Cannot find vulkan adapter on Mesa 25 + RADV

[matteodev8]

Description
After upgrading to Mesa 25 from Mesa 24 today, I am experiences issues with my bevy app. It errors on launch with wgpu complaining about surface configuration failed: incompatible window kind and Device::create_render_pipeline error: Internal error in ShaderStages(VERTEX | FRAGMENT) shader:

Repro steps

1. use Mesa 25.0.1
2. run project

Expected vs observed behavior
Should open the window

Crashes

Extra materials

2025-03-14T15:15:11.165679Z  INFO bevy_render::renderer: AdapterInfo { name: "AMD Radeon RX 7800 XT (radeonsi, navi32, LLVM 19.1.7, DRM 3.61, 6.13.6-arch1-1)", vendor: 4098, device: 0, device_type: Other, driver: "", driver_info: "4.6 (Core Profile) Mesa 25.0.1-arch1.2", backend: Gl }
2025-03-14T15:15:11.224144Z  WARN bevy_pbr::ssao: ScreenSpaceAmbientOcclusionPlugin not loaded. GPU lacks support: TextureFormat::R16Float does not support TextureUsages::STORAGE_BINDING.
2025-03-14T15:15:11.224481Z  INFO bevy_winit::system: Creating new window "App" (0v1#4294967296)
2025-03-14T15:15:11.224506Z  INFO winit::platform_impl::linux::x11::window: Guessed window scale factor: 1
2025-03-14T15:15:11.236518Z ERROR wgpu_core::device::global: surface configuration failed: incompatible window kind

thread 'main' panicked at /home/matteo/.cargo/registry/src/index.crates.io-1949cf8c6b5b557f/wgpu-23.0.1/src/backend/wgpu_core.rs:719:18:
Error in Surface::configure: Validation Error

Caused by:
  Invalid surface

note: run with `RUST_BACKTRACE=1` environment variable to display a backtrace
Encountered a panic in system `bevy_render::view::window::create_surfaces`!
2025-03-14T15:15:11.250487Z ERROR wgpu_core::device::global: Device::create_render_pipeline error: Internal error in ShaderStages(VERTEX | FRAGMENT) shader:
2025-03-14T15:15:11.250502Z ERROR wgpu::backend::wgpu_core: Shader translation error for stage ShaderStages(VERTEX | FRAGMENT):
2025-03-14T15:15:11.250505Z ERROR wgpu::backend::wgpu_core: Please report it to https://github.com/gfx-rs/wgpu
2025-03-14T15:15:11.250510Z ERROR wgpu::backend::wgpu_core: Handling wgpu errors as fatal by default

thread 'Async Compute Task Pool (0)' panicked at /home/matteo/.cargo/registry/src/index.crates.io-1949cf8c6b5b557f/wgpu-23.0.1/src/backend/wgpu_core.rs:1102:18:
wgpu error: Validation Error

Caused by:
  In Device::create_render_pipeline, label = 'sprite_pipeline'
    Internal error in ShaderStages(VERTEX | FRAGMENT) shader:

Platform
OS: Arch Linux (KDE Plasma 6)
Rust: 1.85.0
Mesa: 1:25.0.1-2 (Pacman)
Vulkan Driver: RADV
GPU: Radeon RX 7800 XT
Wgpu: 23.0.1

[matteodev8]

Okay apperently, it crashes with different errors each time! I ran cargo run 3 times

lsim on  dev [?] is 📦 v0.1.0 via 🦀 v1.85.0
❯ cargo run
    Finished `dev` profile [optimized + debuginfo] target(s) in 0.11s
     Running `target/debug/lsim`
2025-03-14T15:26:12.335533Z  INFO bevy_render::renderer: AdapterInfo { name: "AMD Radeon RX 7800 XT (radeonsi, navi32, LLVM 19.1.7, DRM 3.61, 6.13.6-arch1-1)", vendor: 4098, device: 0, device_type: Other, driver: "", driver_info: "4.6 (Core Profile) Mesa 25.0.1-arch1.2", backend:
 Gl }
2025-03-14T15:26:12.394467Z  WARN bevy_pbr::ssao: ScreenSpaceAmbientOcclusionPlugin not loaded. GPU lacks support: TextureFormat::R16Float does not support TextureUsages::STORAGE_BINDING.
2025-03-14T15:26:12.394838Z  INFO bevy_winit::system: Creating new window "App" (0v1#4294967296)
2025-03-14T15:26:12.394863Z  INFO winit::platform_impl::linux::x11::window: Guessed window scale factor: 1
2025-03-14T15:26:12.409044Z ERROR wgpu_core::device::global: surface configuration failed: incompatible window kind

thread 'main' panicked at /home/matteo/.cargo/registry/src/index.crates.io-1949cf8c6b5b557f/wgpu-23.0.1/src/backend/wgpu_core.rs:719:18:
Error in Surface::configure: Validation Error

