Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

Hybrid AMD/NVIDIA system fails #2426

Closed
RAOF opened this issue May 13, 2022 · 0 comments · Fixed by #2430 or #2562
Closed

Hybrid AMD/NVIDIA system fails #2426

RAOF opened this issue May 13, 2022 · 0 comments · Fixed by #2430 or #2562

Comments

@RAOF
Copy link
Contributor

RAOF commented May 13, 2022

Something in Mesa has changed, and now Mir fails to start on my AMD/NVIDIA hybrid system.

[2022-05-13 17:43:28.603698] <information> mirserver: Found rendering driver: mir:eglstream-kms (version 2.8.0)
libEGL warning: DRI2: failed to create dri screen
libEGL warning: DRI2: failed to create dri screen
[2022-05-13 17:43:28.645889] < - debug - > eglstream: Failed to initialise EGL: EGL_NOT_INITIALIZED (0x3001)
[2022-05-13 17:43:28.658006] <information> eglstream: EGLDevice found but unsuitable. Missing extension EGL_KHR_stream_consumer_gltexture
[2022-05-13 17:43:28.658049] <information> eglstream: EGLDevice found but unsuitable. Missing extension EGL_NV_stream_attrib
ERROR: /home/chris/Canonical/Mir/mir/main/src/server/graphics/default_configuration.cpp(335): Throw in function virtual const std::vector<std::shared_ptr<graphics::RenderingPlatform>> &mir::DefaultServerConfiguration::the_rendering_platforms()
Dynamic exception type: boost::wrapexcept<std::runtime_error>
std::exception::what: Exception while creating rendering platform
ERROR: /home/chris/Canonical/Mir/mir/main/src/platforms/eglstream-kms/server/utils.cpp(35): Throw in function dev_t mir::graphics::eglstream::devnum_for_device(EGLDeviceEXT)
Dynamic exception type: boost::wrapexcept<std::system_error>
std::exception::what: Failed to determine DRM device node path from EGLDevice: EGL_BAD_PARAMETER (0x300c)

It looks like Mesa is supporting more of the EGLDevice platform, but is doing so in a way that we don't expect.
This was referenced May 13, 2022
Merged
Merged
@bors bors bot closed this as completed in 23b7b57 May 16, 2022
@bors bors bot closed this as completed in #2430 May 16, 2022
Saviq pushed a commit that referenced this issue May 16, 2022
@bors @RAOF @Saviq
Merge #2429 #2430
cb44628 
2429: eglstream-kms: Kill clients, not Mir, when they submit bad EGLStreams r=AlanGriffiths a=RAOF

This seems to particularly affect XWayland when clients try to use GLX.
It's not great to kill XWayland, but it's better than killing Mir.

We should probably additionally move this code to the mirwayland wrapper
generator, but this particular fix is simple.

Fixes: #2061

2430: eglstream-kms: Handle EGL errors in `devnum_for_device` r=AlanGriffiths a=RAOF

It seems that Mesa is now exposing rendernodes as EGLDevices, but using the EGL query to find the device node of such devices fails.

Handle those exceptions by assuming the device is unsuitable.

Fixes: #2426 

Co-authored-by: Christopher James Halse Rogers <christopher.halse.rogers@canonical.com>
RAOF added a commit that referenced this issue May 23, 2022
@RAOF
eglstream-kms: Retain DRM device access for whole probe.
112a266 
Oops. While fixing up #2426 I changed the scope of `device_holder`, so that
it would almost immediately go out of scope. This broke probing on the VT
console provider, as the VT console provider, and *only* the VT console
provider, drops DRM master when you release the `mir::Device`.

Fixes: #2431.
@RAOF RAOF mentioned this issue May 23, 2022
Merged
bors bot added a commit that referenced this issue May 23, 2022
@bors @RAOF
Merge #2433
2005167 
2433: eglstream-kms: Retain DRM device access for whole probe. r=AlanGriffiths a=RAOF

Oops. While fixing up #2426 I changed the scope of `device_holder`, so that
it would almost immediately go out of scope. This broke probing on the VT
console provider, as the VT console provider, and *only* the VT console
provider, drops DRM master when you release the `mir::Device`.

Fixes: #2431.

