1 files changed, 109 insertions, 84 deletions

diff --git a/drivers/infiniband/core/cm.c b/drivers/infiniband/core/cm.c
index ff58058aeadc..142170473e75 100644
--- a/drivers/infiniband/core/cm.c
+++ b/drivers/infiniband/core/cm.c
@@ -34,6 +34,9 @@ MODULE_AUTHOR("Sean Hefty");
 MODULE_DESCRIPTION("InfiniBand CM");
 MODULE_LICENSE("Dual BSD/GPL");
 
+#define CM_DESTROY_ID_WAIT_TIMEOUT 10000 /* msecs */
+#define CM_DIRECT_RETRY_CTX ((void *) 1UL)
+
 static const char * const ibcm_rej_reason_strs[] = {
 	[IB_CM_REJ_NO_QP]			= "no QP",
 	[IB_CM_REJ_NO_EEC]			= "no EEC",
@@ -92,8 +95,7 @@ static void cm_process_work(struct cm_id_private *cm_id_priv,
 			    struct cm_work *work);
 static int cm_send_sidr_rep_locked(struct cm_id_private *cm_id_priv,
 				   struct ib_cm_sidr_rep_param *param);
-static int cm_send_dreq_locked(struct cm_id_private *cm_id_priv,
-			       const void *private_data, u8 private_data_len);
+static void cm_issue_dreq(struct cm_id_private *cm_id_priv);
 static int cm_send_drep_locked(struct cm_id_private *cm_id_priv,
 			       void *private_data, u8 private_data_len);
 static int cm_send_rej_locked(struct cm_id_private *cm_id_priv,
@@ -306,12 +308,7 @@ static struct ib_mad_send_buf *cm_alloc_msg(struct cm_id_private *cm_id_priv)
 		goto out;
 	}
 
-	/* Timeout set by caller if response is expected. */
 	m->ah = ah;
-	m->retries = cm_id_priv->max_cm_retries;
-
-	refcount_inc(&cm_id_priv->refcount);
-	m->context[0] = cm_id_priv;
 
 out:
 	spin_unlock(&cm_id_priv->av.port->cm_dev->mad_agent_lock);
@@ -320,16 +317,13 @@ out:
 
 static void cm_free_msg(struct ib_mad_send_buf *msg)
 {
-	struct cm_id_private *cm_id_priv = msg->context[0];
-
 	if (msg->ah)
 		rdma_destroy_ah(msg->ah, 0);
-	cm_deref_id(cm_id_priv);
 	ib_free_send_mad(msg);
 }
 
 static struct ib_mad_send_buf *
-cm_alloc_priv_msg(struct cm_id_private *cm_id_priv)
+cm_alloc_priv_msg(struct cm_id_private *cm_id_priv, enum ib_cm_state state)
 {
 	struct ib_mad_send_buf *msg;
 
@@ -338,7 +332,15 @@ cm_alloc_priv_msg(struct cm_id_private *cm_id_priv)
 	msg = cm_alloc_msg(cm_id_priv);
 	if (IS_ERR(msg))
 		return msg;
+
 	cm_id_priv->msg = msg;
+	refcount_inc(&cm_id_priv->refcount);
+	msg->context[0] = cm_id_priv;
+	msg->context[1] = (void *) (unsigned long) state;
+
+	msg->retries = cm_id_priv->max_cm_retries;
+	msg->timeout_ms = cm_id_priv->timeout_ms;
+
 	return msg;
 }
 
@@ -357,13 +359,20 @@ static void cm_free_priv_msg(struct ib_mad_send_buf *msg)
 	ib_free_send_mad(msg);
 }
 
-static struct ib_mad_send_buf *cm_alloc_response_msg_no_ah(struct cm_port *port,
-							   struct ib_mad_recv_wc *mad_recv_wc)
+static struct ib_mad_send_buf *
+cm_alloc_response_msg_no_ah(struct cm_port *port,
+			    struct ib_mad_recv_wc *mad_recv_wc,
+			    bool direct_retry)
 {
-	return ib_create_send_mad(port->mad_agent, 1, mad_recv_wc->wc->pkey_index,
-				  0, IB_MGMT_MAD_HDR, IB_MGMT_MAD_DATA,
-				  GFP_ATOMIC,
-				  IB_MGMT_BASE_VERSION);
+	struct ib_mad_send_buf *m;
+
+	m = ib_create_send_mad(port->mad_agent, 1, mad_recv_wc->wc->pkey_index,
+			       0, IB_MGMT_MAD_HDR, IB_MGMT_MAD_DATA,
+			       GFP_ATOMIC, IB_MGMT_BASE_VERSION);
+	if (!IS_ERR(m))
+		m->context[0] = direct_retry ? CM_DIRECT_RETRY_CTX : NULL;
+
+	return m;
 }
 
 static int cm_create_response_msg_ah(struct cm_port *port,
@@ -383,12 +392,13 @@ static int cm_create_response_msg_ah(struct cm_port *port,
 
 static int cm_alloc_response_msg(struct cm_port *port,
 				 struct ib_mad_recv_wc *mad_recv_wc,
+				 bool direct_retry,
 				 struct ib_mad_send_buf **msg)
 {
 	struct ib_mad_send_buf *m;
 	int ret;
 
-	m = cm_alloc_response_msg_no_ah(port, mad_recv_wc);
+	m = cm_alloc_response_msg_no_ah(port, mad_recv_wc, direct_retry);
 	if (IS_ERR(m))
 		return PTR_ERR(m);
 
@@ -402,13 +412,6 @@ static int cm_alloc_response_msg(struct cm_port *port,
 	return 0;
 }
 
-static void cm_free_response_msg(struct ib_mad_send_buf *msg)
-{
-	if (msg->ah)
-		rdma_destroy_ah(msg->ah, 0);
-	ib_free_send_mad(msg);
-}
-
 static void *cm_copy_private_data(const void *private_data, u8 private_data_len)
 {
 	void *data;
@@ -1025,13 +1028,26 @@ static void cm_reset_to_idle(struct cm_id_private *cm_id_priv)
 	}
 }
 
+static noinline void cm_destroy_id_wait_timeout(struct ib_cm_id *cm_id,
+						enum ib_cm_state old_state)
+{
+	struct cm_id_private *cm_id_priv;
+
+	cm_id_priv = container_of(cm_id, struct cm_id_private, id);
+	pr_err("%s: cm_id=%p timed out. state %d -> %d, refcnt=%d\n", __func__,
+	       cm_id, old_state, cm_id->state, refcount_read(&cm_id_priv->refcount));
+}
+
 static void cm_destroy_id(struct ib_cm_id *cm_id, int err)
 {
 	struct cm_id_private *cm_id_priv;
+	enum ib_cm_state old_state;
 	struct cm_work *work;
+	int ret;
 
 	cm_id_priv = container_of(cm_id, struct cm_id_private, id);
 	spin_lock_irq(&cm_id_priv->lock);
+	old_state = cm_id->state;
 retest:
 	switch (cm_id->state) {
 	case IB_CM_LISTEN:
@@ -1095,7 +1111,8 @@ retest:
 			cm_id->state = IB_CM_IDLE;
 			break;
 		}
-		cm_send_dreq_locked(cm_id_priv, NULL, 0);
+		cm_issue_dreq(cm_id_priv);
+		cm_enter_timewait(cm_id_priv);
 		goto retest;
 	case IB_CM_DREQ_SENT:
 		ib_cancel_mad(cm_id_priv->msg);
@@ -1135,7 +1152,14 @@ retest:
 
 	xa_erase(&cm.local_id_table, cm_local_id(cm_id->local_id));
 	cm_deref_id(cm_id_priv);
-	wait_for_completion(&cm_id_priv->comp);
+	do {
+		ret = wait_for_completion_timeout(&cm_id_priv->comp,
+						  msecs_to_jiffies(
+						  CM_DESTROY_ID_WAIT_TIMEOUT));
+		if (!ret) /* timeout happened */
+			cm_destroy_id_wait_timeout(cm_id, old_state);
+	} while (!ret);
+
 	while ((work = cm_dequeue_work(cm_id_priv)) != NULL)
 		cm_free_work(work);
 
