Merge branches 'misc' and 'rxe' into k.o/for-4.8-1

MAINTAINERS

@@ -7444,6 +7444,15 @@ W:	http://www.mellanox.com
 Q:	http://patchwork.ozlabs.org/project/netdev/list/
 F:	drivers/net/ethernet/mellanox/mlxsw/
 
+SOFT-ROCE DRIVER (rxe)
+M:	Moni Shoua <monis@mellanox.com>
+L:	linux-rdma@vger.kernel.org
+S:	Supported
+W:	https://github.com/SoftRoCE/rxe-dev/wiki/rxe-dev:-Home
+Q:	http://patchwork.kernel.org/project/linux-rdma/list/
+F:	drivers/infiniband/hw/rxe/
+F:	include/uapi/rdma/rdma_user_rxe.h
+
 MEMBARRIER SUPPORT
 M:	Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
 M:	"Paul E. McKenney" <paulmck@linux.vnet.ibm.com>

drivers/infiniband/Kconfig

@@ -84,6 +84,7 @@ source "drivers/infiniband/ulp/iser/Kconfig"
 source "drivers/infiniband/ulp/isert/Kconfig"
 
 source "drivers/infiniband/sw/rdmavt/Kconfig"
+source "drivers/infiniband/sw/rxe/Kconfig"
 
 source "drivers/infiniband/hw/hfi1/Kconfig"
 

drivers/infiniband/core/cma.c

@@ -68,6 +68,7 @@ MODULE_DESCRIPTION("Generic RDMA CM Agent");
 MODULE_LICENSE("Dual BSD/GPL");
 
 #define CMA_CM_RESPONSE_TIMEOUT 20
+#define CMA_QUERY_CLASSPORT_INFO_TIMEOUT 3000
 #define CMA_MAX_CM_RETRIES 15
 #define CMA_CM_MRA_SETTING (IB_CM_MRA_FLAG_DELAY | 24)
 #define CMA_IBOE_PACKET_LIFETIME 18
@@ -162,6 +163,14 @@ struct rdma_bind_list {
 	unsigned short		port;
 };
 
+struct class_port_info_context {
+	struct ib_class_port_info	*class_port_info;
+	struct ib_device		*device;
+	struct completion		done;
+	struct ib_sa_query		*sa_query;
+	u8				port_num;
+};
+
 static int cma_ps_alloc(struct net *net, enum rdma_port_space ps,
 			struct rdma_bind_list *bind_list, int snum)
 {
@@ -306,6 +315,7 @@ struct cma_multicast {
 	struct sockaddr_storage	addr;
 	struct kref		mcref;
 	bool			igmp_joined;
+	u8			join_state;
 };
 
 struct cma_work {
@@ -3754,10 +3764,63 @@ static void cma_set_mgid(struct rdma_id_private *id_priv,
 	}
 }
 
+static void cma_query_sa_classport_info_cb(int status,
+					   struct ib_class_port_info *rec,
+					   void *context)
+{
+	struct class_port_info_context *cb_ctx = context;
+
+	WARN_ON(!context);
+
+	if (status || !rec) {
+		pr_debug("RDMA CM: %s port %u failed query ClassPortInfo status: %d\n",
+			 cb_ctx->device->name, cb_ctx->port_num, status);
+		goto out;
+	}
+
+	memcpy(cb_ctx->class_port_info, rec, sizeof(struct ib_class_port_info));
+
+out:
+	complete(&cb_ctx->done);
+}
+
+static int cma_query_sa_classport_info(struct ib_device *device, u8 port_num,
+				       struct ib_class_port_info *class_port_info)
+{
+	struct class_port_info_context *cb_ctx;
+	int ret;
+
+	cb_ctx = kmalloc(sizeof(*cb_ctx), GFP_KERNEL);
+	if (!cb_ctx)
+		return -ENOMEM;
+
+	cb_ctx->device = device;
+	cb_ctx->class_port_info = class_port_info;
+	cb_ctx->port_num = port_num;
+	init_completion(&cb_ctx->done);
+
+	ret = ib_sa_classport_info_rec_query(&sa_client, device, port_num,
+					     CMA_QUERY_CLASSPORT_INFO_TIMEOUT,
+					     GFP_KERNEL, cma_query_sa_classport_info_cb,
+					     cb_ctx, &cb_ctx->sa_query);
+	if (ret < 0) {
+		pr_err("RDMA CM: %s port %u failed to send ClassPortInfo query, ret: %d\n",
+		       device->name, port_num, ret);
+		goto out;
+	}
+
+	wait_for_completion(&cb_ctx->done);
+
+out:
+	kfree(cb_ctx);
+	return ret;
+}
+
 static int cma_join_ib_multicast(struct rdma_id_private *id_priv,
 				 struct cma_multicast *mc)
 {
 	struct ib_sa_mcmember_rec rec;
+	struct ib_class_port_info class_port_info;
 	struct rdma_dev_addr *dev_addr = &id_priv->id.route.addr.dev_addr;
 	ib_sa_comp_mask comp_mask;
 	int ret;
@@ -3776,7 +3839,24 @@ static int cma_join_ib_multicast(struct rdma_id_private *id_priv,
 	rec.qkey = cpu_to_be32(id_priv->qkey);
 	rdma_addr_get_sgid(dev_addr, &rec.port_gid);
 	rec.pkey = cpu_to_be16(ib_addr_get_pkey(dev_addr));
-	rec.join_state = 1;
+	rec.join_state = mc->join_state;
+
+	if (rec.join_state == BIT(SENDONLY_FULLMEMBER_JOIN)) {
+		ret = cma_query_sa_classport_info(id_priv->id.device,
+						  id_priv->id.port_num,
+						  &class_port_info);
+
+		if (ret)
+			return ret;
+
+		if (!(ib_get_cpi_capmask2(&class_port_info) &
+		      IB_SA_CAP_MASK2_SENDONLY_FULL_MEM_SUPPORT)) {
+			pr_warn("RDMA CM: %s port %u Unable to multicast join\n"
+				"RDMA CM: SM doesn't support Send Only Full Member option\n",
+				id_priv->id.device->name, id_priv->id.port_num);
+			return -EOPNOTSUPP;
+		}
+	}
 
 	comp_mask = IB_SA_MCMEMBER_REC_MGID | IB_SA_MCMEMBER_REC_PORT_GID |
 		    IB_SA_MCMEMBER_REC_PKEY | IB_SA_MCMEMBER_REC_JOIN_STATE |
@@ -3845,6 +3925,9 @@ static int cma_iboe_join_multicast(struct rdma_id_private *id_priv,
 	struct sockaddr *addr = (struct sockaddr *)&mc->addr;
 	struct net_device *ndev = NULL;
 	enum ib_gid_type gid_type;
+	bool send_only;
+
+	send_only = mc->join_state == BIT(SENDONLY_FULLMEMBER_JOIN);
 
 	if (cma_zero_addr((struct sockaddr *)&mc->addr))
 		return -EINVAL;
@@ -3878,12 +3961,14 @@ static int cma_iboe_join_multicast(struct rdma_id_private *id_priv,
 	gid_type = id_priv->cma_dev->default_gid_type[id_priv->id.port_num -
 		   rdma_start_port(id_priv->cma_dev->device)];
 	if (addr->sa_family == AF_INET) {
-		if (gid_type == IB_GID_TYPE_ROCE_UDP_ENCAP)
+		if (gid_type == IB_GID_TYPE_ROCE_UDP_ENCAP) {
+			mc->multicast.ib->rec.hop_limit = IPV6_DEFAULT_HOPLIMIT;
+			if (!send_only) {
 				err = cma_igmp_send(ndev, &mc->multicast.ib->rec.mgid,
 						    true);
-		if (!err) {
+				if (!err)
 					mc->igmp_joined = true;
-			mc->multicast.ib->rec.hop_limit = IPV6_DEFAULT_HOPLIMIT;
+			}
 		}
 	} else {
 		if (gid_type == IB_GID_TYPE_ROCE_UDP_ENCAP)
@@ -3913,7 +3998,7 @@ static int cma_iboe_join_multicast(struct rdma_id_private *id_priv,
 }
 
 int rdma_join_multicast(struct rdma_cm_id *id, struct sockaddr *addr,
-			void *context)
+			u8 join_state, void *context)
 {
 	struct rdma_id_private *id_priv;
 	struct cma_multicast *mc;
@@ -3932,6 +4017,7 @@ int rdma_join_multicast(struct rdma_cm_id *id, struct sockaddr *addr,
 	mc->context = context;
 	mc->id_priv = id_priv;
 	mc->igmp_joined = false;
+	mc->join_state = join_state;
 	spin_lock(&id_priv->lock);
 	list_add(&mc->list, &id_priv->mc_list);
 	spin_unlock(&id_priv->lock);

drivers/infiniband/core/iwpm_util.c

@@ -37,6 +37,7 @@
 #define IWPM_MAPINFO_HASH_MASK	(IWPM_MAPINFO_HASH_SIZE - 1)
 #define IWPM_REMINFO_HASH_SIZE	64
 #define IWPM_REMINFO_HASH_MASK	(IWPM_REMINFO_HASH_SIZE - 1)
+#define IWPM_MSG_SIZE		512
 
 static LIST_HEAD(iwpm_nlmsg_req_list);
 static DEFINE_SPINLOCK(iwpm_nlmsg_req_lock);
@@ -452,7 +453,7 @@ struct sk_buff *iwpm_create_nlmsg(u32 nl_op, struct nlmsghdr **nlh,
 {
 	struct sk_buff *skb = NULL;
 
