attachment "config.txt" of type "text/plain" (63090 bytes)

View attachment "kmemleak.txt" of type "text/plain" (11291 bytes)