[gentoo-user] X crashes with nvidia-173 driver

Maxim Wexler · External Since: Aug 31, 2009 Posts: 22

Hi group

Having installed the Geforce FX5500 drivers,
NVIDIA-Linux-x86-173.14.18-pkg1.run, and ran startx the desktop
appears but the system crashes.

From the X log, (WW):

(WW) warning, (EE) error, (NI) not implemented, (??) unknown.
(WW) The directory "/usr/share/fonts/misc/" does not exist.
(WW) The directory "/usr/share/fonts/TTF/" does not exist.
(WW) The directory "/usr/share/fonts/Type1/" does not exist.
(WW) The directory "/usr/share/fonts/100dpi/" does not exist.
(WW) The directory "/usr/share/fonts/75dpi/" does not exist.
(WW) The directory "/usr/share/fonts/misc/" does not exist.
(WW) The directory "/usr/share/fonts/TTF/" does not exist.
(WW) The directory "/usr/share/fonts/OTF" does not exist.
(WW) The directory "/usr/share/fonts/Type1/" does not exist.
(WW) The directory "/usr/share/fonts/100dpi/" does not exist.
(WW) The directory "/usr/share/fonts/75dpi/" does not exist.
(WW) AllowEmptyInput is on, devices using drivers 'kbd', 'mouse' or
'vmmouse' will be disabled.
(WW) Disabling Mouse1
(WW) Disabling Keyboard1
(WW) Open ACPI failed (/var/run/acpid.socket) (No such file or directory)
(WW) Warning, couldn't open module dri
(WW) Warning, couldn't open module freetype
(WW) Warning, couldn't open module xtt
(WW) Warning, couldn't open module dri2

and (EE)

(WW) warning, (EE) error, (NI) not implemented, (??) unknown.
(EE) Failed to load module "dri" (module does not exist, 0)
(EE) Failed to load module "freetype" (module does not exist, 0)
(EE) Failed to load module "xtt" (module does not exist, 0)
(EE) Failed to load module "dri2" (module does not exist, 0)

Notes:
"dri" was commented out in xorg.conf but uncommenting made no difference.

$ls /usr/share/fonts
default encodings ttf-bitstream-vera

and nothing in those dirs like those mentioned in (WW)

yes, ran "eselect opengl set nvidia" after drivers were emerged.

#nvidia-settings -V
ERROR: The control display is undefined; please run 'nvidia-settings --help'

Is there an option I can pass to nvidia-settings to solve this, or
should I look elsewhere?

Maxim

walt · External Since: Dec 20, 2006 Posts: 27

On 11/01/2009 09:13 AM, Maxim Wexler wrote:
> Hi group
>
> Having installed the Geforce FX5500 drivers,
> NVIDIA-Linux-x86-173.14.18-pkg1.run, and ran startx the desktop
> appears but the system crashes...

Do you mean the entire system, or just the X server?

> #nvidia-settings -V
> ERROR: The control display is undefined; please run 'nvidia-settings --help'

The nvidia-settings applet is an X program and thus requires a running X session.
I assume you are running it from a console, since X crashes?

If you type 'X' from a console (instead of startx) does the crash still happen?

Maxim Wexler · External Since: Aug 31, 2009 Posts: 22

On 11/1/09, walt <w41ter.TakeThisOut@gmail.com> wrote:
> On 11/01/2009 09:13 AM, Maxim Wexler wrote:
>> Hi group
>>
>> Having installed the Geforce FX5500 drivers,
>> NVIDIA-Linux-x86-173.14.18-pkg1.run, and ran startx the desktop
>> appears but the system crashes...
>
> Do you mean the entire system, or just the X server?

Desktop(raw X, default terminals and clock icon) appears, instant
total system crash

> If you type 'X' from a console (instead of startx) does the crash still
> happen?

Black screen, instant total system crash.

Oh, and the green nvidia icon fails to flash briefly after the startx command.

>

walt · External Since: Dec 20, 2006 Posts: 27

On 11/01/2009 02:36 PM, Maxim Wexler wrote:
> On 11/1/09, walt <w41ter.DeleteThis@gmail.com> wrote:
>> On 11/01/2009 09:13 AM, Maxim Wexler wrote:
>>> Hi group
>>>
>>> Having installed the Geforce FX5500 drivers,
>>> NVIDIA-Linux-x86-173.14.18-pkg1.run, and ran startx the desktop
>>> appears but the system crashes...
>>
>> Do you mean the entire system, or just the X server?
>
> Desktop(raw X, default terminals and clock icon) appears, instant
> total system crash
>
>> If you type 'X' from a console (instead of startx) does the crash still
>> happen?
>
> Black screen, instant total system crash.