Caused by:
  Invalid surface

note: run with `RUST_BACKTRACE=1` environment variable to display a backtrace
Encountered a panic in system `bevy_render::view::window::create_surfaces`!
2025-03-14T15:26:12.421399Z ERROR wgpu::backend::wgpu_core: Shader translation error for stage ShaderStages(COMPUTE):
2025-03-14T15:26:12.421413Z ERROR wgpu::backend::wgpu_core: Please report it to https://github.com/gfx-rs/wgpu
2025-03-14T15:26:12.421418Z ERROR wgpu::backend::wgpu_core: Handling wgpu errors as fatal by default

thread 'Async Compute Task Pool (2)' panicked at /home/matteo/.cargo/registry/src/index.crates.io-1949cf8c6b5b557f/wgpu-23.0.1/src/backend/wgpu_core.rs:1147:18:
wgpu error: Validation Error

Caused by:
  In Device::create_compute_pipeline, label = 'mesh preprocessing (GPU culling)'
    Internal error:

lsim on  dev [?] is 📦 v0.1.0 via 🦀 v1.85.0
❯ cargo run
    Finished `dev` profile [optimized + debuginfo] target(s) in 0.11s
     Running `target/debug/lsim`
2025-03-14T15:26:13.272375Z  INFO bevy_render::renderer: AdapterInfo { name: "AMD Radeon RX 7800 XT (radeonsi, navi32, LLVM 19.1.7, DRM 3.61, 6.13.6-arch1-1)", vendor: 4098, device: 0, device_type: Other, driver: "", driver_info: "4.6 (Core Profile) Mesa 25.0.1-arch1.2", backend:
 Gl }
2025-03-14T15:26:13.332049Z  WARN bevy_pbr::ssao: ScreenSpaceAmbientOcclusionPlugin not loaded. GPU lacks support: TextureFormat::R16Float does not support TextureUsages::STORAGE_BINDING.
2025-03-14T15:26:13.332403Z  INFO bevy_winit::system: Creating new window "App" (0v1#4294967296)
2025-03-14T15:26:13.332430Z  INFO winit::platform_impl::linux::x11::window: Guessed window scale factor: 1
2025-03-14T15:26:13.344258Z ERROR wgpu_core::device::global: surface configuration failed: incompatible window kind

thread 'main' panicked at /home/matteo/.cargo/registry/src/index.crates.io-1949cf8c6b5b557f/wgpu-23.0.1/src/backend/wgpu_core.rs:719:18:
Error in Surface::configure: Validation Error

Caused by:
  Invalid surface

note: run with `RUST_BACKTRACE=1` environment variable to display a backtrace
Encountered a panic in system `bevy_render::view::window::create_surfaces`!

lsim on  dev [?] is 📦 v0.1.0 via 🦀 v1.85.0
❯ cargo run
    Finished `dev` profile [optimized + debuginfo] target(s) in 0.11s
     Running `target/debug/lsim`
2025-03-14T15:26:14.539428Z  INFO bevy_render::renderer: AdapterInfo { name: "AMD Radeon RX 7800 XT (radeonsi, navi32, LLVM 19.1.7, DRM 3.61, 6.13.6-arch1-1)", vendor: 4098, device: 0, device_type: Other, driver: "", driver_info: "4.6 (Core Profile) Mesa 25.0.1-arch1.2", backend: Gl }
2025-03-14T15:26:14.598237Z  WARN bevy_pbr::ssao: ScreenSpaceAmbientOcclusionPlugin not loaded. GPU lacks support: TextureFormat::R16Float does not support TextureUsages::STORAGE_BINDING.
2025-03-14T15:26:14.598636Z  INFO bevy_winit::system: Creating new window "App" (0v1#4294967296)
2025-03-14T15:26:14.598662Z  INFO winit::platform_impl::linux::x11::window: Guessed window scale factor: 1
2025-03-14T15:26:14.612069Z ERROR wgpu_core::device::global: surface configuration failed: incompatible window kind

thread 'main' panicked at /home/matteo/.cargo/registry/src/index.crates.io-1949cf8c6b5b557f/wgpu-23.0.1/src/backend/wgpu_core.rs:719:18:
Error in Surface::configure: Validation Error

