X11 segfault after NVIDIA Failed to set DPMS to on (when using accelerated google-chrome)

Over the past few weeks when using OpenSUSE tumbleweed (several recent releases) and google-chrome-beta (also several recent releases) with chrome set to Use hardware acceleration when available, if my desktop switches it’s two monitors to power-saving, when I return to using my desktop X11 segfaults as follows:

[    76.010] (--) NVIDIA(GPU-0): CRT-0: disconnected
[    76.010] (--) NVIDIA(GPU-0): CRT-0: 400.0 MHz maximum pixel clock
[    76.010] (--) NVIDIA(GPU-0): 
[    76.010] (--) NVIDIA(GPU-0): DFP-0: disconnected
[    76.010] (--) NVIDIA(GPU-0): DFP-0: Internal TMDS
[    76.010] (--) NVIDIA(GPU-0): DFP-0: 165.0 MHz maximum pixel clock
[    76.010] (--) NVIDIA(GPU-0): 
[    76.017] (--) NVIDIA(GPU-0): HP ZR24w (DFP-1): connected
[    76.017] (--) NVIDIA(GPU-0): HP ZR24w (DFP-1): Internal TMDS
[    76.017] (--) NVIDIA(GPU-0): HP ZR24w (DFP-1): 330.0 MHz maximum pixel clock
[    76.017] (--) NVIDIA(GPU-0): 
[    76.024] (--) NVIDIA(GPU-0): BenQ T2200HD (DFP-2): connected
[    76.024] (--) NVIDIA(GPU-0): BenQ T2200HD (DFP-2): Internal TMDS
[    76.024] (--) NVIDIA(GPU-0): BenQ T2200HD (DFP-2): 330.0 MHz maximum pixel clock
[    76.024] (--) NVIDIA(GPU-0): 
[  7141.971] (WW) NVIDIA(0): Failed to set the display configuration
[  7141.974] (WW) NVIDIA(0): Failed to set DPMS to on
[  7141.979] (EE) 
[  7141.979] (EE) Backtrace:
[  7141.979] (EE) 0: /usr/bin/X (xorg_backtrace+0x79) [0x559f9440e699]
[  7141.979] (EE) 1: /usr/bin/X (0x559f9425d000+0x1b5368) [0x559f94412368]
[  7141.979] (EE) 2: /lib64/libpthread.so.0 (0x7f7e9b498000+0x142d0) [0x7f7e9b4ac2d0]
[  7141.979] (EE) 3: /usr/lib64/xorg/modules/drivers/nvidia_drv.so (0x7f7e9a288000+0x7af0b) [0x7f7e9a302f0b]
[  7141.979] (EE) 4: /usr/lib64/xorg/modules/drivers/nvidia_drv.so (0x7f7e9a288000+0x4db659) [0x7f7e9a763659]
[  7141.979] (EE) 
[  7141.979] (EE) Segmentation fault at address 0x4
[  7141.979] (EE) 
Fatal server error:
[  7141.979] (EE) Caught signal 11 (Segmentation fault). Server aborting
[  7141.979] (EE) 
[  7141.979] (EE)

If I disable hardware acceleration in google chrome, then the fault does occur. I suspect having web tabs open that using media or 3D might play a part in this (I wonder if it’s because chrome also attempts to enable/disable power saving when media is active).
nvidia-bug-report.log.gz (1.23 MB)

Thanks for reporting this. I filed internal bug 2799585. In the meantime, could you please check whether this problem still occurs with the 440.44 release?

You might be able to work around this problem by disabling the new HardDPMS option in xorg.conf.

Thanks for the response and suggestions.

I upgraded to 440.44 today. I also updated google-chrome. The fault still occurs. I opened a couple of tabs to google-maps and another couple to youtube. I went away for long enough for DPMS to do its thing. On return, the desktop had terminated.

Now that I have 440.44 I’ll see whether HardDPMS makes a difference.

Thanks for checking. My coworker tracked down the crash and it should be fixed in a future release.

I’m not sure why your monitor is failing the first modeset. I wonder if it’s just taking too long to wake from low power mode. Is there an option to disable any kind of “deep sleep” mode in the monitor’s on-screen display?

That’s great news. Thanks.

Disabling 3D in chrome is a good work around for now (easy to toggle). 3D is only missed when using goggle-maps in 3D (which is quite rare).

Are you saying that if the monitor responded quicker the reported bug would be avoided? My HP ZR24w seems to take longer than the T2200HD to start itself up. It’s quite old, perhaps it just has a slow processor. The only option it supports is to turn off power saving completely, which I’m disinclined to do.

Would you mind trying it just as an experiment? It would be a good data point to know whether or not disabling the monitor’s power saving helps.

After resuming the session with power saving on the ZR24 disabled I can resume my session. But I’m seeing a lots of combinations of flickering of black horizontal-stripes, black-triangles, seemly growing from top right (on the ZR24 monitor only, other monitor is fine). Both artifacts often feature some kind of checkerboarding on their lower edge. I’ve seen this before in recent months after some restores, I wasn’t sure whether to blame the kernel or driver or interaction of both.

Using Alt-Ctrl-F1/F7 to switch to a tty console and back to X11 mostly clears the issue, but page draws in chrome are exhibiting some strangeness.

I’m going to try this test again to see in the behaviour is consistent, but not until tomorrow.

I tried again with the ZR24’s power saving disabled, as before, sessions can be resumed, but flickering corruptions of the display make it unusable. I took a photograph of corruption, if it would help I can post it here.

Logging out and starting a new session seems to return things to normal.