Login loop with Ubuntu 16.04.3 with Nvidia GM204GL [Quadro M4000]

I am having trouble getting past the login loop after installing Ubuntu 16.04.3 on a Dell Precision Rack 7910 server.

As many others have had similar issues recently (https://devtalk.nvidia.com/default/topic/1024681/login-loop-with-ubuntu-16-04-with-nvidia-gtx-1060/, and https://devtalk.nvidia.com/default/topic/1027244/-solved-ubuntu-14-04-login-loop-after-installing-nvidia-384/), I tried all the suggested fixes I could find without any luck.

I started out with base Ubuntu 16.04.3 install. I noticed that I had 2 video cards: Matrox mgag200 and the Nvidia GM204GL Quadro M4000. Initially, I could login and the desktop came up as expected. When I looked at System Settings/Software & Updates/Additional Drivers, I noticed it was using the Matrox driver instead of the Nvidia. When I changed it to Nvidia and restarted the box, then the login loop began.

Some of the things I tried include:

  • sudo apt-get dist-upgrade: this put me at Ubuntu 16.04.4
  • blacklisting the mgag200 driver
  • purging nvidia* and reinstalling using sudo ubuntu-drivers autoinstall

I’m including a generated nvidia-bug report. Hopefully someone smarter than I can help shed some light on this.

Thanks in advance!

nvidia-bug-report.log.gz (207 KB)
Xorg.0.log (19.4 KB)
auth.log (42.9 KB)

Since you have two active gpus you have to point the Xserver to the right one. Connect the monitor to the nvidia and use this as xorg.conf:

Section "Device"
    Identifier     "nvidia"
    Driver         "nvidia"
    VendorName     "NVIDIA Corporation"
    BusID          "PCI:3:0:0"
    Option         "AllowEmptyInitialConfiguration"
EndSection

I really appreciate the quick response.

After adding the file contents above in /etc/X11/xorg.conf and assigning the nvidia drivers in settings, rebooting the box resulted in no login screen appearing. Just to be sure, I reinstalled the Ubuntu 16.04.3 OS from scratch and did not disable the mga driver. I then repeated the steps to add the xorg.conf file but still do not get any login screen.

I’m updating the attached result files. From the Xorg.0.log, it appears it is now loading the nvidia driver without any error, so hopefully we are now very close. The only errors I see now are in syslog, which I am also including.

syslog_2.txt (549 KB)
Xorg.0_2.log (21.3 KB)
auth_2.log (11.5 KB)
nvidia-bug-report_2.log.gz (209 KB)

Your monitor is still connected to the mga, connect it to the nvidia. Or do you want to run a remote-only setup?

No, just want to get it to use nvidia driver. I just blacklisted the mgag200 driver in /etc/modprobe.d/blacklist.conf and rebooted. I then got the login loop again, but noticed the /etc/X11/xorg.conf file had been renamed. I copied the renamed file back in place and rebooted. Now, I get no login screen again. Do you want to see the logs now—if so, I will attach.

Ubuntu thinks you’re running on an Optimus laptop, you’ll have to add the kernel parameter
nogpumanager
or disable the mga in bios. Then connect a monitor to the nvidia card and put the xorg.conf back in place.

Here is the latest log:
nvidia-bug-report.log.gz (207 KB)

X comes up fine but still no monitor connected.

[     6.341] (--) NVIDIA(GPU-0): DFP-0: disconnected
[     6.341] (--) NVIDIA(GPU-0): DFP-0: disconnected
[     6.341] (--) NVIDIA(GPU-0): DFP-0: Internal DisplayPort
[     6.341] (--) NVIDIA(GPU-0): DFP-0: 960.0 MHz maximum pixel clock
[     6.341] (--) NVIDIA(GPU-0): 
[     6.341] (--) NVIDIA(GPU-0): DFP-1: disconnected
[     6.341] (--) NVIDIA(GPU-0): DFP-1: Internal TMDS
[     6.341] (--) NVIDIA(GPU-0): DFP-1: 165.0 MHz maximum pixel clock
[     6.341] (--) NVIDIA(GPU-0): 
[     6.341] (--) NVIDIA(GPU-0): DFP-2: disconnected
[     6.341] (--) NVIDIA(GPU-0): DFP-2: Internal DisplayPort
[     6.341] (--) NVIDIA(GPU-0): DFP-2: 960.0 MHz maximum pixel clock
[     6.341] (--) NVIDIA(GPU-0): 
[     6.341] (--) NVIDIA(GPU-0): DFP-3: disconnected
[     6.341] (--) NVIDIA(GPU-0): DFP-3: Internal TMDS
[     6.341] (--) NVIDIA(GPU-0): DFP-3: 165.0 MHz maximum pixel clock
[     6.341] (--) NVIDIA(GPU-0): 
[     6.341] (--) NVIDIA(GPU-0): DFP-4: disconnected
[     6.341] (--) NVIDIA(GPU-0): DFP-4: Internal DisplayPort
[     6.341] (--) NVIDIA(GPU-0): DFP-4: 960.0 MHz maximum pixel clock
[     6.341] (--) NVIDIA(GPU-0): 
[     6.341] (--) NVIDIA(GPU-0): DFP-5: disconnected
[     6.341] (--) NVIDIA(GPU-0): DFP-5: Internal TMDS
[     6.341] (--) NVIDIA(GPU-0): DFP-5: 165.0 MHz maximum pixel clock
[     6.341] (--) NVIDIA(GPU-0): 
[     6.341] (--) NVIDIA(GPU-0): DFP-6: disconnected
[     6.341] (--) NVIDIA(GPU-0): DFP-6: Internal DisplayPort
[     6.341] (--) NVIDIA(GPU-0): DFP-6: 960.0 MHz maximum pixel clock
[     6.341] (--) NVIDIA(GPU-0): 
[     6.341] (--) NVIDIA(GPU-0): DFP-7: disconnected
[     6.341] (--) NVIDIA(GPU-0): DFP-7: Internal TMDS
[     6.341] (--) NVIDIA(GPU-0): DFP-7: 165.0 MHz maximum pixel clock
[     6.341] (--) NVIDIA(GPU-0):

ok, I added nogpumanager to my default grub file and rebooted. Monitor still doesn’t connect.
nvidia-bug-report.log.gz (207 KB)
grub.txt (1.22 KB)

The Quadro has 4 DisplayPorts, what kind of monitor did you connect and where, do you use an adapter?

Just to make things clear, this is not a software problem, it’s set up fine. The problem lies in monitor, cable, connector.

Not using the 4 display ports. Currently plugged into VGA—no adapter to VGA port of monitor.

That doesn’t work. The vga is connected to the Matrox and the mga kms driver is too rudimentary to do a display output redirected from the nvidia.

Thanks so much! Took a while to find the right cable and monitor with display port, but that works great. Wonder if I need all the custom massaging if I use the display port from the beginning—will give that a try.

Again—great info—you are the man!