On 07/04/2024 16:08, Dmitry Baryshkov wrote:
Basically, if for some reason (e.g. because the TCPM returns an error to one of the start functions) the drm_bridge is destroyed, the DRM driver isn't notified about the event. It still keeps the pointer to the bridge pointer and can access freed memory afterwards.
Hmm, my concern/question is about the TCPM code triggered by an IRQ firing here, racing with the bridge code.
If you're happy you've reasoned about that and it won't happen, then apply the ACK with the commit log fixed alone.
--- bod