-	skb = dev_alloc_skb(NLMSG_GOODSIZE);
+	skb = dev_alloc_skb(IWPM_MSG_SIZE);
 	if (!skb) {
 		pr_err("%s Unable to allocate skb\n", __func__);
 		goto create_nlmsg_exit;

drivers/infiniband/core/multicast.c

@@ -93,18 +93,6 @@ enum {
 
 struct mcast_member;
 
-/*
-* There are 4 types of join states:
-* FullMember, NonMember, SendOnlyNonMember, SendOnlyFullMember.
-*/
-enum {
-	FULLMEMBER_JOIN,
-	NONMEMBER_JOIN,
-	SENDONLY_NONMEBER_JOIN,
-	SENDONLY_FULLMEMBER_JOIN,
-	NUM_JOIN_MEMBERSHIP_TYPES,
-};
-
 struct mcast_group {
 	struct ib_sa_mcmember_rec rec;
 	struct rb_node		node;

drivers/infiniband/core/sa_query.c

@@ -65,10 +65,17 @@ struct ib_sa_sm_ah {
 	u8		     src_path_mask;
 };
 
+struct ib_sa_classport_cache {
+	bool valid;
+	struct ib_class_port_info data;
+};
+
 struct ib_sa_port {
 	struct ib_mad_agent *agent;
 	struct ib_sa_sm_ah  *sm_ah;
 	struct work_struct   update_task;
+	struct ib_sa_classport_cache classport_info;
+	spinlock_t                   classport_lock; /* protects class port info set */
 	spinlock_t           ah_lock;
 	u8                   port_num;
 };
@@ -998,6 +1005,13 @@ static void ib_sa_event(struct ib_event_handler *handler, struct ib_event *event
 		port->sm_ah = NULL;
 		spin_unlock_irqrestore(&port->ah_lock, flags);
 
+		if (event->event == IB_EVENT_SM_CHANGE ||
+		    event->event == IB_EVENT_CLIENT_REREGISTER ||
+		    event->event == IB_EVENT_LID_CHANGE) {
+			spin_lock_irqsave(&port->classport_lock, flags);
+			port->classport_info.valid = false;
+			spin_unlock_irqrestore(&port->classport_lock, flags);
+		}
 		queue_work(ib_wq, &sa_dev->port[event->element.port_num -
 					    sa_dev->start_port].update_task);
 	}
@@ -1719,6 +1733,7 @@ static void ib_sa_classport_info_rec_callback(struct ib_sa_query *sa_query,
 					      int status,
 					      struct ib_sa_mad *mad)
 {
+	unsigned long flags;
 	struct ib_sa_classport_info_query *query =
 		container_of(sa_query, struct ib_sa_classport_info_query, sa_query);
 
@@ -1728,6 +1743,16 @@ static void ib_sa_classport_info_rec_callback(struct ib_sa_query *sa_query,
 		ib_unpack(classport_info_rec_table,
 			  ARRAY_SIZE(classport_info_rec_table),
 			  mad->data, &rec);
+
+		spin_lock_irqsave(&sa_query->port->classport_lock, flags);
+		if (!status && !sa_query->port->classport_info.valid) {
+			memcpy(&sa_query->port->classport_info.data, &rec,
+			       sizeof(sa_query->port->classport_info.data));
+
+			sa_query->port->classport_info.valid = true;
+		}
+		spin_unlock_irqrestore(&sa_query->port->classport_lock, flags);
+
 		query->callback(status, &rec, query->context);
 	} else {
 		query->callback(status, NULL, query->context);
@@ -1754,7 +1779,9 @@ int ib_sa_classport_info_rec_query(struct ib_sa_client *client,
 	struct ib_sa_port *port;
 	struct ib_mad_agent *agent;
 	struct ib_sa_mad *mad;
+	struct ib_class_port_info cached_class_port_info;
 	int ret;
+	unsigned long flags;
 
 	if (!sa_dev)
 		return -ENODEV;
@@ -1762,6 +1789,17 @@ int ib_sa_classport_info_rec_query(struct ib_sa_client *client,
 	port  = &sa_dev->port[port_num - sa_dev->start_port];
 	agent = port->agent;
 
+	/* Use cached ClassPortInfo attribute if valid instead of sending mad */
+	spin_lock_irqsave(&port->classport_lock, flags);
+	if (port->classport_info.valid && callback) {
+		memcpy(&cached_class_port_info, &port->classport_info.data,
+		       sizeof(cached_class_port_info));
+		spin_unlock_irqrestore(&port->classport_lock, flags);
+		callback(0, &cached_class_port_info, context);
+		return 0;
+	}
+	spin_unlock_irqrestore(&port->classport_lock, flags);
+
 	query = kzalloc(sizeof(*query), gfp_mask);
 	if (!query)
 		return -ENOMEM;
@@ -1885,6 +1923,9 @@ static void ib_sa_add_one(struct ib_device *device)
 		sa_dev->port[i].sm_ah    = NULL;
 		sa_dev->port[i].port_num = i + s;
 
+		spin_lock_init(&sa_dev->port[i].classport_lock);
+		sa_dev->port[i].classport_info.valid = false;
+
 		sa_dev->port[i].agent =
 			ib_register_mad_agent(device, i + s, IB_QPT_GSI,
 					      NULL, 0, send_handler,

drivers/infiniband/core/ucma.c

@@ -106,6 +106,7 @@ struct ucma_multicast {
 	int			events_reported;
 
 	u64			uid;
+	u8			join_state;
 	struct list_head	list;
 	struct sockaddr_storage	addr;
 };
@@ -1317,12 +1318,20 @@ static ssize_t ucma_process_join(struct ucma_file *file,
 	struct ucma_multicast *mc;
 	struct sockaddr *addr;
 	int ret;
+	u8 join_state;
 
 	if (out_len < sizeof(resp))
 		return -ENOSPC;
 
 	addr = (struct sockaddr *) &cmd->addr;
-	if (cmd->reserved || !cmd->addr_size || (cmd->addr_size != rdma_addr_size(addr)))
+	if (!cmd->addr_size || (cmd->addr_size != rdma_addr_size(addr)))
+		return -EINVAL;
+
+	if (cmd->join_flags == RDMA_MC_JOIN_FLAG_FULLMEMBER)
+		join_state = BIT(FULLMEMBER_JOIN);
+	else if (cmd->join_flags == RDMA_MC_JOIN_FLAG_SENDONLY_FULLMEMBER)
+		join_state = BIT(SENDONLY_FULLMEMBER_JOIN);
+	else
 		return -EINVAL;
 
 	ctx = ucma_get_ctx(file, cmd->id);
@@ -1335,10 +1344,11 @@ static ssize_t ucma_process_join(struct ucma_file *file,
 		ret = -ENOMEM;
 		goto err1;
 	}
-
+	mc->join_state = join_state;
 	mc->uid = cmd->uid;
 	memcpy(&mc->addr, addr, cmd->addr_size);
-	ret = rdma_join_multicast(ctx->cm_id, (struct sockaddr *) &mc->addr, mc);
+	ret = rdma_join_multicast(ctx->cm_id, (struct sockaddr *)&mc->addr,
+				  join_state, mc);
 	if (ret)
 		goto err2;
 
@@ -1382,7 +1392,7 @@ static ssize_t ucma_join_ip_multicast(struct ucma_file *file,
 	join_cmd.uid = cmd.uid;
 	join_cmd.id = cmd.id;
 	join_cmd.addr_size = rdma_addr_size((struct sockaddr *) &cmd.addr);
-	join_cmd.reserved = 0;
+	join_cmd.join_flags = RDMA_MC_JOIN_FLAG_FULLMEMBER;
 	memcpy(&join_cmd.addr, &cmd.addr, join_cmd.addr_size);
 
 	return ucma_process_join(file, &join_cmd, out_len);

drivers/infiniband/core/uverbs.h

@@ -116,6 +116,7 @@ struct ib_uverbs_event_file {
 struct ib_uverbs_file {
 	struct kref				ref;
 	struct mutex				mutex;
+	struct mutex                            cleanup_mutex; /* protect cleanup */
 	struct ib_uverbs_device		       *device;
 	struct ib_ucontext		       *ucontext;
 	struct ib_event_handler			event_handler;

drivers/infiniband/core/uverbs_main.c

@@ -969,6 +969,7 @@ static int ib_uverbs_open(struct inode *inode, struct file *filp)
 	file->async_file = NULL;
 	kref_init(&file->ref);
 	mutex_init(&file->mutex);
+	mutex_init(&file->cleanup_mutex);
 
 	filp->private_data = file;
 	kobject_get(&dev->kobj);
@@ -994,18 +995,20 @@ static int ib_uverbs_close(struct inode *inode, struct file *filp)
 {
 	struct ib_uverbs_file *file = filp->private_data;
 	struct ib_uverbs_device *dev = file->device;
-	struct ib_ucontext *ucontext = NULL;
 
