Nvidia 177.82 30+ seconds hang

[ale.ji]

Please tell us when fix this bug??? Nvidia ...

[kellinwood]

I'm getting this hang with 177.82 and 180.11 on my T61 Thinkpad with an Quadro NVS 140M. As reported by others, the first start of X after power-up is short and sweet, but each successive restart of X results in a 30 second delay after which everything appears and runs normally. I'm running Fedora 9 which starts X twice during a normal boot-up (once for the graphical boot, and again for graphical login). By removing the "rhgb" parameter from the grub kernel parameters, I can get the system to start up very quickly.

FYI, I've tracked the problem down to the open() call of /dev/nvidia0. The full strace output which shows this is attached. Here's an excerpt

19:50:35.001976 chown32("/dev/nvidia0", 0, 0) = 0
19:50:35.002113 open("/dev/nvidia0", O_RDWR|O_LARGEFILE) = 12
19:51:05.710293 ioctl(11, 0xc040464d, 0xbfca1650) = 0
19:51:05.710478 open("/proc/interrupts", O_RDONLY|O_LARGEFILE) = 13

Note that it takes the open() call 30 seconds to complete.

Ken

[NvFuchs]

Not fixed in 180.41, 4 times worse in new series ...

nvidia developers, are you still aware of this bug, and is there a slight chance that this will be fixed in the future?

[ale.ji]

seems they just ignore this issue ...

[pavlinux]

rm -f `modinfo nvidiafb | grep "filename:" | awk '{print $2}'`

DO NOT USE NVIDIAFB with NVIDIA modules simultaneously!!!

1. exec
# Xorg -scanpci

2. find you card pciid

add to xorg.conf

Section "Device"
BusID "2:0:0"
EndSection

and

Section "Screen"
Option "NoInt10" "true"
EndSection

[pferraro]

FYI...

Originally Posted by pavlinux

# Xorg -scanpci

Fatal server error:
Unrecognized option: -scanpci

Originally Posted by pavlinux

Section "Screen"
Option "NoInt10" "true"
EndSection

From /var/log/Xorg.0.log:
[ 7.366155] (WW) NVIDIA(0): Option "NoInt10" is not used

[NvFuchs]

Originally Posted by pavlinux

rm -f `modinfo nvidiafb | grep "filename:" | awk '{print $2}'`

DO NOT USE NVIDIAFB with NVIDIA modules simultaneously!!!

Would you mind not using ridiculous large fonts with colours and multiple exclamation marks? Because I am, probably as most of the other users as well, not using the nvidia framebuffer.

1. exec
# Xorg -scanpci

2. find you card pciid

add to xorg.conf

Section "Device"
BusID "2:0:0"
EndSection

The command is wrong, but by reading my bug report you might find out that I in fact did, as most of the others did as well, specify the PCI BusID in the Xorg.conf

and

Section "Screen"
Option "NoInt10" "true"
EndSection

Is ignored anyway, and I don't see right now why this should help.

Thanks anyway.

Fuchs

[NvFuchs]

*sigh*

Could we at least have a comment on this, nvidia?

Thanks.

Fuchs

[AaronP]

The status is still the same as when I replied before: the bug has been identified and is being investigated.

[craftyguy]

Originally Posted by AaronP

The status is still the same as when I replied before: the bug has been identified and is being investigated.

Lol. Take a couple guys off the VDPAU team to fix some real issues??

[NvFuchs]

Originally Posted by AaronP

The status is still the same as when I replied before: the bug has been identified and is being investigated.

Thanks, so at least I know you did not forget about this.

Thanks in advance for fixing this soon.

Kind regards

[NvFuchs]

After 5 tests I'd say the problem is gone with 180.44.

Could somebody else please confirm?

@Zander or AaronP: Was there any fix regarding this issue?
Because it seems to be gone, and it would be nice to see this ported to new releases (as current beta still has it) as well.

If it really is: thanks a lot nvidia.

Fuchs