[gentoo-user] nouveau loads but freezes later, nvidia won't load

[gentoo-user] nouveau loads but freezes later, nvidia won't load

[Grant]

I've been using the nouveau video driver for a while, but lately video
has been freezing with X at 100% CPU and the only way out seems to be
ssh'ing in and rebooting.  The issue seems to be the same as described
here:

https://bugzilla.redhat.com/show_bug.cgi?id=596330

Disabling acceleration with the kernel parameter nouveau.noaccel=1
works but video quality is terrible.

I tried to switch to the nvidia driver but X won't load and Xorg.0.log
says it can't find a CorePointer device.  I remember having the
CorePointer problem with nvidia in the past but I can't remember how I
fixed it.  Does anyone know how to fix either driver?

- Grant

[gentoo-user] Re: nouveau loads but freezes later, nvidia won't load

[walt]

On 10/19/2010 10:32 AM, Grant wrote:
> I've been using the nouveau video driver for a while, but lately video
> has been freezing with X at 100% CPU and the only way out seems to be
> ssh'ing in and rebooting.

Not that it will help with your bug, but I'm curious to know what video
hardware you're using.  I had the same problem with my ancient nv34 card,
which got fixed a few weeks ago in Linus's git kernel.

OTOH, I've seen the same sort of freeze reported in the nouveau mailing
list for other nvidia chips, so YMMV.  Video drivers all seem to support
a dozen different chips from the same manufacturer, which introduces some
random noise into the bug reports.

> I tried to switch to the nvidia driver but X won't load and Xorg.0.log
> says it can't find a CorePointer device...

Not sure why the nvidia driver would be related to that, unless maybe
you're using a different xorg.conf for nvidia.

Unless you have some weird pointing devices, you should no longer need
any input section in xorg.conf for either keyboard or mouse -- delete any
references to keyboard or mouse, and install the xf86-input-evdev package.
Your X server should find and use the evdev driver automatically, without
any help from xorg.conf.

Re: [gentoo-user] Re: nouveau loads but freezes later, nvidia won't load

[Grant]

>> I've been using the nouveau video driver for a while, but lately video
>> has been freezing with X at 100% CPU and the only way out seems to be
>> ssh'ing in and rebooting.
>
> Not that it will help with your bug, but I'm curious to know what video
> hardware you're using.  I had the same problem with my ancient nv34 card,
> which got fixed a few weeks ago in Linus's git kernel.

I'm not sure of the actual chip, but I get this from lspci -v:

nVidia Corporation C77 [GeForce 8100 / nForce 720a]

It's an onboard chip on my motherboard.  Can you tell me in what
version of the Linux kernel you found the problem to be solved?

>> I tried to switch to the nvidia driver but X won't load and Xorg.0.log
>> says it can't find a CorePointer device...
>
> Not sure why the nvidia driver would be related to that, unless maybe
> you're using a different xorg.conf for nvidia.
>
> Unless you have some weird pointing devices, you should no longer need
> any input section in xorg.conf for either keyboard or mouse -- delete any
> references to keyboard or mouse, and install the xf86-input-evdev package.
> Your X server should find and use the evdev driver automatically, without
> any help from xorg.conf.

I removed nouveau support from the kernel and nvidia loads properly.
Strange that I get CorePointer messages if nouveau is loaded when
nvidia tries to load.  Those messages don't appear in the log after
removing nouveau and loading nvidia.  Weird.

You actually solved this for me when you asked about the video chip.
When I looked in lspci -v, I noticed that it said the nouveau driver
was in use and that is what prompted me to remove it from the kernel.

Thanks,
Grant

[gentoo-user] Re: nouveau loads but freezes later, nvidia won't load

[walt]

On 10/20/2010 12:56 PM, Grant wrote:
>>> I've been using the nouveau video driver for a while, but lately video
>>> has been freezing with X at 100% CPU and the only way out seems to be
>>> ssh'ing in and rebooting.
>>
>> Not that it will help with your bug, but I'm curious to know what video
>> hardware you're using.  I had the same problem with my ancient nv34 card,
>> which got fixed a few weeks ago in Linus's git kernel.
>
> I'm not sure of the actual chip, but I get this from lspci -v:
>
> nVidia Corporation C77 [GeForce 8100 / nForce 720a]
>
> It's an onboard chip on my motherboard.  Can you tell me in what
> version of the Linux kernel you found the problem to be solved?

In my particular case it was this commit, but I get the impression that there
are still unfixed bugs that cause very similar symptoms:

commit 3ba6462355c1c69dde58739a871d13bbb993e2e3
Author: Francisco Jerez
Date:   Sat Aug 28 17:56:33 2010 +0200

     drm/nouveau: Take fence spinlock before reading the last sequence.

     It fixes a race between the TTM delayed work queue and the GEM IOCTLs
     (fdo bug 29583) uncovered by the BKL removal.

That patch was made on Aug 28, so Linus probably committed it in early
September, maybe 2.6.36-rc3 or thereabouts.

If you're serious about testing the nouveau driver I suggest you try the
latest nouveau kernel sources from here:

git://anongit.freedesktop.org/git/nouveau/linux-2.6

I just pulled the latest changes from that repository, approx 40 changesets,
and not one of them applies to my poor old forgotten NV34 chip, so I'm losing
interest in nouveau at this point.  But your hardware is much newer and it may
be worth your while to test the latest changes.  It's certainly easy to do.

Re: [gentoo-user] Re: nouveau loads but freezes later, nvidia won't load

[Grant]

>>>> I've been using the nouveau video driver for a while, but lately video
>>>> has been freezing with X at 100% CPU and the only way out seems to be
>>>> ssh'ing in and rebooting.
>>>
>>> Not that it will help with your bug, but I'm curious to know what video
>>> hardware you're using.  I had the same problem with my ancient nv34 card,
>>> which got fixed a few weeks ago in Linus's git kernel.
>>
>> I'm not sure of the actual chip, but I get this from lspci -v:
>>
>> nVidia Corporation C77 [GeForce 8100 / nForce 720a]
>>
>> It's an onboard chip on my motherboard.  Can you tell me in what
>> version of the Linux kernel you found the problem to be solved?
>
> In my particular case it was this commit, but I get the impression that
> there
> are still unfixed bugs that cause very similar symptoms:
>
> commit 3ba6462355c1c69dde58739a871d13bbb993e2e3
> Author: Francisco Jerez
> Date:   Sat Aug 28 17:56:33 2010 +0200
>
>    drm/nouveau: Take fence spinlock before reading the last sequence.
>
>    It fixes a race between the TTM delayed work queue and the GEM IOCTLs
>    (fdo bug 29583) uncovered by the BKL removal.
>
> That patch was made on Aug 28, so Linus probably committed it in early
> September, maybe 2.6.36-rc3 or thereabouts.
>
> If you're serious about testing the nouveau driver I suggest you try the
> latest nouveau kernel sources from here:
>
> git://anongit.freedesktop.org/git/nouveau/linux-2.6
>
> I just pulled the latest changes from that repository, approx 40 changesets,
> and not one of them applies to my poor old forgotten NV34 chip, so I'm
> losing
> interest in nouveau at this point.  But your hardware is much newer and it
> may
> be worth your while to test the latest changes.  It's certainly easy to do.

Thanks a lot Walt.

- Grant