Re: intel_dp_detect redesign

From: Daniel Vetter <daniel@ffwll.ch>
To: Daniel Stone <daniel@fooishbar.org>
Cc: "Ander Conselvan de Oliveira"
	<ander.conselvan.de.oliveira@intel.com>,
	intel-gfx <intel-gfx@lists.freedesktop.org>,
	"Stéphane Marchesin" <marcheu@google.com>
Subject: Re: intel_dp_detect redesign
Date: Mon, 30 Nov 2015 09:19:48 +0100	[thread overview]
Message-ID: <20151130081948.GJ17050@phenom.ffwll.local> (raw)
In-Reply-To: <CAPj87rOAcTnw4kF-SMeHcvg5t87EHmjNWjRjh9ZZwe8SRxVoHA@mail.gmail.com>

On Fri, Nov 27, 2015 at 04:30:21PM +0000, Daniel Stone wrote:
> Hi,
> +marcheu
> 
> On 26 November 2015 at 10:07, Daniel Vetter <daniel@ffwll.ch> wrote:
> > On Wed, Nov 25, 2015 at 05:09:02PM +0530, Thulasimani, Sivakumar wrote:
> >> however good to explicitly check for this,
> >> following needs to be tested before sending in next patch/merge
> >> 1) MST displays verification (Ander's reported on first set of patches)
> >> 2) check behavior on sleep - resume (dave&danvet)
> >> 3) storm handling needs to be handled as well. (i assume this should be
> >> fine,
> >>      but good to check explicitly) (danvet)
> >>
> >
> > Yeah the storm mitigation will keep on working. What I'm worried about is
> > that polling won't work any more: When a storm happens we disable the hpd
> > and switch all affected connectors completely to polling. Polling happens
> > through the probe helpers in drm_probe_helper.c, and that code exclusively
> > uses ->detect callbacks. Which means if we no longer re-probe in detect
> > (since we assume hpd works correctly) then this will break the storm
> > handling code.
> >
> > Simplest fix (but a bit a hack) would be to check whether polling is
> > enabled at the top of intel_hdmi_detect and if so execute a full probe.
> > And not just return the cached values.
> >
> > Note that storms are only a concern for HDMI, not DP (somehow DP hw is
> > less shit).
> 
> Hmm, from what I understand it's a concern on DP as well. Maybe due to
> DP->HDMI converters which just pass HPD through? Google's Type-C -> DP
> cable crushes all short HPD events - breaking MST short pulses - which
> I assume wasn't for no reason:
> https://chromium.googlesource.com/chromiumos/platform/ec/+/master/board/dingdong/board.c#27

Yeah right, we do need to keep polling working for DP too, but only for
long pulse hpd replacement. As long as a DP cable is plugged in I haven't
seen reports of a storm. But indeed hdmi crap can induce a storm on a DP+
connector too.
-Daniel
-- 
Daniel Vetter
Software Engineer, Intel Corporation
http://blog.ffwll.ch
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/intel-gfx