Re: [Xen-devel] [PATCH 17/17] xen: Switch parameter in get_page_from_gfn to use typesafe gfn

[Julien Grall <julien@xxxxxxx>]

Hi Jan,

On 27/03/2020 13:50, Jan Beulich wrote:
> On 22.03.2020 17:14, julien@xxxxxxx wrote:
> > --- a/xen/arch/x86/hvm/domain.c
> > +++ b/xen/arch/x86/hvm/domain.c
> > @@ -296,8 +296,10 @@ int arch_set_info_hvm_guest(struct vcpu *v, const 
> > vcpu_hvm_context_t *ctx)
> >       if ( hvm_paging_enabled(v) && !paging_mode_hap(v->domain) )
> >       {
> >           /* Shadow-mode CR3 change. Check PDBR and update refcounts. */
> > -        struct page_info *page = get_page_from_gfn(v->domain,
> > -                                 v->arch.hvm.guest_cr[3] >> PAGE_SHIFT,
> > +        struct page_info *page;
> > +
> > +        page = get_page_from_gfn(v->domain,
> > +                                 gaddr_to_gfn(v->arch.hvm.guest_cr[3]),
> My earlier comment on this remains - I thing this conversion makes
> the problem this expression has more hidden than with the shift.
> This would better use a gfn_from_cr3() helper (or whatever it'll
> be that it gets named). Same elsewhere in this patch then.
I will have a closer look the *cr3 helpers and reply with some suggestions.

> > @@ -2363,7 +2364,7 @@ int hvm_set_cr3(unsigned long value, bool may_defer)
> >       {
> >           /* Shadow-mode CR3 change. Check PDBR and update refcounts. */
> >           HVM_DBG_LOG(DBG_LEVEL_VMMU, "CR3 value = %lx", value);
> > -        page = get_page_from_gfn(v->domain, value >> PAGE_SHIFT,
> > +        page = get_page_from_gfn(v->domain, cr3_to_gfn(value),
> Oh, seeing this I recall Paul did point out the above already.
>
> > @@ -508,7 +508,7 @@ int evtchn_fifo_init_control(struct evtchn_init_control 
> > *init_control)
> >   {
> >       struct domain *d = current->domain;
> >       uint32_t vcpu_id;
> > -    uint64_t gfn;
> > +    gfn_t gfn;
> >       uint32_t offset;
> >       struct vcpu *v;
> >       int rc;
> > @@ -516,7 +516,7 @@ int evtchn_fifo_init_control(struct evtchn_init_control 
> > *init_control)
> >       init_control->link_bits = EVTCHN_FIFO_LINK_BITS;
> >   
> >       vcpu_id = init_control->vcpu;
> > -    gfn     = init_control->control_gfn;
> > +    gfn     = _gfn(init_control->control_gfn);
> There's silent truncation here now for Arm32, afaict. Are we really
> okay with this?
Well, the truncation was already silently happening as we call 
get_page_from_gfn() in map_guest_page(). So it is not worse than the 
current situation.
Although, there are a slight advantage with the new code as you can more 
easily spot potential truncation. Indeed, you could add some type check 
in _gfn().
Cheers,

--
Julien Grall