Well, that suggests a kernel panic, I think. What happens if you use an
older kernel?

> Oh, and the green nvidia icon fails to flash briefly after the startx command.

I use the same nvidia drivers for the nVidia Corporation NV34 [GeForce FX 5200]
video card (given to me by my brother-in-law-the-gamer after I installed his new,
upgraded nVidia card for him Surprised

)

For a long time I saw that same green nvidia logo flash during startup, but now
that you mention it, I haven't seen it for a long time. I really don't know when
it stopped, but it was weeks ago, not days.

Dale · External Since: Jun 20, 2009 Posts: 43

walt wrote:
> On 11/01/2009 02:36 PM, Maxim Wexler wrote:
>
>> On 11/1/09, walt <w41ter RemoveThis @gmail.com> wrote:
>>
>>> On 11/01/2009 09:13 AM, Maxim Wexler wrote:
>>>
>>>> Hi group
>>>>
>>>> Having installed the Geforce FX5500 drivers,
>>>> NVIDIA-Linux-x86-173.14.18-pkg1.run, and ran startx the desktop
>>>> appears but the system crashes...
>>>>
>>> Do you mean the entire system, or just the X server?
>>>
>> Desktop(raw X, default terminals and clock icon) appears, instant
>> total system crash
>>
>>
>>> If you type 'X' from a console (instead of startx) does the crash still
>>> happen?
>>>
>> Black screen, instant total system crash.
>>
>
> Well, that suggests a kernel panic, I think. What happens if you use an
> older kernel?
>
>
>> Oh, and the green nvidia icon fails to flash briefly after the startx command.
>>
>
> I use the same nvidia drivers for the nVidia Corporation NV34 [GeForce FX 5200]
> video card (given to me by my brother-in-law-the-gamer after I installed his new,
> upgraded nVidia card for him Surprised

)
>
> For a long time I saw that same green nvidia logo flash during startup, but now
> that you mention it, I haven't seen it for a long time. I really don't know when
> it stopped, but it was weeks ago, not days.
>
>
>

I was informed a while back that the later driver versions only show the
Nvidia splash screen the first time the module is loaded or something
like that. Basically you only see it once. That can be changed somehow
but I can remember how they said to do it. I only asked because I
thought my Nvidia drivers weren't loading.

Dale

Smile

Alan McKinnon · External Since: Jun 26, 2009 Posts: 68

On Monday 02 November 2009 05:11:45 Dale wrote:
> I was informed a while back that the later driver versions only show the
> Nvidia splash screen the first time the module is loaded or something
> like that. Basically you only see it once. That can be changed somehow
> but I can remember how they said to do it. I only asked because I
> thought my Nvidia drivers weren't loading.
>

Option "NoLogo" <boolean>

in xorg.conf

--
alan dot mckinnon at gmail dot com

Maxim Wexler · External Since: Aug 31, 2009 Posts: 22

>> Black screen, instant total system crash.
>
> Well, that suggests a kernel panic, I think. What happens if you use an
> older kernel?
>
There is no older kernel. I knew there was something I forgot to save.
But I saved xorg.conf and it's identical.

The sysrecusecd finds the card and X starts fine. The only difference
in the config is that on the CD CONFIG_AGP_NVIDIA=y and on the PC it's
a module. So I set it to y. Didn't help.

mw

walt · External Since: Dec 20, 2006 Posts: 27

On 11/02/2009 10:55 AM, Maxim Wexler wrote:
>>> Black screen, instant total system crash.
>>
>> Well, that suggests a kernel panic, I think. What happens if you use an
>> older kernel?
>>
> There is no older kernel. I knew there was something I forgot to save.
> But I saved xorg.conf and it's identical.

You can compile one from an older source package and see if it helps. Just
use the same kernel config you have now and do make oldconfig to remove
any new config items.

Maxim Wexler · External Since: Aug 31, 2009 Posts: 22

> You can compile one from an older source package and see if it helps. Just
> use the same kernel config you have now and do make oldconfig to remove
> any new config items.

