gdm3 in Ubuntu 17.10 doesn't list wayland sessions when using hybrid graphics

Hm, there actually is /usr/share/wayland-sessions/ubuntu.desktop - could it be that gdm3 for some reason ignores this folder?

I enabled the debug logs for gdm3 in /etc/gdm3/custom.conf, but the logs are nowhere to be found: /var/log/gdm3 is empty.

Attaching my /etc/gdm3/custom.conf:

I disagree about the duplicate status.

This bug: gdm3 itself starts in Wayland successfully, but it doesn't list the wayland sessions (it only lists the xorg-based sessions).

The bug #1723577 is that gdm3 won't even start and needs to be reconfigured to start in Xorg mode.

Solved! I went to BIOS and switched the Video card setting from Hybrid to Discrete. Now gdm3 offered the wayland option (or, rather, both "Ubuntu" and "Ubuntu on Xorg"). I now run Wayland as confirmed by env|grep wayland:

XDG_SESSION_TYPE=wayland

I believe that nouveau driver is actually being used, so I need to look into the possibility to force Intel, to achieve better support.

Anyways, it seems that Hybrid mode won't pass gdm3 validation routines and it disables Wayland support; switching to Discrete solves the issue.

Confirmed. For me this bug happens with Radeon cards (using the 'radeon' kernel module).

And also confirmed this is not a duplicate of bug 1723577. I was trying to reproduce bug 1723577 when I encountered this one instead.

Can you please provide output from this command?;

  lspci -k

If you happen to have an nvidia card instead then this bug would become invalid. Because the nvidia driver is known to only support Xorg sessions by default. If you have a radeon card though, then this is a new issue which I can confirm.

I actually have a nvidia card, and since I switched to Discrete mode in BIOS, Wayland session works properly; moreover it seems to be working on top of Nouveau (the About dialog says that Graphics is NV117).

I have switched to the Hybrid mode. Now the only session available is the Xorg-based session; About dialog shows that the Intel® HD Graphics 530 (Skylake GT2) is now being used. Could it be that it is not compatible with Wayland?

Anyway, attaching the lspci -k of the Hybrid mode. In the hybrid mode, the Intel graphics is now available:

00:02.0 VGA compatible controller: Intel Corporation HD Graphics 530 (rev 06)
 Subsystem: Lenovo HD Graphics 530
 Kernel driver in use: i915
 Kernel modules: i915

Thanks. In that case this is not a bug. NVIDIA disables KMS support by default so Wayland sessions will be missing.

If you want to try Wayland with the proprietary NVIDIA driver then I believe you need to add:

  nvidia-drm.modeset=1

to your kernel command line...

  https://wiki.archlinux.org/index.php/NVIDIA#DRM_kernel_mode_setting

Also, hybrid support in Gnome is supposedly coming very soon:

  https://mail.gnome.org/archives/gnome-announce-list/2017-October/msg00013.html

So it should work in Ubuntu 18.04.

Thanks. I know Wayland+KMS doesn't work with NVIDIA proprietary. However, I don't use NVIDIA proprietary drivers and I don't have them installed and I don't want to use Wayland on NVIDIA proprietary. Thus, your conclusion is not right.

When I enable "Hybrid" mode in BIOS, I believe that both VGA cards (Intel and Nvidia) are made available to the system (as shown by dmesg); OS is free to use any of those. Typically Linux only uses Intel and ignores Nvidia in this mode - that is completely fine and that's what I want - to run Wayland on top of Intel.

The problem is that Wayland doesn't work with the Intel-based graphic card for me, on this very laptop when both video cards are visible. Or perhaps GDM3 detects more than one graphic card and disables Wayland support. Or perhaps GDM3 detects Skylake Intel chipset and disables Wayland support. I don't know - GDM3 logs nothing usable to its log even when in debug mode, so it is unclear what is the reason behind disabling Wayland support.

To sum it up: I want to use Wayland on Intel, but GDM3 won't allow me to do so, and it doesn't provide any hints as to why. KDE Plasma on Wayland *RUNS FINE* on this very configuration.

