Re: [GIT PULL] Rockchip dts64 changes for 6.16 #1

["Arnd Bergmann" <arnd@arndb.de>]

On Fri, May 9, 2025, at 23:37, Heiko Stübner wrote:
> Am Freitag, 9. Mai 2025, 23:03:54 Mitteleuropäische Sommerzeit schrieb 
>> 
>> but the corresponding nodes are left at disabled. I see
>> that the same mistake is present in the uart nodes.
>> 
>> Please send a fixup to remove these from the .dtsi file
>> here and the similar chips, unless you are sure that every
>> board will have them enabled (like e.g. the gpio nodes).
>> 
>> The aliases are not meant to refer to soc-internal names,
>> but the identifiers on board.
>
> For the uarts and i2c (and spi), all the identifiers are always numerical
> both in the SoC documentation as well as on the boards and board
> schematics.
>
> If you look in a random Rockchip schematic file, the lines for the i2c0
> controller will be called i2c0_scl_foo, i2c0_sda_foo, etc.
> Similar uart0_tx, uart0_rx, etc.
>
> So while I fully understand that mmc0 -> emmc, mmc1 -> sd-card are very
> much board specific, somehow repeating the very same i2c aliases for every
> board feels strange. The 7th i2c controller on the soc, will never be
> called anything else than i2c7 afterall.

It's usually most important for the uart, since that tends to go
to an external connector on the board that may use a different
numbering, or may not even use the one you named serial0, so these are
renumbered most of the time.

For i2c, I can see a reason for keeping the numbering the same
as the on-chip pins, but only have an alias for those that are
actually enabled. We should probably have dtc check that there
are no aliases to disabled devices, but that requires cleaning
up a load of board files first.

> But ok, if you feel strongly about that, I'll move the i2c and uart aliases.

Yes, please do.

      Arnd

_______________________________________________
Linux-rockchip mailing list
Linux-rockchip@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-rockchip