Re: [PATCH v4 14/39] drm/msm/dp: introduce max_streams for DP controller MST support

[Dmitry Baryshkov]

On Fri, May 22, 2026 at 03:58:06PM +0800, Yongxing Mou wrote:
> 
> 
> On 5/21/2026 8:07 PM, Dmitry Baryshkov wrote:
> > On Thu, May 21, 2026 at 07:34:32PM +0800, Yongxing Mou wrote:
> > > 
> > > 
> > > On 4/12/2026 1:59 AM, Dmitry Baryshkov wrote:
> > > > On Fri, Apr 10, 2026 at 05:33:49PM +0800, Yongxing Mou wrote:
> > > > > From: Abhinav Kumar <quic_abhinavk@xxxxxxxxxxx>
> > > > > @@ -2740,7 +2741,8 @@ static int msm_dp_ctrl_clk_init(struct msm_dp_ctrl *msm_dp_ctrl)
> > > > >    	if (rc)
> > > > >    		return rc;
> > > > > -	for (i = DP_STREAM_0; i < DP_STREAM_MAX; i++) {
> > > > > +	ctrl->num_pixel_clks = 0;
> > > > > +	for (i = DP_STREAM_0; i < max_stream; i++) {
> > > > 
> > > > Why? I think the code was fine with getting up to DP_STREAM_MAX clocks.
> > > Sure.. will change max_stream -> DP_STREAM_MAX next version.
> > > > In fact, I'd rather use that instead of the hardcoded value for
> > > > determining how many streams can be there. Think of the old DTs which
> > > > didn't have extra stream clocks. We still need to support those.
> > > > 
> > > The current logic only assigns num streams to dp->max_stream when
> > > desc->mst_streams == num_pixel_clks, otherwise set to 1 which mean sst.
> > > I would like to introduce a switch in the driver to control the number of
> > > streams, rather than having it fully determined by the DTs.
> > 
> > Why?
> > 
> 1. In some scenarios, even if the hardware supports up to 4-stream MST, we
> may not want to enable all streams (e.g., LeMans: 2xDSI + 4xMST + 2xMST, but
> it only has a 6-layer mixer).

6 layer mixers, not 6-layer mixer (the latter means the mixer which
supports 6 blending layers).

Even with 6 LMs, there is no reason to prevent users from using the
hardware in the way they want. Any artifical limitations are NAKed.

> 2. Currently, the DTs for all platforms have already been merged, and for
> some legacy platforms, not sure if all platforms MST will work.

You are a part of the display team of the SoC manufacturer. You can test
the platforms and make sure that MST will work. I should be able to help
you with some of those, maybe except the MSM8998 / SDM660 generation.

> If my understanding is incorrect, please feel free to correct me. Thanks.
> > > Do you think it would be acceptable to use:
> > > dp->max_stream = min(num_pixel_clks, desc->mst_streams)?
> > 
> > We know that num_pixel_clks <= desc->mst_streams. Otherwise DT is wrong
> > and should not have been comitted. So, the min(a,b) has a known return
> > value of num_pixel_clks.
> > 
> > > 
> > > > >    		ctrl->pixel_clk[i] = devm_clk_get(dev, pixel_clks[i]);
> > > > >    		if (i == 0 && IS_ERR(ctrl->pixel_clk[i]))
> > 
> 

-- 
With best wishes
Dmitry