Re: [Xen-devel] [PATCH] fix potential pa_range_info out of bound access

[Stefano Stabellini <sstabellini@xxxxxxxxxx>]

On Fri, 9 Dec 2016, Julien Grall wrote:
> Hi Stefano,
> 
> On 09/12/16 01:40, Stefano Stabellini wrote:
> > On Thu, 8 Dec 2016, Stefano Stabellini wrote:
> > > pa_range_info has only 8 elements and is accessed using pa_range as
> > > index. pa_range is initialized to 16, potentially causing out of bound
> > > access errors. Fix the issue by initializing pa_range to the effective
> > > number of pa_range_info elements.
> > > 
> > > CID 1381865
> > > 
> > > Signed-off-by: Stefano Stabellini <sstabellini@xxxxxxxxxx>
> > > 
> > > diff --git a/xen/arch/arm/p2m.c b/xen/arch/arm/p2m.c
> > > index e4991df..245fcd1 100644
> > > --- a/xen/arch/arm/p2m.c
> > > +++ b/xen/arch/arm/p2m.c
> > > @@ -1629,7 +1629,7 @@ void __init setup_virt_paging(void)
> > >      };
> > > 
> > >      unsigned int cpu;
> > > -    unsigned int pa_range = 0x10; /* Larger than any possible value */
> > > +    unsigned int pa_range = sizeof(pa_range_info) /
> > > sizeof(pa_range_info[0]);
> > > 
> > >      for_each_online_cpu ( cpu )
> > >      {
> > 
> > this is wrong, it should be sizeof(pa_range_info) / sizeof(pa_range_info[0])
> > - 1:
> > 
> > ---
> > pa_range_info has only 8 elements and is accessed using pa_range as
> > index. pa_range is initialized to 16, potentially causing out of bound
> > access errors. Fix the issue by initializing pa_range to the effective
> > number of pa_range_info elements minus 1.
> > 
> > Coverity-ID: 1381865
> > 
> > Signed-off-by: Stefano Stabellini <sstabellini@xxxxxxxxxx>
> > 
> > diff --git a/xen/arch/arm/p2m.c b/xen/arch/arm/p2m.c
> > index e4991df..14901b0 100644
> > --- a/xen/arch/arm/p2m.c
> > +++ b/xen/arch/arm/p2m.c
> > @@ -1629,7 +1629,7 @@ void __init setup_virt_paging(void)
> >      };
> > 
> >      unsigned int cpu;
> > -    unsigned int pa_range = 0x10; /* Larger than any possible value */
> > +    unsigned int pa_range = ARRAY_SIZE(pa_range_info) - 1;
> 
> The previous value was confusing and I think this one is even more.
> 
> But this is not really the problem, it is because the boundary check the later
> on is wrong:
> 
> if ( pa_range&0x8 || !pa_range_info[pa_range].pabits )
> 
> It will only check whether bit 3 is not set. But we want to check that
> pa_range is the range of the array. I.e
> 
> pa_range < ARRAY_SIZE(pa_range_info)

You are right, that is better and I don't think it requires changing the
initial value. Andrew suggested something similar on IRC too.


> If you still want to change the pa_range initial value, then I would prefer to
> see the boot CPU one (i.e boot_cpu_data.mm64.pa_range).


_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxx
https://lists.xen.org/xen-devel