[PATCH v2 00/17] Divorcing irqdomain and device_node

On Tue, Oct 13, 2015 at 12:51:28PM +0100, Marc Zyngier wrote:

[This patch series used to be called "Making the generic ACPI GSI
layer irqdomain aware", but as I've radically changed my approach to
this problem, I've decided to reset the counters...]

The irqdomain code is not entierely ACPI friendly, as it has some
built-in knowledge of the device-tree. Nothing too harmful, but enough
to scare the ARM ACPI developpers which end up with their own version
of the square wheel. And some of the ramifications actually run deeper
than initially expected.

This series adapts the irqdomain code to use 'struct fwnode_handle'
instead of 'struct device_node' as the identifier for a domain
(compatibility interfaces are of course added). It also introduces a
generic IRQ specifier that firmware interfaces (DT or ACPI) can
directly use to configure interrupts, and allow the ACPI GSI code to
be plugged into this.

As examples, we convert the ARM GIC ACPI support to use irqdomains as
originally intended, and rework the MSI code to also move away from
using device nodes when using irqdomains.

Overall, this gives us a way to use irqdomains on both DT and ACPI
enabled platforms, having very little changes made to the actual
drivers (other than the probing infrastructure). Because we keep the
flow of information between the various layers identical between ACPI
and DT, we immediately benefit from the existing infrastructure.

This has been test-booted on Juno, is based on 4.3-rc5 + tip/irq/core
(in order to avoid potential conflicts), and available at:

git://git.kernel.org/pub/scm/linux/kernel/git/maz/arm-platforms.git irq/irq-domain-fwnode-v2

Tested on AMD Supercharger (ACPI PCI+MSI), by applying dependent series [0,1]
and some code soon to be posted to match gic v2m fwnode using this neat
patch series :-)

Tested-by: Lorenzo Pieralisi <redacted>

[0] https://lkml.org/lkml/2015/6/8/443
[1] http://comments.gmane.org/gmane.linux.kernel.pci/40904