On Wed, 24 Jun 2015, David Miller wrote:
From: Govindarajulu Varadarajan <[email protected]> Date: Tue, 23 Jun 2015 21:47:50 +0530-static inline bool enic_poll_unlock_napi(struct vnic_rq *rq) +static inline void enic_poll_unlock_napi(struct vnic_rq *rq, + struct napi_struct *napi) { - bool rc = false; - - spin_lock(&rq->bpoll_lock); - WARN_ON(rq->bpoll_state & - (ENIC_POLL_STATE_POLL | ENIC_POLL_STATE_NAPI_YIELD)); - if (rq->bpoll_state & ENIC_POLL_STATE_POLL_YIELD) - rc = true; - rq->bpoll_state = ENIC_POLL_STATE_IDLE; - spin_unlock(&rq->bpoll_lock); - - return rc; + WARN_ON(atomic_read(&rq->bpoll_state) != ENIC_POLL_STATE_NAPI); + napi_gro_flush(napi, false); + atomic_set(&rq->bpoll_state, ENIC_POLL_STATE_IDLE); }I don't understand why you needed to add this napi_gro_flush() call here, and regardless of the reason you must explain this, in detail, in your commit message,
In napi poll, if gro is enabled, we call napi_gro_receive() to deliver the packets. Before we call napi_complete(), i.e while re-polling, if low latency busy poll is called, we use netif_receive_skb() to deliver the packets. At this point if there are some skb's held in GRO, busy poll could deliver the packets out of order. So we call napi_gro_flush() to flush skbs before we move the napi poll to idle. Let me know if this change is OK, I will send new patch with detailed description. Thanks Govind -- To unsubscribe from this list: send the line "unsubscribe netdev" in the body of a message to [email protected] More majordomo info at http://vger.kernel.org/majordomo-info.html
