Re: [RFC 9/9] of/irq: create interrupts-extended property

[Grant Likely]

On Wed, 13 Nov 2013 09:17:01 +1000, Peter Crosthwaite <peter.crosthwaite@xxxxxxxxxx> wrote:
> On Tue, Nov 12, 2013 at 6:50 PM, Grant Likely <grant.likely@xxxxxxxxxx> wrote:
> > On Tue, 12 Nov 2013 17:49:37 +1000, Peter Crosthwaite <peter.crosthwaite@xxxxxxxxxx> wrote:
> >> Hi Grant,
> >>
> >> On Tue, Nov 12, 2013 at 4:54 PM, Grant Likely <grant.likely@xxxxxxxxxx> wrote:
> >> > On Tue, 12 Nov 2013 08:58:15 +1000, Peter Crosthwaite <peter.crosthwaite@xxxxxxxxxx> wrote:
> >> >> Hi Grant,
> >> >>
> >> >> On Wed, Oct 16, 2013 at 6:39 AM, Grant Likely <grant.likely@xxxxxxxxxx> wrote:
> >> >> > The standard interrupts property in device tree can only handle
> >> >> > interrupts coming from a single interrupt parent. If a device is wired
> >> >> > to multiple interrupt controllers, then it needs to be attached to a
> >> >> > node with an interrupt-map property to demux the interrupt specifiers
> >> >> > which is confusing. It would be a lot easier if there was a form of the
> >> >> > interrupts property that allows for a separate interrupt phandle for
> >> >> > each interrupt specifier.
> >> >> >
> >> >> > This patch does exactly that by creating a new interrupts-extended
> >> >> > property which reuses the phandle+arguments pattern used by GPIOs and
> >> >> > other core bindings.
> >> >> >
> >> >> > Signed-off-by: Grant Likely <grant.likely@xxxxxxxxxx>
> >> >> > Cc: Rob Herring <rob.herring@xxxxxxxxxxx>
> >> >> > ---
> >> >> >  .../bindings/interrupt-controller/interrupts.txt   | 29 +++++++--
> >> >> >  arch/arm/boot/dts/testcases/tests-interrupts.dtsi  | 16 +++++
> >> >> >  arch/arm/boot/dts/versatile-ab.dts                 |  2 +-
> >> >> >  arch/arm/boot/dts/versatile-pb.dts                 |  2 +-
> >> >> >  drivers/of/irq.c                                   | 16 +++--
> >> >> >  drivers/of/selftest.c                              | 70 ++++++++++++++++++++++
> >> >> >  6 files changed, 122 insertions(+), 13 deletions(-)
> >> >> >
> >> >> > diff --git a/Documentation/devicetree/bindings/interrupt-controller/interrupts.txt b/Documentation/devicetree/bindings/interrupt-controller/interrupts.txt
> >> >> > index 72a06c0..1486497 100644
> >> >> > --- a/Documentation/devicetree/bindings/interrupt-controller/interrupts.txt
> >> >> > +++ b/Documentation/devicetree/bindings/interrupt-controller/interrupts.txt
> >> >> > @@ -4,16 +4,33 @@ Specifying interrupt information for devices
> >> >> >  1) Interrupt client nodes
> >> >> >  -------------------------
> >> >> >
> >> >> > -Nodes that describe devices which generate interrupts must contain an
> >> >> > -"interrupts" property. This property must contain a list of interrupt
> >> >> > -specifiers, one per output interrupt. The format of the interrupt specifier is
> >> >> > -determined by the interrupt controller to which the interrupts are routed; see
> >> >> > -section 2 below for details.
> >> >> > +Nodes that describe devices which generate interrupts must contain an either an
> >> >> > +"interrupts" property or an "interrupts-extended" property. These properties
> >> >> > +contain a list of interrupt specifiers, one per output interrupt. The format of
> >> >> > +the interrupt specifier is determined by the interrupt controller to which the
> >> >> > +interrupts are routed; see section 2 below for details.
> >> >> > +
> >> >> > +  Example:
> >> >> > +       interrupt-parent = <&intc1>;
> >> >> > +       interrupts = <5 0>, <6 0>;
> >> >> >
> >> >> >  The "interrupt-parent" property is used to specify the controller to which
> >> >> >  interrupts are routed and contains a single phandle referring to the interrupt
> >> >> >  controller node. This property is inherited, so it may be specified in an
> >> >> > -interrupt client node or in any of its parent nodes.
> >> >> > +interrupt client node or in any of its parent nodes. Interrupts listed in the
> >> >> > +"interrupts" property are always in reference to the node's interrupt parent.
> >> >> > +
> >> >> > +The "interrupts-extended" property is a special form for use when a node needs