-	mutex_lock(&file->device->lists_mutex);
-	ucontext = file->ucontext;
+	mutex_lock(&file->cleanup_mutex);
+	if (file->ucontext) {
+		ib_uverbs_cleanup_ucontext(file, file->ucontext);
 		file->ucontext = NULL;
+	}
+	mutex_unlock(&file->cleanup_mutex);
+
+	mutex_lock(&file->device->lists_mutex);
 	if (!file->is_closed) {
 		list_del(&file->list);
 		file->is_closed = 1;
 	}
 	mutex_unlock(&file->device->lists_mutex);
-	if (ucontext)
-		ib_uverbs_cleanup_ucontext(file, ucontext);
 
 	if (file->async_file)
 		kref_put(&file->async_file->ref, ib_uverbs_release_event_file);
@@ -1219,22 +1222,30 @@ static void ib_uverbs_free_hw_resources(struct ib_uverbs_device *uverbs_dev,
 	mutex_lock(&uverbs_dev->lists_mutex);
 	while (!list_empty(&uverbs_dev->uverbs_file_list)) {
 		struct ib_ucontext *ucontext;
-
 		file = list_first_entry(&uverbs_dev->uverbs_file_list,
 					struct ib_uverbs_file, list);
 		file->is_closed = 1;
-		ucontext = file->ucontext;
 		list_del(&file->list);
-		file->ucontext = NULL;
 		kref_get(&file->ref);
 		mutex_unlock(&uverbs_dev->lists_mutex);
-		/* We must release the mutex before going ahead and calling
-		 * disassociate_ucontext. disassociate_ucontext might end up
-		 * indirectly calling uverbs_close, for example due to freeing
-		 * the resources (e.g mmput).
-		 */
+
 		ib_uverbs_event_handler(&file->event_handler, &event);
+
+		mutex_lock(&file->cleanup_mutex);
+		ucontext = file->ucontext;
+		file->ucontext = NULL;
+		mutex_unlock(&file->cleanup_mutex);
+
+		/* At this point ib_uverbs_close cannot be running
+		 * ib_uverbs_cleanup_ucontext
+		 */
 		if (ucontext) {
+			/* We must release the mutex before going ahead and
+			 * calling disassociate_ucontext. disassociate_ucontext
+			 * might end up indirectly calling uverbs_close,
+			 * for example due to freeing the resources
+			 * (e.g mmput).
+			 */
 			ib_dev->disassociate_ucontext(ucontext);
 			ib_uverbs_cleanup_ucontext(file, ucontext);
 		}

drivers/infiniband/hw/hfi1/Kconfig

@@ -3,7 +3,6 @@ config INFINIBAND_HFI1
 	depends on X86_64 && INFINIBAND_RDMAVT
 	select MMU_NOTIFIER
 	select CRC32
-	default m
 	---help---
 	This is a low-level driver for Intel OPA Gen1 adapter.
 config HFI1_DEBUG_SDMA_ORDER

drivers/infiniband/hw/hfi1/file_ops.c

@@ -225,7 +225,7 @@ static long hfi1_file_ioctl(struct file *fp, unsigned int cmd,
 				    sizeof(struct hfi1_base_info));
 		break;
 	case HFI1_IOCTL_CREDIT_UPD:
-		if (uctxt && uctxt->sc)
+		if (uctxt)
 			sc_return_credits(uctxt->sc);
 		break;
 

drivers/infiniband/hw/mlx4/cq.c

@@ -288,7 +288,7 @@ static int mlx4_alloc_resize_buf(struct mlx4_ib_dev *dev, struct mlx4_ib_cq *cq,
 	if (cq->resize_buf)
 		return -EBUSY;
 
-	cq->resize_buf = kmalloc(sizeof *cq->resize_buf, GFP_ATOMIC);
+	cq->resize_buf = kmalloc(sizeof *cq->resize_buf, GFP_KERNEL);
 	if (!cq->resize_buf)
 		return -ENOMEM;
 
@@ -316,7 +316,7 @@ static int mlx4_alloc_resize_umem(struct mlx4_ib_dev *dev, struct mlx4_ib_cq *cq
 	if (ib_copy_from_udata(&ucmd, udata, sizeof ucmd))
 		return -EFAULT;
 
-	cq->resize_buf = kmalloc(sizeof *cq->resize_buf, GFP_ATOMIC);
+	cq->resize_buf = kmalloc(sizeof *cq->resize_buf, GFP_KERNEL);
 	if (!cq->resize_buf)
 		return -ENOMEM;
 

drivers/infiniband/hw/mlx4/main.c

@@ -2049,6 +2049,195 @@ static struct device_attribute *mlx4_class_attributes[] = {
 	&dev_attr_board_id
 };
 
