Re: [Xen-devel] [PATCH 6 of 9] x86/mm: Rework stale p2m auditing

["Jan Beulich" <JBeulich@xxxxxxxx>]

 >>> On 29.11.11 at 21:21, Andres Lagar-Cavilla <andres@xxxxxxxxxxxxxxxx> wrote:
> @@ -559,13 +560,15 @@ int set_p2m_entry(struct p2m_domain *p2m
>  extern void p2m_pt_init(struct p2m_domain *p2m);
>  
>  /* Debugging and auditing of the P2M code? */
> -#define P2M_AUDIT     0
> +#define P2M_AUDIT     1
>  #define P2M_DEBUGGING 0
>  
>  #if P2M_AUDIT

Was this change of the default really intended? It uncovered a problem
in 22356:cbb6b4b17024 (the code meanwhile moved elsewhere): In a
32-bit build,

        if ( gfn == 0x55555555 || gfn == 0x5555555555555555 )

causes (at least with some compiler versions, didn't check tonight's
stage tester results yet) a compiler warning (which in turn fails the
build). And a comparison like this is wrong on 64-bit too - see other
(sort of proper; "sort of" because this isn't really appropriate anyway,
especially not without at least using a manifest constant that would
force these to be in sync with the memset()-s that actually establish
these values) examples of using these magic values e.g. in
p2m_alloc_table() or guest_physmap_add_entry().

Jan


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