Thanks for the info on Hybrid mode - really nice! However, what I'd like is to run Ubuntu+Wayland on my Intel card only. So, I don't want to run it in hybrid mode, I don't want to run it on NVIDIA proprietary, just on the Intel card. KDE Plasma+Wayland runs fine and uses Intel only; GDM3 forces me to use Xorg for unknown reason. The only workaround is to "set the Discrete mode in BIOS" which only hides Intel card; then GDM3 allows me to run Wayland (on top of Nouveau). Works for me; yet I would still prefer to use just the Intel+Wayland configuration.

Please, let me know how I can make myself more clear.

I believe this is a bug in GDM3 detection algorithm which incorrectly disables Wayland support on a system which runs just fine with Wayland.

I know sorry, but didn't explain myself well enough.

You're right - changing the BIOS setting (or unplugging discrete graphics cards) is the solution right now. And soon in 18.04 you won't need to do either, when Mutter's hybrid support is working.

GDM3 will certainly hide the Wayland options if any part of the Wayland start-up crashed. It catches the crash and just hides the option from you. So maybe we need to debug that and look for a hidden crash...?

OK, please try adding these to your /etc/environment and reboot with the BIOS setup to enable hybrid:

MUTTER_VERBOSE=1
MUTTER_DEBUG=1
MUTTER_USE_LOGFILE=/tmp/mutterlogs
MUTTER_DEBUG_SM=1
MUTTER_DEBUG_BUTTON_GRABS=1
WAYLAND_DEBUG=1

Now do you get any useful information in /tmp/mutterlogs ?

Thank you for your responses, and I apologize if I sounded rude - I didn't meant to.

I rebooted in the hybrid mode and added the abovementioned lines into /etc/environment, unfortunately the /tmp/mutterlogs file is not created. There are other files though:

$ ls -la /tmp/mutter*
-rw------- 1 gdm gdm 468 okt 19 12:12 /tmp/mutter-1444-debug-log-BN1E8Y
-rw------- 1 gdm gdm 77041 okt 19 12:12 /tmp/mutter-1521-debug-log-XFRL8Y
-rw------- 1 mavi mavi 493891 okt 19 12:15 /tmp/mutter-1931-debug-log-POR97Y

I'm not sure if they would be helpful to you. I have verified that the env variables are indeed set:

$ echo $WAYLAND_DEBUG
1

I wondered that the file might be deleted by the auto-cleanup of /tmp, so I created world-writable /foo directory and set MUTTER_USE_LOGFILE=/foo/mutterlogs but nothing got created in /foo.

Please let me know if you'd need the contents of those files. Yet they contain nothing wayland-specific:

$ sudo cat /tmp/mutter* |grep -i wayl
VERBOSE: Using _NET_WM_NAME for new title of 0x3a00001 (Bug #17245): "Bug #1724583 “gdm3 in Ubuntu 17.10 doesn't list wayland sessions...” : Bugs : gdm3 package : Ubuntu - Chromium"
VERBOSE: Ignoring WM_NAME "Bug #1724583 “gdm3 in Ubuntu 17.10 doesn't list wayland sessions...” : Bugs : gdm3 package : Ubuntu - Chromium" as _NET_WM_NAME is set

cat of the first file:

$ sudo cat /tmp/mutter-1444-debug-log-BN1E8Y
VERBOSE: Mutter version 3.26.1 running on 19.10.2017
VERBOSE: Running in locale "LC_CTYPE=en_US.UTF-8;LC_NUMERIC=sk_SK.UTF-8;LC_TIME=sk_SK.UTF-8;LC_COLLATE=en_US.UTF-8;LC_MONETARY=sk_SK.UTF-8;LC_MESSAGES=en_US.UTF-8;LC_PAPER=sk_SK.UTF-8;LC_NAME=sk_SK.UTF-8;LC_ADDRESS=sk_SK.UTF-8;LC_TELEPHONE=sk_SK.UTF-8;LC_MEASUREMENT=sk_SK.UTF-8;LC_IDENTIFICATION=sk_SK.UTF-8" with encoding "UTF-8"
VERBOSE: Compiled with randr extension
VERBOSE: Compiled with startup notification

Perhaps it is as you say - GDM3 sees two graphic cards, assumes hybrid mode and simply disables Wayland right off the bat; perhaps we will need to wait until Mutter 3.27.1 is released.

