On Thu, Oct 07, 2010 at 02:58:26PM +0200, Avi Kivity wrote:
> On 10/04/2010 05:56 PM, Gleb Natapov wrote:
> >+
> >+ Physical address points to 32 bit memory location that will be written
> >+ to by the hypervisor at the time of asynchronous page fault injection to
> >+ indicate type of asynchronous page fault. Value of 1 means that the page
> >+ referred to by the page fault is not present. Value 2 means that the
> >+ page is now available.
>
> "The must not enable interrupts before the reason is read, or it may
> be overwritten by another apf".
>
> Document the fact that disabling interrupts disables APFs.
>
> How does the guest distinguish betweem APFs and ordinary page faults?
>
> What's the role of cr2?
>
> When disabling APF, all pending APFs are flushed and may or may not
> get a completion.
>
> Is a "page available" notification guaranteed to arrive on the same
> vcpu that took the "page not present" fault?
>
You mean documentation is lacking? :)