Re: [PATCH 2/2] firmware: arm_scmi: Support 'reg-io-width' property for shared memory

[Florian Fainelli]

On 8/12/24 11:01, Cristian Marussi wrote:
> > OK, I will think about more about what needs to be done here, but in
> > general, do you agree this is an acceptable approach to support "odd" SRAMs?
>
> Yes, but one question comes up in my mind upfront (maybe similar to Rob remarks):
> is this not, in theory, something general that should be somehow addressed transparently
> by the core SRAM code when dealing with such odd SRAM, since SCMI is
> indeed only onne of the possible users ?
> (not saying to do this in this series that deals with SCMI related issues....)
>
> Anyway, I'll have a though too about the SCMI core transport possible changes that I
> mentiond above, soon-ish... (I tried something already today, hoping to solve it quickly
> ...with poor results :D)

What do you think about keeping the status quo with the scmi_shared_mem 
singleton and instead introduce a per-shmem scmi_shared_mem_io_ops which 
would contain function pointers to read from/write to the shared memory?

This would keep the scmi_shared_mem read-only and a singleton and the 
transport would be responsible for storing and passing that set of 
function pointers around, post setup_iomap() where the determination 
would have been done.
--
Florian