Soft freeze with nvidia 9500 GS and driver 180.18

[helmotz]

Hi to everyone,
since I've bought a new computer which has an integrated GeForce 9500 GS I'm experiencing random lookup problems.
I use Fedora 10 x86_64 and nvidia drivers have been installed using the .run installer.
Apparently without a reason during normal computer usage mouse pointer freezes.
Computer does still work, as I'm able to switch the application using alt-tab and I'm able to type commands in terminals. If I try to switch to a VT console the display gets completely frozen.
I'm still able to ssh inside my PC from another box. If I kill the X-Server I can verify with ps that that process has been really killed but nothing changes on the display.
I followed the instructions reported in sticky posts before submitting this nvidia-bug-report.log in particular:

- I used the pci=nommconf option from my kernel boot line
- I checked that my mother board bios is updated
- I removed vga=0x318 option from my kernel boot line

The nvidia-bug-report.log that I'm sending has been taken when the mouse disappeared but before trying to switch to a virtual console.
Please also notice that pci=nommconf is not present in the kernel boot line and vga=0x318 is present as I verified that the behaviour doesn't change with any combination of these settings.
The problem is very annoying before when this thing happens I don't have any other option than rebooting with 'shutdown -r now'.

Thanks in advance for your help.

P.S. Happy New Year to everybody

[helmotz]

Another bit of information: I use compiz and gnome.
This is the output of the command "rpm -q -a | grep compiz":

libcompizconfig-0.7.8-1.fc10.x86_64
compiz-fusion-extras-gnome-0.7.8-2.fc10.x86_64
compizconfig-python-0.7.8-1.fc10.x86_64
gnome-compiz-manager-0.10.4-4.fc10.x86_64
compiz-fusion-extras-0.7.8-2.fc10.x86_64
compiz-0.7.8-4.fc10.x86_64
compiz-fusion-gnome-0.7.8-2.fc10.x86_64
compiz-manager-0.6.0-8.fc10.noarch
compiz-fusion-0.7.8-2.fc10.x86_64
compiz-kde-0.7.8-4.fc10.x86_64
compiz-gnome-0.7.8-4.fc10.x86_64

[Lethe]

I am getting the same sort of crap on my amdx86_64 running Slamd.

The fix for me is to disable RenderAccel in xorg.conf.

See how you get on.

In 'Section "Device" ', add:

Option "RenderAccel" "false"

That totally stops the similar lockups I have. OK, it makes the desktop run a bit laggy - I use Fluxbox, so that doesn't really affect me - but I guess a KDE/Gnome user would have serious usability issues.

Oh. EDIT: I forgot - enabling noapic to kernel boot line also fixes this, but then I get 100's of IRQ ERR (from interrupt 7) occurring every second - so that is no option at all.

Nick

[helmotz]

Thanks a lot Lethe.
Sadly in my case disabling RenderAccel is not an option as I use most of the times my linux box with Gnome (I gave up using it with KDE4 together with NVidia cards long time ago...).
I remember that once I had this freezing on GDM screen, so I think that (at least this time ) compiz is not guilty.

Thanks a lot for your help,
Francesco

[helmotz]

It seems that performance issues using RenderAccel false are not severe using compiz.
I will give a try and see if this helps with my problem. The soft lookup that I'm experimenting are not so frequent so I will need at least a couple of day for saying if this works or not.
I know how much hardly nvidia folks work for supporting so many GPUs and so many OSes but I'd like to know if they're aware of this problem.

F.

[Lethe]

Originally Posted by helmotz

It seems that performance issues using RenderAccel false are not severe using compiz.
I will give a try and see if this helps with my problem. The soft lookup that I'm experimenting are not so frequent so I will need at least a couple of day for saying if this works or not.
I know how much hardly nvidia folks work for supporting so many GPUs and so many OSes but I'd like to know if they're aware of this problem.

F.

I need to do a bug report, but all I get is one liner in syslogs - and nothing else.

NVRM: Xid (0003:00): 12, COCOD 00000001 80013900 00005039 00001a64 00d9b350

PITA.

Nick

[helmotz]

Originally Posted by Lethe

I need to do a bug report, but all I get is one liner in syslogs - and nothing else.

NVRM: Xid (0003:00): 12, COCOD 00000001 80013900 00005039 00001a64 00d9b350

With the stable driver 177.82 I also got something similar in syslog.
I hope I'll be able to survive with "RenderAccel" disabled... I just noticed that resizing windows can be a real pain this way.

F.

[helmotz]

UPDATE: With 180.18 and the RenderAccel false settings didn't have crashes for a day but resizing was a nightmare, so I reverted back to 180.18 and I experimented again a lockup today.
Of course one day without lookups is not a proof as I also had full days without lookups with 180.18 and with the stable driver.
I'm also experimenting various display corruption problems with Qt4 applications.
I reverted to 173.14.15 to see if both problems where present in that version.
This time I'm using the rpms

xorg-x11-drv-nvidia-173xx-173.14.15-2.fc10.x86_64
xorg-x11-drv-nvidia-173xx-libs-173.14.15-2.fc10.x86_64
kmod-nvidia-173xx-173.14.15-1.fc10.7.x86_64
kmod-nvidia-173xx-2.6.27.9-159.fc10.x86_64-173.14.15-1.fc10.7.x86_64

available for fedora 10.

F.

[helmotz]

UPDATE: I had freezes with 173.xx, 177.xx, 180.xx . It seems that disabling RenderAccel in xorg.conf can improve the situation. The problem is that without RenderAccel 2D performances can become really a nightmare. Moreover it seems that disabling RenderAccel has good effects on various kind of graphical corruption problems that I was experimenting in kde applications (these graphical problems appeared after 173.14.15).
I would want, if possible, to receive some feedback from NVIDIA people at least to know if they are aware of these problems.

[helmotz]

Originally Posted by helmotz

UPDATE: I had freezes with 173.xx, 177.xx, 180.xx . It seems that disabling RenderAccel in xorg.conf can improve the situation. The problem is that without RenderAccel 2D performances can become really a nightmare. Moreover it seems that disabling RenderAccel has good effects on various kind of graphical corruption problems that I was experimenting in kde applications (these graphical problems appeared after 173.14.15).
I would want, if possible, to receive some feedback from NVIDIA people at least to know if they are aware of these problems.

I installed the release 180.22 for testing seeing if the freeze still happens. Meanwhile I discovered that the graphical corruption that I was experimenting in kde application depended on the "--loose-binding" option in compiz. Removing that option those bugs and similar bug that I saw in rdesktop and openoffice disappeared. compiz-manager was responsible for starting compiz with the guilty parameters.