[ofa-general] CM sysfs-related oops on device driver reload

Roland Dreier rdreier at cisco.com
Fri Feb 8 20:48:13 PST 2008


OK, I think the following should make things work better:

diff --git a/drivers/infiniband/core/cm.c b/drivers/infiniband/core/cm.c
index 638b727..017fdcf 100644
--- a/drivers/infiniband/core/cm.c
+++ b/drivers/infiniband/core/cm.c
@@ -159,7 +159,7 @@ struct cm_port {
 	struct ib_mad_agent *mad_agent;
 	struct kobject port_obj;
 	u8 port_num;
-	struct cm_counter_group counter_group[CM_COUNTER_GROUPS];
+	struct cm_counter_group *counter_group[CM_COUNTER_GROUPS];
 };
 
 struct cm_device {
@@ -1363,7 +1363,7 @@ static void cm_dup_req_handler(struct cm_work *work,
 	struct ib_mad_send_buf *msg = NULL;
 	int ret;
 
-	atomic_long_inc(&work->port->counter_group[CM_RECV_DUPLICATES].
+	atomic_long_inc(&work->port->counter_group[CM_RECV_DUPLICATES]->
 			counter[CM_REQ_COUNTER]);
 
 	/* Quick state check to discard duplicate REQs. */
@@ -1741,7 +1741,7 @@ static void cm_dup_rep_handler(struct cm_work *work)
 	if (!cm_id_priv)
 		return;
 
-	atomic_long_inc(&work->port->counter_group[CM_RECV_DUPLICATES].
+	atomic_long_inc(&work->port->counter_group[CM_RECV_DUPLICATES]->
 			counter[CM_REP_COUNTER]);
 	ret = cm_alloc_response_msg(work->port, work->mad_recv_wc, &msg);
 	if (ret)
@@ -1908,7 +1908,7 @@ static int cm_rtu_handler(struct cm_work *work)
 	if (cm_id_priv->id.state != IB_CM_REP_SENT &&
 	    cm_id_priv->id.state != IB_CM_MRA_REP_RCVD) {
 		spin_unlock_irq(&cm_id_priv->lock);
-		atomic_long_inc(&work->port->counter_group[CM_RECV_DUPLICATES].
+		atomic_long_inc(&work->port->counter_group[CM_RECV_DUPLICATES]->
 				counter[CM_RTU_COUNTER]);
 		goto out;
 	}
@@ -2087,7 +2087,7 @@ static int cm_dreq_handler(struct cm_work *work)
 	cm_id_priv = cm_acquire_id(dreq_msg->remote_comm_id,
 				   dreq_msg->local_comm_id);
 	if (!cm_id_priv) {
-		atomic_long_inc(&work->port->counter_group[CM_RECV_DUPLICATES].
+		atomic_long_inc(&work->port->counter_group[CM_RECV_DUPLICATES]->
 				counter[CM_DREQ_COUNTER]);
 		cm_issue_drep(work->port, work->mad_recv_wc);
 		return -EINVAL;
@@ -2108,7 +2108,7 @@ static int cm_dreq_handler(struct cm_work *work)
 	case IB_CM_MRA_REP_RCVD:
 		break;
 	case IB_CM_TIMEWAIT:
-		atomic_long_inc(&work->port->counter_group[CM_RECV_DUPLICATES].
