crosvm graphics not working

[hissssst]

Generated script looks like this

#!/nix/store/q7sqwn7i6w2b67adw0bmix29pxg85x3w-bash-5.3p3/bin/bash
set -eou pipefail
rm -f cage.sock


rm -f cage-gpu.sock
/nix/store/1lw7irxs7x1jv3vnffj4ax9ykhswcy54-crosvm-0-unstable-2025-09-19/bin/crosvm device gpu \
  --socket cage-gpu.sock \
  --wayland-sock $XDG_RUNTIME_DIR/$WAYLAND_DISPLAY\
  --params '{"context-types":"virgl:virgl2:cross-domain","egl":true,"vulkan":true}' \
  &
while ! [ -S cage-gpu.sock ]; do
  sleep .1
done


# Open macvtap interface file descriptors


exec -a "microvm@cage" /nix/store/1lw7irxs7x1jv3vnffj4ax9ykhswcy54-crosvm-0-unstable-2025-09-19/bin/crosvm run -m 512 -c 1 --serial 'type=stdout,console=true,stdin=true' -p 'console=ttyS0 reboot=k panic=1 8250.nr_uarts=1 loglevel=4 lsm=landlock,yama,bpf init=/nix/store/bhvlsdflxk11n2sj6d2n2j8lr2wsy9im-nixos-system-cage-25.11pre-git/init regInfo=/nix/store/96v26angcaxj80y56gp2l37gpi7g33rp-closure-info/registration' --no-balloon -r /nix/store/kjp6dlim0jsbiw85kl6xq674bmgmcnvr-microvm-store-disk.erofs --vhost-user-gpu cage-gpu.sock --seccomp-log-failures -s cage.sock --initrd /nix/store/znkpvkzhyv2dnmjjlcas8r28gz833hi0-initrd-linux-6.12.49/initrd /nix/store/5n8dxv7gca4rhw4sr05hqgif5ha3ykmz-linux-6.12.49-dev/vmlinux  

Running the vm with microvm -r cage

Fails with

[2025-10-02T17:23:38.568620616+00:00 ERROR crosvm] arg parsing failed: Unrecognized argument: --vhost-user-gpu

---

I've tried to change the crosvm to the one present in nixpkgs-unstable (as of the moment of writing the issue)
and this version of crosvm accepted the option but started failing with