@@ -1536,7 +1560,7 @@ int ib_send_cm_req(struct ib_cm_id *cm_id,
 	if (param->alternate_path)
 		cm_move_av_from_path(&cm_id_priv->alt_av, &alt_av);
 
-	msg = cm_alloc_priv_msg(cm_id_priv);
+	msg = cm_alloc_priv_msg(cm_id_priv, IB_CM_REQ_SENT);
 	if (IS_ERR(msg)) {
 		ret = PTR_ERR(msg);
 		goto out_unlock;
@@ -1545,8 +1569,6 @@ int ib_send_cm_req(struct ib_cm_id *cm_id,
 	req_msg = (struct cm_req_msg *)msg->mad;
 	cm_format_req(req_msg, cm_id_priv, param);
 	cm_id_priv->tid = req_msg->hdr.tid;
-	msg->timeout_ms = cm_id_priv->timeout_ms;
-	msg->context[1] = (void *)(unsigned long)IB_CM_REQ_SENT;
 
 	cm_id_priv->local_qpn = cpu_to_be32(IBA_GET(CM_REQ_LOCAL_QPN, req_msg));
 	cm_id_priv->rq_psn = cpu_to_be32(IBA_GET(CM_REQ_STARTING_PSN, req_msg));
@@ -1577,7 +1599,7 @@ static int cm_issue_rej(struct cm_port *port,
 	struct cm_rej_msg *rej_msg, *rcv_msg;
 	int ret;
 
-	ret = cm_alloc_response_msg(port, mad_recv_wc, &msg);
+	ret = cm_alloc_response_msg(port, mad_recv_wc, false, &msg);
 	if (ret)
 		return ret;
 
@@ -1603,7 +1625,7 @@ static int cm_issue_rej(struct cm_port *port,
 		IBA_GET(CM_REJ_REMOTE_COMM_ID, rcv_msg));
 	ret = ib_post_send_mad(msg, NULL);
 	if (ret)
-		cm_free_response_msg(msg);
+		cm_free_msg(msg);
 
 	return ret;
 }
@@ -1930,7 +1952,7 @@ static void cm_dup_req_handler(struct cm_work *work,
 	}
 	spin_unlock_irq(&cm_id_priv->lock);
 
-	ret = cm_alloc_response_msg(work->port, work->mad_recv_wc, &msg);
+	ret = cm_alloc_response_msg(work->port, work->mad_recv_wc, true, &msg);
 	if (ret)
 		return;
 
@@ -1959,7 +1981,7 @@ static void cm_dup_req_handler(struct cm_work *work,
 	return;
 
 unlock:	spin_unlock_irq(&cm_id_priv->lock);
-free:	cm_free_response_msg(msg);
+free:	cm_free_msg(msg);
 }
 
 static struct cm_id_private *cm_match_req(struct cm_work *work,
@@ -2273,7 +2295,7 @@ int ib_send_cm_rep(struct ib_cm_id *cm_id,
 		goto out;
 	}
 
