Looking for a better solution to FC 12 Nvidia 195.36.15 freeze on X startup

pablops · 04-22-10, 07:37 AM

I have a MSI P45T-C51 motherboard with 2 G RAM, and a NVidia 9400
GT. I have Installed on it Fedora 12 and Nvidia 195.36.15. My
experience on this system is that it will normally hang on the
following situations:

1) On X startup, X with the NVidia driver blinks twice then the entire
system freezes, it will not even respond to a ping. On the other hand,
starting X with the VESA driver works just fine.

2) After a fresh boot (without X), the first time I issue on the
console the command "nvidia-smi -d" would result to screen blinks then
after about 8 seconds the gpu found and the temperature are
outputted. The second time I issue a nvidia-smi command would freeze
the system.

There are rare times when X or nvidia-smi will not hang the system. I
found out that if I persevere and keep on booting there will come a
time that the system is able to start X and everything is fine and
stable. In order to be able to do several boots preceded by graceful
shutdown from the previous boot, I boot at runlevel 1. After which I
run "nvidia-smi -d" if the system takes 8 or more seconds to respond I
shutdown, boot again and repeat the process. If the system responds
within 3 seconds I then telinit to runlevel 5 (or telinit to runlevel
3, login and startx), everything will be fine and stable.

Note that I use the time to respond to a nvidia-smi to precisely
predict whether the system will be well behaved or not. A response to
within about 3 seconds signifies that all is well with the gpu
(meaning that I can issue any number of nvidia-smi commands with no
problem or I can launch X), 8 seconds means that the gpu is not
properly initialized (meaning that the next time I issue a nvidia-smi
command or if I startx the system will hang). A warm room helps
considerably. I am in the tropics, if I don't switch on the aircon I
can have X running within the first or second boot if the room is
really sweating hot and it's my lucky day.

Perhaps the X freeze problem with this system is related to a slight
electrical miss timing at gpu initialization and if you keep on trying to boot
you'll eventually get it right. My experience so far is that once the
gpu is properly initialized, as determine by a 3-second response time
to nvidia-smi, it will behave normally, no system hangs or lockups. So
my question is: is there a kernel boot option or maybe a modprobe
option or perhaps a BIOS setting that would solve this problem without
having to do the boot and boot until you succeed thing?

I have tried the clocksource=hpet kernel option and the modprobe
options nvidia NVreg_EnableMSI=1. Both didn't help.

To complete my story, installing FC 12 on this system was
problematic. The normal install would freeze probably because of
nouveau. I manage to do a basic video driver install (the second
choice on the DVD installer). This put in the VESA driver. I installed
the nvidia rpm following the instructions at rpmfusion, after which I
experienced the symptoms above and I found out how it could be
overcome by persistent booting perseverance. Then I installed the
nvidia 195.36.15 and the same behavior manifested.

A number of postings tell of similar problem to what I'm
encountering. As far as I have read there is read no successful work
around yet. I have found a work around for my system, but it's very
tedious, perhaps based on this others can think of a better solution.

quyetnd · 04-22-10, 10:13 AM

Hi Pablops,

I got some experiences messing around with the nvidia driver on fedora 12 before (got the "good enough" solutions from here: http://www.nvnews.net/vbulletin/showthread.php?t=150012 ).
Just something I want you to make sure:
1, You uninstalled/blacklisted the nouveau driver. Some time black list is not enough, you'll need to rebuild the boot image:
# dracut -f /boot/initramfs-$(uname -r).img $(uname -r)
2, You built nvidia module for your kernel (you should mostly already do it since you use the rpmfusion repository)
3, You should attach the bug report (nvidia-bug-report.sh). It'll help alot

pablops · 04-23-10, 12:50 AM

Hi quyetnd !