Caused by:
  Invalid surface

note: run with `RUST_BACKTRACE=1` environment variable to display a backtrace
Encountered a panic in system `bevy_render::view::window::create_surfaces`!
2025-03-14T15:26:14.622852Z ERROR wgpu::backend::wgpu_core: Shader translation error for stage ShaderStages(COMPUTE):
2025-03-14T15:26:14.622861Z ERROR wgpu::backend::wgpu_core: Please report it to https://github.com/gfx-rs/wgpu
2025-03-14T15:26:14.622867Z ERROR wgpu::backend::wgpu_core: Handling wgpu errors as fatal by default

thread 'Async Compute Task Pool (1)' panicked at /home/matteo/.cargo/registry/src/index.crates.io-1949cf8c6b5b557f/wgpu-23.0.1/src/backend/wgpu_core.rs:1147:18:
wgpu error: Validation Error

Caused by:
  In Device::create_compute_pipeline, label = 'mesh preprocessing (direct)'
    Internal error:

[papow65]

I had the same problem using mesa 25. Downgrading mesa and vulkan-radeon to version 24 served as an effective workaround.

[arcaspo]

I am having the same problem. I think it might be because wgpu can't find the vulkan drivers and defaults to gl renderer which causes lots of errors?
When I try force the project to run with vulkan with WGPU_BACKEND=vulkan it runs but at a very low framerate because it uses llvmpipe instead of proper drivers

AdapterInfo { name: "llvmpipe (LLVM 19.1.7, 256 bits)", vendor: 65541, device: 0, device_type: Cpu, driver: "llvmpipe", driver_info: "Mesa 25.0.1 (LLVM 19.1.7)", backend: Vulkan }

wgpu-info returns this:

Adapter 0:
                 Backend: Vulkan
                    Name: llvmpipe (LLVM 19.1.7, 256 bits)
                VendorID: 0x10005
                DeviceID: 0x0
                    Type: Cpu
                  Driver: llvmpipe
              DriverInfo: Mesa 25.0.1 (LLVM 19.1.7)
        WebGPU Compliant: true
Adapter 1:
                 Backend: Gl
                    Name: AMD Radeon RX 6600 XT (radeonsi, navi23, LLVM 19.1.7, DRM 3.61, 6.13.6-200.fc41.x86_64)
                VendorID: 0x1002
                DeviceID: 0x0
                    Type: Other
                  Driver: <empty>
              DriverInfo: 4.6 (Core Profile) Mesa 25.0.1
        WebGPU Compliant: false

but vulkaninfo | grep "GPU id" returns this:

GPU id = 0 (AMD Radeon RX 6600 XT (RADV NAVI23))
GPU id = 1 (llvmpipe (LLVM 19.1.7, 256 bits))

So I think wgpu must not be able to find or use the radv driver for some reason.

[cwfitzgerald]

If you set up env_logger::init(), and then do RUST_LOG=warn in your environment, does wgpu warn of anything. Can you post your full vulkan info?

[arcaspo]

When I set up env_logger wgpu warns

[WARN  wgpu_hal::vulkan::adapter] Adapter is not Vulkan compliant, hiding adapter: AMD Radeon RX 6600 XT (RADV NAVI23)
[WARN  wgpu_hal::gles::egl] Re-initializing Gles context due to Wayland window

This is the full vulkaninfo
vulkaninfo.txt

[cwfitzgerald]

This is most likely https://gitlab.freedesktop.org/mesa/mesa/-/issues/12799 which has already been fixed. In the mean time you can set WGPU_ALLOW_UNDERLYING_NONCOMPLIANT_ADAPTER=1 in your environment or enable the InstanceFlag until this bug is fixed.

[cwfitzgerald]

Closing this as not-our-issue, but if any additional things come up, please let me know and we can re-open.