nvidia-driver rebooting machine on X startup

[Mel]

On Friday 07 September 2007 19:43:03 Darren Spruell wrote:
> On 9/6/07, Derek Ragona <derek at computinginnovations.com> wrote:
> >  At 03:10 PM 9/6/2007, Darren Spruell wrote:
> >
> > Wondering if this is a known issue, or if anyone has suggestions to
> >  get it working...?
> >
> >  Running FreeBSD 6.2-RELEASE-p4 SMP i386.
> >
> >  I installed nvidia-driver-100.14.11 from ports (Sep. 5 snapshot),
> >  Linux compatibility enabled. After loading modules 'linux' and
> >  'nvidia', I do:
> >
> >  # X -configure
> >  # X -config /root/xorg.conf.new
> >
> >  At this point the system spontaneously reboots. 'X -configure' runs
> >  fine and produces a configuration without issue, but the reboot
> >  occcurs when launching X. Nothing is output to screen (it flickers as
> >  if starting X), and nothing is written to /var/log/Xorg.0.log. I do
> >  get the same behavior if running 'xorgcfg'.
> >
> >  Card info:
> >
> >  "NV41 [Quadro FX 3450/4000 SDI] rev 162"
> >
> >  nvidia0 at pci7:0:0: class=0x030000 card=0x029b10de chip=0x00cd10de
> >  rev=0xa2 hdr=0x00
> >      vendor   = 'NVIDIA Corporation'
> >      class    = display
> >      subclass = VGA
> >
> >  I've tried an alternate version of the binary driver
> >  (nvidia-driver-96xx) to see if there was any change, and it also
> >  reboots on X startup.
> >
> >
> >  --
> >  Check your AGP settings and consult the readme, chapter 9, for the
> > latest driver.  You should be running the latest driver with your card.
> >
> >  Make sure you have the correct AGP settings, either letting FreeBSD do
> > it in the kernel or disable the kernel and let the Nvidia driver manage
> > it.
>
> So, I'm not having terrific luck with any combination of settings I try.
>
> The latest FreeBSD driver release from Nvidia is the same as the
> version in the ports tree. This is what I've installed.
>
> Attempting the AGPGART route (building the driver with
> "WITH_FREEBSD_AGP=yes" and disabling NvAgp support in xorg.conf) still
> results in a reboot.
>
> Going the NvAGP route (Building without support for AGPGART, setting
> NvAgp to 1 in the xorg.conf, and setting hint.agp.0.disabled=1 in
> /boot/device.hints) also results in a reboot when X starts. One thing
> I notice here; when I specify hint.agp.0.disabled=1, and check the
> output of 'kldstat -v' after booting, I still see several agp drivers
> loaded:
>
> $ kldstat -v |grep -i agp
>                 387 pci/agp_ali
>                 388 pci/agp_amd
>                 389 pci/agp_amd64
>                 390 pci/agp_ati
>                 391 pci/agp_i810
>                 392 pci/agp_intel
>                 393 pci/agp_nvidia
>                 394 pci/agp_sis
>                 395 pci/agp_via
>
> Is this a problem?
>
> I suspect a couple of problems with what I'm doing. Number one,
> there's several knobs to turn (in the driver, in the OS, etc.) and
> it's possible I'm not getting the correct combination. Other thing is,
> I've been told by a technician that this is a PCI Express video card,
> so would AGP even come into play?

Probably not. I'm seeing the same thing on one machine, no AGP on MB, and agp 
kernel modules loaded, but nvidia-driver doesn't care about it.
Let's kill all red herrings:
- Is COMPAT5X in the kernel?
- Does running `nvidia-xconfig --query-gpu-info' reboot the system?
- If you do not load the nvidia kernel module, but use Xorg's "nv" driver does 
this reboot the system?

Determine cause of reboot:
echo dumpdev="AUTO" >>/etc/rc.conf
/etc/rc.d/dumpon start

Now, if it reboots again and no core is found, the kernel didn't crash but an 
error occured in the hardware layer at which point I'd hit nvidia forums.

If there's a coredump:
http://www.freebsd.org/doc/en_US.ISO8859-1/books/developers-handbook/kerneldebug-gdb.html

and file PR.

-- 
Mel