/*
 * Copyright (c) 2004-2007 Voltaire, Inc. All rights reserved.
 * Copyright (c) 2005 Intel Corporation.  All rights reserved.
 * Copyright (c) 2005 Mellanox Technologies Ltd.  All rights reserved.
 * Copyright (c) 2009 HNR Consulting. 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
 * General Public License (GPL) Version 2, available from the file
 * COPYING in the main directory of this source tree, or the
 * OpenIB.org BSD license below:
 *
 *     Redistribution and use in source and binary forms, with or
 *     without modification, are permitted provided that the following
 *     conditions are met:
 *
 *      - Redistributions of source code must retain the above
 *        copyright notice, this list of conditions and the following
 *        disclaimer.
 *
 *      - Redistributions in binary form must reproduce the above
 *        copyright notice, this list of conditions and the following
 *        disclaimer in the documentation and/or other materials
 *        provided with the distribution.
 *
 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
 * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
 * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
 * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
 * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
 * 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.
 *
 */
#include <linux/dma-mapping.h>
#include <linux/slab.h>
#include <rdma/ib_cache.h>

#include "mad_priv.h"
#include "mad_rmpp.h"
#include "smi.h"
#include "agent.h"

MODULE_LICENSE("Dual BSD/GPL");
MODULE_DESCRIPTION("kernel IB MAD API");
MODULE_AUTHOR("Hal Rosenstock");
MODULE_AUTHOR("Sean Hefty");

static int mad_sendq_size = IB_MAD_QP_SEND_SIZE;
static int mad_recvq_size = IB_MAD_QP_RECV_SIZE;

module_param_named(send_queue_size, mad_sendq_size, int, 0444);
MODULE_PARM_DESC(send_queue_size, "Size of send queue in number of work requests");
module_param_named(recv_queue_size, mad_recvq_size, int, 0444);
MODULE_PARM_DESC(recv_queue_size, "Size of receive queue in number of work requests");

static struct kmem_cache *ib_mad_cache;

static struct list_head ib_mad_port_list;
static u32 ib_mad_client_id = 0;

/* Port list lock */
static DEFINE_SPINLOCK(ib_mad_port_list_lock);

/* Forward declarations */
static int method_in_use(struct ib_mad_mgmt_method_table **method,
			 struct ib_mad_reg_req *mad_reg_req);
static void remove_mad_reg_req(struct ib_mad_agent_private *priv);
static struct ib_mad_agent_private *find_mad_agent(
					struct ib_mad_port_private *port_priv,
					struct ib_mad *mad);
static int ib_mad_post_receive_mads(struct ib_mad_qp_info *qp_info,
				    struct ib_mad_private *mad);
static void cancel_mads(struct ib_mad_agent_private *mad_agent_priv);
static void timeout_sends(struct work_struct *work);
static void local_completions(struct work_struct *work);
static int add_nonoui_reg_req(struct ib_mad_reg_req *mad_reg_req,
			      struct ib_mad_agent_private *agent_priv,
			      u8 mgmt_class);
static int add_oui_reg_req(struct ib_mad_reg_req *mad_reg_req,
			   struct ib_mad_agent_private *agent_priv);

/*
 * Returns a ib_mad_port_private structure or NULL for a device/port
 * Assumes ib_mad_port_list_lock is being held
 */
static inline struct ib_mad_port_private *
__ib_get_mad_port(struct ib_device *device, int port_num)
{
	struct ib_mad_port_private *entry;

	list_for_each_entry(entry, &ib_mad_port_list, port_list) {
		if (entry->device == device && entry->port_num == port_num)
			return entry;
	}
	return NULL;
}

/*
 * Wrapper function to return a ib_mad_port_private structure or NULL
 * for a device/port
 */
static inline struct ib_mad_port_private *
ib_get_mad_port(struct ib_device *device, int port_num)
{
	struct ib_mad_port_private *entry;
	unsigned long flags;

	spin_lock_irqsave(&ib_mad_port_list_lock, flags);
	entry = __ib_get_mad_port(device, port_num);
	spin_unlock_irqrestore(&ib_mad_port_list_lock, flags);

	return entry;
}

static inline u8 convert_mgmt_class(u8 mgmt_class)
{
	/* Alias IB_MGMT_CLASS_SUBN_DIRECTED_ROUTE to 0 */
	return mgmt_class == IB_MGMT_CLASS_SUBN_DIRECTED_ROUTE ?
		0 : mgmt_class;
}

static int get_spl_qp_index(enum ib_qp_type qp_type)
{
	switch (qp_type)
	{
	case IB_QPT_SMI:
		return 0;
	case IB_QPT_GSI:
		return 1;
	default:
		return -1;
	}
}

static int vendor_class_index(u8 mgmt_class)
{
	return mgmt_class - IB_MGMT_CLASS_VENDOR_RANGE2_START;
}

static int is_vendor_class(u8 mgmt_class)
{
	if ((mgmt_class < IB_MGMT_CLASS_VENDOR_RANGE2_START) ||
	    (mgmt_class > IB_MGMT_CLASS_VENDOR_RANGE2_END))
		return 0;
	return 1;
}

static int is_vendor_oui(char *oui)
{
	if (oui[0] || oui[1] || oui[2])
		return 1;
	return 0;
}

static int is_vendor_method_in_use(
		struct ib_mad_mgmt_vendor_class *vendor_class,
		struct ib_mad_reg_req *mad_reg_req)
{
	struct ib_mad_mgmt_method_table *method;
	int i;

	for (i = 0; i < MAX_MGMT_OUI; i++) {
		if (!memcmp(vendor_class->oui[i], mad_reg_req->oui, 3)) {
			method = vendor_class->method_table[i];
			if (method) {
				if (method_in_use(&method, mad_reg_req))
					return 1;
				else
					break;
			}
		}
	}
	return 0;
}

int ib_response_mad(struct ib_mad *mad)
{
	return ((mad->mad_hdr.method & IB_MGMT_METHOD_RESP) ||
		(mad->mad_hdr.method == IB_MGMT_METHOD_TRAP_REPRESS) ||
		((mad->mad_hdr.mgmt_class == IB_MGMT_CLASS_BM) &&
		 (mad->mad_hdr.attr_mod & IB_BM_ATTR_MOD_RESP)));
}
EXPORT_SYMBOL(ib_response_mad);

/*
 * ib_register_mad_agent - Register to send/receive MADs
 */
struct ib_mad_agent *ib_register_mad_agent(struct ib_device *device,
					   u8 port_num,
					   enum ib_qp_type qp_type,
					   struct ib_mad_reg_req *mad_reg_req,
					   u8 rmpp_version,
					   ib_mad_send_handler send_handler,
					   ib_mad_recv_handler recv_handler,
					   void *context)
{
	struct ib_mad_port_private *port_priv;
	struct ib_mad_agent *ret = ERR_PTR(-EINVAL);
	struct ib_mad_agent_private *mad_agent_priv;
	struct ib_mad_reg_req *reg_req = NULL;
	struct ib_mad_mgmt_class_table *class;
	struct ib_mad_mgmt_vendor_class_table *vendor;
	struct ib_mad_mgmt_vendor_class *vendor_class;
	struct ib_mad_mgmt_method_table *method;
	int ret2, qpn;
	unsigned long flags;
	u8 mgmt_class, vclass;

	/* Validate parameters */
	qpn = get_spl_qp_index(qp_type);
	if (qpn == -1)
		goto error1;

	if (rmpp_version && rmpp_version != IB_MGMT_RMPP_VERSION)
		goto error1;

