7-current. is it possible to compile?

[edhunterbg]

I have problems with my 7900gs, and i am trying to install newest driver 100.14.09, but i get an error message:
This driver does not support FreeBSD 7.x/-CURRENT!

i tried the latest legacy driver 1.0-9639 i got the same error too.

I really have 7.0-current (sources and kernel updated about 15.6.2007). Is there any way to run this driver on 7.0.

10x in advance

[zander]

Although FreeBSD 7.0-CURRENT is not supported, the driver is expected to build against and work with this version of FreeBSD if you comment out the #error in ../src/nv-freebsd.h.

[edhunterbg]

Although FreeBSD 7.0-CURRENT is not supported, the driver is expected to build against and work with this version of FreeBSD if you comment out the #error in ../src/nv-freebsd.h.

i think already tried this. still didnt go ... but later i'll post the error because now i am using windows. i'll reboot and see the exact problem, and post again.

[edhunterbg]

here it is the last part of installing
...
...
===> lib/compat/libGLcore (install)
===> x11 (install)
===> x11/driver (install)
install: /usr/X11R6/lib/modules/drivers: No such file or directory
*** Error code 71

Stop in /usr/nvidiatemp/NVIDIA-FreeBSD-x86-100.14.09/x11/driver.
*** Error code 1

Stop in /usr/nvidiatemp/NVIDIA-FreeBSD-x86-100.14.09/x11.
*** Error code 1

Stop in /usr/nvidiatemp/NVIDIA-FreeBSD-x86-100.14.09.
root@geo:/usr/nvidiatemp/NVIDIA-FreeBSD-x86-100.14.09 on ttyv1


I am usign xorg 7.2, may be it is related to the error..

edit... it is ok now, changed some things in Makefiles

[zander]

Right, the current Makefiles don't handle the new ports very well. This will be fixed in future NVIDIA FreeBSD graphics driver releases.

[Ravna]

This might not be the proper solution, but I just changed the line :
#if __FreeBSD_version >= 700000
in ../src/nv-freebsd.h
to: #if __FreeBSD_version >= 800000
(a version that can't obviously be found at present)
and it compiled on a 7-Current - build a few hours ago - without a hitch.

My advice is to not upgrade by using any port-managment tools.
I deleted the whole /usr/X11R6 directory, and nuked any /var/db/pkg X-related entries.
(a dirty method - but we're talking pre-alpha, the whole thing might as well go titsup after the next buildworld)

But that's not the real reason for this post . . .

The reason I tried the new driver is that since June 12 07 I'm not able to shut down the X-server.
I can't even kill it with alt-ctrl-backspace.
The whole system freezes and reboots after a few seconds.
It doesn't matter which Nvidia driver is used - tried all of them - the result is the same.
The problem is reported more than once at the 7-current mailing list, so I dare say it's not my installation.
Maybe you'd get more useful information than mine by reading those posts.

This clearly isn't your fault, the drivers worked fine before 12.06.07.
It seems the FreeBSD committers messed something up.
I compared the pre-after 11.06.-sources for hours and couldn't find anything,
but then again . . . my C knowledge is even worse than my english :cool:
May I dare to ask you to have a look at the sources and tell them what's wrong ?
The xorg nv-driver is so slow, I'd rather use Linux. (shiver)

[edhunterbg]

i have sources from yesterday and running 100.14.09 and everything is ok. before that i also got problems with x.
try to csup your sources, rebuild your kernel and install nvidia's 100.14.09.

[Ravna]

I rebuild world daily, my sources are always fresh.
The drivers work, but the box feezes when I try to shutdown X.
As said above, this is discussed on the current mailing list and not a single occurance.

[zander]

@Ravna: I hadn't been aware of this problem before, thanks for the heads-up; unfortunately, there isn't much to go on in the email thread on freebsd-current (I think you were referring to the 'recent nvidia-driver trouble' thread), other than that it seems to be a panic at device teardown time.

Do you have a serial console setup that you could use to capture the panic? If not, does `cat /dev/nvidia0` (at the console, before starting X) reproduce the problem?

[zander]

@Ravna: also, have you tried the 100.14.09 graphics driver release?

[Ravna]

@Zander
At present I'm on 100.14.09
I did a buildworld with the lastest sources a few minutes ago,
and cat /dev/nvidia0 almost reproduces the problem
The box freezes, it only doesn't reboot anymore.
The output is :

Fatal trap 12: page fault while in Kernel mode
cpuid = 1 ; apic id = 01
fault virtual address = 0x26
fault code = supervisor read, page not found
instruction pointer = 0x20:0xc4b9c70c
stack pointer = 0x28:0xe744aa8c
frame pointer = 0x28:0xe744aaa0
code segment = base 0x0, limit 0xfffff,type 0x1b
= DPL 0, pres 1, def 32, gran 1
processor eflags = interrupt enabled, resume, IOPL=0

Then it tries to dump core and freezes cold - all you can do is pull the plug.

