Re: [linux-sunxi] Re: [PATCH 2/7] dt-bindings: add binding for the Allwinner A64 DE2 bus

[Maxime Ripard]

On Tue, Mar 27, 2018 at 07:31:18AM +0800, Icenowy Zheng wrote:
> 
> 
> ä 2018å3æ27æ GMT+08:00 äå6:22:58, Rob Herring <robh@xxxxxxxxxx> åå:
> >On Thu, Mar 22, 2018 at 02:08:43PM +0800, Icenowy Zheng wrote:
> >> 
> >> 
> >> ä 2018å3æ22æ GMT+08:00 äå1:13:42, "Jernej Åkrabec"
> ><jernej.skrabec@xxxxxxxx> åå:
> >> >Hi all,
> >> >
> >> >Dne sreda, 21. marec 2018 ob 03:18:13 CET je Icenowy Zheng
> >napisal(a):
> >> >> ä 2018å3æ21æ GMT+08:00 äå2:46:46, Maxime Ripard
> >> ><maxime.ripard@xxxxxxxxxxx> 
> >> >åå:
> >> >> >On Sat, Mar 17, 2018 at 01:53:49AM +0800, Icenowy Zheng wrote:
> >> >> >> All the sub-blocks of Allwinner A64 DE2 needs the SRAM C on A64
> >> >SoC
> >> >> >
> >> >> >to
> >> >> >
> >> >> >> be claimed, otherwise the whole DE2 space is inaccessible.
> >> >> >> 
> >> >> >> Add a device tree binding of the DE2 part as a sub-bus.
> >> >> >
> >> >> >Where did you get the info that it was a bus?
> >> >> 
> >> >> There's no direct evidence, just some guess.
> >> >> 
> >> >> The DE2 is a whole part that is just allocated a memory
> >> >> space at the user manual, and the SRAM controls the
> >> >> access to all modules in the DE2.
> >> >> 
> >> >> So it might be a bus.
> >> >> 
> >> >> Implement it as a bus is a clear representation on A64.
> >> >
> >> >Since there is already syscon for same mmio region, we migh as well
> >use
> >> >it 
> >> >when loading ccu-sun8i-de2 driver on A64.
> >> >
> >> >Other options, like SRAM driver or bus driver, might better
> >represent
> >> >HW, but 
> >> 
> >> I think the device tree should properly represent the HW,
> >> it's a basic requirment.
> >> 
> >> >then we would have two DT nodes covering same mmio region, which I
> >> >think is 
> >> >not really acceptable.
> >> 
> >> It's acceptable, and DE2 is not the only user of SRAM controller so
> >far.
> >
> >No, it's not acceptable. Don't create overlapping mmio regions in DT.
> 
> Then should the SRAM controller driver be configured to take the syscon?

We could have a single DT node that would export a syscon yes, just
like you did for the R40 ethernet case.

I'm not sure the SRAM controller itself needs to take the syscon, it
just can export its own.

Maxime

-- 
Maxime Ripard, Bootlin (formerly Free Electrons)
Embedded Linux and Kernel engineering
https://bootlin.com

Attachment: signature.asc
Description: PGP signature