nv problems on FreeBSD 6.1-RELEASE for AMD64

[Matthias Hopf]

On Jan 17, 07 22:43:39 -0600, Chris Radlinski wrote:
> I am able to get a stack trace.  The call to NVSync() that locks the
> system comes from somewhere in libxaa.so.  The particular function

Again, this backtrace doesn't help. It only shows which graphics command
wasn't able to be put into the queue - but in the queue there are
already a bunch of commands which are waiting for the offending one to
finish. It would be necessary to find the offending one, to see whether
the parameters are correct etc.

As NVidia doesn't publish specs, this is almost impossible for us to do
anyway.

I already tried to restart the graphics engine (modulo the commands that
are already in the queue) - so one would only have some graphics
glitches, but not a stalled system. So far I had no luck in doing this.
The engine just doesn't start again. I tried the obvious approach
(KickDMA) and some more, no luck. As it *does* start again if you
restart the Xserver, it probably *is* possible, though. That might be an
interesting approach for you if you want to try this out (maybe you have
more success in that than I had): if the queue doesn't have an empty
slot for - say - 100ms, restart the engine.

> I do have a workaround.  If I set "Options ShadowFB" in xorg.conf, I can
> run the nv driver at reasonable speeds and without problems.  I don't
> know what I'm giving up by doing that but it seems to work OK.

A better workaround would be to try all the "XaaNo*" options as
described in "man xorg.conf" - I guess there is only one you have to
specify in order to get the system running again. For some of these the
impact is pretty low. But if it happens to be XaaNoScreenToScreenCopy
you have bad luck, because that is about the most important one.

We sometimes had good success with XaaNoSolidHorVertLine and/or with
XaaNoPixmapCache and XaaNoOffscreenPixmaps.

> > > I'm not sure what that means.  Do I have a bad card?
> > Possibly. Often cards hang here if they are broken.

Still possible. Even if it works with Gentoo, they probably have a
different driver version, which results in different acceleration
functions being used.

Matthias

-- 
Matthias Hopf <mhopf at suse.de>      __        __   __
Maxfeldstr. 5 / 90409 Nuernberg   (_   | |  (_   |__          mat at mshopf.de
Phone +49-911-74053-715           __)  |_|  __)  |__  R & D   www.mshopf.de