[IB] CM: bind IDs to a specific device

Bind communication identifiers to a device to support device removal.
Export per HCA CM devices to userspace.
Signed-off-by: Sean Hefty <sean.hefty@intel.com>

drivers/infiniband/core/cm.c

@@ -366,9 +366,15 @@ static struct cm_id_private * cm_insert_listen(struct cm_id_private *cm_id_priv)
 		cur_cm_id_priv = rb_entry(parent, struct cm_id_private,
 					  service_node);
 		if ((cur_cm_id_priv->id.service_mask & service_id) ==
-		    (service_mask & cur_cm_id_priv->id.service_id))
-			return cm_id_priv;
-		if (service_id < cur_cm_id_priv->id.service_id)
+		    (service_mask & cur_cm_id_priv->id.service_id) &&
+		    (cm_id_priv->id.device == cur_cm_id_priv->id.device))
+			return cur_cm_id_priv;
+
+		if (cm_id_priv->id.device < cur_cm_id_priv->id.device)
+			link = &(*link)->rb_left;
+		else if (cm_id_priv->id.device > cur_cm_id_priv->id.device)
+			link = &(*link)->rb_right;
+		else if (service_id < cur_cm_id_priv->id.service_id)
 			link = &(*link)->rb_left;
 		else
 			link = &(*link)->rb_right;
@@ -378,7 +384,8 @@ static struct cm_id_private * cm_insert_listen(struct cm_id_private *cm_id_priv)
 	return NULL;
 }
 
-static struct cm_id_private * cm_find_listen(__be64 service_id)
+static struct cm_id_private * cm_find_listen(struct ib_device *device,
+					     __be64 service_id)
 {
 	struct rb_node *node = cm.listen_service_table.rb_node;
 	struct cm_id_private *cm_id_priv;
@@ -386,9 +393,15 @@ static struct cm_id_private * cm_find_listen(__be64 service_id)
 	while (node) {
 		cm_id_priv = rb_entry(node, struct cm_id_private, service_node);
 		if ((cm_id_priv->id.service_mask & service_id) ==
-		    (cm_id_priv->id.service_mask & cm_id_priv->id.service_id))
+		     cm_id_priv->id.service_id &&
+		    (cm_id_priv->id.device == device))
 			return cm_id_priv;
-		if (service_id < cm_id_priv->id.service_id)
+
+		if (device < cm_id_priv->id.device)
+			node = node->rb_left;
+		else if (device > cm_id_priv->id.device)
+			node = node->rb_right;
+		else if (service_id < cm_id_priv->id.service_id)
 			node = node->rb_left;
 		else
 			node = node->rb_right;
@@ -523,7 +536,8 @@ static void cm_reject_sidr_req(struct cm_id_private *cm_id_priv,
 	ib_send_cm_sidr_rep(&cm_id_priv->id, &param);
 }
 
-struct ib_cm_id *ib_create_cm_id(ib_cm_handler cm_handler,
+struct ib_cm_id *ib_create_cm_id(struct ib_device *device,
+				 ib_cm_handler cm_handler,
 				 void *context)
 {
 	struct cm_id_private *cm_id_priv;
@@ -535,6 +549,7 @@ struct ib_cm_id *ib_create_cm_id(ib_cm_handler cm_handler,
 
 	memset(cm_id_priv, 0, sizeof *cm_id_priv);
 	cm_id_priv->id.state = IB_CM_IDLE;
+	cm_id_priv->id.device = device;
 	cm_id_priv->id.cm_handler = cm_handler;
 	cm_id_priv->id.context = context;
 	cm_id_priv->id.remote_cm_qpn = 1;
@@ -1047,7 +1062,6 @@ static void cm_format_req_event(struct cm_work *work,
 	req_msg = (struct cm_req_msg *)work->mad_recv_wc->recv_buf.mad;
 	param = &work->cm_event.param.req_rcvd;
 	param->listen_id = listen_id;
-	param->device = cm_id_priv->av.port->mad_agent->device;
 	param->port = cm_id_priv->av.port->port_num;
 	param->primary_path = &work->path[0];
 	if (req_msg->alt_local_lid)
@@ -1226,7 +1240,8 @@ static struct cm_id_private * cm_match_req(struct cm_work *work,
 	}
 
