Merge branches 'ib_core', 'ib_ipoib', 'srpt', 'drain-cq-v4' and 'net/9p' into k.o/for-4.6

drivers/infiniband/core/verbs.c

@@ -1657,3 +1657,167 @@ int ib_sg_to_pages(struct ib_mr *mr,
 	return i;
 }
 EXPORT_SYMBOL(ib_sg_to_pages);
+
+struct ib_drain_cqe {
+	struct ib_cqe cqe;
+	struct completion done;
+};
+
+static void ib_drain_qp_done(struct ib_cq *cq, struct ib_wc *wc)
+{
+	struct ib_drain_cqe *cqe = container_of(wc->wr_cqe, struct ib_drain_cqe,
+						cqe);
+
+	complete(&cqe->done);
+}
+
+/*
+ * Post a WR and block until its completion is reaped for the SQ.
+ */
+static void __ib_drain_sq(struct ib_qp *qp)
+{
+	struct ib_qp_attr attr = { .qp_state = IB_QPS_ERR };
+	struct ib_drain_cqe sdrain;
+	struct ib_send_wr swr = {}, *bad_swr;
+	int ret;
+
+	if (qp->send_cq->poll_ctx == IB_POLL_DIRECT) {
+		WARN_ONCE(qp->send_cq->poll_ctx == IB_POLL_DIRECT,
+			  "IB_POLL_DIRECT poll_ctx not supported for drain\n");
+		return;
+	}
+
+	swr.wr_cqe = &sdrain.cqe;
+	sdrain.cqe.done = ib_drain_qp_done;
+	init_completion(&sdrain.done);
+
+	ret = ib_modify_qp(qp, &attr, IB_QP_STATE);
+	if (ret) {
+		WARN_ONCE(ret, "failed to drain send queue: %d\n", ret);
+		return;
+	}
+
+	ret = ib_post_send(qp, &swr, &bad_swr);
+	if (ret) {
+		WARN_ONCE(ret, "failed to drain send queue: %d\n", ret);
+		return;
+	}
+
+	wait_for_completion(&sdrain.done);
+}
+
+/*
+ * Post a WR and block until its completion is reaped for the RQ.
+ */
+static void __ib_drain_rq(struct ib_qp *qp)
+{
+	struct ib_qp_attr attr = { .qp_state = IB_QPS_ERR };
+	struct ib_drain_cqe rdrain;
+	struct ib_recv_wr rwr = {}, *bad_rwr;
+	int ret;
+
+	if (qp->recv_cq->poll_ctx == IB_POLL_DIRECT) {
+		WARN_ONCE(qp->recv_cq->poll_ctx == IB_POLL_DIRECT,
+			  "IB_POLL_DIRECT poll_ctx not supported for drain\n");
+		return;
+	}
+
+	rwr.wr_cqe = &rdrain.cqe;
+	rdrain.cqe.done = ib_drain_qp_done;
+	init_completion(&rdrain.done);
+
+	ret = ib_modify_qp(qp, &attr, IB_QP_STATE);
+	if (ret) {
+		WARN_ONCE(ret, "failed to drain recv queue: %d\n", ret);
+		return;
+	}
+
+	ret = ib_post_recv(qp, &rwr, &bad_rwr);
+	if (ret) {
+		WARN_ONCE(ret, "failed to drain recv queue: %d\n", ret);
+		return;
+	}
+
+	wait_for_completion(&rdrain.done);
+}
+
+/**
+ * ib_drain_sq() - Block until all SQ CQEs have been consumed by the
+ *		   application.
+ * @qp:            queue pair to drain
+ *
+ * If the device has a provider-specific drain function, then
+ * call that.  Otherwise call the generic drain function
+ * __ib_drain_sq().
+ *
+ * The caller must:
+ *
+ * ensure there is room in the CQ and SQ for the drain work request and
+ * completion.
+ *
+ * allocate the CQ using ib_alloc_cq() and the CQ poll context cannot be
+ * IB_POLL_DIRECT.
+ *
+ * ensure that there are no other contexts that are posting WRs concurrently.
+ * Otherwise the drain is not guaranteed.
+ */
+void ib_drain_sq(struct ib_qp *qp)
+{
+	if (qp->device->drain_sq)
+		qp->device->drain_sq(qp);
+	else
+		__ib_drain_sq(qp);
+}
+EXPORT_SYMBOL(ib_drain_sq);
+
+/**
+ * ib_drain_rq() - Block until all RQ CQEs have been consumed by the
+ *		   application.
+ * @qp:            queue pair to drain
+ *
+ * If the device has a provider-specific drain function, then
+ * call that.  Otherwise call the generic drain function
+ * __ib_drain_rq().
+ *
+ * The caller must:
+ *
+ * ensure there is room in the CQ and RQ for the drain work request and
+ * completion.
+ *
+ * allocate the CQ using ib_alloc_cq() and the CQ poll context cannot be
+ * IB_POLL_DIRECT.
+ *
+ * ensure that there are no other contexts that are posting WRs concurrently.
+ * Otherwise the drain is not guaranteed.
+ */
+void ib_drain_rq(struct ib_qp *qp)
+{
+	if (qp->device->drain_rq)
+		qp->device->drain_rq(qp);
+	else
+		__ib_drain_rq(qp);
+}
+EXPORT_SYMBOL(ib_drain_rq);
+
+/**
+ * ib_drain_qp() - Block until all CQEs have been consumed by the
+ *		   application on both the RQ and SQ.
+ * @qp:            queue pair to drain
+ *
+ * The caller must:
+ *
+ * ensure there is room in the CQ(s), SQ, and RQ for drain work requests
+ * and completions.
+ *
+ * allocate the CQs using ib_alloc_cq() and the CQ poll context cannot be
+ * IB_POLL_DIRECT.
+ *
+ * ensure that there are no other contexts that are posting WRs concurrently.
+ * Otherwise the drain is not guaranteed.
+ */
+void ib_drain_qp(struct ib_qp *qp)
+{
+	ib_drain_sq(qp);
+	ib_drain_rq(qp);
+}
+EXPORT_SYMBOL(ib_drain_qp);

