Description of problem: AMDGPU driver will fail to launch the graphical login screen (gdm). It is possible to switch to another tty console but launch of gdm.service will fail. (systemctl start gdm.service will flash text on display for maybe about one minute before giving up) Version-Release number of selected component (if applicable): kernel-desktop-4.14.44-2.mga7-1-1.mga7 How reproducible: always Steps to Reproduce: 1. Boot to Mageia with default grub options 2. UI freezes with some text without launching gdm login ui 3. Switch to another tty console with ctrl+f2 4. switch to root (sudo su) 5. systemctl start gdm.service
Created attachment 10208 [details] dmesg from 4.14 based kernel error with amd related boot options added to grub.
Tried to add kernel command line options mentioned on bugs #22284 and #22339 + amdgpu.exp_hw_support=1 option that was requested on dmesg warning when trying without that. So full kernel command line is: BOOT_IMAGE=/vmlinuz-4.14.44-desktop-2.mga7 root=UUID=5e30cfb5-d5ee-480f-bc29-759b0159b4c4 ro nokmsboot amdgpu.exp_hw_support=1 radeon.si_support=0 radeon.cik_support=0 amdgpu.si_support=1 amdgpu.cik_support=1 noiswmd resume=UUID=89590750-eded-4900-b65b-29201c9a48a3 audit=0 vga=791 With this setup the display (over hdmi to 4k tw) freezes totally and it is not possible to switch to another tty console with ctrl+f2. Login with ssh still however works ok.
Created attachment 10209 [details] boot attempt with own build 4.17rc7+ kernel Tried also to add boot options from https://www.reddit.com/r/Amd/comments/8e2fyb/amd_ryzen_2400g_on_arch_linux_gnomewayland/ but still same ui freeze. I can not switch to tty consoles with ctrl+f2 but ssh to box works ok.
Created attachment 10210 [details] working dmesg for amdgpu driver on 2400g with 4.17-rc7+ kernel Got the amdgpu working (display+audio tested) over HDMI with the 4.17.rc7+ kernel. The fix was to enable new kernel option CONFIG_DRM_AMD_DC_DCN1_0=y and then rebuilding the kernel. This option was not available yet on 4.14 kernels. Default grub kernel boot options worked ok, so no need to change them. 0.000000] Linux version 4.17.0-rc7+ (lamikr@localhost) (gcc version 7.3.0 (Mageia 7.3.0-5.mga7)) #2 SMP Thu May 31 01:03:53 PDT 2018 [ 0.000000] Command line: BOOT_IMAGE=/vmlinuz-4.17.0-rc7+ root=UUID=5e30cfb5-d5ee-480f-bc29-759b0159b4c4 ro nokmsboot splash quiet noiswmd resume=UUID=89590750-eded-4900-b65b-29201c9a48a3 audit=0 vga=791
Is this the same hardwar as the one on which you hit bug 23101 ? If so, then please put a reference to this report in that report.
Assignee: bugsquad => kernelCC: (none) => marja11
s/hardwar/hardware/
Yes, it's same (new) hardware on both bugs. And most of the additions and fixes required for supporting this AMD/ATI gpu is coming only to 4.17 kernel, so it's actually not suprise that 4.14 kernels can launch UI with amdgpu driver. But with 4.17 kernel I can get the UI up with amdgpu driver by launching the gddm or xdm manyally with systemctl start command. So this is good information for you to know if you are able to jump to that kernel on Cauldron. But even with the usage of 4.17 has not resolved me the bug 23101.
The following is meant as workaround with mga6's 4.14.x kernel FYI: Ryzen 5 2400G works OK with fbdev driver (but needs to have xorg.conf, for some reason auto-config won't work). But to get resolutions higher than 1024x768 you need to configure different resolution for grub/set an invalid one that will cause auto-detection - X (Gnome/Wayland) will use whatever the frambuffer was during boot. Drawback is that you can only drive a single display with that, but works with VGA (correctly detected 1680x1050 of the monitor), HDMI (1920x1080, attached to AVR/TV) and Displayport (2560x1044, my actual Monitor). Audio via HDMI didn't work though (but I'm using USB-Speakers, so don't care about that for now) So as workaround: xorg.conf → set to use fbdev driver: Section "Device" Identifier "device1" Driver "fbdev" Option "DPMS" EndSection /etc/default/grub → adjust GRUB_GFXMODE GRUB_GFXMODE=1920x1080x32 (note that this apparently is an invalid value/none of the default vesa modes and effectively triggers autodetection, but with the default mode of 1024x768x32 it is stuck with that - if you want to play it safe/be explicit, pick "auto" or use grub's "videoinfo" command to list supported modes - it is possible to supply multiple modes separated by comma or semicolon that will be tried in sequence) Don't forget to run update-grub2 afterwards to apply the changes.
CC: (none) => lohmaier+mageia
Wich is why we will provide a 4.18.12 based kernel in backports :=) https://bugs.mageia.org/show_bug.cgi?id=23643
CC: (none) => tmb
hooray :-)
See Also: (none) => https://bugs.mageia.org/show_bug.cgi?id=23643
FYI: while amdgpu kernel driver from 4.18.12 works and xorg.conf is no longer needed (going straight to xwayland), it is still is using basic graphics modes only/no hardware acceleration using vdpau/vaapi (as driver complains it would need llvm 5 for that – "amdgpu: LLVM 5.0 is required, got LLVM 3.9")
The same issue for AMD Ryzen 2200G (Radeon Vega 8) https://bugs.mageia.org/show_bug.cgi?id=23766
CC: (none) => a.dolata
Closing OLD, Mageia 7 reached EOL in June 2021. Support for this AMDGPU is available in Mageia 8/9.
Resolution: (none) => OLDStatus: NEW => RESOLVED