Kernel 4.4.172 on Slack64 and Nvidia

imitis · 02-03-2019, 06:34 AM

No hope for 32bit users

kingbeowulf · 02-03-2019, 11:16 AM

I'm playing catch-up after a week in the UK for my corporate overlord. As soon as I recover from jet lag (getting too damn old for 10+ hr flights), I'll update my VMs and have a look. Nvidia seems to be concentrating on kernel 4.20 fixes.

Yes, 32-bit is DEAD other than for COMPAT32

Yes, Nvidia is pretty much going to kill off legacy cards. If nouveau works, use that. Although I have a few old cards, I have nothing to plug them into and test.

FlinchX · 02-03-2019, 12:57 PM

Quote:

Originally Posted by kingbeowulf

Yes, Nvidia is pretty much going to kill off legacy cards.

Legacy as in how legacy?

I am using nvidia-legacy390 for GT 730 but I assume this is legacy in context of SBo, to have a package different than the 4xx series. My video card is a few years old, but it's by no means ancient, yet I am worried about not being able to update the kernel if NVidia won't backport the fixes.

This seems to be the relevant thread in their developer forum https://devtalk.nvidia.com/default/t...t_user_pages-/ (and that's probably ponce of Slackware asking the relevant question?) so far looks like they did not comment on their decision regarding this.

camorri · 02-03-2019, 01:38 PM

Flinchx, I'm running NVIDIA-Linux-x86_64-418.30 on a GT 710, and the GT 730 is listed as supported. So, you can use the beta driver now, if you choose to. Its not shown any problems on my system.

FlinchX · 02-03-2019, 01:47 PM

Quote:

Originally Posted by camorri

Flinchx, I'm running NVIDIA-Linux-x86_64-418.30 on a GT 710, and the GT 730 is listed as supported. So, you can use the beta driver now, if you choose to. Its not shown any problems on my system.

Thanks, this means I misunderstood something in the past. I am following the SBo mailing list and when nividia-390 got a legacy tag, for some reason I thought I had to stick with that.

camorri · 02-03-2019, 01:53 PM

Just understand, what I posted is an Nvidia blob run file. It is not a Slackbuild, so when you install it, you will not see any evidence of it in /var/log/packages. I have been using the Nvidia blobs for a long time, I found them easy to use, and much more up to date than the SBo slackbuilds. Just my opinion, you will find many liking the Slackbuild approach better.

FlinchX · 02-03-2019, 02:24 PM

Quote:

Originally Posted by camorri

Just understand, what I posted is an Nvidia blob run file. It is not a Slackbuild, so when you install it, you will not see any evidence of it in /var/log/packages. I have been using the Nvidia blobs for a long time, I found them easy to use, and much more up to date than the SBo slackbuilds. Just my opinion, you will find many liking the Slackbuild approach better.

I have used the blob installer directly in the past but a couple of years ago switched to SBo buildscripts. I don't remember why I did it and I don't want to speculate on it, there's a chance that I did it because the run file provided by NVidia didn't work out of the box, while the SBo buildscript handled things fine by applying a patch. Not to mention that SBo has the useful nvidia-switch script.

kingbeowulf · 02-04-2019, 12:08 AM

Flinchx,
The nvidia-410 series supports GTX 600 series GPUs and newer (see supported products docs for Quadro cards). Nvidia-390 supports GTX 400 series and newer (but not newest RTX 2000 series). That is why SBo moved nvidia-390 to legacy status: nvidia dropped support for GTX 500 and older cards.

https://www.nvidia.com/object/IO_32667.html

Now, when, and if, Nvidia will back port kernel fixes to the legacy series is anyone's guess. So far I haven't found a good patch fot the kernel 4.4.172 compile issue. It is fixed in 418.30 beta but I need to test that one a bit more.

