Re: [Xen-devel] [PATCH for-4.13] xen/arm: Don't use _end in is_xen_fixed_mfn()

[Stefano Stabellini <sstabellini@xxxxxxxxxx>]

On Tue, 15 Oct 2019, Julien Grall wrote:
> Hi,
> 
> On 15/10/2019 20:28, Stefano Stabellini wrote:
> > On Tue, 15 Oct 2019, Julien Grall wrote:
> >> virt_to_maddr() is using the hardware page-table walk instructions to
> >> translate a virtual address to physical address. The function should
> >> only be called on virtual address mapped.
> >>
> >> _end points past the end of Xen binary and may not be mapped when the
> >> binary size is page-aligned. This means virt_to_maddr() will not be able
> >> to do the translation and therefore crash Xen.
> >>
> >> Note there is also an off-by-one issue in this code, but the panic will
> >> trump that.
> >>
> >> Both issues can be fixed by using _end - 1 in the check.
> >>
> >> Signed-off-by: Julien Grall <julien.grall@xxxxxxx>
> >>
> >> ---
> >>
> >> Cc: Andrew Cooper <andrew.cooper3@xxxxxxxxxx>
> >> Cc: George Dunlap <George.Dunlap@xxxxxxxxxxxxx>
> >> Cc: Ian Jackson <ian.jackson@xxxxxxxxxxxxx>
> >> Cc: Jan Beulich <jbeulich@xxxxxxxx>
> >> Cc: Julien Grall <julien@xxxxxxx>
> >> Cc: Konrad Rzeszutek Wilk <konrad.wilk@xxxxxxxxxx>
> >> Cc: Stefano Stabellini <sstabellini@xxxxxxxxxx>
> >> Cc: Tim Deegan <tim@xxxxxxx>
> >> Cc: Wei Liu <wl@xxxxxxx>
> >> Cc: jgross@xxxxxxxx
> >>
> >> x86 seems to be affected by the off-by-one issue. Jan, Andrew?
> >>
> >> This could be reached by a domain via XEN_SYSCTL_page_offline_op.
> >> However, the operation is not security supported (see XSA-77). So we are
> >> fine here.
> >> ---
> >>   xen/include/asm-arm/mm.h | 2 +-
> >>   1 file changed, 1 insertion(+), 1 deletion(-)
> >>
> >> diff --git a/xen/include/asm-arm/mm.h b/xen/include/asm-arm/mm.h
> >> index 262d92f18d..174acd8859 100644
> >> --- a/xen/include/asm-arm/mm.h
> >> +++ b/xen/include/asm-arm/mm.h
> >> @@ -153,7 +153,7 @@ extern unsigned long xenheap_base_pdx;
> >>   
> >>   #define is_xen_fixed_mfn(mfn)                                   \
> >>       ((mfn_to_maddr(mfn) >= virt_to_maddr(&_start)) &&           \
> >> -     (mfn_to_maddr(mfn) <= virt_to_maddr(&_end)))
> >> +     (mfn_to_maddr(mfn) <= virt_to_maddr(_end - 1)))
> > 
> > Thank you for sending the patch and I think that "_end - 1" is the right
> > fix. I am just wondering whether we want/need an explicit cast of some
> > sort here, because technically _end is a char[] and 1 is a integer.
> > Maybe:
> > 
> >    ((vaddr_t)_end - 1)
> > 
> > ?
> 
> I vaguely remember a lengthy discussion about it last year. But I don't 
> think there was any conclusion in it.
>
> In this case, what are you trying to prevent with the cast? Is it 
> underflow of an array? If so, how the cast is actually going to prevent 
> the compiler to do the wrong thing?

Yes, there was a long discussion at the beginning of the year; it was
about how to define _start and _end so that we could avoid compilers
undefined behavior. The main underlying issue is that comparisons
between pointers to different objects are undefined [1] (_start and _end
can be interpreted as pointers to different objects).

This case is a bit different, and easier. The issue is that, because the
result of "_end - 1" is not within the boundaries of the _end array,
then the operation is "undefined" by the C specification (C99 6.5.6).
Undefined is not good.

So, I am not really asking for any complex fix, or hypervisor-wide
rework. I am only asking to avoid introducing new undefined behavior.
Casting to vaddr_t should solve it I think.

[1] https://marc.info/?l=xen-devel&m=154904722227312

_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxxxxxxxxx
https://lists.xenproject.org/mailman/listinfo/xen-devel