Using linux-2.6.29. Changed symlink, re-emerged nvidia-drivers, ran
#modprobe nvidia:

FATAL: error inserting nvidia (/lib/modules/2.6.29/video/nvidia.ko):
No such device

???lspci begs to differ

walt · External Since: Dec 20, 2006 Posts: 27

On 11/02/2009 01:45 PM, Maxim Wexler wrote:
>> You can compile one from an older source package and see if it helps. Just
>> use the same kernel config you have now and do make oldconfig to remove
>> any new config items.
>
> Using linux-2.6.29. Changed symlink, re-emerged nvidia-drivers, ran
> #modprobe nvidia:
>
> FATAL: error inserting nvidia (/lib/modules/2.6.29/video/nvidia.ko):
> No such device
>
> ???lspci begs to differ

<groan> I've never seen that error from nvidia, and I've seen lots of
errors Surprised

(

Are you sure the nvidia module that was actually merged is the same
version you thought you were merging? (I'm reduced to silly questions.)

walt · External Since: Dec 20, 2006 Posts: 27

On 11/02/2009 06:32 PM, Maxim Wexler wrote:
>>> FATAL: error inserting nvidia (/lib/modules/2.6.29/video/nvidia.ko):
>>> No such device
>>>
>>> ???lspci begs to differ
>>
>> <groan> I've never seen that error from nvidia, and I've seen lots of
>> errors Surprised

(
>>
>> Are you sure the nvidia module that was actually merged is the same
>> version you thought you were merging? (I'm reduced to silly questions.)
>>
> 173.14.20 was what emerge -pv called for and that's what's in /var/log/portage
>
> meanwhile attached greps of (EE) and (WW) in Xorg.0.log, FWIW

The error in your Xlog uses the word 'module' in this sense:

(II) LoadModule: "nvidia"
(II) Loading /usr/lib/xorg/modules/drivers//nvidia_drv.so
(II) Module nvidia: vendor="NVIDIA Corporation"
compiled for 4.0.2, module version = 1.0.0
Module class: X.Org Video Driver
(II) NVIDIA dlloader X Driver 173.14.20 Thu Jun 25 19:28:52 PDT 2009
(II) NVIDIA Unified Driver for all Supported NVIDIA GPUs
(II) Primary Device is: PCI 01@00:00:0

I'm guessing that the xorg server couldn't load its 'nvidia' module because
the nvidia kernel module wasn't loaded, so that's no surprise.

Does modprobe -nv nvidia say anything interesting?

Maxim Wexler · External Since: Aug 31, 2009 Posts: 22

> I'm guessing that the xorg server couldn't load its 'nvidia' module because
> the nvidia kernel module wasn't loaded, so that's no surprise.
>
> Does modprobe -nv nvidia say anything interesting?

kyzyl ~ # modprobe -nv nvidia
insmod /lib/modules/2.6.29/video/nvidia.ko NVreg_DeviceFileMode=432
NVreg_DeviceFileUID=0 NVreg_DeviceFileGID=27 NVreg_ModifyDeviceFiles=1
NVreg_DeviceFileMode=432 NVreg_DeviceFileUID=0 NVreg_DeviceFileGID=27
NVreg_ModifyDeviceFiles=1
FATAL: Error inserting nvidia (/lib/modules/2.6.29/video/nvidia.ko):
No such device

walt · External Since: Dec 20, 2006 Posts: 27

On 11/03/2009 06:10 PM, Maxim Wexler wrote:
>> I'm guessing that the xorg server couldn't load its 'nvidia' module because
>> the nvidia kernel module wasn't loaded, so that's no surprise.
>>
>> Does modprobe -nv nvidia say anything interesting?
>
>
> kyzyl ~ # modprobe -nv nvidia
> insmod /lib/modules/2.6.29/video/nvidia.ko NVreg_DeviceFileMode=432
> NVreg_DeviceFileUID=0 NVreg_DeviceFileGID=27 NVreg_ModifyDeviceFiles=1
> NVreg_DeviceFileMode=432 NVreg_DeviceFileUID=0 NVreg_DeviceFileGID=27
> NVreg_ModifyDeviceFiles=1
> FATAL: Error inserting nvidia (/lib/modules/2.6.29/video/nvidia.ko):
> No such device

Do you have /lib/modules/2.6.29/modules.pcimap?

#head modules.pcimap
# pci module vendor device subvendor subdevice class class_mask driver_data
snd-via82xx 0x00001106 0x00003058 0xffffffff 0xffffffff 0x00000000 0x00000000 0x0
snd-via82xx 0x00001106 0x00003059 0xffffffff 0xffffffff 0x00000000 0x00000000 0x0
nvidia 0x000010de 0xffffffff 0xffffffff 0xffffffff 0x00030000 0xffffffff 0x0
nvidia 0x000010de 0xffffffff 0xffffffff 0xffffffff 0x00030200 0xffffffff 0x0
uhci-hcd 0xffffffff 0xffffffff 0xffffffff 0xffffffff 0x000c0300 0xffffffff 0x0

Maxim Wexler · External Since: Aug 31, 2009 Posts: 22

> nvidia 0x000010de 0xffffffff 0xffffffff 0xffffffff 0x00030000
> 0xffffffff 0x0
> nvidia 0x000010de 0xffffffff 0xffffffff 0xffffffff 0x00030200
> 0xffffffff 0x0

The lines in my modules.pcimap are identical.