 	/* Find matching listen request. */
-	listen_cm_id_priv = cm_find_listen(req_msg->service_id);
+	listen_cm_id_priv = cm_find_listen(cm_id_priv->id.device,
+					   req_msg->service_id);
 	if (!listen_cm_id_priv) {
 		spin_unlock_irqrestore(&cm.lock, flags);
 		cm_issue_rej(work->port, work->mad_recv_wc,
@@ -1254,7 +1269,7 @@ static int cm_req_handler(struct cm_work *work)
 
 	req_msg = (struct cm_req_msg *)work->mad_recv_wc->recv_buf.mad;
 
-	cm_id = ib_create_cm_id(NULL, NULL);
+	cm_id = ib_create_cm_id(work->port->cm_dev->device, NULL, NULL);
 	if (IS_ERR(cm_id))
 		return PTR_ERR(cm_id);
 
@@ -2629,7 +2644,6 @@ static void cm_format_sidr_req_event(struct cm_work *work,
 	param = &work->cm_event.param.sidr_req_rcvd;
 	param->pkey = __be16_to_cpu(sidr_req_msg->pkey);
 	param->listen_id = listen_id;
-	param->device = work->port->mad_agent->device;
 	param->port = work->port->port_num;
 	work->cm_event.private_data = &sidr_req_msg->private_data;
 }
@@ -2642,7 +2656,7 @@ static int cm_sidr_req_handler(struct cm_work *work)
 	struct ib_wc *wc;
 	unsigned long flags;
 
-	cm_id = ib_create_cm_id(NULL, NULL);
+	cm_id = ib_create_cm_id(work->port->cm_dev->device, NULL, NULL);
 	if (IS_ERR(cm_id))
 		return PTR_ERR(cm_id);
 	cm_id_priv = container_of(cm_id, struct cm_id_private, id);
@@ -2666,7 +2680,8 @@ static int cm_sidr_req_handler(struct cm_work *work)
 		spin_unlock_irqrestore(&cm.lock, flags);
 		goto out; /* Duplicate message. */
 	}
-	cur_cm_id_priv = cm_find_listen(sidr_req_msg->service_id);
+	cur_cm_id_priv = cm_find_listen(cm_id->device,
+					sidr_req_msg->service_id);
 	if (!cur_cm_id_priv) {
 		rb_erase(&cm_id_priv->sidr_id_node, &cm.remote_sidr_table);
 		spin_unlock_irqrestore(&cm.lock, flags);

drivers/infiniband/core/ucm.c

@@ -52,12 +52,20 @@ MODULE_AUTHOR("Libor Michalek");
 MODULE_DESCRIPTION("InfiniBand userspace Connection Manager access");
 MODULE_LICENSE("Dual BSD/GPL");
 
+struct ib_ucm_device {
+	int			devnum;
+	struct cdev		dev;
+	struct class_device	class_dev;
+	struct ib_device	*ib_dev;
+};
+
 struct ib_ucm_file {
 	struct semaphore mutex;
 	struct file *filp;
+	struct ib_ucm_device *device;
 
-	struct list_head  ctxs;   /* list of active connections */
-	struct list_head  events; /* list of pending events */
+	struct list_head  ctxs;
+	struct list_head  events;
 	wait_queue_head_t poll_wait;
 };
 
@@ -90,14 +98,24 @@ struct ib_ucm_event {
 
 enum {
 	IB_UCM_MAJOR = 231,
-	IB_UCM_MINOR = 255
+	IB_UCM_BASE_MINOR = 224,
+	IB_UCM_MAX_DEVICES = 32
 };
 
