nvidia lockups, crashes

[Jery Wang]

After I update nvidia-glx the related kernel updates on April 01, 2005, every
time I execute konqueror from within my Gnome desktop, my system locks up
immediately after I see the konqueror window appears. Ctrl-Alt-F1, F2, etc,
Ctrl-Alt-Bkspc, Ctrl-ScrlLk, any other key does not work. But I can still move
the mouse over the screen. I have to press the reset button.

I never had problem with this before with nvidia-glx version 6629.
The latest version of nvidia-glx and nvidia-common release today (April 04,
2005) does not solve the problem.

To avoid the above problem I have to modify my /etc/X11/xorg.conf, i.e., I have
to disable:
# RenderAccel "true"

Then the lockup does not occur.

This also happens when I surf some URLs using latest mozilla-firefox 1.0.2 where
the whole system locks up. For example go to this URL:

http://cvs.apache.org/viewcvs.cgi/jakarta-commons/email/src/java/org/apache/

and click URL link on the page:

commons

-> the system locks up.

Other users have reported this problem in ubuntuforums:
http://www.ubuntuforums.org/showthread.php?p=113088#post113088

[Matt Zimmerman (mdz)]

The driver has been updated since then; do you still have problems with the
current version?

[Jery Wang (jery-wang2002)]

(In reply to comment #1)
> The driver has been updated since then; do you still have problems with the
> current version?

Yes, I update only three hours ago. And whenever I enable RenderAccel "true"
the problem occur. I can replicate exactly the same problem, it's very
consistent, i.e., konqueror always locks up the whole system.

Last time I try to check if there is anything in dmesg
There was nothing special (like error, crash, etc.)

The keyboard is totally unresponsive when it locks up.

[Leo (pizbit)]

I can confirm this problem with both 7167 and 7174.
A lengthy thread on the nvidia forums:
http://www.nvnews.net/vbulletin/showthread.php?t=31858
I've seen several in #ubuntu and elsewhere comment on this problem too and most
havn't a clue how to get around it.
As (new) users upgrade or install these drivers they're going to experience
these problems too and needless to say they wont be happy. :)
Perhaps the last known set of drivers without this problem (6629) should be used
again until some drivers come out without very serious problem?

[Jery Wang (jery-wang2002)]

(In reply to comment #3)
> I can confirm this problem with both 7167 and 7174.
> A lengthy thread on the nvidia forums:
> http://www.nvnews.net/vbulletin/showthread.php?t=31858
> I've seen several in #ubuntu and elsewhere comment on this problem too and most
> havn't a clue how to get around it.
> As (new) users upgrade or install these drivers they're going to experience
> these problems too and needless to say they wont be happy. :)
> Perhaps the last known set of drivers without this problem (6629) should be used
> again until some drivers come out without very serious problem?

Yes, and indeed it soon kills my excitement in Linux/Ubuntu although I won't
loose hope that Linux/Ubuntu will be great eventually. My RenderAccel is off now
and I am using 1600x1200 with font anti aliasing. And the performance is
significantly slower than Win$$ that also has anti-aliasing switched on. Of
course, the RenderAccel = off is to avoid locks up, but one of the reasons I use
Linux/Ubuntu is the 'stable as rock' (Linux) and it's user friendliness (Ubuntu)
- it just works.

I started using (on and off) Linux since 1999 (Redhat 5) and my biggest obstacle
in adopting linux is always lack of 'stable' peripheral support (graphics card,
printer). I am not developer, but I am keen to see Linux community grows
healthily, for that I can only contribute with submitting bugs as soon as I find
and as much as I can do to provide any useful info to fix the bug.

Btw, can I downgrade to 6629?

Thanks.
Jery

[Daniel Stone (daniels)]

We are not downgrading to 6629, which had even more serious issues on some
classes of video card. This sort of thing is an example of why it is impossible
to support binary-only drivers.

This bug has been marked as a duplicate of bug 13530.