Hi Chris,

On Wed, 12 Nov 2025 19:06:35 +0000
Chris Brandt <[email protected]> wrote:

> Hi Hugo,
> 
> On Tue, Nov 11, 2025 10:04 PM, Hugo Villeneuve wrote:
> > You still haven't provided the base-commit. There is a ton of branches in 
> > renesas-devel.git, so it is not easy to determine the branch/commit you 
> > used for your patchset.
> > By providing base-commit, you will save everyone a lot of trouble.
> 
> https://git.kernel.org/pub/scm/linux/kernel/git/geert/renesas-devel.git
> master branch
> base-commit: dd30a345f284e0d9b1755e3538f8257cf4deb79f
> 
> 
> > I tried to apply your patches on branch renesas-geert/master, using 
> > base-commit: 211ddde0823f
> 
> That is what I have been using. I just updated this morning to v6.18-rc5
> 
> 
> > The problem seems to be that dsi_div_ab_desired is not properly 
> > initialized, because the number of lanes is uninitialized (0) at this time, 
> > and therefore
> > rzg2l_cpg_dsi_div_set_divider() gets called with its first parameter as 
> > zero.
> 
> Hmmm.
> There are 2 ways to set the number of lanes: Device Tree or hard coded in the 
> panel driver.
>
> What I do not understand is that by the time rzg2l_mipi_dsi_host_attach() is 
> called, the number of lanes should have already been set.

Let me know if there is something specific you wantme to test for this.

> 
> On your system, where is the number of lanes defined? In the panel driver 
> when it is probed?

For all my panels, it is set in the device three like this:

...
        port@1 {
                dsi0_out: endpoint {
                        data-lanes = <1 2 3 4>;
                        remote-endpoint = <&panel_in>;
                };
        };
...


> I am testing with 3 different systems, and I have not seen this issue.

I have 3 different panels, all using different druivers/IC, but I
tested only one (ilitek-ili9881c driver).

> 
> > Like I said, it also happens on 6.17.7 stable tree. If I remove your 
> > patches, everything is fine.
> 
> Things are hard coded in the existing driver.
> They work for you today because you are using the same number of lanes as the 
> Renesas eval board.
> I'm trying to fix support for lanes = 3,2,1
> 
> 
> Thank you,
> Chris
> 


-- 
Hugo Villeneuve

Reply via email to