[PATCH 3/5] drm/dp/mst: change MST detection scheme

Alex Deucher alexdeucher at gmail.com
Wed Jan 25 17:24:35 UTC 2017


On Wed, Jan 25, 2017 at 12:16 PM, Krzysztof Nowicki
<krzysztof.a.nowicki at gmail.com> wrote:
> Hi,
>
> On środa, 15 czerwca 2016 14:14:01 CET  wrote:
>> On 15 June 2016 at 13:49, Dave Airlie <airlied at gmail.com> wrote:
>> > Excuse me for top posting.
>> >
>> > So I finally got back to this patch, still don't like it.
>> >
>> > Apart from the doing 3 things in once which is just annoying, current
>> > userspace for tiled monitors
>> > relies on the behaviour that the cached edids are retrieved before the
>> > connector is show to userspace.
>> >
>> > This is so that for tiled monitors both connectors are shown with
>> > tiling properties so userspace can pick the correct monitor size.
>> >
>> > Now I realise this isn't great, since at least for the two cable 5K
>> > monitors you won't ever plug them in at once.
>> >
>> > So we should probably discuss more what userspace should do in the
>> > presence of a tiled monitor where it can only see one tile,
>> >
>> > If we go ahead with something like this, things will look kinda ugly
>> > as you plug in a 4k monitor and it'll try and set a mode
>> > on the first tile when it gets the hotplug (or it could be racing and
>> > notice the connectors before the hotplug). So it then
>> > sets the 1920x1080 mode on one connector and gets another hotplug to
>> > set the tiled mode up.
>>
>> Though I think the fix is to leave the cached edid test in add port,
>>
>> And maybe the rest of the patch is okay then.
>>
>> Dave.
>
> Is this patch still being considered for pulling in?
>
> Recently I've been fighting with a HP UltraSlim docking station with a MST hub
> inside in order to get it working with the amdgpu DAL/DC code. I couldn't get
> it to work until I stumbled upon this patch. Applying it solved the issue
> instantly.
>
> What I have found during debugging is that the MST hub would return SST sink
> port numbers < 8 which resulted in no EDID being retrieved and in turn the
> status of the outputs was 'disconnected'.

Harry,

Any chance you could re-spin this against something more recent and re-send?

Alex


More information about the dri-devel mailing list