Re: [PATCH 0/5]: Fix kdump under KVM

[Avi Kivity <avi@redhat.com>]

On 10/27/2009 06:41 PM, Chris Lalancette wrote:
> This patch series aims to get kdump working inside a KVM guest.
> The current problem with using kdump is that KVM always delivers
> PIT interrupts to the BSP, and the BSP only.  While this is
> technically allowed by the MPS spec, most motherboards actually
> deliver timer interrupts to *any* LAPIC in virtual wire mode.
> Since a crash can occur on any CPU, timer interrupts must
> be able to reach any CPU in order for kdump to work properly.
>
> Therefore, this patch series kicks all of the relevant vCPUs
> when delivering a timer interrupt.  With these patches in
> place, kdump in a RHEL-5 guest works properly.
>    

This is pretty expensive on large guests.  However I suppose under 
normal conditions we won't be in virtual wire mode?

The kick from i8254 code is pretty bad, as you mention.  I forget why it 
is needed at all - shouldn't kvm_set_irq() end up kicking the correct 
vcpu (and note the pic still insists on the bsp:)

/*
  * callback when PIC0 irq status changed
  */
static void pic_irq_request(void *opaque, int level)
{
     struct kvm *kvm = opaque;
     struct kvm_vcpu *vcpu = kvm->bsp_vcpu;
     struct kvm_pic *s = pic_irqchip(kvm);
     int irq = pic_get_irq(&s->pics[0]);

     s->output = level;
     if (vcpu && level && (s->pics[0].isr_ack & (1 << irq))) {
         s->pics[0].isr_ack &= ~(1 << irq);
         kvm_vcpu_kick(vcpu);
     }
}

-- 
error compiling committee.c: too many arguments to function