-#define IB_UCM_DEV MKDEV(IB_UCM_MAJOR, IB_UCM_MINOR)
+#define IB_UCM_BASE_DEV MKDEV(IB_UCM_MAJOR, IB_UCM_BASE_MINOR)
 
-static struct semaphore ctx_id_mutex;
-static struct idr       ctx_id_table;
+static void ib_ucm_add_one(struct ib_device *device);
+static void ib_ucm_remove_one(struct ib_device *device);
 
+static struct ib_client ucm_client = {
+	.name   = "ucm",
+	.add    = ib_ucm_add_one,
+	.remove = ib_ucm_remove_one
+};
+
+DECLARE_MUTEX(ctx_id_mutex);
+DEFINE_IDR(ctx_id_table);
+static DECLARE_BITMAP(dev_map, IB_UCM_MAX_DEVICES);
 
 static struct ib_ucm_context *ib_ucm_ctx_get(struct ib_ucm_file *file, int id)
 {
@@ -184,10 +202,7 @@ error:
 	kfree(ctx);
 	return NULL;
 }
-/*
- * Event portion of the API, handle CM events
- * and allow event polling.
- */
+
 static void ib_ucm_event_path_get(struct ib_ucm_path_rec *upath,
 				  struct ib_sa_path_rec	 *kpath)
 {
@@ -234,6 +249,7 @@ static void ib_ucm_event_req_get(struct ib_ucm_req_event_resp *ureq,
 	ureq->retry_count                = kreq->retry_count;
 	ureq->rnr_retry_count            = kreq->rnr_retry_count;
 	ureq->srq                        = kreq->srq;
+	ureq->port			 = kreq->port;
 
 	ib_ucm_event_path_get(&ureq->primary_path, kreq->primary_path);
 	ib_ucm_event_path_get(&ureq->alternate_path, kreq->alternate_path);
@@ -320,6 +336,8 @@ static int ib_ucm_event_process(struct ib_cm_event *evt,
 	case IB_CM_SIDR_REQ_RECEIVED:
 		uvt->resp.u.sidr_req_resp.pkey = 
 					evt->param.sidr_req_rcvd.pkey;
+		uvt->resp.u.sidr_req_resp.port = 
+					evt->param.sidr_req_rcvd.port;
 		uvt->data_len = IB_CM_SIDR_REQ_PRIVATE_DATA_SIZE;
 		break;
 	case IB_CM_SIDR_REP_RECEIVED:
@@ -412,9 +430,7 @@ static ssize_t ib_ucm_event(struct ib_ucm_file *file,
 
 	if (copy_from_user(&cmd, inbuf, sizeof(cmd)))
 		return -EFAULT;
-	/*
-	 * wait
-	 */
+
 	down(&file->mutex);
 	while (list_empty(&file->events)) {
 
@@ -496,7 +512,6 @@ done:
 	return result;
 }
 
-
 static ssize_t ib_ucm_create_id(struct ib_ucm_file *file,
 				const char __user *inbuf,
 				int in_len, int out_len)
@@ -519,29 +534,27 @@ static ssize_t ib_ucm_create_id(struct ib_ucm_file *file,
 		return -ENOMEM;
 
 	ctx->uid = cmd.uid;
-	ctx->cm_id = ib_create_cm_id(ib_ucm_event_handler, ctx);
+	ctx->cm_id = ib_create_cm_id(file->device->ib_dev,
+				     ib_ucm_event_handler, ctx);
 	if (IS_ERR(ctx->cm_id)) {
 		result = PTR_ERR(ctx->cm_id);
-		goto err;
+		goto err1;
 	}
 
 	resp.id = ctx->id;
 	if (copy_to_user((void __user *)(unsigned long)cmd.response,
 			 &resp, sizeof(resp))) {
 		result = -EFAULT;
-		goto err;
+		goto err2;
 	}
-
 	return 0;
 
-err:
+err2:
+	ib_destroy_cm_id(ctx->cm_id);
+err1:
 	down(&ctx_id_mutex);
 	idr_remove(&ctx_id_table, ctx->id);
 	up(&ctx_id_mutex);
-
-	if (!IS_ERR(ctx->cm_id))
-		ib_destroy_cm_id(ctx->cm_id);
-
 	kfree(ctx);
 	return result;
 }
@@ -1253,6 +1266,7 @@ static int ib_ucm_open(struct inode *inode, struct file *filp)
 
 	filp->private_data = file;
 	file->filp = filp;
+	file->device = container_of(inode->i_cdev, struct ib_ucm_device, dev);
 
 	return 0;
 }
@@ -1283,7 +1297,17 @@ static int ib_ucm_close(struct inode *inode, struct file *filp)
 	return 0;
 }
 
-static struct file_operations ib_ucm_fops = {
+static void ib_ucm_release_class_dev(struct class_device *class_dev)
+{
+	struct ib_ucm_device *dev;
+
+	dev = container_of(class_dev, struct ib_ucm_device, class_dev);
+	cdev_del(&dev->dev);
+	clear_bit(dev->devnum, dev_map);
+	kfree(dev);
+}
+
+static struct file_operations ucm_fops = {
 	.owner 	 = THIS_MODULE,
 	.open 	 = ib_ucm_open,
 	.release = ib_ucm_close,
@@ -1291,55 +1315,141 @@ static struct file_operations ib_ucm_fops = {
 	.poll    = ib_ucm_poll,
 };
 
+static struct class ucm_class = {
+	.name    = "infiniband_cm",
+	.release = ib_ucm_release_class_dev
+};
 
-static struct class *ib_ucm_class;
-static struct cdev	  ib_ucm_cdev;
+static ssize_t show_dev(struct class_device *class_dev, char *buf)
+{
+	struct ib_ucm_device *dev;
 	
-static int __init ib_ucm_init(void)
+	dev = container_of(class_dev, struct ib_ucm_device, class_dev);
+	return print_dev_t(buf, dev->dev.dev);
+}
+static CLASS_DEVICE_ATTR(dev, S_IRUGO, show_dev, NULL);
+
+static ssize_t show_ibdev(struct class_device *class_dev, char *buf)
 {
-	int result;
+	struct ib_ucm_device *dev;
 	
-	result = register_chrdev_region(IB_UCM_DEV, 1, "infiniband_cm");
-	if (result) {
-		printk(KERN_ERR "ucm: Error <%d> registering dev\n", result);
-		goto err_chr;
-	}
+	dev = container_of(class_dev, struct ib_ucm_device, class_dev);
+	return sprintf(buf, "%s\n", dev->ib_dev->name);
+}
+static CLASS_DEVICE_ATTR(ibdev, S_IRUGO, show_ibdev, NULL);
+
+static void ib_ucm_add_one(struct ib_device *device)
+{
+	struct ib_ucm_device *ucm_dev;
+
+	if (!device->alloc_ucontext)
+		return;
+
+	ucm_dev = kmalloc(sizeof *ucm_dev, GFP_KERNEL);
+	if (!ucm_dev)
+		return;
+
+	memset(ucm_dev, 0, sizeof *ucm_dev);
+	ucm_dev->ib_dev = device;
+
+	ucm_dev->devnum = find_first_zero_bit(dev_map, IB_UCM_MAX_DEVICES);
+	if (ucm_dev->devnum >= IB_UCM_MAX_DEVICES)
+		goto err;
+
+	set_bit(ucm_dev->devnum, dev_map);
 
-	cdev_init(&ib_ucm_cdev, &ib_ucm_fops);
+	cdev_init(&ucm_dev->dev, &ucm_fops);
+	ucm_dev->dev.owner = THIS_MODULE;
+	kobject_set_name(&ucm_dev->dev.kobj, "ucm%d", ucm_dev->devnum);
+	if (cdev_add(&ucm_dev->dev, IB_UCM_BASE_DEV + ucm_dev->devnum, 1))
+		goto err;
 
