Re: [PATCH v3 2/2] spi: cadence: add support for Cadence XSPI controller

[Pratyush Yadav <p.yadav@xxxxxx>]

On 08/09/21 07:27AM, Parshuram Raju Thombare wrote:
> >Depends on SPI_MEM as well.
> 
> Ok
> 
> >I commented on this last time around as well. This does not look right
> >at all. A SPI MEM based driver should *not* need to know anything about
> >the subsystem driving it. That is the entire point of the API.
> >
> >The controller seems to be able to extract the read and write opcodes
> >from the SFDP on its own since you don't pass in that information to
> >cdns_xspi_nor_read(). It looks like it is tied very heavily to a NOR
> >flash, and I am not sure if it can really be used with a NAND flash, or
> >something else entirely.
> >
> >Which makes me wonder how we should handle controllers like these. I
> >don't think they fit in very well with the SPI MEM model, since they
> >can't execute arbitrary SPI MEM commands very well. At the same time we
> >are trying to get rid of mtd/spi-nor/controllers. Dunno...
> >
> >Mark, Tudor, Vignesh, any ideas?
> 
> Ok, then for now I will drop ACMD PIO mode and use only STIG mode.
> In STIG mode driver configures bus width and clock edge mode for
> command, address and data for each operation. 

But it would reduce performance by a lot, no? I think we should try to 
figure out how we can accomodate controllers like this before resorting 
to using the slower modes.

-- 
Regards,
Pratyush Yadav
Texas Instruments Inc.