drivers/infiniband/hw/cxgb4/cq.c

@@ -815,8 +815,15 @@ static int c4iw_poll_cq_one(struct c4iw_cq *chp, struct ib_wc *wc)
 		}
 	}
 out:
-	if (wq)
+	if (wq) {
+		if (unlikely(qhp->attr.state != C4IW_QP_STATE_RTS)) {
+			if (t4_sq_empty(wq))
+				complete(&qhp->sq_drained);
+			if (t4_rq_empty(wq))
+				complete(&qhp->rq_drained);
+		}
 		spin_unlock(&qhp->lock);
+	}
 	return ret;
 }
 

drivers/infiniband/hw/cxgb4/iw_cxgb4.h

@@ -476,6 +476,8 @@ struct c4iw_qp {
 	wait_queue_head_t wait;
 	struct timer_list timer;
 	int sq_sig_all;
+	struct completion rq_drained;
+	struct completion sq_drained;
 };
 
 static inline struct c4iw_qp *to_c4iw_qp(struct ib_qp *ibqp)
@@ -1016,6 +1018,8 @@ extern int c4iw_wr_log;
 extern int db_fc_threshold;
 extern int db_coalescing_threshold;
 extern int use_dsgl;
+void c4iw_drain_rq(struct ib_qp *qp);
+void c4iw_drain_sq(struct ib_qp *qp);
 
 
 #endif

drivers/infiniband/hw/cxgb4/provider.c

@@ -564,6 +564,8 @@ int c4iw_register_device(struct c4iw_dev *dev)
 	dev->ibdev.get_protocol_stats = c4iw_get_mib;
 	dev->ibdev.uverbs_abi_ver = C4IW_UVERBS_ABI_VERSION;
 	dev->ibdev.get_port_immutable = c4iw_port_immutable;
+	dev->ibdev.drain_sq = c4iw_drain_sq;
+	dev->ibdev.drain_rq = c4iw_drain_rq;
 
 	dev->ibdev.iwcm = kmalloc(sizeof(struct iw_cm_verbs), GFP_KERNEL);
 	if (!dev->ibdev.iwcm)

