GPU work never completes on Snapdragon X2 (Adreno X2-45): GL draws produce nothing, VA-API decode and Wayland EGL hang
Maintainers usually reply within 1 day
Nobody has claimed this yet.
Assessment
- Difficulty
- 4/5
- Estimated time
- 3-5 days
- Newbie friendliness
- 34/100
- Issue type
- Bug
- Clarity
- Mostly clear
- Activity status
- Active
- Tech stack
- linux, ubuntu
- Domain
- backend
Research direction
Start with the supplied diagnostic logs and the GPU rendering and VA-API reproductions; compare the Snapdragon X2 behavior with the X Elite report in #13370. The issue does not identify a WSL source file or failing test, so first narrow down whether the hang is in WSL, WSLg, or the Windows GPU driver. Done means shader draws, Wayland EGL, and VA-API decode complete on the X2.
Written by the indexing model from the issue text.
Description
Windows Version
Microsoft Windows [Version 10.0.28000.3231] (Windows 11 Home 26H1, ARM64)
WSL Version
3.0.3.0
Are you using WSL 1, WSL 2, or WSLC?
- WSL 2
- WSL 1
- WSLC
Kernel Version
6.18.54.1-1 (stock). Also reproduced with a custom build of the same 6.18.54.1 source with only binder added.
Distro Version
Ubuntu 26.04.1 LTS
Other Software
- Device: Surface Pro 12-inch 2nd Edition (Snapdragon X2 Plus).
- GPU: Qualcomm Adreno X2-45, driver 32.0.163.3 (2026-07-23).
- WSL loads driver package
qcdx8480.inf_arm64_d5d90950f5d5b257, which containslibqcdx12arm64wslum.soandqcgpuarm64compilercore.so. - Mesa 26.0.8-1ubuntu0.3, libva 2.23.0.
- WSLg 1.0.79, MSRDC 1.2.7391, Direct3D 1.611.1-81528511, DXCore 10.0.26100.1-240331-1435.ge-release.
Repro Steps
Right after wsl --shutdown, in Ubuntu:
sudo apt install -y mesa-utils vainfo ffmpeg
export GALLIUM_DRIVER=d3d12 LIBVA_DRIVER_NAME=d3d12
# 1. Device and context creation work:
eglinfo -B -p surfaceless | grep 'OpenGL core profile renderer'
# -> OpenGL core profile renderer: D3D12 (Qualcomm(R) Adreno(TM) X2-45 GPU)
# 2. EGL on the WSLg Wayland display hangs:
timeout -s KILL 15 eglinfo -B -p wayland; echo "exit $?"
# -> exit 124 (still blocked after 15 s, in a futex wait)
# 3. Rendering never completes a frame (no "FPS" lines, no gears shown):
timeout -s KILL 10 stdbuf -oL es2gears_wayland
# With GALLIUM_DRIVER=llvmpipe the same command renders normally.
# 4. VA-API lists hardware decode profiles, but decoding never finishes:
vainfo --display drm --device /dev/dri/renderD128 | grep VLD
ffmpeg -loglevel error -f lavfi -i testsrc2=size=1920x1080:rate=60 -t 4 -c:v libx264 -y /tmp/test.mp4
timeout -s KILL 40 ffmpeg -loglevel error -hwaccel vaapi -hwaccel_device /dev/dri/renderD128 -hwaccel_output_format vaapi -i /tmp/test.mp4 -f null -; echo "exit $?"
# -> exit 124 (killed after 40 s). Software decoding of the same clip takes about a second.
Expected Behavior
OpenGL rendering and VA-API decoding run on the GPU through Mesa's d3d12 driver, as they do on x64 GPUs. On the Snapdragon X Elite, per #13370, they work once GALLIUM_DRIVER=d3d12 is set.
Actual Behavior
The GPU is found and contexts are created, but any real GPU work never completes:
- Draws: anything that runs shaders produces no output.
