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, 09 Sep 2020 13:31:01 -0700 (PDT)
Subject:     Re: [PATCH v2 2/3] soc: sifive: Add SiFive specific Cadence DDR controller driver
In-Reply-To: <20200909060045.GA13647@infradead.org>
CC:     Christoph Hellwig <hch@infradead.org>, dkangude@cadence.com,
        yash.shah@sifive.com, robh+dt@kernel.org,
        Paul Walmsley <paul.walmsley@sifive.com>, bp@alien8.de,
        mchehab@kernel.org, tony.luck@intel.com,
        devicetree@vger.kernel.org, aou@eecs.berkeley.edu,
        linux-kernel@vger.kernel.org, sachin.ghadi@sifive.com,
        rrichter@marvell.com, james.morse@arm.com,
        linux-riscv@lists.infradead.org, linux-edac@vger.kernel.org
From:   Palmer Dabbelt <palmer@dabbelt.com>
To:     Christoph Hellwig <hch@infradead.org>
Message-ID: <mhng-1b96da02-cbf5-42f4-bc30-a1c5c9cce5fa@palmerdabbelt-glaptop1>
Mime-Version: 1.0 (MHng)
Content-Type: text/plain; charset=utf-8; format=flowed
Content-Transfer-Encoding: 8bit
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk

On Tue, 08 Sep 2020 23:00:45 PDT (-0700), Christoph Hellwig wrote:
> On Tue, Sep 08, 2020 at 08:12:16PM -0700, Palmer Dabbelt wrote:
>> I don't know enough about the block to know if the subtle difference in
>> register names/offsets means.  They look properly jumbled up (ie, not just an
>> offset), so maybe there's just different versions or that's the SiFive-specific
>> part I had bouncing around my head?  Either way, it seems like one driver with
>> some simple configuration could handle both of these -- either sticking the
>> offsets in the DT (if they're going to be different everywhere) or by coming up
>> with some version sort of thing (if there's a handful of these).
>
> regmap can be used to handle non-uniform register layouts for the same
> functionality.

Ah, cool, I hadn't seen that before.  That seems like the way to go if this is
truly an implementatic-specific register mapping.  As I was falling asleep last
night I remembered that we did end up with implementation-specific register
maps for some of the IP we integrated.  That was usually the case for IP where
we had some signals that we just didn't know what to do with, and while I know
the DDR integration was a real trip I'm not sure if that's where these
registers came from.

Hopefully someone who has better access to these hardware implementations can
comment.