Re: [PATCH v3 1/3] Introduce a workqueue to deliver PIT timer interrupts.

[Gleb Natapov <gleb@redhat.com>]

On Tue, Apr 17, 2012 at 07:17:11PM +0300, Avi Kivity wrote:
> On 04/17/2012 07:15 PM, Jan Kiszka wrote:
> > On 2012-04-17 14:06, Avi Kivity wrote:
> > > On 04/17/2012 03:03 PM, Gleb Natapov wrote:
> > >>>
> > >>> KVM_MAX_VCPUS.
> > >>>
> > >> Ah, so you are worried about malicious guest configuring pit to
> > >> broadcast to all its vcpus.
> > > 
> > > Yes - it can introduce huge amounts of latency this way which is exactly
> > > what Jan is trying to prevent.
> > > 
> > > Though I'm not sure spin_lock_irq() in the realtime tree actually
> > > disables irqs (but it's certainly not a good idea in mainline; it's
> > > nasty even with just the spinlock).
> >
> > This depends on how you declare the spin lock type - raw or normal. The
> > former will disable irqs, the latter not even preemption (but become a
> > mutex).
> 
> Yes (and I see no reason to use raw spinlocks here).  Still, for
It was raw spinlock until f4f510508741680e423524c222f615276ca6222c.

> mainline, are we okay with 254*IPIs?  Maybe it's not so bad and I'm
> overinflating the problem.
> 
Isn't 254*IPIs can also happen if application changes memory mapping?

--
			Gleb.