Re: [git pull, take 2] x86 updates for v2.6.28, phase #9, sparse IRQ support

[Ingo Molnar <mingo@elte.hu>]

* Yinghai Lu <yinghai@kernel.org> wrote:

> Ingo Molnar wrote:
> > * Yinghai Lu <yinghai@kernel.org> wrote:
> > 
> >> Ingo Molnar wrote:
> >>> * Ingo Molnar <mingo@elte.hu> wrote:
> >>>
> >>>> Linus,
> >>>>
> >>>> Please pull the latest x86-v28-for-linus-phase9 git tree from:
> >>>>
> >>>> Sparse IRQ support - make NR_IRQS dynamic by turning it into nr_irqs, 
> >>>> an by creating irq descriptors in a sparse way.
> >>>>
> >>>> Includes a number of dependent x86 topics: timers/hpet-percpu, IO-APIC 
> >>>> and local APIC code unification, UV support updates.
> >>>>
> >>>> Has been cross-built on all architectures we could get crosscompilers 
> >>>> for:
> >>>>
> >>>>   http://www.tglx.de/autoqa/index
> >>>>
> >>>> It should be largely uneventful for sparse-irq unaware, non-x86 
> >>>> architectures.
> >>> please find updated pull request below. Relative to phase-N-1. Updates 
> >>> from today in this class of topics:
> >>>
> >>>   b570868: x86: apic - unify APIC_DIVISOR
> >>>
> >>> And i removed:
> >>>
> >>>   eada62c: acpi/x86: introduce __apci_map_table, v4
> >>>
> >>> [ Cc:-ed Yinghai for that - we can live without that patch (i.e. we at
> >>>   most get a leak warning) and can have that via acpi.git, correct? ]
> >> yes. we have several patches that even didn't pick up for a while.
> >>
> >> also please ask Linus to pull pci tree from Jesse before sparseirq.
> >>
> >> because we need
> >> 	PCI: change MSI-x vector to 32bit
> >>
> >> in pci tree.
> > 
> > is it absolutely needed?
> > 
> 
> qla2xxx with MSI-X will crash. only happend with the new card that support msi-x.
> 
> so it is not absolutely needed.

but as Jesse noted we have:

 9e40236: pci: change msi-x vector to 32bit

included in irq/sparseirq so it should be fine and there's no dependency 
on the PCI tree.

	Ingo