NVidia driver misbehaving on FreeBSD
David Holm
I have now tried every combination that I could come up with and find on the FreeBSD
mailing-list, unofficial FAQs etc etc. I have tried getting the driver to work on both STABLE
and CURRENT with the exact same result on both. I have a GeForce2/GTS.
2D wise the nVidia driver seems to work great, my system has been up for 9 days without a
single problem with X. But as soon as I use nvidias OpenGL drivers strange things start
happening.
I have tried the following:
I have lowered the maximum allowed AGP rate in the BIOS from 4 to 2. (I have a via 694X chipset)
I have tried using FreeBSD's agpgart as well as using nvidias, when using FreeBSD's I changed
the #undef to #define in nv-freebsd.h. When using nvidias I have had agp.ko unloaded.
I also tried enabling the ReqAGPRate sysctl so that I was able to change it to set the speed to
2 and 1. I've also tried disabling AGP by adding option "NvAgp" "0" to my XF86Config.
I have also recompiled my kernel and XFree86 without any CPUTYPE or CFLAGS in /etc/make.conf
glxinfo seems to return correct information. I have checked to make sure I have no libGL from
Mesa etc, I have checked the existing libGL to make sure it was nvidias.
I have option "IgnoreEDID" in my XF86Config in order to be able to use 1400x1050, this was never
a problem in Linux but could this cause OpenGL to act up?
I am not loading dri and I see no errors or relevant warnings in /var/log/XFree86.0.log
I'm attaching 3 truss outputs from when glxgears failed (I hope they are useful). In ginfo.log
and ginfo3.log the computer just hung. I had to reboot with reset, when I ran CURRENT it would
hang for about 5 seconds than automatically reboot. With ginfo2.log glxgears just segfaulted.
glxinfo also caused the machine to hang once.
(the logs are gzipped, my uuencode seems broken)
Two times the reboots caused data corruption, one time it even seemed to have broken gcc (I was
compiling code at the same time as running glxgears). When trying to use gcc afterwards it would
fail with "internal compiler error...". As far as I could tell it only corrupted open files.
I have searched /var/log after crashes but I never found anything relating to nvidia or agp etc.
sysctl -a|grep nvidia returns:
hw.nvidia.agp.host-bridge.rates: 2x 1x
hw.nvidia.agp.host-bridge.fw: not supported
hw.nvidia.agp.host-bridge.sba: supported
hw.nvidia.agp.host-bridge.registers: 0x1f000203:0x00000000
hw.nvidia.agp.card.rates: 4x 2x 1x
hw.nvidia.agp.card.fw: supported
hw.nvidia.agp.card.sba: not supported
hw.nvidia.agp.card.registers: 0x1f000017:0x00000000
hw.nvidia.agp.status.status: disabled
hw.nvidia.agp.status.driver: n/a
hw.nvidia.agp.status.rate: n/a
hw.nvidia.agp.status.fw: n/a
hw.nvidia.agp.status.sba: n/a
hw.nvidia.version: NVIDIA FreeBSD x86 nvidia.ko Kernel Module 1.0-3203 Wed Oct 30 06:06:58 PST
2002 hw.nvidia.registry.ReqAGPRate: 4
hw.nvidia.registry.EnableAGPSBA: 0
hw.nvidia.registry.EnableAGPFW: 0
hw.nvidia.registry.SoftEDIDs: 1
hw.nvidia.registry.Mobile: 4294967295
hw.nvidia.cards.0.model: GeForce2 GTS
hw.nvidia.cards.0.irq: 11
hw.nvidia.cards.0.vbios: 02.15.01.07.00
hw.nvidia.cards.0.type: AGP
//David Holm, Reality Rift Studios
Get the logs here: http://www.nvnews.net/vbulletin/showthread.php?s=&threadid=6148
To Unsubscribe: send mail to majo...@FreeBSD.org
with "unsubscribe freebsd-stable" in the body of the message
James Pole
> Hi,
> I have now tried every combination that I could come up with and find on the FreeBSD
> mailing-list, unofficial FAQs etc etc. I have tried getting the driver to work on both STABLE
> and CURRENT with the exact same result on both. I have a GeForce2/GTS.
> 2D wise the nVidia driver seems to work great, my system has been up for 9 days without a
> single problem with X. But as soon as I use nvidias OpenGL drivers strange things start
> happening.
This is just my opinion, but I think you'll get a better chance of
solving your problems if you send your bug reports directly to NVIDIA
since you seem to have done everything you can do. We don't have any
access at all to the source code responsible for the OpenGL rendering so
we have no way of fixing the bugs you are having with your card.
You might get some useful suggestions here, but I think most of the bugs
are mostly in NVIDIA's OpenGL code. I myself have a NVIDIA card and I've
never had this system have an uptime of more than 2.5 days, and so far
I'm blaming NVIDIA's driver -- you're not the only one having problems.
Anyway, Good luck with your problems. Hopefully NVIDIA will make more
effort into improving their drivers.
- James
--
James Pole
ICQ: 21721828
AIM: kiwijames1986
MSN: james...@paradise.net.nz
SMS: +64-210-455-139
David Holm
James Pole <james...@paradise.net.nz> wrote:
Actually I mailed nvidia twice. They never replied to the first mail, and the second time I
mailed them (a CC of this post) it bounced =(. Sent another copy to the linux-bugs@nvid... in
hopes that they would know where to forward it.
I mailed them back when I was using linux once with a bugreport and that team was very helpful,
I got a quick reply and the guy answering me seemed very dedicated to finding a solution (the
next version solved my problem).
I just switched to FreeBSD(the day after nvidia released the
drivers) and I love it. The reason I switched was because nvidia released drivers for it. This
is really frustrating since I'm going to do some 3d development in about one month and I would
really need accelerated graphics and I don't want to switch back to linux =(.
//David Holm
David Holm
I managed to fix the problem.
It turns out that the AGP driver in the FreeBSD kernel does not have the ID for the Via 82C694X
which is an Apollo Pro 133A. Since the AGP driver supports Apollo Pro I assumed it would not
mind supporting the same arch but with a 133MHz bus.
By adding the following lines to src/sys/pci/agp_via.c the kernel detects my AGP chipset and
OpenGL now works perfectly (just add it below the Apollo Pro ID). No lockups or other strange
stuff.
case 0x06051106:
return ("VIA 82C694X (Apollo Pro 133A) host to PCI bridge");
Now, is this enough to get someone to add this to the kernel or how would I go about getting
this into the kernel. I'm guessing this would go into CURRENT before STABLE?
//David Holm
Edwin Groothuis
> Now, is this enough to get someone to add this to the kernel or how would I go about getting
> this into the kernel. I'm guessing this would go into CURRENT before STABLE?
You already did send a PR about it, that's a first step. What a
second un-official step is (for me that is) is that I go through
the file in CVS
(http://www.freebsd.org/cgi/cvsweb.cgi/src/sys/pci/agp_via.c), see
who commited something similair and then inform them about the PR.
Of course, YMWV (your mileage will vary)
Edwin
--
Edwin Groothuis | Personal website: http://www.mavetju.org
ed...@mavetju.org | Weblog: http://www.mavetju.org/weblog/weblog.php