Re: [PATCH 1/2] dt-bindings: reset: imx8mq: Add _N suffix to IMX8MQ_RESET_MIPI_CSI*_RESET

[Robby Cai <robby.cai@nxp.com>]

On Tue, Mar 31, 2026 at 02:45:07PM +0200, Philipp Zabel wrote:
> On Di, 2026-03-31 at 18:13 +0800, Robby Cai wrote:
> > The assert logic of the MIPI CSI reset signals is active-low on i.MX8MQ,
> > but the existing names do not indicate this explicitly. To improve
> > consistency and clarity, append the _N suffix to all
> > IMX8MQ_RESET_MIPI_CSI*_RESET definitions. The deprecated
> > IMX8MQ_RESET_MIPI_CSI*_RESET versions remain temporarily for DT ABI
> > compatibility and will be removed at an appropriate time in the future.
> 
> The register description in the latest reference manual I can download,
> IMX8MDQLQRM Rev. 3.1 (06/2021), still call these bits
> MIPI_CSI1_CORE_RESET and so on (without _N). There is no mention of
> polarity in the bitfield description. Is a documentation update
> planned?

Yes. A documentation update is already planned to clarify the reset polarity
(active-low) and naming. The current RM description is incomplete and will be
corrected in a future revision.

> 
> Right now I'd say this improves clarity, but reduces consistency with
> existing documentation.
> 
> Are these bits self-clearing, or can the reset be asserted by writing
> 0? As it stands, the CSI driver using these resets, imx8mq-mipi-csi2.c,
> only calls reset_control_assert() in imx8mq_mipi_csi_sw_reset():
> 
>           /*                                                          
>            * these are most likely self-clearing reset bits. to make it
>            * more clear, the reset-imx7 driver should implement the   
>            * .reset() operation.                                      
>            */                     
>           ret = reset_control_assert(state->rst);
> 
> This will probably have to be turned into a deassert together with the
> reset driver change.

No, these reset bits are not self-clearing.

According to the design team, the MIPI CSI reset logic on i.MX8MQ is identical
to MIPI DSI: the reset is active-low and must be explicitly deasserted.
Writing ��0�� asserts reset and it will remain asserted until cleared by software.

The current assumption in the CSI driver that these bits are self-clearing is
therefore incorrect. This was exposed by the landing patch [1].

To fix this properly, the reset-imx7 driver needs to reflect the correct polarity,
and the CSI driver should use deassert instead of only calling
reset_control_assert(). The RM will also be updated to clarify this behavior.

[1] https://git.linuxtv.org/media.git/commit/?id=6d79bb8fd2aa25afccbd6aeec2821722fa0b5db5


Regards,
Robby