Re: [PATCH v2 5/7] dt-bindings: pinctrl: airoha: add AN7589

From: John Crispin

Date: Mon Nov 16 2026 - 14:39:34 EST


Hi Conor,

The largest users are EN7523 pcm_spi 14, AN7563 pnand 17, AN7583
pnand 17, EN7528 pnand 17, EN7581 pnand 17 and AN7589 emmc 12 (these
numbers are based on a hackish "grep foo | bar", so they might be
slightly off).

I copied and modified what the other AN/EN bindings do. These set the
limit to the pin count of the SoC: EN7523 38, AN7563 38, AN7583 53,
EN7528 50, EN7581 58. Your call really, I do not mind either way. I can
send a v4 and then a follow up series cleaning up the existing bindings
as well.

John

On 16.11.26 19:15, Conor Dooley wrote:
> On Mon, Nov 16, 2026 at 06:47:27PM +0100, John Crispin wrote:
>> On Fri, Nov 13, 2026 at 05:40:42PM +0000, Conor Dooley wrote:
>>> As Lee requested, please split this out. Two patches, maybe even three.
>>
>> Agreed. v3 splits it into three patches.
>>
>>> Do we even need a child node here? If all this node does is
>>> gpio/pinctrl, why can the driver not bind directly to this node?
>>>
>>> What else does it do that warrants being a syscon?
>>
>> The block also holds the PWM controller. Its registers interleave with
>> the GPIO ones. The PWM series adds it as a second child, as on the
>> EN7581. That series has a depency on this one so I have not posted it yet.
>>
>>> Otherwise, there should only be one. uart1 and uart2 should have
>>> different nodes.
>>
>> Agreed. v3 sets maxItems to 1. Only pcm_spi and pcie keep more, as
>> their groups carry different signals of one device.
>>
>>> So here, maxItems should be 1 for similar reasons.
>>
>> Agreed, v3 does this for i2c too.
>>
>>> I'm also curious about why this goes up to 74. None of the functions
>>> have 74 pins, surely?
>>
>> A conf node can set the same config on any set of pins. 74 is all the
>> pins. The EN7581 binding does the same with 58.
>
> A difference in philosophy I guess - I would be grouping the pins in the
> config node by user. So that there would be something like
>
> spi1-stuff {
> spi1-function {
> function = spi
> group = spi1
> }
>
> spi1-config1 {
> pins = 1, 2
> config1
> config2
> }
>
> spi2-config2 {
> pins = 3
> config1
> }
> }
>
> spi {
> pinctrl-0 = spi1-function, spi-config1, spi-config2
> }
>
> or
>
> spi {
> pinctrl-0 = spi1-stuff
> }
>
> IMO, having config nodes covering more than one "feature" doesn't really
> make sense, as you'd end up with something like:
>
> spi1-function {
> function = spi
> group = spi1
> }
>
> all-config1-config2 {
> pins = 1, 2, 4, 5
> config1
> config2
> }
>
> all-config1 {
> pins = 3
> config1
> }
>
> which kinda falls apart I think when a node can request a setting for
> pins that are not part of it (in this case, that's pins 4 and 5)
>
> spi {
> pinctrl-0 = spi1-function, all-config1, all-config1-config2
> }
>
> So my opinion is that the max number of items should be the size of the
> largest peripheral.
>
> What's the largest actual user on the 7581?
>
>>
>> v3 will follow shortly.
>>
>> John
>>
>>