I was probably wrong: those two cards are not completely separate but need to cooperate somehow (since Intel's connected to the panel while Nvidia's connected to HDMI), and perhaps Mutter needs to do some magic to support that? More info at https://bugzilla.gnome.org/show_bug.cgi?id=785381

Correction: *those two cards are not completely separate but need to cooperate somehow (since Intel's connected to the panel while Nvidia's connected to HDMI, but they need to cooperate, for example when Intel needs to draw over HDMI when an external monitor is connected)

Yes, please attach the log files.

And yes, most hybrid laptops have the Intel GPU wired to the panel. The Nvidia GPU (if enabled) just shares rendered frames with the Intel GPU. And it's the same display system regardless of whether Nvidia is enabled or disabled, so there's never any interruption to the picture on-screen as Nvidia is powered up/down.

It's also normal for the Nvidia GPU to have ownership of one of the HDMI/display ports.

Attaching mutter log; but removing lines such as these:
KEYBINDINGS: Grabbing keybinding keycode 133 mask 0x0 on 0x15a
Since it may reveal my password.

Ah, sorry. Please repeat the test but then also collect the output of:

  journalctl /usr/bin/gnome-shell

and attach it here.

Sorry, make that the output of just:

  journalctl

I can see your system choosing to use X very early...

It looks like you have a partially broken system (from upgrades/nvidia-installer/hacking/something else):

okt 20 13:13:21 mavi-vaadin gpu-manager[1130]: /etc/modprobe.d is not a file
okt 20 13:13:21 mavi-vaadin gpu-manager[1130]: /etc/modprobe.d is not a file
okt 20 13:13:21 mavi-vaadin gpu-manager[1130]: /etc/modprobe.d is not a file
okt 20 13:13:21 mavi-vaadin gpu-manager[1130]: /etc/modprobe.d is not a file
okt 20 13:13:21 mavi-vaadin gpu-manager[1130]: Error: can't open /lib/modules/4.13.0-16-generic/updates/dkms
okt 20 13:13:21 mavi-vaadin gpu-manager[1130]: update-alternatives: error: no alternatives for x86_64-linux-gnu_gfxcore_conf

So you can try rebuilding that with:

sudo apt install --reinstall kmod dkms

Later in the log, your problem becomes more apparent:

okt 20 13:13:28 mavi-vaadin gnome-shell[1666]: Can't initialize KMS backend: could not find drm kms device

That will definitely mean Wayland is not an option. Try the above fix. And if that doesn't work then please log a bug on bugzilla.gnome.org mentioning that:

okt 20 13:13:28 mavi-vaadin gnome-shell[1666]: Can't initialize KMS backend: could not find drm kms device

when you have both:

00:02.0 VGA compatible controller: Intel Corporation HD Graphics 530 (rev 06)
 Subsystem: Lenovo HD Graphics 530
 Kernel driver in use: i915
 Kernel modules: i915
01:00.0 VGA compatible controller: NVIDIA Corporation GM107GLM [Quadro M1000M] (rev a2)
 Subsystem: Lenovo GM107GLM [Quadro M1000M]
 Kernel driver in use: nouveau
 Kernel modules: nvidiafb, nouveau

Daniel, thank you so much for your time investigating this, I appreciate it.

The /etc/modprobe.d definitely exists:

$ ls -la /etc/modprobe.d/
total 76
drwxr-xr-x 2 root root 4096 okt 23 11:51 .
drwxr-xr-x 153 root root 12288 okt 23 11:51 ..
-rw-r--r-- 1 root root 2507 júl 31 2015 alsa-base.conf
-rw-r--r-- 1 root root 325 mar 13 2016 blacklist-ath_pci.conf
-rw-r--r-- 1 root root 1603 mar 13 2016 blacklist.conf
-rw-r--r-- 1 root root 210 mar 13 2016 blacklist-firewire.conf
-rw-r--r-- 1 root root 697 mar 13 2016 blacklist-framebuffer.conf
-rw-r--r-- 1 root root 156 júl 31 2015 blacklist-modem.conf
lrwxrwxrwx 1 root root 41 sep 5 2016 blacklist-oss.conf -> /lib/linux-sound-base/noOSS.modprobe.conf
-rw-r--r-- 1 root root 583 mar 13 2016 blacklist-rare-network.conf
-rw-r--r-- 1 root root 1077 mar 13 2016 blacklist-watchdog.conf
-rw-r--r-- 1 root root 127 feb 7 2017 dkms.conf
-rw-r--r-- 1 root root 154 nov 10 2015 intel-microcode-blacklist.conf
-rw-r--r-- 1 root root 347 mar 13 2016 iwlwifi.conf
-rw-r--r-- 1 root root 379 júl 28 2016 mdadm.conf
-rw-r--r-- 1 root root 104 mar 13 2016 mlx4.conf
-rw-r--r-- 1 root root 68 dec 28 2015 osspd.conf
-rw-r--r-- 1 root root 30 mar 3 2016 vmwgfx-fbdev.conf

