Re: [PATCH] udebs for armmp flavour
On Thu, 2013-09-05 at 03:59 +0100, Ben Hutchings wrote:
> On Sun, 2013-09-01 at 12:43 +0100, Ian Campbell wrote:
> > On Fri, 2013-08-30 at 21:47 +0100, Ian Campbell wrote:
> > > I suppose I should have CCd the ARM kernel maints! I'm not sure if any
> > > of you guys are actually especially interested in server class ARM
> > > though?
> > >
> > > Any objections to my pushing this?
> >
> > Keeping kernel-wedge happy (not complaining about missing modules)
> > requires making more (all in fact) NIC modules options, like the below.
> >
> > Does this make sense? I don't know what most of these are.
> >
> > I don't know if any of the platforms which are supported by armmp have
> > any of these or if any have e.g. a PCI slot which might make them an
> > option. If any do have a PCI slot then I'd expect them to also have an
> > onboard NIC, which is the minimum we would need to support for D-I
> > purposes.
> >
> > Is a blanket "all NICs are optional" a good rule?
>
> No, I don't think so. The common nic-modules list is appropriate for
> configurations where PCI is enabled. Configurations without PCI should
> have their own lists.
Sounds reasonable.
> But why does armmp *not* have CONFIG_PCI=y? The Marvell Armada 370 and
> XP have PCI Express 2.0 interfaces. You can bring them out to slots
> (use <http://cogcomp.com/csb_csb1726.htm> plus
> <http://cogcomp.com/csb_csb1701.htm>) so any standard PCIe card could be
> connected.
I had assumed that some armmp platform was definitely going to have PCI
slots, but I didn't know that we already supported some of them!
Unless they aren't supported by the kernel, but they are listed in e.g.
a dozen ish of arch/arm/boot/dts/armada-*.dts so I would guess they are.
I can't think of any reason not to enable CONFIG_PCI for armmp at this
stage. Arnaud?
Ian.
Reply to: