nV News Forums

 
 

nV News Forums (http://www.nvnews.net/vbulletin/index.php)
-   NVIDIA Linux (http://www.nvnews.net/vbulletin/forumdisplay.php?f=14)
-   -   F12 + Kernel 2.6.32RC6 + nv - not a good plan... (http://www.nvnews.net/vbulletin/showthread.php?t=141378)

mbreuer 11-16-09 11:12 AM

F12 + Kernel 2.6.32RC6 + nv - not a good plan...
 
Perhaps jumping the gun abit... loaded up F12 & upgraded to 2.6.32 (from Rawhide).

For those who might try... 190.42 does not work. Hangs on:

NVIDIA(0): WAIT: (E, 0, 0x507d, 0)

strace shows Xorg hanging on an ioctl:

... lots of this ioctl... then hangs
ioctl(11, 0xc020462a, 0x7fff6fcc5720) = 0
ioctl(11, 0xc020462a, 0x7fff6fcc5720) = 0
ioctl(11, 0xc020462a, 0x7fff6fcc5720) = 0
ioctl(11, 0xc020462a, 0x7fff6fcc5720) = 0
ioctl(11, 0xc020462a, 0x7fff6fcc5720) = 0
ioctl(11, 0xc020462a, 0x7fff6fcc5720) = 0
ioctl(11, 0xc020462a, 0x7fff6fcc5720) = 0
stat("/etc/localtime", {st_mode=S_IFREG|0644, st_size=3519, ...}) = 0
write(2, "(EE) Nov 16 11:06:49 NVIDIA(0): "..., 56(EE) Nov 16 11:06:49 NVIDIA(0): WAIT: (E, 0, 0x507d, 0)
) = 56
write(0, "(EE) Nov 16 11:06:49 NVIDIA(0): "..., 56) = 56
ioctl(11, 0xc020462a, 0x7fff6fcc57b0) = 0
*** Hangs here ***

Also of note to anyone trying this out - the Xorg rpm package is build with a bad modules path - you have to specify the correct modulepath either on the command line or in xorg.conf (reported in bugzilla).

Also note: the native nvidia drivers don't work particularly well in 2.6.32 either.

In 2.6.31.5-127 (shipping with F12)...

Both native and proprietary drivers function... but the proprietary driver seems to cause spurious crashes that *seem* to be related to the screensaver kicking in. As the system freezes (until the watchdog timer reboots) I haven't been able to get much info (hence the upgrade to 2.6.32 to see if that improved things... not hanging the entire system is better, I suppose).

kwizart 11-16-09 12:56 PM

Re: F12 + Kernel 2.6.32RC6 + nv - not a good plan...
 
Quote:

Originally Posted by mbreuer (Post 2124476)
Also of note to anyone trying this out - the Xorg rpm package is build with a bad modules path - you have to specify the correct modulepath either on the command line or in xorg.conf (reported in bugzilla).

Some time ago, the way the nvidia installer used to fix the modulePath need was to overwrite the glx.so provided by the distro. (/usr/lib64/xorg/modules/extensions/glx.so).
For that reason, packager of the driver has choosen to use /usr/lib64/xorg/modules/extensions/nvidia and add this path to the ModulePath in xorg.conf.

It could be nice to have the nvidia-xconfig tool to know about this possibility. But it could be even more interesting (while we are talking to have correct name of the backend driver for the libvdpau wrapper using dri2) to have the driver itself to use such ModulePath so the correct glx.so (or eventually glx.so.%{version} ) would be appropriately choosen. (without hardcoding it in xorg.conf).

Then the same would be needed for the linker to choose the appropriate libGL.so... I wonder if that can be implemented the same as sse library path would be enabled when cpu is capable of.

Once that said, you need to send the output of nvidia-bug-report.sh

mbreuer 11-16-09 01:58 PM

Re: F12 + Kernel 2.6.32RC6 + nv - not a good plan...
 
1 Attachment(s)
Quote:

Originally Posted by kwizart (Post 2124570)
Some time ago, the way the nvidia installer used to fix the modulePath need was to overwrite the glx.so provided by the distro. (/usr/lib64/xorg/modules/extensions/glx.so).
For that reason, packager of the driver has choosen to use /usr/lib64/xorg/modules/extensions/nvidia and add this path to the ModulePath in xorg.conf.

It could be nice to have the nvidia-xconfig tool to know about this possibility. But it could be even more interesting (while we are talking to have correct name of the backend driver for the libvdpau wrapper using dri2) to have the driver itself to use such ModulePath so the correct glx.so (or eventually glx.so.%{version} ) would be appropriately choosen. (without hardcoding it in xorg.conf).

Then the same would be needed for the linker to choose the appropriate libGL.so... I wonder if that can be implemented the same as sse library path would be enabled when cpu is capable of.

Once that said, you need to send the output of nvidia-bug-report.sh

Ok - note that the path error in FC12 is unrelated to Nvidia - it's a misplaced '/' resulting in mangled names.

I've attached the report... not sure how helpful it'll be.


All times are GMT -5. The time now is 06:19 PM.

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