Bug 73445

Summary:	[NV4E] regression with 3.13 -rcX: glxgears etc. hard lockup/freezes machine, on 3.12 everything works
Product:	Mesa	Reporter:	Christian NIemeyer <christian.niemeyer>
Component:	Drivers/DRI/nouveau	Assignee:	Nouveau Project <nouveau>
Status:	RESOLVED INVALID	QA Contact:
Severity:	critical
Priority:	medium	CC:	tesoro302
Version:	10.0
Hardware:	x86-64 (AMD64)
OS:	Linux (All)
Whiteboard:
i915 platform:		i915 features:

Description Christian NIemeyer 2014-01-09 19:25:48 UTC

This seems to be a regression.

Running ubuntu 14.04 development branch. On an old HP G6000 notebook with GeForce Go 6000, this is NV4E according to my /var/log/Xorg.0.log

With 3.12.0-7-generic everything runs fine. Even basic acceleration in Firefox and playing games in Steam (like World of Goo or Snuggle Truck - those run flawlessly). I have OpenGL 2.1 support and Shader Language 1.20 according to glxinfo

I runs fine on 3.12-x with Mesa 10.0 and also fine with xorg-edergs PPA with Mesa 10.1.0-devel

However in both cases booting any 3.13 kernel and running the first GL application like glxgears or starting a game, my system entirely freezes. Whether I use stock repository or xorg-edgers. So to me this doesn't seem to be directly Mesa related, there is some problem with kernel 3.13 definitely. It freezes reproduceable 100%. No problems with 3.12.

Once I could reat a VT after a while, but it didn't react anymore.

3.13 is going to be the kernel for ubuntu 14.04 LTS. 

Sorry, this is my first bug report, please be kind. I'm not able to get any log files after reboot.

Comment 1 Christian NIemeyer 2014-01-09 19:28:05 UTC

PS: It's a GeForce Go 6100 ("6000" was typo) but still NV4E
" NOUVEAU(0): Chipset: "NVIDIA NV4E""

Comment 2 Ilia Mirkin 2014-01-09 19:36:24 UTC

Can you ssh into the machine when this happens? If so, can you look at dmesg? If not, the output of dmesg even before the hang may prove interesting.

This seems like it'd be pretty easy to bisect -- can you give that a shot? (git bisect... if you're not familiar with it, there are dozens of guides online, as well as a good man page)

Comment 3 Ilia Mirkin 2014-01-09 19:38:01 UTC

Oh, and one quick thing to try before you go the bisect route: Try adding nouveau.config=NvMSI=0 to your kernel cmdline, and see if that helps.

Comment 4 Ilia Mirkin 2015-09-11 23:38:57 UTC

Marking as invalid as there has been no response in nearly two years. Assuming the problem fixed itself, esp when we nuked MSI on all the nv4x IGPs.

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.