Re: [PATCH v10 07/10] OF: Introduce helper function for getting PCI domain_nr

[Arnd Bergmann]

On Thursday 11 September 2014 14:11:05 Phil Edworthy wrote:
> On 10 September 2014 19:20, Arnd wrote:
> > On Tuesday 09 September 2014 12:20:54 Catalin Marinas wrote:
> > >
> > > We can assume that if a domain is not specified and there is a single
> > > top level PCIe node, the domain defaults to 0. Are there any arm32
> > > platforms that require multiple domains (and do not specify a number in
> > > the DT)?
> > 
> > In theory, I think all of them could work with a single domain, but then
> > you need to partition the bus number space between the host controllers,
> > so you have the exact same situation that you either need to make up
> > random bus numbers or put them in DT.
> > 
> > Using multiple domains is way cleaner for this, even if we have to
> > make up the numbers.
> 
> Maybe this is a stupid question, but why would you want to specify the domain
> in the DT at all? Doesn't every instance of a driver imply a separate domain?

See Jason Gunthorpe's latest reply on the topic. In short, the domain is
visible to user space and we want it to be stable across boots when
user configuration refers to devices by domain:bus:device:function
numbers.

	Arnd
--
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/