From mboxrd@z Thu Jan  1 00:00:00 1970
From: boris.brezillon@free-electrons.com (Boris Brezillon)
Date: Tue, 16 Dec 2014 15:20:41 +0100
Subject: [PATCH 0/5] ARM: at91: fix irq_pm_install_action WARNING
In-Reply-To: <alpine.DEB.2.11.1412161349420.17382@nanos>
References: <1418660152-17185-1-git-send-email-boris.brezillon@free-electrons.com>
 <2662748.NJbkr6jW7r@vostro.rjw.lan>
 <3542278.8lcJeXLMFm@vostro.rjw.lan>
 <20141216100715.57baeb65@bbrezillon>
 <alpine.DEB.2.11.1412161013570.17382@nanos>
 <20141216122503.77c5632f@bbrezillon>
 <alpine.DEB.2.11.1412161349420.17382@nanos>
Message-ID: <20141216152041.56ff9b2e@bbrezillon>
To: linux-arm-kernel@lists.infradead.org
List-Id: linux-arm-kernel.lists.infradead.org

On Tue, 16 Dec 2014 13:56:26 +0100 (CET)
Thomas Gleixner <tglx@linutronix.de> wrote:

> On Tue, 16 Dec 2014, Boris Brezillon wrote:
> > Thomas Gleixner <tglx@linutronix.de> wrote:
> > > Install the timer interrupt as a demultiplexing interrupt.
> > 
> > I can try to hack the AIC irqchip driver to implement this demux logic,
> > but this logic can't be placed in the PIT (Periodic Interval Timer)
> > driver itself, because the shared IRQ line is used by the at91 clock
> > controller (PMC) which is providing the clock to the PIT device.
> > This gives the following dependency graph:
> > 
> > PIT =depends-on=> Master Clock =provided-by=> PMC =needs=> PMC IRQ.
> 
> It never can be a part of a device driver. This is a property of the
> interrupt controller which fails to provide a proper demux for this
> irq line in the first place.
> 
> So you fake the demux in the irq chip driver, where you provide the
> pseudo chip and the demux logic. None of the device drivers (PIT, PMC,
> UART ...) even knows about that.
> 
> |--------|             |--------|
> | AIC    |-------------| Pseudo |---- PIT
> |        |             | chip   |---- PMC
> |        |             |        |---- UART
> |        |             |--------|
> |        |--- devX
> |        |--- devY
> |        |--- devZ
> |--------|
> 
> So PIT, PMC and UART get interrupt numbers handed out which are
> outside of the interrupt space of AIC.

Okay, that's what I had in mind, thanks for clarifying this aspect.

Regarding the DT representation, it should look like this, right ?

aic: interrupt-controller at fffff000 {
	#interrupt-cells = <3>;
	compatible = "atmel,at91rm9200-aic";
	interrupt-controller;
	reg = <0xfffff000 0x200>;
	atmel,external-irqs = <29 30 31>;

	irq1_demux: irq1-demux {
		#interrupt-cells = <1>;
		interrupt-controller;
	}
};

pit: timer at fffffd30 {
	compatible = "atmel,at91sam9260-pit";
	reg = <0xfffffd30 0xf>;
	/* 0 is the id reserved for the PIT */
	interrupts-extended = <&irq1_demux 0>;
	clocks = <&mck>;
};

pmc: pmc at fffffc00 {
	compatible = "atmel,at91sam9260-pmc";
	reg = <0xfffffc00 0x100>;
	/* 1 is the id reserved for the PMC */
	interrupts-extended = <&irq1_demux 1>;
[...]
};

Regards,

Boris

-- 
Boris Brezillon, Free Electrons
Embedded Linux and Kernel engineering
http://free-electrons.com