drivers/infiniband/hw/cxgb4/qp.c

@@ -1697,6 +1697,8 @@ struct ib_qp *c4iw_create_qp(struct ib_pd *pd, struct ib_qp_init_attr *attrs,
 	qhp->attr.max_ird = 0;
 	qhp->sq_sig_all = attrs->sq_sig_type == IB_SIGNAL_ALL_WR;
 	spin_lock_init(&qhp->lock);
+	init_completion(&qhp->sq_drained);
+	init_completion(&qhp->rq_drained);
 	mutex_init(&qhp->mutex);
 	init_waitqueue_head(&qhp->wait);
 	atomic_set(&qhp->refcnt, 1);
@@ -1888,3 +1890,17 @@ int c4iw_ib_query_qp(struct ib_qp *ibqp, struct ib_qp_attr *attr,
 	init_attr->sq_sig_type = qhp->sq_sig_all ? IB_SIGNAL_ALL_WR : 0;
 	return 0;
 }
+
+void c4iw_drain_sq(struct ib_qp *ibqp)
+{
+	struct c4iw_qp *qp = to_c4iw_qp(ibqp);
+
+	wait_for_completion(&qp->sq_drained);
+}
+
+void c4iw_drain_rq(struct ib_qp *ibqp)
+{
+	struct c4iw_qp *qp = to_c4iw_qp(ibqp);
+
+	wait_for_completion(&qp->rq_drained);
+}

drivers/infiniband/ulp/ipoib/ipoib.h

@@ -244,6 +244,7 @@ struct ipoib_cm_tx {
 	unsigned	     tx_tail;
 	unsigned long	     flags;
 	u32		     mtu;
+	unsigned             max_send_sge;
 };
 
 struct ipoib_cm_rx_buf {
@@ -390,6 +391,7 @@ struct ipoib_dev_priv {
 	int	hca_caps;
 	struct ipoib_ethtool_st ethtool;
 	struct timer_list poll_timer;
+	unsigned max_send_sge;
 };
 
 struct ipoib_ah {

drivers/infiniband/ulp/ipoib/ipoib_cm.c

@@ -710,6 +710,7 @@ void ipoib_cm_send(struct net_device *dev, struct sk_buff *skb, struct ipoib_cm_
 	struct ipoib_dev_priv *priv = netdev_priv(dev);
 	struct ipoib_tx_buf *tx_req;
 	int rc;
+	unsigned usable_sge = tx->max_send_sge - !!skb_headlen(skb);
 
 	if (unlikely(skb->len > tx->mtu)) {
 		ipoib_warn(priv, "packet len %d (> %d) too long to send, dropping\n",
@@ -719,7 +720,23 @@ void ipoib_cm_send(struct net_device *dev, struct sk_buff *skb, struct ipoib_cm_
 		ipoib_cm_skb_too_long(dev, skb, tx->mtu - IPOIB_ENCAP_LEN);
 		return;
 	}
-
+	if (skb_shinfo(skb)->nr_frags > usable_sge) {
+		if (skb_linearize(skb) < 0) {
+			ipoib_warn(priv, "skb could not be linearized\n");
+			++dev->stats.tx_dropped;
+			++dev->stats.tx_errors;
+			dev_kfree_skb_any(skb);
+			return;
+		}
+		/* Does skb_linearize return ok without reducing nr_frags? */
+		if (skb_shinfo(skb)->nr_frags > usable_sge) {
+			ipoib_warn(priv, "too many frags after skb linearize\n");
+			++dev->stats.tx_dropped;
+			++dev->stats.tx_errors;
+			dev_kfree_skb_any(skb);
+			return;
+		}
+	}
 	ipoib_dbg_data(priv, "sending packet: head 0x%x length %d connection 0x%x\n",
 		       tx->tx_head, skb->len, tx->qp->qp_num);
 
@@ -1031,7 +1048,8 @@ static struct ib_qp *ipoib_cm_create_tx_qp(struct net_device *dev, struct ipoib_
 	struct ib_qp *tx_qp;
 
 	if (dev->features & NETIF_F_SG)
-		attr.cap.max_send_sge = MAX_SKB_FRAGS + 1;
+		attr.cap.max_send_sge =
+			min_t(u32, priv->ca->attrs.max_sge, MAX_SKB_FRAGS + 1);
 
