Re: Slow boot with 2.6.37-rc4

[Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>]

On Wed, Dec 01, 2010 at 07:15:29PM +0000, M A Young wrote:
> On Wed, 1 Dec 2010, Konrad Rzeszutek Wilk wrote:
> 
> >>[    1.443315] ata1: SATA max UDMA/133 irq_stat 0x00400040, connection status changed irq 499
> >											    ^^^
> >>[    1.443315] ata2: DUMMY
> >>[    1.443315] ata3: SATA max UDMA/133 abar m2048@0xfe9fb800 port 0xfe9fba00 irq 499
> >										   ^^^^-???
> >
> >That looks odd. Wonder what on baremetal it says?
> 
> [    1.558411] ata1: SATA max UDMA/133 irq_stat 0x00400040,
> connection status changed irq 40
> [    1.558414] ata2: DUMMY
> [    1.558417] ata3: SATA max UDMA/133 abar m2048@0xfe9fb800 port
> 0xfe9fba00 irq 40
> 
> This is with the same kernel but without xen. The full log is attached.

Ok, so it is using legacy IRQs and then it swaps over the the MSI ones.
Under Xen it ends up using IRQ numbers from the back, so that is Ok..
But I wonder if it is just not getting them.

Can you pass in pci=nomsi on the Linux kernel line just to see if it is
indeed MSI (or vector) related.

And also under the Xen hypervisor serial line print out the PCI MSI irq and IRQ vectors
values? Don't remember the debug keys - but the point is to see if for your AHCI controller:
0000:00:1f.2 and 0000:00:1f.1, there is an MSI vector allocated for it.