-	result = cdev_add(&ib_ucm_cdev, IB_UCM_DEV, 1);
-	if (result) {
-		printk(KERN_ERR "ucm: Error <%d> adding cdev\n", result);
+	ucm_dev->class_dev.class = &ucm_class;
+	ucm_dev->class_dev.dev = device->dma_device;
+	snprintf(ucm_dev->class_dev.class_id, BUS_ID_SIZE, "ucm%d",
+		 ucm_dev->devnum);
+	if (class_device_register(&ucm_dev->class_dev))
 		goto err_cdev;
-	}
 
-	ib_ucm_class = class_create(THIS_MODULE, "infiniband_cm");
-	if (IS_ERR(ib_ucm_class)) {
-		result = PTR_ERR(ib_ucm_class);
-		printk(KERN_ERR "Error <%d> creating class\n", result);
+	if (class_device_create_file(&ucm_dev->class_dev,
+				     &class_device_attr_dev))
+		goto err_class;
+	if (class_device_create_file(&ucm_dev->class_dev,
+				     &class_device_attr_ibdev))
 		goto err_class;
+
+	ib_set_client_data(device, &ucm_client, ucm_dev);
+	return;
+
+err_class:
+	class_device_unregister(&ucm_dev->class_dev);
+err_cdev:
+	cdev_del(&ucm_dev->dev);
+	clear_bit(ucm_dev->devnum, dev_map);
+err:
+	kfree(ucm_dev);
+	return;
+}
+
+static void ib_ucm_remove_one(struct ib_device *device)
+{
+	struct ib_ucm_device *ucm_dev = ib_get_client_data(device, &ucm_client);
+
+	if (!ucm_dev)
+		return;
+
+	class_device_unregister(&ucm_dev->class_dev);
+}
+
+static ssize_t show_abi_version(struct class *class, char *buf)
+{
+	return sprintf(buf, "%d\n", IB_USER_CM_ABI_VERSION);
+}
+static CLASS_ATTR(abi_version, S_IRUGO, show_abi_version, NULL);
+
+static int __init ib_ucm_init(void)
+{
+	int ret;
+
+	ret = register_chrdev_region(IB_UCM_BASE_DEV, IB_UCM_MAX_DEVICES,
+				     "infiniband_cm");
+	if (ret) {
+		printk(KERN_ERR "ucm: couldn't register device number\n");
+		goto err;
 	}
 
-	class_device_create(ib_ucm_class, IB_UCM_DEV, NULL, "ucm");
+	ret = class_register(&ucm_class);
+	if (ret) {
+		printk(KERN_ERR "ucm: couldn't create class infiniband_cm\n");
+		goto err_chrdev;
+	}
 
-	idr_init(&ctx_id_table);
-	init_MUTEX(&ctx_id_mutex);
+	ret = class_create_file(&ucm_class, &class_attr_abi_version);
+	if (ret) {
+		printk(KERN_ERR "ucm: couldn't create abi_version attribute\n");
+		goto err_class;
+	}
 
+	ret = ib_register_client(&ucm_client);
+	if (ret) {
+		printk(KERN_ERR "ucm: couldn't register client\n");
+		goto err_class;
+	}
 	return 0;
+
 err_class:
-	cdev_del(&ib_ucm_cdev);
-err_cdev:
-	unregister_chrdev_region(IB_UCM_DEV, 1);
-err_chr:
-	return result;
+	class_unregister(&ucm_class);
+err_chrdev:
+	unregister_chrdev_region(IB_UCM_BASE_DEV, IB_UCM_MAX_DEVICES);
+err:
+	return ret;
 }
 