-	msg = cm_alloc_priv_msg(cm_id_priv);
+	msg = cm_alloc_priv_msg(cm_id_priv, IB_CM_REP_SENT);
 	if (IS_ERR(msg)) {
 		ret = PTR_ERR(msg);
 		goto out;
@@ -2281,8 +2303,6 @@ int ib_send_cm_rep(struct ib_cm_id *cm_id,
 
 	rep_msg = (struct cm_rep_msg *) msg->mad;
 	cm_format_rep(rep_msg, cm_id_priv, param);
-	msg->timeout_ms = cm_id_priv->timeout_ms;
-	msg->context[1] = (void *) (unsigned long) IB_CM_REP_SENT;
 
 	trace_icm_send_rep(cm_id);
 	ret = ib_post_send_mad(msg, NULL);
@@ -2423,7 +2443,7 @@ static void cm_dup_rep_handler(struct cm_work *work)
 
 	atomic_long_inc(
 		&work->port->counters[CM_RECV_DUPLICATES][CM_REP_COUNTER]);
-	ret = cm_alloc_response_msg(work->port, work->mad_recv_wc, &msg);
+	ret = cm_alloc_response_msg(work->port, work->mad_recv_wc, true, &msg);
 	if (ret)
 		goto deref;
 
@@ -2448,7 +2468,7 @@ static void cm_dup_rep_handler(struct cm_work *work)
 	goto deref;
 
 unlock:	spin_unlock_irq(&cm_id_priv->lock);
-free:	cm_free_response_msg(msg);
+free:	cm_free_msg(msg);
 deref:	cm_deref_id(cm_id_priv);
 }
 
@@ -2632,59 +2652,68 @@ static void cm_format_dreq(struct cm_dreq_msg *dreq_msg,
 			    private_data_len);
 }
 
-static int cm_send_dreq_locked(struct cm_id_private *cm_id_priv,
-			       const void *private_data, u8 private_data_len)
+static void cm_issue_dreq(struct cm_id_private *cm_id_priv)
 {
 	struct ib_mad_send_buf *msg;
 	int ret;
 
 	lockdep_assert_held(&cm_id_priv->lock);
 
+	msg = cm_alloc_msg(cm_id_priv);
+	if (IS_ERR(msg))
+		return;
+
+	cm_format_dreq((struct cm_dreq_msg *) msg->mad, cm_id_priv, NULL, 0);
+
+	trace_icm_send_dreq(&cm_id_priv->id);
+	ret = ib_post_send_mad(msg, NULL);
+	if (ret)
+		cm_free_msg(msg);
+}
+
+int ib_send_cm_dreq(struct ib_cm_id *cm_id, const void *private_data,
+		    u8 private_data_len)
+{
+	struct cm_id_private *cm_id_priv =
+		container_of(cm_id, struct cm_id_private, id);
+	struct ib_mad_send_buf *msg;
+	unsigned long flags;
+	int ret;
+
 	if (private_data && private_data_len > IB_CM_DREQ_PRIVATE_DATA_SIZE)
 		return -EINVAL;
 
+	spin_lock_irqsave(&cm_id_priv->lock, flags);
 	if (cm_id_priv->id.state != IB_CM_ESTABLISHED) {
 		trace_icm_dreq_skipped(&cm_id_priv->id);
-		return -EINVAL;
+		ret = -EINVAL;
+		goto unlock;
 	}
 
 	if (cm_id_priv->id.lap_state == IB_CM_LAP_SENT ||
 	    cm_id_priv->id.lap_state == IB_CM_MRA_LAP_RCVD)
 		ib_cancel_mad(cm_id_priv->msg);
 
-	msg = cm_alloc_priv_msg(cm_id_priv);
+	msg = cm_alloc_priv_msg(cm_id_priv, IB_CM_DREQ_SENT);
 	if (IS_ERR(msg)) {
 		cm_enter_timewait(cm_id_priv);
-		return PTR_ERR(msg);
+		ret = PTR_ERR(msg);
+		goto unlock;
 	}
 
 	cm_format_dreq((struct cm_dreq_msg *) msg->mad, cm_id_priv,
 		       private_data, private_data_len);
-	msg->timeout_ms = cm_id_priv->timeout_ms;
-	msg->context[1] = (void *) (unsigned long) IB_CM_DREQ_SENT;
 
