Received-SPF: pass (google.com: best guess record for domain of linux-kernel-owner@vger.kernel.org designates 209.132.180.67 as permitted sender) client-ip=209.132.180.67;
MIME-Version: 1.0
References: <20181204181550.29122-1-brgl@bgdev.pl> <20181204181550.29122-2-brgl@bgdev.pl>
 <20181205153534.GF6205@sirena.org.uk>
In-Reply-To: <20181205153534.GF6205@sirena.org.uk>
From:   Bartosz Golaszewski <bgolaszewski@baylibre.com>
Date:   Wed, 5 Dec 2018 21:44:37 +0100
Message-ID: <CAMpxmJUoC0sU4K-Qo6QGnNUMVYXKZOuyvuGmZiv6aUWk0x+MmA@mail.gmail.com>
Subject: Re: [PATCH 1/1] regmap: irq: handle HW using separate mask bits for edges
To:     Mark Brown <broonie@kernel.org>
Cc:     Bartosz Golaszewski <brgl@bgdev.pl>,
        Greg KH <gregkh@linuxfoundation.org>, rafael@kernel.org,
        LKML <linux-kernel@vger.kernel.org>
Content-Type: text/plain; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk

=C5=9Br., 5 gru 2018 o 16:35 Mark Brown <broonie@kernel.org> napisa=C5=82(a=
):
>
> On Tue, Dec 04, 2018 at 07:15:50PM +0100, Bartosz Golaszewski wrote:
>
> > Let's reuse the existing type fields in struct regmap_irq to make
> > regmap_irq_chip available to such HW.
>
> I'm not sure this is ideal, it makes the interface less clear for users
> especially since there's nothing in the comments in the header that
> users will look at which mentions the feature.
>
> > If the type_base and mask_base offsets are the same - assume there
> > are separate bits for falling and rising edge interrupts and use
> > the value previously written to the type buffer by the set_type()
> > callback instead of the entire mask specified for this interrupt
> > so that we only enable the requested edge interrupts.
>
> This feels like it's very strongly tied to a specific implementation of
> the feature and TBH I'm somewhat unclear on what this ends up concretely
> meaning.  It sounds like this hardware represents the two edges as
> separate interrupts but you want to combine them into one but I can't
> see exactly how the interrupt number gets mapped with your change.

Yes, it's two edges represented as separate interrupts. I will post a
patch with a different (hopefully clearer) approach together with an
example snippet from the code actually using it.

Bart