Re: [RFC PATCH v1 1/4] arm/pci: PCI setup and PCI host bridge discovery within XEN on ARM.

[Julien Grall <julien@xxxxxxx>]

Hi,

On 24/07/2020 00:38, Stefano Stabellini wrote:
> > +    bridge->dt_node = dev;
> > +    bridge->sysdata = cfg;
> > +    bridge->ops = &ops->pci_ops;
> > +
> > +    if( !dt_property_read_u32(dev, "linux,pci-domain", &segment) )
> > +    {
> > +        printk(XENLOG_ERR "\"linux,pci-domain\" property in not available in 
> > DT\n");
> > +        return -ENODEV;
> > +    }
> > +
> > +    bridge->segment = (u16)segment;
> My understanding is that a Linux pci-domain doesn't correspond exactly
> to a PCI segment. See for instance:
>
> https://lists.gnu.org/archive/html/qemu-devel/2018-04/msg03885.html
>
> Do we need to care about the difference? If we mean pci-domain here,
> should we just call them as such instead of calling them "segments" in
> Xen (if they are not segments)?
So we definitely need a segment number in hand because this is what the 
admin will use to assign a PCI device to a guest.
The segment number is just a value defined by the software. So as long 
as Linux and Xen agrees with the number, then we should be ok.
Looking at the code, I think Linux is using 'segment' as 'domain'. So 
they should be interchangeably. I am not sure what other OS does though.
Cheers,

--
Julien Grall