+		atomic_long_inc(&work->port->counter_group[CM_RECV_DUPLICATES]->
 				counter[CM_DREQ_COUNTER]);
 		if (cm_alloc_response_msg(work->port, work->mad_recv_wc, &msg))
 			goto unlock;
@@ -2122,7 +2122,7 @@ static int cm_dreq_handler(struct cm_work *work)
 			cm_free_msg(msg);
 		goto deref;
 	case IB_CM_DREQ_RCVD:
-		atomic_long_inc(&work->port->counter_group[CM_RECV_DUPLICATES].
+		atomic_long_inc(&work->port->counter_group[CM_RECV_DUPLICATES]->
 				counter[CM_DREQ_COUNTER]);
 		goto unlock;
 	default:
@@ -2479,7 +2479,7 @@ static int cm_mra_handler(struct cm_work *work)
 				  cm_id_priv->msg, timeout)) {
 			if (cm_id_priv->id.lap_state == IB_CM_MRA_LAP_RCVD)
 				atomic_long_inc(&work->port->
-						counter_group[CM_RECV_DUPLICATES].
+						counter_group[CM_RECV_DUPLICATES]->
 						counter[CM_MRA_COUNTER]);
 			goto out;
 		}
@@ -2487,7 +2487,7 @@ static int cm_mra_handler(struct cm_work *work)
 		break;
 	case IB_CM_MRA_REQ_RCVD:
 	case IB_CM_MRA_REP_RCVD:
-		atomic_long_inc(&work->port->counter_group[CM_RECV_DUPLICATES].
+		atomic_long_inc(&work->port->counter_group[CM_RECV_DUPLICATES]->
 				counter[CM_MRA_COUNTER]);
 		/* fall through */
 	default:
@@ -2649,7 +2649,7 @@ static int cm_lap_handler(struct cm_work *work)
 	case IB_CM_LAP_IDLE:
 		break;
 	case IB_CM_MRA_LAP_SENT:
-		atomic_long_inc(&work->port->counter_group[CM_RECV_DUPLICATES].
+		atomic_long_inc(&work->port->counter_group[CM_RECV_DUPLICATES]->
 				counter[CM_LAP_COUNTER]);
 		if (cm_alloc_response_msg(work->port, work->mad_recv_wc, &msg))
 			goto unlock;
@@ -2665,7 +2665,7 @@ static int cm_lap_handler(struct cm_work *work)
 			cm_free_msg(msg);
 		goto deref;
 	case IB_CM_LAP_RCVD:
-		atomic_long_inc(&work->port->counter_group[CM_RECV_DUPLICATES].
+		atomic_long_inc(&work->port->counter_group[CM_RECV_DUPLICATES]->
 				counter[CM_LAP_COUNTER]);
 		goto unlock;
 	default:
@@ -2949,7 +2949,7 @@ static int cm_sidr_req_handler(struct cm_work *work)
 	cur_cm_id_priv = cm_insert_remote_sidr(cm_id_priv);
 	if (cur_cm_id_priv) {
 		spin_unlock_irq(&cm.lock);
-		atomic_long_inc(&work->port->counter_group[CM_RECV_DUPLICATES].
+		atomic_long_inc(&work->port->counter_group[CM_RECV_DUPLICATES]->
 				counter[CM_SIDR_REQ_COUNTER]);
 		goto out; /* Duplicate message. */
 	}
@@ -3161,10 +3161,10 @@ static void cm_send_handler(struct ib_mad_agent *mad_agent,
 		msg->retries = 1;
 