	/* Validate MAD registration request if supplied */
	if (mad_reg_req) {
		if (mad_reg_req->mgmt_class_version >= MAX_MGMT_VERSION)
			goto error1;
		if (!recv_handler)
			goto error1;
		if (mad_reg_req->mgmt_class >= MAX_MGMT_CLASS) {
			/*
			 * IB_MGMT_CLASS_SUBN_DIRECTED_ROUTE is the only
			 * one in this range currently allowed
			 */
			if (mad_reg_req->mgmt_class !=
			    IB_MGMT_CLASS_SUBN_DIRECTED_ROUTE)
				goto error1;
		} else if (mad_reg_req->mgmt_class == 0) {
			/*
			 * Class 0 is reserved in IBA and is used for
			 * aliasing of IB_MGMT_CLASS_SUBN_DIRECTED_ROUTE
			 */
			goto error1;
		} else if (is_vendor_class(mad_reg_req->mgmt_class)) {
			/*
			 * If class is in "new" vendor range,
			 * ensure supplied OUI is not zero
			 */
			if (!is_vendor_oui(mad_reg_req->oui))
				goto error1;
		}
		/* Make sure class supplied is consistent with RMPP */
		if (!ib_is_mad_class_rmpp(mad_reg_req->mgmt_class)) {
			if (rmpp_version)
				goto error1;
		}
		/* Make sure class supplied is consistent with QP type */
		if (qp_type == IB_QPT_SMI) {
			if ((mad_reg_req->mgmt_class !=
					IB_MGMT_CLASS_SUBN_LID_ROUTED) &&
			    (mad_reg_req->mgmt_class !=
					IB_MGMT_CLASS_SUBN_DIRECTED_ROUTE))
				goto error1;
		} else {
			if ((mad_reg_req->mgmt_class ==
					IB_MGMT_CLASS_SUBN_LID_ROUTED) ||
			    (mad_reg_req->mgmt_class ==
					IB_MGMT_CLASS_SUBN_DIRECTED_ROUTE))
				goto error1;
		}
	} else {
		/* No registration request supplied */
		if (!send_handler)
			goto error1;
	}

	/* Validate device and port */
	port_priv = ib_get_mad_port(device, port_num);
	if (!port_priv) {
		ret = ERR_PTR(-ENODEV);
		goto error1;
	}

	/* Allocate structures */
	mad_agent_priv = kzalloc(sizeof *mad_agent_priv, GFP_KERNEL);
	if (!mad_agent_priv) {
		ret = ERR_PTR(-ENOMEM);
		goto error1;
	}

	mad_agent_priv->agent.mr = ib_get_dma_mr(port_priv->qp_info[qpn].qp->pd,
						 IB_ACCESS_LOCAL_WRITE);
	if (IS_ERR(mad_agent_priv->agent.mr)) {
		ret = ERR_PTR(-ENOMEM);
		goto error2;
	}

	if (mad_reg_req) {
		reg_req = kmemdup(mad_reg_req, sizeof *reg_req, GFP_KERNEL);
		if (!reg_req) {
			ret = ERR_PTR(-ENOMEM);
			goto error3;
		}
	}

	/* Now, fill in the various structures */
	mad_agent_priv->qp_info = &port_priv->qp_info[qpn];
	mad_agent_priv->reg_req = reg_req;
	mad_agent_priv->agent.rmpp_version = rmpp_version;
	mad_agent_priv->agent.device = device;
	mad_agent_priv->agent.recv_handler = recv_handler;
	mad_agent_priv->agent.send_handler = send_handler;
	mad_agent_priv->agent.context = context;
	mad_agent_priv->agent.qp = port_priv->qp_info[qpn].qp;
	mad_agent_priv->agent.port_num = port_num;
	spin_lock_init(&mad_agent_priv->lock);
	INIT_LIST_HEAD(&mad_agent_priv->send_list);
	INIT_LIST_HEAD(&mad_agent_priv->wait_list);
	INIT_LIST_HEAD(&mad_agent_priv->done_list);
	INIT_LIST_HEAD(&mad_agent_priv->rmpp_list);
	INIT_DELAYED_WORK(&mad_agent_priv->timed_work, timeout_sends);
	INIT_LIST_HEAD(&mad_agent_priv->local_list);
	INIT_WORK(&mad_agent_priv->local_work, local_completions);
	atomic_set(&mad_agent_priv->refcount, 1);
	init_completion(&mad_agent_priv->comp);

	spin_lock_irqsave(&port_priv->reg_lock, flags);
	mad_agent_priv->agent.hi_tid = ++ib_mad_client_id;

	/*
	 * Make sure MAD registration (if supplied)
	 * is non overlapping with any existing ones
	 */
	if (mad_reg_req) {
		mgmt_class = convert_mgmt_class(mad_reg_req->mgmt_class);
		if (!is_vendor_class(mgmt_class)) {
			class = port_priv->version[mad_reg_req->
						   mgmt_class_version].class;
			if (class) {
				method = class->method_table[mgmt_class];
				if (method) {
					if (method_in_use(&method,
							   mad_reg_req))
						goto error4;
				}
			}
			ret2 = add_nonoui_reg_req(mad_reg_req, mad_agent_priv,
						  mgmt_class);
		} else {
			/* "New" vendor class range */
			vendor = port_priv->version[mad_reg_req->
						    mgmt_class_version].vendor;
			if (vendor) {
				vclass = vendor_class_index(mgmt_class);
				vendor_class = vendor->vendor_class[vclass];
				if (vendor_class) {
					if (is_vendor_method_in_use(
							vendor_class,
							mad_reg_req))
						goto error4;
				}
			}
			ret2 = add_oui_reg_req(mad_reg_req, mad_agent_priv);
		}
		if (ret2) {
			ret = ERR_PTR(ret2);
			goto error4;
		}
	}

	/* Add mad agent into port's agent list */
	list_add_tail(&mad_agent_priv->agent_list, &port_priv->agent_list);
	spin_unlock_irqrestore(&port_priv->reg_lock, flags);

	return &mad_agent_priv->agent;

error4:
	spin_unlock_irqrestore(&port_priv->reg_lock, flags);
	kfree(reg_req);
error3:
	ib_dereg_mr(mad_agent_priv->agent.mr);
error2:
	kfree(mad_agent_priv);
error1:
	return ret;
}
EXPORT_SYMBOL(ib_register_mad_agent);

static inline int is_snooping_sends(int mad_snoop_flags)
{
	return (mad_snoop_flags &
		(/*IB_MAD_SNOOP_POSTED_SENDS |
		 IB_MAD_SNOOP_RMPP_SENDS |*/
		 IB_MAD_SNOOP_SEND_COMPLETIONS /*|
		 IB_MAD_SNOOP_RMPP_SEND_COMPLETIONS*/));
}

static inline int is_snooping_recvs(int mad_snoop_flags)
{
	return (mad_snoop_flags &
		(IB_MAD_SNOOP_RECVS /*|
		 IB_MAD_SNOOP_RMPP_RECVS*/));
}

static int register_snoop_agent(struct ib_mad_qp_info *qp_info,
				struct ib_mad_snoop_private *mad_snoop_priv)
{
	struct ib_mad_snoop_private **new_snoop_table;
	unsigned long flags;
	int i;

	spin_lock_irqsave(&qp_info->snoop_lock, flags);
	/* Check for empty slot in array. */
	for (i = 0; i < qp_info->snoop_table_size; i++)
		if (!qp_info->snoop_table[i])
			break;

	if (i == qp_info->snoop_table_size) {
		/* Grow table. */
		new_snoop_table = krealloc(qp_info->snoop_table,
					   sizeof mad_snoop_priv *
					   (qp_info->snoop_table_size + 1),
					   GFP_ATOMIC);
		if (!new_snoop_table) {
			i = -ENOMEM;
			goto out;
		}

		qp_info->snoop_table = new_snoop_table;
		qp_info->snoop_table_size++;
	}
	qp_info->snoop_table[i] = mad_snoop_priv;
	atomic_inc(&qp_info->snoop_count);
out:
	spin_unlock_irqrestore(&qp_info->snoop_lock, flags);
	return i;
}

struct ib_mad_agent *ib_register_mad_snoop(struct ib_device *device,
					   u8 port_num,
					   enum ib_qp_type qp_type,
					   int mad_snoop_flags,
					   ib_mad_snoop_handler snoop_handler,
					   ib_mad_recv_handler recv_handler,
					   void *context)
{
	struct ib_mad_port_private *port_priv;
	struct ib_mad_agent *ret;
	struct ib_mad_snoop_private *mad_snoop_priv;
	int qpn;

	/* Validate parameters */
	if ((is_snooping_sends(mad_snoop_flags) && !snoop_handler) ||
	    (is_snooping_recvs(mad_snoop_flags) && !recv_handler)) {
		ret = ERR_PTR(-EINVAL);
		goto error1;
	}
	qpn = get_spl_qp_index(qp_type);
	if (qpn == -1) {
		ret = ERR_PTR(-EINVAL);
		goto error1;
	}
	port_priv = ib_get_mad_port(device, port_num);
	if (!port_priv) {
		ret = ERR_PTR(-ENODEV);
		goto error1;
	}
	/* Allocate structures */
	mad_snoop_priv = kzalloc(sizeof *mad_snoop_priv, GFP_KERNEL);
	if (!mad_snoop_priv) {
		ret = ERR_PTR(-ENOMEM);
		goto error1;
	}

	/* Now, fill in the various structures */
	mad_snoop_priv->qp_info = &port_priv->qp_info[qpn];
	mad_snoop_priv->agent.device = device;
	mad_snoop_priv->agent.recv_handler = recv_handler;
	mad_snoop_priv->agent.snoop_handler = snoop_handler;
	mad_snoop_priv->agent.context = context;
	mad_snoop_priv->agent.qp = port_priv->qp_info[qpn].qp;
	mad_snoop_priv->agent.port_num = port_num;
	mad_snoop_priv->mad_snoop_flags = mad_snoop_flags;
	init_completion(&mad_snoop_priv->comp);
	mad_snoop_priv->snoop_index = register_snoop_agent(
						&port_priv->qp_info[qpn],
						mad_snoop_priv);
	if (mad_snoop_priv->snoop_index < 0) {
		ret = ERR_PTR(mad_snoop_priv->snoop_index);
		goto error2;
	}

	atomic_set(&mad_snoop_priv->refcount, 1);
	return &mad_snoop_priv->agent;

error2:
	kfree(mad_snoop_priv);
error1:
	return ret;
}
EXPORT_SYMBOL(ib_register_mad_snoop);

static inline void deref_mad_agent(struct ib_mad_agent_private *mad_agent_priv)
{
	if (atomic_dec_and_test(&mad_agent_priv->refcount))
		complete(&mad_agent_priv->comp);
}

static inline void deref_snoop_agent(struct ib_mad_snoop_private *mad_snoop_priv)
{
	if (atomic_dec_and_test(&mad_snoop_priv->refcount))
		complete(&mad_snoop_priv->comp);
}

static void unregister_mad_agent(struct ib_mad_agent_private *mad_agent_priv)
{
	struct ib_mad_port_private *port_priv;
	unsigned long flags;

	/* Note that we could still be handling received MADs */

	/*
	 * Canceling all sends results in dropping received response
	 * MADs, preventing us from queuing additional work
	 */
	cancel_mads(mad_agent_priv);
	port_priv = mad_agent_priv->qp_info->port_priv;
	cancel_delayed_work(&mad_agent_priv->timed_work);

	spin_lock_irqsave(&port_priv->reg_lock, flags);
	remove_mad_reg_req(mad_agent_priv);
	list_del(&mad_agent_priv->agent_list);
	spin_unlock_irqrestore(&port_priv->reg_lock, flags);

	flush_workqueue(port_priv->wq);
	ib_cancel_rmpp_recvs(mad_agent_priv);

	deref_mad_agent(mad_agent_priv);
	wait_for_completion(&mad_agent_priv->comp);

	kfree(mad_agent_priv->reg_req);
	ib_dereg_mr(mad_agent_priv->agent.mr);
	kfree(mad_agent_priv);
}

static void unregister_mad_snoop(struct ib_mad_snoop_private *mad_snoop_priv)
{
	struct ib_mad_qp_info *qp_info;
	unsigned long flags;

	qp_info = mad_snoop_priv->qp_info;
	spin_lock_irqsave(&qp_info->snoop_lock, flags);
	qp_info->snoop_table[mad_snoop_priv->snoop_index] = NULL;
	atomic_dec(&qp_info->snoop_count);
	spin_unlock_irqrestore(&qp_info->snoop_lock, flags);

	deref_snoop_agent(mad_snoop_priv);
	wait_for_completion(&mad_snoop_priv->comp);

	kfree(mad_snoop_priv);
}

/*
 * ib_unregister_mad_agent - Unregisters a client from using MAD services
 */
int ib_unregister_mad_agent(struct ib_mad_agent *mad_agent)
{
	struct ib_mad_agent_private *mad_agent_priv;
	struct ib_mad_snoop_private *mad_snoop_priv;

	/* If the TID is zero, the agent can only snoop. */
	if (mad_agent->hi_tid) {
		mad_agent_priv = container_of(mad_agent,
					      struct ib_mad_agent_private,
					      agent);
		unregister_mad_agent(mad_agent_priv);
	} else {
		mad_snoop_priv = container_of(mad_agent,
					      struct ib_mad_snoop_private,
					      agent);
		unregister_mad_snoop(mad_snoop_priv);
	}
	return 0;
}
EXPORT_SYMBOL(ib_unregister_mad_agent);

static void dequeue_mad(struct ib_mad_list_head *mad_list)
{
	struct ib_mad_queue *mad_queue;
	unsigned long flags;

	BUG_ON(!mad_list->mad_queue);
	mad_queue = mad_list->mad_queue;
	spin_lock_irqsave(&mad_queue->lock, flags);
	list_del(&mad_list->list);
	mad_queue->count--;
	spin_unlock_irqrestore(&mad_queue->lock, flags);
}

static void snoop_send(struct ib_mad_qp_info *qp_info,
		       struct ib_mad_send_buf *send_buf,
		       struct ib_mad_send_wc *mad_send_wc,
		       int mad_snoop_flags)
{
	struct ib_mad_snoop_private *mad_snoop_priv;
	unsigned long flags;
	int i;

	spin_lock_irqsave(&qp_info->snoop_lock, flags);
	for (i = 0; i < qp_info->snoop_table_size; i++) {
		mad_snoop_priv = qp_info->snoop_table[i];
		if (!mad_snoop_priv ||
		    !(mad_snoop_priv->mad_snoop_flags & mad_snoop_flags))
			continue;

		atomic_inc(&mad_snoop_priv->refcount);
		spin_unlock_irqrestore(&qp_info->snoop_lock, flags);
		mad_snoop_priv->agent.snoop_handler(&mad_snoop_priv->agent,
						    send_buf, mad_send_wc);
		deref_snoop_agent(mad_snoop_priv);
		spin_lock_irqsave(&qp_info->snoop_lock, flags);
	}
	spin_unlock_irqrestore(&qp_info->snoop_lock, flags);
}

static void snoop_recv(struct ib_mad_qp_info *qp_info,
		       struct ib_mad_recv_wc *mad_recv_wc,
		       int mad_snoop_flags)
{
	struct ib_mad_snoop_private *mad_snoop_priv;
	unsigned long flags;
	int i;

	spin_lock_irqsave(&qp_info->snoop_lock, flags);
	for (i = 0; i < qp_info->snoop_table_size; i++) {
		mad_snoop_priv = qp_info->snoop_table[i];
		if (!mad_snoop_priv ||
		    !(mad_snoop_priv->mad_snoop_flags & mad_snoop_flags))
			continue;

		atomic_inc(&mad_snoop_priv->refcount);
		spin_unlock_irqrestore(&qp_info->snoop_lock, flags);
		mad_snoop_priv->agent.recv_handler(&mad_snoop_priv->agent,
						   mad_recv_wc);
		deref_snoop_agent(mad_snoop_priv);
		spin_lock_irqsave(&qp_info->snoop_lock, flags);
	}
	spin_unlock_irqrestore(&qp_info->snoop_lock, flags);
}

static void build_smp_wc(struct ib_qp *qp,
			 u64 wr_id, u16 slid, u16 pkey_index, u8 port_num,
			 struct ib_wc *wc)
{
	memset(wc, 0, sizeof *wc);
	wc->wr_id = wr_id;
	wc->status = IB_WC_SUCCESS;
	wc->opcode = IB_WC_RECV;
	wc->pkey_index = pkey_index;
	wc->byte_len = sizeof(struct ib_mad) + sizeof(struct ib_grh);
	wc->src_qp = IB_QP0;
	wc->qp = qp;
	wc->slid = slid;
	wc->sl = 0;
	wc->dlid_path_bits = 0;
	wc->port_num = port_num;
}

/*
 * Return 0 if SMP is to be sent
 * Return 1 if SMP was consumed locally (whether or not solicited)
 * Return < 0 if error
 */
static int handle_outgoing_dr_smp(struct ib_mad_agent_private *mad_agent_priv,
				  struct ib_mad_send_wr_private *mad_send_wr)
{
	int ret = 0;
	struct ib_smp *smp = mad_send_wr->send_buf.mad;
	unsigned long flags;
	struct ib_mad_local_private *local;
	struct ib_mad_private *mad_priv;
	struct ib_mad_port_private *port_priv;
	struct ib_mad_agent_private *recv_mad_agent = NULL;
	struct ib_device *device = mad_agent_priv->agent.device;
	u8 port_num;
	struct ib_wc mad_wc;
	struct ib_send_wr *send_wr = &mad_send_wr->send_wr;

	if (device->node_type == RDMA_NODE_IB_SWITCH &&
	    smp->mgmt_class == IB_MGMT_CLASS_SUBN_DIRECTED_ROUTE)
		port_num = send_wr->wr.ud.port_num;
	else
		port_num = mad_agent_priv->agent.port_num;

	/*
	 * Directed route handling starts if the initial LID routed part of
	 * a request or the ending LID routed part of a response is empty.
	 * If we are at the start of the LID routed part, don't update the
	 * hop_ptr or hop_cnt.  See section 14.2.2, Vol 1 IB spec.
	 */
	if ((ib_get_smp_direction(smp) ? smp->dr_dlid : smp->dr_slid) ==
	     IB_LID_PERMISSIVE &&
	     smi_handle_dr_smp_send(smp, device->node_type, port_num) ==
	     IB_SMI_DISCARD) {
		ret = -EINVAL;
		printk(KERN_ERR PFX "Invalid directed route\n");
		goto out;
	}

	/* Check to post send on QP or process locally */
	if (smi_check_local_smp(smp, device) == IB_SMI_DISCARD &&
	    smi_check_local_returning_smp(smp, device) == IB_SMI_DISCARD)
		goto out;

	local = kmalloc(sizeof *local, GFP_ATOMIC);
	if (!local) {
		ret = -ENOMEM;
		printk(KERN_ERR PFX "No memory for ib_mad_local_private\n");
		goto out;
	}
	local->mad_priv = NULL;
	local->recv_mad_agent = NULL;
	mad_priv = kmem_cache_alloc(ib_mad_cache, GFP_ATOMIC);
	if (!mad_priv) {
		ret = -ENOMEM;
		printk(KERN_ERR PFX "No memory for local response MAD\n");
		kfree(local);
		goto out;
	}

	build_smp_wc(mad_agent_priv->agent.qp,
		     send_wr->wr_id, be16_to_cpu(smp->dr_slid),
		     send_wr->wr.ud.pkey_index,
		     send_wr->wr.ud.port_num, &mad_wc);

	/* No GRH for DR SMP */
	ret = device->process_mad(device, 0, port_num, &mad_wc, NULL,
				  (struct ib_mad *)smp,
				  (struct ib_mad *)&mad_priv->mad);
	switch (ret)
	{
	case IB_MAD_RESULT_SUCCESS | IB_MAD_RESULT_REPLY:
		if (ib_response_mad(&mad_priv->mad.mad) &&
		    mad_agent_priv->agent.recv_handler) {
			local->mad_priv = mad_priv;
			local->recv_mad_agent = mad_agent_priv;
			/*
			 * Reference MAD agent until receive
			 * side of local completion handled
			 */
			atomic_inc(&mad_agent_priv->refcount);
		} else
			kmem_cache_free(ib_mad_cache, mad_priv);
		break;
	case IB_MAD_RESULT_SUCCESS | IB_MAD_RESULT_CONSUMED:
		kmem_cache_free(ib_mad_cache, mad_priv);
		break;
	case IB_MAD_RESULT_SUCCESS:
		/* Treat like an incoming receive MAD */
		port_priv = ib_get_mad_port(mad_agent_priv->agent.device,
					    mad_agent_priv->agent.port_num);
		if (port_priv) {
			memcpy(&mad_priv->mad.mad, smp, sizeof(struct ib_mad));
			recv_mad_agent = find_mad_agent(port_priv,
						        &mad_priv->mad.mad);
		}
		if (!port_priv || !recv_mad_agent) {
			/*
			 * No receiving agent so drop packet and
			 * generate send completion.
			 */
			kmem_cache_free(ib_mad_cache, mad_priv);
			break;
		}
		local->mad_priv = mad_priv;
		local->recv_mad_agent = recv_mad_agent;
		break;
	default:
		kmem_cache_free(ib_mad_cache, mad_priv);
		kfree(local);
		ret = -EINVAL;
		goto out;
	}

	local->mad_send_wr = mad_send_wr;
	/* Reference MAD agent until send side of local completion handled */
	atomic_inc(&mad_agent_priv->refcount);
	/* Queue local completion to local list */
	spin_lock_irqsave(&mad_agent_priv->lock, flags);
	list_add_tail(&local->completion_list, &mad_agent_priv->local_list);
	spin_unlock_irqrestore(&mad_agent_priv->lock, flags);
	queue_work(mad_agent_priv->qp_info->port_priv->wq,
		   &mad_agent_priv->local_work);
	ret = 1;
out:
	return ret;
}

static int get_pad_size(int hdr_len, int data_len)
{
	int seg_size, pad;

	seg_size = sizeof(struct ib_mad) - hdr_len;
	if (data_len && seg_size) {
		pad = seg_size - data_len % seg_size;
		return pad == seg_size ? 0 : pad;
	} else
		return seg_size;
}

static void free_send_rmpp_list(struct ib_mad_send_wr_private *mad_send_wr)
{
	struct ib_rmpp_segment *s, *t;

	list_for_each_entry_safe(s, t, &mad_send_wr->rmpp_list, list) {
		list_del(&s->list);
		kfree(s);
	}
}

static int alloc_send_rmpp_list(struct ib_mad_send_wr_private *send_wr,
				gfp_t gfp_mask)
{
	struct ib_mad_send_buf *send_buf = &send_wr->send_buf;
	struct ib_rmpp_mad *rmpp_mad = send_buf->mad;
	struct ib_rmpp_segment *seg = NULL;
	int left, seg_size, pad;

	send_buf->seg_size = sizeof (struct ib_mad) - send_buf->hdr_len;
	seg_size = send_buf->seg_size;
	pad = send_wr->pad;

	/* Allocate data segments. */
	for (left = send_buf->data_len + pad; left > 0; left -= seg_size) {
		seg = kmalloc(sizeof (*seg) + seg_size, gfp_mask);
		if (!seg) {
			printk(KERN_ERR "alloc_send_rmpp_segs: RMPP mem "
			       "alloc failed for len %zd, gfp %#x\n",
			       sizeof (*seg) + seg_size, gfp_mask);
			free_send_rmpp_list(send_wr);
			return -ENOMEM;
		}
		seg->num = ++send_buf->seg_count;
		list_add_tail(&seg->list, &send_wr->rmpp_list);
	}

	/* Zero any padding */
	if (pad)
		memset(seg->data + seg_size - pad, 0, pad);

	rmpp_mad->rmpp_hdr.rmpp_version = send_wr->mad_agent_priv->
					  agent.rmpp_version;
	rmpp_mad->rmpp_hdr.rmpp_type = IB_MGMT_RMPP_TYPE_DATA;
	ib_set_rmpp_flags(&rmpp_mad->rmpp_hdr, IB_MGMT_RMPP_FLAG_ACTIVE);

	send_wr->cur_seg = container_of(send_wr->rmpp_list.next,
					struct ib_rmpp_segment, list);
	send_wr->last_ack_seg = send_wr->cur_seg;
	return 0;
}

struct ib_mad_send_buf * ib_create_send_mad(struct ib_mad_agent *mad_agent,
					    u32 remote_qpn, u16 pkey_index,
					    int rmpp_active,
					    int hdr_len, int data_len,
					    gfp_t gfp_mask)
{
	struct ib_mad_agent_private *mad_agent_priv;
	struct ib_mad_send_wr_private *mad_send_wr;
	int pad, message_size, ret, size;
	void *buf;

	mad_agent_priv = container_of(mad_agent, struct ib_mad_agent_private,
				      agent);
	pad = get_pad_size(hdr_len, data_len);
	message_size = hdr_len + data_len + pad;

	if ((!mad_agent->rmpp_version &&
	     (rmpp_active || message_size > sizeof(struct ib_mad))) ||
	    (!rmpp_active && message_size > sizeof(struct ib_mad)))
		return ERR_PTR(-EINVAL);

	size = rmpp_active ? hdr_len : sizeof(struct ib_mad);
	buf = kzalloc(sizeof *mad_send_wr + size, gfp_mask);
	if (!buf)
		return ERR_PTR(-ENOMEM);

	mad_send_wr = buf + size;
	INIT_LIST_HEAD(&mad_send_wr->rmpp_list);
	mad_send_wr->send_buf.mad = buf;
	mad_send_wr->send_buf.hdr_len = hdr_len;
	mad_send_wr->send_buf.data_len = data_len;
	mad_send_wr->pad = pad;

	mad_send_wr->mad_agent_priv = mad_agent_priv;
	mad_send_wr->sg_list[0].length = hdr_len;
	mad_send_wr->sg_list[0].lkey = mad_agent->mr->lkey;
	mad_send_wr->sg_list[1].length = sizeof(struct ib_mad) - hdr_len;
	mad_send_wr->sg_list[1].lkey = mad_agent->mr->lkey;

	mad_send_wr->send_wr.wr_id = (unsigned long) mad_send_wr;
	mad_send_wr->send_wr.sg_list = mad_send_wr->sg_list;
	mad_send_wr->send_wr.num_sge = 2;
	mad_send_wr->send_wr.opcode = IB_WR_SEND;
	mad_send_wr->send_wr.send_flags = IB_SEND_SIGNALED;
	mad_send_wr->send_wr.wr.ud.remote_qpn = remote_qpn;
	mad_send_wr->send_wr.wr.ud.remote_qkey = IB_QP_SET_QKEY;
	mad_send_wr->send_wr.wr.ud.pkey_index = pkey_index;

	if (rmpp_active) {
		ret = alloc_send_rmpp_list(mad_send_wr, gfp_mask);
		if (ret) {
			kfree(buf);
			return ERR_PTR(ret);
		}
	}

	mad_send_wr->send_buf.mad_agent = mad_agent;
	atomic_inc(&mad_agent_priv->refcount);
	return &mad_send_wr->send_buf;
}
EXPORT_SYMBOL(ib_create_send_mad);

int ib_get_mad_data_offset(u8 mgmt_class)
{
	if (mgmt_class == IB_MGMT_CLASS_SUBN_ADM)
		return IB_MGMT_SA_HDR;
	else if ((mgmt_class == IB_MGMT_CLASS_DEVICE_MGMT) ||
		 (mgmt_class == IB_MGMT_CLASS_DEVICE_ADM) ||
		 (mgmt_class == IB_MGMT_CLASS_BIS))
		return IB_MGMT_DEVICE_HDR;
	else if ((mgmt_class >= IB_MGMT_CLASS_VENDOR_RANGE2_START) &&
		 (mgmt_class <= IB_MGMT_CLASS_VENDOR_RANGE2_END))
		return IB_MGMT_VENDOR_HDR;
	else
		return IB_MGMT_MAD_HDR;
}
EXPORT_SYMBOL(ib_get_mad_data_offset);

int ib_is_mad_class_rmpp(u8 mgmt_class)
{
	if ((mgmt_class == IB_MGMT_CLASS_SUBN_ADM) ||
	    (mgmt_class == IB_MGMT_CLASS_DEVICE_MGMT) ||
	    (mgmt_class == IB_MGMT_CLASS_DEVICE_ADM) ||
	    (mgmt_class == IB_MGMT_CLASS_BIS) ||
	    ((mgmt_class >= IB_MGMT_CLASS_VENDOR_RANGE2_START) &&
	     (mgmt_class <= IB_MGMT_CLASS_VENDOR_RANGE2_END)))
		return 1;
	return 0;
}
EXPORT_SYMBOL(ib_is_mad_class_rmpp);

void *ib_get_rmpp_segment(struct ib_mad_send_buf *send_buf, int seg_num)
{
	struct ib_mad_send_wr_private *mad_send_wr;
	struct list_head *list;

	mad_send_wr = container_of(send_buf, struct ib_mad_send_wr_private,
				   send_buf);
	list = &mad_send_wr->cur_seg->list;

	if (mad_send_wr->cur_seg->num < seg_num) {
		list_for_each_entry(mad_send_wr->cur_seg, list, list)
			if (mad_send_wr->cur_seg->num == seg_num)
				break;
	} else if (mad_send_wr->cur_seg->num > seg_num) {
		list_for_each_entry_reverse(mad_send_wr->cur_seg, list, list)
			if (mad_send_wr->cur_seg->num == seg_num)
				break;
	}
	return mad_send_wr->cur_seg->data;
}
EXPORT_SYMBOL(ib_get_rmpp_segment);

static inline void *ib_get_payload(struct ib_mad_send_wr_private *mad_send_wr)
{
	if (mad_send_wr->send_buf.seg_count)
		return ib_get_rmpp_segment(&mad_send_wr->send_buf,
					   mad_send_wr->seg_num);
	else
		return mad_send_wr->send_buf.mad +
		       mad_send_wr->send_buf.hdr_len;
}

void ib_free_send_mad(struct ib_mad_send_buf *send_buf)
{
	struct ib_mad_agent_private *mad_agent_priv;
	struct ib_mad_send_wr_private *mad_send_wr;

	mad_agent_priv = container_of(send_buf->mad_agent,
				      struct ib_mad_agent_private, agent);
	mad_send_wr = container_of(send_buf, struct ib_mad_send_wr_private,
				   send_buf);

	free_send_rmpp_list(mad_send_wr);
	kfree(send_buf->mad);
	deref_mad_agent(mad_agent_priv);
}
EXPORT_SYMBOL(ib_free_send_mad);

int ib_send_mad(struct ib_mad_send_wr_private *mad_send_wr)
{
	struct ib_mad_qp_info *qp_info;
	struct list_head *list;
	struct ib_send_wr *bad_send_wr;
	struct ib_mad_agent *mad_agent;
	struct ib_sge *sge;
	unsigned long flags;
	int ret;

	/* Set WR ID to find mad_send_wr upon completion */
	qp_info = mad_send_wr->mad_agent_priv->qp_info;
	mad_send_wr->send_wr.wr_id = (unsigned long)&mad_send_wr->mad_list;
	mad_send_wr->mad_list.mad_queue = &qp_info->send_queue;

	mad_agent = mad_send_wr->send_buf.mad_agent;
	sge = mad_send_wr->sg_list;
	sge[0].addr = ib_dma_map_single(mad_agent->device,
					mad_send_wr->send_buf.mad,
					sge[0].length,
					DMA_TO_DEVICE);
	mad_send_wr->header_mapping = sge[0].addr;

	sge[1].addr = ib_dma_map_single(mad_agent->device,
					ib_get_payload(mad_send_wr),
					sge[1].length,
					DMA_TO_DEVICE);
	mad_send_wr->payload_mapping = sge[1].addr;

	spin_lock_irqsave(&qp_info->send_queue.lock, flags);
	if (qp_info->send_queue.count < qp_info->send_queue.max_active) {
		ret = ib_post_send(mad_agent->qp, &mad_send_wr->send_wr,
				   &bad_send_wr);
		list = &qp_info->send_queue.list;
	} else {
		ret = 0;
		list = &qp_info->overflow_list;
	}

	if (!ret) {
		qp_info->send_queue.count++;
		list_add_tail(&mad_send_wr->mad_list.list, list);
	}
	spin_unlock_irqrestore(&qp_info->send_queue.lock, flags);
	if (ret) {
		ib_dma_unmap_single(mad_agent->device,
				    mad_send_wr->header_mapping,
				    sge[0].length, DMA_TO_DEVICE);
		ib_dma_unmap_single(mad_agent->device,
				    mad_send_wr->payload_mapping,
				    sge[1].length, DMA_TO_DEVICE);
	}
	return ret;
}

/*
 * ib_post_send_mad - Posts MAD(s) to the send queue of the QP associated
 *  with the registered client
 */
int ib_post_send_mad(struct ib_mad_send_buf *send_buf,
		     struct ib_mad_send_buf **bad_send_buf)
{
	struct ib_mad_agent_private *mad_agent_priv;
	struct ib_mad_send_buf *next_send_buf;
	struct ib_mad_send_wr_private *mad_send_wr;
	unsigned long flags;
	int ret = -EINVAL;

	/* Walk list of send WRs and post each on send list */
	for (; send_buf; send_buf = next_send_buf) {

		mad_send_wr = container_of(send_buf,
					   struct ib_mad_send_wr_private,
					   send_buf);
		mad_agent_priv = mad_send_wr->mad_agent_priv;

		if (!send_buf->mad_agent->send_handler ||
		    (send_buf->timeout_ms &&
		     !send_buf->mad_agent->recv_handler)) {
			ret = -EINVAL;
			goto error;
		}

		if (!ib_is_mad_class_rmpp(((struct ib_mad_hdr *) send_buf->mad)->mgmt_class)) {
			if (mad_agent_priv->agent.rmpp_version) {
				ret = -EINVAL;
				goto error;
			}
		}

		/*
		 * Save pointer to next work request to post in case the
		 * current one completes, and the user modifies the work
		 * request associated with the completion
		 */
		next_send_buf = send_buf->next;
		mad_send_wr->send_wr.wr.ud.ah = send_buf->ah;

		if (((struct ib_mad_hdr *) send_buf->mad)->mgmt_class ==
		    IB_MGMT_CLASS_SUBN_DIRECTED_ROUTE) {
			ret = handle_outgoing_dr_smp(mad_agent_priv,
						     mad_send_wr);
			if (ret < 0)		/* error */
				goto error;
			else if (ret == 1)	/* locally consumed */
				continue;
		}

		mad_send_wr->tid = ((struct ib_mad_hdr *) send_buf->mad)->tid;
		/* Timeout will be updated after send completes */
		mad_send_wr->timeout = msecs_to_jiffies(send_buf->timeout_ms);
		mad_send_wr->max_retries = send_buf->retries;
		mad_send_wr->retries_left = send_buf->retries;
		send_buf->retries = 0;
		/* Reference for work request to QP + response */
		mad_send_wr->refcount = 1 + (mad_send_wr->timeout > 0);
		mad_send_wr->status = IB_WC_SUCCESS;

		/* Reference MAD agent until send completes */
		atomic_inc(&mad_agent_priv->refcount);
		spin_lock_irqsave(&mad_agent_priv->lock, flags);
		list_add_tail(&mad_send_wr->agent_list,
			      &mad_agent_priv->send_list);
		spin_unlock_irqrestore(&mad_agent_priv->lock, flags);

		if (mad_agent_priv->agent.rmpp_version) {
			ret = ib_send_rmpp_mad(mad_send_wr);
			if (ret >= 0 && ret != IB_RMPP_RESULT_CONSUMED)
				ret = ib_send_mad(mad_send_wr);
		} else
			ret = ib_send_mad(mad_send_wr);
		if (ret < 0) {
			/* Fail send request */
			spin_lock_irqsave(&mad_agent_priv->lock, flags);
			list_del(&mad_send_wr->agent_list);
			spin_unlock_irqrestore(&mad_agent_priv->lock, flags);
			atomic_dec(&mad_agent_priv->refcount);
			goto error;
		}
	}
	return 0;
error:
	if (bad_send_buf)
		*bad_send_buf = send_buf;
	return ret;
}
EXPORT_SYMBOL(ib_post_send_mad);

/*
 * ib_free_recv_mad - Returns data buffers used to receive
 *  a MAD to the access layer
 */
void ib_free_recv_mad(struct ib_mad_recv_wc *mad_recv_wc)
{
	struct ib_mad_recv_buf *mad_recv_buf, *temp_recv_buf;
	struct ib_mad_private_header *mad_priv_hdr;
	struct ib_mad_private *priv;
	struct list_head free_list;

	INIT_LIST_HEAD(&free_list);
	list_splice_init(&mad_recv_wc->rmpp_list, &free_list);

	list_for_each_entry_safe(mad_recv_buf, temp_recv_buf,
					&free_list, list) {
		mad_recv_wc = container_of(mad_recv_buf, struct ib_mad_recv_wc,
					   recv_buf);
		mad_priv_hdr = container_of(mad_recv_wc,
					    struct ib_mad_private_header,
					    recv_wc);
		priv = container_of(mad_priv_hdr, struct ib_mad_private,
				    header);
		kmem_cache_free(ib_mad_cache, priv);
	}
}
EXPORT_SYMBOL(ib_free_recv_mad);

struct ib_mad_agent *ib_redirect_mad_qp(struct ib_qp *qp,
					u8 rmpp_version,
					ib_mad_send_handler send_handler,
					ib_mad_recv_handler recv_handler,
					void *context)
{
	return ERR_PTR(-EINVAL);	/* XXX: for now */
}
EXPORT_SYMBOL(ib_redirect_mad_qp);

int ib_process_mad_wc(struct ib_mad_agent *mad_agent,
		      struct ib_wc *wc)
{
	printk(KERN_ERR PFX "ib_process_mad_wc() not implemented yet\n");
	return 0;
}
EXPORT_SYMBOL(ib_process_mad_wc);

static int method_in_use(struct ib_mad_mgmt_method_table **method,
			 struct ib_mad_reg_req *mad_reg_req)
{
	int i;

	for_each_set_bit(i, mad_reg_req->method_mask, IB_MGMT_MAX_METHODS) {
		if ((*method)->agent[i]) {
			printk(KERN_ERR PFX "Method %d already in use\n", i);
			return -EINVAL;
		}
	}
	return 0;
}

static int allocate_method_table(struct ib_mad_mgmt_method_table **method)
{
	/* Allocate management method table */
	*method = kzalloc(sizeof **method, GFP_ATOMIC);
	if (!*method) {
		printk(KERN_ERR PFX "No memory for "
		       "ib_mad_mgmt_method_table\n");
		return -ENOMEM;
	}

	return 0;
}

/*
 * Check to see if there are any methods still in use
 */
static int check_method_table(struct ib_mad_mgmt_method_table *method)
{
	int i;

	for (i = 0; i < IB_MGMT_MAX_METHODS; i++)
		if (method->agent[i])
			return 1;
	return 0;
}

/*
 * Check to see if there are any method tables for this class still in use
 */
static int check_class_table(struct ib_mad_mgmt_class_table *class)
{
	int i;

	for (i = 0; i < MAX_MGMT_CLASS; i++)
		if (class->method_table[i])
			return 1;
	return 0;
}

static int check_vendor_class(struct ib_mad_mgmt_vendor_class *vendor_class)
{
	int i;

	for (i = 0; i < MAX_MGMT_OUI; i++)
		if (vendor_class->method_table[i])
			return 1;
	return 0;
}

static int find_vendor_oui(struct ib_mad_mgmt_vendor_class *vendor_class,
			   char *oui)
{
	int i;

	for (i = 0; i < MAX_MGMT_OUI; i++)
		/* Is there matching OUI for this vendor class ? */
		if (!memcmp(vendor_class->oui[i], oui, 3))
			return i;

	return -1;
}

static int check_vendor_table(struct ib_mad_mgmt_vendor_class_table *vendor)
{
	int i;

	for (i = 0; i < MAX_MGMT_VENDOR_RANGE2; i++)
		if (vendor->vendor_class[i])
			return 1;

	return 0;
}

static void remove_methods_mad_agent(struct ib_mad_mgmt_method_table *method,
				     struct ib_mad_agent_private *agent)
{
	int i;

	/* Remove any methods for this mad agent */
	for (i = 0; i < IB_MGMT_MAX_METHODS; i++) {
		if (method->agent[i] == agent) {
			method->agent[i] = NULL;
		}
	}
}

static int add_nonoui_reg_req(struct ib_mad_reg_req *mad_reg_req,
			      struct ib_mad_agent_private *agent_priv,
			      u8 mgmt_class)
{
	struct ib_mad_port_private *port_priv;
	struct ib_mad_mgmt_class_table **class;
	struct ib_mad_mgmt_method_table **method;
	int i, ret;

	port_priv = agent_priv->qp_info->port_priv;
	class = &port_priv->version[mad_reg_req->mgmt_class_version].class;
	if (!*class) {
		/* Allocate management class table for "new" class version */
		*class = kzalloc(sizeof **class, GFP_ATOMIC);
		if (!*class) {
			printk(KERN_ERR PFX "No memory for "
			       "ib_mad_mgmt_class_table\n");
			ret = -ENOMEM;
			goto error1;
		}

		/* Allocate method table for this management class */
		method = &(*class)->method_table[mgmt_class];
		if ((ret = allocate_method_table(method)))
			goto error2;
	} else {
		method = &(*class)->method_table[mgmt_class];
		if (!*method) {
			/* Allocate method table for this management class */
			if ((ret = allocate_method_table(method)))
				goto error1;
		}
	}

	/* Now, make sure methods are not already in use */
	if (method_in_use(method, mad_reg_req))
		goto error3;

	/* Finally, add in methods being registered */
	for_each_set_bit(i, mad_reg_req->method_mask, IB_MGMT_MAX_METHODS)
		(*method)->agent[i] = agent_priv;

	return 0;

error3:
	/* Remove any methods for this mad agent */
	remove_methods_mad_agent(*method, agent_priv);
	/* Now, check to see if there are any methods in use */
	if (!check_method_table(*method)) {
		/* If not, release management method table */
		kfree(*method);
		*method = NULL;
	}
	ret = -EINVAL;
	goto error1;
error2:
	kfree(*class);
	*class = NULL;
error1:
	return ret;
}

static int add_oui_reg_req(struct ib_mad_reg_req *mad_reg_req,
			   struct ib_mad_agent_private *agent_priv)
{
	struct ib_mad_port_private *port_priv;
	struct ib_mad_mgmt_vendor_class_table **vendor_table;
	struct ib_mad_mgmt_vendor_class_table *vendor = NULL;
	struct ib_mad_mgmt_vendor_class *vendor_class = NULL;
	struct ib_mad_mgmt_method_table **method;
	int i, ret = -ENOMEM;
	u8 vclass;

	/* "New" vendor (with OUI) class */
	vclass = vendor_class_index(mad_reg_req->mgmt_class);
	port_priv = agent_priv->qp_info->port_priv;
	vendor_table = &port_priv->version[
				mad_reg_req->mgmt_class_version].vendor;
	if (!*vendor_table) {
		/* Allocate mgmt vendor class table for "new" class version */
		vendor = kzalloc(sizeof *vendor, GFP_ATOMIC);
		if (!vendor) {
			printk(KERN_ERR PFX "No memory for "
			       "ib_mad_mgmt_vendor_class_table\n");
			goto error1;
		}

		*vendor_table = vendor;
	}
	if (!(*vendor_table)->vendor_class[vclass]) {
		/* Allocate table for this management vendor class */
		vendor_class = kzalloc(sizeof *vendor_class, GFP_ATOMIC);
		if (!vendor_class) {
			printk(KERN_ERR PFX "No memory for "
			       "ib_mad_mgmt_vendor_class\n");
			goto error2;
		}

		(*vendor_table)->vendor_class[vclass] = vendor_class;
	}
	for (i = 0; i < MAX_MGMT_OUI; i++) {
		/* Is there matching OUI for this vendor class ? */
		if (!memcmp((*vendor_table)->vendor_class[vclass]->oui[i],
			    mad_reg_req->oui, 3)) {
			method = &(*vendor_table)->vendor_class[
						vclass]->method_table[i];
			BUG_ON(!*method);
			goto check_in_use;
		}
	}
	for (i = 0; i < MAX_MGMT_OUI; i++) {
		/* OUI slot available ? */
		if (!is_vendor_oui((*vendor_table)->vendor_class[
				vclass]->oui[i])) {
			method = &(*vendor_table)->vendor_class[
				vclass]->method_table[i];
			BUG_ON(*method);
			/* Allocate method table for this OUI */
			if ((ret = allocate_method_table(method)))
				goto error3;
			memcpy((*vendor_table)->vendor_class[vclass]->oui[i],
			       mad_reg_req->oui, 3);
			goto check_in_use;
		}
	}
	printk(KERN_ERR PFX "All OUI slots in use\n");
	goto error3;

check_in_use:
	/* Now, make sure methods are not already in use */
	if (method_in_use(method, mad_reg_req))
		goto error4;

	/* Finally, add in methods being registered */
	for_each_set_bit(i, mad_reg_req->method_mask, IB_MGMT_MAX_METHODS)
		(*method)->agent[i] = agent_priv;

	return 0;

error4:
	/* Remove any methods for this mad agent */
	remove_methods_mad_agent(*method, agent_priv);
	/* Now, check to see if there are any methods in use */
	if (!check_method_table(*method)) {
		/* If not, release management method table */
		kfree(*method);
		*method = NULL;
	}
	ret = -EINVAL;
error3:
	if (vendor_class) {
		(*vendor_table)->vendor_class[vclass] = NULL;
		kfree(vendor_class);
	}
error2:
	if (vendor) {
		*vendor_table = NULL;
		kfree(vendor);
	}
error1:
	return ret;
}

static void remove_mad_reg_req(struct ib_mad_agent_private *agent_priv)
{
	struct ib_mad_port_private *port_priv;
	struct ib_mad_mgmt_class_table *class;
	struct ib_mad_mgmt_method_table *method;
	struct ib_mad_mgmt_vendor_class_table *vendor;
	struct ib_mad_mgmt_vendor_class *vendor_class;
	int index;
	u8 mgmt_class;

	/*
	 * Was MAD registration request supplied
	 * with original registration ?
	 */
	if (!agent_priv->reg_req) {
		goto out;
	}

	port_priv = agent_priv->qp_info->port_priv;
	mgmt_class = convert_mgmt_class(agent_priv->reg_req->mgmt_class);
	class = port_priv->version[
			agent_priv->reg_req->mgmt_class_version].class;
	if (!class)
		goto vendor_check;

	method = class->method_table[mgmt_class];
	if (method) {
		/* Remove any methods for this mad agent */
		remove_methods_mad_agent(method, agent_priv);
		/* Now, check to see if there are any methods still in use */
		if (!check_method_table(method)) {
			/* If not, release management method table */
			 kfree(method);
			 class->method_table[mgmt_class] = NULL;
			 /* Any management classes left ? */
			if (!check_class_table(class)) {
				/* If not, release management class table */
				kfree(class);
				port_priv->version[
					agent_priv->reg_req->
					mgmt_class_version].class = NULL;
			}
		}
	}

vendor_check:
	if (!is_vendor_class(mgmt_class))
		goto out;

	/* normalize mgmt_class to vendor range 2 */
	mgmt_class = vendor_class_index(agent_priv->reg_req->mgmt_class);
	vendor = port_priv->version[
			agent_priv->reg_req->mgmt_class_version].vendor;

	if (!vendor)
		goto out;

	vendor_class = vendor->vendor_class[mgmt_class];
	if (vendor_class) {
		index = find_vendor_oui(vendor_class, agent_priv->reg_req->oui);
		if (index < 0)
			goto out;
		method = vendor_class->method_table[index];
		if (method) {
			/* Remove any methods for this mad agent */
			remove_methods_mad_agent(method, agent_priv);
			/*
			 * Now, check to see if there are
			 * any methods still in use
			 */
			if (!check_method_table(method)) {
				/* If not, release management method table */
				kfree(method);
				vendor_class->method_table[index] = NULL;
				memset(vendor_class->oui[index], 0, 3);
				/* Any OUIs left ? */
				if (!check_vendor_class(vendor_class)) {
					/* If not, release vendor class table */
					kfree(vendor_class);
					vendor->vendor_class[mgmt_class] = NULL;
					/* Any other vendor classes left ? */
					if (!check_vendor_table(vendor)) {
						kfree(vendor);
						port_priv->version[
							agent_priv->reg_req->
							mgmt_class_version].
							vendor = NULL;
					}
				}
			}
		}
	}

out:
	return;
}

static struct ib_mad_agent_private *
find_mad_agent(struct ib_mad_port_private *port_priv,
	       struct ib_mad *mad)
{
	struct ib_mad_agent_private *mad_agent = NULL;
	unsigned long flags;

	spin_lock_irqsave(&port_priv->reg_lock, flags);
	if (ib_response_mad(mad)) {
		u32 hi_tid;
		struct ib_mad_agent_private *entry;

		/*
		 * Routing is based on high 32 bits of transaction ID
		 * of MAD.
		 */
		hi_tid = be64_to_cpu(mad->mad_hdr.tid) >> 32;
		list_for_each_entry(entry, &port_priv->agent_list, agent_list) {
			if (entry->agent.hi_tid == hi_tid) {
				mad_agent = entry;
				break;
			}
		}
	} else {
		struct ib_mad_mgmt_class_table *class;
		struct ib_mad_mgmt_method_table *method;
		struct ib_mad_mgmt_vendor_class_table *vendor;
		struct ib_mad_mgmt_vendor_class *vendor_class;
		struct ib_vendor_mad *vendor_mad;
		int index;

		/*
		 * Routing is based on version, class, and method
		 * For "newer" vendor MADs, also based on OUI
		 */
		if (mad->mad_hdr.class_version >= MAX_MGMT_VERSION)
			goto out;
		if (!is_vendor_class(mad->mad_hdr.mgmt_class)) {
			class = port_priv->version[
					mad->mad_hdr.class_version].class;
			if (!class)
				goto out;
			method = class->method_table[convert_mgmt_class(
							mad->mad_hdr.mgmt_class)];
			if (method)
				mad_agent = method->agent[mad->mad_hdr.method &
							  ~IB_MGMT_METHOD_RESP];
		} else {
			vendor = port_priv->version[
					mad->mad_hdr.class_version].vendor;
			if (!vendor)
				goto out;
			vendor_class = vendor->vendor_class[vendor_class_index(
						mad->mad_hdr.mgmt_class)];
			if (!vendor_class)
				goto out;
			/* Find matching OUI */
			vendor_mad = (struct ib_vendor_mad *)mad;
			index = find_vendor_oui(vendor_class, vendor_mad->oui);
			if (index == -1)
				goto out;
			method = vendor_class->method_table[index];
			if (method) {
				mad_agent = method->agent[mad->mad_hdr.method &
							  ~IB_MGMT_METHOD_RESP];
			}
		}
	}

	if (mad_agent) {
		if (mad_agent->agent.recv_handler)
			atomic_inc(&mad_agent->refcount);
		else {
			printk(KERN_NOTICE PFX "No receive handler for client "
			       "%p on port %d\n",
			       &mad_agent->agent, port_priv->port_num);
			mad_agent = NULL;
		}
	}
out:
	spin_unlock_irqrestore(&port_priv->reg_lock, flags);

	return mad_agent;
}

static int validate_mad(struct ib_mad *mad, u32 qp_num)
{
	int valid = 0;

	/* Make sure MAD base version is understood */
	if (mad->mad_hdr.base_version != IB_MGMT_BASE_VERSION) {
		printk(KERN_ERR PFX "MAD received with unsupported base "
		       "version %d\n", mad->mad_hdr.base_version);
		goto out;
	}

	/* Filter SMI packets sent to other than QP0 */
	if ((mad->mad_hdr.mgmt_class == IB_MGMT_CLASS_SUBN_LID_ROUTED) ||
	    (mad->mad_hdr.mgmt_class == IB_MGMT_CLASS_SUBN_DIRECTED_ROUTE)) {
		if (qp_num == 0)
			valid = 1;
	} else {
		/* Filter GSI packets sent to QP0 */
		if (qp_num != 0)
			valid = 1;
	}

out:
	return valid;
}

static int is_data_mad(struct ib_mad_agent_private *mad_agent_priv,
		       struct ib_mad_hdr *mad_hdr)
{
	struct ib_rmpp_mad *rmpp_mad;

	rmpp_mad = (struct ib_rmpp_mad *)mad_hdr;
	return !mad_agent_priv->agent.rmpp_version ||
		!(ib_get_rmpp_flags(&rmpp_mad->rmpp_hdr) &
				    IB_MGMT_RMPP_FLAG_ACTIVE) ||
		(rmpp_mad->rmpp_hdr.rmpp_type == IB_MGMT_RMPP_TYPE_DATA);
}

static inline int rcv_has_same_class(struct ib_mad_send_wr_private *wr,
				     struct ib_mad_recv_wc *rwc)
{
	return ((struct ib_mad *)(wr->send_buf.mad))->mad_hdr.mgmt_class ==
		rwc->recv_buf.mad->mad_hdr.mgmt_class;
}

static inline int rcv_has_same_gid(struct ib_mad_agent_private *mad_agent_priv,
				   struct ib_mad_send_wr_private *wr,
				   struct ib_mad_recv_wc *rwc )
{
	struct ib_ah_attr attr;
	u8 send_resp, rcv_resp;
	union ib_gid sgid;
	struct ib_device *device = mad_agent_priv->agent.device;
	u8 port_num = mad_agent_priv->agent.port_num;
	u8 lmc;

	send_resp = ib_response_mad((struct ib_mad *)wr->send_buf.mad);
	rcv_resp = ib_response_mad(rwc->recv_buf.mad);

	if (send_resp == rcv_resp)
		/* both requests, or both responses. GIDs different */
		return 0;

	if (ib_query_ah(wr->send_buf.ah, &attr))
		/* Assume not equal, to avoid false positives. */
		return 0;

	if (!!(attr.ah_flags & IB_AH_GRH) !=
	    !!(rwc->wc->wc_flags & IB_WC_GRH))
		/* one has GID, other does not.  Assume different */
		return 0;

	if (!send_resp && rcv_resp) {
		/* is request/response. */
		if (!(attr.ah_flags & IB_AH_GRH)) {
			if (ib_get_cached_lmc(device, port_num, &lmc))
				return 0;
			return (!lmc || !((attr.src_path_bits ^
					   rwc->wc->dlid_path_bits) &
					  ((1 << lmc) - 1)));
		} else {
			if (ib_get_cached_gid(device, port_num,
					      attr.grh.sgid_index, &sgid))
				return 0;
			return !memcmp(sgid.raw, rwc->recv_buf.grh->dgid.raw,
				       16);
		}
	}

	if (!(attr.ah_flags & IB_AH_GRH))
		return attr.dlid == rwc->wc->slid;
	else
		return !memcmp(attr.grh.dgid.raw, rwc->recv_buf.grh->sgid.raw,
			       16);
}

static inline int is_direct(u8 class)
{
	return (class == IB_MGMT_CLASS_SUBN_DIRECTED_ROUTE);
}

struct ib_mad_send_wr_private*
ib_find_send_mad(struct ib_mad_agent_private *mad_agent_priv,
		 struct ib_mad_recv_wc *wc)
{
	struct ib_mad_send_wr_private *wr;
	struct ib_mad *mad;

	mad = (struct ib_mad *)wc->recv_buf.mad;

	list_for_each_entry(wr, &mad_agent_priv->wait_list, agent_list) {
		if ((wr->tid == mad->mad_hdr.tid) &&
		    rcv_has_same_class(wr, wc) &&
		    /*
		     * Don't check GID for direct routed MADs.
		     * These might have permissive LIDs.
		     */
		    (is_direct(wc->recv_buf.mad->mad_hdr.mgmt_class) ||
		     rcv_has_same_gid(mad_agent_priv, wr, wc)))
			return (wr->status == IB_WC_SUCCESS) ? wr : NULL;
	}

	/*
	 * It's possible to receive the response before we've
	 * been notified that the send has completed
	 */
	list_for_each…