On Sun, Sep 20, 2015 at 06:22:39PM -0400, Nicholas Krause wrote: > This fixes the incorrect use of the non wrapper version of the > function skb_dequeue in fcoe_percpu_receive_thread to use the > wrapper version as we need to protect avoid other users from > concurrently access on the sk_buff_head tmp used as the link > list head for this call to skb_queue.
The sk_buff_head tmp is a local reference on the stack, so not accessed concurrently from anywhere. It's set with skb_queue_splice_init while holding the lock for the real fcoe_rx_list, specifically to batch process without repeatedly taking and dropping the lock. Nak, flawed reasoning for justification. - Chris _______________________________________________ fcoe-devel mailing list fcoe-devel@open-fcoe.org http://lists.open-fcoe.org/mailman/listinfo/fcoe-devel
