surface: Don't crash due to weird state after resume

[Drakulix]

@jacobgkau Should fix resume issues from #694 (unable to confirm myself)

[jacobgkau]

With this installed on the Mira B4 w/ 4060 Ti, resume still isn't working correctly.

The first change of behavior is I now get a mouse cursor on top of an otherwise blank screen after resuming; the cursor doesn't move around smoothly, but if I move the mouse, the cursor will pop to a new location a few seconds later.

With this fix, I am able to get to a TTY after the resume (which I wasn't able to do before). If I restart cosmic-greeter.service, then I can log back into the GUI successfully.

Here's the system journal after the resume: after-resume-d580294.txt

I see these logs after suspend is started but before it's finished suspending:

Aug 09 20:54:51 pop-os cosmic-comp[2028]: Failed to submit rendering: Failed to submit result for display
                                          
                                          Caused by:
                                              0: The underlying drm surface encountered an error: DRM access error: Page flip commit failed on device `Some("/dev/dri/card1")` (Permission denied (os error 13))
                                              1: DRM access error: Page flip commit failed on device `Some("/dev/dri/card1")` (Permission denied (os error 13))
                                              2: Permission denied (os error 13)

After resume, I see this line repeated a lot in one place:

Aug 09 20:55:42 pop-os cosmic-session[1958]: [2024-08-09T20:55:42Z WARN  iced_futures::subscription::tracker] Error sending event to subscription: TrySendError { kind: Full }

And then one of these just before I switch to a TTY:

Aug 09 20:56:28 pop-os cosmic-comp[2028]: Failed to submit rendering: Failed to submit result for display
                                          
                                          Caused by:
                                              0: The underlying drm surface encountered an error: Device is currently paused, operation rejected
                                              1: Device is currently paused, operation rejected

I'm approving this since it's an improvement in behavior, but it seems more work is still required, possibly in cosmic-comp.

[Drakulix]

With this installed on the Mira B4 w/ 4060 Ti, resume still isn't working correctly.

Here's the system journal after the resume: after-resume-d580294.txt

Difficult to decode what exactly is happening here.

I see these logs after suspend is started but before it's finished suspending:

Aug 09 20:54:51 pop-os cosmic-comp[2028]: Failed to submit rendering: Failed to submit result for display
                                          
                                          Caused by:
                                              0: The underlying drm surface encountered an error: DRM access error: Page flip commit failed on device `Some("/dev/dri/card1")` (Permission denied (os error 13))
                                              1: DRM access error: Page flip commit failed on device `Some("/dev/dri/card1")` (Permission denied (os error 13))
                                              2: Permission denied (os error 13)

That seems correct, as the currently queued render likely fails while the session is being deactivate, the error code also matches up here. cosmic-comp will catch this and given this line doesn't repeat it clearly also doesn't try to redraw, so success on the suspend side of things.

After resume, I see this line repeated a lot in one place:

Aug 09 20:55:42 pop-os cosmic-session[1958]: [2024-08-09T20:55:42Z WARN  iced_futures::subscription::tracker] Error sending event to subscription: TrySendError { kind: Full }

Seems like a separate cosmic-session bug we should track.

And then one of these just before I switch to a TTY:

Aug 09 20:56:28 pop-os cosmic-comp[2028]: Failed to submit rendering: Failed to submit result for display
                                          
                                          Caused by:
                                              0: The underlying drm surface encountered an error: Device is currently paused, operation rejected
                                              1: Device is currently paused, operation rejected

This one is puzzling. I wonder if the log message actually originates from before suspend, as it would make sense that the surface still exists but the device is already paused.

Otherwise I don't know how we would get into this state, as we first resume the device and then startup all the surfaces again (and at that point only one thread is involved).

[Drakulix]

Pushed a commit to address the last log message.

@jacobgkau If you have the time to retest, that would be very appreciated. The system journal would again help with debugging.