 	trace_icm_send_dreq(&cm_id_priv->id);
 	ret = ib_post_send_mad(msg, NULL);
 	if (ret) {
 		cm_enter_timewait(cm_id_priv);
 		cm_free_priv_msg(msg);
-		return ret;
+		goto unlock;
 	}
 
 	cm_id_priv->id.state = IB_CM_DREQ_SENT;
-	return 0;
-}
-
-int ib_send_cm_dreq(struct ib_cm_id *cm_id, const void *private_data,
-		    u8 private_data_len)
-{
-	struct cm_id_private *cm_id_priv =
-		container_of(cm_id, struct cm_id_private, id);
-	unsigned long flags;
-	int ret;
-
-	spin_lock_irqsave(&cm_id_priv->lock, flags);
-	ret = cm_send_dreq_locked(cm_id_priv, private_data, private_data_len);
+unlock:
 	spin_unlock_irqrestore(&cm_id_priv->lock, flags);
 	return ret;
 }
@@ -2770,7 +2799,7 @@ static int cm_issue_drep(struct cm_port *port,
 	struct cm_drep_msg *drep_msg;
 	int ret;
 
-	ret = cm_alloc_response_msg(port, mad_recv_wc, &msg);
+	ret = cm_alloc_response_msg(port, mad_recv_wc, true, &msg);
 	if (ret)
 		return ret;
 
@@ -2788,7 +2817,7 @@ static int cm_issue_drep(struct cm_port *port,
 		IBA_GET(CM_DREQ_REMOTE_COMM_ID, dreq_msg));
 	ret = ib_post_send_mad(msg, NULL);
 	if (ret)
-		cm_free_response_msg(msg);
+		cm_free_msg(msg);
 
 	return ret;
 }
@@ -2835,7 +2864,8 @@ static int cm_dreq_handler(struct cm_work *work)
 	case IB_CM_TIMEWAIT:
 		atomic_long_inc(&work->port->counters[CM_RECV_DUPLICATES]
 						     [CM_DREQ_COUNTER]);
-		msg = cm_alloc_response_msg_no_ah(work->port, work->mad_recv_wc);
+		msg = cm_alloc_response_msg_no_ah(work->port, work->mad_recv_wc,
+						  true);
 		if (IS_ERR(msg))
 			goto unlock;
 
@@ -2846,7 +2876,7 @@ static int cm_dreq_handler(struct cm_work *work)
 
 		if (cm_create_response_msg_ah(work->port, work->mad_recv_wc, msg) ||
 		    ib_post_send_mad(msg, NULL))
-			cm_free_response_msg(msg);
+			cm_free_msg(msg);
 		goto deref;
 	case IB_CM_DREQ_RCVD:
 		atomic_long_inc(&work->port->counters[CM_RECV_DUPLICATES]
@@ -3340,7 +3370,8 @@ static int cm_lap_handler(struct cm_work *work)
 	case IB_CM_MRA_LAP_SENT:
 		atomic_long_inc(&work->port->counters[CM_RECV_DUPLICATES]
 						     [CM_LAP_COUNTER]);
-		msg = cm_alloc_response_msg_no_ah(work->port, work->mad_recv_wc);
+		msg = cm_alloc_response_msg_no_ah(work->port, work->mad_recv_wc,
+						  true);
 		if (IS_ERR(msg))
 			goto unlock;
 
@@ -3353,7 +3384,7 @@ static int cm_lap_handler(struct cm_work *work)
 
 		if (cm_create_response_msg_ah(work->port, work->mad_recv_wc, msg) ||
 		    ib_post_send_mad(msg, NULL))
-			cm_free_response_msg(msg);
+			cm_free_msg(msg);
 		goto deref;
 	case IB_CM_LAP_RCVD:
 		atomic_long_inc(&work->port->counters[CM_RECV_DUPLICATES]
@@ -3492,7 +3523,7 @@ int ib_send_cm_sidr_req(struct ib_cm_id *cm_id,
 		goto out_unlock;
 	}
 
