xen/xenbus: Don't leak memory when unmapping the ring on HVM backend
authorJulien Grall <julien.grall@citrix.com>
Mon, 10 Aug 2015 18:10:38 +0000 (19:10 +0100)
committerDavid Vrabel <david.vrabel@citrix.com>
Tue, 11 Aug 2015 10:05:58 +0000 (11:05 +0100)
The commit ccc9d90a9a8b5c4ad7e9708ec41f75ff9e98d61d "xenbus_client:
Extend interface to support multi-page ring" removes the call to
free_xenballooned_pages() in xenbus_unmap_ring_vfree_hvm(), leaking a
page for every shared ring.

Only with backends running in HVM domains were affected.

Signed-off-by: Julien Grall <julien.grall@citrix.com>
Cc: <stable@vger.kernel.org>
Reviewed-by: Boris Ostrovsky <boris.ostrovsky@oracle.com>
Reviewed-by: Wei Liu <wei.liu2@citrix.com>
Signed-off-by: David Vrabel <david.vrabel@citrix.com>
drivers/xen/xenbus/xenbus_client.c

index 9ad3272..e303535 100644 (file)
@@ -814,8 +814,10 @@ static int xenbus_unmap_ring_vfree_hvm(struct xenbus_device *dev, void *vaddr)
 
        rv = xenbus_unmap_ring(dev, node->handles, node->nr_handles,
                               addrs);
-       if (!rv)
+       if (!rv) {
                vunmap(vaddr);
+               free_xenballooned_pages(node->nr_handles, node->hvm.pages);
+       }
        else
                WARN(1, "Leaking %p, size %u page(s)\n", vaddr,
                     node->nr_handles);