Re: [RFC PATCH] arm64: dts: qcom: add device tree for DB845c working in dual DSI mode

[Bjorn Andersson <andersson@kernel.org>]

On Fri, Apr 07, 2023 at 08:09:23PM +0300, Dmitry Baryshkov wrote:
> On 07/04/2023 19:48, Bjorn Andersson wrote:
> > On Wed, Jan 18, 2023 at 08:47:53PM +0200, Dmitry Baryshkov wrote:
> > > 18 января 2023 г. 19:32:42 GMT+02:00, Bjorn Andersson <andersson@kernel.org> пишет:
> > > > On Wed, Jan 18, 2023 at 10:20:48AM +0200, Dmitry Baryshkov wrote:
> > > > > While we are testing the DB845c working in dual DSI mode (which is
> > > > > required for the 4k support), create a separate device tree using this
> > > > > configuration.
> > > > > 
> > > > 
> > > > I don't think it's right to change the DTB on your device depending on
> > > > your TV's ability to do 4k or not...
> > > > 
> > > > I ran into this problem when playing around with 4k-support on the RB3
> > > > as well a while back. I was not able to figure out a way to run the
> > > > lt9611 off two inputs for the lower resolutions and ended never posted
> > > > anything because I couldn't come up with any good solution.
> > > 
> > > 
> > > This is implemented with the latest lt9611 patchset, referenced in the patch annotation.
> > > 
> > > I posted this as an RFC with the separate dts to let people test dual DSI, DPU wide planes, etc. and to compare that with plain old single link DSI/lt9611.
> > > 
> > > > 
> > > > 
> > > > Would it be possible somehow to describe the two ports in DT, but have
> > > > the display driver use one or both depending on the number of inputs
> > > > needed by the lt9611?
> > > 
> > > This would be very problemsome. Even switching the number of lanes
> > > goes against the standard and is not supported by Linux kernel.
> > > Supporting switching the host off and on would be a real troublle.
> > > 
> > 
> > Does this imply that switching resolution in runtime is not possible for
> > any of these DSI->HDMI/DP/eDP bridges?
> > 
> > The immediate worry is that we need 4k and non-4k variants of every
> > mezzanine (unless we start doing overlays). But beyond that I still
> > don't like the fact that you need to change firmware depending on which
> > monitor/TV you're going to connect the board to. (And with the current
> > boot.img-based design, this is overly cumbersome)
> No. With the latest lt9611 driver and with the current msm/msm-next it
> should not be necessary to have a separate dtsi. I submitted this patch as
> an RFC to let people compare 2xDSI vs single-link cases. Currently I'm happy
> with the way it works, so I can probably send this as a patch to main
> sdm845-db845c.dtb.
> 
> As a background info:
> - lt9611 driver can utilize either one or two DSI links
> - drm/msm can output 4k over two DSI links in bonded mode.
> - drm/msm will still utilize two DSI links for smaller resolutions
> (switching between single and bonded DSI modes doesn't seem practical).
> - If just one DSI link is supplied, 1080p is the maximum resolution that can
> be supported.
> 

Sorry, I missed your reply earlier. If we can run the lower resolutions
off two bonded DSI I think we should make that the default
configuration - it's just that when I tried it a few years ago I had
problems getting a working output... But much has changed since then.

There is a dip switch to allow one of the DSI links to be passed onto
one of the HS-connectors, but for anyone to use that they would need a
mezzaine dts anyways, so they can unbond the DSIs there.

Regards,
Bjorn