nV News Forums

 
 

nV News Forums (http://www.nvnews.net/vbulletin/index.php)
-   NVIDIA Linux (http://www.nvnews.net/vbulletin/forumdisplay.php?f=14)
-   -   kvm windows 7 guest segfaults under centos 5.7 with 285.05.09 driver from elrepo (http://www.nvnews.net/vbulletin/showthread.php?t=167271)

q7joey 10-11-11 03:12 PM

kvm windows 7 guest segfaults under centos 5.7 with 285.05.09 driver from elrepo
 
basically, kvm guests would randomly segfault even when no console was being displayed. i normally use rdp to talk to my windows guest anyway, and i use vnc as the console, so very little activity should be going through the nvidia libraries.

i'm not going to be able to give much more detail since i had to drop back to 280.13 in order to keep my windows guest stable, but i was able to get the following stack trace from one of the segfaults.

#0 0x0000003a4a200a08 in _nv007tls ()
from /usr/lib64/nvidia/tls/libnvidia-tls.so.285.05.09
#1 0x00000030d4e05ad9 in __nptl_deallocate_tsd () from
/lib64/libpthread.so.0
#2 0x00000030d4e0674b in start_thread () from /lib64/libpthread.so.0
#3 0x00000030d42d44bd in clone () from /lib64/libc.so.6

and that address has been consistent through numerous segfaults:

Oct 10 08:15:16 charon kernel: qemu-kvm[25987]: segfault at 00000000000140a0 rip 0000003a4a200a08 rsp 00000000462d7080 error 4
Oct 10 10:33:34 charon kernel: qemu-kvm[26961]: segfault at 0000000000000088 rip 0000003a4a200a08 rsp 000000005f996080 error 4
Oct 10 11:34:14 charon kernel: qemu-kvm[27386]: segfault at 0000000000000001 rip 0000003a4a200a08 rsp 000000005a6fc080 error 4
Oct 11 08:09:17 charon kernel: qemu-kvm[32216]: segfault at 0000000000000001 rip 0000003a4a200a08 rsp 0000000044afa080 error 4
Oct 11 10:32:22 charon kernel: qemu-kvm[22121]: segfault at 0000000000000001 rip 0000003a4a200a08 rsp 0000000058c9c080 error 4

let me know if sending the nvidia-bug-report data would be useful given that i've gone back to the previous release.

sandipt 10-12-11 02:31 AM

Re: kvm windows 7 guest segfaults under centos 5.7 with 285.05.09 driver from elrepo
 
Please describe your hardware setup and reproduction steps, also attach nvidia bug report.

q7joey 10-12-11 10:22 AM

Re: kvm windows 7 guest segfaults under centos 5.7 with 285.05.09 driver from elrepo
 
1 Attachment(s)
i've attached the nvidia-bug-report file. (btw, the nvidia-bug-report script needs quotes around [:digit:] in the 'tr' invocation, otherwise it matches files like t).

i don't know what other config info to give that isn't in the bug-report file.

reproduction was just running my windows 7 kvm guest. after a while (even with no interaction on my part), qemu-kvm would segfault. one other piece of info is that a linux centos 6 guest did not trigger the bug.

sandipt 10-20-11 12:40 AM

Re: kvm windows 7 guest segfaults under centos 5.7 with 285.05.09 driver from elrepo
 
Today I installed CentOS 5.7 x64 with Win 7 x64 kvm guest + G96, win7 is running since 24 hours without any issue. Both centos and win7 are fully updated.

q7joey 10-21-11 10:51 AM

Re: kvm windows 7 guest segfaults under centos 5.7 with 285.05.09 driver from elrepo
 
one other piece that could be related is that i do have the redhat kvm virtio drivers loaded in windows 7. i can't really tell what version things are, so i'm updating them to the latest from http://alt.fedoraproject.org/pub/alt...st/images/bin/ and i'll try updating the nvidia drivers again. it seems unlikely that they would be related, but it's worth a try.

q7joey 10-21-11 05:07 PM

Re: kvm windows 7 guest segfaults under centos 5.7 with 285.05.09 driver from elrepo
 
i just updated to the newest centos kernel (2.6.18-274.7.1.el5) and have the latest virtio drivers from fedora (virtio-win-0.1-15.iso) and my win7 virtual died after a few hours, and i wasn't even using it. the crash message is at a different location this time, but it sure seems to be the same situation as before:

qemu-kvm[7416]: segfault at 000000000000000e rip 00002aac0d0d1a08 rsp 0000000059137080 error 4

gasmantwo 10-27-11 09:46 AM

Re: kvm windows 7 guest segfaults under centos 5.7 with 285.05.09 driver from elrepo
 
Hi,

Just jumping in here because this is the only thread that is similar to a problem I recently experienced with qemu-kvm on CentOS 5.7 x86_64 host (2.6.18-274.7.1.el5) running Nvidia driver version 285.05.09.

I was getting segfaults with CentOS 6.0 x86_64 guests only. Windows based guests were not having any problems. I used gdb on the core dump from the last segfault.

Core was generated by `/usr/libexec/qemu-kvm -S -M rhel5.4.0 -m 2048 -smp 1,sockets=1,cores=1,threads='.
Program terminated with signal 11, Segmentation fault.
#0 0x0000003e5fe00a08 in _nv007tls ()
from /usr/lib64/tls/libnvidia-tls.so.285.05.09

The segfaults stopped happening when I rolled back to Nvidia driver version 275.28. I will try again and run the nvidia-bug-report.sh script when I get a moment.

q7joey 11-02-11 01:47 PM

Re: kvm windows 7 guest segfaults under centos 5.7 with 285.05.09 driver from elrepo
 
this may have been a hardware issue. i upgraded to centos 6 and had random lockups using nouveau, as well as both the 285 and 280 nvidia drivers from elrepo. so i finally just pulled the card and went back to onboard video for now. so far things are stable. so my issue could be a sick video card that only gets tickled by certain access patterns.


All times are GMT -5. The time now is 06:17 AM.

Powered by vBulletin® Version 3.7.1
Copyright ©2000 - 2014, Jelsoft Enterprises Ltd.
Copyright 1998 - 2014, nV News.