Re: [PATCH v2 5/8] hw/arm/virt: Fix devicetree warnings about the GPIO node

From: Peter Maydell <peter.maydell@linaro.org>
To: Rob Herring <robh@kernel.org>
Cc: Jean-Philippe Brucker <jean-philippe@linaro.org>,
	qemu-arm@nongnu.org, qemu-devel@nongnu.org
Subject: Re: [PATCH v2 5/8] hw/arm/virt: Fix devicetree warnings about the GPIO node
Date: Wed, 30 Nov 2022 12:27:23 +0000	[thread overview]
Message-ID: <CAFEAcA_t2hRPLcG-dyg4U5EzBX16FWWRzqMGy=ovnQpC1iB6Vg@mail.gmail.com> (raw)
In-Reply-To: <CAL_JsqK+NngZMheR7zR2oTVmgNFKKUhZtOAys=NmPQK_FdSq_A@mail.gmail.com>

On Tue, 29 Nov 2022 at 20:56, Rob Herring <robh@kernel.org> wrote:
>
> On Tue, Sep 27, 2022 at 6:25 AM Peter Maydell <peter.maydell@linaro.org> wrote:
> >
> > On Tue, 27 Sept 2022 at 11:12, Jean-Philippe Brucker
> > <jean-philippe@linaro.org> wrote:
> > >
> > > Since the pl061 device can be used as interrupt controller, its node
> > > should contain "interrupt-controller" and "#interrupt-cells" properties.
> >
> > It *can* be, but this PL061 is *not* an interrupt controller.
> > I don't see any reason why we should claim so in the DT.
>
> Taking another look, it is an interrupt controller. The GPIOs are
> connected to the 'gpio-keys' node which is interrupt based (there's a
> polled version too). That binding happens to be pretty lax and allows
> the GPIO to be specified either with 'gpios' or 'interrupts' property.
> The Linux PL061 driver happens to work only because it always
> registers an interrupt controller regardless of having
> "interrupt-controller" and "#interrupt-cells" properties or not.

No, it really isn't an interrupt controller. The interrupt controller
in this system is the GIC. The PL061 is a GPIO controller. It *has*
an interrupt line, which it connects to the GIC, but that doesn't make
it an interrupt controller any more than it makes a UART an interrupt
controller. It just means you can use the GPIOs in either a polled
or an interrupt-driven mode, same as you can use a PL011 UART in
either polled or interrupt-driven mode. And the guest knows it
can do that because we've told it "this is a PL061" and that's part
of the PL061's functionality.

The gpio-keys stuff is just "here is a wire which is tracking
the state of an emulated power button". This isn't an interrupt,
it's just a wire that has some status the guest probably
cares about.

The second PL061 which this bit of dtb-generation code also
applies to happens to currently be being used purely for
output GPIOs, so calling that one an interrupt controller
makes even less sense.

thanks
-- PMM