I am typing this from my wayland-enabled session, and the journalctl also contains those lines:

okt 23 08:41:32 mavi-vaadin gpu-manager[1074]: /etc/modprobe.d is not a file
okt 23 08:41:32 mavi-vaadin kernel: new mount options do not match the existing superblock, will be ignored
okt 23 08:41:32 mavi-vaadin anacron[1079]: Will run job `cron.daily' in 5 min.
okt 23 08:41:32 mavi-vaadin anacron[1079]: Jobs will be executed sequentially
okt 23 08:41:32 mavi-vaadin gpu-manager[1074]: /etc/modprobe.d is not a file
okt 23 08:41:32 mavi-vaadin gpu-manager[1074]: /etc/modprobe.d is not a file
okt 23 08:41:32 mavi-vaadin avahi-daemon[1078]: Found user 'avahi' (UID 113) and group 'avahi' (GID 122).
okt 23 08:41:32 mavi-vaadin avahi-daemon[1078]: Successfully dropped root privileges.
okt 23 08:41:32 mavi-vaadin gpu-manager[1074]: /etc/modprobe.d is not a file
okt 23 08:41:32 mavi-vaadin avahi-daemon[1078]: avahi-daemon 0.6.32 starting up.
okt 23 08:41:32 mavi-vaadin ModemManager[1080]: <info> ModemManager (version 1.6.8) starting in system bus...
okt 23 08:41:32 mavi-vaadin gpu-manager[1074]: /etc/modprobe.d is not a file
okt 23 08:41:32 mavi-vaadin gpu-manager[1074]: Error: can't open /lib/modules/4.13.0-16-generic/updates/dkms
okt 23 08:41:32 mavi-vaadin gpu-manager[1074]: update-alternatives: error: no alternatives for x86_64-linux-gnu_gfxcore_

I have the root FS encrypted by LLVM/dmcrypt, and it couldn't have been mounted so soon, since it takes cca 2 seconds to enter my password. Could that be the reason of seeing those error messages?

Anyways, since those messages are present and I'm able to run wayland, this leads me to believe that those lines are unconnected to this issue.

Anyways, googling for the gnome-shell error message did in fact found something very interesting: https://bugzilla.redhat.com/show_bug.cgi?id=1402283

It seems that even on non-hybrid desktop system...

It worked! I switched my machine back to hybrid mode (so that both Intel and Nvidia cards are visible in lspci) and added the nouveau.modeset=0 kernel parameters. And now I am able to select the Ubuntu (wayland) option and indeed it works: XDG_SESSION_TYPE=wayland, and the About dialog shows "Intel® HD Graphics 530 (Skylake GT2)" as the Graphics card.

Thanks! So the bug is indeed in gnome-shell.

modeset=0 sounds similar to the workaround for bug 1705369. Although the symptoms are a bit different.

Status changed to 'Confirmed' because the bug affects multiple users.

I had the same problem with Intel + AMD tandem on a desktop. I solved it by adding in /etc/environment:

MUTTER_ALLOW_HYBRID_GPUS=1

Adrian, thank you very much. Your advice saved me a lot of money.

I unfortunately own an MSI GX70 notebook with two GPUS: Radeon HD 8650 G Richland and Radeon HA 8970 G Neptune.

It wouldn't load without nomodest option in grub configuration. While fglrx lasted, it was bearable. After 16.04, however, nomodeset started to block installing amdgpu driver.

With MUTTER_ALLOW_HYBRID_GPUS=1, however, the notebook booted without nomodeset, also it switched to integrated Radeon HD 8650 G Richland automatically.

