Re: [PATCH 0/5] arm64, pci: Add ECAM/PCIe support for Cavium ThunderX

[Will Deacon <will.deacon@arm.com>]

On Wed, Jul 15, 2015 at 06:07:01PM +0100, Mark Rutland wrote:
> On Wed, Jul 15, 2015 at 05:54:40PM +0100, David Daney wrote:
> > From: David Daney <david.daney@cavium.com>
> > 
> > The subject pretty much says it all.  The first four patches tweak the
> > infrastructure a little so that we can get required behavior.  The
> > final patch adds the drivers.
> > 
> > David Daney (5):
> >   pci: Add is_pcierc element to struct pci_bus
> >   gic-its: Allow pci_requester_id to be overridden.
> >   arm64, pci: Allow RC drivers to supply pcibios_add_device()
> >     implementation.
> >   irqchip: gic-v3: Add gic_get_irq_domain() to get the irqdomain of the
> >     GIC.
> >   PCI: Add host drivers for Cavium ThunderX processors.
> > 
> >  arch/arm64/include/asm/pci.h        |   3 +
> >  arch/arm64/kernel/pci.c             |  10 +
> >  drivers/irqchip/irq-gic-v3-its.c    |  14 +-
> >  drivers/irqchip/irq-gic-v3.c        |   5 +
> >  drivers/pci/host/Kconfig            |  12 +
> >  drivers/pci/host/Makefile           |   2 +
> >  drivers/pci/host/pcie-thunder-pem.c | 462 ++++++++++++++++++++++++++++++++++++
> >  drivers/pci/host/pcie-thunder.c     | 422 ++++++++++++++++++++++++++++++++
> >  drivers/pci/probe.c                 |   2 +
> >  include/linux/irqchip/arm-gic-v3.h  |   3 +
> >  include/linux/pci.h                 |   1 +
> >  11 files changed, 935 insertions(+), 1 deletion(-)
> 
> I note that the driver attempts to probe with DT, yet there is no DT
> binding in this patch, nor has the devicetree mailing list been placed
> on Cc.
> 
> A DT binding document is mandatory for anything probing via DT.
> 
> Please write one, and submit it (as its own patch) with the next version
> of this series as per
> Documentation/devicetree/bindings/submitting-patches.txt.

Furthermore, the RequesterID -> DeviceID transformation should be described
in device-tree as an offset on each PCI host controller node, rather than
computed dynamically by picking apart the (software-assigned) PCI domain
ID.

For ACPI, IORT should be used.

Will