diff mbox series

[rdma-next,3/3] IB/cm: Rework sending DREQ when destroying a cm_id

Message ID a288a098b8e0550305755fd4a7937431699317f4.1731495873.git.leon@kernel.org (mailing list archive)
State New
Headers show
Series Batch of IBCM improvements | expand

Commit Message

Leon Romanovsky Nov. 13, 2024, 11:12 a.m. UTC 
From: Sean Hefty <shefty@nvidia.com>

A DREQ is sent in 2 situations:

  1. When requested by the user.
     This DREQ has to wait for a DREP, which will be routed to the user.

  2. When the cm_id is destroyed.
     This DREQ is generated by the CM to notify the peer that the
     connection has been destroyed.

In the latter case, any DREP that is received will be discarded.
There's no need to hold a reference on the cm_id.  Today, both
situations are covered by the same function: cm_send_dreq_locked().
When invoked in the cm_id destroy path, the cm_id reference would be
held until the DREQ completes, blocking the destruction.  Because it
could take several seconds to minutes before the DREQ receives a DREP,
the destroy call posts a send for the DREQ then immediately cancels the
MAD.  However, cancellation is not immediate in the MAD layer.  There
could still be a delay before the MAD layer returns the DREQ to the CM.
Moreover, the only guarantee is that the DREQ will be sent at most once.

Introduce a separate flow for sending a DREQ when destroying the cm_id.
The new flow will not hold a reference on the cm_id, allowing it to be
cleaned up immediately.  The cancellation trick is no longer needed.
The MAD layer will send the DREQ exactly once.

Signed-off-by: Sean Hefty <shefty@nvidia.com>
Signed-off-by: Or Har-Toov <ohartoov@nvidia.com>
Signed-off-by: Vlad Dumitrescu <vdumitrescu@nvidia.com>
Signed-off-by: Leon Romanovsky <leonro@nvidia.com>
---
 drivers/infiniband/core/cm.c | 53 ++++++++++++++++++++++--------------
 1 file changed, 32 insertions(+), 21 deletions(-)
diff mbox series

Patch

diff --git a/drivers/infiniband/core/cm.c b/drivers/infiniband/core/cm.c
index 2517bfebcfd5..142170473e75 100644
--- a/drivers/infiniband/core/cm.c
+++ b/drivers/infiniband/core/cm.c
@@ -95,8 +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,
@@ -1112,7 +1111,8 @@  static void cm_destroy_id(struct ib_cm_id *cm_id, int err)
 			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);
@@ -2652,20 +2652,42 @@  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 ||
@@ -2675,7 +2697,8 @@  static int cm_send_dreq_locked(struct cm_id_private *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,
@@ -2686,23 +2709,11 @@  static int cm_send_dreq_locked(struct cm_id_private *cm_id_priv,
 	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;
 }