Re: [PATCH v2 14/15] spi: dt-bindings: snps,dw-ahb-ssi: Add generic dw-ahb-ssi version

[Serge Semin <fancer.lancer@xxxxxxxxx>]

On Tue, Dec 13, 2022 at 05:47:53PM +0000, Sudip Mukherjee wrote:
> On Tue, Dec 13, 2022 at 4:59 PM Mark Brown <broonie@xxxxxxxxxx> wrote:
> >
> > On Tue, Dec 13, 2022 at 10:32:09AM -0600, Rob Herring wrote:
> > > On Mon, Dec 12, 2022 at 06:07:31PM +0000, Sudip Mukherjee wrote:
> >
> > > > Add new snps,dw-ahb-ssi version to the bindings.
> >
> > > Really, this should probably only be a fallback with an SoC specific
> > > compatible. Future quirk properties which are not board specific only
> > > will be rejected. You've been warned.
> >
> > Given how widely used DesignWare stuff is and usage in FPGAs it does
> > seem reasonable to have compatibles for just the IP rather than SoC
> > specific ones - we do have quirked versions that have been modified but
> > these are things that people manage to deploy without needing that and
> > SoC specific compatibles for FPGA instantiations would get painful.
> 

> Also, this patchset adds the autodetect procedure as discussed in the review
> of the previous series at
> https://lore.kernel.org/lkml/20220826233116.uulisbo663cxiadt@mobilestation/
> 
> So, we should be able to replace "snps,dw-apb-ssi" and
> "snps,dwc-ssi-1.01a" with "snps,dw-ahb-ssi" after this.

Just "snps,dwc-ssi-1.01a". That is the IP-core
https://www.synopsys.com/dw/ipdir.php?c=dwc_ssi
which support was added in
https://lore.kernel.org/linux-spi/20200505130618.554-4-wan.ahmad.zainie.wan.mohamad@xxxxxxxxx/
Since the IP-core version is auto-detectable there is no need in
having the version attached to the compatible string. That's why I
asked @Sudip to introduce a new generic device name free of the
version suffix. It should be used instead of "snps,dwc-ssi-1.01a"
from now.

The "snps,dw-apb-ssi" compatible string will stay since it corresponds
to another IP-core:
https://www.synopsys.com/dw/ipdir.php?c=DW_apb_ssi

Answering to the @Rob note regarding the quirk properties. All the
features added by Sudip here were supposed to be auto-detectable.
So the generic IP-core name will be still useful as before with no
need in adding any quirks.

@Rob AFAIR you used to be against the generic fallback compatible
names, but had nothing against just generic compatibles. Has something
changed?

-Serge(y)

> And, also this generic compatible has been tested with the new 1.03a
> version we are working with, which was
> mentioned in my v1 at
> https://lore.kernel.org/lkml/20220826233305.5ugpukokzldum7y5@mobilestation/
> 
> -- 
> Regards
> Sudip