+struct diag_counter {
+	const char *name;
+	u32 offset;
+};
+
+#define DIAG_COUNTER(_name, _offset)			\
+	{ .name = #_name, .offset = _offset }
+
+static const struct diag_counter diag_basic[] = {
+	DIAG_COUNTER(rq_num_lle, 0x00),
+	DIAG_COUNTER(sq_num_lle, 0x04),
+	DIAG_COUNTER(rq_num_lqpoe, 0x08),
+	DIAG_COUNTER(sq_num_lqpoe, 0x0C),
+	DIAG_COUNTER(rq_num_lpe, 0x18),
+	DIAG_COUNTER(sq_num_lpe, 0x1C),
+	DIAG_COUNTER(rq_num_wrfe, 0x20),
+	DIAG_COUNTER(sq_num_wrfe, 0x24),
+	DIAG_COUNTER(sq_num_mwbe, 0x2C),
+	DIAG_COUNTER(sq_num_bre, 0x34),
+	DIAG_COUNTER(sq_num_rire, 0x44),
+	DIAG_COUNTER(rq_num_rire, 0x48),
+	DIAG_COUNTER(sq_num_rae, 0x4C),
+	DIAG_COUNTER(rq_num_rae, 0x50),
+	DIAG_COUNTER(sq_num_roe, 0x54),
+	DIAG_COUNTER(sq_num_tree, 0x5C),
+	DIAG_COUNTER(sq_num_rree, 0x64),
+	DIAG_COUNTER(rq_num_rnr, 0x68),
+	DIAG_COUNTER(sq_num_rnr, 0x6C),
+	DIAG_COUNTER(rq_num_oos, 0x100),
+	DIAG_COUNTER(sq_num_oos, 0x104),
+};
+
+static const struct diag_counter diag_ext[] = {
+	DIAG_COUNTER(rq_num_dup, 0x130),
+	DIAG_COUNTER(sq_num_to, 0x134),
+};
+
+static const struct diag_counter diag_device_only[] = {
+	DIAG_COUNTER(num_cqovf, 0x1A0),
+	DIAG_COUNTER(rq_num_udsdprd, 0x118),
+};
+
+static struct rdma_hw_stats *mlx4_ib_alloc_hw_stats(struct ib_device *ibdev,
+						    u8 port_num)
+{
+	struct mlx4_ib_dev *dev = to_mdev(ibdev);
+	struct mlx4_ib_diag_counters *diag = dev->diag_counters;
+
+	if (!diag[!!port_num].name)
+		return NULL;
+
+	return rdma_alloc_hw_stats_struct(diag[!!port_num].name,
+					  diag[!!port_num].num_counters,
+					  RDMA_HW_STATS_DEFAULT_LIFESPAN);
+}
+
+static int mlx4_ib_get_hw_stats(struct ib_device *ibdev,
+				struct rdma_hw_stats *stats,
+				u8 port, int index)
+{
+	struct mlx4_ib_dev *dev = to_mdev(ibdev);
+	struct mlx4_ib_diag_counters *diag = dev->diag_counters;
+	u32 hw_value[ARRAY_SIZE(diag_device_only) +
+		ARRAY_SIZE(diag_ext) + ARRAY_SIZE(diag_basic)] = {};
+	int ret;
+	int i;
+
+	ret = mlx4_query_diag_counters(dev->dev,
+				       MLX4_OP_MOD_QUERY_TRANSPORT_CI_ERRORS,
+				       diag[!!port].offset, hw_value,
+				       diag[!!port].num_counters, port);
+
+	if (ret)
+		return ret;
+
+	for (i = 0; i < diag[!!port].num_counters; i++)
+		stats->value[i] = hw_value[i];
+
+	return diag[!!port].num_counters;
+}
+
+static int __mlx4_ib_alloc_diag_counters(struct mlx4_ib_dev *ibdev,
+					 const char ***name,
+					 u32 **offset,
+					 u32 *num,
+					 bool port)
+{
+	u32 num_counters;
+
+	num_counters = ARRAY_SIZE(diag_basic);
+
+	if (ibdev->dev->caps.flags2 & MLX4_DEV_CAP_FLAG2_DIAG_PER_PORT)
+		num_counters += ARRAY_SIZE(diag_ext);
+
+	if (!port)
+		num_counters += ARRAY_SIZE(diag_device_only);
+
+	*name = kcalloc(num_counters, sizeof(**name), GFP_KERNEL);
+	if (!*name)
+		return -ENOMEM;
+
+	*offset = kcalloc(num_counters, sizeof(**offset), GFP_KERNEL);
+	if (!*offset)
+		goto err_name;
+
+	*num = num_counters;
+
+	return 0;
+
+err_name:
+	kfree(*name);
+	return -ENOMEM;
+}
+
+static void mlx4_ib_fill_diag_counters(struct mlx4_ib_dev *ibdev,
+				       const char **name,
+				       u32 *offset,
+				       bool port)
+{
+	int i;
+	int j;
+
+	for (i = 0, j = 0; i < ARRAY_SIZE(diag_basic); i++, j++) {
+		name[i] = diag_basic[i].name;
+		offset[i] = diag_basic[i].offset;
+	}
+
+	if (ibdev->dev->caps.flags2 & MLX4_DEV_CAP_FLAG2_DIAG_PER_PORT) {
+		for (i = 0; i < ARRAY_SIZE(diag_ext); i++, j++) {
+			name[j] = diag_ext[i].name;
+			offset[j] = diag_ext[i].offset;
+		}
+	}
+
+	if (!port) {
+		for (i = 0; i < ARRAY_SIZE(diag_device_only); i++, j++) {
+			name[j] = diag_device_only[i].name;
+			offset[j] = diag_device_only[i].offset;
+		}
+	}
+}
+
+static int mlx4_ib_alloc_diag_counters(struct mlx4_ib_dev *ibdev)
+{
+	struct mlx4_ib_diag_counters *diag = ibdev->diag_counters;
+	int i;
+	int ret;
+	bool per_port = !!(ibdev->dev->caps.flags2 &
+		MLX4_DEV_CAP_FLAG2_DIAG_PER_PORT);
+
+	for (i = 0; i < MLX4_DIAG_COUNTERS_TYPES; i++) {
+		/* i == 1 means we are building port counters */
+		if (i && !per_port)
+			continue;
+
+		ret = __mlx4_ib_alloc_diag_counters(ibdev, &diag[i].name,
+						    &diag[i].offset,
+						    &diag[i].num_counters, i);
+		if (ret)
+			goto err_alloc;
+
+		mlx4_ib_fill_diag_counters(ibdev, diag[i].name,
+					   diag[i].offset, i);
+	}
+
+	ibdev->ib_dev.get_hw_stats	= mlx4_ib_get_hw_stats;
+	ibdev->ib_dev.alloc_hw_stats	= mlx4_ib_alloc_hw_stats;
+
+	return 0;
+
+err_alloc:
+	if (i) {
+		kfree(diag[i - 1].name);
+		kfree(diag[i - 1].offset);
+	}
+
+	return ret;
+}
+
+static void mlx4_ib_diag_cleanup(struct mlx4_ib_dev *ibdev)
+{
+	int i;
+
+	for (i = 0; i < MLX4_DIAG_COUNTERS_TYPES; i++) {
+		kfree(ibdev->diag_counters[i].offset);
+		kfree(ibdev->diag_counters[i].name);
+	}
+}
+
 #define MLX4_IB_INVALID_MAC	((u64)-1)
 static void mlx4_ib_update_qps(struct mlx4_ib_dev *ibdev,
 			       struct net_device *dev,
@@ -2552,9 +2741,12 @@ static void *mlx4_ib_add(struct mlx4_dev *dev)
 	for (j = 1; j <= ibdev->dev->caps.num_ports; j++)
 		atomic64_set(&iboe->mac[j - 1], ibdev->dev->caps.def_mac[j]);
 
-	if (ib_register_device(&ibdev->ib_dev, NULL))
+	if (mlx4_ib_alloc_diag_counters(ibdev))
 		goto err_steer_free_bitmap;
 
+	if (ib_register_device(&ibdev->ib_dev, NULL))
+		goto err_diag_counters;
+
 	if (mlx4_ib_mad_init(ibdev))
 		goto err_reg;
 
@@ -2620,6 +2812,9 @@ static void *mlx4_ib_add(struct mlx4_dev *dev)
 err_reg:
 	ib_unregister_device(&ibdev->ib_dev);
 
+err_diag_counters:
+	mlx4_ib_diag_cleanup(ibdev);
+
 err_steer_free_bitmap:
 	kfree(ibdev->ib_uc_qpns_bitmap);
 
@@ -2723,6 +2918,7 @@ static void mlx4_ib_remove(struct mlx4_dev *dev, void *ibdev_ptr)
 	mlx4_ib_close_sriov(ibdev);
 	mlx4_ib_mad_cleanup(ibdev);
 	ib_unregister_device(&ibdev->ib_dev);
+	mlx4_ib_diag_cleanup(ibdev);
 	if (ibdev->iboe.nb.notifier_call) {
 		if (unregister_netdevice_notifier(&ibdev->iboe.nb))
 			pr_warn("failure unregistering notifier\n");

drivers/infiniband/hw/mlx4/mlx4_ib.h

@@ -549,6 +549,14 @@ struct mlx4_ib_counters {
 	u32			default_counter;
 };
 
+#define MLX4_DIAG_COUNTERS_TYPES 2
+
+struct mlx4_ib_diag_counters {
+	const char **name;
+	u32 *offset;
+	u32 num_counters;
+};
+
 struct mlx4_ib_dev {
 	struct ib_device	ib_dev;
 	struct mlx4_dev	       *dev;
@@ -585,6 +593,7 @@ struct mlx4_ib_dev {
 	/* protect resources needed as part of reset flow */
 	spinlock_t		reset_flow_resource_lock;
 	struct list_head		qp_list;
+	struct mlx4_ib_diag_counters diag_counters[MLX4_DIAG_COUNTERS_TYPES];
 };
 
 struct ib_event_work {

drivers/infiniband/hw/mthca/mthca_reset.c

@@ -98,7 +98,7 @@ int mthca_reset(struct mthca_dev *mdev)
 		err = -ENOMEM;
 		mthca_err(mdev, "Couldn't allocate memory to save HCA "
 			  "PCI header, aborting.\n");
-		goto out;
+		goto put_dev;
 	}
 
 	for (i = 0; i < 64; ++i) {
@@ -108,7 +108,7 @@ int mthca_reset(struct mthca_dev *mdev)
 			err = -ENODEV;
 			mthca_err(mdev, "Couldn't save HCA "
 				  "PCI header, aborting.\n");
-			goto out;
+			goto free_hca;
 		}
 	}
 
@@ -121,7 +121,7 @@ int mthca_reset(struct mthca_dev *mdev)
 			err = -ENOMEM;
 			mthca_err(mdev, "Couldn't allocate memory to save HCA "
 				  "bridge PCI header, aborting.\n");
-			goto out;
+			goto free_hca;
 		}
 
 		for (i = 0; i < 64; ++i) {
@@ -131,7 +131,7 @@ int mthca_reset(struct mthca_dev *mdev)
 				err = -ENODEV;
 				mthca_err(mdev, "Couldn't save HCA bridge "
 					  "PCI header, aborting.\n");
-				goto out;
+				goto free_bh;
 			}
 		}
 		bridge_pcix_cap = pci_find_capability(bridge, PCI_CAP_ID_PCIX);
@@ -139,7 +139,7 @@ int mthca_reset(struct mthca_dev *mdev)
 				err = -ENODEV;
 				mthca_err(mdev, "Couldn't locate HCA bridge "
 					  "PCI-X capability, aborting.\n");
-				goto out;
+				goto free_bh;
 		}
 	}
 
@@ -152,7 +152,7 @@ int mthca_reset(struct mthca_dev *mdev)
 			err = -ENOMEM;
 			mthca_err(mdev, "Couldn't map HCA reset register, "
 				  "aborting.\n");
-			goto out;
+			goto free_bh;
 		}
 
 		writel(MTHCA_RESET_VALUE, reset);
@@ -172,7 +172,7 @@ int mthca_reset(struct mthca_dev *mdev)
 				err = -ENODEV;
 				mthca_err(mdev, "Couldn't access HCA after reset, "
 					  "aborting.\n");
-				goto out;
+				goto free_bh;
 			}
 
 			if (v != 0xffffffff)
@@ -184,7 +184,7 @@ int mthca_reset(struct mthca_dev *mdev)
 		err = -ENODEV;
 		mthca_err(mdev, "PCI device did not come back after reset, "
 			  "aborting.\n");
-		goto out;
+		goto free_bh;
 	}
 
 good:
@@ -195,14 +195,14 @@ int mthca_reset(struct mthca_dev *mdev)
 			err = -ENODEV;
 			mthca_err(mdev, "Couldn't restore HCA bridge Upstream "
 				  "split transaction control, aborting.\n");
-			goto out;
+			goto free_bh;
 		}
 		if (pci_write_config_dword(bridge, bridge_pcix_cap + 0xc,
 				 bridge_header[(bridge_pcix_cap + 0xc) / 4])) {
 			err = -ENODEV;
 			mthca_err(mdev, "Couldn't restore HCA bridge Downstream "
 				  "split transaction control, aborting.\n");
-			goto out;
+			goto free_bh;
 		}
 		/*
 		 * Bridge control register is at 0x3e, so we'll
@@ -216,7 +216,7 @@ int mthca_reset(struct mthca_dev *mdev)
 				err = -ENODEV;
 				mthca_err(mdev, "Couldn't restore HCA bridge reg %x, "
 					  "aborting.\n", i);
-				goto out;
+				goto free_bh;
 			}
 		}
 
@@ -225,7 +225,7 @@ int mthca_reset(struct mthca_dev *mdev)
 			err = -ENODEV;
 			mthca_err(mdev, "Couldn't restore HCA bridge COMMAND, "
 				  "aborting.\n");
-			goto out;
+			goto free_bh;
 		}
 	}
 
@@ -235,7 +235,7 @@ int mthca_reset(struct mthca_dev *mdev)
 			err = -ENODEV;
 			mthca_err(mdev, "Couldn't restore HCA PCI-X "
 				  "command register, aborting.\n");
-			goto out;
+			goto free_bh;
 		}
 	}
 
@@ -246,7 +246,7 @@ int mthca_reset(struct mthca_dev *mdev)
 			err = -ENODEV;
 			mthca_err(mdev, "Couldn't restore HCA PCI Express "
 				  "Device Control register, aborting.\n");
-			goto out;
+			goto free_bh;
 		}
 		linkctl = hca_header[(hca_pcie_cap + PCI_EXP_LNKCTL) / 4];
 		if (pcie_capability_write_word(mdev->pdev, PCI_EXP_LNKCTL,
@@ -254,7 +254,7 @@ int mthca_reset(struct mthca_dev *mdev)
 			err = -ENODEV;
 			mthca_err(mdev, "Couldn't restore HCA PCI Express "
 				  "Link control register, aborting.\n");
-			goto out;
+			goto free_bh;
 		}
 	}
 
@@ -266,7 +266,7 @@ int mthca_reset(struct mthca_dev *mdev)
 			err = -ENODEV;
 			mthca_err(mdev, "Couldn't restore HCA reg %x, "
 				  "aborting.\n", i);
-			goto out;
+			goto free_bh;
 		}
 	}
 
@@ -275,14 +275,12 @@ int mthca_reset(struct mthca_dev *mdev)
 		err = -ENODEV;
 		mthca_err(mdev, "Couldn't restore HCA COMMAND, "
 			  "aborting.\n");
-		goto out;
 	}
-
-out:
-	if (bridge)
-		pci_dev_put(bridge);
+free_bh:
 	kfree(bridge_header);
+free_hca:
 	kfree(hca_header);
-
+put_dev:
+	pci_dev_put(bridge);
 	return err;
 }

drivers/infiniband/sw/Makefile

 obj-$(CONFIG_INFINIBAND_RDMAVT)		+= rdmavt/
+obj-$(CONFIG_RDMA_RXE)			+= rxe/

drivers/infiniband/sw/rdmavt/Kconfig

 config INFINIBAND_RDMAVT
 	tristate "RDMA verbs transport library"
 	depends on 64BIT
-	default m
 	---help---
 	This is a common software verbs provider for RDMA networks.

drivers/infiniband/sw/rxe/Kconfig

+config RDMA_RXE
+	tristate "Software RDMA over Ethernet (RoCE) driver"
+	depends on INET && PCI && INFINIBAND
+	depends on NET_UDP_TUNNEL
+	---help---
+	This driver implements the InfiniBand RDMA transport over
+	the Linux network stack. It enables a system with a
+	standard Ethernet adapter to interoperate with a RoCE
+	adapter or with another system running the RXE driver.
+	Documentation on InfiniBand and RoCE can be downloaded at
+	www.infinibandta.org and www.openfabrics.org. (See also
+	siw which is a similar software driver for iWARP.)
+
+	The driver is split into two layers, one interfaces with the
+	Linux RDMA stack and implements a kernel or user space
+	verbs API. The user space verbs API requires a support
+	library named librxe which is loaded by the generic user
+	space verbs API, libibverbs. The other layer interfaces
+	with the Linux network stack at layer 3.
+
+	To configure and work with soft-RoCE driver please use the
+	following wiki page under "configure Soft-RoCE (RXE)" section:
+
+	https://github.com/SoftRoCE/rxe-dev/wiki/rxe-dev:-Home

drivers/infiniband/sw/rxe/Makefile

+obj-$(CONFIG_RDMA_RXE) += rdma_rxe.o
+
+rdma_rxe-y := \
+	rxe.o \
+	rxe_comp.o \
+	rxe_req.o \
+	rxe_resp.o \
+	rxe_recv.o \
+	rxe_pool.o \
+	rxe_queue.o \
+	rxe_verbs.o \
+	rxe_av.o \
+	rxe_srq.o \
+	rxe_qp.o \
+	rxe_cq.o \
+	rxe_mr.o \
+	rxe_dma.o \
+	rxe_opcode.o \
+	rxe_mmap.o \
+	rxe_icrc.o \
+	rxe_mcast.o \
+	rxe_task.o \
+	rxe_net.o \
+	rxe_sysfs.o

drivers/infiniband/sw/rxe/rxe.c

+/*
+ * Copyright (c) 2016 Mellanox Technologies Ltd. All rights reserved.
+ * Copyright (c) 2015 System Fabric Works, 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.
+ */
+
+#include "rxe.h"
+#include "rxe_loc.h"
+
+MODULE_AUTHOR("Bob Pearson, Frank Zago, John Groves, Kamal Heib");
+MODULE_DESCRIPTION("Soft RDMA transport");
+MODULE_LICENSE("Dual BSD/GPL");
+MODULE_VERSION("0.2");
+
+/* free resources for all ports on a device */
+static void rxe_cleanup_ports(struct rxe_dev *rxe)
+{
+	kfree(rxe->port.pkey_tbl);
+	rxe->port.pkey_tbl = NULL;
+
+}
+
+/* free resources for a rxe device all objects created for this device must
+ * have been destroyed
+ */
+static void rxe_cleanup(struct rxe_dev *rxe)
+{
+	rxe_pool_cleanup(&rxe->uc_pool);
+	rxe_pool_cleanup(&rxe->pd_pool);
+	rxe_pool_cleanup(&rxe->ah_pool);
+	rxe_pool_cleanup(&rxe->srq_pool);
+	rxe_pool_cleanup(&rxe->qp_pool);
+	rxe_pool_cleanup(&rxe->cq_pool);
+	rxe_pool_cleanup(&rxe->mr_pool);
+	rxe_pool_cleanup(&rxe->mw_pool);
+	rxe_pool_cleanup(&rxe->mc_grp_pool);
+	rxe_pool_cleanup(&rxe->mc_elem_pool);
+
+	rxe_cleanup_ports(rxe);
+}
+
+/* called when all references have been dropped */
+void rxe_release(struct kref *kref)
+{
+	struct rxe_dev *rxe = container_of(kref, struct rxe_dev, ref_cnt);
+
+	rxe_cleanup(rxe);
+	ib_dealloc_device(&rxe->ib_dev);
+}
+
+void rxe_dev_put(struct rxe_dev *rxe)
+{
+	kref_put(&rxe->ref_cnt, rxe_release);
+}
+EXPORT_SYMBOL_GPL(rxe_dev_put);
+
+/* initialize rxe device parameters */
+static int rxe_init_device_param(struct rxe_dev *rxe)
+{
+	rxe->max_inline_data			= RXE_MAX_INLINE_DATA;
+
+	rxe->attr.fw_ver			= RXE_FW_VER;
+	rxe->attr.max_mr_size			= RXE_MAX_MR_SIZE;
+	rxe->attr.page_size_cap			= RXE_PAGE_SIZE_CAP;
+	rxe->attr.vendor_id			= RXE_VENDOR_ID;
+	rxe->attr.vendor_part_id		= RXE_VENDOR_PART_ID;
+	rxe->attr.hw_ver			= RXE_HW_VER;
+	rxe->attr.max_qp			= RXE_MAX_QP;
+	rxe->attr.max_qp_wr			= RXE_MAX_QP_WR;
+	rxe->attr.device_cap_flags		= RXE_DEVICE_CAP_FLAGS;
+	rxe->attr.max_sge			= RXE_MAX_SGE;
+	rxe->attr.max_sge_rd			= RXE_MAX_SGE_RD;
+	rxe->attr.max_cq			= RXE_MAX_CQ;
+	rxe->attr.max_cqe			= (1 << RXE_MAX_LOG_CQE) - 1;
+	rxe->attr.max_mr			= RXE_MAX_MR;
+	rxe->attr.max_pd			= RXE_MAX_PD;
+	rxe->attr.max_qp_rd_atom		= RXE_MAX_QP_RD_ATOM;
+	rxe->attr.max_ee_rd_atom		= RXE_MAX_EE_RD_ATOM;
+	rxe->attr.max_res_rd_atom		= RXE_MAX_RES_RD_ATOM;
+	rxe->attr.max_qp_init_rd_atom		= RXE_MAX_QP_INIT_RD_ATOM;
+	rxe->attr.max_ee_init_rd_atom		= RXE_MAX_EE_INIT_RD_ATOM;
+	rxe->attr.atomic_cap			= RXE_ATOMIC_CAP;
+	rxe->attr.max_ee			= RXE_MAX_EE;
+	rxe->attr.max_rdd			= RXE_MAX_RDD;
+	rxe->attr.max_mw			= RXE_MAX_MW;
+	rxe->attr.max_raw_ipv6_qp		= RXE_MAX_RAW_IPV6_QP;
+	rxe->attr.max_raw_ethy_qp		= RXE_MAX_RAW_ETHY_QP;
+	rxe->attr.max_mcast_grp			= RXE_MAX_MCAST_GRP;
+	rxe->attr.max_mcast_qp_attach		= RXE_MAX_MCAST_QP_ATTACH;
+	rxe->attr.max_total_mcast_qp_attach	= RXE_MAX_TOT_MCAST_QP_ATTACH;
+	rxe->attr.max_ah			= RXE_MAX_AH;
+	rxe->attr.max_fmr			= RXE_MAX_FMR;
+	rxe->attr.max_map_per_fmr		= RXE_MAX_MAP_PER_FMR;
+	rxe->attr.max_srq			= RXE_MAX_SRQ;
+	rxe->attr.max_srq_wr			= RXE_MAX_SRQ_WR;
+	rxe->attr.max_srq_sge			= RXE_MAX_SRQ_SGE;
+	rxe->attr.max_fast_reg_page_list_len	= RXE_MAX_FMR_PAGE_LIST_LEN;
+	rxe->attr.max_pkeys			= RXE_MAX_PKEYS;
+	rxe->attr.local_ca_ack_delay		= RXE_LOCAL_CA_ACK_DELAY;
+
+	rxe->max_ucontext			= RXE_MAX_UCONTEXT;
+
+	return 0;
+}
+
+/* initialize port attributes */
+static int rxe_init_port_param(struct rxe_port *port)
+{
+	port->attr.state		= RXE_PORT_STATE;
+	port->attr.max_mtu		= RXE_PORT_MAX_MTU;
+	port->attr.active_mtu		= RXE_PORT_ACTIVE_MTU;
+	port->attr.gid_tbl_len		= RXE_PORT_GID_TBL_LEN;
+	port->attr.port_cap_flags	= RXE_PORT_PORT_CAP_FLAGS;
+	port->attr.max_msg_sz		= RXE_PORT_MAX_MSG_SZ;
+	port->attr.bad_pkey_cntr	= RXE_PORT_BAD_PKEY_CNTR;
+	port->attr.qkey_viol_cntr	= RXE_PORT_QKEY_VIOL_CNTR;
+	port->attr.pkey_tbl_len		= RXE_PORT_PKEY_TBL_LEN;
+	port->attr.lid			= RXE_PORT_LID;
+	port->attr.sm_lid		= RXE_PORT_SM_LID;
+	port->attr.lmc			= RXE_PORT_LMC;
+	port->attr.max_vl_num		= RXE_PORT_MAX_VL_NUM;
+	port->attr.sm_sl		= RXE_PORT_SM_SL;
+	port->attr.subnet_timeout	= RXE_PORT_SUBNET_TIMEOUT;
+	port->attr.init_type_reply	= RXE_PORT_INIT_TYPE_REPLY;
+	port->attr.active_width		= RXE_PORT_ACTIVE_WIDTH;
+	port->attr.active_speed		= RXE_PORT_ACTIVE_SPEED;
+	port->attr.phys_state		= RXE_PORT_PHYS_STATE;
+	port->mtu_cap			=
+				ib_mtu_enum_to_int(RXE_PORT_ACTIVE_MTU);
+	port->subnet_prefix		= cpu_to_be64(RXE_PORT_SUBNET_PREFIX);
+
+	return 0;
+}
+
+/* initialize port state, note IB convention that HCA ports are always
+ * numbered from 1
+ */
+static int rxe_init_ports(struct rxe_dev *rxe)
+{
+	struct rxe_port *port = &rxe->port;
+
+	rxe_init_port_param(port);
+
+	if (!port->attr.pkey_tbl_len || !port->attr.gid_tbl_len)
+		return -EINVAL;
+
+	port->pkey_tbl = kcalloc(port->attr.pkey_tbl_len,
+			sizeof(*port->pkey_tbl), GFP_KERNEL);
+
+	if (!port->pkey_tbl)
+		return -ENOMEM;
+
+	port->pkey_tbl[0] = 0xffff;
+	port->port_guid = rxe->ifc_ops->port_guid(rxe);
+
+	spin_lock_init(&port->port_lock);
+
+	return 0;
+}
+
+/* init pools of managed objects */
+static int rxe_init_pools(struct rxe_dev *rxe)
+{
+	int err;
+
+	err = rxe_pool_init(rxe, &rxe->uc_pool, RXE_TYPE_UC,
+			    rxe->max_ucontext);
+	if (err)
+		goto err1;
+
+	err = rxe_pool_init(rxe, &rxe->pd_pool, RXE_TYPE_PD,
+			    rxe->attr.max_pd);
+	if (err)
+		goto err2;
+
+	err = rxe_pool_init(rxe, &rxe->ah_pool, RXE_TYPE_AH,
+			    rxe->attr.max_ah);
+	if (err)
+		goto err3;
+
+	err = rxe_pool_init(rxe, &rxe->srq_pool, RXE_TYPE_SRQ,
+			    rxe->attr.max_srq);
+	if (err)
+		goto err4;
+
+	err = rxe_pool_init(rxe, &rxe->qp_pool, RXE_TYPE_QP,
+			    rxe->attr.max_qp);
+	if (err)
+		goto err5;
+
+	err = rxe_pool_init(rxe, &rxe->cq_pool, RXE_TYPE_CQ,
+			    rxe->attr.max_cq);
+	if (err)
+		goto err6;
+
+	err = rxe_pool_init(rxe, &rxe->mr_pool, RXE_TYPE_MR,
+			    rxe->attr.max_mr);
+	if (err)
+		goto err7;
+
+	err = rxe_pool_init(rxe, &rxe->mw_pool, RXE_TYPE_MW,
+			    rxe->attr.max_mw);
+	if (err)
+		goto err8;
+
+	err = rxe_pool_init(rxe, &rxe->mc_grp_pool, RXE_TYPE_MC_GRP,
+			    rxe->attr.max_mcast_grp);
+	if (err)
+		goto err9;
+
+	err = rxe_pool_init(rxe, &rxe->mc_elem_pool, RXE_TYPE_MC_ELEM,
+			    rxe->attr.max_total_mcast_qp_attach);
+	if (err)
+		goto err10;
+
+	return 0;
+
+err10:
+	rxe_pool_cleanup(&rxe->mc_grp_pool);
+err9:
+	rxe_pool_cleanup(&rxe->mw_pool);
+err8:
+	rxe_pool_cleanup(&rxe->mr_pool);
+err7:
+	rxe_pool_cleanup(&rxe->cq_pool);
+err6:
+	rxe_pool_cleanup(&rxe->qp_pool);
+err5:
+	rxe_pool_cleanup(&rxe->srq_pool);
+err4:
+	rxe_pool_cleanup(&rxe->ah_pool);
+err3:
+	rxe_pool_cleanup(&rxe->pd_pool);
+err2:
+	rxe_pool_cleanup(&rxe->uc_pool);
+err1:
+	return err;
+}
+
+/* initialize rxe device state */
+static int rxe_init(struct rxe_dev *rxe)
+{
+	int err;
+
+	/* init default device parameters */
+	rxe_init_device_param(rxe);
+
+	err = rxe_init_ports(rxe);
+	if (err)
+		goto err1;
+
+	err = rxe_init_pools(rxe);
+	if (err)
+		goto err2;
+
+	/* init pending mmap list */
+	spin_lock_init(&rxe->mmap_offset_lock);
+	spin_lock_init(&rxe->pending_lock);
+	INIT_LIST_HEAD(&rxe->pending_mmaps);
+	INIT_LIST_HEAD(&rxe->list);
+
+	mutex_init(&rxe->usdev_lock);
+
+	return 0;
+
+err2:
+	rxe_cleanup_ports(rxe);
+err1:
+	return err;
+}
+
+int rxe_set_mtu(struct rxe_dev *rxe, unsigned int ndev_mtu)
+{
+	struct rxe_port *port = &rxe->port;
+	enum ib_mtu mtu;
+
+	mtu = eth_mtu_int_to_enum(ndev_mtu);
+
+	/* Make sure that new MTU in range */
+	mtu = mtu ? min_t(enum ib_mtu, mtu, RXE_PORT_MAX_MTU) : IB_MTU_256;
+
+	port->attr.active_mtu = mtu;
+	port->mtu_cap = ib_mtu_enum_to_int(mtu);
+
+	return 0;
+}
+EXPORT_SYMBOL(rxe_set_mtu);
+
+/* called by ifc layer to create new rxe device.
+ * The caller should allocate memory for rxe by calling ib_alloc_device.
+ */
+int rxe_add(struct rxe_dev *rxe, unsigned int mtu)
+{
+	int err;
+
+	kref_init(&rxe->ref_cnt);
+
+	err = rxe_init(rxe);
+	if (err)
+		goto err1;
+
+	err = rxe_set_mtu(rxe, mtu);
+	if (err)
+		goto err1;
+
+	err = rxe_register_device(rxe);
+	if (err)
+		goto err1;
+
+	return 0;
+
+err1:
+	rxe_dev_put(rxe);
+	return err;
+}
+EXPORT_SYMBOL(rxe_add);
+
+/* called by the ifc layer to remove a device */
+void rxe_remove(struct rxe_dev *rxe)
+{
+	rxe_unregister_device(rxe);
+
+	rxe_dev_put(rxe);
+}
+EXPORT_SYMBOL(rxe_remove);
+
+static int __init rxe_module_init(void)
+{
+	int err;
+
+	/* initialize slab caches for managed objects */
+	err = rxe_cache_init();
+	if (err) {
+		pr_err("rxe: unable to init object pools\n");
+		return err;
+	}
+
+	err = rxe_net_init();
+	if (err) {
+		pr_err("rxe: unable to init\n");
+		rxe_cache_exit();
+		return err;
+	}
+	pr_info("rxe: loaded\n");
+
+	return 0;
+}
+
+static void __exit rxe_module_exit(void)
+{
+	rxe_remove_all();
+	rxe_net_exit();
+	rxe_cache_exit();
+
+	pr_info("rxe: unloaded\n");
+}
+
+module_init(rxe_module_init);
+module_exit(rxe_module_exit);