 static void __exit ib_ucm_cleanup(void)
 {
-	class_device_destroy(ib_ucm_class, IB_UCM_DEV);
-	class_destroy(ib_ucm_class);
-	cdev_del(&ib_ucm_cdev);
-	unregister_chrdev_region(IB_UCM_DEV, 1);
+	ib_unregister_client(&ucm_client);
+	class_unregister(&ucm_class);
+	unregister_chrdev_region(IB_UCM_BASE_DEV, IB_UCM_MAX_DEVICES);
 }
 
 module_init(ib_ucm_init);

include/rdma/ib_cm.h

 /*
- * Copyright (c) 2004 Intel Corporation.  All rights reserved.
+ * Copyright (c) 2004, 2005 Intel Corporation.  All rights reserved.
  * Copyright (c) 2004 Topspin Corporation.  All rights reserved.
  * Copyright (c) 2004 Voltaire Corporation.  All rights reserved.
  * Copyright (c) 2005 Sun Microsystems, Inc. All rights reserved.
@@ -109,7 +109,6 @@ struct ib_cm_id;
 
 struct ib_cm_req_event_param {
 	struct ib_cm_id		*listen_id;
-	struct ib_device	*device;
 	u8			port;
 
 	struct ib_sa_path_rec	*primary_path;
@@ -220,7 +219,6 @@ struct ib_cm_apr_event_param {
 
 struct ib_cm_sidr_req_event_param {
 	struct ib_cm_id		*listen_id;
-	struct ib_device	*device;
 	u8			port;
 	u16			pkey;
 };
@@ -284,6 +282,7 @@ typedef int (*ib_cm_handler)(struct ib_cm_id *cm_id,
 struct ib_cm_id {
 	ib_cm_handler		cm_handler;
 	void			*context;
+	struct ib_device	*device;
 	__be64			service_id;
 	__be64			service_mask;
 	enum ib_cm_state	state;		/* internal CM/debug use */
@@ -295,6 +294,8 @@ struct ib_cm_id {
 
 /**
  * ib_create_cm_id - Allocate a communication identifier.
+ * @device: Device associated with the cm_id.  All related communication will
+ * be associated with the specified device.
  * @cm_handler: Callback invoked to notify the user of CM events.
  * @context: User specified context associated with the communication
  *   identifier.
@@ -302,7 +303,8 @@ struct ib_cm_id {
  * Communication identifiers are used to track connection states, service
  * ID resolution requests, and listen requests.
  */
-struct ib_cm_id *ib_create_cm_id(ib_cm_handler cm_handler,
+struct ib_cm_id *ib_create_cm_id(struct ib_device *device,
+				 ib_cm_handler cm_handler,
 				 void *context);
 
 /**

include/rdma/ib_user_cm.h

@@ -38,7 +38,7 @@
 
 #include <linux/types.h>
 
-#define IB_USER_CM_ABI_VERSION 2
+#define IB_USER_CM_ABI_VERSION 3
 
 enum {
 	IB_USER_CM_CMD_CREATE_ID,
@@ -299,8 +299,6 @@ struct ib_ucm_event_get {
 };
 
 struct ib_ucm_req_event_resp {
-	/* device */
-	/* port */
 	struct ib_ucm_path_rec primary_path;
 	struct ib_ucm_path_rec alternate_path;
 	__be64                 remote_ca_guid;
@@ -316,6 +314,7 @@ struct ib_ucm_req_event_resp {
 	__u8  retry_count;
 	__u8  rnr_retry_count;
 	__u8  srq;
+	__u8  port;
 };
 
 struct ib_ucm_rep_event_resp {
@@ -353,10 +352,9 @@ struct ib_ucm_apr_event_resp {
 };
 
 struct ib_ucm_sidr_req_event_resp {
-	/* device */
-	/* port */
 	__u16 pkey;
-	__u8  reserved[2];
+	__u8  port;
+	__u8  reserved;
 };
 
 struct ib_ucm_sidr_rep_event_resp {