Re: [PATCH] arm64: iort: take _DMA methods into account for named components

From: Lorenzo Pieralisi <lorenzo.pieralisi@arm.com>
To: Ard Biesheuvel <ardb@kernel.org>
Cc: Catalin Marinas <catalin.marinas@arm.com>,
	Will Deacon <will@kernel.org>,
	robin.murphy@arm.com,
	Linux ARM <linux-arm-kernel@lists.infradead.org>
Subject: Re: [PATCH] arm64: iort: take _DMA methods into account for named components
Date: Mon, 6 Apr 2020 14:14:32 +0100	[thread overview]
Message-ID: <20200406131432.GC4650@red-moon.cambridge.arm.com> (raw)
In-Reply-To: <CAMj1kXHt0yhnFE=fVZm0p3iyC20s=j=8Z4Gzcmp_cqx3P3N5Dw@mail.gmail.com>

On Mon, Apr 06, 2020 at 01:59:07PM +0200, Ard Biesheuvel wrote:
> On Mon, 6 Apr 2020 at 13:32, Lorenzo Pieralisi
> <lorenzo.pieralisi@arm.com> wrote:
> >
> > On Mon, Apr 06, 2020 at 01:16:15PM +0200, Ard Biesheuvel wrote:
> > > On Mon, 6 Apr 2020 at 13:04, Lorenzo Pieralisi
> > > <lorenzo.pieralisi@arm.com> wrote:
> > > >
> > > > On Sat, Apr 04, 2020 at 09:30:47AM +0200, Ard Biesheuvel wrote:
> > > > > Where IORT nodes for named components can describe simple DMA limits
> > > > > expressed as the number of address bits a device can driver, _DMA methods
> > > > > in AML can express more complex topologies, involving DMA translation in
> > > > > particular.
> > > > >
> > > > > Currently, we only take this _DMA method into account if it appears on a
> > > > > ACPI device node describing a PCIe root complex, but it is perfectly
> > > > > acceptable to attach them to named components as well, so let's ensure
> > > > > we take them into account in those cases too.
> > > >
> > > > ACPI spec v6.3, 6.2.4 _DMA:
> > > >
> > > > "_DMA is only defined under devices that represent buses"
> > > >
> > >
> > > Sure. But ACPI0004 module devices are also bus nodes, so that
> > > statement does not exclude named components that are defined under
> > > such a module device.
> >
> > Yes. _DMA is valid only if a _CRS is present, ACPI0004 does not seem
> > to _require_ a _CRS to be considered valid.
> >
> 
> How is that relevant? Any node that has a _DMA must have a _CRS as
> well. Some nodes that don't have a _DMA may not have a _CRS either.
> But that does not disqualify a ACPI0004 that *does* have both from
> being considered a bus node, no? Or is that not what you are saying?

I am just trying to prevent firmware developers from adding ACPI0004
nodes with *just* a _DMA object (because the _CRS is optional) which
will become unusable in OS context (where we do check the _CRS
presence), that's all I am saying.

Just trying to make specs foolproof :)

Thanks,
Lorenzo

_______________________________________________
linux-arm-kernel mailing list
linux-arm-kernel@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel