Merge branches 'core', 'cxgb3', 'cxgb4', 'iser', 'iwpm', 'misc', 'mlx4',...

Merge branches 'core', 'cxgb3', 'cxgb4', 'iser', 'iwpm', 'misc', 'mlx4', 'mlx5', 'noio', 'ocrdma', 'qib', 'srp' and 'usnic' into for-next

drivers/infiniband/Makefile

 obj-$(CONFIG_INFINIBAND)		+= core/
-obj-$(CONFIG_INFINIBAND_MTHCA)		+= hw/mthca/
-obj-$(CONFIG_INFINIBAND_IPATH)		+= hw/ipath/
-obj-$(CONFIG_INFINIBAND_QIB)		+= hw/qib/
-obj-$(CONFIG_INFINIBAND_EHCA)		+= hw/ehca/
-obj-$(CONFIG_INFINIBAND_AMSO1100)	+= hw/amso1100/
-obj-$(CONFIG_INFINIBAND_CXGB3)		+= hw/cxgb3/
-obj-$(CONFIG_INFINIBAND_CXGB4)		+= hw/cxgb4/
-obj-$(CONFIG_MLX4_INFINIBAND)		+= hw/mlx4/
-obj-$(CONFIG_MLX5_INFINIBAND)		+= hw/mlx5/
-obj-$(CONFIG_INFINIBAND_NES)		+= hw/nes/
-obj-$(CONFIG_INFINIBAND_OCRDMA)		+= hw/ocrdma/
-obj-$(CONFIG_INFINIBAND_USNIC)		+= hw/usnic/
-obj-$(CONFIG_INFINIBAND_IPOIB)		+= ulp/ipoib/
-obj-$(CONFIG_INFINIBAND_SRP)		+= ulp/srp/
-obj-$(CONFIG_INFINIBAND_SRPT)		+= ulp/srpt/
-obj-$(CONFIG_INFINIBAND_ISER)		+= ulp/iser/
-obj-$(CONFIG_INFINIBAND_ISERT)		+= ulp/isert/
+obj-$(CONFIG_INFINIBAND)		+= hw/
+obj-$(CONFIG_INFINIBAND)		+= ulp/

drivers/infiniband/core/Makefile

@@ -18,7 +18,7 @@ ib_sa-y :=			sa_query.o multicast.o
 
 ib_cm-y :=			cm.o
 
-iw_cm-y :=			iwcm.o
+iw_cm-y :=			iwcm.o iwpm_util.o iwpm_msg.o
 
 rdma_cm-y :=			cma.o
 

drivers/infiniband/core/cma.c

@@ -3607,7 +3607,8 @@ static int cma_get_id_stats(struct sk_buff *skb, struct netlink_callback *cb)
 
 			id_stats = ibnl_put_msg(skb, &nlh, cb->nlh->nlmsg_seq,
 						sizeof *id_stats, RDMA_NL_RDMA_CM,
-						RDMA_NL_RDMA_CM_ID_STATS);
+						RDMA_NL_RDMA_CM_ID_STATS,
+						NLM_F_MULTI);
 			if (!id_stats)
 				goto out;
 

drivers/infiniband/core/iwpm_util.c

+/*
+ * Copyright (c) 2014 Chelsio, Inc. All rights reserved.
+ * Copyright (c) 2014 Intel Corporation. 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 "iwpm_util.h"
+
+#define IWPM_HASH_BUCKET_SIZE	512
+#define IWPM_HASH_BUCKET_MASK	(IWPM_HASH_BUCKET_SIZE - 1)
+
+static LIST_HEAD(iwpm_nlmsg_req_list);
+static DEFINE_SPINLOCK(iwpm_nlmsg_req_lock);
+
+static struct hlist_head *iwpm_hash_bucket;
+static DEFINE_SPINLOCK(iwpm_mapinfo_lock);
+
+static DEFINE_MUTEX(iwpm_admin_lock);
+static struct iwpm_admin_data iwpm_admin;
+
+int iwpm_init(u8 nl_client)
+{
+	if (iwpm_valid_client(nl_client))
+		return -EINVAL;
+	mutex_lock(&iwpm_admin_lock);
+	if (atomic_read(&iwpm_admin.refcount) == 0) {
+		iwpm_hash_bucket = kzalloc(IWPM_HASH_BUCKET_SIZE *
+					sizeof(struct hlist_head), GFP_KERNEL);
+		if (!iwpm_hash_bucket) {
+			mutex_unlock(&iwpm_admin_lock);
+			pr_err("%s Unable to create mapinfo hash table\n", __func__);
+			return -ENOMEM;
+		}
+	}
+	atomic_inc(&iwpm_admin.refcount);
+	mutex_unlock(&iwpm_admin_lock);
+	iwpm_set_valid(nl_client, 1);
+	return 0;
+}
+EXPORT_SYMBOL(iwpm_init);
+
+static void free_hash_bucket(void);
+
+int iwpm_exit(u8 nl_client)
+{
+
+	if (!iwpm_valid_client(nl_client))
+		return -EINVAL;
+	mutex_lock(&iwpm_admin_lock);
+	if (atomic_read(&iwpm_admin.refcount) == 0) {
+		mutex_unlock(&iwpm_admin_lock);
+		pr_err("%s Incorrect usage - negative refcount\n", __func__);
+		return -EINVAL;
+	}
+	if (atomic_dec_and_test(&iwpm_admin.refcount)) {
+		free_hash_bucket();
+		pr_debug("%s: Mapinfo hash table is destroyed\n", __func__);
+	}
+	mutex_unlock(&iwpm_admin_lock);
+	iwpm_set_valid(nl_client, 0);
+	return 0;
+}
+EXPORT_SYMBOL(iwpm_exit);
+
+static struct hlist_head *get_hash_bucket_head(struct sockaddr_storage *,
+					       struct sockaddr_storage *);
+
+int iwpm_create_mapinfo(struct sockaddr_storage *local_sockaddr,
+			struct sockaddr_storage *mapped_sockaddr,
+			u8 nl_client)
+{
+	struct hlist_head *hash_bucket_head;
+	struct iwpm_mapping_info *map_info;
+	unsigned long flags;
+
+	if (!iwpm_valid_client(nl_client))
+		return -EINVAL;
+	map_info = kzalloc(sizeof(struct iwpm_mapping_info), GFP_KERNEL);
+	if (!map_info) {
+		pr_err("%s: Unable to allocate a mapping info\n", __func__);
+		return -ENOMEM;
+	}
+	memcpy(&map_info->local_sockaddr, local_sockaddr,
+	       sizeof(struct sockaddr_storage));
+	memcpy(&map_info->mapped_sockaddr, mapped_sockaddr,
+	       sizeof(struct sockaddr_storage));
+	map_info->nl_client = nl_client;
+
+	spin_lock_irqsave(&iwpm_mapinfo_lock, flags);
+	if (iwpm_hash_bucket) {
+		hash_bucket_head = get_hash_bucket_head(
+					&map_info->local_sockaddr,
+					&map_info->mapped_sockaddr);
+		hlist_add_head(&map_info->hlist_node, hash_bucket_head);
+	}
+	spin_unlock_irqrestore(&iwpm_mapinfo_lock, flags);
+	return 0;
+}
+EXPORT_SYMBOL(iwpm_create_mapinfo);
+
+int iwpm_remove_mapinfo(struct sockaddr_storage *local_sockaddr,
+			struct sockaddr_storage *mapped_local_addr)
+{
+	struct hlist_node *tmp_hlist_node;
+	struct hlist_head *hash_bucket_head;
+	struct iwpm_mapping_info *map_info = NULL;
+	unsigned long flags;
+	int ret = -EINVAL;
+
+	spin_lock_irqsave(&iwpm_mapinfo_lock, flags);
+	if (iwpm_hash_bucket) {
+		hash_bucket_head = get_hash_bucket_head(
+					local_sockaddr,
+					mapped_local_addr);
+		hlist_for_each_entry_safe(map_info, tmp_hlist_node,
+					hash_bucket_head, hlist_node) {
+
+			if (!iwpm_compare_sockaddr(&map_info->mapped_sockaddr,
+						mapped_local_addr)) {
+
+				hlist_del_init(&map_info->hlist_node);
+				kfree(map_info);
+				ret = 0;
+				break;
+			}
+		}
+	}
+	spin_unlock_irqrestore(&iwpm_mapinfo_lock, flags);
+	return ret;
+}
+EXPORT_SYMBOL(iwpm_remove_mapinfo);
+
+static void free_hash_bucket(void)
+{
+	struct hlist_node *tmp_hlist_node;
+	struct iwpm_mapping_info *map_info;
+	unsigned long flags;
+	int i;
+
+	/* remove all the mapinfo data from the list */
+	spin_lock_irqsave(&iwpm_mapinfo_lock, flags);
+	for (i = 0; i < IWPM_HASH_BUCKET_SIZE; i++) {
+		hlist_for_each_entry_safe(map_info, tmp_hlist_node,
+			&iwpm_hash_bucket[i], hlist_node) {
+
+				hlist_del_init(&map_info->hlist_node);
+				kfree(map_info);
+			}
+	}
+	/* free the hash list */
+	kfree(iwpm_hash_bucket);
+	iwpm_hash_bucket = NULL;
+	spin_unlock_irqrestore(&iwpm_mapinfo_lock, flags);
+}
+
+struct iwpm_nlmsg_request *iwpm_get_nlmsg_request(__u32 nlmsg_seq,
+					u8 nl_client, gfp_t gfp)
+{
+	struct iwpm_nlmsg_request *nlmsg_request = NULL;
+	unsigned long flags;
+
+	nlmsg_request = kzalloc(sizeof(struct iwpm_nlmsg_request), gfp);
+	if (!nlmsg_request) {
+		pr_err("%s Unable to allocate a nlmsg_request\n", __func__);
+		return NULL;
+	}
+	spin_lock_irqsave(&iwpm_nlmsg_req_lock, flags);
+	list_add_tail(&nlmsg_request->inprocess_list, &iwpm_nlmsg_req_list);
+	spin_unlock_irqrestore(&iwpm_nlmsg_req_lock, flags);
+
+	kref_init(&nlmsg_request->kref);
+	kref_get(&nlmsg_request->kref);
+	nlmsg_request->nlmsg_seq = nlmsg_seq;
+	nlmsg_request->nl_client = nl_client;
+	nlmsg_request->request_done = 0;
+	nlmsg_request->err_code = 0;
+	return nlmsg_request;
+}
+
+void iwpm_free_nlmsg_request(struct kref *kref)
+{
+	struct iwpm_nlmsg_request *nlmsg_request;
+	unsigned long flags;
+
+	nlmsg_request = container_of(kref, struct iwpm_nlmsg_request, kref);
+
+	spin_lock_irqsave(&iwpm_nlmsg_req_lock, flags);
+	list_del_init(&nlmsg_request->inprocess_list);
+	spin_unlock_irqrestore(&iwpm_nlmsg_req_lock, flags);
+
+	if (!nlmsg_request->request_done)
+		pr_debug("%s Freeing incomplete nlmsg request (seq = %u).\n",
+			__func__, nlmsg_request->nlmsg_seq);
+	kfree(nlmsg_request);
+}
+
+struct iwpm_nlmsg_request *iwpm_find_nlmsg_request(__u32 echo_seq)
+{
+	struct iwpm_nlmsg_request *nlmsg_request;
+	struct iwpm_nlmsg_request *found_request = NULL;
+	unsigned long flags;
+
+	spin_lock_irqsave(&iwpm_nlmsg_req_lock, flags);
+	list_for_each_entry(nlmsg_request, &iwpm_nlmsg_req_list,
+			    inprocess_list) {
+		if (nlmsg_request->nlmsg_seq == echo_seq) {
+			found_request = nlmsg_request;
+			kref_get(&nlmsg_request->kref);
+			break;
+		}
+	}
+	spin_unlock_irqrestore(&iwpm_nlmsg_req_lock, flags);
+	return found_request;
+}
+
+int iwpm_wait_complete_req(struct iwpm_nlmsg_request *nlmsg_request)
+{
+	int ret;
+	init_waitqueue_head(&nlmsg_request->waitq);
+
+	ret = wait_event_timeout(nlmsg_request->waitq,
+			(nlmsg_request->request_done != 0), IWPM_NL_TIMEOUT);
+	if (!ret) {
+		ret = -EINVAL;
+		pr_info("%s: Timeout %d sec for netlink request (seq = %u)\n",
+			__func__, (IWPM_NL_TIMEOUT/HZ), nlmsg_request->nlmsg_seq);
+	} else {
+		ret = nlmsg_request->err_code;
+	}
+	kref_put(&nlmsg_request->kref, iwpm_free_nlmsg_request);
+	return ret;
+}
+
+int iwpm_get_nlmsg_seq(void)
+{
+	return atomic_inc_return(&iwpm_admin.nlmsg_seq);
+}
+
+int iwpm_valid_client(u8 nl_client)
+{
+	if (nl_client >= RDMA_NL_NUM_CLIENTS)
+		return 0;
+	return iwpm_admin.client_list[nl_client];
+}
+
+void iwpm_set_valid(u8 nl_client, int valid)
+{
+	if (nl_client >= RDMA_NL_NUM_CLIENTS)
+		return;
+	iwpm_admin.client_list[nl_client] = valid;
+}
+
+/* valid client */
+int iwpm_registered_client(u8 nl_client)
+{
+	return iwpm_admin.reg_list[nl_client];
+}
+
+/* valid client */
+void iwpm_set_registered(u8 nl_client, int reg)
+{
+	iwpm_admin.reg_list[nl_client] = reg;
+}
+
+int iwpm_compare_sockaddr(struct sockaddr_storage *a_sockaddr,
+				struct sockaddr_storage *b_sockaddr)
+{
+	if (a_sockaddr->ss_family != b_sockaddr->ss_family)
+		return 1;
+	if (a_sockaddr->ss_family == AF_INET) {
+		struct sockaddr_in *a4_sockaddr =
+			(struct sockaddr_in *)a_sockaddr;
+		struct sockaddr_in *b4_sockaddr =
+			(struct sockaddr_in *)b_sockaddr;
+		if (!memcmp(&a4_sockaddr->sin_addr,
+			&b4_sockaddr->sin_addr, sizeof(struct in_addr))
+			&& a4_sockaddr->sin_port == b4_sockaddr->sin_port)
+				return 0;
+
+	} else if (a_sockaddr->ss_family == AF_INET6) {
+		struct sockaddr_in6 *a6_sockaddr =
+			(struct sockaddr_in6 *)a_sockaddr;
+		struct sockaddr_in6 *b6_sockaddr =
+			(struct sockaddr_in6 *)b_sockaddr;
+		if (!memcmp(&a6_sockaddr->sin6_addr,
+			&b6_sockaddr->sin6_addr, sizeof(struct in6_addr))
+			&& a6_sockaddr->sin6_port == b6_sockaddr->sin6_port)
+				return 0;
+
+	} else {
+		pr_err("%s: Invalid sockaddr family\n", __func__);
+	}
+	return 1;
+}
+
+struct sk_buff *iwpm_create_nlmsg(u32 nl_op, struct nlmsghdr **nlh,
+						int nl_client)
+{
+	struct sk_buff *skb = NULL;
+
+	skb = dev_alloc_skb(NLMSG_GOODSIZE);
+	if (!skb) {
+		pr_err("%s Unable to allocate skb\n", __func__);
+		goto create_nlmsg_exit;
+	}
+	if (!(ibnl_put_msg(skb, nlh, 0, 0, nl_client, nl_op,
+			   NLM_F_REQUEST))) {
+		pr_warn("%s: Unable to put the nlmsg header\n", __func__);
+		dev_kfree_skb(skb);
+		skb = NULL;
+	}
+create_nlmsg_exit:
+	return skb;
+}
+
+int iwpm_parse_nlmsg(struct netlink_callback *cb, int policy_max,
+				   const struct nla_policy *nlmsg_policy,
+				   struct nlattr *nltb[], const char *msg_type)
+{
+	int nlh_len = 0;
+	int ret;
+	const char *err_str = "";
+
+	ret = nlmsg_validate(cb->nlh, nlh_len, policy_max-1, nlmsg_policy);
+	if (ret) {
+		err_str = "Invalid attribute";
+		goto parse_nlmsg_error;
+	}
+	ret = nlmsg_parse(cb->nlh, nlh_len, nltb, policy_max-1, nlmsg_policy);
+	if (ret) {
+		err_str = "Unable to parse the nlmsg";
+		goto parse_nlmsg_error;
+	}
+	ret = iwpm_validate_nlmsg_attr(nltb, policy_max);
+	if (ret) {
+		err_str = "Invalid NULL attribute";
+		goto parse_nlmsg_error;
+	}
+	return 0;
+parse_nlmsg_error:
+	pr_warn("%s: %s (msg type %s ret = %d)\n",
+			__func__, err_str, msg_type, ret);
+	return ret;
+}
+
+void iwpm_print_sockaddr(struct sockaddr_storage *sockaddr, char *msg)
+{
+	struct sockaddr_in6 *sockaddr_v6;
+	struct sockaddr_in *sockaddr_v4;
+
+	switch (sockaddr->ss_family) {
+	case AF_INET:
+		sockaddr_v4 = (struct sockaddr_in *)sockaddr;
+		pr_debug("%s IPV4 %pI4: %u(0x%04X)\n",
+			msg, &sockaddr_v4->sin_addr,
+			ntohs(sockaddr_v4->sin_port),
+			ntohs(sockaddr_v4->sin_port));
+		break;
+	case AF_INET6:
+		sockaddr_v6 = (struct sockaddr_in6 *)sockaddr;
+		pr_debug("%s IPV6 %pI6: %u(0x%04X)\n",
+			msg, &sockaddr_v6->sin6_addr,
+			ntohs(sockaddr_v6->sin6_port),
+			ntohs(sockaddr_v6->sin6_port));
+		break;
+	default:
+		break;
+	}
+}
+
+static u32 iwpm_ipv6_jhash(struct sockaddr_in6 *ipv6_sockaddr)
+{
+	u32 ipv6_hash = jhash(&ipv6_sockaddr->sin6_addr, sizeof(struct in6_addr), 0);
+	u32 hash = jhash_2words(ipv6_hash, (__force u32) ipv6_sockaddr->sin6_port, 0);
+	return hash;
+}
+
+static u32 iwpm_ipv4_jhash(struct sockaddr_in *ipv4_sockaddr)
+{
+	u32 ipv4_hash = jhash(&ipv4_sockaddr->sin_addr, sizeof(struct in_addr), 0);
+	u32 hash = jhash_2words(ipv4_hash, (__force u32) ipv4_sockaddr->sin_port, 0);
+	return hash;
+}
+
+static struct hlist_head *get_hash_bucket_head(struct sockaddr_storage
+					       *local_sockaddr,
+					       struct sockaddr_storage
+					       *mapped_sockaddr)
+{
+	u32 local_hash, mapped_hash, hash;
+
+	if (local_sockaddr->ss_family == AF_INET) {
+		local_hash = iwpm_ipv4_jhash((struct sockaddr_in *) local_sockaddr);
+		mapped_hash = iwpm_ipv4_jhash((struct sockaddr_in *) mapped_sockaddr);
+
+	} else if (local_sockaddr->ss_family == AF_INET6) {
+		local_hash = iwpm_ipv6_jhash((struct sockaddr_in6 *) local_sockaddr);
+		mapped_hash = iwpm_ipv6_jhash((struct sockaddr_in6 *) mapped_sockaddr);
+	} else {
+		pr_err("%s: Invalid sockaddr family\n", __func__);
+		return NULL;
+	}
+
+	if (local_hash == mapped_hash) /* if port mapper isn't available */
+		hash = local_hash;
+	else
+		hash = jhash_2words(local_hash, mapped_hash, 0);
+
+	return &iwpm_hash_bucket[hash & IWPM_HASH_BUCKET_MASK];
+}
+
+static int send_mapinfo_num(u32 mapping_num, u8 nl_client, int iwpm_pid)
+{
+	struct sk_buff *skb = NULL;
+	struct nlmsghdr *nlh;
+	u32 msg_seq;
+	const char *err_str = "";
+	int ret = -EINVAL;
+
+	skb = iwpm_create_nlmsg(RDMA_NL_IWPM_MAPINFO_NUM, &nlh, nl_client);
+	if (!skb) {
+		err_str = "Unable to create a nlmsg";
+		goto mapinfo_num_error;
+	}
+	nlh->nlmsg_seq = iwpm_get_nlmsg_seq();
+	msg_seq = 0;
+	err_str = "Unable to put attribute of mapinfo number nlmsg";
+	ret = ibnl_put_attr(skb, nlh, sizeof(u32), &msg_seq, IWPM_NLA_MAPINFO_SEQ);
+	if (ret)
+		goto mapinfo_num_error;
+	ret = ibnl_put_attr(skb, nlh, sizeof(u32),
+				&mapping_num, IWPM_NLA_MAPINFO_SEND_NUM);
+	if (ret)
+		goto mapinfo_num_error;
+	ret = ibnl_unicast(skb, nlh, iwpm_pid);
+	if (ret) {
+		skb = NULL;
+		err_str = "Unable to send a nlmsg";
+		goto mapinfo_num_error;
+	}
+	pr_debug("%s: Sent mapping number = %d\n", __func__, mapping_num);
+	return 0;
+mapinfo_num_error:
+	pr_info("%s: %s\n", __func__, err_str);
+	if (skb)
+		dev_kfree_skb(skb);
+	return ret;
+}
+
+static int send_nlmsg_done(struct sk_buff *skb, u8 nl_client, int iwpm_pid)
+{
+	struct nlmsghdr *nlh = NULL;
+	int ret = 0;
+
+	if (!skb)
+		return ret;
+	if (!(ibnl_put_msg(skb, &nlh, 0, 0, nl_client,
+			   RDMA_NL_IWPM_MAPINFO, NLM_F_MULTI))) {
+		pr_warn("%s Unable to put NLMSG_DONE\n", __func__);
+		return -ENOMEM;
+	}
+	nlh->nlmsg_type = NLMSG_DONE;
+	ret = ibnl_unicast(skb, (struct nlmsghdr *)skb->data, iwpm_pid);
+	if (ret)
+		pr_warn("%s Unable to send a nlmsg\n", __func__);
+	return ret;
+}
+
+int iwpm_send_mapinfo(u8 nl_client, int iwpm_pid)
+{
+	struct iwpm_mapping_info *map_info;
+	struct sk_buff *skb = NULL;
+	struct nlmsghdr *nlh;
+	int skb_num = 0, mapping_num = 0;
+	int i = 0, nlmsg_bytes = 0;
+	unsigned long flags;
+	const char *err_str = "";
+	int ret;
+
+	skb = dev_alloc_skb(NLMSG_GOODSIZE);
+	if (!skb) {
+		ret = -ENOMEM;
+		err_str = "Unable to allocate skb";
+		goto send_mapping_info_exit;
+	}
+	skb_num++;
+	spin_lock_irqsave(&iwpm_mapinfo_lock, flags);
+	for (i = 0; i < IWPM_HASH_BUCKET_SIZE; i++) {
+		hlist_for_each_entry(map_info, &iwpm_hash_bucket[i],
+				     hlist_node) {
+			if (map_info->nl_client != nl_client)
+				continue;
+			nlh = NULL;
+			if (!(ibnl_put_msg(skb, &nlh, 0, 0, nl_client,
+					RDMA_NL_IWPM_MAPINFO, NLM_F_MULTI))) {
+				ret = -ENOMEM;
+				err_str = "Unable to put the nlmsg header";
+				goto send_mapping_info_unlock;
+			}
+			err_str = "Unable to put attribute of the nlmsg";
+			ret = ibnl_put_attr(skb, nlh,
+					sizeof(struct sockaddr_storage),
+					&map_info->local_sockaddr,
+					IWPM_NLA_MAPINFO_LOCAL_ADDR);
+			if (ret)
+				goto send_mapping_info_unlock;
+
+			ret = ibnl_put_attr(skb, nlh,
+					sizeof(struct sockaddr_storage),
+					&map_info->mapped_sockaddr,
+					IWPM_NLA_MAPINFO_MAPPED_ADDR);
+			if (ret)
+				goto send_mapping_info_unlock;
+
+			iwpm_print_sockaddr(&map_info->local_sockaddr,
+				"send_mapping_info: Local sockaddr:");
+			iwpm_print_sockaddr(&map_info->mapped_sockaddr,
+				"send_mapping_info: Mapped local sockaddr:");
+			mapping_num++;
+			nlmsg_bytes += nlh->nlmsg_len;
+
+			/* check if all mappings can fit in one skb */
+			if (NLMSG_GOODSIZE - nlmsg_bytes < nlh->nlmsg_len * 2) {
+				/* and leave room for NLMSG_DONE */
+				nlmsg_bytes = 0;
+				skb_num++;
+				spin_unlock_irqrestore(&iwpm_mapinfo_lock,
+						       flags);
+				/* send the skb */
+				ret = send_nlmsg_done(skb, nl_client, iwpm_pid);
+				skb = NULL;
+				if (ret) {
+					err_str = "Unable to send map info";
+					goto send_mapping_info_exit;
+				}
+				if (skb_num == IWPM_MAPINFO_SKB_COUNT) {
+					ret = -ENOMEM;
+					err_str = "Insufficient skbs for map info";
+					goto send_mapping_info_exit;
+				}
+				skb = dev_alloc_skb(NLMSG_GOODSIZE);
+				if (!skb) {
+					ret = -ENOMEM;
+					err_str = "Unable to allocate skb";
+					goto send_mapping_info_exit;
+				}
+				spin_lock_irqsave(&iwpm_mapinfo_lock, flags);
+			}
+		}
+	}
+send_mapping_info_unlock:
+	spin_unlock_irqrestore(&iwpm_mapinfo_lock, flags);
+send_mapping_info_exit:
+	if (ret) {
+		pr_warn("%s: %s (ret = %d)\n", __func__, err_str, ret);
+		if (skb)
+			dev_kfree_skb(skb);
+		return ret;
+	}
+	send_nlmsg_done(skb, nl_client, iwpm_pid);
+	return send_mapinfo_num(mapping_num, nl_client, iwpm_pid);
+}
+
+int iwpm_mapinfo_available(void)
+{
+	unsigned long flags;
+	int full_bucket = 0, i = 0;
+
+	spin_lock_irqsave(&iwpm_mapinfo_lock, flags);
+	if (iwpm_hash_bucket) {
+		for (i = 0; i < IWPM_HASH_BUCKET_SIZE; i++) {
+			if (!hlist_empty(&iwpm_hash_bucket[i])) {
+				full_bucket = 1;
+				break;
+			}
+		}
+	}
+	spin_unlock_irqrestore(&iwpm_mapinfo_lock, flags);
+	return full_bucket;
+}

drivers/infiniband/core/iwpm_util.h

+/*
+ * Copyright (c) 2014 Intel Corporation. All rights reserved.
+ * Copyright (c) 2014 Chelsio, Inc. 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.
+ */
+#ifndef _IWPM_UTIL_H
+#define _IWPM_UTIL_H
+
+#include <linux/module.h>
+#include <linux/io.h>
+#include <linux/in.h>
+#include <linux/in6.h>
+#include <linux/spinlock.h>
+#include <linux/kernel.h>
+#include <linux/netdevice.h>
+#include <linux/delay.h>
+#include <linux/workqueue.h>
+#include <linux/mutex.h>
+#include <linux/jhash.h>
+#include <linux/kref.h>
+#include <net/netlink.h>
+#include <linux/errno.h>
+#include <rdma/iw_portmap.h>
+#include <rdma/rdma_netlink.h>
+
+
+#define IWPM_NL_RETRANS		3
+#define IWPM_NL_TIMEOUT		(10*HZ)
+#define IWPM_MAPINFO_SKB_COUNT	20
+
+#define IWPM_PID_UNDEFINED     -1
+#define IWPM_PID_UNAVAILABLE   -2
+
+struct iwpm_nlmsg_request {
+	struct list_head    inprocess_list;
+	__u32               nlmsg_seq;
+	void                *req_buffer;
+	u8	            nl_client;
+	u8                  request_done;
+	u16                 err_code;
+	wait_queue_head_t   waitq;
+	struct kref         kref;
+};
+
+struct iwpm_mapping_info {
+	struct hlist_node hlist_node;
+	struct sockaddr_storage local_sockaddr;
+	struct sockaddr_storage mapped_sockaddr;
+	u8     nl_client;
+};
+
+struct iwpm_admin_data {
+	atomic_t refcount;
+	atomic_t nlmsg_seq;
+	int      client_list[RDMA_NL_NUM_CLIENTS];
+	int      reg_list[RDMA_NL_NUM_CLIENTS];
+};
+
+/**
+ * iwpm_get_nlmsg_request - Allocate and initialize netlink message request
+ * @nlmsg_seq: Sequence number of the netlink message
+ * @nl_client: The index of the netlink client
+ * @gfp: Indicates how the memory for the request should be allocated
+ *
+ * Returns the newly allocated netlink request object if successful,
+ * otherwise returns NULL
+ */
+struct iwpm_nlmsg_request *iwpm_get_nlmsg_request(__u32 nlmsg_seq,
+						u8 nl_client, gfp_t gfp);
+
+/**
+ * iwpm_free_nlmsg_request - Deallocate netlink message request
+ * @kref: Holds reference of netlink message request
+ */
+void iwpm_free_nlmsg_request(struct kref *kref);
+
+/**
+ * iwpm_find_nlmsg_request - Find netlink message request in the request list
+ * @echo_seq: Sequence number of the netlink request to find
+ *
+ * Returns the found netlink message request,
+ * if not found, returns NULL
+ */
+struct iwpm_nlmsg_request *iwpm_find_nlmsg_request(__u32 echo_seq);
+
+/**
+ * iwpm_wait_complete_req - Block while servicing the netlink request
+ * @nlmsg_request: Netlink message request to service
+ *
+ * Wakes up, after the request is completed or expired
+ * Returns 0 if the request is complete without error
+ */
+int iwpm_wait_complete_req(struct iwpm_nlmsg_request *nlmsg_request);
+
+/**
+ * iwpm_get_nlmsg_seq - Get the sequence number for a netlink
+ *			message to send to the port mapper
+ *
+ * Returns the sequence number for the netlink message.
+ */
+int iwpm_get_nlmsg_seq(void);
+
+/**
+ * iwpm_valid_client - Check if the port mapper client is valid
+ * @nl_client: The index of the netlink client
+ *
+ * Valid clients need to call iwpm_init() before using
+ * the port mapper
+ */
+int iwpm_valid_client(u8 nl_client);
+
+/**
+ * iwpm_set_valid - Set the port mapper client to valid or not
+ * @nl_client: The index of the netlink client
+ * @valid: 1 if valid or 0 if invalid
+ */
+void iwpm_set_valid(u8 nl_client, int valid);
+
+/**
+ * iwpm_registered_client - Check if the port mapper client is registered
+ * @nl_client: The index of the netlink client
+ *
+ * Call iwpm_register_pid() to register a client
+ */
+int iwpm_registered_client(u8 nl_client);
+
+/**
+ * iwpm_set_registered - Set the port mapper client to registered or not
+ * @nl_client: The index of the netlink client
+ * @reg: 1 if registered or 0 if not
+ */
+void iwpm_set_registered(u8 nl_client, int reg);
+
+/**
+ * iwpm_send_mapinfo - Send local and mapped IPv4/IPv6 address info of
+ *                     a client to the user space port mapper
+ * @nl_client: The index of the netlink client
+ * @iwpm_pid: The pid of the user space port mapper
+ *
+ * If successful, returns the number of sent mapping info records
+ */
+int iwpm_send_mapinfo(u8 nl_client, int iwpm_pid);
+
+/**
+ * iwpm_mapinfo_available - Check if any mapping info records is available
+ *		            in the hash table
+ *
+ * Returns 1 if mapping information is available, otherwise returns 0
+ */
+int iwpm_mapinfo_available(void);
+
+/**
+ * iwpm_compare_sockaddr - Compare two sockaddr storage structs
+ *
+ * Returns 0 if they are holding the same ip/tcp address info,
+ * otherwise returns 1
+ */
+int iwpm_compare_sockaddr(struct sockaddr_storage *a_sockaddr,
+			struct sockaddr_storage *b_sockaddr);
+
+/**
+ * iwpm_validate_nlmsg_attr - Check for NULL netlink attributes
+ * @nltb: Holds address of each netlink message attributes
+ * @nla_count: Number of netlink message attributes
+ *
+ * Returns error if any of the nla_count attributes is NULL
+ */
+static inline int iwpm_validate_nlmsg_attr(struct nlattr *nltb[],
+					   int nla_count)
+{
+	int i;
+	for (i = 1; i < nla_count; i++) {
+		if (!nltb[i])
+			return -EINVAL;
+	}
+	return 0;
+}
+
+/**
+ * iwpm_create_nlmsg - Allocate skb and form a netlink message
+ * @nl_op: Netlink message opcode
+ * @nlh: Holds address of the netlink message header in skb
+ * @nl_client: The index of the netlink client
+ *
+ * Returns the newly allcated skb, or NULL if the tailroom of the skb
+ * is insufficient to store the message header and payload
+ */
+struct sk_buff *iwpm_create_nlmsg(u32 nl_op, struct nlmsghdr **nlh,
+					int nl_client);
+
+/**
+ * iwpm_parse_nlmsg - Validate and parse the received netlink message
+ * @cb: Netlink callback structure
+ * @policy_max: Maximum attribute type to be expected
+ * @nlmsg_policy: Validation policy
+ * @nltb: Array to store policy_max parsed elements
+ * @msg_type: Type of netlink message
+ *
+ * Returns 0 on success or a negative error code
+ */
+int iwpm_parse_nlmsg(struct netlink_callback *cb, int policy_max,
+				const struct nla_policy *nlmsg_policy,
+				struct nlattr *nltb[], const char *msg_type);
+
+/**
+ * iwpm_print_sockaddr - Print IPv4/IPv6 address and TCP port
+ * @sockaddr: Socket address to print
+ * @msg: Message to print
+ */
+void iwpm_print_sockaddr(struct sockaddr_storage *sockaddr, char *msg);
+#endif

drivers/infiniband/core/netlink.c

@@ -103,13 +103,13 @@ int ibnl_remove_client(int index)
 EXPORT_SYMBOL(ibnl_remove_client);
 
 void *ibnl_put_msg(struct sk_buff *skb, struct nlmsghdr **nlh, int seq,
-		   int len, int client, int op)
+		   int len, int client, int op, int flags)
 {
 	unsigned char *prev_tail;
 
 	prev_tail = skb_tail_pointer(skb);
 	*nlh = nlmsg_put(skb, 0, seq, RDMA_NL_GET_TYPE(client, op),
-			 len, NLM_F_MULTI);
+			 len, flags);
 	if (!*nlh)
 		goto out_nlmsg_trim;
 	(*nlh)->nlmsg_len = skb_tail_pointer(skb) - prev_tail;
@@ -172,6 +172,20 @@ static void ibnl_rcv(struct sk_buff *skb)
 	mutex_unlock(&ibnl_mutex);
 }
 
+int ibnl_unicast(struct sk_buff *skb, struct nlmsghdr *nlh,
+			__u32 pid)
+{
+	return nlmsg_unicast(nls, skb, pid);
+}
+EXPORT_SYMBOL(ibnl_unicast);
+
+int ibnl_multicast(struct sk_buff *skb, struct nlmsghdr *nlh,
+			unsigned int group, gfp_t flags)
+{
+	return nlmsg_multicast(nls, skb, 0, group, flags);
+}
+EXPORT_SYMBOL(ibnl_multicast);
+
 int __init ibnl_init(void)
 {
 	struct netlink_kernel_cfg cfg = {

drivers/infiniband/core/sa_query.c

@@ -618,7 +618,7 @@ static void init_mad(struct ib_sa_mad *mad, struct ib_mad_agent *agent)
 
 static int send_mad(struct ib_sa_query *query, int timeout_ms, gfp_t gfp_mask)
 {
-	bool preload = gfp_mask & __GFP_WAIT;
+	bool preload = !!(gfp_mask & __GFP_WAIT);
 	unsigned long flags;
 	int ret, id;
 

drivers/infiniband/core/verbs.c

@@ -48,7 +48,7 @@
 
 #include "core_priv.h"
 
-int ib_rate_to_mult(enum ib_rate rate)
+__attribute_const__ int ib_rate_to_mult(enum ib_rate rate)
 {
 	switch (rate) {
 	case IB_RATE_2_5_GBPS: return  1;
@@ -65,7 +65,7 @@ int ib_rate_to_mult(enum ib_rate rate)
 }
 EXPORT_SYMBOL(ib_rate_to_mult);
 
-enum ib_rate mult_to_ib_rate(int mult)
+__attribute_const__ enum ib_rate mult_to_ib_rate(int mult)
 {
 	switch (mult) {
 	case 1:  return IB_RATE_2_5_GBPS;
@@ -82,7 +82,7 @@ enum ib_rate mult_to_ib_rate(int mult)
 }
 EXPORT_SYMBOL(mult_to_ib_rate);
 
-int ib_rate_to_mbps(enum ib_rate rate)
+__attribute_const__ int ib_rate_to_mbps(enum ib_rate rate)
 {
 	switch (rate) {
 	case IB_RATE_2_5_GBPS: return 2500;
@@ -107,7 +107,7 @@ int ib_rate_to_mbps(enum ib_rate rate)
 }
 EXPORT_SYMBOL(ib_rate_to_mbps);
 
-enum rdma_transport_type
+__attribute_const__ enum rdma_transport_type
 rdma_node_get_transport(enum rdma_node_type node_type)
 {
 	switch (node_type) {

drivers/infiniband/hw/Makefile

+obj-$(CONFIG_INFINIBAND_MTHCA)		+= mthca/
+obj-$(CONFIG_INFINIBAND_IPATH)		+= ipath/
+obj-$(CONFIG_INFINIBAND_QIB)		+= qib/
+obj-$(CONFIG_INFINIBAND_EHCA)		+= ehca/
+obj-$(CONFIG_INFINIBAND_AMSO1100)	+= amso1100/
+obj-$(CONFIG_INFINIBAND_CXGB3)		+= cxgb3/
+obj-$(CONFIG_INFINIBAND_CXGB4)		+= cxgb4/
+obj-$(CONFIG_MLX4_INFINIBAND)		+= mlx4/
+obj-$(CONFIG_MLX5_INFINIBAND)		+= mlx5/
+obj-$(CONFIG_INFINIBAND_NES)		+= nes/
+obj-$(CONFIG_INFINIBAND_OCRDMA)		+= ocrdma/
+obj-$(CONFIG_INFINIBAND_USNIC)		+= usnic/

drivers/infiniband/hw/cxgb3/cxio_hal.c

@@ -735,14 +735,12 @@ static int __cxio_tpt_op(struct cxio_rdev *rdev_p, u32 reset_tpt_entry,
 			((perm & TPT_MW_BIND) ? F_TPT_MW_BIND_ENABLE : 0) |
 			V_TPT_ADDR_TYPE((zbva ? TPT_ZBTO : TPT_VATO)) |
 			V_TPT_PAGE_SIZE(page_size));
-		tpt.rsvd_pbl_addr = reset_tpt_entry ? 0 :
-				    cpu_to_be32(V_TPT_PBL_ADDR(PBL_OFF(rdev_p, pbl_addr)>>3));
+		tpt.rsvd_pbl_addr = cpu_to_be32(V_TPT_PBL_ADDR(PBL_OFF(rdev_p, pbl_addr)>>3));
 		tpt.len = cpu_to_be32(len);
 		tpt.va_hi = cpu_to_be32((u32) (to >> 32));
 		tpt.va_low_or_fbo = cpu_to_be32((u32) (to & 0xFFFFFFFFULL));
 		tpt.rsvd_bind_cnt_or_pstag = 0;
-		tpt.rsvd_pbl_size = reset_tpt_entry ? 0 :
-				  cpu_to_be32(V_TPT_PBL_SIZE(pbl_size >> 2));
+		tpt.rsvd_pbl_size = cpu_to_be32(V_TPT_PBL_SIZE(pbl_size >> 2));
 	}
 	err = cxio_hal_ctrl_qp_write_mem(rdev_p,
 				       stag_idx +

drivers/infiniband/hw/cxgb3/iwch_cm.c

@@ -418,6 +418,7 @@ static int send_abort(struct iwch_ep *ep, struct sk_buff *skb, gfp_t gfp)
 	skb->priority = CPL_PRIORITY_DATA;
 	set_arp_failure_handler(skb, abort_arp_failure);
 	req = (struct cpl_abort_req *) skb_put(skb, sizeof(*req));
+	memset(req, 0, sizeof(*req));
 	req->wr.wr_hi = htonl(V_WR_OP(FW_WROPCODE_OFLD_HOST_ABORT_CON_REQ));
 	req->wr.wr_lo = htonl(V_WR_TID(ep->hwtid));
 	OPCODE_TID(req) = htonl(MK_OPCODE_TID(CPL_ABORT_REQ, ep->hwtid));

drivers/infiniband/hw/cxgb4/cm.c

 /*
- * Copyright (c) 2009-2010 Chelsio, Inc. All rights reserved.
+ * Copyright (c) 2009-2014 Chelsio, Inc. 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
@@ -47,6 +47,8 @@
 #include <net/ip6_route.h>
 #include <net/addrconf.h>
 
+#include <rdma/ib_addr.h>
+
 #include "iw_cxgb4.h"
 
 static char *states[] = {
@@ -294,6 +296,12 @@ void _c4iw_free_ep(struct kref *kref)
 		dst_release(ep->dst);
 		cxgb4_l2t_release(ep->l2t);
 	}
+	if (test_bit(RELEASE_MAPINFO, &ep->com.flags)) {
+		print_addr(&ep->com, __func__, "remove_mapinfo/mapping");
+		iwpm_remove_mapinfo(&ep->com.local_addr,
+				    &ep->com.mapped_local_addr);
+		iwpm_remove_mapping(&ep->com.local_addr, RDMA_NL_C4IW);
+	}
 	kfree(ep);
 }
 
@@ -341,10 +349,7 @@ static struct sk_buff *get_skb(struct sk_buff *skb, int len, gfp_t gfp)
 
 static struct net_device *get_real_dev(struct net_device *egress_dev)
 {
-	struct net_device *phys_dev = egress_dev;
-	if (egress_dev->priv_flags & IFF_802_1Q_VLAN)
-		phys_dev = vlan_dev_real_dev(egress_dev);
-	return phys_dev;
+	return rdma_vlan_dev_real_dev(egress_dev) ? : egress_dev;
 }
 
 static int our_interface(struct c4iw_dev *dev, struct net_device *egress_dev)
@@ -528,6 +533,38 @@ static int send_abort(struct c4iw_ep *ep, struct sk_buff *skb, gfp_t gfp)
 	return c4iw_l2t_send(&ep->com.dev->rdev, skb, ep->l2t);
 }
 
+/*
+ * c4iw_form_pm_msg - Form a port mapper message with mapping info
+ */
+static void c4iw_form_pm_msg(struct c4iw_ep *ep,
+				struct iwpm_sa_data *pm_msg)
+{
+	memcpy(&pm_msg->loc_addr, &ep->com.local_addr,
+		sizeof(ep->com.local_addr));
+	memcpy(&pm_msg->rem_addr, &ep->com.remote_addr,
+		sizeof(ep->com.remote_addr));
+}
+
+/*
+ * c4iw_form_reg_msg - Form a port mapper message with dev info
+ */
+static void c4iw_form_reg_msg(struct c4iw_dev *dev,
+				struct iwpm_dev_data *pm_msg)
+{
+	memcpy(pm_msg->dev_name, dev->ibdev.name, IWPM_DEVNAME_SIZE);
+	memcpy(pm_msg->if_name, dev->rdev.lldi.ports[0]->name,
+				IWPM_IFNAME_SIZE);
+}
+
+static void c4iw_record_pm_msg(struct c4iw_ep *ep,
+			struct iwpm_sa_data *pm_msg)
+{
+	memcpy(&ep->com.mapped_local_addr, &pm_msg->mapped_loc_addr,
+		sizeof(ep->com.mapped_local_addr));
+	memcpy(&ep->com.mapped_remote_addr, &pm_msg->mapped_rem_addr,
+		sizeof(ep->com.mapped_remote_addr));
+}
+
 static int send_connect(struct c4iw_ep *ep)
 {
 	struct cpl_act_open_req *req;
@@ -546,10 +583,14 @@ static int send_connect(struct c4iw_ep *ep)
 	int sizev6 = is_t4(ep->com.dev->rdev.lldi.adapter_type) ?
 				sizeof(struct cpl_act_open_req6) :
 				sizeof(struct cpl_t5_act_open_req6);
-	struct sockaddr_in *la = (struct sockaddr_in *)&ep->com.local_addr;
-	struct sockaddr_in *ra = (struct sockaddr_in *)&ep->com.remote_addr;
-	struct sockaddr_in6 *la6 = (struct sockaddr_in6 *)&ep->com.local_addr;
-	struct sockaddr_in6 *ra6 = (struct sockaddr_in6 *)&ep->com.remote_addr;
+	struct sockaddr_in *la = (struct sockaddr_in *)
+				 &ep->com.mapped_local_addr;
+	struct sockaddr_in *ra = (struct sockaddr_in *)
+				 &ep->com.mapped_remote_addr;
+	struct sockaddr_in6 *la6 = (struct sockaddr_in6 *)
+				   &ep->com.mapped_local_addr;
+	struct sockaddr_in6 *ra6 = (struct sockaddr_in6 *)
+				   &ep->com.mapped_remote_addr;
 
 	wrlen = (ep->com.remote_addr.ss_family == AF_INET) ?
 			roundup(sizev4, 16) :
@@ -1627,10 +1668,10 @@ static void send_fw_act_open_req(struct c4iw_ep *ep, unsigned int atid)
 	req->le.filter = cpu_to_be32(cxgb4_select_ntuple(
 				     ep->com.dev->rdev.lldi.ports[0],
 				     ep->l2t));
-	sin = (struct sockaddr_in *)&ep->com.local_addr;
+	sin = (struct sockaddr_in *)&ep->com.mapped_local_addr;
 	req->le.lport = sin->sin_port;
 	req->le.u.ipv4.lip = sin->sin_addr.s_addr;
-	sin = (struct sockaddr_in *)&ep->com.remote_addr;
+	sin = (struct sockaddr_in *)&ep->com.mapped_remote_addr;
 	req->le.pport = sin->sin_port;
 	req->le.u.ipv4.pip = sin->sin_addr.s_addr;
 	req->tcb.t_state_to_astid =
@@ -1746,16 +1787,16 @@ static int import_ep(struct c4iw_ep *ep, int iptype, __u8 *peer_ip,
 		if (!ep->l2t)
 			goto out;
 		ep->mtu = dst_mtu(dst);
-		ep->tx_chan = cxgb4_port_chan(n->dev);
-		ep->smac_idx = (cxgb4_port_viid(n->dev) & 0x7F) << 1;
+		ep->tx_chan = cxgb4_port_chan(pdev);
+		ep->smac_idx = (cxgb4_port_viid(pdev) & 0x7F) << 1;
 		step = cdev->rdev.lldi.ntxq /
 			cdev->rdev.lldi.nchan;
-		ep->txq_idx = cxgb4_port_idx(n->dev) * step;
-		ep->ctrlq_idx = cxgb4_port_idx(n->dev);
+		ep->txq_idx = cxgb4_port_idx(pdev) * step;
+		ep->ctrlq_idx = cxgb4_port_idx(pdev);
 		step = cdev->rdev.lldi.nrxq /
 			cdev->rdev.lldi.nchan;
 		ep->rss_qid = cdev->rdev.lldi.rxq_ids[
-			cxgb4_port_idx(n->dev) * step];
+			cxgb4_port_idx(pdev) * step];
 
 		if (clear_mpa_v1) {
 			ep->retry_with_mpa_v1 = 0;
@@ -1870,10 +1911,10 @@ static int act_open_rpl(struct c4iw_dev *dev, struct sk_buff *skb)
 	struct sockaddr_in6 *ra6;
 
 	ep = lookup_atid(t, atid);
-	la = (struct sockaddr_in *)&ep->com.local_addr;
-	ra = (struct sockaddr_in *)&ep->com.remote_addr;
-	la6 = (struct sockaddr_in6 *)&ep->com.local_addr;
-	ra6 = (struct sockaddr_in6 *)&ep->com.remote_addr;
+	la = (struct sockaddr_in *)&ep->com.mapped_local_addr;
+	ra = (struct sockaddr_in *)&ep->com.mapped_remote_addr;
+	la6 = (struct sockaddr_in6 *)&ep->com.mapped_local_addr;
+	ra6 = (struct sockaddr_in6 *)&ep->com.mapped_remote_addr;
 
 	PDBG("%s ep %p atid %u status %u errno %d\n", __func__, ep, atid,
 	     status, status2errno(status));
@@ -2730,13 +2771,15 @@ int c4iw_connect(struct iw_cm_id *cm_id, struct iw_cm_conn_param *conn_param)
 	struct c4iw_dev *dev = to_c4iw_dev(cm_id->device);
 	struct c4iw_ep *ep;
 	int err = 0;
-	struct sockaddr_in *laddr = (struct sockaddr_in *)&cm_id->local_addr;
-	struct sockaddr_in *raddr = (struct sockaddr_in *)&cm_id->remote_addr;
-	struct sockaddr_in6 *laddr6 = (struct sockaddr_in6 *)&cm_id->local_addr;
-	struct sockaddr_in6 *raddr6 = (struct sockaddr_in6 *)
-				      &cm_id->remote_addr;
+	struct sockaddr_in *laddr;
+	struct sockaddr_in *raddr;
+	struct sockaddr_in6 *laddr6;
+	struct sockaddr_in6 *raddr6;
+	struct iwpm_dev_data pm_reg_msg;
+	struct iwpm_sa_data pm_msg;
 	__u8 *ra;
 	int iptype;
+	int iwpm_err = 0;
 
 	if ((conn_param->ord > c4iw_max_read_depth) ||
 	    (conn_param->ird > c4iw_max_read_depth)) {
@@ -2767,7 +2810,7 @@ int c4iw_connect(struct iw_cm_id *cm_id, struct iw_cm_conn_param *conn_param)
 	if (!ep->com.qp) {
 		PDBG("%s qpn 0x%x not found!\n", __func__, conn_param->qpn);
 		err = -EINVAL;
-		goto fail2;
+		goto fail1;
 	}
 	ref_qp(ep);
 	PDBG("%s qpn 0x%x qp %p cm_id %p\n", __func__, conn_param->qpn,
@@ -2780,10 +2823,50 @@ int c4iw_connect(struct iw_cm_id *cm_id, struct iw_cm_conn_param *conn_param)
 	if (ep->atid == -1) {
 		printk(KERN_ERR MOD "%s - cannot alloc atid.\n", __func__);
 		err = -ENOMEM;
-		goto fail2;
+		goto fail1;
 	}
 	insert_handle(dev, &dev->atid_idr, ep, ep->atid);
 
+	memcpy(&ep->com.local_addr, &cm_id->local_addr,
+	       sizeof(ep->com.local_addr));
+	memcpy(&ep->com.remote_addr, &cm_id->remote_addr,
+	       sizeof(ep->com.remote_addr));
+
+	/* No port mapper available, go with the specified peer information */
+	memcpy(&ep->com.mapped_local_addr, &cm_id->local_addr,
+	       sizeof(ep->com.mapped_local_addr));
+	memcpy(&ep->com.mapped_remote_addr, &cm_id->remote_addr,
+	       sizeof(ep->com.mapped_remote_addr));
+
+	c4iw_form_reg_msg(dev, &pm_reg_msg);
+	iwpm_err = iwpm_register_pid(&pm_reg_msg, RDMA_NL_C4IW);
+	if (iwpm_err) {
+		PDBG("%s: Port Mapper reg pid fail (err = %d).\n",
+			__func__, iwpm_err);
+	}
+	if (iwpm_valid_pid() && !iwpm_err) {
+		c4iw_form_pm_msg(ep, &pm_msg);
+		iwpm_err = iwpm_add_and_query_mapping(&pm_msg, RDMA_NL_C4IW);
+		if (iwpm_err)
+			PDBG("%s: Port Mapper query fail (err = %d).\n",
+				__func__, iwpm_err);
+		else
+			c4iw_record_pm_msg(ep, &pm_msg);
+	}
+	if (iwpm_create_mapinfo(&ep->com.local_addr,
+				&ep->com.mapped_local_addr, RDMA_NL_C4IW)) {
+		iwpm_remove_mapping(&ep->com.local_addr, RDMA_NL_C4IW);
+		err = -ENOMEM;
+		goto fail1;
+	}
+	print_addr(&ep->com, __func__, "add_query/create_mapinfo");
+	set_bit(RELEASE_MAPINFO, &ep->com.flags);
+
+	laddr = (struct sockaddr_in *)&ep->com.mapped_local_addr;
+	raddr = (struct sockaddr_in *)&ep->com.mapped_remote_addr;
+	laddr6 = (struct sockaddr_in6 *)&ep->com.mapped_local_addr;
+	raddr6 = (struct sockaddr_in6 *) &ep->com.mapped_remote_addr;
+
 	if (cm_id->remote_addr.ss_family == AF_INET) {
 		iptype = 4;
 		ra = (__u8 *)&raddr->sin_addr;
@@ -2794,7 +2877,7 @@ int c4iw_connect(struct iw_cm_id *cm_id, struct iw_cm_conn_param *conn_param)
 		if ((__force int)raddr->sin_addr.s_addr == INADDR_ANY) {
 			err = pick_local_ipaddrs(dev, cm_id);
 			if (err)
-				goto fail2;
+				goto fail1;
 		}
 
 		/* find a route */
@@ -2814,7 +2897,7 @@ int c4iw_connect(struct iw_cm_id *cm_id, struct iw_cm_conn_param *conn_param)
 		if (ipv6_addr_type(&raddr6->sin6_addr) == IPV6_ADDR_ANY) {
 			err = pick_local_ip6addrs(dev, cm_id);
 			if (err)
-				goto fail2;
+				goto fail1;
 		}
 
 		/* find a route */
@@ -2830,13 +2913,13 @@ int c4iw_connect(struct iw_cm_id *cm_id, struct iw_cm_conn_param *conn_param)
 	if (!ep->dst) {
 		printk(KERN_ERR MOD "%s - cannot find route.\n", __func__);
 		err = -EHOSTUNREACH;
-		goto fail3;
+		goto fail2;
 	}
 
 	err = import_ep(ep, iptype, ra, ep->dst, ep->com.dev, true);
 	if (err) {
 		printk(KERN_ERR MOD "%s - cannot alloc l2e.\n", __func__);
-		goto fail4;
+		goto fail3;
 	}
 
 	PDBG("%s txq_idx %u tx_chan %u smac_idx %u rss_qid %u l2t_idx %u\n",
@@ -2845,10 +2928,6 @@ int c4iw_connect(struct iw_cm_id *cm_id, struct iw_cm_conn_param *conn_param)
 
 	state_set(&ep->com, CONNECTING);
 	ep->tos = 0;
-	memcpy(&ep->com.local_addr, &cm_id->local_addr,
-	       sizeof(ep->com.local_addr));
-	memcpy(&ep->com.remote_addr, &cm_id->remote_addr,
-	       sizeof(ep->com.remote_addr));
 
 	/* send connect request to rnic */
 	err = send_connect(ep);
@@ -2856,12 +2935,12 @@ int c4iw_connect(struct iw_cm_id *cm_id, struct iw_cm_conn_param *conn_param)
 		goto out;
 
 	cxgb4_l2t_release(ep->l2t);
-fail4:
-	dst_release(ep->dst);
 fail3:
+	dst_release(ep->dst);
+fail2:
 	remove_handle(ep->com.dev, &ep->com.dev->atid_idr, ep->atid);
 	cxgb4_free_atid(ep->com.dev->rdev.lldi.tids, ep->atid);
-fail2:
+fail1:
 	cm_id->rem_ref(cm_id);
 	c4iw_put_ep(&ep->com);
 out:
@@ -2871,7 +2950,8 @@ int c4iw_connect(struct iw_cm_id *cm_id, struct iw_cm_conn_param *conn_param)
 static int create_server6(struct c4iw_dev *dev, struct c4iw_listen_ep *ep)
 {
 	int err;
-	struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)&ep->com.local_addr;
+	struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)
+				    &ep->com.mapped_local_addr;
 
 	c4iw_init_wr_wait(&ep->com.wr_wait);
 	err = cxgb4_create_server6(ep->com.dev->rdev.lldi.ports[0],
@@ -2892,7 +2972,8 @@ static int create_server6(struct c4iw_dev *dev, struct c4iw_listen_ep *ep)
 static int create_server4(struct c4iw_dev *dev, struct c4iw_listen_ep *ep)
 {
 	int err;
-	struct sockaddr_in *sin = (struct sockaddr_in *)&ep->com.local_addr;
+	struct sockaddr_in *sin = (struct sockaddr_in *)
+				  &ep->com.mapped_local_addr;
 
 	if (dev->rdev.lldi.enable_fw_ofld_conn) {
 		do {
@@ -2927,6 +3008,9 @@ int c4iw_create_listen(struct iw_cm_id *cm_id, int backlog)
 	int err = 0;
 	struct c4iw_dev *dev = to_c4iw_dev(cm_id->device);
 	struct c4iw_listen_ep *ep;
+	struct iwpm_dev_data pm_reg_msg;
+	struct iwpm_sa_data pm_msg;
+	int iwpm_err = 0;
 
 	might_sleep();
 
@@ -2961,6 +3045,37 @@ int c4iw_create_listen(struct iw_cm_id *cm_id, int backlog)
 		goto fail2;
 	}
 	insert_handle(dev, &dev->stid_idr, ep, ep->stid);
+
+	/* No port mapper available, go with the specified info */
+	memcpy(&ep->com.mapped_local_addr, &cm_id->local_addr,
+	       sizeof(ep->com.mapped_local_addr));
+
+	c4iw_form_reg_msg(dev, &pm_reg_msg);
+	iwpm_err = iwpm_register_pid(&pm_reg_msg, RDMA_NL_C4IW);
+	if (iwpm_err) {
+		PDBG("%s: Port Mapper reg pid fail (err = %d).\n",
+			__func__, iwpm_err);
+	}
+	if (iwpm_valid_pid() && !iwpm_err) {
+		memcpy(&pm_msg.loc_addr, &ep->com.local_addr,
+				sizeof(ep->com.local_addr));
+		iwpm_err = iwpm_add_mapping(&pm_msg, RDMA_NL_C4IW);
+		if (iwpm_err)
+			PDBG("%s: Port Mapper query fail (err = %d).\n",
+				__func__, iwpm_err);
+		else
+			memcpy(&ep->com.mapped_local_addr,
+				&pm_msg.mapped_loc_addr,
+				sizeof(ep->com.mapped_local_addr));
+	}
+	if (iwpm_create_mapinfo(&ep->com.local_addr,
+				&ep->com.mapped_local_addr, RDMA_NL_C4IW)) {
+		err = -ENOMEM;
+		goto fail3;
+	}
+	print_addr(&ep->com, __func__, "add_mapping/create_mapinfo");
+
+	set_bit(RELEASE_MAPINFO, &ep->com.flags);
 	state_set(&ep->com, LISTEN);
 	if (ep->com.local_addr.ss_family == AF_INET)
 		err = create_server4(dev, ep);
@@ -2970,6 +3085,8 @@ int c4iw_create_listen(struct iw_cm_id *cm_id, int backlog)
 		cm_id->provider_data = ep;
 		goto out;
 	}
+
+fail3:
 	cxgb4_free_stid(ep->com.dev->rdev.lldi.tids, ep->stid,
 			ep->com.local_addr.ss_family);
 fail2:

drivers/infiniband/hw/cxgb4/cq.c

@@ -940,7 +940,6 @@ struct ib_cq *c4iw_create_cq(struct ib_device *ibdev, int entries,
 		if (!mm2)
 			goto err4;
 
-		memset(&uresp, 0, sizeof(uresp));
 		uresp.qid_mask = rhp->rdev.cqmask;
 		uresp.cqid = chp->cq.cqid;
 		uresp.size = chp->cq.size;
@@ -951,7 +950,8 @@ struct ib_cq *c4iw_create_cq(struct ib_device *ibdev, int entries,
 		uresp.gts_key = ucontext->key;
 		ucontext->key += PAGE_SIZE;
 		spin_unlock(&ucontext->mmap_lock);
-		ret = ib_copy_to_udata(udata, &uresp, sizeof uresp);
+		ret = ib_copy_to_udata(udata, &uresp,
+				       sizeof(uresp) - sizeof(uresp.reserved));
 		if (ret)
 			goto err5;
 

drivers/infiniband/hw/cxgb4/device.c

@@ -77,6 +77,16 @@ struct c4iw_debugfs_data {
 	int pos;
 };
 
+/* registered cxgb4 netlink callbacks */
+static struct ibnl_client_cbs c4iw_nl_cb_table[] = {
+	[RDMA_NL_IWPM_REG_PID] = {.dump = iwpm_register_pid_cb},
+	[RDMA_NL_IWPM_ADD_MAPPING] = {.dump = iwpm_add_mapping_cb},
+	[RDMA_NL_IWPM_QUERY_MAPPING] = {.dump = iwpm_add_and_query_mapping_cb},
+	[RDMA_NL_IWPM_HANDLE_ERR] = {.dump = iwpm_mapping_error_cb},
+	[RDMA_NL_IWPM_MAPINFO] = {.dump = iwpm_mapping_info_cb},
+	[RDMA_NL_IWPM_MAPINFO_NUM] = {.dump = iwpm_ack_mapping_info_cb}
+};
+
 static int count_idrs(int id, void *p, void *data)
 {
 	int *countp = data;
@@ -113,35 +123,49 @@ static int dump_qp(int id, void *p, void *data)
 				&qp->ep->com.local_addr;
 			struct sockaddr_in *rsin = (struct sockaddr_in *)
 				&qp->ep->com.remote_addr;
+			struct sockaddr_in *mapped_lsin = (struct sockaddr_in *)
+				&qp->ep->com.mapped_local_addr;
+			struct sockaddr_in *mapped_rsin = (struct sockaddr_in *)
+				&qp->ep->com.mapped_remote_addr;
 
 			cc = snprintf(qpd->buf + qpd->pos, space,
 				      "rc qp sq id %u rq id %u state %u "
 				      "onchip %u ep tid %u state %u "
-				      "%pI4:%u->%pI4:%u\n",
+				      "%pI4:%u/%u->%pI4:%u/%u\n",
 				      qp->wq.sq.qid, qp->wq.rq.qid,
 				      (int)qp->attr.state,
 				      qp->wq.sq.flags & T4_SQ_ONCHIP,
 				      qp->ep->hwtid, (int)qp->ep->com.state,
 				      &lsin->sin_addr, ntohs(lsin->sin_port),
-				      &rsin->sin_addr, ntohs(rsin->sin_port));
+				      ntohs(mapped_lsin->sin_port),
+				      &rsin->sin_addr, ntohs(rsin->sin_port),
+				      ntohs(mapped_rsin->sin_port));
 		} else {
 			struct sockaddr_in6 *lsin6 = (struct sockaddr_in6 *)
 				&qp->ep->com.local_addr;
 			struct sockaddr_in6 *rsin6 = (struct sockaddr_in6 *)
 				&qp->ep->com.remote_addr;
+			struct sockaddr_in6 *mapped_lsin6 =
+				(struct sockaddr_in6 *)
+				&qp->ep->com.mapped_local_addr;
+			struct sockaddr_in6 *mapped_rsin6 =
+				(struct sockaddr_in6 *)
+				&qp->ep->com.mapped_remote_addr;
 
 			cc = snprintf(qpd->buf + qpd->pos, space,
 				      "rc qp sq id %u rq id %u state %u "
 				      "onchip %u ep tid %u state %u "
-				      "%pI6:%u->%pI6:%u\n",
+				      "%pI6:%u/%u->%pI6:%u/%u\n",
 				      qp->wq.sq.qid, qp->wq.rq.qid,
 				      (int)qp->attr.state,
 				      qp->wq.sq.flags & T4_SQ_ONCHIP,
 				      qp->ep->hwtid, (int)qp->ep->com.state,
 				      &lsin6->sin6_addr,
 				      ntohs(lsin6->sin6_port),
+				      ntohs(mapped_lsin6->sin6_port),
 				      &rsin6->sin6_addr,
-				      ntohs(rsin6->sin6_port));
+				      ntohs(rsin6->sin6_port),
+				      ntohs(mapped_rsin6->sin6_port));
 		}
 	} else
 		cc = snprintf(qpd->buf + qpd->pos, space,
@@ -386,31 +410,43 @@ static int dump_ep(int id, void *p, void *data)
 			&ep->com.local_addr;
 		struct sockaddr_in *rsin = (struct sockaddr_in *)
 			&ep->com.remote_addr;
+		struct sockaddr_in *mapped_lsin = (struct sockaddr_in *)
+			&ep->com.mapped_local_addr;
+		struct sockaddr_in *mapped_rsin = (struct sockaddr_in *)
+			&ep->com.mapped_remote_addr;
 
 		cc = snprintf(epd->buf + epd->pos, space,
 			      "ep %p cm_id %p qp %p state %d flags 0x%lx "
 			      "history 0x%lx hwtid %d atid %d "
-			      "%pI4:%d <-> %pI4:%d\n",
+			      "%pI4:%d/%d <-> %pI4:%d/%d\n",
 			      ep, ep->com.cm_id, ep->com.qp,
 			      (int)ep->com.state, ep->com.flags,
 			      ep->com.history, ep->hwtid, ep->atid,
 			      &lsin->sin_addr, ntohs(lsin->sin_port),
-			      &rsin->sin_addr, ntohs(rsin->sin_port));
+			      ntohs(mapped_lsin->sin_port),
+			      &rsin->sin_addr, ntohs(rsin->sin_port),
+			      ntohs(mapped_rsin->sin_port));
 	} else {
 		struct sockaddr_in6 *lsin6 = (struct sockaddr_in6 *)
 			&ep->com.local_addr;
 		struct sockaddr_in6 *rsin6 = (struct sockaddr_in6 *)
 			&ep->com.remote_addr;
+		struct sockaddr_in6 *mapped_lsin6 = (struct sockaddr_in6 *)
+			&ep->com.mapped_local_addr;
+		struct sockaddr_in6 *mapped_rsin6 = (struct sockaddr_in6 *)
+			&ep->com.mapped_remote_addr;
 
 		cc = snprintf(epd->buf + epd->pos, space,
 			      "ep %p cm_id %p qp %p state %d flags 0x%lx "
 			      "history 0x%lx hwtid %d atid %d "
-			      "%pI6:%d <-> %pI6:%d\n",
+			      "%pI6:%d/%d <-> %pI6:%d/%d\n",
 			      ep, ep->com.cm_id, ep->com.qp,
 			      (int)ep->com.state, ep->com.flags,
 			      ep->com.history, ep->hwtid, ep->atid,
 			      &lsin6->sin6_addr, ntohs(lsin6->sin6_port),
-			      &rsin6->sin6_addr, ntohs(rsin6->sin6_port));
+			      ntohs(mapped_lsin6->sin6_port),
+			      &rsin6->sin6_addr, ntohs(rsin6->sin6_port),
+			      ntohs(mapped_rsin6->sin6_port));
 	}
 	if (cc < space)
 		epd->pos += cc;
@@ -431,23 +467,29 @@ static int dump_listen_ep(int id, void *p, void *data)
 	if (ep->com.local_addr.ss_family == AF_INET) {
 		struct sockaddr_in *lsin = (struct sockaddr_in *)
 			&ep->com.local_addr;
+		struct sockaddr_in *mapped_lsin = (struct sockaddr_in *)
+			&ep->com.mapped_local_addr;
 
 		cc = snprintf(epd->buf + epd->pos, space,
 			      "ep %p cm_id %p state %d flags 0x%lx stid %d "
-			      "backlog %d %pI4:%d\n",
+			      "backlog %d %pI4:%d/%d\n",
 			      ep, ep->com.cm_id, (int)ep->com.state,
 			      ep->com.flags, ep->stid, ep->backlog,
-			      &lsin->sin_addr, ntohs(lsin->sin_port));
+			      &lsin->sin_addr, ntohs(lsin->sin_port),
+			      ntohs(mapped_lsin->sin_port));
 	} else {
 		struct sockaddr_in6 *lsin6 = (struct sockaddr_in6 *)
 			&ep->com.local_addr;
+		struct sockaddr_in6 *mapped_lsin6 = (struct sockaddr_in6 *)
+			&ep->com.mapped_local_addr;
 
 		cc = snprintf(epd->buf + epd->pos, space,
 			      "ep %p cm_id %p state %d flags 0x%lx stid %d "
-			      "backlog %d %pI6:%d\n",
+			      "backlog %d %pI6:%d/%d\n",
 			      ep, ep->com.cm_id, (int)ep->com.state,
 			      ep->com.flags, ep->stid, ep->backlog,
-			      &lsin6->sin6_addr, ntohs(lsin6->sin6_port));
+			      &lsin6->sin6_addr, ntohs(lsin6->sin6_port),
+			      ntohs(mapped_lsin6->sin6_port));
 	}
 	if (cc < space)
 		epd->pos += cc;
@@ -687,6 +729,7 @@ static void c4iw_dealloc(struct uld_ctx *ctx)
 	if (ctx->dev->rdev.oc_mw_kva)
 		iounmap(ctx->dev->rdev.oc_mw_kva);
 	ib_dealloc_device(&ctx->dev->ibdev);
+	iwpm_exit(RDMA_NL_C4IW);
 	ctx->dev = NULL;
 }
 
@@ -736,6 +779,7 @@ static struct c4iw_dev *c4iw_alloc(const struct cxgb4_lld_info *infop)
 			pci_resource_len(devp->rdev.lldi.pdev, 2));
 		if (!devp->rdev.bar2_kva) {
 			pr_err(MOD "Unable to ioremap BAR2\n");
+			ib_dealloc_device(&devp->ibdev);
 			return ERR_PTR(-EINVAL);
 		}
 	} else if (ocqp_supported(infop)) {
@@ -747,6 +791,7 @@ static struct c4iw_dev *c4iw_alloc(const struct cxgb4_lld_info *infop)
 			devp->rdev.lldi.vr->ocq.size);
 		if (!devp->rdev.oc_mw_kva) {
 			pr_err(MOD "Unable to ioremap onchip mem\n");
+			ib_dealloc_device(&devp->ibdev);
 			return ERR_PTR(-EINVAL);
 		}
 	}
@@ -780,6 +825,14 @@ static struct c4iw_dev *c4iw_alloc(const struct cxgb4_lld_info *infop)
 					c4iw_debugfs_root);
 		setup_debugfs(devp);
 	}
+
+	ret = iwpm_init(RDMA_NL_C4IW);
+	if (ret) {
+		pr_err("port mapper initialization failed with %d\n", ret);
+		ib_dealloc_device(&devp->ibdev);
+		return ERR_PTR(ret);
+	}
+
 	return devp;
 }
 
@@ -1274,6 +1327,11 @@ static int __init c4iw_init_module(void)
 		printk(KERN_WARNING MOD
 		       "could not create debugfs entry, continuing\n");
 
+	if (ibnl_add_client(RDMA_NL_C4IW, RDMA_NL_IWPM_NUM_OPS,
+			    c4iw_nl_cb_table))
+		pr_err("%s[%u]: Failed to add netlink callback\n"
+		       , __func__, __LINE__);
+
 	cxgb4_register_uld(CXGB4_ULD_RDMA, &c4iw_uld_info);
 
 	return 0;
@@ -1291,6 +1349,7 @@ static void __exit c4iw_exit_module(void)
 	}
 	mutex_unlock(&dev_mutex);
 	cxgb4_unregister_uld(CXGB4_ULD_RDMA);
+	ibnl_remove_client(RDMA_NL_C4IW);
 	c4iw_cm_term();
 	debugfs_remove_recursive(c4iw_debugfs_root);
 }

drivers/infiniband/hw/cxgb4/iw_cxgb4.h

@@ -52,6 +52,8 @@
 
 #include <rdma/ib_verbs.h>
 #include <rdma/iw_cm.h>
+#include <rdma/rdma_netlink.h>
+#include <rdma/iw_portmap.h>
 
 #include "cxgb4.h"
 #include "cxgb4_uld.h"
@@ -728,6 +730,7 @@ enum c4iw_ep_flags {
 	CLOSE_SENT		= 3,
 	TIMEOUT                 = 4,
 	QP_REFERENCED           = 5,
+	RELEASE_MAPINFO		= 6,
 };
 
 enum c4iw_ep_history {
@@ -764,6 +767,8 @@ struct c4iw_ep_common {
 	struct mutex mutex;
 	struct sockaddr_storage local_addr;
 	struct sockaddr_storage remote_addr;
+	struct sockaddr_storage mapped_local_addr;
+	struct sockaddr_storage mapped_remote_addr;
 	struct c4iw_wr_wait wr_wait;
 	unsigned long flags;
 	unsigned long history;
@@ -807,6 +812,45 @@ struct c4iw_ep {
 	unsigned int retry_count;
 };
 
+static inline void print_addr(struct c4iw_ep_common *epc, const char *func,
+			      const char *msg)
+{
+
+#define SINA(a) (&(((struct sockaddr_in *)(a))->sin_addr.s_addr))
+#define SINP(a) ntohs(((struct sockaddr_in *)(a))->sin_port)
+#define SIN6A(a) (&(((struct sockaddr_in6 *)(a))->sin6_addr))
+#define SIN6P(a) ntohs(((struct sockaddr_in6 *)(a))->sin6_port)
+
+	if (c4iw_debug) {
+		switch (epc->local_addr.ss_family) {
+		case AF_INET:
+			PDBG("%s %s %pI4:%u/%u <-> %pI4:%u/%u\n",
+			     func, msg, SINA(&epc->local_addr),
+			     SINP(&epc->local_addr),
+			     SINP(&epc->mapped_local_addr),
+			     SINA(&epc->remote_addr),
+			     SINP(&epc->remote_addr),
+			     SINP(&epc->mapped_remote_addr));
+			break;
+		case AF_INET6:
+			PDBG("%s %s %pI6:%u/%u <-> %pI6:%u/%u\n",
+			     func, msg, SIN6A(&epc->local_addr),
+			     SIN6P(&epc->local_addr),
+			     SIN6P(&epc->mapped_local_addr),
+			     SIN6A(&epc->remote_addr),
+			     SIN6P(&epc->remote_addr),
+			     SIN6P(&epc->mapped_remote_addr));
+			break;
+		default:
+			break;
+		}
+	}
+#undef SINA
+#undef SINP
+#undef SIN6A
+#undef SIN6P
+}
+
 static inline struct c4iw_ep *to_ep(struct iw_cm_id *cm_id)
 {
 	return cm_id->provider_data;

drivers/infiniband/hw/cxgb4/provider.c

@@ -122,7 +122,7 @@ static struct ib_ucontext *c4iw_alloc_ucontext(struct ib_device *ibdev,
 	INIT_LIST_HEAD(&context->mmaps);
 	spin_lock_init(&context->mmap_lock);
 
-	if (udata->outlen < sizeof(uresp)) {
+	if (udata->outlen < sizeof(uresp) - sizeof(uresp.reserved)) {
 		if (!warned++)
 			pr_err(MOD "Warning - downlevel libcxgb4 (non-fatal), device status page disabled.");
 		rhp->rdev.flags |= T4_STATUS_PAGE_DISABLED;
@@ -140,7 +140,8 @@ static struct ib_ucontext *c4iw_alloc_ucontext(struct ib_device *ibdev,
 		context->key += PAGE_SIZE;
 		spin_unlock(&context->mmap_lock);
 
-		ret = ib_copy_to_udata(udata, &uresp, sizeof(uresp));
+		ret = ib_copy_to_udata(udata, &uresp,
+				       sizeof(uresp) - sizeof(uresp.reserved));
 		if (ret)
 			goto err_mm;
 

drivers/infiniband/hw/cxgb4/user.h

@@ -48,6 +48,7 @@ struct c4iw_create_cq_resp {
 	__u32 cqid;
 	__u32 size;
 	__u32 qid_mask;
+	__u32 reserved; /* explicit padding (optional for i386) */
 };
 
 
@@ -74,5 +75,6 @@ struct c4iw_create_qp_resp {
 struct c4iw_alloc_ucontext_resp {
 	__u64 status_page_key;
 	__u32 status_page_size;
+	__u32 reserved; /* explicit padding (optional for i386) */
 };
 #endif

drivers/infiniband/hw/ipath/ipath_diag.c

@@ -346,6 +346,10 @@ static ssize_t ipath_diagpkt_write(struct file *fp,
 			ret = -EFAULT;
 			goto bail;
 		}
+		dp.len = odp.len;
+		dp.unit = odp.unit;
+		dp.data = odp.data;
+		dp.pbc_wd = 0;
 	} else {
 		ret = -EINVAL;
 		goto bail;

drivers/infiniband/hw/ipath/ipath_intr.c

@@ -70,7 +70,7 @@ void ipath_disarm_senderrbufs(struct ipath_devdata *dd)
 	if (sbuf[0] || sbuf[1] || (piobcnt > 128 && (sbuf[2] || sbuf[3]))) {
 		int i;
 		if (ipath_debug & (__IPATH_PKTDBG|__IPATH_DBG) &&
-			dd->ipath_lastcancel > jiffies) {
+			time_after(dd->ipath_lastcancel, jiffies)) {
 			__IPATH_DBG_WHICH(__IPATH_PKTDBG|__IPATH_DBG,
 					  "SendbufErrs %lx %lx", sbuf[0],
 					  sbuf[1]);
@@ -755,7 +755,7 @@ static int handle_errors(struct ipath_devdata *dd, ipath_err_t errs)
 
 	/* likely due to cancel; so suppress message unless verbose */
 	if ((errs & (INFINIPATH_E_SPKTLEN | INFINIPATH_E_SPIOARMLAUNCH)) &&
-		dd->ipath_lastcancel > jiffies) {
+		time_after(dd->ipath_lastcancel, jiffies)) {
 		/* armlaunch takes precedence; it often causes both. */
 		ipath_cdbg(VERBOSE,
 			"Suppressed %s error (%llx) after sendbuf cancel\n",

drivers/infiniband/hw/ipath/ipath_sdma.c

@@ -247,7 +247,7 @@ static void sdma_abort_task(unsigned long opaque)
 
 	/* ipath_sdma_abort() is done, waiting for interrupt */
 	if (status == IPATH_SDMA_ABORT_DISARMED) {
-		if (jiffies < dd->ipath_sdma_abort_intr_timeout)
+		if (time_before(jiffies, dd->ipath_sdma_abort_intr_timeout))
 			goto resched_noprint;
 		/* give up, intr got lost somewhere */
 		ipath_dbg("give up waiting for SDMADISABLED intr\n");
@@ -341,7 +341,7 @@ static void sdma_abort_task(unsigned long opaque)
 	 * JAG - this is bad to just have default be a loop without
 	 * state change
 	 */
-	if (jiffies > dd->ipath_sdma_abort_jiffies) {
+	if (time_after(jiffies, dd->ipath_sdma_abort_jiffies)) {
 		ipath_dbg("looping with status 0x%08lx\n",
 			  dd->ipath_sdma_status);
 		dd->ipath_sdma_abort_jiffies = jiffies + 5 * HZ;

drivers/infiniband/hw/mlx4/ah.c

@@ -73,7 +73,7 @@ static struct ib_ah *create_iboe_ah(struct ib_pd *pd, struct ib_ah_attr *ah_attr
 {
 	struct mlx4_ib_dev *ibdev = to_mdev(pd->device);
 	struct mlx4_dev *dev = ibdev->dev;
-	int is_mcast;
+	int is_mcast = 0;
 	struct in6_addr in6;
 	u16 vlan_tag;
 

drivers/infiniband/hw/mlx4/cq.c

@@ -102,7 +102,7 @@ static int mlx4_ib_alloc_cq_buf(struct mlx4_ib_dev *dev, struct mlx4_ib_cq_buf *
 	int err;
 
 	err = mlx4_buf_alloc(dev->dev, nent * dev->dev->caps.cqe_size,
-			     PAGE_SIZE * 2, &buf->buf);
+			     PAGE_SIZE * 2, &buf->buf, GFP_KERNEL);
 
 	if (err)
 		goto out;
@@ -113,7 +113,7 @@ static int mlx4_ib_alloc_cq_buf(struct mlx4_ib_dev *dev, struct mlx4_ib_cq_buf *
 	if (err)
 		goto err_buf;
 
-	err = mlx4_buf_write_mtt(dev->dev, &buf->mtt, &buf->buf);
+	err = mlx4_buf_write_mtt(dev->dev, &buf->mtt, &buf->buf, GFP_KERNEL);
 	if (err)
 		goto err_mtt;
 
@@ -209,7 +209,7 @@ struct ib_cq *mlx4_ib_create_cq(struct ib_device *ibdev, int entries, int vector
 
 		uar = &to_mucontext(context)->uar;
 	} else {
-		err = mlx4_db_alloc(dev->dev, &cq->db, 1);
+		err = mlx4_db_alloc(dev->dev, &cq->db, 1, GFP_KERNEL);
 		if (err)
 			goto err_cq;
 

drivers/infiniband/hw/mlx4/mad.c

@@ -478,10 +478,6 @@ int mlx4_ib_send_to_slave(struct mlx4_ib_dev *dev, int slave, u8 port,
 	if (!tun_ctx || tun_ctx->state != DEMUX_PV_STATE_ACTIVE)
 		return -EAGAIN;
 
-	/* QP0 forwarding only for Dom0 */
-	if (!dest_qpt && (mlx4_master_func_num(dev->dev) != slave))
-		return -EINVAL;
-
 	if (!dest_qpt)
 		tun_qp = &tun_ctx->qp[0];
 	else
@@ -667,6 +663,21 @@ static int mlx4_ib_demux_mad(struct ib_device *ibdev, u8 port,
 	}
 	/* Class-specific handling */
 	switch (mad->mad_hdr.mgmt_class) {
+	case IB_MGMT_CLASS_SUBN_LID_ROUTED:
+	case IB_MGMT_CLASS_SUBN_DIRECTED_ROUTE:
+		/* 255 indicates the dom0 */
+		if (slave != 255 && slave != mlx4_master_func_num(dev->dev)) {
+			if (!mlx4_vf_smi_enabled(dev->dev, slave, port))
+				return -EPERM;
+			/* for a VF. drop unsolicited MADs */
+			if (!(mad->mad_hdr.method & IB_MGMT_METHOD_RESP)) {
+				mlx4_ib_warn(ibdev, "demux QP0. rejecting unsolicited mad for slave %d class 0x%x, method 0x%x\n",
+					     slave, mad->mad_hdr.mgmt_class,
+					     mad->mad_hdr.method);
+				return -EINVAL;
+			}
+		}
+		break;
 	case IB_MGMT_CLASS_SUBN_ADM:
 		if (mlx4_ib_demux_sa_handler(ibdev, port, slave,
 					     (struct ib_sa_mad *) mad))
@@ -1165,10 +1176,6 @@ int mlx4_ib_send_to_wire(struct mlx4_ib_dev *dev, int slave, u8 port,
 	if (!sqp_ctx || sqp_ctx->state != DEMUX_PV_STATE_ACTIVE)
 		return -EAGAIN;
 
-	/* QP0 forwarding only for Dom0 */
-	if (dest_qpt == IB_QPT_SMI && (mlx4_master_func_num(dev->dev) != slave))
-		return -EINVAL;
-
 	if (dest_qpt == IB_QPT_SMI) {
 		src_qpnum = 0;
 		sqp = &sqp_ctx->qp[0];
@@ -1285,11 +1292,6 @@ static void mlx4_ib_multiplex_mad(struct mlx4_ib_demux_pv_ctx *ctx, struct ib_wc
 			     "belongs to another slave\n", wc->src_qp);
 		return;
 	}
-	if (slave != mlx4_master_func_num(dev->dev) && !(wc->src_qp & 0x2)) {
-		mlx4_ib_warn(ctx->ib_dev, "can't multiplex bad sqp:%d: "
-			     "non-master trying to send QP0 packets\n", wc->src_qp);
-		return;
-	}
 
 	/* Map transaction ID */
 	ib_dma_sync_single_for_cpu(ctx->ib_dev, tun_qp->ring[wr_ix].map,
@@ -1317,6 +1319,12 @@ static void mlx4_ib_multiplex_mad(struct mlx4_ib_demux_pv_ctx *ctx, struct ib_wc
 
 	/* Class-specific handling */
 	switch (tunnel->mad.mad_hdr.mgmt_class) {
+	case IB_MGMT_CLASS_SUBN_LID_ROUTED:
+	case IB_MGMT_CLASS_SUBN_DIRECTED_ROUTE:
+		if (slave != mlx4_master_func_num(dev->dev) &&
+		    !mlx4_vf_smi_enabled(dev->dev, slave, ctx->port))
+			return;
+		break;
 	case IB_MGMT_CLASS_SUBN_ADM:
 		if (mlx4_ib_multiplex_sa_handler(ctx->ib_dev, ctx->port, slave,
 			      (struct ib_sa_mad *) &tunnel->mad))
@@ -1749,9 +1757,9 @@ static int create_pv_resources(struct ib_device *ibdev, int slave, int port,
 		return -EEXIST;
 
 	ctx->state = DEMUX_PV_STATE_STARTING;
-	/* have QP0 only on port owner, and only if link layer is IB */
-	if (ctx->slave == mlx4_master_func_num(to_mdev(ctx->ib_dev)->dev) &&
-	    rdma_port_get_link_layer(ibdev, ctx->port) == IB_LINK_LAYER_INFINIBAND)
+	/* have QP0 only if link layer is IB */
+	if (rdma_port_get_link_layer(ibdev, ctx->port) ==
+	    IB_LINK_LAYER_INFINIBAND)
 		ctx->has_smi = 1;
 
 	if (ctx->has_smi) {

drivers/infiniband/hw/mlx4/main.c

@@ -544,12 +544,11 @@ static int mlx4_ib_modify_device(struct ib_device *ibdev, int mask,
 	return 0;
 }
 
-static int mlx4_SET_PORT(struct mlx4_ib_dev *dev, u8 port, int reset_qkey_viols,
-			 u32 cap_mask)
+static int mlx4_ib_SET_PORT(struct mlx4_ib_dev *dev, u8 port, int reset_qkey_viols,
+			    u32 cap_mask)
 {
 	struct mlx4_cmd_mailbox *mailbox;
 	int err;
-	u8 is_eth = dev->dev->caps.port_type[port] == MLX4_PORT_TYPE_ETH;
 
 	mailbox = mlx4_alloc_cmd_mailbox(dev->dev);
 	if (IS_ERR(mailbox))
@@ -563,8 +562,8 @@ static int mlx4_SET_PORT(struct mlx4_ib_dev *dev, u8 port, int reset_qkey_viols,
 		((__be32 *) mailbox->buf)[1] = cpu_to_be32(cap_mask);
 	}
 
-	err = mlx4_cmd(dev->dev, mailbox->dma, port, is_eth, MLX4_CMD_SET_PORT,
-		       MLX4_CMD_TIME_CLASS_B, MLX4_CMD_NATIVE);
+	err = mlx4_cmd(dev->dev, mailbox->dma, port, 0, MLX4_CMD_SET_PORT,
+		       MLX4_CMD_TIME_CLASS_B, MLX4_CMD_WRAPPED);
 
 	mlx4_free_cmd_mailbox(dev->dev, mailbox);
 	return err;
@@ -573,11 +572,20 @@ static int mlx4_SET_PORT(struct mlx4_ib_dev *dev, u8 port, int reset_qkey_viols,
 static int mlx4_ib_modify_port(struct ib_device *ibdev, u8 port, int mask,
 			       struct ib_port_modify *props)
 {
+	struct mlx4_ib_dev *mdev = to_mdev(ibdev);
+	u8 is_eth = mdev->dev->caps.port_type[port] == MLX4_PORT_TYPE_ETH;
 	struct ib_port_attr attr;
 	u32 cap_mask;
 	int err;
 
-	mutex_lock(&to_mdev(ibdev)->cap_mask_mutex);
+	/* return OK if this is RoCE. CM calls ib_modify_port() regardless
+	 * of whether port link layer is ETH or IB. For ETH ports, qkey
+	 * violations and port capabilities are not meaningful.
+	 */
+	if (is_eth)
+		return 0;
+
+	mutex_lock(&mdev->cap_mask_mutex);
 
 	err = mlx4_ib_query_port(ibdev, port, &attr);
 	if (err)
@@ -586,9 +594,9 @@ static int mlx4_ib_modify_port(struct ib_device *ibdev, u8 port, int mask,
 	cap_mask = (attr.port_cap_flags | props->set_port_cap_mask) &
 		~props->clr_port_cap_mask;
 
-	err = mlx4_SET_PORT(to_mdev(ibdev), port,
-			    !!(mask & IB_PORT_RESET_QKEY_CNTR),
-			    cap_mask);
+	err = mlx4_ib_SET_PORT(mdev, port,
+			       !!(mask & IB_PORT_RESET_QKEY_CNTR),
+			       cap_mask);
 
 out:
 	mutex_unlock(&to_mdev(ibdev)->cap_mask_mutex);

drivers/infiniband/hw/mlx4/mlx4_ib.h

@@ -156,6 +156,7 @@ enum mlx4_ib_qp_flags {
 	MLX4_IB_QP_LSO = IB_QP_CREATE_IPOIB_UD_LSO,
 	MLX4_IB_QP_BLOCK_MULTICAST_LOOPBACK = IB_QP_CREATE_BLOCK_MULTICAST_LOOPBACK,
 	MLX4_IB_QP_NETIF = IB_QP_CREATE_NETIF_QP,
+	MLX4_IB_QP_CREATE_USE_GFP_NOIO = IB_QP_CREATE_USE_GFP_NOIO,
 	MLX4_IB_SRIOV_TUNNEL_QP = 1 << 30,
 	MLX4_IB_SRIOV_SQP = 1 << 31,
 };

drivers/infiniband/hw/mlx4/qp.c

@@ -608,9 +608,20 @@ static int qp_has_rq(struct ib_qp_init_attr *attr)
 	return !attr->srq;
 }
 
+static int qp0_enabled_vf(struct mlx4_dev *dev, int qpn)
+{
+	int i;
+	for (i = 0; i < dev->caps.num_ports; i++) {
+		if (qpn == dev->caps.qp0_proxy[i])
+			return !!dev->caps.qp0_qkey[i];
+	}
+	return 0;
+}
+
 static int create_qp_common(struct mlx4_ib_dev *dev, struct ib_pd *pd,
 			    struct ib_qp_init_attr *init_attr,
-			    struct ib_udata *udata, int sqpn, struct mlx4_ib_qp **caller_qp)
+			    struct ib_udata *udata, int sqpn, struct mlx4_ib_qp **caller_qp,
+			    gfp_t gfp)
 {
 	int qpn;
 	int err;
@@ -625,10 +636,13 @@ static int create_qp_common(struct mlx4_ib_dev *dev, struct ib_pd *pd,
 		     !(init_attr->create_flags & MLX4_IB_SRIOV_SQP))) {
 			if (init_attr->qp_type == IB_QPT_GSI)
 				qp_type = MLX4_IB_QPT_PROXY_GSI;
-			else if (mlx4_is_master(dev->dev))
-				qp_type = MLX4_IB_QPT_PROXY_SMI_OWNER;
-			else
-				qp_type = MLX4_IB_QPT_PROXY_SMI;
+			else {
+				if (mlx4_is_master(dev->dev) ||
+				    qp0_enabled_vf(dev->dev, sqpn))
+					qp_type = MLX4_IB_QPT_PROXY_SMI_OWNER;
+				else
+					qp_type = MLX4_IB_QPT_PROXY_SMI;
+			}
 		}
 		qpn = sqpn;
 		/* add extra sg entry for tunneling */
@@ -643,7 +657,9 @@ static int create_qp_common(struct mlx4_ib_dev *dev, struct ib_pd *pd,
 			return -EINVAL;
 		if (tnl_init->proxy_qp_type == IB_QPT_GSI)
 			qp_type = MLX4_IB_QPT_TUN_GSI;
-		else if (tnl_init->slave == mlx4_master_func_num(dev->dev))
+		else if (tnl_init->slave == mlx4_master_func_num(dev->dev) ||
+			 mlx4_vf_smi_enabled(dev->dev, tnl_init->slave,
+					     tnl_init->port))
 			qp_type = MLX4_IB_QPT_TUN_SMI_OWNER;
 		else
 			qp_type = MLX4_IB_QPT_TUN_SMI;
@@ -658,14 +674,14 @@ static int create_qp_common(struct mlx4_ib_dev *dev, struct ib_pd *pd,
 		if (qp_type == MLX4_IB_QPT_SMI || qp_type == MLX4_IB_QPT_GSI ||
 		    (qp_type & (MLX4_IB_QPT_PROXY_SMI | MLX4_IB_QPT_PROXY_SMI_OWNER |
 				MLX4_IB_QPT_PROXY_GSI | MLX4_IB_QPT_TUN_SMI_OWNER))) {
-			sqp = kzalloc(sizeof (struct mlx4_ib_sqp), GFP_KERNEL);
+			sqp = kzalloc(sizeof (struct mlx4_ib_sqp), gfp);
 			if (!sqp)
 				return -ENOMEM;
 			qp = &sqp->qp;
 			qp->pri.vid = 0xFFFF;
 			qp->alt.vid = 0xFFFF;
 		} else {
-			qp = kzalloc(sizeof (struct mlx4_ib_qp), GFP_KERNEL);
+			qp = kzalloc(sizeof (struct mlx4_ib_qp), gfp);
 			if (!qp)
 				return -ENOMEM;
 			qp->pri.vid = 0xFFFF;
@@ -748,14 +764,14 @@ static int create_qp_common(struct mlx4_ib_dev *dev, struct ib_pd *pd,
 			goto err;
 
 		if (qp_has_rq(init_attr)) {
-			err = mlx4_db_alloc(dev->dev, &qp->db, 0);
+			err = mlx4_db_alloc(dev->dev, &qp->db, 0, gfp);
 			if (err)
 				goto err;
 
 			*qp->db.db = 0;
 		}
 
-		if (mlx4_buf_alloc(dev->dev, qp->buf_size, PAGE_SIZE * 2, &qp->buf)) {
+		if (mlx4_buf_alloc(dev->dev, qp->buf_size, PAGE_SIZE * 2, &qp->buf, gfp)) {
 			err = -ENOMEM;
 			goto err_db;
 		}
@@ -765,13 +781,12 @@ static int create_qp_common(struct mlx4_ib_dev *dev, struct ib_pd *pd,
 		if (err)
 			goto err_buf;
 
-		err = mlx4_buf_write_mtt(dev->dev, &qp->mtt, &qp->buf);
+		err = mlx4_buf_write_mtt(dev->dev, &qp->mtt, &qp->buf, gfp);
 		if (err)
 			goto err_mtt;
 
-		qp->sq.wrid  = kmalloc(qp->sq.wqe_cnt * sizeof (u64), GFP_KERNEL);
-		qp->rq.wrid  = kmalloc(qp->rq.wqe_cnt * sizeof (u64), GFP_KERNEL);
-
+		qp->sq.wrid  = kmalloc(qp->sq.wqe_cnt * sizeof (u64), gfp);
+		qp->rq.wrid  = kmalloc(qp->rq.wqe_cnt * sizeof (u64), gfp);
 		if (!qp->sq.wrid || !qp->rq.wrid) {
 			err = -ENOMEM;
 			goto err_wrid;
@@ -801,7 +816,7 @@ static int create_qp_common(struct mlx4_ib_dev *dev, struct ib_pd *pd,
 			goto err_proxy;
 	}
 
-	err = mlx4_qp_alloc(dev->dev, qpn, &qp->mqp);
+	err = mlx4_qp_alloc(dev->dev, qpn, &qp->mqp, gfp);
 	if (err)
 		goto err_qpn;
 
@@ -1040,7 +1055,10 @@ struct ib_qp *mlx4_ib_create_qp(struct ib_pd *pd,
 	struct mlx4_ib_qp *qp = NULL;
 	int err;
 	u16 xrcdn = 0;
+	gfp_t gfp;
 
+	gfp = (init_attr->create_flags & MLX4_IB_QP_CREATE_USE_GFP_NOIO) ?
+		GFP_NOIO : GFP_KERNEL;
 	/*
 	 * We only support LSO, vendor flag1, and multicast loopback blocking,
 	 * and only for kernel UD QPs.
@@ -1049,7 +1067,8 @@ struct ib_qp *mlx4_ib_create_qp(struct ib_pd *pd,
 					MLX4_IB_QP_BLOCK_MULTICAST_LOOPBACK |
 					MLX4_IB_SRIOV_TUNNEL_QP |
 					MLX4_IB_SRIOV_SQP |
-					MLX4_IB_QP_NETIF))
+					MLX4_IB_QP_NETIF |
+					MLX4_IB_QP_CREATE_USE_GFP_NOIO))
 		return ERR_PTR(-EINVAL);
 
 	if (init_attr->create_flags & IB_QP_CREATE_NETIF_QP) {
@@ -1059,7 +1078,7 @@ struct ib_qp *mlx4_ib_create_qp(struct ib_pd *pd,
 
 	if (init_attr->create_flags &&
 	    (udata ||
-	     ((init_attr->create_flags & ~MLX4_IB_SRIOV_SQP) &&
+	     ((init_attr->create_flags & ~(MLX4_IB_SRIOV_SQP | MLX4_IB_QP_CREATE_USE_GFP_NOIO)) &&
 	      init_attr->qp_type != IB_QPT_UD) ||
 	     ((init_attr->create_flags & MLX4_IB_SRIOV_SQP) &&
 	      init_attr->qp_type > IB_QPT_GSI)))
@@ -1079,7 +1098,7 @@ struct ib_qp *mlx4_ib_create_qp(struct ib_pd *pd,
 	case IB_QPT_RC:
 	case IB_QPT_UC:
 	case IB_QPT_RAW_PACKET:
-		qp = kzalloc(sizeof *qp, GFP_KERNEL);
+		qp = kzalloc(sizeof *qp, gfp);
 		if (!qp)
 			return ERR_PTR(-ENOMEM);
 		qp->pri.vid = 0xFFFF;
@@ -1088,7 +1107,7 @@ struct ib_qp *mlx4_ib_create_qp(struct ib_pd *pd,
 	case IB_QPT_UD:
 	{
 		err = create_qp_common(to_mdev(pd->device), pd, init_attr,
-				       udata, 0, &qp);
+				       udata, 0, &qp, gfp);
 		if (err)
 			return ERR_PTR(err);
 
@@ -1106,7 +1125,7 @@ struct ib_qp *mlx4_ib_create_qp(struct ib_pd *pd,
 
 		err = create_qp_common(to_mdev(pd->device), pd, init_attr, udata,
 				       get_sqp_num(to_mdev(pd->device), init_attr),
-				       &qp);
+				       &qp, gfp);
 		if (err)
 			return ERR_PTR(err);
 
@@ -1930,6 +1949,19 @@ int mlx4_ib_modify_qp(struct ib_qp *ibqp, struct ib_qp_attr *attr,
 	return err;
 }
 
+static int vf_get_qp0_qkey(struct mlx4_dev *dev, int qpn, u32 *qkey)
+{
+	int i;
+	for (i = 0; i < dev->caps.num_ports; i++) {
+		if (qpn == dev->caps.qp0_proxy[i] ||
+		    qpn == dev->caps.qp0_tunnel[i]) {
+			*qkey = dev->caps.qp0_qkey[i];
+			return 0;
+		}
+	}
+	return -EINVAL;
+}
+
 static int build_sriov_qp0_header(struct mlx4_ib_sqp *sqp,
 				  struct ib_send_wr *wr,
 				  void *wqe, unsigned *mlx_seg_len)
@@ -1987,8 +2019,13 @@ static int build_sriov_qp0_header(struct mlx4_ib_sqp *sqp,
 			cpu_to_be32(mdev->dev->caps.qp0_tunnel[sqp->qp.port - 1]);
 
 	sqp->ud_header.bth.psn = cpu_to_be32((sqp->send_psn++) & ((1 << 24) - 1));
-	if (mlx4_get_parav_qkey(mdev->dev, sqp->qp.mqp.qpn, &qkey))
-		return -EINVAL;
+	if (mlx4_is_master(mdev->dev)) {
+		if (mlx4_get_parav_qkey(mdev->dev, sqp->qp.mqp.qpn, &qkey))
+			return -EINVAL;
+	} else {
+		if (vf_get_qp0_qkey(mdev->dev, sqp->qp.mqp.qpn, &qkey))
+			return -EINVAL;
+	}
 	sqp->ud_header.deth.qkey = cpu_to_be32(qkey);
 	sqp->ud_header.deth.source_qpn = cpu_to_be32(sqp->qp.mqp.qpn);
 
@@ -2370,7 +2407,8 @@ static void set_datagram_seg(struct mlx4_wqe_datagram_seg *dseg,
 
 static void set_tunnel_datagram_seg(struct mlx4_ib_dev *dev,
 				    struct mlx4_wqe_datagram_seg *dseg,
-				    struct ib_send_wr *wr, enum ib_qp_type qpt)
+				    struct ib_send_wr *wr,
+				    enum mlx4_ib_qp_type qpt)
 {
 	union mlx4_ext_av *av = &to_mah(wr->wr.ud.ah)->av;
 	struct mlx4_av sqp_av = {0};
@@ -2383,8 +2421,10 @@ static void set_tunnel_datagram_seg(struct mlx4_ib_dev *dev,
 			cpu_to_be32(0xf0000000);
 
 	memcpy(dseg->av, &sqp_av, sizeof (struct mlx4_av));
-	/* This function used only for sending on QP1 proxies */
-	dseg->dqpn = cpu_to_be32(dev->dev->caps.qp1_tunnel[port - 1]);
+	if (qpt == MLX4_IB_QPT_PROXY_GSI)
+		dseg->dqpn = cpu_to_be32(dev->dev->caps.qp1_tunnel[port - 1]);
+	else
+		dseg->dqpn = cpu_to_be32(dev->dev->caps.qp0_tunnel[port - 1]);
 	/* Use QKEY from the QP context, which is set by master */
 	dseg->qkey = cpu_to_be32(IB_QP_SET_QKEY);
 }
@@ -2679,11 +2719,6 @@ int mlx4_ib_post_send(struct ib_qp *ibqp, struct ib_send_wr *wr,
 			break;
 
 		case MLX4_IB_QPT_PROXY_SMI_OWNER:
-			if (unlikely(!mlx4_is_master(to_mdev(ibqp->device)->dev))) {
-				err = -ENOSYS;
-				*bad_wr = wr;
-				goto out;
-			}
 			err = build_sriov_qp0_header(to_msqp(qp), wr, ctrl, &seglen);
 			if (unlikely(err)) {
 				*bad_wr = wr;
@@ -2700,16 +2735,13 @@ int mlx4_ib_post_send(struct ib_qp *ibqp, struct ib_send_wr *wr,
 			size += seglen / 16;
 			break;
 		case MLX4_IB_QPT_PROXY_SMI:
-			/* don't allow QP0 sends on guests */
-			err = -ENOSYS;
-			*bad_wr = wr;
-			goto out;
 		case MLX4_IB_QPT_PROXY_GSI:
 			/* If we are tunneling special qps, this is a UD qp.
 			 * In this case we first add a UD segment targeting
 			 * the tunnel qp, and then add a header with address
 			 * information */
-			set_tunnel_datagram_seg(to_mdev(ibqp->device), wqe, wr, ibqp->qp_type);
+			set_tunnel_datagram_seg(to_mdev(ibqp->device), wqe, wr,
+						qp->mlx4_ib_qp_type);
 			wqe  += sizeof (struct mlx4_wqe_datagram_seg);
 			size += sizeof (struct mlx4_wqe_datagram_seg) / 16;
 			build_tunnel_header(wr, wqe, &seglen);

drivers/infiniband/hw/mlx4/srq.c

@@ -134,13 +134,14 @@ struct ib_srq *mlx4_ib_create_srq(struct ib_pd *pd,
 		if (err)
 			goto err_mtt;
 	} else {
-		err = mlx4_db_alloc(dev->dev, &srq->db, 0);
+		err = mlx4_db_alloc(dev->dev, &srq->db, 0, GFP_KERNEL);
 		if (err)
 			goto err_srq;
 
 		*srq->db.db = 0;
 
-		if (mlx4_buf_alloc(dev->dev, buf_size, PAGE_SIZE * 2, &srq->buf)) {
+		if (mlx4_buf_alloc(dev->dev, buf_size, PAGE_SIZE * 2, &srq->buf,
+				   GFP_KERNEL)) {
 			err = -ENOMEM;
 			goto err_db;
 		}
@@ -165,7 +166,7 @@ struct ib_srq *mlx4_ib_create_srq(struct ib_pd *pd,
 		if (err)
 			goto err_buf;
 
-		err = mlx4_buf_write_mtt(dev->dev, &srq->mtt, &srq->buf);
+		err = mlx4_buf_write_mtt(dev->dev, &srq->mtt, &srq->buf, GFP_KERNEL);
 		if (err)
 			goto err_mtt;
 

drivers/infiniband/hw/mlx4/sysfs.c

@@ -389,8 +389,10 @@ struct mlx4_port {
 	struct mlx4_ib_dev    *dev;
 	struct attribute_group pkey_group;
 	struct attribute_group gid_group;
-	u8                     port_num;
+	struct device_attribute	enable_smi_admin;
+	struct device_attribute	smi_enabled;
 	int		       slave;
+	u8                     port_num;
 };
 
 
@@ -558,6 +560,101 @@ alloc_group_attrs(ssize_t (*show)(struct mlx4_port *,
 	return NULL;
 }
 
+static ssize_t sysfs_show_smi_enabled(struct device *dev,
+				      struct device_attribute *attr, char *buf)
+{
+	struct mlx4_port *p =
+		container_of(attr, struct mlx4_port, smi_enabled);
+	ssize_t len = 0;
+
+	if (mlx4_vf_smi_enabled(p->dev->dev, p->slave, p->port_num))
+		len = sprintf(buf, "%d\n", 1);
+	else
+		len = sprintf(buf, "%d\n", 0);
+
+	return len;
+}
+
+static ssize_t sysfs_show_enable_smi_admin(struct device *dev,
+					   struct device_attribute *attr,
+					   char *buf)
+{
+	struct mlx4_port *p =
+		container_of(attr, struct mlx4_port, enable_smi_admin);
+	ssize_t len = 0;
+
+	if (mlx4_vf_get_enable_smi_admin(p->dev->dev, p->slave, p->port_num))
+		len = sprintf(buf, "%d\n", 1);
+	else
+		len = sprintf(buf, "%d\n", 0);
+
+	return len;
+}
+
+static ssize_t sysfs_store_enable_smi_admin(struct device *dev,
+					    struct device_attribute *attr,
+					    const char *buf, size_t count)
+{
+	struct mlx4_port *p =
+		container_of(attr, struct mlx4_port, enable_smi_admin);
+	int enable;
+
+	if (sscanf(buf, "%i", &enable) != 1 ||
+	    enable < 0 || enable > 1)
+		return -EINVAL;
+
+	if (mlx4_vf_set_enable_smi_admin(p->dev->dev, p->slave, p->port_num, enable))
+		return -EINVAL;
+	return count;
+}
+
+static int add_vf_smi_entries(struct mlx4_port *p)
+{
+	int is_eth = rdma_port_get_link_layer(&p->dev->ib_dev, p->port_num) ==
+			IB_LINK_LAYER_ETHERNET;
+	int ret;
+
+	/* do not display entries if eth transport, or if master */
+	if (is_eth || p->slave == mlx4_master_func_num(p->dev->dev))
+		return 0;
+
+	sysfs_attr_init(&p->smi_enabled.attr);
+	p->smi_enabled.show = sysfs_show_smi_enabled;
+	p->smi_enabled.store = NULL;
+	p->smi_enabled.attr.name = "smi_enabled";
+	p->smi_enabled.attr.mode = 0444;
+	ret = sysfs_create_file(&p->kobj, &p->smi_enabled.attr);
+	if (ret) {
+		pr_err("failed to create smi_enabled\n");
+		return ret;
+	}
+
+	sysfs_attr_init(&p->enable_smi_admin.attr);
+	p->enable_smi_admin.show = sysfs_show_enable_smi_admin;
+	p->enable_smi_admin.store = sysfs_store_enable_smi_admin;
+	p->enable_smi_admin.attr.name = "enable_smi_admin";
+	p->enable_smi_admin.attr.mode = 0644;
+	ret = sysfs_create_file(&p->kobj, &p->enable_smi_admin.attr);
+	if (ret) {
+		pr_err("failed to create enable_smi_admin\n");
+		sysfs_remove_file(&p->kobj, &p->smi_enabled.attr);
+		return ret;
+	}
+	return 0;
+}
+
+static void remove_vf_smi_entries(struct mlx4_port *p)
+{
+	int is_eth = rdma_port_get_link_layer(&p->dev->ib_dev, p->port_num) ==
+			IB_LINK_LAYER_ETHERNET;
+
+	if (is_eth || p->slave == mlx4_master_func_num(p->dev->dev))
+		return;
+
+	sysfs_remove_file(&p->kobj, &p->smi_enabled.attr);
+	sysfs_remove_file(&p->kobj, &p->enable_smi_admin.attr);
+}
+
 static int add_port(struct mlx4_ib_dev *dev, int port_num, int slave)
 {
 	struct mlx4_port *p;
@@ -602,6 +699,10 @@ static int add_port(struct mlx4_ib_dev *dev, int port_num, int slave)
 	if (ret)
 		goto err_free_gid;
 
+	ret = add_vf_smi_entries(p);
+	if (ret)
+		goto err_free_gid;
+
 	list_add_tail(&p->kobj.entry, &dev->pkeys.pkey_port_list[slave]);
 	return 0;
 
@@ -669,6 +770,7 @@ static int register_one_pkey_tree(struct mlx4_ib_dev *dev, int slave)
 		mport = container_of(p, struct mlx4_port, kobj);
 		sysfs_remove_group(p, &mport->pkey_group);
 		sysfs_remove_group(p, &mport->gid_group);
+		remove_vf_smi_entries(mport);
 		kobject_put(p);
 	}
 	kobject_put(dev->dev_ports_parent[slave]);
@@ -713,6 +815,7 @@ static void unregister_pkey_tree(struct mlx4_ib_dev *device)
 			port = container_of(p, struct mlx4_port, kobj);
 			sysfs_remove_group(p, &port->pkey_group);
 			sysfs_remove_group(p, &port->gid_group);
+			remove_vf_smi_entries(port);
 			kobject_put(p);
 			kobject_put(device->dev_ports_parent[slave]);
 		}

drivers/infiniband/hw/mlx5/cq.c

@@ -32,6 +32,7 @@
 
 #include <linux/kref.h>
 #include <rdma/ib_umem.h>
+#include <rdma/ib_user_verbs.h>
 #include "mlx5_ib.h"
 #include "user.h"
 
@@ -602,14 +603,24 @@ static int create_cq_user(struct mlx5_ib_dev *dev, struct ib_udata *udata,
 			  int *cqe_size, int *index, int *inlen)
 {
 	struct mlx5_ib_create_cq ucmd;
+	size_t ucmdlen;
 	int page_shift;
 	int npages;
 	int ncont;
 	int err;
 
-	if (ib_copy_from_udata(&ucmd, udata, sizeof(ucmd)))
+	ucmdlen =
+		(udata->inlen - sizeof(struct ib_uverbs_cmd_hdr) <
+		 sizeof(ucmd)) ? (sizeof(ucmd) -
+				  sizeof(ucmd.reserved)) : sizeof(ucmd);
+
+	if (ib_copy_from_udata(&ucmd, udata, ucmdlen))
 		return -EFAULT;
 
+	if (ucmdlen == sizeof(ucmd) &&
+	    ucmd.reserved != 0)
+		return -EINVAL;
+
 	if (ucmd.cqe_size != 64 && ucmd.cqe_size != 128)
 		return -EINVAL;
 

drivers/infiniband/hw/mlx5/mlx5_ib.h

@@ -264,8 +264,6 @@ struct mlx5_ib_mr {
 	__be64			*pas;
 	dma_addr_t		dma;
 	int			npages;
-	struct completion	done;
-	enum ib_wc_status	status;
 	struct mlx5_ib_dev     *dev;
 	struct mlx5_create_mkey_mbox_out out;
 	struct mlx5_core_sig_ctx    *sig;
@@ -277,6 +275,17 @@ struct mlx5_ib_fast_reg_page_list {
 	dma_addr_t			map;
 };
 
+struct mlx5_ib_umr_context {
+	enum ib_wc_status	status;
+	struct completion	done;
+};
+
+static inline void mlx5_ib_init_umr_context(struct mlx5_ib_umr_context *context)
+{
+	context->status = -1;
+	init_completion(&context->done);
+}
+
 struct umr_common {
 	struct ib_pd	*pd;
 	struct ib_cq	*cq;

drivers/infiniband/hw/mlx5/mr.c

@@ -73,6 +73,8 @@ static void reg_mr_callback(int status, void *context)
 	struct mlx5_cache_ent *ent = &cache->ent[c];
 	u8 key;
 	unsigned long flags;
+	struct mlx5_mr_table *table = &dev->mdev.priv.mr_table;
+	int err;
 
 	spin_lock_irqsave(&ent->lock, flags);
 	ent->pending--;
@@ -107,6 +109,13 @@ static void reg_mr_callback(int status, void *context)
 	ent->cur++;
 	ent->size++;
 	spin_unlock_irqrestore(&ent->lock, flags);
+
+	write_lock_irqsave(&table->lock, flags);
+	err = radix_tree_insert(&table->tree, mlx5_base_mkey(mr->mmr.key),
+				&mr->mmr);
+	if (err)
+		pr_err("Error inserting to mr tree. 0x%x\n", -err);
+	write_unlock_irqrestore(&table->lock, flags);
 }
 
 static int add_keys(struct mlx5_ib_dev *dev, int c, int num)
@@ -699,7 +708,7 @@ static void prep_umr_unreg_wqe(struct mlx5_ib_dev *dev,
 
 void mlx5_umr_cq_handler(struct ib_cq *cq, void *cq_context)
 {
-	struct mlx5_ib_mr *mr;
+	struct mlx5_ib_umr_context *context;
 	struct ib_wc wc;
 	int err;
 
@@ -712,9 +721,9 @@ void mlx5_umr_cq_handler(struct ib_cq *cq, void *cq_context)
 		if (err == 0)
 			break;
 
-		mr = (struct mlx5_ib_mr *)(unsigned long)wc.wr_id;
-		mr->status = wc.status;
-		complete(&mr->done);
+		context = (struct mlx5_ib_umr_context *) (unsigned long) wc.wr_id;
+		context->status = wc.status;
+		complete(&context->done);
 	}
 	ib_req_notify_cq(cq, IB_CQ_NEXT_COMP);
 }
@@ -726,11 +735,12 @@ static struct mlx5_ib_mr *reg_umr(struct ib_pd *pd, struct ib_umem *umem,
 	struct mlx5_ib_dev *dev = to_mdev(pd->device);
 	struct device *ddev = dev->ib_dev.dma_device;
 	struct umr_common *umrc = &dev->umrc;
+	struct mlx5_ib_umr_context umr_context;
 	struct ib_send_wr wr, *bad;
 	struct mlx5_ib_mr *mr;
 	struct ib_sge sg;
 	int size = sizeof(u64) * npages;
-	int err;
+	int err = 0;
 	int i;
 
 	for (i = 0; i < 1; i++) {
@@ -751,7 +761,7 @@ static struct mlx5_ib_mr *reg_umr(struct ib_pd *pd, struct ib_umem *umem,
 	mr->pas = kmalloc(size + MLX5_UMR_ALIGN - 1, GFP_KERNEL);
 	if (!mr->pas) {
 		err = -ENOMEM;
-		goto error;
+		goto free_mr;
 	}
 
 	mlx5_ib_populate_pas(dev, umem, page_shift,
@@ -760,44 +770,46 @@ static struct mlx5_ib_mr *reg_umr(struct ib_pd *pd, struct ib_umem *umem,
 	mr->dma = dma_map_single(ddev, mr_align(mr->pas, MLX5_UMR_ALIGN), size,
 				 DMA_TO_DEVICE);
 	if (dma_mapping_error(ddev, mr->dma)) {
-		kfree(mr->pas);
 		err = -ENOMEM;
-		goto error;
+		goto free_pas;
 	}
 
 	memset(&wr, 0, sizeof(wr));
-	wr.wr_id = (u64)(unsigned long)mr;
+	wr.wr_id = (u64)(unsigned long)&umr_context;
 	prep_umr_reg_wqe(pd, &wr, &sg, mr->dma, npages, mr->mmr.key, page_shift, virt_addr, len, access_flags);
 
-	/* We serialize polls so one process does not kidnap another's
-	 * completion. This is not a problem since wr is completed in
-	 * around 1 usec
-	 */
+	mlx5_ib_init_umr_context(&umr_context);
 	down(&umrc->sem);
-	init_completion(&mr->done);
 	err = ib_post_send(umrc->qp, &wr, &bad);
 	if (err) {
 		mlx5_ib_warn(dev, "post send failed, err %d\n", err);
-		up(&umrc->sem);
-		goto error;
+		goto unmap_dma;
+	} else {
+		wait_for_completion(&umr_context.done);
+		if (umr_context.status != IB_WC_SUCCESS) {
+			mlx5_ib_warn(dev, "reg umr failed\n");
+			err = -EFAULT;
+		}
 	}
-	wait_for_completion(&mr->done);
-	up(&umrc->sem);
 
+	mr->mmr.iova = virt_addr;
+	mr->mmr.size = len;
+	mr->mmr.pd = to_mpd(pd)->pdn;
+
+unmap_dma:
+	up(&umrc->sem);
 	dma_unmap_single(ddev, mr->dma, size, DMA_TO_DEVICE);
+
+free_pas:
 	kfree(mr->pas);
 
-	if (mr->status != IB_WC_SUCCESS) {
-		mlx5_ib_warn(dev, "reg umr failed\n");
-		err = -EFAULT;
-		goto error;
+free_mr:
+	if (err) {
+		free_cached_mr(dev, mr);
+		return ERR_PTR(err);
 	}
 
 	return mr;
-
-error:
-	free_cached_mr(dev, mr);
-	return ERR_PTR(err);
 }
 
 static struct mlx5_ib_mr *reg_create(struct ib_pd *pd, u64 virt_addr,
@@ -926,24 +938,26 @@ struct ib_mr *mlx5_ib_reg_user_mr(struct ib_pd *pd, u64 start, u64 length,
 static int unreg_umr(struct mlx5_ib_dev *dev, struct mlx5_ib_mr *mr)
 {
 	struct umr_common *umrc = &dev->umrc;
+	struct mlx5_ib_umr_context umr_context;
 	struct ib_send_wr wr, *bad;
 	int err;
 
 	memset(&wr, 0, sizeof(wr));
-	wr.wr_id = (u64)(unsigned long)mr;
+	wr.wr_id = (u64)(unsigned long)&umr_context;
 	prep_umr_unreg_wqe(dev, &wr, mr->mmr.key);
 
+	mlx5_ib_init_umr_context(&umr_context);
 	down(&umrc->sem);
-	init_completion(&mr->done);
 	err = ib_post_send(umrc->qp, &wr, &bad);
 	if (err) {
 		up(&umrc->sem);
 		mlx5_ib_dbg(dev, "err %d\n", err);
 		goto error;
+	} else {
+		wait_for_completion(&umr_context.done);
+		up(&umrc->sem);
 	}
-	wait_for_completion(&mr->done);
-	up(&umrc->sem);
-	if (mr->status != IB_WC_SUCCESS) {
+	if (umr_context.status != IB_WC_SUCCESS) {
 		mlx5_ib_warn(dev, "unreg umr failed\n");
 		err = -EFAULT;
 		goto error;

drivers/infiniband/hw/mlx5/qp.c

@@ -574,6 +574,10 @@ static int create_user_qp(struct mlx5_ib_dev *dev, struct ib_pd *pd,
 	uar_index = uuarn_to_uar_index(&context->uuari, uuarn);
 	mlx5_ib_dbg(dev, "uuarn 0x%x, uar_index 0x%x\n", uuarn, uar_index);
 
+	qp->rq.offset = 0;
+	qp->sq.wqe_shift = ilog2(MLX5_SEND_WQE_BB);
+	qp->sq.offset = qp->rq.wqe_cnt << qp->rq.wqe_shift;
+
 	err = set_user_buf_size(dev, qp, &ucmd);
 	if (err)
 		goto err_uuar;
@@ -2078,6 +2082,7 @@ static int mlx5_set_bsf(struct ib_mr *sig_mr,
 	struct ib_sig_domain *wire = &sig_attrs->wire;
 	int ret, selector;
 
+	memset(bsf, 0, sizeof(*bsf));
 	switch (sig_attrs->mem.sig_type) {
 	case IB_SIG_TYPE_T10_DIF:
 		if (sig_attrs->wire.sig_type != IB_SIG_TYPE_T10_DIF)
@@ -2090,9 +2095,11 @@ static int mlx5_set_bsf(struct ib_mr *sig_mr,
 			/* Same block structure */
 			basic->bsf_size_sbs = 1 << 4;
 			if (mem->sig.dif.bg_type == wire->sig.dif.bg_type)
-				basic->wire.copy_byte_mask = 0xff;
-			else
-				basic->wire.copy_byte_mask = 0x3f;
+				basic->wire.copy_byte_mask |= 0xc0;
+			if (mem->sig.dif.app_tag == wire->sig.dif.app_tag)
+				basic->wire.copy_byte_mask |= 0x30;
+			if (mem->sig.dif.ref_tag == wire->sig.dif.ref_tag)
+				basic->wire.copy_byte_mask |= 0x0f;
 		} else
 			basic->wire.bs_selector = bs_selector(wire->sig.dif.pi_interval);
 
@@ -2131,9 +2138,13 @@ static int set_sig_data_segment(struct ib_send_wr *wr, struct mlx5_ib_qp *qp,
 	int ret;
 	int wqe_size;
 
-	if (!wr->wr.sig_handover.prot) {
+	if (!wr->wr.sig_handover.prot ||
+	    (data_key == wr->wr.sig_handover.prot->lkey &&
+	     data_va == wr->wr.sig_handover.prot->addr &&
+	     data_len == wr->wr.sig_handover.prot->length)) {
 		/**
 		 * Source domain doesn't contain signature information
+		 * or data and protection are interleaved in memory.
 		 * So need construct:
 		 *                  ------------------
 		 *                 |     data_klm     |
@@ -2187,23 +2198,13 @@ static int set_sig_data_segment(struct ib_send_wr *wr, struct mlx5_ib_qp *qp,
 		data_sentry->bcount = cpu_to_be16(block_size);
 		data_sentry->key = cpu_to_be32(data_key);
 		data_sentry->va = cpu_to_be64(data_va);
+		data_sentry->stride = cpu_to_be16(block_size);
+
 		prot_sentry->bcount = cpu_to_be16(prot_size);
 		prot_sentry->key = cpu_to_be32(prot_key);
+		prot_sentry->va = cpu_to_be64(prot_va);
+		prot_sentry->stride = cpu_to_be16(prot_size);
 
-		if (prot_key == data_key && prot_va == data_va) {
-			/**
-			 * The data and protection are interleaved
-			 * in a single memory region
-			 **/
-			prot_sentry->va = cpu_to_be64(data_va + block_size);
-			prot_sentry->stride = cpu_to_be16(block_size + prot_size);
-			data_sentry->stride = prot_sentry->stride;
-		} else {
-			/* The data and protection are two different buffers */
-			prot_sentry->va = cpu_to_be64(prot_va);
-			data_sentry->stride = cpu_to_be16(block_size);
-			prot_sentry->stride = cpu_to_be16(prot_size);
-		}
 		wqe_size = ALIGN(sizeof(*sblock_ctrl) + sizeof(*data_sentry) +
 				 sizeof(*prot_sentry), 64);
 	}
@@ -2275,7 +2276,10 @@ static int set_sig_umr_wr(struct ib_send_wr *wr, struct mlx5_ib_qp *qp,
 
 	/* length of the protected region, data + protection */
 	region_len = wr->sg_list->length;
-	if (wr->wr.sig_handover.prot)
+	if (wr->wr.sig_handover.prot &&
+	    (wr->wr.sig_handover.prot->lkey != wr->sg_list->lkey  ||
+	     wr->wr.sig_handover.prot->addr != wr->sg_list->addr  ||
+	     wr->wr.sig_handover.prot->length != wr->sg_list->length))
 		region_len += wr->wr.sig_handover.prot->length;
 
 	/**

drivers/infiniband/hw/mlx5/srq.c

@@ -35,6 +35,7 @@
 #include <linux/mlx5/srq.h>
 #include <linux/slab.h>
 #include <rdma/ib_umem.h>
+#include <rdma/ib_user_verbs.h>
 
 #include "mlx5_ib.h"
 #include "user.h"
@@ -78,16 +79,27 @@ static int create_srq_user(struct ib_pd *pd, struct mlx5_ib_srq *srq,
 {
 	struct mlx5_ib_dev *dev = to_mdev(pd->device);
 	struct mlx5_ib_create_srq ucmd;
+	size_t ucmdlen;
 	int err;
 	int npages;
 	int page_shift;
 	int ncont;
 	u32 offset;
 
-	if (ib_copy_from_udata(&ucmd, udata, sizeof(ucmd))) {
+	ucmdlen =
+		(udata->inlen - sizeof(struct ib_uverbs_cmd_hdr) <
+		 sizeof(ucmd)) ? (sizeof(ucmd) -
+				  sizeof(ucmd.reserved)) : sizeof(ucmd);
+
+	if (ib_copy_from_udata(&ucmd, udata, ucmdlen)) {
 		mlx5_ib_dbg(dev, "failed copy udata\n");
 		return -EFAULT;
 	}
+
+	if (ucmdlen == sizeof(ucmd) &&
+	    ucmd.reserved != 0)
+		return -EINVAL;
+
 	srq->wq_sig = !!(ucmd.flags & MLX5_SRQ_FLAG_SIGNATURE);
 
 	srq->umem = ib_umem_get(pd->uobject->context, ucmd.buf_addr, buf_size,

drivers/infiniband/hw/mlx5/user.h

@@ -91,6 +91,7 @@ struct mlx5_ib_create_cq {
 	__u64	buf_addr;
 	__u64	db_addr;
 	__u32	cqe_size;
+	__u32	reserved; /* explicit padding (optional on i386) */
 };
 
 struct mlx5_ib_create_cq_resp {
@@ -109,6 +110,7 @@ struct mlx5_ib_create_srq {
 	__u64	buf_addr;
 	__u64	db_addr;
 	__u32	flags;
+	__u32	reserved; /* explicit padding (optional on i386) */
 };
 
 struct mlx5_ib_create_srq_resp {

drivers/infiniband/hw/nes/nes.c

@@ -68,7 +68,6 @@ MODULE_VERSION(DRV_VERSION);
 int max_mtu = 9000;
 int interrupt_mod_interval = 0;
 
-
 /* Interoperability */
 int mpa_version = 1;
 module_param(mpa_version, int, 0644);
@@ -112,6 +111,16 @@ static struct pci_device_id nes_pci_table[] = {
 
 MODULE_DEVICE_TABLE(pci, nes_pci_table);
 
+/* registered nes netlink callbacks */
+static struct ibnl_client_cbs nes_nl_cb_table[] = {
+	[RDMA_NL_IWPM_REG_PID] = {.dump = iwpm_register_pid_cb},
+	[RDMA_NL_IWPM_ADD_MAPPING] = {.dump = iwpm_add_mapping_cb},
+	[RDMA_NL_IWPM_QUERY_MAPPING] = {.dump = iwpm_add_and_query_mapping_cb},
+	[RDMA_NL_IWPM_HANDLE_ERR] = {.dump = iwpm_mapping_error_cb},
+	[RDMA_NL_IWPM_MAPINFO] = {.dump = iwpm_mapping_info_cb},
+	[RDMA_NL_IWPM_MAPINFO_NUM] = {.dump = iwpm_ack_mapping_info_cb}
+};
+
 static int nes_inetaddr_event(struct notifier_block *, unsigned long, void *);
 static int nes_net_event(struct notifier_block *, unsigned long, void *);
 static int nes_notifiers_registered;
@@ -672,6 +681,17 @@ static int nes_probe(struct pci_dev *pcidev, const struct pci_device_id *ent)
 	}
 	nes_notifiers_registered++;
 
+	if (ibnl_add_client(RDMA_NL_NES, RDMA_NL_IWPM_NUM_OPS, nes_nl_cb_table))
+		printk(KERN_ERR PFX "%s[%u]: Failed to add netlink callback\n",
+			__func__, __LINE__);
+
+	ret = iwpm_init(RDMA_NL_NES);
+	if (ret) {
+		printk(KERN_ERR PFX "%s: port mapper initialization failed\n",
+				pci_name(pcidev));
+		goto bail7;
+	}
+
 	INIT_DELAYED_WORK(&nesdev->work, nes_recheck_link_status);
 
 	/* Initialize network devices */
@@ -710,6 +730,7 @@ static int nes_probe(struct pci_dev *pcidev, const struct pci_device_id *ent)
 
 	nes_debug(NES_DBG_INIT, "netdev_count=%d, nesadapter->netdev_count=%d\n",
 			nesdev->netdev_count, nesdev->nesadapter->netdev_count);
+	ibnl_remove_client(RDMA_NL_NES);
 
 	nes_notifiers_registered--;
 	if (nes_notifiers_registered == 0) {
@@ -773,6 +794,8 @@ static void nes_remove(struct pci_dev *pcidev)
 				nesdev->nesadapter->netdev_count--;
 			}
 		}
+	ibnl_remove_client(RDMA_NL_NES);
+	iwpm_exit(RDMA_NL_NES);
 
 	nes_notifiers_registered--;
 	if (nes_notifiers_registered == 0) {

drivers/infiniband/hw/nes/nes.h

@@ -51,6 +51,8 @@
 #include <rdma/ib_pack.h>
 #include <rdma/rdma_cm.h>
 #include <rdma/iw_cm.h>
+#include <rdma/rdma_netlink.h>
+#include <rdma/iw_portmap.h>
 
 #define NES_SEND_FIRST_WRITE
 
@@ -130,6 +132,7 @@
 #define NES_DBG_IW_TX       0x00040000
 #define NES_DBG_SHUTDOWN    0x00080000
 #define NES_DBG_PAU         0x00100000
+#define NES_DBG_NLMSG       0x00200000
 #define NES_DBG_RSVD1       0x10000000
 #define NES_DBG_RSVD2       0x20000000
 #define NES_DBG_RSVD3       0x40000000

drivers/infiniband/hw/nes/nes_cm.h

 /*
- * Copyright (c) 2006 - 2011 Intel Corporation.  All rights reserved.
+ * Copyright (c) 2006 - 2014 Intel Corporation.  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
@@ -293,8 +293,8 @@ struct nes_cm_listener {
 	struct list_head           list;
 	struct nes_cm_core         *cm_core;
 	u8                         loc_mac[ETH_ALEN];
-	nes_addr_t                 loc_addr;
-	u16                        loc_port;
+	nes_addr_t                 loc_addr, mapped_loc_addr;
+	u16                        loc_port, mapped_loc_port;
 	struct iw_cm_id            *cm_id;
 	enum nes_cm_conn_type      conn_type;
 	atomic_t                   ref_count;
@@ -308,7 +308,9 @@ struct nes_cm_listener {
 /* per connection node and node state information */
 struct nes_cm_node {
 	nes_addr_t                loc_addr, rem_addr;
+	nes_addr_t                mapped_loc_addr, mapped_rem_addr;
 	u16                       loc_port, rem_port;
+	u16                       mapped_loc_port, mapped_rem_port;
 
 	u8                        loc_mac[ETH_ALEN];
 	u8                        rem_mac[ETH_ALEN];
@@ -364,6 +366,10 @@ struct nes_cm_info {
 	u16 rem_port;
 	nes_addr_t loc_addr;
 	nes_addr_t rem_addr;
+	u16 mapped_loc_port;
+	u16 mapped_rem_port;
+	nes_addr_t mapped_loc_addr;
+	nes_addr_t mapped_rem_addr;
 
 	enum nes_cm_conn_type  conn_type;
 	int backlog;

drivers/infiniband/hw/ocrdma/ocrdma_stats.c

@@ -510,16 +510,9 @@ static ssize_t ocrdma_dbgfs_ops_read(struct file *filp, char __user *buffer,
 	return status;
 }
 
-static int ocrdma_debugfs_open(struct inode *inode, struct file *file)
-{
-	if (inode->i_private)
-		file->private_data = inode->i_private;
-	return 0;
-}
-
 static const struct file_operations ocrdma_dbg_ops = {
 	.owner = THIS_MODULE,
-	.open = ocrdma_debugfs_open,
+	.open = simple_open,
 	.read = ocrdma_dbgfs_ops_read,
 };
 

drivers/infiniband/hw/qib/qib_init.c

@@ -1272,7 +1272,7 @@ static int qib_notify_dca(struct notifier_block *nb, unsigned long event,
  * Do all the generic driver unit- and chip-independent memory
  * allocation and initialization.
  */
-static int __init qlogic_ib_init(void)
+static int __init qib_ib_init(void)
 {
 	int ret;
 
@@ -1316,12 +1316,12 @@ static int __init qlogic_ib_init(void)
 	return ret;
 }
 
-module_init(qlogic_ib_init);
+module_init(qib_ib_init);
 
 /*
  * Do the non-unit driver cleanup, memory free, etc. at unload.
  */
-static void __exit qlogic_ib_cleanup(void)
+static void __exit qib_ib_cleanup(void)
 {
 	int ret;
 
@@ -1346,7 +1346,7 @@ static void __exit qlogic_ib_cleanup(void)
 	qib_dev_cleanup();
 }
 
-module_exit(qlogic_ib_cleanup);
+module_exit(qib_ib_cleanup);
 
 /* this can only be called after a successful initialization */
 static void cleanup_device_data(struct qib_devdata *dd)

drivers/infiniband/hw/qib/qib_mad.c

@@ -1028,7 +1028,7 @@ static int set_pkeys(struct qib_devdata *dd, u8 port, u16 *pkeys)
 
 		event.event = IB_EVENT_PKEY_CHANGE;
 		event.device = &dd->verbs_dev.ibdev;
-		event.element.port_num = 1;
+		event.element.port_num = port;
 		ib_dispatch_event(&event);
 	}
 	return 0;

drivers/infiniband/hw/qib/qib_qp.c

@@ -985,7 +985,8 @@ struct ib_qp *qib_create_qp(struct ib_pd *ibpd,
 	struct ib_qp *ret;
 
 	if (init_attr->cap.max_send_sge > ib_qib_max_sges ||
-	    init_attr->cap.max_send_wr > ib_qib_max_qp_wrs) {
+	    init_attr->cap.max_send_wr > ib_qib_max_qp_wrs ||
+	    init_attr->create_flags) {
 		ret = ERR_PTR(-EINVAL);
 		goto bail;
 	}

drivers/infiniband/hw/usnic/usnic_ib_verbs.c

@@ -466,6 +466,9 @@ struct ib_qp *usnic_ib_create_qp(struct ib_pd *pd,
 	ucontext = to_uucontext(pd->uobject->context);
 	us_ibdev = to_usdev(pd->device);
 
+	if (init_attr->create_flags)
+		return ERR_PTR(-EINVAL);
+
 	err = ib_copy_from_udata(&cmd, udata, sizeof(cmd));
 	if (err) {
 		usnic_err("%s: cannot copy udata for create_qp\n",

drivers/infiniband/hw/usnic/usnic_uiom_interval_tree.c

+/*
+ * Copyright (c) 2014, Cisco Systems, Inc. All rights reserved.
+ *
+ * This program is free software; you may redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; version 2 of the License.
+ *
+ * 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/init.h>
 #include <linux/list.h>
 #include <linux/slab.h>

drivers/infiniband/ulp/Makefile

+obj-$(CONFIG_INFINIBAND_IPOIB)		+= ipoib/
+obj-$(CONFIG_INFINIBAND_SRP)		+= srp/
+obj-$(CONFIG_INFINIBAND_SRPT)		+= srpt/
+obj-$(CONFIG_INFINIBAND_ISER)		+= iser/
+obj-$(CONFIG_INFINIBAND_ISERT)		+= isert/

drivers/infiniband/ulp/ipoib/ipoib_cm.c

@@ -1030,10 +1030,20 @@ static struct ib_qp *ipoib_cm_create_tx_qp(struct net_device *dev, struct ipoib_
 		.cap.max_send_sge	= 1,
 		.sq_sig_type		= IB_SIGNAL_ALL_WR,
 		.qp_type		= IB_QPT_RC,
-		.qp_context		= tx
+		.qp_context		= tx,
+		.create_flags		= IB_QP_CREATE_USE_GFP_NOIO
 	};
 
-	return ib_create_qp(priv->pd, &attr);
+	struct ib_qp *tx_qp;
+
+	tx_qp = ib_create_qp(priv->pd, &attr);
+	if (PTR_ERR(tx_qp) == -EINVAL) {
+		ipoib_warn(priv, "can't use GFP_NOIO for QPs on device %s, using GFP_KERNEL\n",
+			   priv->ca->name);
+		attr.create_flags &= ~IB_QP_CREATE_USE_GFP_NOIO;
+		tx_qp = ib_create_qp(priv->pd, &attr);
+	}
+	return tx_qp;
 }
 
 static int ipoib_cm_send_req(struct net_device *dev,
@@ -1104,12 +1114,14 @@ static int ipoib_cm_tx_init(struct ipoib_cm_tx *p, u32 qpn,
 	struct ipoib_dev_priv *priv = netdev_priv(p->dev);
 	int ret;
 
-	p->tx_ring = vzalloc(ipoib_sendq_size * sizeof *p->tx_ring);
+	p->tx_ring = __vmalloc(ipoib_sendq_size * sizeof *p->tx_ring,
+			       GFP_NOIO, PAGE_KERNEL);
 	if (!p->tx_ring) {
 		ipoib_warn(priv, "failed to allocate tx ring\n");
 		ret = -ENOMEM;
 		goto err_tx;
 	}
+	memset(p->tx_ring, 0, ipoib_sendq_size * sizeof *p->tx_ring);
 
 	p->qp = ipoib_cm_create_tx_qp(p->dev, p);
 	if (IS_ERR(p->qp)) {

drivers/infiniband/ulp/iser/iscsi_iser.c

@@ -99,6 +99,7 @@ MODULE_PARM_DESC(pi_enable, "Enable T10-PI offload support (default:disabled)");
 module_param_named(pi_guard, iser_pi_guard, int, 0644);
 MODULE_PARM_DESC(pi_guard, "T10-PI guard_type, 0:CRC|1:IP_CSUM (default:CRC)");
 
+static struct workqueue_struct *release_wq;
 struct iser_global ig;
 
 void
@@ -337,24 +338,6 @@ iscsi_iser_conn_create(struct iscsi_cls_session *cls_session, uint32_t conn_idx)
 	return cls_conn;
 }
 
-static void
-iscsi_iser_conn_destroy(struct iscsi_cls_conn *cls_conn)
-{
-	struct iscsi_conn *conn = cls_conn->dd_data;
-	struct iser_conn *ib_conn = conn->dd_data;
-
-	iscsi_conn_teardown(cls_conn);
-	/*
-	 * Userspace will normally call the stop callback and
-	 * already have freed the ib_conn, but if it goofed up then
-	 * we free it here.
-	 */
-	if (ib_conn) {
-		ib_conn->iscsi_conn = NULL;
-		iser_conn_put(ib_conn, 1); /* deref iscsi/ib conn unbinding */
-	}
-}
-
 static int
 iscsi_iser_conn_bind(struct iscsi_cls_session *cls_session,
 		     struct iscsi_cls_conn *cls_conn, uint64_t transport_eph,
@@ -392,29 +375,39 @@ iscsi_iser_conn_bind(struct iscsi_cls_session *cls_session,
 	conn->dd_data = ib_conn;
 	ib_conn->iscsi_conn = conn;
 
-	iser_conn_get(ib_conn); /* ref iscsi/ib conn binding */
 	return 0;
 }
 
+static int
+iscsi_iser_conn_start(struct iscsi_cls_conn *cls_conn)
+{
+	struct iscsi_conn *iscsi_conn;
+	struct iser_conn *ib_conn;
+
+	iscsi_conn = cls_conn->dd_data;
+	ib_conn = iscsi_conn->dd_data;
+	reinit_completion(&ib_conn->stop_completion);
+
+	return iscsi_conn_start(cls_conn);
+}
+
 static void
 iscsi_iser_conn_stop(struct iscsi_cls_conn *cls_conn, int flag)
 {
 	struct iscsi_conn *conn = cls_conn->dd_data;
 	struct iser_conn *ib_conn = conn->dd_data;
 
+	iser_dbg("stopping iscsi_conn: %p, ib_conn: %p\n", conn, ib_conn);
+	iscsi_conn_stop(cls_conn, flag);
+
 	/*
 	 * Userspace may have goofed up and not bound the connection or
 	 * might have only partially setup the connection.
 	 */
 	if (ib_conn) {
-		iscsi_conn_stop(cls_conn, flag);
-		/*
-		 * There is no unbind event so the stop callback
-		 * must release the ref from the bind.
-		 */
-		iser_conn_put(ib_conn, 1); /* deref iscsi/ib conn unbinding */
+		conn->dd_data = NULL;
+		complete(&ib_conn->stop_completion);
 	}
-	conn->dd_data = NULL;
 }
 
 static void iscsi_iser_session_destroy(struct iscsi_cls_session *cls_session)
@@ -515,28 +508,28 @@ iscsi_iser_set_param(struct iscsi_cls_conn *cls_conn,
 	case ISCSI_PARAM_HDRDGST_EN:
 		sscanf(buf, "%d", &value);
 		if (value) {
-			iser_err("DataDigest wasn't negotiated to None");
+			iser_err("DataDigest wasn't negotiated to None\n");
 			return -EPROTO;
 		}
 		break;
 	case ISCSI_PARAM_DATADGST_EN:
 		sscanf(buf, "%d", &value);
 		if (value) {
-			iser_err("DataDigest wasn't negotiated to None");
+			iser_err("DataDigest wasn't negotiated to None\n");
 			return -EPROTO;
 		}
 		break;
 	case ISCSI_PARAM_IFMARKER_EN:
 		sscanf(buf, "%d", &value);
 		if (value) {
-			iser_err("IFMarker wasn't negotiated to No");
+			iser_err("IFMarker wasn't negotiated to No\n");
 			return -EPROTO;
 		}
 		break;
 	case ISCSI_PARAM_OFMARKER_EN:
 		sscanf(buf, "%d", &value);
 		if (value) {
-			iser_err("OFMarker wasn't negotiated to No");
+			iser_err("OFMarker wasn't negotiated to No\n");
 			return -EPROTO;
 		}
 		break;
@@ -652,19 +645,20 @@ iscsi_iser_ep_disconnect(struct iscsi_endpoint *ep)
 	struct iser_conn *ib_conn;
 
 	ib_conn = ep->dd_data;
-	if (ib_conn->iscsi_conn)
-		/*
-		 * Must suspend xmit path if the ep is bound to the
-		 * iscsi_conn, so we know we are not accessing the ib_conn
-		 * when we free it.
-		 *
-		 * This may not be bound if the ep poll failed.
-		 */
-		iscsi_suspend_tx(ib_conn->iscsi_conn);
-
-
-	iser_info("ib conn %p state %d\n", ib_conn, ib_conn->state);
+	iser_info("ep %p ib conn %p state %d\n", ep, ib_conn, ib_conn->state);
 	iser_conn_terminate(ib_conn);
+
+	/*
+	 * if iser_conn and iscsi_conn are bound, we must wait iscsi_conn_stop
+	 * call and ISER_CONN_DOWN state before freeing the iser resources.
+	 * otherwise we are safe to free resources immediately.
+	 */
+	if (ib_conn->iscsi_conn) {
+		INIT_WORK(&ib_conn->release_work, iser_release_work);
+		queue_work(release_wq, &ib_conn->release_work);
+	} else {
+		iser_conn_release(ib_conn);
+	}
 }
 
 static umode_t iser_attr_is_visible(int param_type, int param)
@@ -748,13 +742,13 @@ static struct iscsi_transport iscsi_iser_transport = {
 	/* connection management */
 	.create_conn            = iscsi_iser_conn_create,
 	.bind_conn              = iscsi_iser_conn_bind,
-	.destroy_conn           = iscsi_iser_conn_destroy,
+	.destroy_conn           = iscsi_conn_teardown,
 	.attr_is_visible	= iser_attr_is_visible,
 	.set_param              = iscsi_iser_set_param,
 	.get_conn_param		= iscsi_conn_get_param,
 	.get_ep_param		= iscsi_iser_get_ep_param,
 	.get_session_param	= iscsi_session_get_param,
-	.start_conn             = iscsi_conn_start,
+	.start_conn             = iscsi_iser_conn_start,
 	.stop_conn              = iscsi_iser_conn_stop,
 	/* iscsi host params */
 	.get_host_param		= iscsi_host_get_param,
@@ -801,6 +795,12 @@ static int __init iser_init(void)
 	mutex_init(&ig.connlist_mutex);
 	INIT_LIST_HEAD(&ig.connlist);
 
+	release_wq = alloc_workqueue("release workqueue", 0, 0);
+	if (!release_wq) {
+		iser_err("failed to allocate release workqueue\n");
+		return -ENOMEM;
+	}
+
 	iscsi_iser_scsi_transport = iscsi_register_transport(
 							&iscsi_iser_transport);
 	if (!iscsi_iser_scsi_transport) {
@@ -819,7 +819,24 @@ static int __init iser_init(void)
 
 static void __exit iser_exit(void)
 {
+	struct iser_conn *ib_conn, *n;
+	int connlist_empty;
+
 	iser_dbg("Removing iSER datamover...\n");
+	destroy_workqueue(release_wq);
+
+	mutex_lock(&ig.connlist_mutex);
+	connlist_empty = list_empty(&ig.connlist);
+	mutex_unlock(&ig.connlist_mutex);
+
+	if (!connlist_empty) {
+		iser_err("Error cleanup stage completed but we still have iser "
+			 "connections, destroying them anyway.\n");
+		list_for_each_entry_safe(ib_conn, n, &ig.connlist, conn_list) {
+			iser_conn_release(ib_conn);
+		}
+	}
+
 	iscsi_unregister_transport(&iscsi_iser_transport);
 	kmem_cache_destroy(ig.desc_cache);
 }

drivers/infiniband/ulp/iser/iscsi_iser.h

@@ -69,7 +69,7 @@
 
 #define DRV_NAME	"iser"
 #define PFX		DRV_NAME ": "
-#define DRV_VER		"1.3"
+#define DRV_VER		"1.4"
 
 #define iser_dbg(fmt, arg...)				\
 	do {						\
@@ -333,6 +333,8 @@ struct iser_conn {
 	int                          post_recv_buf_count; /* posted rx count  */
 	atomic_t                     post_send_buf_count; /* posted tx count   */
 	char 			     name[ISER_OBJECT_NAME_SIZE];
+	struct work_struct	     release_work;
+	struct completion	     stop_completion;
 	struct list_head	     conn_list;       /* entry in ig conn list */
 
 	char  			     *login_buf;
@@ -417,12 +419,12 @@ void iscsi_iser_recv(struct iscsi_conn *conn,
 
 void iser_conn_init(struct iser_conn *ib_conn);
 
-void iser_conn_get(struct iser_conn *ib_conn);
-
-int iser_conn_put(struct iser_conn *ib_conn, int destroy_cma_id_allowed);
+void iser_conn_release(struct iser_conn *ib_conn);
 
 void iser_conn_terminate(struct iser_conn *ib_conn);
 
+void iser_release_work(struct work_struct *work);
+
 void iser_rcv_completion(struct iser_rx_desc *desc,
 			 unsigned long    dto_xfer_len,
 			struct iser_conn *ib_conn);

drivers/infiniband/ulp/iser/iser_verbs.c

@@ -581,14 +581,30 @@ static int iser_conn_state_comp_exch(struct iser_conn *ib_conn,
 	return ret;
 }
 
+void iser_release_work(struct work_struct *work)
+{
+	struct iser_conn *ib_conn;
+
+	ib_conn = container_of(work, struct iser_conn, release_work);
+
+	/* wait for .conn_stop callback */
+	wait_for_completion(&ib_conn->stop_completion);
+
+	/* wait for the qp`s post send and post receive buffers to empty */
+	wait_event_interruptible(ib_conn->wait,
+				 ib_conn->state == ISER_CONN_DOWN);
+
+	iser_conn_release(ib_conn);
+}
+
 /**
  * Frees all conn objects and deallocs conn descriptor
  */
-static void iser_conn_release(struct iser_conn *ib_conn, int can_destroy_id)
+void iser_conn_release(struct iser_conn *ib_conn)
 {
 	struct iser_device  *device = ib_conn->device;
 
-	BUG_ON(ib_conn->state != ISER_CONN_DOWN);
+	BUG_ON(ib_conn->state == ISER_CONN_UP);
 
 	mutex_lock(&ig.connlist_mutex);
 	list_del(&ib_conn->conn_list);
@@ -600,27 +616,13 @@ static void iser_conn_release(struct iser_conn *ib_conn, int can_destroy_id)
 	if (device != NULL)
 		iser_device_try_release(device);
 	/* if cma handler context, the caller actually destroy the id */
-	if (ib_conn->cma_id != NULL && can_destroy_id) {
+	if (ib_conn->cma_id != NULL) {
 		rdma_destroy_id(ib_conn->cma_id);
 		ib_conn->cma_id = NULL;
 	}
 	iscsi_destroy_endpoint(ib_conn->ep);
 }
 
-void iser_conn_get(struct iser_conn *ib_conn)
-{
-	atomic_inc(&ib_conn->refcount);
-}
-
-int iser_conn_put(struct iser_conn *ib_conn, int can_destroy_id)
-{
-	if (atomic_dec_and_test(&ib_conn->refcount)) {
-		iser_conn_release(ib_conn, can_destroy_id);
-		return 1;
-	}
-	return 0;
-}
-
 /**
  * triggers start of the disconnect procedures and wait for them to be done
  */
@@ -638,24 +640,19 @@ void iser_conn_terminate(struct iser_conn *ib_conn)
 	if (err)
 		iser_err("Failed to disconnect, conn: 0x%p err %d\n",
 			 ib_conn,err);
-
-	wait_event_interruptible(ib_conn->wait,
-				 ib_conn->state == ISER_CONN_DOWN);
-
-	iser_conn_put(ib_conn, 1); /* deref ib conn deallocate */
 }
 
-static int iser_connect_error(struct rdma_cm_id *cma_id)
+static void iser_connect_error(struct rdma_cm_id *cma_id)
 {
 	struct iser_conn *ib_conn;
+
 	ib_conn = (struct iser_conn *)cma_id->context;
 
 	ib_conn->state = ISER_CONN_DOWN;
 	wake_up_interruptible(&ib_conn->wait);
-	return iser_conn_put(ib_conn, 0); /* deref ib conn's cma id */
 }
 
-static int iser_addr_handler(struct rdma_cm_id *cma_id)
+static void iser_addr_handler(struct rdma_cm_id *cma_id)
 {
 	struct iser_device *device;
 	struct iser_conn   *ib_conn;
@@ -664,7 +661,8 @@ static int iser_addr_handler(struct rdma_cm_id *cma_id)
 	device = iser_device_find_by_ib_device(cma_id);
 	if (!device) {
 		iser_err("device lookup/creation failed\n");
-		return iser_connect_error(cma_id);
+		iser_connect_error(cma_id);
+		return;
 	}
 
 	ib_conn = (struct iser_conn *)cma_id->context;
@@ -686,13 +684,12 @@ static int iser_addr_handler(struct rdma_cm_id *cma_id)
 	ret = rdma_resolve_route(cma_id, 1000);
 	if (ret) {
 		iser_err("resolve route failed: %d\n", ret);
-		return iser_connect_error(cma_id);
+		iser_connect_error(cma_id);
+		return;
 	}
-
-	return 0;
 }
 
-static int iser_route_handler(struct rdma_cm_id *cma_id)
+static void iser_route_handler(struct rdma_cm_id *cma_id)
 {
 	struct rdma_conn_param conn_param;
 	int    ret;
@@ -720,9 +717,9 @@ static int iser_route_handler(struct rdma_cm_id *cma_id)
 		goto failure;
 	}
 
-	return 0;
+	return;
 failure:
-	return iser_connect_error(cma_id);
+	iser_connect_error(cma_id);
 }
 
 static void iser_connected_handler(struct rdma_cm_id *cma_id)
@@ -735,14 +732,13 @@ static void iser_connected_handler(struct rdma_cm_id *cma_id)
 	iser_info("remote qpn:%x my qpn:%x\n", attr.dest_qp_num, cma_id->qp->qp_num);
 
 	ib_conn = (struct iser_conn *)cma_id->context;
-	ib_conn->state = ISER_CONN_UP;
-	wake_up_interruptible(&ib_conn->wait);
+	if (iser_conn_state_comp_exch(ib_conn, ISER_CONN_PENDING, ISER_CONN_UP))
+		wake_up_interruptible(&ib_conn->wait);
 }
 
-static int iser_disconnected_handler(struct rdma_cm_id *cma_id)
+static void iser_disconnected_handler(struct rdma_cm_id *cma_id)
 {
 	struct iser_conn *ib_conn;
-	int ret;
 
 	ib_conn = (struct iser_conn *)cma_id->context;
 
@@ -762,24 +758,19 @@ static int iser_disconnected_handler(struct rdma_cm_id *cma_id)
 		ib_conn->state = ISER_CONN_DOWN;
 		wake_up_interruptible(&ib_conn->wait);
 	}
-
-	ret = iser_conn_put(ib_conn, 0); /* deref ib conn's cma id */
-	return ret;
 }
 
 static int iser_cma_handler(struct rdma_cm_id *cma_id, struct rdma_cm_event *event)
 {
-	int ret = 0;
-
 	iser_info("event %d status %d conn %p id %p\n",
 		  event->event, event->status, cma_id->context, cma_id);
 
 	switch (event->event) {
 	case RDMA_CM_EVENT_ADDR_RESOLVED:
-		ret = iser_addr_handler(cma_id);
+		iser_addr_handler(cma_id);
 		break;
 	case RDMA_CM_EVENT_ROUTE_RESOLVED:
-		ret = iser_route_handler(cma_id);
+		iser_route_handler(cma_id);
 		break;
 	case RDMA_CM_EVENT_ESTABLISHED:
 		iser_connected_handler(cma_id);
@@ -789,18 +780,18 @@ static int iser_cma_handler(struct rdma_cm_id *cma_id, struct rdma_cm_event *eve
 	case RDMA_CM_EVENT_CONNECT_ERROR:
 	case RDMA_CM_EVENT_UNREACHABLE:
 	case RDMA_CM_EVENT_REJECTED:
-		ret = iser_connect_error(cma_id);
+		iser_connect_error(cma_id);
 		break;
 	case RDMA_CM_EVENT_DISCONNECTED:
 	case RDMA_CM_EVENT_DEVICE_REMOVAL:
 	case RDMA_CM_EVENT_ADDR_CHANGE:
-		ret = iser_disconnected_handler(cma_id);
+		iser_disconnected_handler(cma_id);
 		break;
 	default:
 		iser_err("Unexpected RDMA CM event (%d)\n", event->event);
 		break;
 	}
-	return ret;
+	return 0;
 }
 
 void iser_conn_init(struct iser_conn *ib_conn)
@@ -809,7 +800,7 @@ void iser_conn_init(struct iser_conn *ib_conn)
 	init_waitqueue_head(&ib_conn->wait);
 	ib_conn->post_recv_buf_count = 0;
 	atomic_set(&ib_conn->post_send_buf_count, 0);
-	atomic_set(&ib_conn->refcount, 1); /* ref ib conn allocation */
+	init_completion(&ib_conn->stop_completion);
 	INIT_LIST_HEAD(&ib_conn->conn_list);
 	spin_lock_init(&ib_conn->lock);
 }
@@ -837,7 +828,6 @@ int iser_connect(struct iser_conn   *ib_conn,
 
 	ib_conn->state = ISER_CONN_PENDING;
 
-	iser_conn_get(ib_conn); /* ref ib conn's cma id */
 	ib_conn->cma_id = rdma_create_id(iser_cma_handler,
 					     (void *)ib_conn,
 					     RDMA_PS_TCP, IB_QPT_RC);
@@ -874,9 +864,8 @@ int iser_connect(struct iser_conn   *ib_conn,
 	ib_conn->cma_id = NULL;
 addr_failure:
 	ib_conn->state = ISER_CONN_DOWN;
-	iser_conn_put(ib_conn, 1); /* deref ib conn's cma id */
 connect_failure:
-	iser_conn_put(ib_conn, 1); /* deref ib conn deallocate */
+	iser_conn_release(ib_conn);
 	return err;
 }
 

drivers/infiniband/ulp/srp/ib_srp.h

@@ -66,13 +66,10 @@ enum {
 	SRP_TAG_NO_REQ		= ~0U,
 	SRP_TAG_TSK_MGMT	= 1U << 31,
 
-	SRP_FMR_SIZE		= 512,
-	SRP_FMR_MIN_SIZE	= 128,
-	SRP_FMR_POOL_SIZE	= 1024,
-	SRP_FMR_DIRTY_SIZE	= SRP_FMR_POOL_SIZE / 4,
+	SRP_MAX_PAGES_PER_MR	= 512,
 
-	SRP_MAP_ALLOW_FMR	= 0,
-	SRP_MAP_NO_FMR		= 1,
+	LOCAL_INV_WR_ID_MASK	= 1,
+	FAST_REG_WR_ID_MASK	= 2,
 };
 
 enum srp_target_state {
@@ -86,15 +83,24 @@ enum srp_iu_type {
 	SRP_IU_RSP,
 };
 
+/*
+ * @mr_page_mask: HCA memory registration page mask.
+ * @mr_page_size: HCA memory registration page size.
+ * @mr_max_size: Maximum size in bytes of a single FMR / FR registration
+ *   request.
+ */
 struct srp_device {
 	struct list_head	dev_list;
 	struct ib_device       *dev;
 	struct ib_pd	       *pd;
 	struct ib_mr	       *mr;
-	struct ib_fmr_pool     *fmr_pool;
-	u64			fmr_page_mask;
-	int			fmr_page_size;
-	int			fmr_max_size;
+	u64			mr_page_mask;
+	int			mr_page_size;
+	int			mr_max_size;
+	int			max_pages_per_mr;
+	bool			has_fmr;
+	bool			has_fr;
+	bool			use_fast_reg;
 };
 
 struct srp_host {
@@ -112,11 +118,14 @@ struct srp_request {
 	struct list_head	list;
 	struct scsi_cmnd       *scmnd;
 	struct srp_iu	       *cmd;
-	struct ib_pool_fmr    **fmr_list;
+	union {
+		struct ib_pool_fmr **fmr_list;
+		struct srp_fr_desc **fr_list;
+	};
 	u64		       *map_page;
 	struct srp_direct_buf  *indirect_desc;
 	dma_addr_t		indirect_dma_addr;
-	short			nfmr;
+	short			nmdesc;
 	short			index;
 };
 
@@ -131,6 +140,10 @@ struct srp_target_port {
 	struct ib_cq	       *send_cq ____cacheline_aligned_in_smp;
 	struct ib_cq	       *recv_cq;
 	struct ib_qp	       *qp;
+	union {
+		struct ib_fmr_pool     *fmr_pool;
+		struct srp_fr_pool     *fr_pool;
+	};
 	u32			lkey;
 	u32			rkey;
 	enum srp_target_state	state;
@@ -197,15 +210,66 @@ struct srp_iu {
 	enum dma_data_direction	direction;
 };
 
+/**
+ * struct srp_fr_desc - fast registration work request arguments
+ * @entry: Entry in srp_fr_pool.free_list.
+ * @mr:    Memory region.
+ * @frpl:  Fast registration page list.
+ */
+struct srp_fr_desc {
+	struct list_head		entry;
+	struct ib_mr			*mr;
+	struct ib_fast_reg_page_list	*frpl;
+};
+
+/**
+ * struct srp_fr_pool - pool of fast registration descriptors
+ *
+ * An entry is available for allocation if and only if it occurs in @free_list.
+ *
+ * @size:      Number of descriptors in this pool.
+ * @max_page_list_len: Maximum fast registration work request page list length.
+ * @lock:      Protects free_list.
+ * @free_list: List of free descriptors.
+ * @desc:      Fast registration descriptor pool.
+ */
+struct srp_fr_pool {
+	int			size;
+	int			max_page_list_len;
+	spinlock_t		lock;
+	struct list_head	free_list;
+	struct srp_fr_desc	desc[0];
+};
+
+/**
+ * struct srp_map_state - per-request DMA memory mapping state
+ * @desc:	    Pointer to the element of the SRP buffer descriptor array
+ *		    that is being filled in.
+ * @pages:	    Array with DMA addresses of pages being considered for
+ *		    memory registration.
+ * @base_dma_addr:  DMA address of the first page that has not yet been mapped.
+ * @dma_len:	    Number of bytes that will be registered with the next
+ *		    FMR or FR memory registration call.
+ * @total_len:	    Total number of bytes in the sg-list being mapped.
+ * @npages:	    Number of page addresses in the pages[] array.
+ * @nmdesc:	    Number of FMR or FR memory descriptors used for mapping.
+ * @ndesc:	    Number of SRP buffer descriptors that have been filled in.
+ * @unmapped_sg:    First element of the sg-list that is mapped via FMR or FR.
+ * @unmapped_index: Index of the first element mapped via FMR or FR.
+ * @unmapped_addr:  DMA address of the first element mapped via FMR or FR.
+ */
 struct srp_map_state {
-	struct ib_pool_fmr    **next_fmr;
+	union {
+		struct ib_pool_fmr **next_fmr;
+		struct srp_fr_desc **next_fr;
+	};
 	struct srp_direct_buf  *desc;
 	u64		       *pages;
 	dma_addr_t		base_dma_addr;
-	u32			fmr_len;
+	u32			dma_len;
 	u32			total_len;
 	unsigned int		npages;
-	unsigned int		nfmr;
+	unsigned int		nmdesc;
 	unsigned int		ndesc;
 	struct scatterlist     *unmapped_sg;
 	int			unmapped_index;

drivers/net/ethernet/mellanox/mlx4/alloc.c

@@ -171,7 +171,7 @@ void mlx4_bitmap_cleanup(struct mlx4_bitmap *bitmap)
  */
 
 int mlx4_buf_alloc(struct mlx4_dev *dev, int size, int max_direct,
-		   struct mlx4_buf *buf)
+		   struct mlx4_buf *buf, gfp_t gfp)
 {
 	dma_addr_t t;
 
@@ -180,7 +180,7 @@ int mlx4_buf_alloc(struct mlx4_dev *dev, int size, int max_direct,
 		buf->npages       = 1;
 		buf->page_shift   = get_order(size) + PAGE_SHIFT;
 		buf->direct.buf   = dma_alloc_coherent(&dev->pdev->dev,
-						       size, &t, GFP_KERNEL);
+						       size, &t, gfp);
 		if (!buf->direct.buf)
 			return -ENOMEM;
 
@@ -200,14 +200,14 @@ int mlx4_buf_alloc(struct mlx4_dev *dev, int size, int max_direct,
 		buf->npages      = buf->nbufs;
 		buf->page_shift  = PAGE_SHIFT;
 		buf->page_list   = kcalloc(buf->nbufs, sizeof(*buf->page_list),
-					   GFP_KERNEL);
+					   gfp);
 		if (!buf->page_list)
 			return -ENOMEM;
 
 		for (i = 0; i < buf->nbufs; ++i) {
 			buf->page_list[i].buf =
 				dma_alloc_coherent(&dev->pdev->dev, PAGE_SIZE,
-						   &t, GFP_KERNEL);
+						   &t, gfp);
 			if (!buf->page_list[i].buf)
 				goto err_free;
 
@@ -218,7 +218,7 @@ int mlx4_buf_alloc(struct mlx4_dev *dev, int size, int max_direct,
 
 		if (BITS_PER_LONG == 64) {
 			struct page **pages;
-			pages = kmalloc(sizeof *pages * buf->nbufs, GFP_KERNEL);
+			pages = kmalloc(sizeof *pages * buf->nbufs, gfp);
 			if (!pages)
 				goto err_free;
 			for (i = 0; i < buf->nbufs; ++i)
@@ -260,11 +260,12 @@ void mlx4_buf_free(struct mlx4_dev *dev, int size, struct mlx4_buf *buf)
 }
 EXPORT_SYMBOL_GPL(mlx4_buf_free);
 
-static struct mlx4_db_pgdir *mlx4_alloc_db_pgdir(struct device *dma_device)
+static struct mlx4_db_pgdir *mlx4_alloc_db_pgdir(struct device *dma_device,
+						 gfp_t gfp)
 {
 	struct mlx4_db_pgdir *pgdir;
 
-	pgdir = kzalloc(sizeof *pgdir, GFP_KERNEL);
+	pgdir = kzalloc(sizeof *pgdir, gfp);
 	if (!pgdir)
 		return NULL;
 
@@ -272,7 +273,7 @@ static struct mlx4_db_pgdir *mlx4_alloc_db_pgdir(struct device *dma_device)
 	pgdir->bits[0] = pgdir->order0;
 	pgdir->bits[1] = pgdir->order1;
 	pgdir->db_page = dma_alloc_coherent(dma_device, PAGE_SIZE,
-					    &pgdir->db_dma, GFP_KERNEL);
+					    &pgdir->db_dma, gfp);
 	if (!pgdir->db_page) {
 		kfree(pgdir);
 		return NULL;
@@ -312,7 +313,7 @@ static int mlx4_alloc_db_from_pgdir(struct mlx4_db_pgdir *pgdir,
 	return 0;
 }
 
-int mlx4_db_alloc(struct mlx4_dev *dev, struct mlx4_db *db, int order)
+int mlx4_db_alloc(struct mlx4_dev *dev, struct mlx4_db *db, int order, gfp_t gfp)
 {
 	struct mlx4_priv *priv = mlx4_priv(dev);
 	struct mlx4_db_pgdir *pgdir;
@@ -324,7 +325,7 @@ int mlx4_db_alloc(struct mlx4_dev *dev, struct mlx4_db *db, int order)
 		if (!mlx4_alloc_db_from_pgdir(pgdir, db, order))
 			goto out;
 
-	pgdir = mlx4_alloc_db_pgdir(&(dev->pdev->dev));
+	pgdir = mlx4_alloc_db_pgdir(&(dev->pdev->dev), gfp);
 	if (!pgdir) {
 		ret = -ENOMEM;
 		goto out;
@@ -376,13 +377,13 @@ int mlx4_alloc_hwq_res(struct mlx4_dev *dev, struct mlx4_hwq_resources *wqres,
 {
 	int err;
 
-	err = mlx4_db_alloc(dev, &wqres->db, 1);
+	err = mlx4_db_alloc(dev, &wqres->db, 1, GFP_KERNEL);
 	if (err)
 		return err;
 
 	*wqres->db.db = 0;
 
-	err = mlx4_buf_alloc(dev, size, max_direct, &wqres->buf);
+	err = mlx4_buf_alloc(dev, size, max_direct, &wqres->buf, GFP_KERNEL);
 	if (err)
 		goto err_db;
 
@@ -391,7 +392,7 @@ int mlx4_alloc_hwq_res(struct mlx4_dev *dev, struct mlx4_hwq_resources *wqres,
 	if (err)
 		goto err_buf;
 
-	err = mlx4_buf_write_mtt(dev, &wqres->mtt, &wqres->buf);
+	err = mlx4_buf_write_mtt(dev, &wqres->mtt, &wqres->buf, GFP_KERNEL);
 	if (err)
 		goto err_mtt;
 

drivers/net/ethernet/mellanox/mlx4/cmd.c

@@ -705,20 +705,28 @@ static int mlx4_MAD_IFC_wrapper(struct mlx4_dev *dev, int slave,
 	struct ib_smp *smp = inbox->buf;
 	u32 index;
 	u8 port;
+	u8 opcode_modifier;
 	u16 *table;
 	int err;
 	int vidx, pidx;
+	int network_view;
 	struct mlx4_priv *priv = mlx4_priv(dev);
 	struct ib_smp *outsmp = outbox->buf;
 	__be16 *outtab = (__be16 *)(outsmp->data);
 	__be32 slave_cap_mask;
 	__be64 slave_node_guid;
+
 	port = vhcr->in_modifier;
 
+	/* network-view bit is for driver use only, and should not be passed to FW */
+	opcode_modifier = vhcr->op_modifier & ~0x8; /* clear netw view bit */
+	network_view = !!(vhcr->op_modifier & 0x8);
+
 	if (smp->base_version == 1 &&
 	    smp->mgmt_class == IB_MGMT_CLASS_SUBN_LID_ROUTED &&
 	    smp->class_version == 1) {
-		if (smp->method	== IB_MGMT_METHOD_GET) {
+		/* host view is paravirtualized */
+		if (!network_view && smp->method == IB_MGMT_METHOD_GET) {
 			if (smp->attr_id == IB_SMP_ATTR_PKEY_TABLE) {
 				index = be32_to_cpu(smp->attr_mod);
 				if (port < 1 || port > dev->caps.num_ports)
@@ -743,7 +751,7 @@ static int mlx4_MAD_IFC_wrapper(struct mlx4_dev *dev, int slave,
 				/*get the slave specific caps:*/
 				/*do the command */
 				err = mlx4_cmd_box(dev, inbox->dma, outbox->dma,
-					    vhcr->in_modifier, vhcr->op_modifier,
+					    vhcr->in_modifier, opcode_modifier,
 					    vhcr->op, MLX4_CMD_TIME_CLASS_C, MLX4_CMD_NATIVE);
 				/* modify the response for slaves */
 				if (!err && slave != mlx4_master_func_num(dev)) {
@@ -760,7 +768,7 @@ static int mlx4_MAD_IFC_wrapper(struct mlx4_dev *dev, int slave,
 				smp->attr_mod = cpu_to_be32(slave / 8);
 				/* execute cmd */
 				err = mlx4_cmd_box(dev, inbox->dma, outbox->dma,
-					     vhcr->in_modifier, vhcr->op_modifier,
+					     vhcr->in_modifier, opcode_modifier,
 					     vhcr->op, MLX4_CMD_TIME_CLASS_C, MLX4_CMD_NATIVE);
 				if (!err) {
 					/* if needed, move slave gid to index 0 */
@@ -774,7 +782,7 @@ static int mlx4_MAD_IFC_wrapper(struct mlx4_dev *dev, int slave,
 			}
 			if (smp->attr_id == IB_SMP_ATTR_NODE_INFO) {
 				err = mlx4_cmd_box(dev, inbox->dma, outbox->dma,
-					     vhcr->in_modifier, vhcr->op_modifier,
+					     vhcr->in_modifier, opcode_modifier,
 					     vhcr->op, MLX4_CMD_TIME_CLASS_C, MLX4_CMD_NATIVE);
 				if (!err) {
 					slave_node_guid =  mlx4_get_slave_node_guid(dev, slave);
@@ -784,19 +792,24 @@ static int mlx4_MAD_IFC_wrapper(struct mlx4_dev *dev, int slave,
 			}
 		}
 	}
+
+	/* Non-privileged VFs are only allowed "host" view LID-routed 'Get' MADs.
+	 * These are the MADs used by ib verbs (such as ib_query_gids).
+	 */
 	if (slave != mlx4_master_func_num(dev) &&
-	    ((smp->mgmt_class == IB_MGMT_CLASS_SUBN_DIRECTED_ROUTE) ||
-	     (smp->mgmt_class == IB_MGMT_CLASS_SUBN_LID_ROUTED &&
-	      smp->method == IB_MGMT_METHOD_SET))) {
-		mlx4_err(dev, "slave %d is trying to execute a Subnet MGMT MAD, "
-			 "class 0x%x, method 0x%x for attr 0x%x. Rejecting\n",
-			 slave, smp->method, smp->mgmt_class,
-			 be16_to_cpu(smp->attr_id));
-		return -EPERM;
+	    !mlx4_vf_smi_enabled(dev, slave, port)) {
+		if (!(smp->mgmt_class == IB_MGMT_CLASS_SUBN_LID_ROUTED &&
+		      smp->method == IB_MGMT_METHOD_GET) || network_view) {
+			mlx4_err(dev, "Unprivileged slave %d is trying to execute a Subnet MGMT MAD, class 0x%x, method 0x%x, view=%s for attr 0x%x. Rejecting\n",
+				 slave, smp->method, smp->mgmt_class,
+				 network_view ? "Network" : "Host",
+				 be16_to_cpu(smp->attr_id));
+			return -EPERM;
+		}
 	}
-	/*default:*/
+
 	return mlx4_cmd_box(dev, inbox->dma, outbox->dma,
-				    vhcr->in_modifier, vhcr->op_modifier,
+				    vhcr->in_modifier, opcode_modifier,
 				    vhcr->op, MLX4_CMD_TIME_CLASS_C, MLX4_CMD_NATIVE);
 }
 
@@ -1653,6 +1666,8 @@ static int mlx4_master_activate_admin_state(struct mlx4_priv *priv, int slave)
 	for (port = min_port; port <= max_port; port++) {
 		if (!test_bit(port - 1, actv_ports.ports))
 			continue;
+		priv->mfunc.master.vf_oper[slave].smi_enabled[port] =
+			priv->mfunc.master.vf_admin[slave].enable_smi[port];
 		vp_oper = &priv->mfunc.master.vf_oper[slave].vport[port];
 		vp_admin = &priv->mfunc.master.vf_admin[slave].vport[port];
 		vp_oper->state = *vp_admin;
@@ -1704,6 +1719,8 @@ static void mlx4_master_deactivate_admin_state(struct mlx4_priv *priv, int slave
 	for (port = min_port; port <= max_port; port++) {
 		if (!test_bit(port - 1, actv_ports.ports))
 			continue;
+		priv->mfunc.master.vf_oper[slave].smi_enabled[port] =
+			MLX4_VF_SMI_DISABLED;
 		vp_oper = &priv->mfunc.master.vf_oper[slave].vport[port];
 		if (NO_INDX != vp_oper->vlan_idx) {
 			__mlx4_unregister_vlan(&priv->dev,
@@ -2537,3 +2554,50 @@ int mlx4_set_vf_link_state(struct mlx4_dev *dev, int port, int vf, int link_stat
 	return 0;
 }
 EXPORT_SYMBOL_GPL(mlx4_set_vf_link_state);
+
+int mlx4_vf_smi_enabled(struct mlx4_dev *dev, int slave, int port)
+{
+	struct mlx4_priv *priv = mlx4_priv(dev);
+
+	if (slave < 1 || slave >= dev->num_slaves ||
+	    port < 1 || port > MLX4_MAX_PORTS)
+		return 0;
+
+	return priv->mfunc.master.vf_oper[slave].smi_enabled[port] ==
+		MLX4_VF_SMI_ENABLED;
+}
+EXPORT_SYMBOL_GPL(mlx4_vf_smi_enabled);
+
+int mlx4_vf_get_enable_smi_admin(struct mlx4_dev *dev, int slave, int port)
+{
+	struct mlx4_priv *priv = mlx4_priv(dev);
+
+	if (slave == mlx4_master_func_num(dev))
+		return 1;
+
+	if (slave < 1 || slave >= dev->num_slaves ||
+	    port < 1 || port > MLX4_MAX_PORTS)
+		return 0;
+
+	return priv->mfunc.master.vf_admin[slave].enable_smi[port] ==
+		MLX4_VF_SMI_ENABLED;
+}
+EXPORT_SYMBOL_GPL(mlx4_vf_get_enable_smi_admin);
+
+int mlx4_vf_set_enable_smi_admin(struct mlx4_dev *dev, int slave, int port,
+				 int enabled)
+{
+	struct mlx4_priv *priv = mlx4_priv(dev);
+
+	if (slave == mlx4_master_func_num(dev))
+		return 0;
+
+	if (slave < 1 || slave >= dev->num_slaves ||
+	    port < 1 || port > MLX4_MAX_PORTS ||
+	    enabled < 0 || enabled > 1)
+		return -EINVAL;
+
+	priv->mfunc.master.vf_admin[slave].enable_smi[port] = enabled;
+	return 0;
+}
+EXPORT_SYMBOL_GPL(mlx4_vf_set_enable_smi_admin);

drivers/net/ethernet/mellanox/mlx4/cq.c

@@ -173,11 +173,11 @@ int __mlx4_cq_alloc_icm(struct mlx4_dev *dev, int *cqn)
 	if (*cqn == -1)
 		return -ENOMEM;
 
-	err = mlx4_table_get(dev, &cq_table->table, *cqn);
+	err = mlx4_table_get(dev, &cq_table->table, *cqn, GFP_KERNEL);
 	if (err)
 		goto err_out;
 
-	err = mlx4_table_get(dev, &cq_table->cmpt_table, *cqn);
+	err = mlx4_table_get(dev, &cq_table->cmpt_table, *cqn, GFP_KERNEL);
 	if (err)
 		goto err_put;
 	return 0;

drivers/net/ethernet/mellanox/mlx4/en_rx.c

@@ -972,7 +972,7 @@ static int mlx4_en_config_rss_qp(struct mlx4_en_priv *priv, int qpn,
 	if (!context)
 		return -ENOMEM;
 
-	err = mlx4_qp_alloc(mdev->dev, qpn, qp);
+	err = mlx4_qp_alloc(mdev->dev, qpn, qp, GFP_KERNEL);
 	if (err) {
 		en_err(priv, "Failed to allocate qp #%x\n", qpn);
 		goto out;
@@ -1012,7 +1012,7 @@ int mlx4_en_create_drop_qp(struct mlx4_en_priv *priv)
 		en_err(priv, "Failed reserving drop qpn\n");
 		return err;
 	}
-	err = mlx4_qp_alloc(priv->mdev->dev, qpn, &priv->drop_qp);
+	err = mlx4_qp_alloc(priv->mdev->dev, qpn, &priv->drop_qp, GFP_KERNEL);
 	if (err) {
 		en_err(priv, "Failed allocating drop qp\n");
 		mlx4_qp_release_range(priv->mdev->dev, qpn, 1);
@@ -1071,7 +1071,7 @@ int mlx4_en_config_rss_steer(struct mlx4_en_priv *priv)
 	}
 
 	/* Configure RSS indirection qp */
-	err = mlx4_qp_alloc(mdev->dev, priv->base_qpn, &rss_map->indir_qp);
+	err = mlx4_qp_alloc(mdev->dev, priv->base_qpn, &rss_map->indir_qp, GFP_KERNEL);
 	if (err) {
 		en_err(priv, "Failed to allocate RSS indirection QP\n");
 		goto rss_err;

drivers/net/ethernet/mellanox/mlx4/en_tx.c

@@ -113,7 +113,7 @@ int mlx4_en_create_tx_ring(struct mlx4_en_priv *priv,
 	       ring->buf_size, (unsigned long long) ring->wqres.buf.direct.map);
 
 	ring->qpn = qpn;
-	err = mlx4_qp_alloc(mdev->dev, ring->qpn, &ring->qp);
+	err = mlx4_qp_alloc(mdev->dev, ring->qpn, &ring->qp, GFP_KERNEL);
 	if (err) {
 		en_err(priv, "Failed allocating qp %d\n", ring->qpn);
 		goto err_map;

drivers/net/ethernet/mellanox/mlx4/fw.h

@@ -134,6 +134,7 @@ struct mlx4_func_cap {
 	int	max_eq;
 	int	reserved_eq;
 	int	mcg_quota;
+	u32	qp0_qkey;
 	u32	qp0_tunnel_qpn;
 	u32	qp0_proxy_qpn;
 	u32	qp1_tunnel_qpn;

drivers/net/ethernet/mellanox/mlx4/icm.c

@@ -245,7 +245,8 @@ int mlx4_UNMAP_ICM_AUX(struct mlx4_dev *dev)
 			MLX4_CMD_TIME_CLASS_B, MLX4_CMD_NATIVE);
 }
 
-int mlx4_table_get(struct mlx4_dev *dev, struct mlx4_icm_table *table, u32 obj)
+int mlx4_table_get(struct mlx4_dev *dev, struct mlx4_icm_table *table, u32 obj,
+		   gfp_t gfp)
 {
 	u32 i = (obj & (table->num_obj - 1)) /
 			(MLX4_TABLE_CHUNK_SIZE / table->obj_size);
@@ -259,7 +260,7 @@ int mlx4_table_get(struct mlx4_dev *dev, struct mlx4_icm_table *table, u32 obj)
 	}
 
 	table->icm[i] = mlx4_alloc_icm(dev, MLX4_TABLE_CHUNK_SIZE >> PAGE_SHIFT,
-				       (table->lowmem ? GFP_KERNEL : GFP_HIGHUSER) |
+				       (table->lowmem ? gfp : GFP_HIGHUSER) |
 				       __GFP_NOWARN, table->coherent);
 	if (!table->icm[i]) {
 		ret = -ENOMEM;
@@ -356,7 +357,7 @@ int mlx4_table_get_range(struct mlx4_dev *dev, struct mlx4_icm_table *table,
 	u32 i;
 
 	for (i = start; i <= end; i += inc) {
-		err = mlx4_table_get(dev, table, i);
+		err = mlx4_table_get(dev, table, i, GFP_KERNEL);
 		if (err)
 			goto fail;
 	}

drivers/net/ethernet/mellanox/mlx4/icm.h

@@ -71,7 +71,8 @@ struct mlx4_icm *mlx4_alloc_icm(struct mlx4_dev *dev, int npages,
 				gfp_t gfp_mask, int coherent);
 void mlx4_free_icm(struct mlx4_dev *dev, struct mlx4_icm *icm, int coherent);
 
-int mlx4_table_get(struct mlx4_dev *dev, struct mlx4_icm_table *table, u32 obj);
+int mlx4_table_get(struct mlx4_dev *dev, struct mlx4_icm_table *table, u32 obj,
+		   gfp_t gfp);
 void mlx4_table_put(struct mlx4_dev *dev, struct mlx4_icm_table *table, u32 obj);
 int mlx4_table_get_range(struct mlx4_dev *dev, struct mlx4_icm_table *table,
 			 u32 start, u32 end);

drivers/net/ethernet/mellanox/mlx4/srq.c

@@ -103,11 +103,11 @@ int __mlx4_srq_alloc_icm(struct mlx4_dev *dev, int *srqn)
 	if (*srqn == -1)
 		return -ENOMEM;
 
-	err = mlx4_table_get(dev, &srq_table->table, *srqn);
+	err = mlx4_table_get(dev, &srq_table->table, *srqn, GFP_KERNEL);
 	if (err)
 		goto err_out;
 
-	err = mlx4_table_get(dev, &srq_table->cmpt_table, *srqn);
+	err = mlx4_table_get(dev, &srq_table->cmpt_table, *srqn, GFP_KERNEL);
 	if (err)
 		goto err_put;
 	return 0;