Re: [Qemu-devel] [RFC PATCH v2 07/49] kvmapic: fixing loading vmstate

[Paolo Bonzini]

Il 29/07/2014 14:03, Pavel Dovgaluk ha scritto:
>> > 
>> > This could be a subsection.  sipi_vector is only used (needed) if 
>> > wait_for_sipi != 0.
>   Right, sipi_vector is used when wait_for_sipi != 0. But we can set 
> sipi_vector to non-zero,
> save the snapshot, and then set wait_for_sipi. If that snapshot will be 
> loaded, sipi_vector become incorrect.
> Isn't this scenario possible?

sipi_vector will not be used until CPU_INTERRUPT_SIPI is set, and then
sipi_vector will have been overwritten with a new value.  The
architecture guarantees that.

Paolo