Re: [PATCH v4 18/18] drm/panthor: Add debugfs knobs to simulate reset failures

[Boris Brezillon]

On Sat, 12 Sep 2026 20:27:28 +0100
Adrian Larumbe <adrian.larumbe@xxxxxxxxxxxxx> wrote:

> I've tried poking the knob that fakes an error in the reset path and got two different kinds of oputput:
> 
> Either
> ```
> [ 1950.080236] panthor fb000000.gpu: [drm] *ERROR* Failed to boot MCU after reset, making device unusable.
> [ 1951.167384] panthor fb000000.gpu: [drm] Timed out waiting for MCU to halt
> [ 1951.167543] panthor fb000000.gpu: [drm] Failed to cleanly suspend MCU

Hm, I guess that happens when the unplug work has been queued but not
executed yet, and something tries to interact with the FW after that. I
suspect it's some autosuspend request kicking in after that, and asking
the FW to suspend, which it can't do because it's not even booted.
What's weird is the 1sec between those two events (I don't expect
unplug to take that long to execute).

As for the race, I'd say that's not a big deal if the final outcome
still is that the device is unplug. This being said, the more I look at
this, the more I think triggering an unplug on a FW boot failure is not
the right thing to do, but that's another can of worms I don't want to
open now.