Black Screen on eDP-1 After Aquamarine Update in Hyprland with AMD/NVIDIA Hybrid GPU Setup #8308
Comments
|
have you tried setting |
|
Update:
|
when you build in debug mode Hyprland will used hyprlandd.log, if you build in release mode it will use hyprland.log
logs are kinda weird, but it seems like you have outdated or the wrong gpu drivers as a sanity check can you make sure other compositors still currently work? (sway, plasma wayland, GNOME wayland) |
|
The NVIDIA drivers are version 560.31.02, while the AMDGPU driver is included with Linux, currently at version 6.11.2. To my understanding, these are fairly recent versions, and I don't believe the NVIDIA drivers should play any significant role here. I've explicitly set Aquamarine to utilize the integrated graphics, but upon reviewing the logs, it seems that the NVIDIA GPU is still somehow being considered. Plasma works, but unless the NVIDIA drivers are blacklisted, it always defaults to the dGPU. Reports from others suggest this is true for other DEs/WM as well. I am confident this is a quirk specific to my device. For reference, similar issues are discussed in Weird ordering of PCI bus, KWin’s response, Gnomes heuristic for gpu selection. One of Hyprland's most attractive feature for me was that it allowed me to bypass this automatic selection by allowing me to select the gpu of preference via BIOS is set to hybrid. And that worked pre-0.42. By the way, I have a new set of logs after changing the compilation back to |
|
I'm seeing the same issue |
|
I'm seeing a similar issue. Laptop: ASUS TUF Everything worked fine in the last release of 0.44.1 (arch linux's extra repo). |
|
I'm having a similar issue Laptop: ASUS ROG Strix Scar 18 hyprctl systeminfo -cHyprland 0.45.0 built from branch main at commit 1930a95 (shm: fix shm fd size check before creating or resizing shm_pool (8572)). Date: Mon Nov 25 01:50:35 2024 Tag: v0.45.0-62-g1930a950, commits: 5495 built against aquamarine 0.5.0flags set: System Information: GPU information: os-release: NAME="CachyOS Linux" plugins: ======Config-Start====== Config File: /home/mrgeek/.config/hypr/hyprland.conf: Read Succeeded Please note not all available settings / options are set here.For a full list, see the wikienv = HYPRCURSOR_THEME,rose-pine-hyprcursor source = ~/.config/hypr/monitors.conf source = ~/.config/hypr/env.confSlow app launch fixexec-once = systemctl --user import-environment & ## Idle configurationexec-once = swayidle -w timeout 300 'swaylock -f -c 000000' before-sleep 'swaylock -f -c 000000' Source a file (multi-file configs)source = ~/.config/hypr/myColors.conf## Input configurationdebug { cursor { input { follow_mouse = 1 # 0|1|2|3 touchpad { repeat_rate=60 sensitivity = -0.3 sensitivity = 1.3accel_profile = flat general { } decoration { } Blur for waybar#blurls = waybar animations { } See https://wiki.hyprland.org/Configuring/Dwindle-Layout/ for moredwindle { See https://wiki.hyprland.org/Configuring/Master-Layout/ for moremaster { misc { gestures { device { See https://wiki.hyprland.org/Configuring/Keywords/ for more$mainMod = SUPER Example binds, see https://wiki.hyprland.org/Configuring/Binds/ for morebind = $mainMod, RETURN, exec, kitty bind = $mainMod, SPACE, exec, wofibind = $mainMod, F, fullscreen, 1 bind = $mainMod, G, togglegroup, bind = $mainMod SHIFT, G,exec,hyprctl --batch "keyword general:gaps_out 5;keyword general:gaps_in 3" bind = $mainMod , G,exec,hyprctl --batch "keyword general:gaps_out 0;keyword general:gaps_in 0"bind = $mainMod , G,exec, rofi -show cr -modi "cr:chromrofi --order-by visit_count -l 50 -s" -config ~/.config/rofi/themes/catppuccin.rasibind = CTRL SHIFT, SPACE, exec, makoctl dismiss Volume controlbind=,XF86AudioLowerVolume,exec,pamixer -ud 3 && pamixer --get-volume > /tmp/$HYPRLAND_INSTANCE_SIGNATURE.wob mute soundbind=,XF86AudioMute,exec,amixer sset Master toggle | sed -En '/[on]/ s/.[([0-9]+)%]./\1/ p; /[off]/ s/.