Re: [RFC PATCH 1/2] dt-bindings: PCI: Add Apple PCI controller

From: Marc Zyngier <maz@kernel.org>
To: Mark Kettenis <openbsd@xs4all.nl>
Cc: "Alyssa Rosenzweig" <alyssa@rosenzweig.io>,
	linux-pci@vger.kernel.org, "Bjorn Helgaas" <bhelgaas@google.com>,
	"Rob Herring" <robh+dt@kernel.org>,
	"Lorenzo Pieralisi" <lorenzo.pieralisi@arm.com>,
	"Krzysztof Wilczyński" <kw@linux.com>,
	"Stan Skowronek" <stan@corellium.com>,
	"Mark Kettenis" <kettenis@openbsd.org>,
	"Sven Peter" <sven@svenpeter.dev>,
	"Hector Martin" <marcan@marcan.st>,
	devicetree@vger.kernel.org, linux-kernel@vger.kernel.org
Subject: Re: [RFC PATCH 1/2] dt-bindings: PCI: Add Apple PCI controller
Date: Sun, 15 Aug 2021 10:12:44 +0100	[thread overview]
Message-ID: <878s131377.wl-maz@kernel.org> (raw)
In-Reply-To: <1566004903.6140692.1629015053757@ox-webmail.xs4all.nl>

Hi Mark,

On Sun, 15 Aug 2021 09:10:53 +0100,
Mark Kettenis <openbsd@xs4all.nl> wrote:
> 
> Hi Marc,
> 
> What can I do to make progress with my binding proposal? It seems we're stuck
> on the MSI issue where you and robh disagree. I still think your idea of
> describing the MSIs as a range makes much more sense than describing them
> individually and bunching them together with the host bridge port interrupts.
> 

It looks like I missed an email from Rob, which explains why we're in
limbo (it was left unread and unmarked, which in my flow means "read
once I have too much time on my hands"). Apologies for that, I'll try
and reply tonight (travelling at the moment).

>     Op 15-08-2021 09:09 schreef Marc Zyngier <maz@kernel.org>:
> 
>     Hi Alyssa,
>    
>     On Sun, 15 Aug 2021 05:25:24 +0100,
>     Alyssa Rosenzweig <alyssa@rosenzweig.io> wrote:
>    
>         Document the properties used by the Apple PCI controller. This is a
>         fairly standard PCI controller, although it is not derived from any
>         known non-Apple IP.
>        
>         Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
>    
>     I would rather you post something as an extension to Mark's work, for
>     multiple reasons:
>    
>     - Mark's patch is still being discussed, and is the current
>     reference (specially given that it is already in use in OpenBSD and
>     u-boot).
>    
>     - we cannot have multiple bindings. There can only be one, shared
>     across implementations. Otherwise, you need a different kernel
>     depending on whether you are booting from m1n1 or u-boot.
>    
>     - what you have here is vastly inconsistent (you are describing the
>     MSIs twice, using two different methods).
> 
> That's probably my fault. The current u-boot device tree is a bit of a
> Frankenstein thing to ease the transition from my initial binding to the
> current proposal. I should clean that up at some point.

That would certainly help. There are a lot of moving pieces at the
moment, and it is getting hard to get a clear picture of what is using
what.

Thanks,

	M.

-- 
Without deviation from the norm, progress is not possible.