Re: [Xen-devel] [PATCH] hvmloader: cast to avoid potential overflow in shadow_gs_test

["Jan Beulich" <JBeulich@xxxxxxxx>]

>>> On 11.09.17 at 16:07, <wei.liu2@xxxxxxxxxx> wrote:
> e2fc5bb5cb4 ("hvmloader: dynamically determine scratch memory range
> for tests") makes the test dependent on _end. Coverity reported that
> the shift might overflow and suggested we cast i to uint64_t.
> 
> Coverity-ID: 1417660
> 
> Signed-off-by: Wei Liu <wei.liu2@xxxxxxxxxx>
> ---
> Cc: Jan Beulich <jbeulich@xxxxxxxx>
> Cc: Andrew Cooper <andrew.cooper3@xxxxxxxxxx>
> ---
>  tools/firmware/hvmloader/tests.c | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
> 
> diff --git a/tools/firmware/hvmloader/tests.c 
> b/tools/firmware/hvmloader/tests.c
> index a70c72dffb..3c4c29a6c7 100644
> --- a/tools/firmware/hvmloader/tests.c
> +++ b/tools/firmware/hvmloader/tests.c
> @@ -231,7 +231,7 @@ static int shadow_gs_test(void)
>      pd += 512;
>      /* Level 2: */
>      for ( i = 0; i <= (unsigned long)(_end - 1) >> (PAGE_SHIFT + 9); i++ )

With the shift here there's no way ...

> -        *pd++ = (i << (PAGE_SHIFT + 9)) + 0x1e3;
> +        *pd++ = ((uint64_t)i << (PAGE_SHIFT + 9)) + 0x1e3;

... this shift (or the add) can overflow, irrespective of the actual
value of _end, and with my dislike of casts I'm a little hesitant to
give my ack for such a tool weakness workaround.

Jan


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