Re: [Xen-devel][PATCH 2/2] Enhance MTRR/PAT virtualization for EPT & VT-d enabled both

[Keir Fraser <keir.fraser@xxxxxxxxxxxxx>]

On 23/01/2009 12:03, "Keir Fraser" <keir.fraser@xxxxxxxxxxxxx> wrote:

> Would it not be quite easy for ept_set_entry() to determine whether any
> gfn->mfn mapping has changed, since it can see the old EPTE as it modifies
> it? In which case why not have that function determine whether it needs to
> modify VT-d tables (i.e, default to 0 at top of function, and flip to 1 as
> soon as you see a gfn->mfn translation change)?
> 
> Now, I could accept your current patch if I'm missing something, because I
> prefer this new interface to the way you did it before, but perhaps you can
> hide this new flag entirely inside ept_set_entry() and avoid any extra
> interface complexity at all?

Or indeed would it be simpler to have ept_change_entry_emt_with_range()
contain the bits of ept_set_entry() it needs, rather than calling it? I'm
not sure about that -- I suppose it depends on how much of ept_set_entry()
would disappear if you basically inlined it. It might not make sense to take
this approach.

 -- Keir



_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxxxxxxxx
http://lists.xensource.com/xen-devel