On 09/04/2019 16:17, Jan Beulich wrote:
On 09.04.19 at 15:38, <andrew.coop...@citrix.com> wrote:
On 09/04/2019 14:01, Jan Beulich wrote:
@@ -463,9 +463,14 @@ static void timer_softirq_action(void)
if ( unlikely(ts->list != NULL) )
{
/* old_limit == (2^n)-1; new_limit == (2^(n+4))-1 */
- int old_limit = heap_metadata(heap)->limit;
- int new_limit = ((old_limit + 1) << 4) - 1;
- struct timer **newheap = xmalloc_array(struct timer *, new_limit +
1);
+ unsigned int old_limit = heap_metadata(heap)->limit;
+ unsigned int new_limit = ((old_limit + 1) << 4) - 1;
+ struct timer **newheap = NULL;
+
+ /* Don't grow the heap beyond what is representable in its metadata. */
+ if ( new_limit == (typeof(heap_metadata(heap)->limit))new_limit &&
+ new_limit + 1 )
+ newheap = xmalloc_array(struct timer *, new_limit + 1);
It would probably be helpful to have a warn_once/print_once in the case
that we do hit the metadata limit
I can do this, albeit the lack of the constructs you suggest will
make this a little ugly.
Macros implementing such behavior would be more than welcomed. We tend to
open-code WARN_ONCE/PRINT_ONCE logic in every place which is not very nice.
Cheers,
--
Julien Grall
_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xenproject.org
https://lists.xenproject.org/mailman/listinfo/xen-devel