I browsed the forum and it seems whe had that before.

[zander]

Could you get a backtrace using ddb?

[Ravna]

" Could you get a backtrace using ddb?" - Huh ?

Thank goddess, the FreeBSD Handbook leaves nothing unexplained.

I'd build my Kernel as sleek as possible - which is obviously of no use right now . . .
<< building a Kernel with debugging features as we speak - gimme a couple of hours to sort this out.

[zander]

OK, great. Thanks for your help.

[Ravna]

Had to re-install the system, all those panics left the box in an unusable state.
Did a new build- kernel/world today, and it's not rebooting / freezing anymore.
The X-server starts, the Nvidia logo flashes briefly, and the server segfaults with the good old signal 11
The Xorg.log shows no other errors than that.
a cat/dev/nvidia0 only shows the same signal 11 error.

Out of interests sake i tried the 9746 driver from the ports, and it shows the old behavior
Freezing - crashing - rebooting etc.
Backtracing is not possible, coz the db> prompt is frozen and there's no core dump.
Building an X-server with debugging symbols fails at the moment.
My C - knowledge is too limited to deal with all the error messages.
I have to admit that I'm at a loss here.

( reading - learning - cursing)

[jmelo]

Hello, this is my first post in this forum and i have the same problem.

I updated my system to FreeBSD 7-CURRENT yesterday, built all xorg 7.2 ports and nvidia-driver, but when i start the X and try to go back to console (via crtl+alt+fx) the system freeze.

I also tried to install all nvidia drivers versions without success, all of them give me the same problem.

Anyone know what could be?

Thanks

[edhunterbg]

try the newest 100.14.11 i thing it fixes the problem :)

[johnjen]

I can confirm that this driver does indeed fix the issue that I too was seeing with -current (jun 20th kernel). Whew! Now I don't have to fsck 300Gb of disk each time I exit Xorg :)

Curiously as I dug into things I realized that x11/nvidia-driver is a port of the 96xx series driver but not this "bleeding edge" driver. I'm not sure why. I haven't asked the ports mailing list yet. But as Zander helpfully said, the driver compiles and runs flawlessly on -current (as of Jun 22 kernel), so the stage is set for success once 7.0-CURRENT -> 7.0-RELEASE "real soon now." :captnkill: Thanks Zander for your input on this thread!

[jmelo]

Yes, the last driver fixed the freeze problem, thanks!

[decemplex]

Ok, fine - but what's with people like me, with a GeForce 2 MX400 and limited to driver version 9631?

[Ravna]

. . not only working - it's faster, too.
Still had to symlink ../xorg/modules on a fresh X-build, but that's not an issue.

[gcc]

. . not only working - it's faster, too.
Still had to symlink ../xorg/modules on a fresh X-build, but that's not an issue.

I welcome!

Xorg 7.2
7.0 CURRENT
Geforce 2 Mx400 9639
on leaving from Xorg or "Cat /dev/nvidia0" the system is overloaded!

This problem in Xorg 7.2 or freebsd 7.0 CURRENT??? (I STABLE did not try, yet a long ago there were Xorg6.9, an update in 6.2 will occupy a lot of time...) :)
is it possible to correct? what is the matter?

log is ok

[Grogan]

I'm running 7.0-CURRENT-200706 and I did manage to get the 100.14.11 driver to compile and install by commenting out the #ifdef and creating the xorg/modules symlink.

I haven't got a problem starting, using or quitting X, but where it locks and panics for me is if I try to use the 3D accelerator (e.g. run a 3D screensaver to test). It freezes, panics and reboots. The message is:

"spin locks can only use msleep_spin"

[zander]

@Grogan: is this a recent regression? Please provide the complete panic message (if possible, with a KDB backtrace).

[Grogan]

I honestly don't know if it's a recent regression. The last time I used FreeBSD was 6.1 on an older system and the Nvidia driver at the time worked perfectly. I'm relatively new to FreeBSD (mostly a Linux user) and the only reason I installed FreeBSD from a snapshot is because the 6.2 kernel doesn't have new enough drivers for my hardware.

I didn't see the actual panic, the machine froze for several seconds and then spontaneously rebooted. After the reboot I got the message about the spinlocks (when it was saving the core) I've got a core dump, but I don't think it will be useful because I haven't got debugging options enabled and I haven't a clue how to debug a kernel in FreeBSD.

I'm about to try this again and I'll try to get better info.