*/0/p' | head -1 > /tmp/$HYPRLAND_INSTANCE_SIGNATURE.wob Playback controlbind=,XF86AudioPlay,exec, playerctl play-pause Screen brightnessbind = , XF86MonBrightnessUp, exec, brightnessctl s +5% bind = $mainMod SHIFT, P, exec, qalculate-gtk bind = $mainMod, O, exec, killall -SIGUSR2 waybar Move focus with mainMod + arrow keysbind = $mainMod, H, movefocus, l Switch workspaces with mainMod + [0-9]bind = $mainMod, 1, workspace, 1 bind = $mainMod, minus, movetoworkspace,special bind = $mainMod SHIFT,H ,movewindow, l Move active window to a workspace with mainMod + CTRL + [0-9]bind = $mainMod CTRL, 1, movetoworkspace, 1 same as above, but doesnt switch to the workspacebind = $mainMod SHIFT, 1, movetoworkspacesilent, 1 Scroll through existing workspaces with mainMod + scrollbind = $mainMod, mouse_down, workspace, e+1 binds { bind = $mainMod,R,submap,resize bind=CTRL SHIFT, left, resizeactive,-15 0bind=CTRL SHIFT, right, resizeactive,15 0bind=CTRL SHIFT, up, resizeactive,0 -15bind=CTRL SHIFT, down, resizeactive,0 15bind=CTRL SHIFT, l, resizeactive, 15 0bind=CTRL SHIFT, h, resizeactive,-15 0bind=CTRL SHIFT, k, resizeactive, 0 -15bind=CTRL SHIFT, j, resizeactive, 0 15Move/resize windows with mainMod + LMB/RMB and draggingbindm = $mainMod, mouse:272, movewindow #exec-once = alacritty auto start#------------# Float Necessary Windowswindowrule=float,Rofi Increase the opacitywindowrule=opacity 0.92,Thunar #---------------# windows rules#---------------# Config File: /home/mrgeek/.config/hypr/monitors.conf: Read Succeeded Generated by nwg-displays on 2024-11-26 at 12:36:36. Do not edit manually.monitor=eDP-2,[email protected],2560x0,1.0monitor = eDP-2, preferred, 0x0, 1 Config File: /home/mrgeek/.config/hypr/workspaces.conf: Read Succeeded Generated by nwg-displays on 2024-08-12 at 08:57:26. Do not edit manually.workspace=1,monitor:HDMI-A-1,default:true Config File: /home/mrgeek/.config/hypr/nvidia.conf: Read Succeeded env = GBM_BACKEND,nvidia-drm # was commentedenv = GBM_BACKEND,nvidia env = __NV_PRIME_RENDER_OFFLOAD,1 env = __GL_GSYNC_ALLOWED,1 # was commentedenv = __GL_VRR_ALLOWED,1 env = __VK_LAYER_NV_optimus,NVIDIA_only # was commentedenv = NVD_BACKEND,direct # was commentedenv = ELECTRON_ENABLE_WAYLAND,1 env = AQ_NO_MODIFIERS,0 env = AQ_DRM_DEVICES,/dev/dri/card0:/dev/dri/card1 env = AQ_DRM_DEVICES,/dev/dri/card1======Config-End======== Hyprland --systeminfoHyprland 0.45.0 built from branch main at commit 1930a95 (shm: fix shm fd size check before creating or resizing shm_pool (8572)). flags set: System Information: GPU information: os-release: NAME="CachyOS Linux" plugins: hyprctl monitorsMonitor HDMI-A-1 (ID 0): Monitor eDP-2 (ID 1): snippet of $XDG_RUNTIME_DIR/hypr/1930a95000d336b76d18c0c95ef77e138c9a4cd0_1732625874_854059302/hyprland.log | grep ERRI'm noticing even more strange issues when i set these environment variables:
|
|
Same issue, same hardware as @ottersome |
|
@ottersome is your problem that Hyprland crashes when setting AQ_DRM_DEVICES to amd iGPU? Or that eDP-1 is blank? if so it's probably because only the dGPU exists according to crash report (below output is from and based on the tty logs you gave, it could be that the iGPU driver (mesa) is installed incorrectly causing Hyprland to not be able to use it, and therefore not starting since it's the only whitelisted gpu either way, if you can give a full trace hyprland.log that would be helpful to diagnose the issue |
|
No crashing happening. I can confirm by using my keybindings to access kitty and touch a file. The lspci thing: this goes back to the problem I mentioned earlier where dGPU gets marked as VGA. I hope I don't come off as pedantic but I am quite sure it is related to the problem I linked earlier. Which apparently many systems take as a heuristic to chose the integrated graphics. ➜ ~ lspci -vnn | grep -E 'VGA|Display|3D'
pcilib: Error reading /sys/bus/pci/devices/0000:00:08.3/label: Operation not permitted
65:00.0 VGA compatible controller [0300]: NVIDIA Corporation AD107M [GeForce RTX 4060 Max-Q / Mobile] [10de:28e0] (rev a1) (prog-if 00 [VGA controller])
66:00.0 Display controller [0380]: Advanced Micro Devices, Inc. [AMD/ATI] Strix [Radeon 880M / 890M] [1002:150e] (rev c1)Dump from And in case it helps I am dumping output of |
|
