Another workaround for the glitch in the network layer, whereby one call
ignores the (otherwise kernel-wide) convention that free() calls should
not oops when passed nulls.  This code already handles that API glitch in
most other paths.
From: Erik Hovland <erik@hovland.org>
Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
                defer_kevent (dev, WORK_RX_MEMORY);
        if (retval) {
                DEBUG (dev, "rx submit --> %d\n", retval);
-               dev_kfree_skb_any (skb);
+               if (skb)
+                       dev_kfree_skb_any(skb);
                spin_lock(&dev->req_lock);
                list_add (&req->list, &dev->rx_reqs);
                spin_unlock(&dev->req_lock);