[Intel-gfx] [PATCH] drm/i915: delay hotplug scheduling

[Chris Wilson]

Quoting Jani Nikula (2018-09-26 08:06:56)
> On Tue, 25 Sep 2018, Chris Wilson <chris at chris-wilson.co.uk> wrote:
> > Quoting Jani Nikula (2018-09-25 08:18:36)
> >> On some systems we get the hotplug irq on unplug before the DDC pins
> >> have been disconnected, resulting in connector status remaining
> >> connected. Since previous attempts at using hotplug live status before
> >> DDC have failed, the only viable option is to reduce the window for
> >> mistakes. Delay the hotplug processing.
> >
> > The only concern I would raise is there some expectation for latency of
> > hotplug? 300ms doesn't seem like it would irk the user -- it won't be
> > noticeable unless using a kvm or equivalent. I would be more concerned
> > if there was a line in the conformance tests that expect a response to a
> > hotplug interrupt within a certain time window.
> 
> Frankly, I don't know. But I suppose a conformance test like that would
> like the retry approach even less?

The retry at least has variable backoff, so in the case where the
unplug completed before the interrupt handler, the latency would be
minimal. And then it could do a 1,2,..N jiffie backoff before giving up.
(As I understood the problem at least.)
-Chris