Another correction. For the one workaround, switching to discrete graphics in the BIOS only seemed to get past the initial gnome login. But resuming from DPMS off/suspend, the same error loop happens even when hybrid graphics is disabled in the BIOS.
Even worse, seems hardware acceleration ends up being disabled:
$ journalctl --no-hostname -b -p warning _COMM=gnome-shell
-- Logs begin at Wed 2019-02-06 14:13:52 SAST, end at Sun 2020-05-03 16:09:45 SAST. --
May 03 14:51:40 org.gnome.Shell.desktop[2284]: glamor: 'wl_drm' not supported
May 03 14:51:40 org.gnome.Shell.desktop[2284]: Missing Wayland requirements for glamor GBM backend
May 03 14:51:40 org.gnome.Shell.desktop[2284]: Failed to initialize glamor, falling back to sw
...
May 03 14:53:25 gnome-shell[6549]: setup_framebuffers: assertion 'width > 0' failed
The "EGL failed to allocate resources" error stops happening probably because gnome-shell isn't even using hardware acceleration now.
I've also disabled all gnome shell extensions to avoid noise/errors in the logs and will try weed out the problem better.
I checked `grep -C2 -E '(EE)|(WW)' /var/log/Xorg.0.log` and things seem fine:
[ 45.181] (II) NVIDIA dlloader X Driver 440.59 Thu Jan 30 01:08:17 UTC 2020
[ 45.181] (II) NVIDIA Unified Driver for all Supported NVIDIA GPUs
--
[ 45.182] (II) LoadModule: "ramdac"
[ 45.182] (II) Module "ramdac" already built-in
[ 45.183] (WW) Falling back to old probe method for modesetting
[ 45.183] (II) NVIDIA(0): Creating default Display subsection in Screen section
"Default Screen Section" for depth/fbbpp 24/32
--
[ 46.392] (==) NVIDIA(0): Silken mouse enabled
[ 46.392] (==) NVIDIA(0): DPMS enabled
[ 46.393] (WW) NVIDIA(0): Option "PrimaryGPU" is not used
[ 46.393] (II) Loading sub module "dri2"
[ 46.393] (II) LoadModule: "dri2"
Note, I've done no customisation to X config - trying to stick to defaults.
xrandr confirms only one provider / hybrid is indeed off and Intel not in use:
So there seems to be, on current Ubuntu 18.04.4 LTS (fully up to date as of today), a 4K / UHD 60Hz issue with Nvidia, as seen in X log
[ 45.228] (II) NVIDIA GLX Module 440.59 Thu Jan 30 01:05:38 UTC 2020
Lastly, I think due to `needs_root_rights=yes` in /etc/X11/Xwrapper.config, X logs to /var/log/Xorg.0.log again instead of the X without root which used to log at ~/.local/share/xorg/Xorg.0.log
Another correction. For the one workaround, switching to discrete graphics in the BIOS only seemed to get past the initial gnome login. But resuming from DPMS off/suspend, the same error loop happens even when hybrid graphics is disabled in the BIOS.
Even worse, seems hardware acceleration ends up being disabled:
$ journalctl --no-hostname -b -p warning _COMM=gnome-shell Shell.desktop[ 2284]: glamor: 'wl_drm' not supported Shell.desktop[ 2284]: Missing Wayland requirements for glamor GBM backend Shell.desktop[ 2284]: Failed to initialize glamor, falling back to sw
-- Logs begin at Wed 2019-02-06 14:13:52 SAST, end at Sun 2020-05-03 16:09:45 SAST. --
May 03 14:51:40 org.gnome.
May 03 14:51:40 org.gnome.
May 03 14:51:40 org.gnome.
...
May 03 14:53:25 gnome-shell[6549]: setup_framebuffers: assertion 'width > 0' failed
The "EGL failed to allocate resources" error stops happening probably because gnome-shell isn't even using hardware acceleration now.
I've also disabled all gnome shell extensions to avoid noise/errors in the logs and will try weed out the problem better.
I checked `grep -C2 -E '(EE)|(WW)' /var/log/ Xorg.0. log` and things seem fine:
[ 45.181] (II) NVIDIA dlloader X Driver 440.59 Thu Jan 30 01:08:17 UTC 2020
[ 45.181] (II) NVIDIA Unified Driver for all Supported NVIDIA GPUs
--
[ 45.182] (II) LoadModule: "ramdac"
[ 45.182] (II) Module "ramdac" already built-in
[ 45.183] (WW) Falling back to old probe method for modesetting
[ 45.183] (II) NVIDIA(0): Creating default Display subsection in Screen section
"Default Screen Section" for depth/fbbpp 24/32
--
[ 46.392] (==) NVIDIA(0): Silken mouse enabled
[ 46.392] (==) NVIDIA(0): DPMS enabled
[ 46.393] (WW) NVIDIA(0): Option "PrimaryGPU" is not used
[ 46.393] (II) Loading sub module "dri2"
[ 46.393] (II) LoadModule: "dri2"
Note, I've done no customisation to X config - trying to stick to defaults.
xrandr confirms only one provider / hybrid is indeed off and Intel not in use:
$ xrandr --listproviders
Providers: number : 1
Provider 0: id: 0x218 cap: 0x1, Source Output crtcs: 4 outputs: 6 associated providers: 0 name:NVIDIA-0
I'm guessing the Lenovo P53 impliments a mux/switch to connect the laptop panel to the NVidia card so that Intel is disabled in this mode.
And double checking, also no i915 modules are loaded:
$ lsmod | grep '^drm' helper, nvidia_ drm
drm_kms_helper 180224 1 nvidia_drm
drm 491520 10 drm_kms_
So there seems to be, on current Ubuntu 18.04.4 LTS (fully up to date as of today), a 4K / UHD 60Hz issue with Nvidia, as seen in X log
[ 45.228] (II) NVIDIA GLX Module 440.59 Thu Jan 30 01:05:38 UTC 2020
Lastly, I think due to `needs_ root_rights= yes` in /etc/X11/ Xwrapper. config, X logs to /var/log/Xorg.0.log again instead of the X without root which used to log at ~/.local/ share/xorg/ Xorg.0. log