I will still have to figure out how to switch to the discrete GPU when it is required but the notebook is already getting to be usable again after one and a half years.

I am using XPS 13 9360, it happens often that the login session goes without wayland.

Here is the log I got from running journalctl /usr/bin/gnome-shell

Dec 15 12:45:45 cico-XPS-13-9360 gnome-shell[883]: Failed to apply DRM plane transform 0: Permission denied
Dec 15 12:45:45 cico-XPS-13-9360 org.gnome.Shell.desktop[883]: (EE)
Dec 15 12:45:45 cico-XPS-13-9360 org.gnome.Shell.desktop[883]: Fatal server error:
Dec 15 12:45:45 cico-XPS-13-9360 org.gnome.Shell.desktop[883]: (EE) wl_drm@4: error 0: authenicate failed
Dec 15 12:45:45 cico-XPS-13-9360 org.gnome.Shell.desktop[883]: (EE)
Dec 15 12:45:46 cico-XPS-13-9360 gnome-shell[883]: X Wayland crashed; aborting
Dec 15 12:45:47 cico-XPS-13-9360 gnome-shell[1030]: JS WARNING: [resource:///org/gnome/shell/ui/main.js 315]: reference to undefined property "MetaStage"
Dec 15 12:45:47 cico-XPS-13-9360 gnome-shell[1030]: JS WARNING: [resource:///org/gnome/shell/ui/layout.js 221]: reference to undefined property "MetaWindowGroup"
Dec 15 12:45:47 cico-XPS-13-9360 gnome-shell[1030]: JS WARNING: [resource:///org/gnome/shell/ui/osdMonitorLabeler.js 59]: reference to undefined property "MetaDBusDisplayConfig
Dec 15 12:45:48 cico-XPS-13-9360 gnome-shell[1030]: JS WARNING: [resource:///org/gnome/shell/ui/slider.js 38]: reference to undefined property "CallyActor"
Dec 15 12:45:48 cico-XPS-13-9360 gnome-shell[1030]: JS WARNING: [resource:///org/gnome/gjs/modules/tweener/tweener.js 540]: reference to undefined property "isSpecialProperty"
Dec 15 12:45:48 cico-XPS-13-9360 gnome-shell[1030]: Error looking up permission: GDBus.Error:org.freedesktop.DBus.Error.ServiceUnknown: The name org.freedesktop.impl.portal.Per
Dec 15 12:45:48 cico-XPS-13-9360 org.gnome.Shell.desktop[1030]: Window manager warning: "XF86RFKill" is not a valid accelerator
Dec 15 12:45:49 cico-XPS-13-9360 gnome-shell[1030]: JS WARNING: [resource:///org/gnome/shell/ui/layout.js 29]: reference to undefined property "MetaWindowX11"
Dec 15 12:46:02 cico-XPS-13-9360 gnome-shell[1809]: Failed to read monitors config file '/home/cico/.config/monitors.xml': Logical monitor scales must be identical
Dec 15 12:46:03 cico-XPS-13-9360 gnome-shell[1809]: JS WARNING: [resource:///org/gnome/shell/ui/main.js 315]: reference to undefined property "MetaStage"
Dec 15 12:46:03 cico-XPS-13-9360 gnome-shell[1809]: JS WARNING: [resource:///org/gnome/shell/ui/layout.js 221]: reference to undefined property "MetaWindowGroup"
Dec 15 12:46:03 cico-XPS-13-9360 gnome-shell[1809]: JS WARNING: [resource:///org/gnome/shell/ui/osdMonitorLabeler.js 59]: reference to undefined property "MetaDBusDisplayConfig
Dec 15 12:46:03 cico-XPS-13-9360 gnome-shell[1809]: JS WARNING: [resource:///org/gnome/shell/ui/slider.js 38]: reference to undefined property "CallyActor"

Thank you for reporting this bug to Ubuntu.
Ubuntu 17.10 (artful) reached end-of-life on July 19, 2018.

See this document for currently supported Ubuntu releases:
https://wiki.ubuntu.com/Releases

We appreciate that this bug may be old and you might not be interested in discussing it any more. But if you are then please upgrade to the latest Ubuntu version and re-test. If you then find the bug is still present in the newer Ubuntu version, please add a comment here telling us which new version it is in and change the bug status to Confirmed.