Re: [RFC PATCH] PCI: remove access to pci_[enable|disable]_msi() for drivers - take 2

[Roland Dreier <roland@topspin.com>]

    Greg> I've also seen some complaints that it is very arch specific
    Greg> (x86 based).  But as no other arches seem to want to support
    Greg> MSI, I don't really see any need to split it up.  Any
    Greg> comments about this?

    Grant> IA64 supports MSI/MSI-X. Roland, didn't PPC as well?

    Grant> I don't think it's *that* x86 specific.  The base address
    Grant> of the Processor interrupt Block (IIRC 0xfee0000) is
    Grant> implemented the same on several arches AFAIK. Even on some
    Grant> parisc chipsets.  I been constantly chasing other basic
    Grant> issues on parisc and just haven't had time to implement it
    Grant> in the past 3 years.

Some PPC does support MSI.  For example the PCI-X host bridge inside
the IBM (now AMCC) PPC 440GP supports MSI.  I don't know if any
non-embedded PPC or PPC64 supports MSI though -- for example do any
IBM POWER-based systems support MSI??

The current code isn't exactly x86-specific.  It's more Intel APIC
specific, which means that chipsets that for some reason or another
support the same address/message format will work.  So ia64 works as
well.  Following from that, HP parisc boxes that use the same chipset
as HP ia64 boxes have a chance at working.

But the PPC 440GP uses a different address and a different message
format.  So if anyone cared about supporting that, the current code
would have to be made more generic.

 - R.