 	tx_qp = ib_create_qp(priv->pd, &attr);
 	if (PTR_ERR(tx_qp) == -EINVAL) {
@@ -1040,6 +1058,7 @@ static struct ib_qp *ipoib_cm_create_tx_qp(struct net_device *dev, struct ipoib_
 		attr.create_flags &= ~IB_QP_CREATE_USE_GFP_NOIO;
 		tx_qp = ib_create_qp(priv->pd, &attr);
 	}
+	tx->max_send_sge = attr.cap.max_send_sge;
 	return tx_qp;
 }
 

drivers/infiniband/ulp/ipoib/ipoib_ib.c

@@ -538,6 +538,7 @@ void ipoib_send(struct net_device *dev, struct sk_buff *skb,
 	struct ipoib_tx_buf *tx_req;
 	int hlen, rc;
 	void *phead;
+	unsigned usable_sge = priv->max_send_sge - !!skb_headlen(skb);
 
 	if (skb_is_gso(skb)) {
 		hlen = skb_transport_offset(skb) + tcp_hdrlen(skb);
@@ -561,6 +562,23 @@ void ipoib_send(struct net_device *dev, struct sk_buff *skb,
 		phead = NULL;
 		hlen  = 0;
 	}
+	if (skb_shinfo(skb)->nr_frags > usable_sge) {
+		if (skb_linearize(skb) < 0) {
+			ipoib_warn(priv, "skb could not be linearized\n");
+			++dev->stats.tx_dropped;
+			++dev->stats.tx_errors;
+			dev_kfree_skb_any(skb);
+			return;
+		}
+		/* Does skb_linearize return ok without reducing nr_frags? */
+		if (skb_shinfo(skb)->nr_frags > usable_sge) {
+			ipoib_warn(priv, "too many frags after skb linearize\n");
+			++dev->stats.tx_dropped;
+			++dev->stats.tx_errors;
+			dev_kfree_skb_any(skb);
+			return;
+		}
+	}
 
 	ipoib_dbg_data(priv, "sending packet, length=%d address=%p qpn=0x%06x\n",
 		       skb->len, address, qpn);

drivers/infiniband/ulp/ipoib/ipoib_verbs.c

@@ -206,7 +206,8 @@ int ipoib_transport_dev_init(struct net_device *dev, struct ib_device *ca)
 		init_attr.create_flags |= IB_QP_CREATE_NETIF_QP;
 
 	if (dev->features & NETIF_F_SG)
-		init_attr.cap.max_send_sge = MAX_SKB_FRAGS + 1;
+		init_attr.cap.max_send_sge =
+			min_t(u32, priv->ca->attrs.max_sge, MAX_SKB_FRAGS + 1);
 
 	priv->qp = ib_create_qp(priv->pd, &init_attr);
 	if (IS_ERR(priv->qp)) {
@@ -233,6 +234,8 @@ int ipoib_transport_dev_init(struct net_device *dev, struct ib_device *ca)
 	priv->rx_wr.next = NULL;
 	priv->rx_wr.sg_list = priv->rx_sge;
 
+	priv->max_send_sge = init_attr.cap.max_send_sge;
+
 	return 0;
 
 out_free_send_cq:

drivers/infiniband/ulp/iser/iscsi_iser.h

@@ -458,9 +458,6 @@ struct iser_fr_pool {
  * @comp:                iser completion context
  * @fr_pool:             connection fast registration poool
  * @pi_support:          Indicate device T10-PI support
- * @last:                last send wr to signal all flush errors were drained
- * @last_cqe:            cqe handler for last wr
- * @last_comp:           completes when all connection completions consumed
  */
 struct ib_conn {
 	struct rdma_cm_id           *cma_id;
@@ -472,10 +469,7 @@ struct ib_conn {
 	struct iser_comp	    *comp;
 	struct iser_fr_pool          fr_pool;
 	bool			     pi_support;
-	struct ib_send_wr	     last;
-	struct ib_cqe		     last_cqe;
 	struct ib_cqe		     reg_cqe;
-	struct completion	     last_comp;
 };
 
 /**
@@ -617,7 +611,6 @@ void iser_cmd_comp(struct ib_cq *cq, struct ib_wc *wc);
 void iser_ctrl_comp(struct ib_cq *cq, struct ib_wc *wc);
 void iser_dataout_comp(struct ib_cq *cq, struct ib_wc *wc);
 void iser_reg_comp(struct ib_cq *cq, struct ib_wc *wc);
-void iser_last_comp(struct ib_cq *cq, struct ib_wc *wc);
 
 void iser_task_rdma_init(struct iscsi_iser_task *task);
 

drivers/infiniband/ulp/iser/iser_initiator.c

@@ -729,13 +729,6 @@ void iser_dataout_comp(struct ib_cq *cq, struct ib_wc *wc)
 	kmem_cache_free(ig.desc_cache, desc);
 }
 