pls give the log from as for the sway.log, it's spamming as for the dGPU being marked as vga and being automatically selected, yes that is an issue |
|
Apologies. My one brain cell is working overtime atm. Anyways, here's the correct Hyprland.log. Yeah you are right on dGPU being marked as vga being a separate but related issue. Best narrow down on manual selection for now. Let me know if you want me to try some changes on the codebase and compile them. I've got it setup. |
|
ok so it seems the iGPU is being taken by another process/seat? meaning Hyprland can't use it at all however, that doesn't affect eDP-1 since that's on your amdgpu not sure how that would happen by upgrading to aquamarine but getting a dmesg.log will probably help find the issue |
|
Bus Id 65:00 is the dGPU, not iGPU. Which makes this statement even weirder since I am manually setting Nonetheless here is the dmesg.log |
|
intresting... You can try setting a lower resolution or refresh rate, or get a different cable as for the gpu stuff, not sure why Hyprland can't use your dGPU, but it doesn't affect eDP-1 so oh well |
|
Plasma has no issue with it. In fact, Hyprland also had no problem with it previously. This is the same configuration that powered my Hyprland setup before switching to Aquamarine. Regardless, I’ve tried changing it to 60Hz, but that doesn’t work either (although it does in Plasma). Since this is a laptop, I can’t modify any of the connections. To clarify, I have no intention of running Hyprland with the dGPU; I only use dGPU for gaming or training machine learning models. I still don't understand why logs keep referencing Nvidia. Additionally, this issue isn’t consistent across all window managers or desktop environments; Plasma and GNOME work perfectly fine. With regards to not enough VRAM: I noticed that the UEFI has a When going though the logs this line caught my attention: [ERR] [AQ] drm: No format for outputWhich led me to this issue which might prove useful. I'll play with it when I get a bit of time, but for now I share it in case it helps. Edit: My mesa version is 24.2.6 whereas the issue author's is 24.3.0. He found regressing to 24.2.7 fixed it. |
|
interesting that wlroots works and aq doesnt, can you give a HL log from before aquamarine (ex. v0.41.0) also try #8804 branch and test with AQ_NO_MODIFIERS=1 and without as for the |
|
Running I see no improvements with branch #8804 :
|
|
does pre aq Hyprland still work? from the logs it seems to be failing just the same |
|
Yeah it doesn't crash. Writing this comment on pre-aq hyprland at the moment. Perhaps It was me actively killing the process (doubt it) or me making a typo for |
Already reported ? *
Regression?
Yes
System Info and Version
System/Version info
Description
After introducing aquamarine in commit 016da23, my computer displays a black screen on eDP-1 as soon as Hyprland launches, though it renders normally on DP-2. Hyprland itself remains responsive, with no noticeable errors.
Sanity checks performed:
My best lead on the cause may relate to Laptop GPU order reversed from normal causes boot_vga/drm issues and KWin’s specific approach to the problem. TL;DR: The PCI bus order places the dGPU before the iGPU on my specific laptop model, causing confusion in many WMs, like Cosmic and KWin.
Additionally, I’ve noticed increased logging of NVIDIA references when launching Hyprland instances since aquamarine’s introduction.
I've attached the bisection file (hypr_bisection.log), as well as text files for the output of ./build/Hyprland &> (compiled_commit).txt and any accompanying crash report.
How to reproduce
git clone --no-checkout https://github.com/hyprwm/Hyprlandgit checkout 016da234git submodule update --init --recursivenix developmake allAQ_DRM_DEVICES=/dev/dri/card1where card1 corresponds to amd gpu../build/HyrplandCrash reports, logs, images, videos
016da234d0e852de3ef20eb2e89ac58d2a85f6e7.txt
hyprlandCrashReport205751.txt
077494ee85c8fa4c6ae74ad8d749feea826294d2.txt
hyprlandCrashReport192884.txt
04b40ea2ec85dd1d74ff18edc046a233b65024ac.txt
hypr_bisection.log
The text was updated successfully, but these errors were encountered: