1 files changed, 563 insertions, 324 deletions

diff --git a/drivers/infiniband/core/user_mad.c b/drivers/infiniband/core/user_mad.c
index c908de8db5a..1acb9910055 100644
--- a/drivers/infiniband/core/user_mad.c
+++ b/drivers/infiniband/core/user_mad.c
@@ -1,7 +1,8 @@
 /*
  * Copyright (c) 2004 Topspin Communications.  All rights reserved.
- * Copyright (c) 2005 Voltaire, Inc. All rights reserved. 
+ * Copyright (c) 2005 Voltaire, Inc. All rights reserved.
  * Copyright (c) 2005 Sun Microsystems, Inc. All rights reserved.
+ * Copyright (c) 2008 Cisco. All rights reserved.
  *
  * This software is available to you under a choice of one of two
  * licenses.  You may choose to be licensed under the terms of the GNU
@@ -30,8 +31,6 @@
  * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
  * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
  * SOFTWARE.
- *
- * $Id: user_mad.c 4010 2005-11-09 23:11:56Z roland $
  */
 
 #include <linux/module.h>
@@ -40,14 +39,16 @@
 #include <linux/err.h>
 #include <linux/fs.h>
 #include <linux/cdev.h>
-#include <linux/pci.h>
 #include <linux/dma-mapping.h>
 #include <linux/poll.h>
-#include <linux/rwsem.h>
+#include <linux/mutex.h>
 #include <linux/kref.h>
+#include <linux/compat.h>
+#include <linux/sched.h>
+#include <linux/semaphore.h>
+#include <linux/slab.h>
 
 #include <asm/uaccess.h>
-#include <asm/semaphore.h>
 
 #include <rdma/ib_mad.h>
 #include <rdma/ib_user_mad.h>
@@ -65,12 +66,9 @@ enum {
 };
 
 /*
- * Our lifetime rules for these structs are the following: each time a
- * device special file is opened, we look up the corresponding struct
- * ib_umad_port by minor in the umad_port[] table while holding the
- * port_lock.  If this lookup succeeds, we take a reference on the
- * ib_umad_port's struct ib_umad_device while still holding the
- * port_lock; if the lookup fails, we fail the open().  We drop these
+ * Our lifetime rules for these structs are the following:
+ * device special file is opened, we take a reference on the
+ * ib_umad_port's struct ib_umad_device. We drop these
  * references in the corresponding close().
  *
  * In addition to references coming from open character devices, there
@@ -78,23 +76,18 @@ enum {
  * module's reference taken when allocating the ib_umad_device in
  * ib_umad_add_one().
  *
- * When destroying an ib_umad_device, we clear all of its
- * ib_umad_ports from umad_port[] while holding port_lock before
- * dropping the module's reference to the ib_umad_device.  This is
- * always safe because any open() calls will either succeed and obtain
- * a reference before we clear the umad_port[] entries, or fail after
- * we clear the umad_port[] entries.
+ * When destroying an ib_umad_device, we drop the module's reference.
  */
 
 struct ib_umad_port {
-	struct cdev           *dev;
-	struct class_device   *class_dev;
+	struct cdev           cdev;
+	struct device	      *dev;
 
-	struct cdev           *sm_dev;
-	struct class_device   *sm_class_dev;
+	struct cdev           sm_cdev;
+	struct device	      *sm_dev;
 	struct semaphore       sm_sem;
 
-	struct rw_semaphore    mutex;
+	struct mutex	       file_mutex;
 	struct list_head       file_list;
 
 	struct ib_device      *ib_dev;
@@ -105,22 +98,27 @@ struct ib_umad_port {
 
 struct ib_umad_device {
 	int                  start_port, end_port;
-	struct kref          ref;
+	struct kobject       kobj;
 	struct ib_umad_port  port[0];
 };
 
 struct ib_umad_file {
+	struct mutex		mutex;
 	struct ib_umad_port    *port;
 	struct list_head	recv_list;
+	struct list_head	send_list;
 	struct list_head	port_list;
-	spinlock_t		recv_lock;
+	spinlock_t		send_lock;
 	wait_queue_head_t	recv_wait;
 	struct ib_mad_agent    *agent[IB_UMAD_MAX_AGENTS];
 	int			agents_dead;
+	u8			use_pkey_index;
+	u8			already_used;
 };
 
 struct ib_umad_packet {
 	struct ib_mad_send_buf *msg;
+	struct ib_mad_recv_wc  *recv_wc;
 	struct list_head   list;
 	int		   length;
 	struct ib_user_mad mad;
@@ -131,21 +129,30 @@ static struct class *umad_class;
 static const dev_t base_dev = MKDEV(IB_UMAD_MAJOR, IB_UMAD_MINOR_BASE);
 
 static DEFINE_SPINLOCK(port_lock);
-static struct ib_umad_port *umad_port[IB_UMAD_MAX_PORTS];
-static DECLARE_BITMAP(dev_map, IB_UMAD_MAX_PORTS * 2);
+static DECLARE_BITMAP(dev_map, IB_UMAD_MAX_PORTS);
 
 static void ib_umad_add_one(struct ib_device *device);
 static void ib_umad_remove_one(struct ib_device *device);
 
-static void ib_umad_release_dev(struct kref *ref)
+static void ib_umad_release_dev(struct kobject *kobj)
 {
 	struct ib_umad_device *dev =
-		container_of(ref, struct ib_umad_device, ref);
+		container_of(kobj, struct ib_umad_device, kobj);
 
 	kfree(dev);
 }
 
-/* caller must hold port->mutex at least for reading */
+static struct kobj_type ib_umad_dev_ktype = {
+	.release = ib_umad_release_dev,
+};
+
+static int hdr_size(struct ib_umad_file *file)
+{
+	return file->use_pkey_index ? sizeof (struct ib_user_mad_hdr) :
+		sizeof (struct ib_user_mad_hdr_old);
+}
+
+/* caller must hold file->mutex */
 static struct ib_mad_agent *__get_agent(struct ib_umad_file *file, int id)
 {
 	return file->agents_dead ? NULL : file->agent[id];
@@ -157,50 +164,47 @@ static int queue_packet(struct ib_umad_file *file,
 {
 	int ret = 1;
 
-	down_read(&file->port->mutex);
+	mutex_lock(&file->mutex);
 
 	for (packet->mad.hdr.id = 0;
 	     packet->mad.hdr.id < IB_UMAD_MAX_AGENTS;
 	     packet->mad.hdr.id++)
 		if (agent == __get_agent(file, packet->mad.hdr.id)) {
-			spin_lock_irq(&file->recv_lock);
 			list_add_tail(&packet->list, &file->recv_list);
-			spin_unlock_irq(&file->recv_lock);
 			wake_up_interruptible(&file->recv_wait);
 			ret = 0;
 			break;
 		}
 
-	up_read(&file->port->mutex);
+	mutex_unlock(&file->mutex);
 
 	return ret;
 }
 
+static void dequeue_send(struct ib_umad_file *file,
+			 struct ib_umad_packet *packet)
+{
+	spin_lock_irq(&file->send_lock);
+	list_del(&packet->list);
+	spin_unlock_irq(&file->send_lock);
+}
+
 static void send_handler(struct ib_mad_agent *agent,
 			 struct ib_mad_send_wc *send_wc)
 {
 	struct ib_umad_file *file = agent->context;
-	struct ib_umad_packet *timeout;
 	struct ib_umad_packet *packet = send_wc->send_buf->context[0];
 
+	dequeue_send(file, packet);
 	ib_destroy_ah(packet->msg->ah);
 	ib_free_send_mad(packet->msg);
 
 	if (send_wc->status == IB_WC_RESP_TIMEOUT_ERR) {
-		timeout = kzalloc(sizeof *timeout + IB_MGMT_MAD_HDR, GFP_KERNEL);
-		if (!timeout)
-			goto out;
-
-		timeout->length 	= IB_MGMT_MAD_HDR;
-		timeout->mad.hdr.id 	= packet->mad.hdr.id;
-		timeout->mad.hdr.status = ETIMEDOUT;
-		memcpy(timeout->mad.data, packet->mad.data,
-		       sizeof (struct ib_mad_hdr));
-
-		if (queue_packet(file, agent, timeout))
-			kfree(timeout);
+		packet->length = IB_MGMT_MAD_HDR;
+		packet->mad.hdr.status = ETIMEDOUT;
+		if (!queue_packet(file, agent, packet))
+			return;
 	}
-out:
 	kfree(packet);
 }
 
@@ -209,43 +213,118 @@ static void recv_handler(struct ib_mad_agent *agent,
 {
 	struct ib_umad_file *file = agent->context;
 	struct ib_umad_packet *packet;
-	int length;
 
 	if (mad_recv_wc->wc->status != IB_WC_SUCCESS)
-		goto out;
+		goto err1;
 
-	length = mad_recv_wc->mad_len;
-	packet = kzalloc(sizeof *packet + length, GFP_KERNEL);
+	packet = kzalloc(sizeof *packet, GFP_KERNEL);
 	if (!packet)
-		goto out;
-
-	packet->length = length;
-
-	ib_coalesce_recv_mad(mad_recv_wc, packet->mad.data);
-
-	packet->mad.hdr.status    = 0;
-	packet->mad.hdr.length    = length + sizeof (struct ib_user_mad);
-	packet->mad.hdr.qpn 	  = cpu_to_be32(mad_recv_wc->wc->src_qp);
-	packet->mad.hdr.lid 	  = cpu_to_be16(mad_recv_wc->wc->slid);
-	packet->mad.hdr.sl  	  = mad_recv_wc->wc->sl;
-	packet->mad.hdr.path_bits = mad_recv_wc->wc->dlid_path_bits;
+		goto err1;
+
+	packet->length = mad_recv_wc->mad_len;
+	packet->recv_wc = mad_recv_wc;
+
+	packet->mad.hdr.status	   = 0;
+	packet->mad.hdr.length	   = hdr_size(file) + mad_recv_wc->mad_len;
+	packet->mad.hdr.qpn	   = cpu_to_be32(mad_recv_wc->wc->src_qp);
+	packet->mad.hdr.lid	   = cpu_to_be16(mad_recv_wc->wc->slid);
+	packet->mad.hdr.sl	   = mad_recv_wc->wc->sl;
+	packet->mad.hdr.path_bits  = mad_recv_wc->wc->dlid_path_bits;
+	packet->mad.hdr.pkey_index = mad_recv_wc->wc->pkey_index;
 	packet->mad.hdr.grh_present = !!(mad_recv_wc->wc->wc_flags & IB_WC_GRH);
 	if (packet->mad.hdr.grh_present) {
-		/* XXX parse GRH */
-		packet->mad.hdr.gid_index 	= 0;
-		packet->mad.hdr.hop_limit 	= 0;
-		packet->mad.hdr.traffic_class	= 0;
-		memset(packet->mad.hdr.gid, 0, 16);
-		packet->mad.hdr.flow_label	= 0;
+		struct ib_ah_attr ah_attr;
+
+		ib_init_ah_from_wc(agent->device, agent->port_num,
+				   mad_recv_wc->wc, mad_recv_wc->recv_buf.grh,
+				   &ah_attr);
+
+		packet->mad.hdr.gid_index = ah_attr.grh.sgid_index;
+		packet->mad.hdr.hop_limit = ah_attr.grh.hop_limit;
+		packet->mad.hdr.traffic_class = ah_attr.grh.traffic_class;
+		memcpy(packet->mad.hdr.gid, &ah_attr.grh.dgid, 16);
+		packet->mad.hdr.flow_label = cpu_to_be32(ah_attr.grh.flow_label);
 	}
 
 	if (queue_packet(file, agent, packet))
-		kfree(packet);
+		goto err2;
+	return;
 
-out:
+err2:
+	kfree(packet);
+err1:
 	ib_free_recv_mad(mad_recv_wc);
 }
 
+static ssize_t copy_recv_mad(struct ib_umad_file *file, char __user *buf,
+			     struct ib_umad_packet *packet, size_t count)
+{
+	struct ib_mad_recv_buf *recv_buf;
+	int left, seg_payload, offset, max_seg_payload;
+
+	/* We need enough room to copy the first (or only) MAD segment. */
+	recv_buf = &packet->recv_wc->recv_buf;
+	if ((packet->length <= sizeof (*recv_buf->mad) &&
+	     count < hdr_size(file) + packet->length) ||
+	    (packet->length > sizeof (*recv_buf->mad) &&
+	     count < hdr_size(file) + sizeof (*recv_buf->mad)))
+		return -EINVAL;
+
+	if (copy_to_user(buf, &packet->mad, hdr_size(file)))
+		return -EFAULT;
+
+	buf += hdr_size(file);
+	seg_payload = min_t(int, packet->length, sizeof (*recv_buf->mad));
+	if (copy_to_user(buf, recv_buf->mad, seg_payload))
+		return -EFAULT;
+
+	if (seg_payload < packet->length) {
+		/*
+		 * Multipacket RMPP MAD message. Copy remainder of message.
+		 * Note that last segment may have a shorter payload.
+		 */
+		if (count < hdr_size(file) + packet->length) {
+			/*
+			 * The buffer is too small, return the first RMPP segment,
+			 * which includes the RMPP message length.
+			 */
+			return -ENOSPC;
+		}
+		offset = ib_get_mad_data_offset(recv_buf->mad->mad_hdr.mgmt_class);
+		max_seg_payload = sizeof (struct ib_mad) - offset;
+
+		for (left = packet->length - seg_payload, buf += seg_payload;
+		     left; left -= seg_payload, buf += seg_payload) {
+			recv_buf = container_of(recv_buf->list.next,
+						struct ib_mad_recv_buf, list);
+			seg_payload = min(left, max_seg_payload);
+			if (copy_to_user(buf, ((void *) recv_buf->mad) + offset,
+					 seg_payload))
+				return -EFAULT;
+		}
+	}
+	return hdr_size(file) + packet->length;
+}
+
+static ssize_t copy_send_mad(struct ib_umad_file *file, char __user *buf,
+			     struct ib_umad_packet *packet, size_t count)
+{
+	ssize_t size = hdr_size(file) + packet->length;
+
+	if (count < size)
+		return -EINVAL;
+
+	if (copy_to_user(buf, &packet->mad, hdr_size(file)))
+		return -EFAULT;
+
+	buf += hdr_size(file);
+
+	if (copy_to_user(buf, packet->mad.data, packet->length))
+		return -EFAULT;
+
+	return size;
+}
+
 static ssize_t ib_umad_read(struct file *filp, char __user *buf,
 			    size_t count, loff_t *pos)
 {
@@ -253,13 +332,13 @@ static ssize_t ib_umad_read(struct file *filp, char __user *buf,
 	struct ib_umad_packet *packet;
 	ssize_t ret;
 
-	if (count < sizeof (struct ib_user_mad) + sizeof (struct ib_mad))
+	if (count < hdr_size(file))
 		return -EINVAL;
 
-	spin_lock_irq(&file->recv_lock);
+	mutex_lock(&file->mutex);
 
 	while (list_empty(&file->recv_list)) {
-		spin_unlock_irq(&file->recv_lock);
+		mutex_unlock(&file->mutex);
 
 		if (filp->f_flags & O_NONBLOCK)
 			return -EAGAIN;
@@ -268,36 +347,97 @@ static ssize_t ib_umad_read(struct file *filp, char __user *buf,
 					     !list_empty(&file->recv_list)))
 			return -ERESTARTSYS;
 
-		spin_lock_irq(&file->recv_lock);
+		mutex_lock(&file->mutex);
 	}
 
 	packet = list_entry(file->recv_list.next, struct ib_umad_packet, list);
 	list_del(&packet->list);
 
-	spin_unlock_irq(&file->recv_lock);
+	mutex_unlock(&file->mutex);
 
-	if (count < packet->length + sizeof (struct ib_user_mad)) {
-		/* Return length needed (and first RMPP segment) if too small */
-		if (copy_to_user(buf, &packet->mad,
-				 sizeof (struct ib_user_mad) + sizeof (struct ib_mad)))
-			ret = -EFAULT;
-		else
-			ret = -ENOSPC;
-	} else if (copy_to_user(buf, &packet->mad,
-				packet->length + sizeof (struct ib_user_mad)))
-		ret = -EFAULT;
+	if (packet->recv_wc)
+		ret = copy_recv_mad(file, buf, packet, count);
 	else
-		ret = packet->length + sizeof (struct ib_user_mad);
+		ret = copy_send_mad(file, buf, packet, count);
+
 	if (ret < 0) {
 		/* Requeue packet */
-		spin_lock_irq(&file->recv_lock);
+		mutex_lock(&file->mutex);
 		list_add(&packet->list, &file->recv_list);
-		spin_unlock_irq(&file->recv_lock);
-	} else
+		mutex_unlock(&file->mutex);
+	} else {
+		if (packet->recv_wc)
+			ib_free_recv_mad(packet->recv_wc);
 		kfree(packet);
+	}
 	return ret;
 }
 
+static int copy_rmpp_mad(struct ib_mad_send_buf *msg, const char __user *buf)
+{
+	int left, seg;
+
+	/* Copy class specific header */
+	if ((msg->hdr_len > IB_MGMT_RMPP_HDR) &&
+	    copy_from_user(msg->mad + IB_MGMT_RMPP_HDR, buf + IB_MGMT_RMPP_HDR,
+			   msg->hdr_len - IB_MGMT_RMPP_HDR))
+		return -EFAULT;
+
+	/* All headers are in place.  Copy data segments. */
+	for (seg = 1, left = msg->data_len, buf += msg->hdr_len; left > 0;
+	     seg++, left -= msg->seg_size, buf += msg->seg_size) {
+		if (copy_from_user(ib_get_rmpp_segment(msg, seg), buf,
+				   min(left, msg->seg_size)))
+			return -EFAULT;
+	}
+	return 0;
+}
+
+static int same_destination(struct ib_user_mad_hdr *hdr1,
+			    struct ib_user_mad_hdr *hdr2)
+{
+	if (!hdr1->grh_present && !hdr2->grh_present)
+	   return (hdr1->lid == hdr2->lid);
+
+	if (hdr1->grh_present && hdr2->grh_present)
+	   return !memcmp(hdr1->gid, hdr2->gid, 16);
+
+	return 0;
+}
+
+static int is_duplicate(struct ib_umad_file *file,
+			struct ib_umad_packet *packet)
+{
+	struct ib_umad_packet *sent_packet;
+	struct ib_mad_hdr *sent_hdr, *hdr;
+
+	hdr = (struct ib_mad_hdr *) packet->mad.data;
+	list_for_each_entry(sent_packet, &file->send_list, list) {
+		sent_hdr = (struct ib_mad_hdr *) sent_packet->mad.data;
+
+		if ((hdr->tid != sent_hdr->tid) ||
+		    (hdr->mgmt_class != sent_hdr->mgmt_class))
+			continue;
+
+		/*
+		 * No need to be overly clever here.  If two new operations have
+		 * the same TID, reject the second as a duplicate.  This is more
+		 * restrictive than required by the spec.
+		 */
+		if (!ib_response_mad((struct ib_mad *) hdr)) {
+			if (!ib_response_mad((struct ib_mad *) sent_hdr))
+				return 1;
+			continue;
+		} else if (!ib_response_mad((struct ib_mad *) sent_hdr))
+			continue;
+
+		if (same_destination(&packet->mad.hdr, &sent_packet->mad.hdr))
+			return 1;
+	}
+
+	return 0;
+}
+
 static ssize_t ib_umad_write(struct file *filp, const char __user *buf,
 			     size_t count, loff_t *pos)
 {
@@ -307,32 +447,34 @@ static ssize_t ib_umad_write(struct file *filp, const char __user *buf,
 	struct ib_ah_attr ah_attr;
 	struct ib_ah *ah;
 	struct ib_rmpp_mad *rmpp_mad;
-	u8 method;
 	__be64 *tid;
-	int ret, length, hdr_len, copy_offset;
-	int rmpp_active, has_rmpp_header;
+	int ret, data_len, hdr_len, copy_offset, rmpp_active;
 
-	if (count < sizeof (struct ib_user_mad) + IB_MGMT_RMPP_HDR)
+	if (count < hdr_size(file) + IB_MGMT_RMPP_HDR)
 		return -EINVAL;
 
-	length = count - sizeof (struct ib_user_mad);
-	packet = kmalloc(sizeof *packet + IB_MGMT_RMPP_HDR, GFP_KERNEL);
+	packet = kzalloc(sizeof *packet + IB_MGMT_RMPP_HDR, GFP_KERNEL);
 	if (!packet)
 		return -ENOMEM;
 
-	if (copy_from_user(&packet->mad, buf,
-			    sizeof (struct ib_user_mad) + IB_MGMT_RMPP_HDR)) {
+	if (copy_from_user(&packet->mad, buf, hdr_size(file))) {
 		ret = -EFAULT;
 		goto err;
 	}
 
-	if (packet->mad.hdr.id < 0 ||
-	    packet->mad.hdr.id >= IB_UMAD_MAX_AGENTS) {
+	if (packet->mad.hdr.id >= IB_UMAD_MAX_AGENTS) {
 		ret = -EINVAL;
 		goto err;
 	}
 
-	down_read(&file->port->mutex);
+	buf += hdr_size(file);
+
+	if (copy_from_user(packet->mad.data, buf, IB_MGMT_RMPP_HDR)) {
+		ret = -EFAULT;
+		goto err;
+	}
+
+	mutex_lock(&file->mutex);
 
 	agent = __get_agent(file, packet->mad.hdr.id);
 	if (!agent) {
@@ -348,8 +490,9 @@ static ssize_t ib_umad_write(struct file *filp, const char __user *buf,
 	if (packet->mad.hdr.grh_present) {
 		ah_attr.ah_flags = IB_AH_GRH;
 		memcpy(ah_attr.grh.dgid.raw, packet->mad.hdr.gid, 16);
-		ah_attr.grh.flow_label 	   = be32_to_cpu(packet->mad.hdr.flow_label);
-		ah_attr.grh.hop_limit  	   = packet->mad.hdr.hop_limit;
+		ah_attr.grh.sgid_index	   = packet->mad.hdr.gid_index;
+		ah_attr.grh.flow_label	   = be32_to_cpu(packet->mad.hdr.flow_label);
+		ah_attr.grh.hop_limit	   = packet->mad.hdr.hop_limit;
 		ah_attr.grh.traffic_class  = packet->mad.hdr.traffic_class;
 	}
 
@@ -360,91 +503,84 @@ static ssize_t ib_umad_write(struct file *filp, const char __user *buf,
 	}
 
 	rmpp_mad = (struct ib_rmpp_mad *) packet->mad.data;
-	if (rmpp_mad->mad_hdr.mgmt_class == IB_MGMT_CLASS_SUBN_ADM) {
-		hdr_len = IB_MGMT_SA_HDR;
-		copy_offset = IB_MGMT_RMPP_HDR;
-		has_rmpp_header = 1;
-	} else if (rmpp_mad->mad_hdr.mgmt_class >= IB_MGMT_CLASS_VENDOR_RANGE2_START &&
-		   rmpp_mad->mad_hdr.mgmt_class <= IB_MGMT_CLASS_VENDOR_RANGE2_END) {
-			hdr_len = IB_MGMT_VENDOR_HDR;
-			copy_offset = IB_MGMT_RMPP_HDR;
-			has_rmpp_header = 1;
-	} else {
-		hdr_len = IB_MGMT_MAD_HDR;
+	hdr_len = ib_get_mad_data_offset(rmpp_mad->mad_hdr.mgmt_class);
+	if (!ib_is_mad_class_rmpp(rmpp_mad->mad_hdr.mgmt_class)) {
 		copy_offset = IB_MGMT_MAD_HDR;
-		has_rmpp_header = 0;
-	}
-
-	if (has_rmpp_header)
+		rmpp_active = 0;
+	} else {
+		copy_offset = IB_MGMT_RMPP_HDR;
 		rmpp_active = ib_get_rmpp_flags(&rmpp_mad->rmpp_hdr) &
 			      IB_MGMT_RMPP_FLAG_ACTIVE;
-	else
-		rmpp_active = 0;
-
-	/* Validate that the management class can support RMPP */
-	if (rmpp_active && !agent->rmpp_version) {
-		ret = -EINVAL;
-		goto err_ah;
 	}
 
+	data_len = count - hdr_size(file) - hdr_len;
 	packet->msg = ib_create_send_mad(agent,
 					 be32_to_cpu(packet->mad.hdr.qpn),
-					 0, rmpp_active,
-					 hdr_len, length - hdr_len,
-					 GFP_KERNEL);
+					 packet->mad.hdr.pkey_index, rmpp_active,
+					 hdr_len, data_len, GFP_KERNEL);
 	if (IS_ERR(packet->msg)) {
 		ret = PTR_ERR(packet->msg);
 		goto err_ah;
 	}
 
-	packet->msg->ah 	= ah;
+	packet->msg->ah		= ah;
 	packet->msg->timeout_ms = packet->mad.hdr.timeout_ms;
-	packet->msg->retries 	= packet->mad.hdr.retries;
+	packet->msg->retries	= packet->mad.hdr.retries;
 	packet->msg->context[0] = packet;
 
-	/* Copy MAD headers (RMPP header in place) */
+	/* Copy MAD header.  Any RMPP header is already in place. */
 	memcpy(packet->msg->mad, packet->mad.data, IB_MGMT_MAD_HDR);
-	/* Now, copy rest of message from user into send buffer */
-	if (copy_from_user(packet->msg->mad + copy_offset,
-			   buf + sizeof (struct ib_user_mad) + copy_offset,
-			   length - copy_offset)) {
-		ret = -EFAULT;
-		goto err_msg;
+
+	if (!rmpp_active) {
+		if (copy_from_user(packet->msg->mad + copy_offset,
+				   buf + copy_offset,
+				   hdr_len + data_len - copy_offset)) {
+			ret = -EFAULT;
+			goto err_msg;
+		}
+	} else {
+		ret = copy_rmpp_mad(packet->msg, buf);
+		if (ret)
+			goto err_msg;
 	}
 
 	/*
-	 * If userspace is generating a request that will generate a
-	 * response, we need to make sure the high-order part of the
-	 * transaction ID matches the agent being used to send the
-	 * MAD.
+	 * Set the high-order part of the transaction ID to make MADs from
+	 * different agents unique, and allow routing responses back to the
+	 * original requestor.
 	 */
-	method = ((struct ib_mad_hdr *) packet->msg->mad)->method;
-
-	if (!(method & IB_MGMT_METHOD_RESP)       &&
-	    method != IB_MGMT_METHOD_TRAP_REPRESS &&
-	    method != IB_MGMT_METHOD_SEND) {
+	if (!ib_response_mad(packet->msg->mad)) {
 		tid = &((struct ib_mad_hdr *) packet->msg->mad)->tid;
 		*tid = cpu_to_be64(((u64) agent->hi_tid) << 32 |
 				   (be64_to_cpup(tid) & 0xffffffff));
+		rmpp_mad->mad_hdr.tid = *tid;
 	}
 
-	ret = ib_post_send_mad(packet->msg, NULL);
-	if (ret)
+	spin_lock_irq(&file->send_lock);
+	ret = is_duplicate(file, packet);
+	if (!ret)
+		list_add_tail(&packet->list, &file->send_list);
+	spin_unlock_irq(&file->send_lock);
+	if (ret) {
+		ret = -EINVAL;
 		goto err_msg;
+	}
 
-	up_read(&file->port->mutex);
+	ret = ib_post_send_mad(packet->msg, NULL);
+	if (ret)
+		goto err_send;
 
+	mutex_unlock(&file->mutex);
 	return count;
 
+err_send:
+	dequeue_send(file, packet);
 err_msg:
 	ib_free_send_mad(packet->msg);
-
 err_ah:
 	ib_destroy_ah(ah);
-
 err_up:
-	up_read(&file->port->mutex);
-
+	mutex_unlock(&file->mutex);
 err:
 	kfree(packet);
 	return ret;
@@ -465,22 +601,24 @@ static unsigned int ib_umad_poll(struct file *filp, struct poll_table_struct *wa
 	return mask;
 }
 
-static int ib_umad_reg_agent(struct ib_umad_file *file, unsigned long arg)
+static int ib_umad_reg_agent(struct ib_umad_file *file, void __user *arg,
+			     int compat_method_mask)
 {
 	struct ib_user_mad_reg_req ureq;
 	struct ib_mad_reg_req req;
-	struct ib_mad_agent *agent;
+	struct ib_mad_agent *agent = NULL;
 	int agent_id;
 	int ret;
 
-	down_write(&file->port->mutex);
+	mutex_lock(&file->port->file_mutex);
+	mutex_lock(&file->mutex);
 
 	if (!file->port->ib_dev) {
 		ret = -EPIPE;
 		goto out;
 	}
 
-	if (copy_from_user(&ureq, (void __user *) arg, sizeof ureq)) {
+	if (copy_from_user(&ureq, arg, sizeof ureq)) {
 		ret = -EFAULT;
 		goto out;
 	}
@@ -501,8 +639,18 @@ found:
 	if (ureq.mgmt_class) {
 		req.mgmt_class         = ureq.mgmt_class;
 		req.mgmt_class_version = ureq.mgmt_class_version;
-		memcpy(req.method_mask, ureq.method_mask, sizeof req.method_mask);
-		memcpy(req.oui,         ureq.oui,         sizeof req.oui);
+		memcpy(req.oui, ureq.oui, sizeof req.oui);
+
+		if (compat_method_mask) {
+			u32 *umm = (u32 *) ureq.method_mask;
+			int i;
+
+			for (i = 0; i < BITS_TO_LONGS(IB_MGMT_MAX_METHODS); ++i)
+				req.method_mask[i] =
+					umm[i * 2] | ((u64) umm[i * 2 + 1] << 32);
+		} else
+			memcpy(req.method_mask, ureq.method_mask,
+			       sizeof req.method_mask);
 	}
 
 	agent = ib_register_mad_agent(file->port->ib_dev, file->port->port_num,
@@ -512,36 +660,53 @@ found:
 				      send_handler, recv_handler, file);
 	if (IS_ERR(agent)) {
 		ret = PTR_ERR(agent);
+		agent = NULL;
 		goto out;
 	}
 
 	if (put_user(agent_id,
 		     (u32 __user *) (arg + offsetof(struct ib_user_mad_reg_req, id)))) {
 		ret = -EFAULT;
-		ib_unregister_mad_agent(agent);
 		goto out;
 	}
 
+	if (!file->already_used) {
+		file->already_used = 1;
+		if (!file->use_pkey_index) {
+			printk(KERN_WARNING "user_mad: process %s did not enable "
+			       "P_Key index support.\n", current->comm);
+			printk(KERN_WARNING "user_mad:   Documentation/infiniband/user_mad.txt "
+			       "has info on the new ABI.\n");
+		}
+	}
+
 	file->agent[agent_id] = agent;
 	ret = 0;
 
 out:
-	up_write(&file->port->mutex);
+	mutex_unlock(&file->mutex);
+
+	if (ret && agent)
+		ib_unregister_mad_agent(agent);
+
+	mutex_unlock(&file->port->file_mutex);
+
 	return ret;
 }
 
-static int ib_umad_unreg_agent(struct ib_umad_file *file, unsigned long arg)
+static int ib_umad_unreg_agent(struct ib_umad_file *file, u32 __user *arg)
 {
 	struct ib_mad_agent *agent = NULL;
 	u32 id;
 	int ret = 0;
 
-	if (get_user(id, (u32 __user *) arg))
+	if (get_user(id, arg))
 		return -EFAULT;
 
-	down_write(&file->port->mutex);
+	mutex_lock(&file->port->file_mutex);
+	mutex_lock(&file->mutex);
 
-	if (id < 0 || id >= IB_UMAD_MAX_AGENTS || !__get_agent(file, id)) {
+	if (id >= IB_UMAD_MAX_AGENTS || !__get_agent(file, id)) {
 		ret = -EINVAL;
 		goto out;
 	}
@@ -550,11 +715,27 @@ static int ib_umad_unreg_agent(struct ib_umad_file *file, unsigned long arg)
 	file->agent[id] = NULL;
 
 out:
-	up_write(&file->port->mutex);
+	mutex_unlock(&file->mutex);
 
 	if (agent)
 		ib_unregister_mad_agent(agent);
 
+	mutex_unlock(&file->port->file_mutex);
+
+	return ret;
+}
+
+static long ib_umad_enable_pkey(struct ib_umad_file *file)
+{
+	int ret = 0;
+
+	mutex_lock(&file->mutex);
+	if (file->already_used)
+		ret = -EINVAL;
+	else
+		file->use_pkey_index = 1;
+	mutex_unlock(&file->mutex);
+
 	return ret;
 }
 
@@ -563,45 +744,64 @@ static long ib_umad_ioctl(struct file *filp, unsigned int cmd,
 {
 	switch (cmd) {
 	case IB_USER_MAD_REGISTER_AGENT:
-		return ib_umad_reg_agent(filp->private_data, arg);
+		return ib_umad_reg_agent(filp->private_data, (void __user *) arg, 0);
 	case IB_USER_MAD_UNREGISTER_AGENT:
-		return ib_umad_unreg_agent(filp->private_data, arg);
+		return ib_umad_unreg_agent(filp->private_data, (__u32 __user *) arg);
+	case IB_USER_MAD_ENABLE_PKEY:
+		return ib_umad_enable_pkey(filp->private_data);
 	default:
 		return -ENOIOCTLCMD;
 	}
 }
 
+#ifdef CONFIG_COMPAT
+static long ib_umad_compat_ioctl(struct file *filp, unsigned int cmd,
+				 unsigned long arg)
+{
+	switch (cmd) {
+	case IB_USER_MAD_REGISTER_AGENT:
+		return ib_umad_reg_agent(filp->private_data, compat_ptr(arg), 1);
+	case IB_USER_MAD_UNREGISTER_AGENT:
+		return ib_umad_unreg_agent(filp->private_data, compat_ptr(arg));
+	case IB_USER_MAD_ENABLE_PKEY:
+		return ib_umad_enable_pkey(filp->private_data);
+	default:
+		return -ENOIOCTLCMD;
+	}
+}
+#endif
+
+/*
+ * ib_umad_open() does not need the BKL:
+ *
+ *  - the ib_umad_port structures are properly reference counted, and
+ *    everything else is purely local to the file being created, so
+ *    races against other open calls are not a problem;
+ *  - the ioctl method does not affect any global state outside of the
+ *    file structure being operated on;
+ */
 static int ib_umad_open(struct inode *inode, struct file *filp)
 {
 	struct ib_umad_port *port;
 	struct ib_umad_file *file;
-	int ret = 0;
-
-	spin_lock(&port_lock);
-	port = umad_port[iminor(inode) - IB_UMAD_MINOR_BASE];
-	if (port)
-		kref_get(&port->umad_dev->ref);
-	spin_unlock(&port_lock);
+	int ret = -ENXIO;
 
-	if (!port)
-		return -ENXIO;
+	port = container_of(inode->i_cdev, struct ib_umad_port, cdev);
 
-	down_write(&port->mutex);
+	mutex_lock(&port->file_mutex);
 
-	if (!port->ib_dev) {
-		ret = -ENXIO;
+	if (!port->ib_dev)
 		goto out;
-	}
 
+	ret = -ENOMEM;
 	file = kzalloc(sizeof *file, GFP_KERNEL);
-	if (!file) {
-		kref_put(&port->umad_dev->ref, ib_umad_release_dev);
-		ret = -ENOMEM;
+	if (!file)
 		goto out;
-	}
 
-	spin_lock_init(&file->recv_lock);
+	mutex_init(&file->mutex);
+	spin_lock_init(&file->send_lock);
 	INIT_LIST_HEAD(&file->recv_list);
+	INIT_LIST_HEAD(&file->send_list);
 	init_waitqueue_head(&file->recv_wait);
 
 	file->port = port;
@@ -609,8 +809,17 @@ static int ib_umad_open(struct inode *inode, struct file *filp)
 
 	list_add_tail(&file->port_list, &port->file_list);
 
+	ret = nonseekable_open(inode, filp);
+	if (ret) {
+		list_del(&file->port_list);
+		kfree(file);
+		goto out;
+	}
+
+	kobject_get(&port->umad_dev->kobj);
+
 out:
-	up_write(&port->mutex);
+	mutex_unlock(&port->file_mutex);
 	return ret;
 }
 
@@ -622,40 +831,47 @@ static int ib_umad_close(struct inode *inode, struct file *filp)
 	int already_dead;
 	int i;
 
-	down_write(&file->port->mutex);
+	mutex_lock(&file->port->file_mutex);
+	mutex_lock(&file->mutex);
 
 	already_dead = file->agents_dead;
 	file->agents_dead = 1;
 
-	list_for_each_entry_safe(packet, tmp, &file->recv_list, list)
+	list_for_each_entry_safe(packet, tmp, &file->recv_list, list) {
+		if (packet->recv_wc)
+			ib_free_recv_mad(packet->recv_wc);
 		kfree(packet);
+	}
 
 	list_del(&file->port_list);
 
-	downgrade_write(&file->port->mutex);
+	mutex_unlock(&file->mutex);
 
 	if (!already_dead)
 		for (i = 0; i < IB_UMAD_MAX_AGENTS; ++i)
 			if (file->agent[i])
 				ib_unregister_mad_agent(file->agent[i]);
 
-	up_read(&file->port->mutex);
+	mutex_unlock(&file->port->file_mutex);
 
 	kfree(file);
-	kref_put(&dev->ref, ib_umad_release_dev);
+	kobject_put(&dev->kobj);
 
 	return 0;
 }
 
-static struct file_operations umad_fops = {
-	.owner 	 	= THIS_MODULE,
-	.read 	 	= ib_umad_read,
-	.write 	 	= ib_umad_write,
-	.poll 	 	= ib_umad_poll,
+static const struct file_operations umad_fops = {
+	.owner		= THIS_MODULE,
+	.read		= ib_umad_read,
+	.write		= ib_umad_write,
+	.poll		= ib_umad_poll,
 	.unlocked_ioctl = ib_umad_ioctl,
-	.compat_ioctl 	= ib_umad_ioctl,
-	.open 	 	= ib_umad_open,
-	.release 	= ib_umad_close
+#ifdef CONFIG_COMPAT
+	.compat_ioctl	= ib_umad_compat_ioctl,
+#endif
+	.open		= ib_umad_open,
+	.release	= ib_umad_close,
+	.llseek		= no_llseek,
 };
 
 static int ib_umad_sm_open(struct inode *inode, struct file *filp)
@@ -666,14 +882,7 @@ static int ib_umad_sm_open(struct inode *inode, struct file *filp)
 	};
 	int ret;
 
-	spin_lock(&port_lock);
-	port = umad_port[iminor(inode) - IB_UMAD_MINOR_BASE - IB_UMAD_MAX_PORTS];
-	if (port)
-		kref_get(&port->umad_dev->ref);
-	spin_unlock(&port_lock);
-
-	if (!port)
-		return -ENXIO;
+	port = container_of(inode->i_cdev, struct ib_umad_port, sm_cdev);
 
 	if (filp->f_flags & O_NONBLOCK) {
 		if (down_trylock(&port->sm_sem)) {
@@ -688,17 +897,27 @@ static int ib_umad_sm_open(struct inode *inode, struct file *filp)
 	}
 
 	ret = ib_modify_port(port->ib_dev, port->port_num, 0, &props);
-	if (ret) {
-		up(&port->sm_sem);
-		goto fail;
-	}
+	if (ret)
+		goto err_up_sem;
 
 	filp->private_data = port;
 
+	ret = nonseekable_open(inode, filp);
+	if (ret)
+		goto err_clr_sm_cap;
+
+	kobject_get(&port->umad_dev->kobj);
+
 	return 0;
 
+err_clr_sm_cap:
+	swap(props.set_port_cap_mask, props.clr_port_cap_mask);
+	ib_modify_port(port->ib_dev, port->port_num, 0, &props);
+
+err_up_sem:
+	up(&port->sm_sem);
+
 fail:
-	kref_put(&port->umad_dev->ref, ib_umad_release_dev);
 	return ret;
 }
 
@@ -710,22 +929,23 @@ static int ib_umad_sm_close(struct inode *inode, struct file *filp)
 	};
 	int ret = 0;
 
-	down_write(&port->mutex);
+	mutex_lock(&port->file_mutex);
 	if (port->ib_dev)
 		ret = ib_modify_port(port->ib_dev, port->port_num, 0, &props);
-	up_write(&port->mutex);
+	mutex_unlock(&port->file_mutex);
 
 	up(&port->sm_sem);
 
-	kref_put(&port->umad_dev->ref, ib_umad_release_dev);
+	kobject_put(&port->umad_dev->kobj);
 
 	return ret;
 }
 
-static struct file_operations umad_sm_fops = {
-	.owner 	 = THIS_MODULE,
-	.open 	 = ib_umad_sm_open,
-	.release = ib_umad_sm_close
+static const struct file_operations umad_sm_fops = {
+	.owner	 = THIS_MODULE,
+	.open	 = ib_umad_sm_open,
+	.release = ib_umad_sm_close,
+	.llseek	 = no_llseek,
 };
 
 static struct ib_client umad_client = {
@@ -734,113 +954,141 @@ static struct ib_client umad_client = {
 	.remove = ib_umad_remove_one
 };
 
-static ssize_t show_ibdev(struct class_device *class_dev, char *buf)
+static ssize_t show_ibdev(struct device *dev, struct device_attribute *attr,
+			  char *buf)
 {
-	struct ib_umad_port *port = class_get_devdata(class_dev);
+	struct ib_umad_port *port = dev_get_drvdata(dev);
 
 	if (!port)
 		return -ENODEV;
 
 	return sprintf(buf, "%s\n", port->ib_dev->name);
 }
-static CLASS_DEVICE_ATTR(ibdev, S_IRUGO, show_ibdev, NULL);
+static DEVICE_ATTR(ibdev, S_IRUGO, show_ibdev, NULL);
 
-static ssize_t show_port(struct class_device *class_dev, char *buf)
+static ssize_t show_port(struct device *dev, struct device_attribute *attr,
+			 char *buf)
 {
-	struct ib_umad_port *port = class_get_devdata(class_dev);
+	struct ib_umad_port *port = dev_get_drvdata(dev);
 
 	if (!port)
 		return -ENODEV;
 
 	return sprintf(buf, "%d\n", port->port_num);
 }
-static CLASS_DEVICE_ATTR(port, S_IRUGO, show_port, NULL);
+static DEVICE_ATTR(port, S_IRUGO, show_port, NULL);
 
-static ssize_t show_abi_version(struct class *class, char *buf)
+static CLASS_ATTR_STRING(abi_version, S_IRUGO,
+			 __stringify(IB_USER_MAD_ABI_VERSION));
+
+static dev_t overflow_maj;
+static DECLARE_BITMAP(overflow_map, IB_UMAD_MAX_PORTS);
+static int find_overflow_devnum(void)
 {
-	return sprintf(buf, "%d\n", IB_USER_MAD_ABI_VERSION);
+	int ret;
+
+	if (!overflow_maj) {
+		ret = alloc_chrdev_region(&overflow_maj, 0, IB_UMAD_MAX_PORTS * 2,
+					  "infiniband_mad");
+		if (ret) {
+			printk(KERN_ERR "user_mad: couldn't register dynamic device number\n");
+			return ret;
+		}
+	}
+
+	ret = find_first_zero_bit(overflow_map, IB_UMAD_MAX_PORTS);
+	if (ret >= IB_UMAD_MAX_PORTS)
+		return -1;
+
+	return ret;
 }
-static CLASS_ATTR(abi_version, S_IRUGO, show_abi_version, NULL);
 
 static int ib_umad_init_port(struct ib_device *device, int port_num,
+			     struct ib_umad_device *umad_dev,
 			     struct ib_umad_port *port)
 {
+	int devnum;
+	dev_t base;
+
 	spin_lock(&port_lock);
-	port->dev_num = find_first_zero_bit(dev_map, IB_UMAD_MAX_PORTS);
-	if (port->dev_num >= IB_UMAD_MAX_PORTS) {
+	devnum = find_first_zero_bit(dev_map, IB_UMAD_MAX_PORTS);
+	if (devnum >= IB_UMAD_MAX_PORTS) {
 		spin_unlock(&port_lock);
-		return -1;
+		devnum = find_overflow_devnum();
+		if (devnum < 0)
+			return -1;
+
+		spin_lock(&port_lock);
+		port->dev_num = devnum + IB_UMAD_MAX_PORTS;
+		base = devnum + overflow_maj;
+		set_bit(devnum, overflow_map);
+	} else {
+		port->dev_num = devnum;
+		base = devnum + base_dev;
+		set_bit(devnum, dev_map);
 	}
-	set_bit(port->dev_num, dev_map);
 	spin_unlock(&port_lock);
 
 	port->ib_dev   = device;
 	port->port_num = port_num;
-	init_MUTEX(&port->sm_sem);
-	init_rwsem(&port->mutex);
+	sema_init(&port->sm_sem, 1);
+	mutex_init(&port->file_mutex);
 	INIT_LIST_HEAD(&port->file_list);
 
-	port->dev = cdev_alloc();
-	if (!port->dev)
-		return -1;
-	port->dev->owner = THIS_MODULE;
-	port->dev->ops   = &umad_fops;
-	kobject_set_name(&port->dev->kobj, "umad%d", port->dev_num);
-	if (cdev_add(port->dev, base_dev + port->dev_num, 1))
+	cdev_init(&port->cdev, &umad_fops);
+	port->cdev.owner = THIS_MODULE;
+	port->cdev.kobj.parent = &umad_dev->kobj;
+	kobject_set_name(&port->cdev.kobj, "umad%d", port->dev_num);
+	if (cdev_add(&port->cdev, base, 1))
 		goto err_cdev;
 
-	port->class_dev = class_device_create(umad_class, NULL, port->dev->dev,
-					      device->dma_device,
-					      "umad%d", port->dev_num);
-	if (IS_ERR(port->class_dev))
+	port->dev = device_create(umad_class, device->dma_device,
+				  port->cdev.dev, port,
+				  "umad%d", port->dev_num);
+	if (IS_ERR(port->dev))
 		goto err_cdev;
 
-	if (class_device_create_file(port->class_dev, &class_device_attr_ibdev))
-		goto err_class;
-	if (class_device_create_file(port->class_dev, &class_device_attr_port))
-		goto err_class;
-
-	port->sm_dev = cdev_alloc();
-	if (!port->sm_dev)
-		goto err_class;
-	port->sm_dev->owner = THIS_MODULE;
-	port->sm_dev->ops   = &umad_sm_fops;
-	kobject_set_name(&port->sm_dev->kobj, "issm%d", port->dev_num);
-	if (cdev_add(port->sm_dev, base_dev + port->dev_num + IB_UMAD_MAX_PORTS, 1))
+	if (device_create_file(port->dev, &dev_attr_ibdev))
+		goto err_dev;
+	if (device_create_file(port->dev, &dev_attr_port))
+		goto err_dev;
+
+	base += IB_UMAD_MAX_PORTS;
+	cdev_init(&port->sm_cdev, &umad_sm_fops);
+	port->sm_cdev.owner = THIS_MODULE;
+	port->sm_cdev.kobj.parent = &umad_dev->kobj;
+	kobject_set_name(&port->sm_cdev.kobj, "issm%d", port->dev_num);
+	if (cdev_add(&port->sm_cdev, base, 1))
 		goto err_sm_cdev;
 
-	port->sm_class_dev = class_device_create(umad_class, NULL, port->sm_dev->dev,
-						 device->dma_device,
-						 "issm%d", port->dev_num);
-	if (IS_ERR(port->sm_class_dev))
+	port->sm_dev = device_create(umad_class, device->dma_device,
+				     port->sm_cdev.dev, port,
+				     "issm%d", port->dev_num);
+	if (IS_ERR(port->sm_dev))
 		goto err_sm_cdev;
 
-	class_set_devdata(port->class_dev,    port);
-	class_set_devdata(port->sm_class_dev, port);
-
-	if (class_device_create_file(port->sm_class_dev, &class_device_attr_ibdev))
-		goto err_sm_class;
-	if (class_device_create_file(port->sm_class_dev, &class_device_attr_port))
-		goto err_sm_class;
-
-	spin_lock(&port_lock);
-	umad_port[port->dev_num] = port;
-	spin_unlock(&port_lock);
+	if (device_create_file(port->sm_dev, &dev_attr_ibdev))
+		goto err_sm_dev;
+	if (device_create_file(port->sm_dev, &dev_attr_port))
+		goto err_sm_dev;
 
 	return 0;
 
-err_sm_class:
-	class_device_destroy(umad_class, port->sm_dev->dev);
+err_sm_dev:
+	device_destroy(umad_class, port->sm_cdev.dev);
 
 err_sm_cdev:
-	cdev_del(port->sm_dev);
+	cdev_del(&port->sm_cdev);
 
-err_class:
-	class_device_destroy(umad_class, port->dev->dev);
+err_dev:
+	device_destroy(umad_class, port->cdev.dev);
 
 err_cdev:
-	cdev_del(port->dev);
-	clear_bit(port->dev_num, dev_map);
+	cdev_del(&port->cdev);
+	if (port->dev_num < IB_UMAD_MAX_PORTS)
+		clear_bit(devnum, dev_map);
+	else
+		clear_bit(devnum, overflow_map);
 
 	return -1;
 }
@@ -850,57 +1098,35 @@ static void ib_umad_kill_port(struct ib_umad_port *port)
 	struct ib_umad_file *file;
 	int id;
 
-	class_set_devdata(port->class_dev,    NULL);
-	class_set_devdata(port->sm_class_dev, NULL);
+	dev_set_drvdata(port->dev,    NULL);
+	dev_set_drvdata(port->sm_dev, NULL);
 
-	class_device_destroy(umad_class, port->dev->dev);
-	class_device_destroy(umad_class, port->sm_dev->dev);
+	device_destroy(umad_class, port->cdev.dev);
+	device_destroy(umad_class, port->sm_cdev.dev);
 
-	cdev_del(port->dev);
-	cdev_del(port->sm_dev);
-
-	spin_lock(&port_lock);
-	umad_port[port->dev_num] = NULL;
-	spin_unlock(&port_lock);
+	cdev_del(&port->cdev);
+	cdev_del(&port->sm_cdev);
 
-	down_write(&port->mutex);
+	mutex_lock(&port->file_mutex);
 
 	port->ib_dev = NULL;
 
-	/*
-	 * Now go through the list of files attached to this port and
-	 * unregister all of their MAD agents.  We need to hold
-	 * port->mutex while doing this to avoid racing with
-	 * ib_umad_close(), but we can't hold the mutex for writing
-	 * while calling ib_unregister_mad_agent(), since that might
-	 * deadlock by calling back into queue_packet().  So we
-	 * downgrade our lock to a read lock, and then drop and
-	 * reacquire the write lock for the next iteration.
-	 *
-	 * We do list_del_init() on the file's list_head so that the
-	 * list_del in ib_umad_close() is still OK, even after the
-	 * file is removed from the list.
-	 */
-	while (!list_empty(&port->file_list)) {
-		file = list_entry(port->file_list.next, struct ib_umad_file,
-				  port_list);
-
+	list_for_each_entry(file, &port->file_list, port_list) {
+		mutex_lock(&file->mutex);
 		file->agents_dead = 1;
-		list_del_init(&file->port_list);
-
-		downgrade_write(&port->mutex);
+		mutex_unlock(&file->mutex);
 
 		for (id = 0; id < IB_UMAD_MAX_AGENTS; ++id)
 			if (file->agent[id])
 				ib_unregister_mad_agent(file->agent[id]);
-
-		up_read(&port->mutex);
-		down_write(&port->mutex);
 	}
 
-	up_write(&port->mutex);
+	mutex_unlock(&port->file_mutex);
 
-	clear_bit(port->dev_num, dev_map);
+	if (port->dev_num < IB_UMAD_MAX_PORTS)
+		clear_bit(port->dev_num, dev_map);
+	else
+		clear_bit(port->dev_num - IB_UMAD_MAX_PORTS, overflow_map);
 }
 
 static void ib_umad_add_one(struct ib_device *device)
@@ -908,7 +1134,10 @@ static void ib_umad_add_one(struct ib_device *device)
 	struct ib_umad_device *umad_dev;
 	int s, e, i;
 
-	if (device->node_type == IB_NODE_SWITCH)
+	if (rdma_node_get_transport(device->node_type) != RDMA_TRANSPORT_IB)
+		return;
+
+	if (device->node_type == RDMA_NODE_IB_SWITCH)
 		s = e = 0;
 	else {
 		s = 1;
@@ -921,7 +1150,7 @@ static void ib_umad_add_one(struct ib_device *device)
 	if (!umad_dev)
 		return;
 
-	kref_init(&umad_dev->ref);
+	kobject_init(&umad_dev->kobj, &ib_umad_dev_ktype);
 
 	umad_dev->start_port = s;
 	umad_dev->end_port   = e;
@@ -929,7 +1158,8 @@ static void ib_umad_add_one(struct ib_device *device)
 	for (i = s; i <= e; ++i) {
 		umad_dev->port[i - s].umad_dev = umad_dev;
 
-		if (ib_umad_init_port(device, i, &umad_dev->port[i - s]))
+		if (ib_umad_init_port(device, i, umad_dev,
+				      &umad_dev->port[i - s]))
 			goto err;
 	}
 
@@ -941,7 +1171,7 @@ err:
 	while (--i >= s)
 		ib_umad_kill_port(&umad_dev->port[i - s]);
 
-	kref_put(&umad_dev->ref, ib_umad_release_dev);
+	kobject_put(&umad_dev->kobj);
 }
 
 static void ib_umad_remove_one(struct ib_device *device)
@@ -955,7 +1185,12 @@ static void ib_umad_remove_one(struct ib_device *device)
 	for (i = 0; i <= umad_dev->end_port - umad_dev->start_port; ++i)
 		ib_umad_kill_port(&umad_dev->port[i]);
 
-	kref_put(&umad_dev->ref, ib_umad_release_dev);
+	kobject_put(&umad_dev->kobj);
+}
+
+static char *umad_devnode(struct device *dev, umode_t *mode)
+{
+	return kasprintf(GFP_KERNEL, "infiniband/%s", dev_name(dev));
 }
 
 static int __init ib_umad_init(void)
@@ -976,7 +1211,9 @@ static int __init ib_umad_init(void)
 		goto out_chrdev;
 	}
 
-	ret = class_create_file(umad_class, &class_attr_abi_version);
+	umad_class->devnode = umad_devnode;
+
+	ret = class_create_file(umad_class, &class_attr_abi_version.attr);
 	if (ret) {
 		printk(KERN_ERR "user_mad: couldn't create abi_version attribute\n");
 		goto out_class;
@@ -1005,6 +1242,8 @@ static void __exit ib_umad_cleanup(void)
 	ib_unregister_client(&umad_client);
 	class_destroy(umad_class);
 	unregister_chrdev_region(base_dev, IB_UMAD_MAX_PORTS * 2);
+	if (overflow_maj)
+		unregister_chrdev_region(overflow_maj, IB_UMAD_MAX_PORTS * 2);
 }
 
 module_init(ib_umad_init);