Re: [RFC PATCH V4 06/17] x86/hyperv: decrypt VMBus pages for sev-snp enlightened guest

[Tianyu Lan]

 On 4/12/2023 10:32 PM, Michael Kelley (LINUX) wrote:
> > @@ -191,6 +222,15 @@ void hv_synic_free(void)
> >   		free_page((unsigned long)hv_cpu->synic_event_page);
> >   		free_page((unsigned long)hv_cpu->synic_message_page);
> >   		free_page((unsigned long)hv_cpu->post_msg_page);
> > +
> > +		if (hv_isolation_type_en_snp()) {
> > +			set_memory_encrypted((unsigned long)
> > +				hv_cpu->synic_message_page, 1);
> > +			set_memory_encrypted((unsigned long)
> > +				hv_cpu->synic_event_page, 1);
> > +			set_memory_encrypted((unsigned long)
> > +				hv_cpu->post_msg_page, 1);
> > +		}
> The re-encryption must be done*before*  pages are freed!
>
> Furthermore, if the re-encryption fails, we should not free
> the page as it would pollute the free memory pool.  The best
> we can do is leak the memory.  See Patch 5 in Dexuan's
> TDX series, which does the same thing (but still doesn't
> get it quite right, per my comments).

You are right. The order is wrong. we should figure out a right solution 
to handle such case.