 	atomic_long_add(1 + msg->retries,
-			&port->counter_group[CM_XMIT].counter[attr_index]);
+			&port->counter_group[CM_XMIT]->counter[attr_index]);
 	if (msg->retries)
 		atomic_long_add(msg->retries,
-				&port->counter_group[CM_XMIT_RETRIES].
+				&port->counter_group[CM_XMIT_RETRIES]->
 				counter[attr_index]);
 
 	switch (mad_send_wc->status) {
@@ -3373,7 +3373,7 @@ static void cm_recv_handler(struct ib_mad_agent *mad_agent,
 	}
 
 	attr_id = be16_to_cpu(mad_recv_wc->recv_buf.mad->mad_hdr.attr_id);
-	atomic_long_inc(&port->counter_group[CM_RECV].
+	atomic_long_inc(&port->counter_group[CM_RECV]->
 			counter[attr_id - CM_ATTR_ID_OFFSET]);
 
 	work = kmalloc(sizeof *work + sizeof(struct ib_sa_path_rec) * paths,
@@ -3561,6 +3561,14 @@ static void cm_get_ack_delay(struct cm_device *cm_dev)
 		cm_dev->ack_delay = attr.local_ca_ack_delay;
 }
 
+static void cm_release_counter_obj(struct kobject *obj)
+{
+	struct cm_counter_group *counter_group;
+
+	counter_group = container_of(obj, struct cm_counter_group, obj);
+	kfree(counter_group);
+}
+
 static ssize_t cm_show_counter(struct kobject *obj, struct attribute *attr,
 			       char *buf)
 {
@@ -3579,6 +3587,7 @@ static struct sysfs_ops cm_counter_ops = {
 };
 
 static struct kobj_type cm_counter_obj_type = {
+	.release = cm_release_counter_obj,
 	.sysfs_ops = &cm_counter_ops,
 	.default_attrs = cm_counter_default_attrs
 };
@@ -3587,8 +3596,6 @@ static void cm_release_port_obj(struct kobject *obj)
 {
 	struct cm_port *cm_port;
 
-	printk(KERN_ERR "free cm port\n");
-
 	cm_port = container_of(obj, struct cm_port, port_obj);
 	kfree(cm_port);
 }
@@ -3601,8 +3608,6 @@ static void cm_release_dev_obj(struct kobject *obj)
 {
 	struct cm_device *cm_dev;
 
-	printk(KERN_ERR "free cm dev\n");
-
 	cm_dev = container_of(obj, struct cm_device, dev_obj);
 	kfree(cm_dev);
 }
@@ -3616,18 +3621,12 @@ struct class cm_class = {
 };
 EXPORT_SYMBOL(cm_class);
 
-static void cm_remove_fs_obj(struct kobject *obj)
-{
-	kobject_put(obj->parent);
-	kobject_put(obj);
-}
-
 static int cm_create_port_fs(struct cm_port *port)
 {
 	int i, ret;
 
 	ret = kobject_init_and_add(&port->port_obj, &cm_port_obj_type,
-				   kobject_get(&port->cm_dev->dev_obj),
+				   &port->cm_dev->dev_obj,
 				   "%d", port->port_num);
 	if (ret) {
 		kfree(port);
@@ -3635,20 +3634,30 @@ static int cm_create_port_fs(struct cm_port *port)
 	}
 
 	for (i = 0; i < CM_COUNTER_GROUPS; i++) {
-		ret = kobject_init_and_add(&port->counter_group[i].obj,
+		port->counter_group[i] = kzalloc(sizeof *port->counter_group[i],
+						 GFP_KERNEL);
+		if (!port->counter_group[i]) {
+			ret = -ENOMEM;
+			goto error;
+		}
+
+		ret = kobject_init_and_add(&port->counter_group[i]->obj,
 					   &cm_counter_obj_type,
-					   kobject_get(&port->port_obj),
+					   &port->port_obj,
 					   "%s", counter_group_names[i]);
-		if (ret)
+		if (ret) {
+			kfree(port->counter_group[i]);
 			goto error;
+		}
 	}
 
 	return 0;
 
 error:
 	while (i--)
-		cm_remove_fs_obj(&port->counter_group[i].obj);
-	cm_remove_fs_obj(&port->port_obj);
+		kobject_put(&port->counter_group[i]->obj);
+
+	kobject_put(&port->port_obj);
 	return ret;
 
 }
@@ -3658,9 +3667,9 @@ static void cm_remove_port_fs(struct cm_port *port)
 	int i;
 
 	for (i = 0; i < CM_COUNTER_GROUPS; i++)
-		cm_remove_fs_obj(&port->counter_group[i].obj);
+		kobject_put(&port->counter_group[i]->obj);
 
-	cm_remove_fs_obj(&port->port_obj);
+	kobject_put(&port->port_obj);
 }
 
 static void cm_add_one(struct ib_device *device)
@@ -3744,7 +3753,7 @@ error1:
 		ib_unregister_mad_agent(port->mad_agent);
 		cm_remove_port_fs(port);
 	}
-	cm_remove_fs_obj(&cm_dev->dev_obj);
+	kobject_put(&cm_dev->dev_obj);
 }
 
 static void cm_remove_one(struct ib_device *device)
@@ -3771,7 +3780,7 @@ static void cm_remove_one(struct ib_device *device)
 		ib_unregister_mad_agent(port->mad_agent);
 		cm_remove_port_fs(port);
 	}
-	cm_remove_fs_obj(&cm_dev->dev_obj);
+	kobject_put(&cm_dev->dev_obj);
 }
 
 static int __init ib_cm_init(void)



More information about the general mailing list