-void iser_last_comp(struct ib_cq *cq, struct ib_wc *wc)
-{
-	struct ib_conn *ib_conn = wc->qp->qp_context;
-
-	complete(&ib_conn->last_comp);
-}
-
 void iser_task_rdma_init(struct iscsi_iser_task *iser_task)
 
 {

drivers/infiniband/ulp/iser/iser_verbs.c

@@ -663,7 +663,6 @@ void iser_conn_release(struct iser_conn *iser_conn)
 int iser_conn_terminate(struct iser_conn *iser_conn)
 {
 	struct ib_conn *ib_conn = &iser_conn->ib_conn;
-	struct ib_send_wr *bad_wr;
 	int err = 0;
 
 	/* terminate the iser conn only if the conn state is UP */
@@ -688,14 +687,8 @@ int iser_conn_terminate(struct iser_conn *iser_conn)
 			iser_err("Failed to disconnect, conn: 0x%p err %d\n",
 				 iser_conn, err);
 
-		/* post an indication that all flush errors were consumed */
-		err = ib_post_send(ib_conn->qp, &ib_conn->last, &bad_wr);
-		if (err) {
-			iser_err("conn %p failed to post last wr", ib_conn);
-			return 1;
-		}
-
-		wait_for_completion(&ib_conn->last_comp);
+		/* block until all flush errors are consumed */
+		ib_drain_sq(ib_conn->qp);
 	}
 
 	return 1;
@@ -954,10 +947,6 @@ void iser_conn_init(struct iser_conn *iser_conn)
 
 	ib_conn->post_recv_buf_count = 0;
 	ib_conn->reg_cqe.done = iser_reg_comp;
-	ib_conn->last_cqe.done = iser_last_comp;
-	ib_conn->last.wr_cqe = &ib_conn->last_cqe;
-	ib_conn->last.opcode = IB_WR_SEND;
-	init_completion(&ib_conn->last_comp);
 }
 
  /**

drivers/infiniband/ulp/srp/ib_srp.c

@@ -446,49 +446,17 @@ static struct srp_fr_pool *srp_alloc_fr_pool(struct srp_target_port *target)
 				  dev->max_pages_per_mr);
 }
 
-static void srp_drain_done(struct ib_cq *cq, struct ib_wc *wc)
-{
-	struct srp_rdma_ch *ch = cq->cq_context;
-
-	complete(&ch->done);
-}
-
-static struct ib_cqe srp_drain_cqe = {
-	.done		= srp_drain_done,
-};
-
 /**
  * srp_destroy_qp() - destroy an RDMA queue pair
  * @ch: SRP RDMA channel.
  *
- * Change a queue pair into the error state and wait until all receive
- * completions have been processed before destroying it. This avoids that
- * the receive completion handler can access the queue pair while it is
+ * Drain the qp before destroying it.  This avoids that the receive
+ * completion handler can access the queue pair while it is
  * being destroyed.
  */
 static void srp_destroy_qp(struct srp_rdma_ch *ch)
 {
-	static struct ib_qp_attr attr = { .qp_state = IB_QPS_ERR };
-	static struct ib_recv_wr wr = { 0 };
-	struct ib_recv_wr *bad_wr;
-	int ret;
-
-	wr.wr_cqe = &srp_drain_cqe;
-	/* Destroying a QP and reusing ch->done is only safe if not connected */
-	WARN_ON_ONCE(ch->connected);
-
-	ret = ib_modify_qp(ch->qp, &attr, IB_QP_STATE);
-	WARN_ONCE(ret, "ib_cm_init_qp_attr() returned %d\n", ret);
-	if (ret)
-		goto out;
-
-	init_completion(&ch->done);
-	ret = ib_post_recv(ch->qp, &wr, &bad_wr);
-	WARN_ONCE(ret, "ib_post_recv() returned %d\n", ret);
-	if (ret == 0)
-		wait_for_completion(&ch->done);
-
-out:
+	ib_drain_rq(ch->qp);
 	ib_destroy_qp(ch->qp);
 }
 
