On 11/07/2025, Ahmad Fatoum wrote:
> Hi,

Hi,

> 
> On 04.11.25 21:44, Fabio Estevam wrote:
>> On Mon, Oct 20, 2025 at 11:53 PM Liu Ying <[email protected]> wrote:
>>> On 10/20/2025, Fabio Estevam wrote:
>>>> On Tue, Oct 14, 2025 at 1:12 AM Liu Ying <[email protected]> wrote:
>>>>
>>>>> Strictly speaking, I don't think i.MX6SX LCDIF is compatible with i.MX28 
>>>>> LCDIF
>>>>> because at least i.MX28 LCDIF has the version and debug{0,1,2} 
>>>>> registers(at
>>>>> 0x1c0, 0x1d0, 0x1e0 and 0x1f0) while i.MX6SX LCDIF hasn't.
> 
> Thanks for pointing this out. In my opinion, these registers are auxiliary
> and don't really change the compatibility situation as a functional driver
> can be written without their use, evidenced by the Linux driver doing just
> fine without using these registers.

DT bindings should after all describe hardwares and in theory they should not
consider software implementation, that's why I said 'strictly speaking'.

People may argue that potential software would access those "auxiliary"
registers and hence i.MX28 LCDIF is not a fallback for i.MX6SX LCDIF.

And, to me, register at 0x1e0, i.e., HW_LCDIF_DEBUG1 is not that "auxiliary"
at least for Linux, because Linux DRM supports getting current display scanout
position to generate accurate vblank timestamp.  HW_LCDIF_DEBUG1 actually
provides the interface to read scanout position.

https://elixir.bootlin.com/linux/v6.18-rc1/source/include/drm/drm_modeset_helper_vtables.h#L448

> 
>>>> There are some DT users, such as Barebox that matches against
>>>> fsl,imx28-lcdif, so we cannot remove it.
>>>
>>> Hmmm, it looks like software projects like Barebox don't really follow this 
>>> DT
>>> binding.  Is it possible to fix Barebox to avoid changing this DT binding by
>>> this patch?  I'm assuming that Uboot has already been fixed.
>>> What do you think?
> 
> I am sorry my prior feedback ended up stalling this series.
> 
> There is a lot of regressions happening due to upstream DT changes and I am
> just trying to raise awareness. Another example I stumbled over today:
> https://lore.kernel.org/all/[email protected]/
> 
> 
> I have submitted a patch[1] to barebox adding explicit i.MX6 SoloX support,
> so, having expressed my opinion above, please proceed as you see fit.

Fabio, it seems after Barebox is fixed, you can fix the LCDIF compatible
strings in imx6sx.dtsi?  I don't know if there is any other software project
which is taking fsl,imx28-lcdif as a fallback for i.MX6SX LCDIF...

> 
> [1]: 
> https://lore.barebox.org/barebox/[email protected]/
> 
> Cheers,
> Ahmad
> 
>>
>> Thanks
>>
> 
> 


-- 
Regards,
Liu Ying

Reply via email to