drivers/infiniband/sw/rxe/rxe.h

+/*
+ * Copyright (c) 2016 Mellanox Technologies Ltd. All rights reserved.
+ * Copyright (c) 2015 System Fabric Works, 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 RXE_H
+#define RXE_H
+
+#include <linux/module.h>
+#include <linux/skbuff.h>
+#include <linux/crc32.h>
+
+#include <rdma/ib_verbs.h>
+#include <rdma/ib_user_verbs.h>
+#include <rdma/ib_pack.h>
+#include <rdma/ib_smi.h>
+#include <rdma/ib_umem.h>
+#include <rdma/ib_cache.h>
+#include <rdma/ib_addr.h>
+
+#include "rxe_net.h"
+#include "rxe_opcode.h"
+#include "rxe_hdr.h"
+#include "rxe_param.h"
+#include "rxe_verbs.h"
+
+#define RXE_UVERBS_ABI_VERSION		(1)
+
+#define IB_PHYS_STATE_LINK_UP		(5)
+#define IB_PHYS_STATE_LINK_DOWN		(3)
+
+#define RXE_ROCE_V2_SPORT		(0xc000)
+
+int rxe_set_mtu(struct rxe_dev *rxe, unsigned int dev_mtu);
+
+int rxe_add(struct rxe_dev *rxe, unsigned int mtu);
+void rxe_remove(struct rxe_dev *rxe);
+void rxe_remove_all(void);
+
+int rxe_rcv(struct sk_buff *skb);
+
+void rxe_dev_put(struct rxe_dev *rxe);
+struct rxe_dev *net_to_rxe(struct net_device *ndev);
+struct rxe_dev *get_rxe_by_name(const char* name);
+
+void rxe_port_up(struct rxe_dev *rxe);
+void rxe_port_down(struct rxe_dev *rxe);
+
+#endif /* RXE_H */

drivers/infiniband/sw/rxe/rxe_av.c

+/*
+ * Copyright (c) 2016 Mellanox Technologies Ltd. All rights reserved.
+ * Copyright (c) 2015 System Fabric Works, 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.
+ */
+
+#include "rxe.h"
+#include "rxe_loc.h"
+
+int rxe_av_chk_attr(struct rxe_dev *rxe, struct ib_ah_attr *attr)
+{
+	struct rxe_port *port;
+
+	if (attr->port_num != 1) {
+		pr_info("rxe: invalid port_num = %d\n", attr->port_num);
+		return -EINVAL;
+	}
+
+	port = &rxe->port;
+
+	if (attr->ah_flags & IB_AH_GRH) {
+		if (attr->grh.sgid_index > port->attr.gid_tbl_len) {
+			pr_info("rxe: invalid sgid index = %d\n",
+				attr->grh.sgid_index);
+			return -EINVAL;
+		}
+	}
+
+	return 0;
+}
+
+int rxe_av_from_attr(struct rxe_dev *rxe, u8 port_num,
+		     struct rxe_av *av, struct ib_ah_attr *attr)
+{
+	memset(av, 0, sizeof(*av));
+	memcpy(&av->grh, &attr->grh, sizeof(attr->grh));
+	av->port_num = port_num;
+	return 0;
+}
+
+int rxe_av_to_attr(struct rxe_dev *rxe, struct rxe_av *av,
+		   struct ib_ah_attr *attr)
+{
+	memcpy(&attr->grh, &av->grh, sizeof(av->grh));
+	attr->port_num = av->port_num;
+	return 0;
+}
+
+int rxe_av_fill_ip_info(struct rxe_dev *rxe,
+			struct rxe_av *av,
+			struct ib_ah_attr *attr,
+			struct ib_gid_attr *sgid_attr,
+			union ib_gid *sgid)
+{
+	rdma_gid2ip(&av->sgid_addr._sockaddr, sgid);
+	rdma_gid2ip(&av->dgid_addr._sockaddr, &attr->grh.dgid);
+	av->network_type = ib_gid_to_network_type(sgid_attr->gid_type, sgid);
+
+	return 0;
+}
+
+struct rxe_av *rxe_get_av(struct rxe_pkt_info *pkt)
+{
+	if (!pkt || !pkt->qp)
+		return NULL;
+
+	if (qp_type(pkt->qp) == IB_QPT_RC || qp_type(pkt->qp) == IB_QPT_UC)
+		return &pkt->qp->pri_av;
+
+	return (pkt->wqe) ? &pkt->wqe->av : NULL;
+}