@@ -508,7 +476,7 @@ static int srp_create_ch_ib(struct srp_rdma_ch *ch)
 	if (!init_attr)
 		return -ENOMEM;
 
-	/* queue_size + 1 for ib_drain_qp */
+	/* queue_size + 1 for ib_drain_rq() */
 	recv_cq = ib_alloc_cq(dev->dev, ch, target->queue_size + 1,
 				ch->comp_vector, IB_POLL_SOFTIRQ);
 	if (IS_ERR(recv_cq)) {

drivers/infiniband/ulp/srpt/ib_srpt.h

@@ -218,20 +218,20 @@ struct srpt_send_ioctx {
 
 /**
  * enum rdma_ch_state - SRP channel state.
- * @CH_CONNECTING:	 QP is in RTR state; waiting for RTU.
- * @CH_LIVE:		 QP is in RTS state.
- * @CH_DISCONNECTING:    DREQ has been received; waiting for DREP
- *                       or DREQ has been send and waiting for DREP
- *                       or .
- * @CH_DRAINING:	 QP is in ERR state; waiting for last WQE event.
- * @CH_RELEASING:	 Last WQE event has been received; releasing resources.
+ * @CH_CONNECTING:    QP is in RTR state; waiting for RTU.
+ * @CH_LIVE:	      QP is in RTS state.
+ * @CH_DISCONNECTING: DREQ has been sent and waiting for DREP or DREQ has
+ *                    been received.
+ * @CH_DRAINING:      DREP has been received or waiting for DREP timed out
+ *                    and last work request has been queued.
+ * @CH_DISCONNECTED:  Last completion has been received.
  */
 enum rdma_ch_state {
 	CH_CONNECTING,
 	CH_LIVE,
 	CH_DISCONNECTING,
 	CH_DRAINING,
-	CH_RELEASING
+	CH_DISCONNECTED,
 };
 
 /**
@@ -267,6 +267,8 @@ struct srpt_rdma_ch {
 	struct ib_cm_id		*cm_id;
 	struct ib_qp		*qp;
 	struct ib_cq		*cq;
+	struct ib_cqe		zw_cqe;
+	struct kref		kref;
 	int			rq_size;
 	u32			rsp_size;
 	atomic_t		sq_wr_avail;
@@ -286,7 +288,6 @@ struct srpt_rdma_ch {
 	u8			sess_name[36];
 	struct work_struct	release_work;
 	struct completion	*release_done;
-	bool			in_shutdown;
 };
 
 /**
@@ -343,7 +344,7 @@ struct srpt_port {
  * @ioctx_ring:    Per-HCA SRQ.
  * @rch_list:      Per-device channel list -- see also srpt_rdma_ch.list.
  * @ch_releaseQ:   Enables waiting for removal from rch_list.
- * @spinlock:      Protects rch_list and tpg.
+ * @mutex:         Protects rch_list.
  * @port:          Information about the ports owned by this HCA.
  * @event_handler: Per-HCA asynchronous IB event handler.
  * @list:          Node in srpt_dev_list.
@@ -357,18 +358,10 @@ struct srpt_device {
 	struct srpt_recv_ioctx	**ioctx_ring;
 	struct list_head	rch_list;
 	wait_queue_head_t	ch_releaseQ;
-	spinlock_t		spinlock;
+	struct mutex		mutex;
 	struct srpt_port	port[2];
 	struct ib_event_handler	event_handler;
 	struct list_head	list;
 };
 
-/**
- * struct srpt_node_acl - Per-initiator ACL data (managed via configfs).
- * @nacl:      Target core node ACL information.
- */
-struct srpt_node_acl {
-	struct se_node_acl	nacl;
-};
-
 #endif				/* IB_SRPT_H */

include/rdma/ib_verbs.h

@@ -1846,6 +1846,8 @@ struct ib_device {
 	int			   (*check_mr_status)(struct ib_mr *mr, u32 check_mask,
 						      struct ib_mr_status *mr_status);
 	void			   (*disassociate_ucontext)(struct ib_ucontext *ibcontext);
+	void			   (*drain_rq)(struct ib_qp *qp);
+	void			   (*drain_sq)(struct ib_qp *qp);
 
 	struct ib_dma_mapping_ops   *dma_ops;
 
