Bug 111868 - xorg with nouveau crash when starting mate-session in debian 10
Summary: xorg with nouveau crash when starting mate-session in debian 10
Status: NEW
Alias: None
Product: xorg
Classification: Unclassified
Component: Driver/nouveau (show other bugs)
Version: unspecified
Hardware: Other All
: not set not set
Assignee: Nouveau Project
QA Contact: Xorg Project Team
URL:
Whiteboard:
Keywords:
Depends on:
Blocks:
 
Reported: 2019-09-30 11:57 UTC by alain.prignet
Modified: 2019-10-13 16:02 UTC (History)
0 users

See Also:
i915 platform:
i915 features:


Attachments
xorg.0.log with xorg backtrace (34.19 KB, text/plain)
2019-09-30 11:57 UTC, alain.prignet
no flags Details
startx with .xinitrc "exec mate-session". it fails. (31.22 KB, text/plain)
2019-10-06 07:57 UTC, alain.prignet
no flags Details
startx with .xinitrc "exec xfce4-session". it works. (29.04 KB, text/plain)
2019-10-06 07:58 UTC, alain.prignet
no flags Details

Note You need to log in before you can comment on or make changes to this bug.
Description alain.prignet 2019-09-30 11:57:03 UTC
Created attachment 145592 [details]
xorg.0.log with xorg backtrace

Geforce 2 since debian 10 crash when starting mate-session (but not with xfce4-session or showing gdm3 or lightdm).

I have also downgrade xorg-server-core to debian 9 and I get the same crash. All was working with debian 9 (and 4.9 kernel).

I have tried 4.19 and 4.9 kernel : both crash.

If I launch mate-session from command line in a xfce4-session, mate-session works.

The backtrace of Xorg.0.log is the same as
https://termbin.com/7avc
I put it in attachment since I have not the non working computer with me. Later I will give a working Xorg.0.log (with xfce4) and the non working (with mate).
Comment 1 Ilia Mirkin 2019-09-30 12:06:01 UTC
You have a bigger problem ... your GeForce2 is reporting as a Riva TNT2!

That's a weird stack trace -- looks like it got corrupted. The "nouveau" address doesn't map to a function at all, and the present_register_complete_notify frames appear questionable to me.

How much debugging are you able to do? Do you have a second computer and know how to work gdb?
Comment 2 alain.prignet 2019-09-30 12:54:01 UTC
Thanks for your answer.

Sorry, I give you this "pastebin" since I'm away from my computer.
This log has a Riva TNT2, but mine a Geforce 2. But they gives the
"same" backtrace.

I will give you the two log files (mate-session and xfce4-session)
from my computer in few days.

gdb is difficult for me, I will need help. I'm not sure if I need two
computers since I can still use the computer in console mode after the
crash. The crash does not freeze the computer : with lightdm, I get
the infinite connect loop.
Comment 3 alain.prignet 2019-09-30 12:56:23 UTC
Le 30/09/2019 a 12:06:01 +0000, bugzilla-daemon@freedesktop.org a ecrit :
> https://bugs.freedesktop.org/show_bug.cgi?id=111868
> 
> --- Comment #1 from Ilia Mirkin <imirkin@alum.mit.edu> ---
> You have a bigger problem ... your GeForce2 is reporting as a Riva TNT2!

Sorry, I give you this "pastebin" since I'm away from my computer.
This log has a Riva TNT2, but mine a Geforce 2. But they gives the
"same" backtrace.

> 
> That's a weird stack trace -- looks like it got corrupted. The "nouveau"
> address doesn't map to a function at all, and the
> present_register_complete_notify frames appear questionable to me.

So I will give you the two log file from my computer in few days.

> 
> How much debugging are you able to do? Do you have a second computer and know
> how to work gdb?
> 

gdb is difficult for me, I will need help. I'm not sure I need two
computer since I can still use the computer in console mode after the
crash. The crash does not freeze the computer : with lightdm, I get
the infinite connect loop.

Thanks,

alain prignet
Comment 4 Ilia Mirkin 2019-09-30 13:16:09 UTC
OK, alternatively do you have a fool-proof sequence of steps to reproduce this? You originally made it sound like mate-session was the thing, but but then in later comments you say that it's actually fine.

Like let's say I run xinit and then run mate-session from within the terminal -- is that enough? (I'd like to avoid dealing with gdm3 as much as possible.)
Comment 5 alain.prignet 2019-09-30 13:31:19 UTC
Yes I have a fool-proof sequence to reproduce it, even not using gdm3 or
lightdm.

I can reproduce it using "startx" and an .xinitrc containing only
"exec mate-session". Since the computer does not freeze, I can reproducing at demand.

(The bug is not show when having in .xinitrc "exec xfce4-session", or when using gdm3 or lightdm with xfce, even if launching mate-session during the xfce4 session)
Comment 6 Ilia Mirkin 2019-09-30 13:48:44 UTC
Great, thanks. I'll work on reproducing locally.

I think we spoke on IRC a while back -- sorry I didn't actually get around to investigating -- you said that it only reproduced consistently with like TNT2, not with the later GPUs like NV34 (which I happen to have plugged in, which is why I'm asking).
Comment 7 alain.prignet 2019-09-30 13:56:56 UTC
I have a very old Geforce 2. But according to the file I found on termbin, it is the same with a Riva TNT2. This is the only nvidia card I have. My other are Intel.
Comment 8 alain.prignet 2019-10-06 07:57:47 UTC
Created attachment 145664 [details]
startx with .xinitrc "exec mate-session". it fails.
Comment 9 alain.prignet 2019-10-06 07:58:49 UTC
Created attachment 145665 [details]
startx with .xinitrc "exec xfce4-session". it works.
Comment 10 alain.prignet 2019-10-06 08:01:55 UTC
Above the 2 "real" log files.

In console mode, (real user, not root), I launch startx with .xinitrc as describe.
Comment 11 Ilia Mirkin 2019-10-13 16:02:08 UTC
This patch may help:

https://patchwork.freedesktop.org/patch/335616/

I believe that primarily GPUs with 32MB of VRAM (or less) would be affected. However it may also have a rarer effect on larger-VRAM devices.


Use of freedesktop.org services, including Bugzilla, is subject to our Code of Conduct. How we collect and use information is described in our Privacy Policy.