Re: [PATCH 09/24] PCI, powerpc: Register busn_res for root buses

On Wed, Feb 8, 2012 at 2:02 PM, Benjamin Herrenschmidt
[off-list ref] wrote:

On Wed, 2012-02-08 at 07:58 -0800, Bjorn Helgaas wrote:

quoted

The only architecture-specific thing here is discovering the range of
bus numbers below a host bridge. =A0The architecture should not have to
mess around with pci_bus_update_busn_res_end() like this. =A0It should
be able to say "here's my bus number range" (and of course the PCI
core can default to 0-255 if the arch doesn't supply a range) and the
core should take care of the rest.

So it's a bit messy in here because we deal with several things.

What the firmware gives us is the range it assigned, but that isn't
necessarily the HW limits (almost never is in fact).

In some cases we honor it, for example when in "probe only" mode where
we prevent any reassigning, and in some case, we ignore it and let the
PCI core renumber things (typically because the FW "forgot" to set aside
bus numbers for a cardbus slot for example, that sort of things).

So it's a bit of a tricky situation.

Off the top of my head, I'm pretty sure that most if not all of our PCI
host bridges simply support a full 0...255 range and there is no sharing
between bridges like on x86, they are just different domains.

My point is that the interface between the arch and the PCI core
should be simply the arch telling the core "this is the range of bus
numbers you can use."  If the firmware doesn't give you the HW limits,
that's the arch's problem.  If you want to assume 0..255 are
available, again, that's the arch's decision.

But the answer to the question "what bus numbers are available to me"
depends only on the host bridge HW configuration.  It does not depend
on what pci_scan_child_bus() found.  Therefore, I think we can come up
with a design where pci_bus_update_busn_res_end() is unnecessary.

Bjorn