If nouveau has suffient support for your GPU, I recommend nouveau instead of nvidia for the older cards (< GTX 600).

akimmet · 02-05-2019, 11:51 AM

In my opinion both nVidia and the Linux kernel are at fault for this, not Slackware. Since this was a security related upgrade I believe Pat made the right call.
nVidia's fault for always being behind on kernel development.
Linux kernel devs for sometimes going out of there way to break or limit close source drivers.

I'll be holding off upgrading the kernel on my machine running 14.2 until there are some non-beta drivers that fix this.

I have switched to AMD for my other machine that is on -current. AMD's drivers have gotten much better in the last few years. In my experience RX 5XX series and earlier now work without issue. Unfortunately, Vega is still a little buggy for some people yet.

kingbeowulf · 02-05-2019, 09:23 PM

I'm keeping on eye on AMD as well. It would be spiffy to be able to feed my gaming addiction without wrestling with nvidia.

I've submitted an update to SBo. I'm running 418.30 on kernel 4.4.172 now and so far so good.

Code:

$ nvidia-smi   
Tue Feb  5 19:22:38 2019       
+-----------------------------------------------------------------------------+
| NVIDIA-SMI 418.30       Driver Version: 418.30       CUDA Version: 10.1     |
|-------------------------------+----------------------+----------------------+
| GPU  Name        Persistence-M| Bus-Id        Disp.A | Volatile Uncorr. ECC |
| Fan  Temp  Perf  Pwr:Usage/Cap|         Memory-Usage | GPU-Util  Compute M. |
|===============================+======================+======================|
|   0  GeForce GTX 106...  Off  | 00000000:03:00.0  On |                  N/A |
| 48%   29C    P8     6W / 120W |     91MiB /  6078MiB |      0%      Default |
+-------------------------------+----------------------+----------------------+
|   1  GeForce GTX 660     Off  | 00000000:04:00.0 N/A |                  N/A |
| 40%   30C    P8    N/A /  N/A |     73MiB /  1999MiB |     N/A      Default |
+-------------------------------+----------------------+----------------------+
                                                                               
+-----------------------------------------------------------------------------+
| Processes:                                                       GPU Memory |
|  GPU       PID   Type   Process name                             Usage      |
|=============================================================================|
|    0      1517      G   /usr/libexec/Xorg                             88MiB |
|    1                    Not Supported                                       |
+-----------------------------------------------------------------------------+

NathanBarley · 02-07-2019, 04:53 PM

Quote:

Originally Posted by kingbeowulf

I'm keeping on eye on AMD as well. It would be spiffy to be able to feed my gaming addiction without wrestling with nvidia.

I've submitted an update to SBo. I'm running 418.30 on kernel 4.4.172 now and so far so good.

Code:

$ nvidia-smi   
Tue Feb  5 19:22:38 2019       
+-----------------------------------------------------------------------------+
| NVIDIA-SMI 418.30       Driver Version: 418.30       CUDA Version: 10.1     |
|-------------------------------+----------------------+----------------------+
| GPU  Name        Persistence-M| Bus-Id        Disp.A | Volatile Uncorr. ECC |
| Fan  Temp  Perf  Pwr:Usage/Cap|         Memory-Usage | GPU-Util  Compute M. |
|===============================+======================+======================|
|   0  GeForce GTX 106...  Off  | 00000000:03:00.0  On |                  N/A |
| 48%   29C    P8     6W / 120W |     91MiB /  6078MiB |      0%      Default |
+-------------------------------+----------------------+----------------------+
|   1  GeForce GTX 660     Off  | 00000000:04:00.0 N/A |                  N/A |
| 40%   30C    P8    N/A /  N/A |     73MiB /  1999MiB |     N/A      Default |
+-------------------------------+----------------------+----------------------+
                                                                               
+-----------------------------------------------------------------------------+
| Processes:                                                       GPU Memory |
|  GPU       PID   Type   Process name                             Usage      |
|=============================================================================|
|    0      1517      G   /usr/libexec/Xorg                             88MiB |
|    1                    Not Supported                                       |
+-----------------------------------------------------------------------------+

Already very decent, my friend.

kingbeowulf · 02-09-2019, 02:57 PM

For those monitoring this thread, see
https://www.linuxquestions.org/quest...ml#post5959634
for 418.30 issues running 2 different nvidia GPUs.