- Fence waits:
eglinfoon Wayland, the firstes2gearsframe, and VA-API decoding all block forever in futex waits. The stuckffmpegdoesn't even respond to SIGTERM. - What works: clears, copies and readbacks. Through virglrenderer (
virgl_test_serverwithGALLIUM_DRIVER=d3d12),glClearplusglReadPixelsreturn correct values, but every draw call is silently dropped, with no GL errors. With llvmpipe on the server side, the same draws are pixel-perfect.
Things ruled out:
- It reproduces right after
wsl --shutdownon Microsoft's stock kernel, so it's not the kernel or a long-running session. - Windows logs no display driver resets (TDR) in the System event log during the hangs.
- Windows Developer Mode is on, and
D3D12_DEBUG=experimentalmakes no difference.
dmesg shows the same lines on the stock and custom kernels:
misc dxg: dxgk: dxgkio_query_adapter_info: Ioctl failed: -22
misc dxg: dxgk: dxgkio_query_adapter_info: Ioctl failed: -2
misc dxg: dxgk: dxgkio_is_feature_enabled: Ioctl failed: -22
Possibly related:
- #13370: on the X Elite (Adreno X1-85),
GALLIUM_DRIVER=d3d12restores acceleration. On the X2 it selects the GPU, but work never completes. - microsoft/wslg#1513: on the Adreno X1-85, GPU processes hang or crash at exit inside
qcgpuarm64compilercore.so.
Diagnostic Logs
- Dominant language
- C++
- Stars
- 34.4k
- Forks
- 1.9k
- Avg merge
- 1d 23h
- Merged PRs (30d)
- 166
Getting set up
- No Dockerfile or Docker Compose file
- Has a pull request template
- Read the contributing guide
First steps
- Read the whole issue, then the project's contributing guide.
- Comment on the issue to say you are picking it up — it saves two people doing the same work.
- Fork the repository and make your change on a branch.
- Open a pull request that references the issue number.
More from microsoft/WSL
-
feature wsl2
Difficulty 2/5 1-3 hours Newbie friendliness 80/100
microsoft/WSL#42306 · 3 comments ·
Maintainers usually reply within 1 day
-
feature
Difficulty 2/5 1-3 hours Newbie friendliness 72/100
microsoft/WSL#42114 · 3 comments ·
Maintainers usually reply within 1 day
-
feature kernel wsl2
Difficulty 2/5 1-3 hours Newbie friendliness 68/100
microsoft/WSL#41107 · 3 comments · 1 reaction ·
Maintainers usually reply within 1 day
-
feature kernel
Difficulty 2/5 1-3 hours Newbie friendliness 82/100
microsoft/WSL#41054 · 9 comments · 1 reaction ·
Maintainers usually reply within 1 day
-
feature GPU
Difficulty 1/5 1-3 hours Newbie friendliness 70/100
microsoft/WSL#13592 · 4 comments ·
Maintainers usually reply within 1 day
Similar issues
-
bug
Difficulty 2/5 1-3 hours Newbie friendliness 62/100
nextcloud/passman-android#214 ·
Maintainers usually reply within 1 day
-
Difficulty 1/5 1-3 hours Newbie friendliness 62/100
Maintainers usually reply within 1 day
-
2026.11.0 fails to install on Intel macOS: requirements.txt pins cryptography==50.0.2 for Darwin x86_64May be free again A pull request for this issue was closed without being merged. Open
Difficulty 1/5 Under an hour Newbie friendliness 87/100
esphome/esphome#22892 · 1 comment ·
Maintainers usually reply within 1 day
-
bug
Difficulty 2/5 1-3 hours Newbie friendliness 77/100
distante/esphome-components#50 · 1 comment ·
Maintainers usually reply within 1 day
-
agent:Windows enhancement pri:medium release windows
Difficulty 2/5 1-3 hours Newbie friendliness 84/100
Maintainers usually reply within 1 day