Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 23.128.96.18 as permitted sender) client-ip=23.128.96.18;
Date:   Wed, 6 Oct 2021 00:33:25 +0200
From:   Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
To:     Pali =?UTF-8?B?Um9ow6Fy?= <pali@kernel.org>
Cc:     Bjorn Helgaas <bhelgaas@google.com>, linuxarm@huawei.com,
        mauro.chehab@huawei.com,
        Krzysztof =?UTF-8?B?V2lsY3p5xYRza2k=?= <kw@linux.com>,
        Binghui Wang <wangbinghui@hisilicon.com>,
        Lorenzo Pieralisi <lorenzo.pieralisi@arm.com>,
        Rob Herring <robh@kernel.org>,
        Xiaowei Song <songxiaowei@hisilicon.com>,
        linux-kernel@vger.kernel.org, linux-pci@vger.kernel.org
Subject: Re: [PATCH v12 03/11] PCI: kirin: Add support for a PHY layer
Message-ID: <20211006003325.6106cfab@coco.lan>
In-Reply-To: <20211005203148.gn2f34pfvm62w6ca@pali>
References: <cover.1632814194.git.mchehab+huawei@kernel.org>
        <8a6d353145c0ec169d212094f5d534f93e2597f8.1632814194.git.mchehab+huawei@kernel.org>
        <20211005203148.gn2f34pfvm62w6ca@pali>
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: quoted-printable
Precedence: bulk

Hi Pali,

Em Tue, 5 Oct 2021 22:31:48 +0200
Pali Roh=C3=A1r <pali@kernel.org> escreveu:

> Hello!
>=20
> On Tuesday 28 September 2021 09:34:13 Mauro Carvalho Chehab wrote:
> > The pcie-kirin driver contains both PHY and generic PCI driver
> > on it.
> >=20
> > The best would be, instead, to support a PCI PHY driver, making
> > the driver more generic.
> >=20
> > However, it is too late to remove the Kirin 960 PHY, as a change
> > like that would make the DT schema incompatible with past versions. =20
>=20
> I have not looked deeply at it. But is not it really possible to declare
> PHY node in DTS file with backward compatible manner? Or cannot Rob help
> with it (maybe there was similar issue in past with other driver)?

It would be possible to split the Kirin 960 PHY into drivers/phy. It is
also possible to hack the phy driver to search for the PHY-specific data
inside the PCI compatible string (I did wrote some patches doing that
and sent as a RFC several months ago), but the problem is that the
PHY driver won't be probed without adding a new compatible inside the
DT schema. By doing that, the schema will be incompatible.

In any case, the patches on this series split all PHY-specific code
inside the driver on a separate part of the source code. Moving it
to a new driver would be easy once someone comes with a solution
to add some new method at the PHY layer that would allow to load a
new module without having a compatible for it.

> I was fixing something similar, address space defined in DTS was used by
> two HW blocks: clock and UART. And I was able to make both DTS file and
> driver backward compatible.

Thanks,
Mauro