DRM: Explicit synchorisation support #3717
Open
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
|
At least two things should be resolved before landing this:
|
We want to be able to load real hardware rendering platforms on top in tests, so have the stub `DisplayPlatform` implement the simplest of the concrete interfaces
When we're making a `StubBuffer` with non-empty `size`, make sure we also set the stride to a reasonable value.
None of our code *really* cares that we don't want to get events before calling `monitor.enable()`, and UDev in Ubuntu 25.04 has changed behaviour so that events *are* received before enabling the monitor, so just drop the test.
RAOF
commented
Jan 17, 2025
| sync_timeline.value().resource, | ||
| SyncTimeline::Error::no_buffer, | ||
| "Timeline release sync point set, but no buffer committed"}; | ||
| } | ||
| frame_callback_executor->spawn(std::move(executor_send_frame_callbacks)); |
There was a problem hiding this comment.
Aha! Here is the race causing BadBufferTest.test_truncated_shm_file to fail, sometimes.
The process goes:
- The WLCS test in surface creation makes a buffer-less commit while setting up the surface roles
- That commit goes down this codepath, which primes a 16ms timer here. When that expires, it will call
send_frame_callbacks(), which sends all the pending frame callbacks at the time of the 16ms-later call (not at the time of this call). - The WLCS test attaches a buffer, requests a frame event, commits, and waits for that frame event.
- The buffer goes through here, gets a composition pass is triggered.
- In the compositor thread the buffer is consumed triggering the the
send_frame_callbacks, Mir sends the frame event. - The WLCS test attaches the bad buffer, requests a frame event, and waits, expecting dispatch on that frame event to receive the protocol error.
- Sometimes the 16ms timer fires after the new buffer has been committed, but before the compositor thread consumes it. This results in the WLCS test seeing a frame event, and the test fails (because it didn't protocol error).
This whole 16ms-timer infrastructure is because Firefox has a frame/commit loop if we immediately dispatch the frame event on an empty commit.
There was a problem hiding this comment.
- it will call
send_frame_callbacks(), which sends all the pending frame callbacks at the time of the 16ms-later call (not at the time of this call).
That seems wrong but I'm not sure chosing the right "right" is straightforward
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
This implements the
linux_drm_syncobj_v1explicit synchronisation protocol.Currently only the simplest support - we accept
acquirefences and wait for them to become ready before we process the associatedwl_surface.commit, and we signal thereleasefence once we have finished with the buffer.There are optimisation opportunities at both ends here - we could push the
acquirefence deeper into the pipeline, and trigger thereleasefence sooner.