Re: [PATCH v5 15/22] x86/virt/tdx: Allocate and set up PAMTs for TDMRs

[Kai Huang]

On Mon, 2022-06-27 at 13:41 -0700, Dave Hansen wrote:
> On 6/27/22 03:31, Kai Huang wrote:
> > > > +/* Page sizes supported by TDX */
> > > > +enum tdx_page_sz {
> > > > +	TDX_PG_4K,
> > > > +	TDX_PG_2M,
> > > > +	TDX_PG_1G,
> > > > +	TDX_PG_MAX,
> > > > +};
> > > Are these the same constants as the magic numbers in Kirill's
> > > try_accept_one()?
> > try_accept_once() uses 'enum pg_level' PG_LEVEL_{4K,2M,1G} directly.  They can
> > be used directly too, but 'enum pg_level' has more than we need here:
> 
> I meant this:
> 
> +       switch (level) {
> +       case PG_LEVEL_4K:
> +               page_size = 0;
> +               break;
> 
> Because TDX_PG_4K==page_size==0, and for this:
> 
> +       case PG_LEVEL_2M:
> +               page_size = 1;
> 
> where TDX_PG_2M==page_size==1
> 
> See?
> 
> Are Kirill's magic 0/1/2 numbers the same as
> 
> 	TDX_PG_4K,
> 	TDX_PG_2M,
> 	TDX_PG_1G,
> 
> ?

Yes they are the same.  Kirill uses 0/1/2 as input of TDX_ACCEPT_PAGE TDCALL. 
Here I only need them to distinguish different page sizes.

Do you mean we should put TDX_PG_4K/2M/1G definition to asm/tdx.h, and
try_accept_one() should use them instead of magic 0/1/2?


-- 
Thanks,
-Kai