@@ -3094,4 +3096,7 @@ int ib_sg_to_pages(struct ib_mr *mr,
 		   int sg_nents,
 		   int (*set_page)(struct ib_mr *, u64));
 
+void ib_drain_rq(struct ib_qp *qp);
+void ib_drain_sq(struct ib_qp *qp);
+void ib_drain_qp(struct ib_qp *qp);
 #endif /* IB_VERBS_H */

net/9p/trans_rdma.c

@@ -109,14 +109,13 @@ struct p9_trans_rdma {
 /**
  * p9_rdma_context - Keeps track of in-process WR
  *
- * @wc_op: The original WR op for when the CQE completes in error.
  * @busa: Bus address to unmap when the WR completes
  * @req: Keeps track of requests (send)
  * @rc: Keepts track of replies (receive)
  */
 struct p9_rdma_req;
 struct p9_rdma_context {
-	enum ib_wc_opcode wc_op;
+	struct ib_cqe cqe;
 	dma_addr_t busa;
 	union {
 		struct p9_req_t *req;
@@ -284,9 +283,12 @@ p9_cm_event_handler(struct rdma_cm_id *id, struct rdma_cm_event *event)
 }
 
 static void
-handle_recv(struct p9_client *client, struct p9_trans_rdma *rdma,
-	    struct p9_rdma_context *c, enum ib_wc_status status, u32 byte_len)
+recv_done(struct ib_cq *cq, struct ib_wc *wc)
 {
+	struct p9_client *client = cq->cq_context;
+	struct p9_trans_rdma *rdma = client->trans;
+	struct p9_rdma_context *c =
+		container_of(wc->wr_cqe, struct p9_rdma_context, cqe);
 	struct p9_req_t *req;
 	int err = 0;
 	int16_t tag;
@@ -295,7 +297,7 @@ handle_recv(struct p9_client *client, struct p9_trans_rdma *rdma,
 	ib_dma_unmap_single(rdma->cm_id->device, c->busa, client->msize,
 							 DMA_FROM_DEVICE);
 
-	if (status != IB_WC_SUCCESS)
+	if (wc->status != IB_WC_SUCCESS)
 		goto err_out;
 
 	err = p9_parse_header(c->rc, NULL, NULL, &tag, 1);
@@ -316,21 +318,32 @@ handle_recv(struct p9_client *client, struct p9_trans_rdma *rdma,
 	req->rc = c->rc;
 	p9_client_cb(client, req, REQ_STATUS_RCVD);
 
+ out:
+	up(&rdma->rq_sem);
+	kfree(c);
 	return;
 
  err_out:
-	p9_debug(P9_DEBUG_ERROR, "req %p err %d status %d\n", req, err, status);
+	p9_debug(P9_DEBUG_ERROR, "req %p err %d status %d\n",
+			req, err, wc->status);
 	rdma->state = P9_RDMA_FLUSHING;
 	client->status = Disconnected;
+	goto out;
 }
 
 static void
-handle_send(struct p9_client *client, struct p9_trans_rdma *rdma,
-	    struct p9_rdma_context *c, enum ib_wc_status status, u32 byte_len)
+send_done(struct ib_cq *cq, struct ib_wc *wc)
 {
+	struct p9_client *client = cq->cq_context;
+	struct p9_trans_rdma *rdma = client->trans;
+	struct p9_rdma_context *c =
+		container_of(wc->wr_cqe, struct p9_rdma_context, cqe);
+
 	ib_dma_unmap_single(rdma->cm_id->device,
 			    c->busa, c->req->tc->size,
 			    DMA_TO_DEVICE);
+	up(&rdma->sq_sem);
+	kfree(c);
 }
 
 static void qp_event_handler(struct ib_event *event, void *context)
@@ -339,42 +352,6 @@ static void qp_event_handler(struct ib_event *event, void *context)
 		 event->event, context);
 }
 
