[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Message-ID: <1c8f39015eae7cc71fbfb9136af8a728d743a991.camel@codeconstruct.com.au>
Date: Fri, 08 Nov 2024 10:07:31 +1030
From: Andrew Jeffery <andrew@...econstruct.com.au>
To: Potin Lai <potin.lai.pt@...il.com>
Cc: Rob Herring <robh@...nel.org>, Krzysztof Kozlowski <krzk+dt@...nel.org>,
Conor Dooley <conor+dt@...nel.org>, Joel Stanley <joel@....id.au>, Patrick
Williams <patrick@...cx.xyz>, devicetree@...r.kernel.org,
linux-arm-kernel@...ts.infradead.org, linux-aspeed@...ts.ozlabs.org,
linux-kernel@...r.kernel.org, Potin Lai <potin.lai@...ntatw.com>, Cosmo
Chou <cosmo.chou@...ntatw.com>
Subject: Re: [PATCH 1/2] ARM: dts: aspeed: catalina: update pdb board cpld
ioexp linename
On Thu, 2024-11-07 at 19:22 +0800, Potin Lai wrote:
> On Thu, Nov 7, 2024 at 7:41 AM Andrew Jeffery
> <andrew@...econstruct.com.au> wrote:
> >
> > On Wed, 2024-11-06 at 16:58 +0800, Potin Lai wrote:
> > > Update the GPIO linename of each PDB CPLD IO expander based on
> > > latest
> > > CPLD firmware.
> >
> > What version is the latest CPLD firmware? What was the previous
> > version
> > with the old pin assignments?
>
> Because the hardware changes from EVT to DVT, the CPLD firmware
> reallocated the IOEXP pin mapping in DVT version.
> I will add more description into the commit message in the next
> version.
If you have different revisions of the board, it would seem sensible to
have separate devicetrees, one for each, rather than constantly
evolving one devicetree? Tack on an `-evt`/`-dvt` suffix as required?
>From there you can always have a suffix-less dts file that #includes
the most recent board revision. See some of the Aspeed EVB devicetrees
for an example.
>
> >
> > I'm also interested in some discussion of the coordination between
> > CPLD
> > firmware, the devicetree and the BMC userspace configuration. This
> > change feels pretty painful.
>
> I am not from the CPLD firmware team,
I don't see why you need to be? This is a cross-component concern, and
you need to make all the pieces of the puzzle line up.
> I only know our CPLD team was
> redesigning the entire struct which causes the huge changes of IOEXP
> pins.
>
> This is probably a different topic, I am curious about is it possible
> to assign the linename in userspace?
> In OpenBMC, there are many services that depend on GPIO linename, it
> will be more flexible if I can assign the linename before service
> starts.
Not that I'm aware, and to determine otherwise I'd probably need to
read the implementation as much as you :)
However, separating the devicetrees would go a long way here if the
CPLD firmware is tied to the board revision...
Andrew
Powered by blists - more mailing lists