> >> >> > +to reference multiple interrupt parents. Each entry in this property contains
> >> >> > +both the parent phandle and the interrupt specifier. "interrupts-extended"
> >> >> > +should only be used when a device has multiple interrupt parents.
> >> >> > +
> >> >> > +  Example:
> >> >> > +       interrupts-extended = <&intc1 5 1>, <&intc2 1 0>;
> >> >> > +
> >> >>
> >> >> A slightly different but related problem I am trying to solve with
> >> >> interrupt bindings, is to describe connection of a single device
> >> >> interrupt line to multiple interrupts controller, whereas this binding
> >> >> seems to be describing connection of different interrupt lines to
> >> >> different controllers. Can this syntax accommodate my case is any way?
> >> >>
> >> >> Bascially I think I'm asking for multiple interrupt specifiers for a
> >> >> single IRQ output if that's possible.
> >> >>
> >> >> Should I just be using the interrupt map syntax instead?
> >> >>
> >> >> interrupt-map-mask = < 0 0 63>;
> >> >> interrupt-map = < 0 0 25 &intc0 ... >, < 0 0 25 &intc1 ... >,
> >> >
> >> > Interrupt map doesn't actually help you here either since the core code
> >> > doesn't know what to do with it. It will probably only match and return
> >> > one of the options.
> >> >
> >>
> >> Yes I noticed that. Although we are a fair way off Linux patches for
> >> this just yet and all I want to solve near-term is the hardware
> >> description problem (what core Linux IRQ code is actually supposed to
> >> do when given such a wiring setup is probably an open question).
> >>
> >> > What I would do is describe both in your interrupts property and make
> >> > the driver know that the extra interrupt specifier is for the same
> >> > interrupt output.
> >> >
> >>
> >> So this is a little annoying, as ideally device drivers should not be
> >> aware of their external connections. I'm trying to describe board (or
> >> SoC) level wiring in hopefully a way that doesn't require individual
> >> driver awareness. It also gets very messy when you have multiple
> >> interrupt outputs that are each in themselves connected to multiple
> >> interrupt controllers (which happens to be my situation). The meaning
> >> of multiple tuples becomes ambiguous.
> >>
> >> Is the interrupt-map system I proposed acceptable long-term from just
> >> a pure binding acceptability standpoint? The driver side solution is
> >> also a large change pattern as I'm using this with Xilinx Zynq which
> >> has a large number of peripherals, any of which could be
> >> multiply-connected to both A9GIC and soft intc's in FPGA fabric.
> >
> > Well, it isn't actively dangerous, but it really isn't a very good
> > description either. There is no information about what the difference
> > would be between the different connection options. The kernel certainly
> > cannot do anything intelligent with it.
> >
> > You'd probably be better off with an IRQ mux node and associated driver
> > that chooses the appropriate connection and makes the behaviour
> > transparent to the device driver. That way you'd have a place to embed
> > the knowledge of how to make good decisions.
> >
> 
> Hi Grant,
> 
> That sounds good, and it avoids me having to deal with that
> interrupt-map crazyness. Here's a first attempt at this form of
> syntax:
> 
> foo_irq_mux {
>     #interrupt-cells = 0;
>     comaptible = "irq-mux";
>     interrupts-extended = < &intc0 0 1 >,< &intc1 2 3 >;
>     /* descision making info goes here */
> };
> 
> foo : foo@deadbeef {
>     compatible = "vendor,foo";
>     interrupts-extended = <&foo_irq_mux>;
> }
> 
> 
> It's going to get a little verbose once you start making multiple
> connections as you need one mux per wire. Perhaps it could be cleaned
> up by making the foo_irq_mux node(s) a child of foo?

It could, but then you need some way of attaching a driver to that node,
and that would require building knowledge into the driver again.

Can you boil it down to a couple of concrete examples? What is a
specific example of how the platform should decide which interrupt line
to use?

g.

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/