Re: [PATCH 2/2] firmware: smccc: Add ARCH_SOC_ID support

From: Arnd Bergmann <arnd@arndb.de>
To: Sudeep Holla <sudeep.holla@arm.com>
Cc: Linux ARM <linux-arm-kernel@lists.infradead.org>,
	Will Deacon <will@kernel.org>,
	Mark Rutland <mark.rutland@arm.com>,
	Lorenzo Pieralisi <lorenzo.pieralisi@arm.com>,
	"linux-kernel@vger.kernel.org" <linux-kernel@vger.kernel.org>,
	harb@amperecomputing.com, Jose.Marinho@arm.com,
	Greg Kroah-Hartman <gregkh@linuxfoundation.org>,
	Francois Ozog <francois.ozog@linaro.org>
Subject: Re: [PATCH 2/2] firmware: smccc: Add ARCH_SOC_ID support
Date: Sat, 23 May 2020 21:40:17 +0200	[thread overview]
Message-ID: <CAK8P3a1+jjgOyJcRQm60RULjwtvWzvYK1QwrjOX_aRAbDGKuJg@mail.gmail.com> (raw)
In-Reply-To: <20200523172721.GC18810@bogus>

On Sat, May 23, 2020 at 7:27 PM Sudeep Holla <sudeep.holla@arm.com> wrote:
> On Fri, May 22, 2020 at 08:41:59PM +0200, Arnd Bergmann wrote:
> > On Fri, May 22, 2020 at 6:54 PM Sudeep Holla <sudeep.holla@arm.com> wrote:
>
> > jep106:5678 (the IMP_DEF_SOC_ID field in my example) would
> > probably be sufficient to not conflict with a another soc_device
> > driver, but is quite likely to clash with an ID used by another
> > manufacturer.
> >
>
> IIUC, you are fine with "jep106:1234:5678" where 1234 is jep106 manufacture
> id code and 5678 is soc_id as it may avoid all the conflicts across
> the manufacture namespaces.

I think either jep106:5678 or jep106:1234:5678 (or some variation with
different field separators if you prefer) would be fine here.

> > jep106:1234 (the manufacturer ID) in turn seems too broad from
> > the soc_id field, as that would include every chip made by one
> > company.
> >
>
> I understand, and IIUC you prefer this to be embedded with soc_id
> especially to avoid namespace conflicts which makes sense.
>
> Thanks for all the discussion and valuable inputs. I am now bit nervous
> to add SoC info from SMCCC ARCH_SOC_ID to sysfs yet as we need more info
> especially "machine" and "serial_number" for elsewhere(OEM firmware mostly
> from DT or ACPI).

I probably wouldn't mix those in with the same driver. If machine and
serial_number have no smccc interface, then they should be left out,
but there could be a separate soc_device that gets that information
by other means, usually using one of the existing hardware id register
drivers.

> TBH, the mix might be bit of a mess as there are soc drivers that rely
> on DT completely today. Anyways, this SOC_ID can be added as library that
> can be used by a *generic* soc driver once we define a standard way to
> fetch other information("machine" and "serial_number"). I am happy to
> get suggestions on that front especially from you and Francois as you
> have got some context already.

Well, I suppose we could have the entire data from the smccc interface
in a central place somewhere, such as (to stay with my example)
"1234:5678:9abcdef0" in an attribute of any soc_device driver that
adds a call to a library function for the jep106 ID, or possibly in
/sys/firmware or even a field added to /proc/cpuinfo.

       Arnd