Re: [PATCH 2/10] x86: convert to generic helpers for IPI function calls

[Jens Axboe <jens.axboe@oracle.com>]

On Thu, May 29 2008, Jeremy Fitzhardinge wrote:
> Jens Axboe wrote:
> >@@ -122,6 +109,17 @@ static int xen_smp_intr_init(unsigned int cpu)
> > 		goto fail;
> > 	per_cpu(debug_irq, cpu) = rc;
> > 
> >+	callfunc_name = kasprintf(GFP_KERNEL, "callfuncsingle%d", 
> >cpu);
> >+	rc = bind_ipi_to_irqhandler(XEN_CALL_FUNCTION_SINGLE_VECTOR,
> >+				    cpu,
> >+				    
> >xen_call_function_single_interrupt,
> >+				    
> >IRQF_DISABLED|IRQF_PERCPU|IRQF_NOBALANCING,
> >+				    callfunc_name,
> >+				    NULL);
> >+	if (rc < 0)
> >+		goto fail;
> >+	per_cpu(callfuncsingle_irq, cpu) = rc;
> >  
> 
> Tiny nit: could you shift this up below the other callfunc 
> registration, so they appear next to each other in /proc?

Certainly, will do.

> >+void xen_smp_send_call_function_ipi(cpumask_t mask)
> >+{
> >+	int cpu;
> >+
> >+	xen_send_IPI_mask(mask, XEN_CALL_FUNCTION_VECTOR);
> >+
> >+	/* Make sure other vcpus get a chance to run if they need to. 
> >*/
> >+	for_each_cpu_mask(cpu, mask) {
> >+		if (xen_vcpu_stolen(cpu)) {
> >+			HYPERVISOR_sched_op(SCHEDOP_yield, 0);
> >+			break;
> >+		}
> >+	}
> >  
> 
> I think you should just drop this loop for now; I want to do some 
> measurements before putting it in.

If you don't mind, I'd like to keep it as-is. Then the patch should
not have a functional change (there, at least), which I think is
important. I can add a later patch removing this stolen bit, it would
be even better if you send me such a patch :-)

-- 
Jens Axboe