kernel.3045 and MSI/e1000?

[Tom London]

On 4/7/07, Dave Jones <davej at redhat.com> wrote:
> On Sat, Apr 07, 2007 at 11:07:30AM -0700, Tom London wrote:
>  > On 4/7/07, dragoran dragoran <drago01 at gmail.com> wrote:
>  > >
>  > > On 4/7/07, Tom London <selinux at gmail.com> wrote:
>  > > >
>  > > > * Wed Apr 04 2007 Dave Jones <davej at redhat.com >
>  > > > - Disable PCI MSI and MMCONFIG by default (cebbert)
>  > > >
>  > > > And I now get
>  > > > Apr  7 09:55:55  localhost kernel: assign_interrupt_mode Found MSI
>  > > capability
>  > > > Apr  7 09:55:56  localhost kernel: e1000: eth1: e1000_request_irq:
>  > > > Unable to allocate MSI interrupt Error: -22
>  > > > Apr  7 09:56:09  localhost kernel: e1000: eth1: e1000_request_irq:
>  > > > Unable to allocate MSI interrupt Error: -22
>  > >
>  > > try booting with pci=msi
>  > Yeah, this makes the messages 'go away'.  Thanks.
>  > Not sure I understand the apparent confusion here.....
>
> It's harmless, just noisy.
> In the 'Unable to allocate' case, it'll fall back to allocating non-MSI interrupts.
> For the majority of users, MSI will make absolutely no discernable difference.
> The patch referenced in the changelog inverts the usual logic the
> kernel uses (Enable MSI by default, and offer 'pci=nomsi') to
> disable by default, and offer with pci=msi for the minority cases where it
> may be necessary.
>
> This was done because the MSI code is still evolving at a rate where breakage occurs
> frequently, even when drivers aren't using it.
>
>         Dave
>
Thanks for the explanation. I'm presuming e1000 doesn't really 'need'
MSI then...

Any chance this is related to e1000 stability issues?

tom
-- 
Tom London