I’m trying to upgrade a SuSE box from Leap 42.3 to Leap 15.2.
I can only get the X server to work with one GPU and 2 displays. The nvidia-settings applet initially displays the two displays on the second GPU as disabled. As soon as i enable them and insert the generated conf file into xorg.conf.d Xorg fails to start. Xorg.0.log (21.5 KB)
In OpenSuse Leap 42.3 all 4 displays on all 2 GPUs are working as expected. Xorg.0.log (44.3 KB)
Any help would be appreciated as the issue blocks an already overdue update.
Well, the X server crashes, as you stated, so sorry cannot help you there.
But your logs show this:
[ 36.018] (WW) NVIDIA: The Composite and Xinerama extensions are both enabled, which
[ 36.018] (WW) NVIDIA: is an unsupported configuration. The driver will continue
[ 36.018] (WW) NVIDIA: to load, but may behave strangely.
Maybe worth trying to disable compositing in your window manager, though that might not be very nice either.
Edit:
You setup your xorg.conf using nvidia-settings, but you didn’t provide that file.
Read the above link and compare the gathered knowledge about xinerama setup with what nvidia-settings has provided.
I will go through the suggested Wiki article and report back. However, most of the examples are mentioning just two displays. With two displays i have it working already.
I have attached my /etc/X11/xorg.conf.d/20-nvidia.conf file that is currently working in Leap 42.3: 20-nvidia.conf (4.6 KB)
I was hoping to get the xrandr output in this report.
But it doesn’t get info from X screen 0.
I can see that a default X screen config is created.
Is the X server running? Did you create the bug report via ssh?
Please show the output of xrandr --version.
I was hoping we could use this as a starting point: https://wiki.archlinux.org/index.php/Multihead#Combine_screens_into_virtual_display
As you can see above xrandr completely ignores the second GPU.
In the failed state i created the bug report from a root tty. X crashes and therefore doesn’t start at all. In the “half working” state i created the bug report from within an xterm after a ‘sudo su’.
Well odd!
In the bug report I can see the nvidia driver recognizing both GPUs and the four monitors connected to them.
I have no idea why xrandr cannot see them.
In the dmesg output I see this:
[ 75.229160] resource sanity check: requesting [mem 0x000c0000-0x000fffff], which spans more than PCI Bus 0000:00 [mem 0x000d0000-0x000dffff window]
[ 75.229473] caller _nv001015rm+0x1bf/0x1f0 [nvidia] mapping multiple BARs
But I’m uncertain if that’s connected and how exactly to deal with it.
Maybe a BIOS setting. Look for “Above 4G decoding”.
I don’t think I can help here anymore :-(
Maybe @generix has an idea?
I’ve created an xorg.conf using Xinerama. Besides the RanDr problem, would you mind trying it?
You might need to modify the layout section to your whishes.
EDIT: Sorry forgot the UseDisplayDevice options.
I guess in this case it is an Nvidia bug. The Open Source driver detects all GPUs and all monitors. The OSS driver has other problems, unfortunately :-(
Seems I’ve eventually become unlucky with my GPU selection. Burned a lot of $$.
Xinerama+nvidia driver has become crashy on newer Xservers, in your case xorg 1.18 + nvidia 390 worked fine but now you got xorg 1.20 and it crashes.
You could try two things:
I tried disabling composite with the minimal Xinerama configuration.
Xorg still crashes. I’ve spent days on getting this to work.
I’m giving up, and go buy GPUs from a different vendor.