[2025-10-02T17:26:49.603817390+00:00 WARN  rutabaga_gfx::rutabaga_core] error initializing gpu backend=virglrenderer, falling back to 2d.
[2025-10-02T17:26:49.604217364+00:00 WARN  devices::virtio::gpu::virtio_gpu] virtio-gpu get_capset_info(index=1) failed. intentionally poisoning response
[    1.257035] [drm:virtio_gpu_init.cold [virtio_gpu]] *ERROR* invalid capset id 4294967295
[2025-10-02T17:26:49.610001077+00:00 ERROR devices::virtio::gpu] error processing gpu command ResourceCreate2d: ErrRutabaga(InvalidComponent)
[2025-10-02T17:26:49.610025095+00:00 ERROR devices::virtio::gpu] error processing gpu command ResourceAttachBacking: ErrInvalidResourceId
[2025-10-02T17:26:49.610031162+00:00 ERROR devices::virtio::gpu] error processing gpu command CtxCreate: ErrRutabaga(InvalidComponent)
[2025-10-02T17:26:49.610035284+00:00 ERROR devices::virtio::gpu] error processing gpu command CtxAttachResource: ErrRutabaga(InvalidContextId)
[2025-10-02T17:26:49.610039108+00:00 ERROR devices::virtio::gpu] error processing gpu command CtxDetachResource: ErrRutabaga(InvalidContextId)
[    1.262955] [drm:virtio_gpu_dequeue_ctrl_func [virtio_gpu]] *ERROR* response 0x1200 (command 0x101)
[    1.263649] [drm:virtio_gpu_dequeue_ctrl_func [virtio_gpu]] *ERROR* response 0x1203 (command 0x106)
[    1.264323] [drm:virtio_gpu_dequeue_ctrl_func [virtio_gpu]] *ERROR* response 0x1200 (command 0x200)
[    1.264879] [drm:virtio_gpu_dequeue_ctrl_func [virtio_gpu]] *ERROR* response 0x1200 (command 0x202)
[    1.265397] [drm:virtio_gpu_dequeue_ctrl_func [virtio_gpu]] *ERROR* response 0x1200 (command 0x203)
...
[2025-10-02T17:26:50.799081829+00:00 ERROR devices::virtio::gpu] error processing gpu command TransferToHost2d: ErrRutabaga(InvalidComponent)
[2025-10-02T17:26:50.799110218+00:00 ERROR devices::virtio::gpu] error processing gpu command SetScanout: ErrInvalidResourceId
[2025-10-02T17:26:50.799116226+00:00 ERROR devices::virtio::gpu] error processing gpu command ResourceFlush: ErrInvalidResourceId
[2025-10-02T17:26:50.799313514+00:00 ERROR devices::virtio::gpu] error processing gpu command TransferToHost2d: ErrRutabaga(InvalidComponent)
[2025-10-02T17:26:50.799319468+00:00 ERROR devices::virtio::gpu] error processing gpu command SetScanout: ErrInvalidResourceId
[2025-10-02T17:26:50.799324705+00:00 ERROR devices::virtio::gpu] error processing gpu command ResourceFlush: ErrInvalidResourceId
[    2.452614] [drm:virtio_gpu_dequeue_ctrl_func [virtio_gpu]] *ERROR* response 0x1200 (command 0x105)
[    2.453225] [drm:virtio_gpu_dequeue_ctrl_func [virtio_gpu]] *ERROR* response 0x1203 (command 0x103)
[    2.453706] [drm:virtio_gpu_dequeue_ctrl_func [virtio_gpu]] *ERROR* response 0x1203 (command 0x104)
[    2.454216] [drm:virtio_gpu_dequeue_ctrl_func [virtio_gpu]] *ERROR* response 0x1200 (command 0x105)
[    2.454675] [drm:virtio_gpu_dequeue_ctrl_func [virtio_gpu]] *ERROR* response 0x1203 (command 0x103)
[2025-10-02T17:26:50.808759456+00:00 ERROR devices::virtio::gpu] error processing gpu command TransferToHost2d: ErrRutabaga(InvalidComponent)
[2025-10-02T17:26:50.808785144+00:00 ERROR devices::virtio::gpu] error processing gpu command ResourceFlush: ErrInvalidResourceId
[2025-10-02T17:26:50.839715361+00:00 ERROR devices::virtio::gpu] error processing gpu command TransferToHost2d: ErrRutabaga(InvalidComponent)
[2025-10-02T17:26:50.839779601+00:00 ERROR devices::virtio::gpu] error processing gpu command ResourceFlush: ErrInvalidResourceId
[2025-10-02T17:26:50.840358611+00:00 ERROR devices::virtio::gpu] error processing gpu command TransferToHost2d: ErrRutabaga(InvalidComponent)
[2025-10-02T17:26:50.840376840+00:00 ERROR devices::virtio::gpu] error processing gpu command ResourceFlush: ErrInvalidResourceId
[2025-10-02T17:26:50.840383730+00:00 ERROR devices::virtio::gpu] error processing gpu command TransferToHost2d: ErrRutabaga(InvalidComponent)

cc @alyssais

[alyssais]

[2025-10-02T17:23:38.568620616+00:00 ERROR crosvm] arg parsing failed: Unrecognized argument: --vhost-user-gpu

Yep, this needs to be updated to --vhost-user gpu,, like this: https://spectrum-os.org/git/spectrum/commit/?id=cd2ccb597b2d3de088b247a9b0ac6bd630c8d3f2

I've tried to change the crosvm to the one present in nixpkgs-unstable (as of the moment of writing the issue) and this version of crosvm accepted the option but started failing with

[2025-10-02T17:26:49.603817390+00:00 WARN  rutabaga_gfx::rutabaga_core] error initializing gpu backend=virglrenderer, falling back to 2d.
[2025-10-02T17:26:49.604217364+00:00 WARN  devices::virtio::gpu::virtio_gpu] virtio-gpu get_capset_info(index=1) failed. intentionally poisoning response

This doesn't look good. Has this ever worked for you?

[hissssst]

Yep, this needs to be updated to --vhost-user gpu,, like this: https://spectrum-os.org/git/spectrum/commit/?id=cd2ccb597b2d3de088b247a9b0ac6bd630c8d3f2

Thanks, I'll try that

This doesn't look good. Has this ever worked for you?

Yep, some long time ago I managed to make it work. My cloud-hypervisor setup is working too

[alyssais]

This doesn't look good. Has this ever worked for you?

Yep, some long time ago I managed to make it work.

If you could find a past version that worked, you could try bisecting.

[hissssst]

Oh, that's impossible to find, it was like 2-3 years ago. I have several things in mind though which may fix this