[RFC PATCH] ARM64: PCI: inherit root controller's dma-coherent

[arnd@arndb.de (Arnd Bergmann)]

On Thursday 27 November 2014 11:36:36 Catalin Marinas wrote:
> On Thu, Nov 27, 2014 at 09:05:57AM +0000, Arnd Bergmann wrote:
> > On Thursday 27 November 2014 02:39:59 Jon Masters wrote:
> > > We're looking at this at the moment (in addition to having the ACPI specification
> > > clarified to indicate that there is no safe default assumption, requiring that
> > > DMA masters always indicate their coherency or otherwise via a _CCA whether true
> > > or false). 
> > 
> > I think for arm64, this should be easy: since ACPI is only for servers, we can
> > rely on the fact that the devices have cache-coherent DMA all the time. Let's
> > not over-complicate the ACPI case with all the special hacks we need for embedded.
> 
> As usual, we need hw vendors to say what they expect/build here.
> 
> I think in the past (non-ARM) the assumption was that the DMA is
> coherent on ACPI-capable systems unless otherwise stated but I've seen
> discussions (as Jon mentioned above) that ACPI should always indicate
> the coherency on ARM without any assumption.

Yes, that is fine, but I think Linux should just BUG_ON() any device
on ACPI that doesn't have this flag set. You definitely want to have
the flags work both ways for Windows Phone, but we don't need to support
that hardware with Linux.

The SBSA is quite specific about requiring coherent DMA to work,
so this is a safe assumption to make. We can support all other systems
with DT without problems.

> Either way, I don't think it's a problem for the kernel. We just need to
> change the default DMA ops to coherent when booting with ACPI (using
> non-coherent ops for a coherent device is not safe as the CPU can
> corrupt cache lines written by the device).

We should probably default to the coherent IOMMU dma ops with ACPI
eventually, I'd assume that these machines also come with an IOMMU,
though if we ever want to support virtual machines with ACPI as well,
that probably wants the linear operations, and until the SMMU support
for PCI is done, we have to start out using the linear operations
as well.

	Arnd