FreeBSD-amd64 driver ?

[Gnusix]

Originally Posted by zander

I've tested the FreeBSD/amd64 driver internally on a system with 8GB of RAM and didn't see any problems specific to that system memory configuration. The Linux/x86-64 driver has been used on configurations with 64GB of RAM for a very long time, so there's a high level of confidence in the core driver. However, I haven't really had a chance to test PAE-enabled FreeBSD/i386 on large memory configurations.

Well, then maybe is a problem I have with my laptop. It is a DELL D630 with an nvidia NVS-135 with FreeBSD8.0/AMD64 STABLE.

I remember my laptop booted fine with 4Gb without X11 (I could log in text mode), but X11 was unable to start with 4Gb. So of course it is not a system problem. I'll attach the message I got with 4Gb. I checked xorg.0.log and I remember I saw nvidia module complained about memory remapping or something similar (I can't tell right now).

I remember trying some configs in /boot/loader.conf (machdep.disable_mttrs, hw.pci.allow_unsupported_io_range, hw.physmem=4G) someone suggested somewhere, but none of them did the trick.

[Gnusix]

Originally Posted by Gnusix

I remember my laptop booted fine with 4Gb without X11 (I could log in text mode), but X11 was unable to start with 4Gb. So of course it is not a system problem. I'll attach the message I got with 4Gb. I checked xorg.0.log and I remember I saw nvidia module complained about memory remapping or something similar (I can't tell right now).

This is what I get in my dmesg :

Mar  3 09:58:56 gusiport kernel: vgapci0: <VGA-compatible display> port 0xef00-0
xef7f mem 0xfd000000-0xfdffffff,0xfa000000-0xfbffffff irq 16 at device 0.0 on pc
i1
Mar  3 09:58:56 gusiport kernel: nvidia0: <Quadro NVS 135M> on vgapci0
Mar  3 09:58:56 gusiport kernel: vgapci0: child nvidia0 requested pci_enable_bus
master
Mar  3 09:58:56 gusiport kernel: vgapci0: child nvidia0 requested pci_enable_io
Mar  3 09:58:56 gusiport kernel: vgapci0: 0x10000000 bytes of rid 0x14 res 3 fai
led (0, 0xffffffffffffffff).
Mar  3 09:58:56 gusiport kernel: nvidia0: NVRM: NVIDIA MEM resource alloc failed
, BAR1 @ 0x14.
Mar  3 09:58:56 gusiport kernel: nvidia0: NVRM: NVIDIA hardware alloc failed.
Mar  3 09:58:56 gusiport kernel: device_attach: nvidia0 attach returned 6

And this is what I got in my Xorg.X.log (you'll see different hours, I can't find the one correspoding to the message in dmesg. Sorry :

(EE) Mar 03 10:22:49 NVIDIA(0): Failed to initialize the NVIDIA kernel module. P
lease see the
(EE) Mar 03 10:22:49 NVIDIA(0):     system's kernel log for additional error mes
sages and
(EE) Mar 03 10:22:49 NVIDIA(0):     consult the NVIDIA README for details.
(EE) NVIDIA(0):  *** Aborting ***

Any help would be appreciated.

[rnejdl]

These tweaks were only applicable if you were running 4GB of memory on a 32bit system:
machdep.disable_mttrs, hw.pci.allow_unsupported_io_range, hw.physmem=4G

My loader.conf is very simple:

kern.cam.scsi_delay="1000"
autoboot_delay="3"
ahci_load="YES"
linux_load="YES"
ipfw_load="YES"
ipdivert_load="YES"
accf_http_load="YES"
accf_data_load="YES"
atapicam_load="YES"
vboxdrv_load="YES"
nvidia_load="YES"

Did you verify that your card is supported with the 195.22 driver?

Rusty Nejdl
http://networking.ringofsaturn.com

[zander]

@Gnusix: please see http://www.nvnews.net/vbulletin/showthread.php?t=142391.

[rhurlin]

Yes, I can confirm last posting from zander@ with his hint to Jonathan Chen@.

I use exactly the same machine like Gnusix@: Dell D630 with 4GB and Quadro NVS 135M.

Because of a bug in the BIOS you have to set "debug.acpi.disabled=1" in /boot/loader.conf, reboot and all should work :-)

[Gnusix]

Originally Posted by rhurlin

Yes, I can confirm last posting from zander@ with his hint to Jonathan Chen@.

I use exactly the same machine like Gnusix@: Dell D630 with 4GB and Quadro NVS 135M.

Because of a bug in the BIOS you have to set "debug.acpi.disabled=1" in /boot/loader.conf, reboot and all should work :-)

@Zander, thanks for the link it did the trick with my nvidia card. I'd like to ask, will the driver work with AMD64/HEAD ?

@rhurlin : setting the mib you said did not work for me. I Don't know why but it keeps failing. I checked it with another D630 a fellow at work has and doesn't work too. Different bios revisions (mine has A03, the other has A15).

Anyway, setting what the link Zander sent breaks others things, like cardbus handling. It causes a conflict (didn't googled). Will check it sometime in the future.

Thansk !

[ChuckAtkins]

I finally resolved the FreeBSD + 64bit + Dell D630 + >= 4GB of RAM issue!

So, I'm running 8-STABLE (works w/ 8-RELEASE as well, I checked) and on a D630 which uses an NVS135M and I have 8GB of RAM. As others have experienced, Everything works just fine and dandy with 2GB but as soon as 4 or more go in then everything goes to hell. Just completely disabling ACPI causes many other problems to show up as well so that's not really a solution.

After looking through logs from the working Linux install and comparing them to the FreeBSD logs, I was able to narrow down the system resource conflict. It appears that the FreeBSD kernel is reserving the memory needed by BAR1 to another resource first, where as on Linux, the card is getting the resource first and the other system component can't get it's resources. I haven't been able to track down exactly what piece of hardware is causing the conflict, but I have been able to work around it. By setting:

debug.acpi.disabled="sysresource"

in the loader.conf, then you still get all the rest of the ACPI system but the conflicting resource is disabled. After I set that then everything worked like a champ. FreeBSD + Dell D630 + 8GB RAM + NVidia driver = happy times now.

[ChuckAtkins]

Whoops, looks like I was late to the game. I totally missed the response in the other thread addressing this. And I was so proud of myself for finding on my own too! Oh well. At least it's workable now.

http://www.nvnews.net/vbulletin/showthread.php?t=142391.

I haven't had any problems though using external monitors either.