nvidia-driver-195.36.15 hangs [was Re: cvs commit: ports/x11/nvidia-driver Makefile distinfo pkg-plist]

[Garrett Cooper]

On Mon, May 3, 2010 at 11:00 AM, Garrett Cooper <yanefbsd at gmail.com> wrote:
> On Mon, Apr 26, 2010 at 4:26 PM, Garrett Cooper <yanefbsd at gmail.com> wrote:
>> 2010/4/26 Rene Ladan <rene at freebsd.org>:
>>> On 26-04-2010 22:41, Max Laier wrote:
>>>> On Monday 26 April 2010 16:35:12 Alexey Dokuchaev wrote:
>>>>> On Sun, Apr 25, 2010 at 10:30:00PM +0200, Max Laier wrote:
>>>>>> On Sunday 25 April 2010 22:20:49 Doug Barton wrote:
>>>>>>> On 04/25/10 11:17, Max Laier wrote:
>>>>>>>> On Saturday 10 April 2010 15:40:07 Alexey Dokuchaev wrote:
>>>>>>>>> danfe       2010-04-10 13:40:07 UTC
>>>>>>>>>
>>>>>>>>>   FreeBSD ports repository
>>>>>>>>>
>>>>>>>>>   Modified files:
>>>>>>>>>     x11/nvidia-driver    Makefile distinfo pkg-plist
>>>>>>>>>   Log:
>>>>>>>>>   - Update nVidia BETA drivers to version 195.36.15
>>>>>>>>
>>>>>>>> might be me, but it looks like the BETA driver has been more stable.
>>>>>>>> I just downgraded back to BETA and now X doesn't hang anymore after
>>>>>>>> ~30min.
>>>>>>>
>>>>>>> I had to downgrade back to 195.22. When running the latest version I
>>>>>>> had freezes similar to what Max described. If it ran successfully for
>>>>>>> any length of time it would lock up as soon as xscreensaver kicked in.
>>>>>>
>>>>>> not related to xscreensaver in my case, I don't think.  It was locking up
>>>>>> while playing video in xine which - afaik - disables the screensaver.  As
>>>>>> always, X lock ups are hard to track down unfortunately.  On the plus
>>>>>> side, it doesn't seem to be a hard lock up - i.e. the power button would
>>>>>> still get me to a clean shutdown eventually, but interrupts didn't fire
>>>>>> anymore (at least the network stops working).  I can't attach a serial
>>>>>> console at the moment to get into DDB.  Will try to, tomorrow.
>>>>>
>>>>> Please do.  I am still not able to reproduce it here on Februarish
>>>>> -CURRENT/amd64 with GFX260.  You might also want to lurk on nVidia
>>>>> forums at http://www.nvnews.net/vbulletin/forumdisplay.php?f=47 to see
>>>>> if someone has similar issues.
>>>>
>>>> There are quite a few reports.  I did a diff on the glue code and there are a
>>>> few changes between 22 and 36 that could be the culprit.  I'll investigate
>>>> further and let you know.  8.0-RELEASE here, by the way - does anyone know if
>>>> there are (missing) MFCs regarding the cdevpriv API?  That's one of the bigger
>>>> changes in the glue code.
>>>>
>>> The drviver seems to run ok here. X.org 7.4 has been running for 27
>>> hours now, without lockups.  I'm using the blank screensaver inside
>>> X.org (first blank the screen, then power down the screen).
>>>
>>> This is on an Acer Aspire 7738G with this video card:
>>> (II) Apr 25 20:10:17 NVIDIA(0): NVIDIA GPU GeForce GT 240M (GT216) at
>>> PCI:1:0:0 (GPU-0)
>>
>>    Hmmm... this is part of an issue I need to triage on my two
>> machines. I have two machines with similar hardware specs:
>>
>> Intel W3520 in both
>> 12GB RAM, same vendor in both
>> Different nvidia cards; same model -- 9600GT, different memory
>> amounts, different vendors
>> Different DVD drives (shouldn't matter)
>> Different hard drives
>> An LSI RAID card in one
>> A snd_emu10kx sound card in the server/workstation.
>>
>>    There are a few variables in the mix, but my machines hardlock
>> from approximately a week usage with sources based off r206031, using
>> ahci(4) [instead of ata(4) -- I switched over my machines to use
>> ahci(4) 2 weeks ago], and this particular driver version... I'll
>> downgrade nvidia-drivers and see what happens.
>
>    Downgrading nvidia-driver to 195.22 has been pointing a smoking
> gun at a regression in 195.36.15 with my configuration, as my system
> has not hung up once since I downgraded both machines, whereas I ruled
> out the issue by disabling all of the other experimental features I
> may have been tweaking around with rather quickly (my monkeying around
> with the power settings and the rtc was causing some variance as well
> between the two machines).
>    Others have reported similar issues with panics on #bsdports.

(switching from cvs-all@ to xorg@)
    Here's my xorg.conf, for future reference. I'll try and get
firewire up and going on the box so I can reproduce the panic or
hardlock condition on the console.
Thanks,
-Garrett
-------------- next part --------------
A non-text attachment was scrubbed...
Name: xorg.conf
Type: application/octet-stream
Size: 2583 bytes
Desc: not available
Url : http://lists.freebsd.org/pipermail/freebsd-x11/attachments/20100503/450a2c67/xorg.obj