Unfortunately when the system hangs it doesn't have time to write anything on the logs.

quyetnd · 04-23-10, 10:30 AM

For logging you can boot the system in lv3. (in grub press "a" and add 3 into the boot command)

04-22-10, 07:37 AM	#1
pablops Registered User Join Date: Apr 2010 Posts: 2	Looking for a better solution to FC 12 Nvidia 195.36.15 freeze on X startup I have a MSI P45T-C51 motherboard with 2 G RAM, and a NVidia 9400 GT. I have Installed on it Fedora 12 and Nvidia 195.36.15. My experience on this system is that it will normally hang on the following situations: 1) On X startup, X with the NVidia driver blinks twice then the entire system freezes, it will not even respond to a ping. On the other hand, starting X with the VESA driver works just fine. 2) After a fresh boot (without X), the first time I issue on the console the command "nvidia-smi -d" would result to screen blinks then after about 8 seconds the gpu found and the temperature are outputted. The second time I issue a nvidia-smi command would freeze the system. There are rare times when X or nvidia-smi will not hang the system. I found out that if I persevere and keep on booting there will come a time that the system is able to start X and everything is fine and stable. In order to be able to do several boots preceded by graceful shutdown from the previous boot, I boot at runlevel 1. After which I run "nvidia-smi -d" if the system takes 8 or more seconds to respond I shutdown, boot again and repeat the process. If the system responds within 3 seconds I then telinit to runlevel 5 (or telinit to runlevel 3, login and startx), everything will be fine and stable. Note that I use the time to respond to a nvidia-smi to precisely predict whether the system will be well behaved or not. A response to within about 3 seconds signifies that all is well with the gpu (meaning that I can issue any number of nvidia-smi commands with no problem or I can launch X), 8 seconds means that the gpu is not properly initialized (meaning that the next time I issue a nvidia-smi command or if I startx the system will hang). A warm room helps considerably. I am in the tropics, if I don't switch on the aircon I can have X running within the first or second boot if the room is really sweating hot and it's my lucky day. Perhaps the X freeze problem with this system is related to a slight electrical miss timing at gpu initialization and if you keep on trying to boot you'll eventually get it right. My experience so far is that once the gpu is properly initialized, as determine by a 3-second response time to nvidia-smi, it will behave normally, no system hangs or lockups. So my question is: is there a kernel boot option or maybe a modprobe option or perhaps a BIOS setting that would solve this problem without having to do the boot and boot until you succeed thing? I have tried the clocksource=hpet kernel option and the modprobe options nvidia NVreg_EnableMSI=1. Both didn't help. To complete my story, installing FC 12 on this system was problematic. The normal install would freeze probably because of nouveau. I manage to do a basic video driver install (the second choice on the DVD installer). This put in the VESA driver. I installed the nvidia rpm following the instructions at rpmfusion, after which I experienced the symptoms above and I found out how it could be overcome by persistent booting perseverance. Then I installed the nvidia 195.36.15 and the same behavior manifested. A number of postings tell of similar problem to what I'm encountering. As far as I have read there is read no successful work around yet. I have found a work around for my system, but it's very tedious, perhaps based on this others can think of a better solution.

04-22-10, 10:13 AM	#2
quyetnd Registered User Join Date: Mar 2010 Posts: 10	Re: Looking for a better solution to FC 12 Nvidia 195.36.15 freeze on X startup Hi Pablops, I got some experiences messing around with the nvidia driver on fedora 12 before (got the "good enough" solutions from here: http://www.nvnews.net/vbulletin/showthread.php?t=150012 ). Just something I want you to make sure: 1, You uninstalled/blacklisted the nouveau driver. Some time black list is not enough, you'll need to rebuild the boot image: # dracut -f /boot/initramfs-$(uname -r).img $(uname -r) 2, You built nvidia module for your kernel (you should mostly already do it since you use the rpmfusion repository) 3, You should attach the bug report (nvidia-bug-report.sh). It'll help alot

04-23-10, 12:50 AM	#3
pablops Registered User Join Date: Apr 2010 Posts: 2	Re: Looking for a better solution to FC 12 Nvidia 195.36.15 freeze on X startup Hi quyetnd ! Unfortunately when the system hangs it doesn't have time to write anything on the logs.

04-23-10, 10:30 AM	#4
quyetnd Registered User Join Date: Mar 2010 Posts: 10	Re: Looking for a better solution to FC 12 Nvidia 195.36.15 freeze on X startup For logging you can boot the system in lv3. (in grub press "a" and add 3 into the boot command)