FWIW, I mentioned previously that startx precipitated a total crash;
not so, having installed the sound drivers it is clear that only X
crashes and SysRq works to get me backto a console. Also, been reading
the recent hal + nvidia thread. Definite similarities to my situation.
I emerged xorg with hal support but I tried with the hal daemon off
and it didn't help. I also tried the very brief, five line, xorg.conf
file suggested no help, no change.

I also note something eerily similar in my netbook which has only a
builtin intel vid chip. In this case X starts fine but then, after a
period of time, could be as short as 15 mins or as long as 8 hours the
desktop goes blank, ususally black but once grey. I can migrate to the
first console and turn off theX server but I have to reboot to get
back to the desktop. Running startx again just leads back to a blank
screen. Every thing else, sound card, wireless seems to work OK.

mw

walt · External Since: Dec 20, 2006 Posts: 27

On 11/05/2009 01:17 PM, Maxim Wexler wrote:
> ...
> I also note something eerily similar in my netbook which has only a
> builtin intel vid chip. In this case X starts fine but then, after a
> period of time, could be as short as 15 mins or as long as 8 hours the
> desktop goes blank, ususally black but once grey. I can migrate to the
> first console and turn off theX server but I have to reboot to get
> back to the desktop. Running startx again just leads back to a blank
> screen...

There are definitely bugs in the intel video driver, as posted to the
freedesktop.xorg mailing list:

Release 2.9.1 (2009-10-26)
==========================
We are pleased to announce a maintenance release (2.9.1) of the
xf86-video-intel driver. This release comes one month after 2.9.0 and
consists only of a few, hand-picked bug fixes since that release.

Thanks to everyone for your help in putting together a fine release.

-Carl

PS. If you want to nominate fixes for future 2.9.x releases, please do
so at: http://wiki.x.org/wiki/Intel29Branch

Where to get xf86-video-intel 2.9.1
-----------------------------------
git tag: 2.9.1

http://xorg.freedesktop.org/archive/individual/driver/xf86-video-intel...9.1.tar
MD5: 8951d0366c16991badb7f9050556f4f3 xf86-video-intel-2.9.1.tar.bz2
SHA1: ca5887df67ec46f27eeeaf8158048d6500333ccf xf86-video-intel-2.9.1.tar.bz2

http://xorg.freedesktop.org/archive/individual/driver/xf86-video-intel...9.1.tar
MD5: b1926dc68cc10209fc2200729fb6162b xf86-video-intel-2.9.1.tar.gz
SHA1: a1810874c6bff487232cbea26ec02a45bbe6e9a2 xf86-video-intel-2.9.1.tar.gz

Complete list of fixes in 2.9.1 compared to 2.9.0
-------------------------------------------------
* Fix corruption and artifacts due to wrong colors in the colormap
with X server 1.7

* Fix incorrect rendering, such as missing scrollbar arrows in some
themes (http://bugs.freedesktop.org/show_bug.cgi?id=24459)

* Fix black screen when X server is reset
(https://bugs.freedesktop.org/show_bug.cgi?id=24383)

* Fix regressions detecting DVI monitors

http://bugs.freedesktop.org/show_bug.cgi?id=24255
http://bugs.freedesktop.org/show_bug.cgi?id=24282
http://bugs.freedesktop.org/show_bug.cgi?id=24458

Maxim Wexler · External Since: Aug 31, 2009 Posts: 22

> There are definitely bugs in the intel video driver, as posted to the
> freedesktop.xorg mailing list:
>
> Release 2.9.1 (2009-10-26)

Thanks. Just did an eix-sync followed by ACCEPT_KEYWORDS="~x86"
emerge -v xf86-video-intel and the package cae down.

Will see how it works.

BTW how come package.mask doesn't like <xf86-video-intel-2.9.1? I want
to mask everything earlier than 2.9.1, no?

mw

walt · External Since: Dec 20, 2006 Posts: 27

On 11/05/2009 07:19 PM, Maxim Wexler wrote:
>> There are definitely bugs in the intel video driver, as posted to the
>> freedesktop.xorg mailing list:
>>
>> Release 2.9.1 (2009-10-26)
>
> Thanks. Just did an eix-sync followed by ACCEPT_KEYWORDS="~x86"
> emerge -v xf86-video-intel and the package cae down.
>
> Will see how it works.
>
> BTW how come package.mask doesn't like <xf86-video-intel-2.9.1? I want
> to mask everything earlier than 2.9.1, no?

You need to use the package category also:
<x11-drivers/xf86-video-intel-2.9.1