On 02.09.2011, at 20:14, Scott Wood wrote: > On 09/02/2011 10:12 AM, Alexander Graf wrote: >> >> Am 02.09.2011 um 01:08 schrieb Scott Wood <scottwood@xxxxxxxxxxxxx>: >> >>> Signed-off-by: Scott Wood <scottwood@xxxxxxxxxxxxx> >> >> Patch description missing. > > It's not missing, it's just brief. :-) > > I suppose you could add "The hardcoded behavior prevents SMP support. > QEMU shall specify the vcpu's PIR as the vcpu id". Ok, let me get my head around this. Before, PIR was forced to 0 by the setup code and set_sregs with PIR != 0 failed. Now it's simply vcpu_id which is already the correct value. Why didn't I run into this failure? Why did SMP work for me at all then? Shouldn't the guest be completely confused and find two CPU 0s? > >> Also, since pir == vcpu_id now, can't we just remove pir? > > From sregs? Is that worth the compatibility breakage? We could define > a new bit for ARCH206 without PIR, but older QEMU would then not see the > other ARCH206 stuff. Ugh. This is only in sregs. In KVM we already do use vcpu_id. Yes, all is fine there. I misread things as if we were having vcpu->pir :). Alex -- To unsubscribe from this list: send the line "unsubscribe kvm" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