Co-authored-by: Christopher James Halse Rogers <christopher.halse.rogers@canonical.com>
AlanGriffiths pushed a commit that referenced this issue May 23, 2022
@bors @RAOF @AlanGriffiths
Merge #2433
c2aba00 
2433: eglstream-kms: Retain DRM device access for whole probe. r=AlanGriffiths a=RAOF

Oops. While fixing up #2426 I changed the scope of `device_holder`, so that
it would almost immediately go out of scope. This broke probing on the VT
console provider, as the VT console provider, and *only* the VT console
provider, drops DRM master when you release the `mir::Device`.

Fixes: #2431.

Co-authored-by: Christopher James Halse Rogers <christopher.halse.rogers@canonical.com>
@AlanGriffiths AlanGriffiths mentioned this issue Aug 18, 2022
Merged
bors bot added a commit that referenced this issue Aug 30, 2022
@bors @AlanGriffiths
Merge #2562
06d3144 
2562: Release 2.9.0 r=AlanGriffiths a=AlanGriffiths

## ABI summary:
* miral ABI bumped to 5
* mircommon ABI unchanged at 9
* mircookie ABI unchanged at 2
* mircore ABI bumped to 2
* miroil ABI bumped to 2
* mirplatform ABI unchanged at 23
* mirserver ABI unchanged at 58
* mirwayland ABI unchanged at 3
* mirplatformgraphics ABI unchanged at 20
* mirinputplatform ABI unchanged at 8
## Enhancements:
* [Wayland] Implement zwp_idle_inhibit_manager_v1
* [Wayland] Implement zwlr_virtual_pointer_v1
* [Wayland] Implement zwp_text_input_manager_v1 (Electron works with OSK)
* [Wayland] Bump wl_seat to v8 and implement hi-res scrolling (Fixes: #2176, Fixes: #2499)
* [Wayland platform] improve failed to connect error
* [Wayland platform] Port to xdg-shell (Fixes #1903, Fixes: #2434)
* [MirAL] Allow `--add-wayland-extenions all`
* [MirAL] Allow servers to get repeated string options
* [MirAL] Improvement to ExternalClientLauncher: Don't force clients to split command lines themselves
* [MirAL] Tidy up event filtering API
* [MirAL] Expose miral::Zone::id()
* [Input] Filter 2 distinct bogus touch event scenarios (UBports)
* [gbm-kms] new driver quirk to disable KMS modeset probe
* [gbm-kms] Add defaults to driver-quirks for nvidia and evdi (Fixes: #2467)
* [gbm-kms] Add defaults to driver-quirks for vc4-drm and v3d
* [mir-smoke-test-runner] Enable working in a Wayland only environment
## Bugs fixed:
* Fix ABI breakage of libmircore.so.1 with v2.8.0 vs v1.8.2 (Fixes #2465)
* [Wayland] Text Input v2: do not use commit count as serial
* [Wayland] Send keyboard modifiers after keyboard enter (Fixes: #2535, Fixes: #2025)
* [Wayland] wlr-screencopy-v1: send .damage event as required (wayvnc fix)
* [Wayland] wlr-screencopy-v1: Waits until the copy area has been damaged in `.copy_with_damage` request (wayvnc fix)
* [MirAL] Unblock signals before execing child processes (Fixes: #1284)
* [eglstream-kms] Kill clients, not Mir, when they submit bad EGLStreams (Fixes: #2061)
* [eglstream-kms] Handle EGL errors in devnum_for_device (Fixes: #2426)
* [test clients] Explicitly ask for a GLESv2 context (Fixes: #2440)
* [renderers/gl] Clear framebuffer to opaque black (Fixes: #2427)
* CMake cleanup for locally built dependencies (Fixes: #2507, Fixes: #2261)
* Fix event timestamps (Fixes: #2577)

Co-authored-by: Alan Griffiths <alan@octopull.co.uk>
Co-authored-by: bors[bot] <26634292+bors[bot]@users.noreply.github.com>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Assignees
No one assigned
Labels
None yet
Projects
None yet
Milestone
No milestone
Development

Successfully merging a pull request may close this issue.

eglstream-kms: Handle EGL errors in devnum_for_device canonical/mir
Release 2.9.0 canonical/mir
1 participant
@RAOF