Re: [PATCH] x86/PCI: adjust section annotations

[Jesper Nilsson <jesper.nilsson@axis.com>]

On Thu, Jun 21, 2012 at 02:06:55AM +0200, Bjorn Helgaas wrote:
> On Mon, Jun 18, 2012 at 4:34 AM, Jan Beulich <JBeulich@suse.com> wrote:
> > DMI tables referenced from __init code only can be __initconst, and as
> > a result the functions referenced from there can become __init.
> >
> > pcibios_setup() can be __init as being a command line parsing function
> > only.
> >
> > A few other variables can then also have their attributes adjusted.
> 
> This seems OK as far as it goes.
> 
> However, if you're going to make pcibios_setup() __init for x86, I'd
> really encourage you to make it consistent across all the other
> architectures.  And if you do *that*, I think it would be cool if you
> supplied a generic do-nothing "weak" version in the PCI core.  That
> would allow you to remove it altogether from alpha, ia64, microblaze,
> mips pmc-sierra, parisc, powerpc, sh, sparc, tile, and xtensa.
> 
> CRIS-folk: It would also fix what looks like a bug in cris, which
> implements pcibios_setup() such that pci_setup() doesn't even look for
> all the supposedly generic options.

Ok, thanks for the heads-up!

I'm going to take a look at the PCI code in the CRIS-port.

My current feeling is that the PCI support for CRIS probably
should be dropped, both due to rotted code, but also that the
hardware is hard to come by. I need to do some more digging around...

/^JN - Jesper Nilsson
-- 
               Jesper Nilsson -- jesper.nilsson@axis.com