-	msg = cm_alloc_priv_msg(cm_id_priv);
+	msg = cm_alloc_priv_msg(cm_id_priv, IB_CM_SIDR_REQ_SENT);
 	if (IS_ERR(msg)) {
 		ret = PTR_ERR(msg);
 		goto out_unlock;
@@ -3500,8 +3531,6 @@ int ib_send_cm_sidr_req(struct ib_cm_id *cm_id,
 
 	cm_format_sidr_req((struct cm_sidr_req_msg *)msg->mad, cm_id_priv,
 			   param);
-	msg->timeout_ms = cm_id_priv->timeout_ms;
-	msg->context[1] = (void *)(unsigned long)IB_CM_SIDR_REQ_SENT;
 
 	trace_icm_send_sidr_req(&cm_id_priv->id);
 	ret = ib_post_send_mad(msg, NULL);
@@ -3747,17 +3776,17 @@ out:
 
 static void cm_process_send_error(struct cm_id_private *cm_id_priv,
 				  struct ib_mad_send_buf *msg,
-				  enum ib_cm_state state,
 				  enum ib_wc_status wc_status)
 {
+	enum ib_cm_state state = (unsigned long) msg->context[1];
 	struct ib_cm_event cm_event = {};
 	int ret;
 
-	/* Discard old sends or ones without a response. */
+	/* Discard old sends. */
 	spin_lock_irq(&cm_id_priv->lock);
 	if (msg != cm_id_priv->msg) {
 		spin_unlock_irq(&cm_id_priv->lock);
-		cm_free_msg(msg);
+		cm_free_priv_msg(msg);
 		return;
 	}
 	cm_free_priv_msg(msg);
@@ -3805,9 +3834,7 @@ static void cm_send_handler(struct ib_mad_agent *mad_agent,
 			    struct ib_mad_send_wc *mad_send_wc)
 {
 	struct ib_mad_send_buf *msg = mad_send_wc->send_buf;
-	struct cm_id_private *cm_id_priv = msg->context[0];
-	enum ib_cm_state state =
-		(enum ib_cm_state)(unsigned long)msg->context[1];
+	struct cm_id_private *cm_id_priv;
 	struct cm_port *port;
 	u16 attr_index;
 
@@ -3815,13 +3842,12 @@ static void cm_send_handler(struct ib_mad_agent *mad_agent,
 	attr_index = be16_to_cpu(((struct ib_mad_hdr *)
 				  msg->mad)->attr_id) - CM_ATTR_ID_OFFSET;
 
-	/*
-	 * If the send was in response to a received message (context[0] is not
-	 * set to a cm_id), and is not a REJ, then it is a send that was
-	 * manually retried.
-	 */
-	if (!cm_id_priv && (attr_index != CM_REJ_COUNTER))
+	if (msg->context[0] == CM_DIRECT_RETRY_CTX) {
 		msg->retries = 1;
+		cm_id_priv = NULL;
+	} else {
+		cm_id_priv = msg->context[0];
+	}
 
 	atomic_long_add(1 + msg->retries, &port->counters[CM_XMIT][attr_index]);
 	if (msg->retries)
@@ -3829,10 +3855,9 @@ static void cm_send_handler(struct ib_mad_agent *mad_agent,
 				&port->counters[CM_XMIT_RETRIES][attr_index]);
 
 	if (cm_id_priv)
-		cm_process_send_error(cm_id_priv, msg, state,
-				      mad_send_wc->status);
+		cm_process_send_error(cm_id_priv, msg, mad_send_wc->status);
 	else
-		cm_free_response_msg(msg);
+		cm_free_msg(msg);
 }
 
 static void cm_work_handler(struct work_struct *_work)