FlinchX · 02-14-2019, 04:32 AM

Quote:

Originally Posted by camorri

Flinchx, I'm running NVIDIA-Linux-x86_64-418.30 on a GT 710, and the GT 730 is listed as supported. So, you can use the beta driver now, if you choose to. Its not shown any problems on my system.

I've just finally updated 14.2 to the latest 4.4.172 kernel then got the 418.30 blob via SBo. There was a minor trouble with the download link, but I follow the mailing list and I was aware of it, so I just fixed the link manually in the info file and everything worked ... but I've still ended with a toasted system

I have no idea how do you manage to run this for GT 710 while GT 730 is clearly not supported - I've got triple slapped with warnings about it.

I do indeed have a GT 730 GPU:

Code:

# lspci -v | grep GT  
01:00.0 VGA compatible controller: NVIDIA Corporation GF108 [GeForce GT 730] (rev a1) (prog-if 00 [VGA controller])
	Subsystem: ZOTAC International (MCO) Ltd. GF108 [GeForce GT 730]

First slap: came from sbopkg-build-log at package build time:

Code:

WARNING: The NVIDIA GeForce GT 730 GPU installed in this system is
         supported through the NVIDIA 390.xx legacy Linux graphics drivers.
         Please visit http://www.nvidia.com/object/unix.html for more
         information.  The 418.30 NVIDIA Linux graphics driver will ignore
         this GPU.

Second slap came after I did reboot the system, below is the relevant snippet from the output of dmesg:

Code:

[   23.786454] NVRM: The NVIDIA GeForce GT 730 GPU installed in this system is
               NVRM:  supported through the NVIDIA 390.xx Legacy drivers. Please
               NVRM:  visit http://www.nvidia.com/object/unix.html for more
               NVRM:  information.  The 418.30 NVIDIA driver will ignore
               NVRM:  this GPU.  Continuing probe...
[   23.788988] NVRM: No NVIDIA graphics adapter found!

Third slap came when I actually tried to execute startx (at this point I was already almost sure it won't work), below is the relevant snippet from Xorg.0.log:

Code:

[   308.996] (WW) NVIDIA(0): The NVIDIA GeForce GT 730 GPU installed in this system is
[   308.996] (WW) NVIDIA(0):     supported through the NVIDIA 390.xx Legacy drivers. Please
[   308.996] (WW) NVIDIA(0):     visit http://www.nvidia.com/object/unix.html for more
[   308.996] (WW) NVIDIA(0):     information.  The 418.30 NVIDIA driver will ignore this
[   308.996] (WW) NVIDIA(0):     GPU.  Continuing probe... 
[   308.996] (EE) No devices detected.

What are my options now? I have reverted back to nouveau as a temporary solution. The image quality feels slightly worse. But I still need to run software that requires the blob. Is my only option to upgrade the GPU now?

Wheelerof4te · 02-14-2019, 06:06 AM

^Reasons like those are why you shouldn't support non-free software. Planned obsolescence is almost never an issue with open-drivers. Your only options are to continue using older, working kernel or upgrade to the latest while using nouveau.
There should be some nouveau improvements in new kernels.

FlinchX · 02-14-2019, 06:41 AM

I don't understand what's going on

https://download.nvidia.com/XFree86/...rtedchips.html

lists GT 730 as supported, just as camorri said above

Edit: I have just spent more than one hour waiting in nvidia's live chat queue to ask about this. Their support staff member immediately told me that they don't support Linux in live chat and redirected me to the forum instead. So it looks like nothing has changed since Linus did show them the middle finger - they are still hostile towards Linux and don't admit it on their site. If https://www.nvidia.com/object/support.html would've stated they don't support Linux in live chat, I wouldn't bother to wait for so long to talk to somebody there.