xen-netfront: Remove BUGs on paged skb data which crosses a page boundary
authorSeth Forshee <seth.forshee@canonical.com>
Wed, 26 Nov 2014 02:28:24 +0000 (20:28 -0600)
committerGreg Kroah-Hartman <gregkh@linuxfoundation.org>
Tue, 16 Dec 2014 17:09:42 +0000 (09:09 -0800)
commit 8d609725d4357f499e2103e46011308b32f53513 upstream.

These BUGs can be erroneously triggered by frags which refer to
tail pages within a compound page. The data in these pages may
overrun the hardware page while still being contained within the
compound page, but since compound_order() evaluates to 0 for tail
pages the assertion fails. The code already iterates through
subsequent pages correctly in this scenario, so the BUGs are
unnecessary and can be removed.

Fixes: f36c374782e4 ("xen/netfront: handle compound page fragments on transmit")
Signed-off-by: Seth Forshee <seth.forshee@canonical.com>
Reviewed-by: David Vrabel <david.vrabel@citrix.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
drivers/net/xen-netfront.c

index 82e0f1fd22542ef95fa1c68119201c1cbe10940b..a1db958df4a4b6f517599ecb123dffcd283dae9d 100644 (file)
@@ -459,9 +459,6 @@ static void xennet_make_frags(struct sk_buff *skb, struct net_device *dev,
                len = skb_frag_size(frag);
                offset = frag->page_offset;
 
-               /* Data must not cross a page boundary. */
-               BUG_ON(len + offset > PAGE_SIZE<<compound_order(page));
-
                /* Skip unused frames from start of page */
                page += offset >> PAGE_SHIFT;
                offset &= ~PAGE_MASK;
@@ -469,8 +466,6 @@ static void xennet_make_frags(struct sk_buff *skb, struct net_device *dev,
                while (len > 0) {
                        unsigned long bytes;
 
-                       BUG_ON(offset >= PAGE_SIZE);
-
                        bytes = PAGE_SIZE - offset;
                        if (bytes > len)
                                bytes = len;