-static void cq_comp_handler(struct ib_cq *cq, void *cq_context)
-{
-	struct p9_client *client = cq_context;
-	struct p9_trans_rdma *rdma = client->trans;
-	int ret;
-	struct ib_wc wc;
-
-	ib_req_notify_cq(rdma->cq, IB_CQ_NEXT_COMP);
-	while ((ret = ib_poll_cq(cq, 1, &wc)) > 0) {
-		struct p9_rdma_context *c = (void *) (unsigned long) wc.wr_id;
-
-		switch (c->wc_op) {
-		case IB_WC_RECV:
-			handle_recv(client, rdma, c, wc.status, wc.byte_len);
-			up(&rdma->rq_sem);
-			break;
-
-		case IB_WC_SEND:
-			handle_send(client, rdma, c, wc.status, wc.byte_len);
-			up(&rdma->sq_sem);
-			break;
-
-		default:
-			pr_err("unexpected completion type, c->wc_op=%d, wc.opcode=%d, status=%d\n",
-			       c->wc_op, wc.opcode, wc.status);
-			break;
-		}
-		kfree(c);
-	}
-}
-
-static void cq_event_handler(struct ib_event *e, void *v)
-{
-	p9_debug(P9_DEBUG_ERROR, "CQ event %d context %p\n", e->event, v);
-}
-
 static void rdma_destroy_trans(struct p9_trans_rdma *rdma)
 {
 	if (!rdma)
@@ -387,7 +364,7 @@ static void rdma_destroy_trans(struct p9_trans_rdma *rdma)
 		ib_dealloc_pd(rdma->pd);
 
 	if (rdma->cq && !IS_ERR(rdma->cq))
-		ib_destroy_cq(rdma->cq);
+		ib_free_cq(rdma->cq);
 
 	if (rdma->cm_id && !IS_ERR(rdma->cm_id))
 		rdma_destroy_id(rdma->cm_id);
@@ -408,13 +385,14 @@ post_recv(struct p9_client *client, struct p9_rdma_context *c)
 	if (ib_dma_mapping_error(rdma->cm_id->device, c->busa))
 		goto error;
 
+	c->cqe.done = recv_done;
+
 	sge.addr = c->busa;
 	sge.length = client->msize;
 	sge.lkey = rdma->pd->local_dma_lkey;
 
 	wr.next = NULL;
-	c->wc_op = IB_WC_RECV;
-	wr.wr_id = (unsigned long) c;
+	wr.wr_cqe = &c->cqe;
 	wr.sg_list = &sge;
 	wr.num_sge = 1;
 	return ib_post_recv(rdma->qp, &wr, &bad_wr);
@@ -499,13 +477,14 @@ static int rdma_request(struct p9_client *client, struct p9_req_t *req)
 		goto send_error;
 	}
 
+	c->cqe.done = send_done;
+
 	sge.addr = c->busa;
 	sge.length = c->req->tc->size;
 	sge.lkey = rdma->pd->local_dma_lkey;
 
 	wr.next = NULL;
-	c->wc_op = IB_WC_SEND;
-	wr.wr_id = (unsigned long) c;
+	wr.wr_cqe = &c->cqe;
 	wr.opcode = IB_WR_SEND;
 	wr.send_flags = IB_SEND_SIGNALED;
 	wr.sg_list = &sge;
@@ -642,7 +621,6 @@ rdma_create_trans(struct p9_client *client, const char *addr, char *args)
 	struct p9_trans_rdma *rdma;
 	struct rdma_conn_param conn_param;
 	struct ib_qp_init_attr qp_attr;
-	struct ib_cq_init_attr cq_attr = {};
 
 	/* Parse the transport specific mount options */
 	err = parse_opts(args, &opts);
@@ -695,13 +673,11 @@ rdma_create_trans(struct p9_client *client, const char *addr, char *args)
 		goto error;
 
 	/* Create the Completion Queue */
-	cq_attr.cqe = opts.sq_depth + opts.rq_depth + 1;
-	rdma->cq = ib_create_cq(rdma->cm_id->device, cq_comp_handler,
-				cq_event_handler, client,
-				&cq_attr);
+	rdma->cq = ib_alloc_cq(rdma->cm_id->device, client,
+			opts.sq_depth + opts.rq_depth + 1,
+			0, IB_POLL_SOFTIRQ);
 	if (IS_ERR(rdma->cq))
 		goto error;
-	ib_req_notify_cq(rdma->cq, IB_CQ_NEXT_COMP);
 
 	/* Create the Protection Domain */
 	rdma->pd = ib_alloc_pd(rdma->cm_id->device);