drivers/infiniband/sw/rxe/rxe_comp.c

+/*
+ * Copyright (c) 2016 Mellanox Technologies Ltd. All rights reserved.
+ * Copyright (c) 2015 System Fabric Works, 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.
+ */
+
+#include <linux/skbuff.h>
+
+#include "rxe.h"
+#include "rxe_loc.h"
+#include "rxe_queue.h"
+#include "rxe_task.h"
+
+enum comp_state {
+	COMPST_GET_ACK,
+	COMPST_GET_WQE,
+	COMPST_COMP_WQE,
+	COMPST_COMP_ACK,
+	COMPST_CHECK_PSN,
+	COMPST_CHECK_ACK,
+	COMPST_READ,
+	COMPST_ATOMIC,
+	COMPST_WRITE_SEND,
+	COMPST_UPDATE_COMP,
+	COMPST_ERROR_RETRY,
+	COMPST_RNR_RETRY,
+	COMPST_ERROR,
+	COMPST_EXIT, /* We have an issue, and we want to rerun the completer */
+	COMPST_DONE, /* The completer finished successflly */
+};
+
+static char *comp_state_name[] =  {
+	[COMPST_GET_ACK]		= "GET ACK",
+	[COMPST_GET_WQE]		= "GET WQE",
+	[COMPST_COMP_WQE]		= "COMP WQE",
+	[COMPST_COMP_ACK]		= "COMP ACK",
+	[COMPST_CHECK_PSN]		= "CHECK PSN",
+	[COMPST_CHECK_ACK]		= "CHECK ACK",
+	[COMPST_READ]			= "READ",
+	[COMPST_ATOMIC]			= "ATOMIC",
+	[COMPST_WRITE_SEND]		= "WRITE/SEND",
+	[COMPST_UPDATE_COMP]		= "UPDATE COMP",
+	[COMPST_ERROR_RETRY]		= "ERROR RETRY",
+	[COMPST_RNR_RETRY]		= "RNR RETRY",
+	[COMPST_ERROR]			= "ERROR",
+	[COMPST_EXIT]			= "EXIT",
+	[COMPST_DONE]			= "DONE",
+};
+
+static unsigned long rnrnak_usec[32] = {
+	[IB_RNR_TIMER_655_36] = 655360,
+	[IB_RNR_TIMER_000_01] = 10,
+	[IB_RNR_TIMER_000_02] = 20,
+	[IB_RNR_TIMER_000_03] = 30,
+	[IB_RNR_TIMER_000_04] = 40,
+	[IB_RNR_TIMER_000_06] = 60,
+	[IB_RNR_TIMER_000_08] = 80,
+	[IB_RNR_TIMER_000_12] = 120,
+	[IB_RNR_TIMER_000_16] = 160,
+	[IB_RNR_TIMER_000_24] = 240,
+	[IB_RNR_TIMER_000_32] = 320,
+	[IB_RNR_TIMER_000_48] = 480,
+	[IB_RNR_TIMER_000_64] = 640,
+	[IB_RNR_TIMER_000_96] = 960,
+	[IB_RNR_TIMER_001_28] = 1280,
+	[IB_RNR_TIMER_001_92] = 1920,
+	[IB_RNR_TIMER_002_56] = 2560,
+	[IB_RNR_TIMER_003_84] = 3840,
+	[IB_RNR_TIMER_005_12] = 5120,
+	[IB_RNR_TIMER_007_68] = 7680,
+	[IB_RNR_TIMER_010_24] = 10240,
+	[IB_RNR_TIMER_015_36] = 15360,
+	[IB_RNR_TIMER_020_48] = 20480,
+	[IB_RNR_TIMER_030_72] = 30720,
+	[IB_RNR_TIMER_040_96] = 40960,
+	[IB_RNR_TIMER_061_44] = 61410,
+	[IB_RNR_TIMER_081_92] = 81920,
+	[IB_RNR_TIMER_122_88] = 122880,
+	[IB_RNR_TIMER_163_84] = 163840,
+	[IB_RNR_TIMER_245_76] = 245760,
+	[IB_RNR_TIMER_327_68] = 327680,
+	[IB_RNR_TIMER_491_52] = 491520,
+};
+
+static inline unsigned long rnrnak_jiffies(u8 timeout)
+{
+	return max_t(unsigned long,
+		usecs_to_jiffies(rnrnak_usec[timeout]), 1);
+}
+
+static enum ib_wc_opcode wr_to_wc_opcode(enum ib_wr_opcode opcode)
+{
+	switch (opcode) {
+	case IB_WR_RDMA_WRITE:			return IB_WC_RDMA_WRITE;
+	case IB_WR_RDMA_WRITE_WITH_IMM:		return IB_WC_RDMA_WRITE;
+	case IB_WR_SEND:			return IB_WC_SEND;
+	case IB_WR_SEND_WITH_IMM:		return IB_WC_SEND;
+	case IB_WR_RDMA_READ:			return IB_WC_RDMA_READ;
+	case IB_WR_ATOMIC_CMP_AND_SWP:		return IB_WC_COMP_SWAP;
+	case IB_WR_ATOMIC_FETCH_AND_ADD:	return IB_WC_FETCH_ADD;
+	case IB_WR_LSO:				return IB_WC_LSO;
+	case IB_WR_SEND_WITH_INV:		return IB_WC_SEND;
+	case IB_WR_RDMA_READ_WITH_INV:		return IB_WC_RDMA_READ;
+	case IB_WR_LOCAL_INV:			return IB_WC_LOCAL_INV;
+	case IB_WR_REG_MR:			return IB_WC_REG_MR;
+
+	default:
+		return 0xff;
+	}
+}
+
+void retransmit_timer(unsigned long data)
+{
+	struct rxe_qp *qp = (struct rxe_qp *)data;
+
+	if (qp->valid) {
+		qp->comp.timeout = 1;
+		rxe_run_task(&qp->comp.task, 1);
+	}
+}
+
+void rxe_comp_queue_pkt(struct rxe_dev *rxe, struct rxe_qp *qp,
+			struct sk_buff *skb)
+{
+	int must_sched;
+
+	skb_queue_tail(&qp->resp_pkts, skb);
+
+	must_sched = skb_queue_len(&qp->resp_pkts) > 1;
+	rxe_run_task(&qp->comp.task, must_sched);
+}
+
+static inline enum comp_state get_wqe(struct rxe_qp *qp,
+				      struct rxe_pkt_info *pkt,
+				      struct rxe_send_wqe **wqe_p)
+{
+	struct rxe_send_wqe *wqe;
+
+	/* we come here whether or not we found a response packet to see if
+	 * there are any posted WQEs
+	 */
+	wqe = queue_head(qp->sq.queue);
+	*wqe_p = wqe;
+
+	/* no WQE or requester has not started it yet */
+	if (!wqe || wqe->state == wqe_state_posted)
+		return pkt ? COMPST_DONE : COMPST_EXIT;
+
+	/* WQE does not require an ack */
+	if (wqe->state == wqe_state_done)
+		return COMPST_COMP_WQE;
+
+	/* WQE caused an error */
+	if (wqe->state == wqe_state_error)
+		return COMPST_ERROR;
+
+	/* we have a WQE, if we also have an ack check its PSN */
+	return pkt ? COMPST_CHECK_PSN : COMPST_EXIT;
+}
+
+static inline void reset_retry_counters(struct rxe_qp *qp)
+{
+	qp->comp.retry_cnt = qp->attr.retry_cnt;
+	qp->comp.rnr_retry = qp->attr.rnr_retry;
+}
+
+static inline enum comp_state check_psn(struct rxe_qp *qp,
+					struct rxe_pkt_info *pkt,
+					struct rxe_send_wqe *wqe)
+{
+	s32 diff;
+
+	/* check to see if response is past the oldest WQE. if it is, complete
+	 * send/write or error read/atomic
+	 */
+	diff = psn_compare(pkt->psn, wqe->last_psn);
+	if (diff > 0) {
+		if (wqe->state == wqe_state_pending) {
+			if (wqe->mask & WR_ATOMIC_OR_READ_MASK)
+				return COMPST_ERROR_RETRY;
+
+			reset_retry_counters(qp);
+			return COMPST_COMP_WQE;
+		} else {
+			return COMPST_DONE;
+		}
+	}
+
+	/* compare response packet to expected response */
+	diff = psn_compare(pkt->psn, qp->comp.psn);
+	if (diff < 0) {
+		/* response is most likely a retried packet if it matches an
+		 * uncompleted WQE go complete it else ignore it
+		 */
+		if (pkt->psn == wqe->last_psn)
+			return COMPST_COMP_ACK;
+		else
+			return COMPST_DONE;
+	} else if ((diff > 0) && (wqe->mask & WR_ATOMIC_OR_READ_MASK)) {
+		return COMPST_ERROR_RETRY;
+	} else {
+		return COMPST_CHECK_ACK;
+	}
+}
+
+static inline enum comp_state check_ack(struct rxe_qp *qp,
+					struct rxe_pkt_info *pkt,
+					struct rxe_send_wqe *wqe)
+{
+	unsigned int mask = pkt->mask;
+	u8 syn;
+
+	/* Check the sequence only */
+	switch (qp->comp.opcode) {
+	case -1:
+		/* Will catch all *_ONLY cases. */
+		if (!(mask & RXE_START_MASK))
+			return COMPST_ERROR;
+
+		break;
+
+	case IB_OPCODE_RC_RDMA_READ_RESPONSE_FIRST:
+	case IB_OPCODE_RC_RDMA_READ_RESPONSE_MIDDLE:
+		if (pkt->opcode != IB_OPCODE_RC_RDMA_READ_RESPONSE_MIDDLE &&
+		    pkt->opcode != IB_OPCODE_RC_RDMA_READ_RESPONSE_LAST) {
+			return COMPST_ERROR;
+		}
+		break;
+	default:
+		WARN_ON(1);
+	}
+
+	/* Check operation validity. */
+	switch (pkt->opcode) {
+	case IB_OPCODE_RC_RDMA_READ_RESPONSE_FIRST:
+	case IB_OPCODE_RC_RDMA_READ_RESPONSE_LAST:
+	case IB_OPCODE_RC_RDMA_READ_RESPONSE_ONLY:
+		syn = aeth_syn(pkt);
+
+		if ((syn & AETH_TYPE_MASK) != AETH_ACK)
+			return COMPST_ERROR;
+
+		/* Fall through (IB_OPCODE_RC_RDMA_READ_RESPONSE_MIDDLE
+		 * doesn't have an AETH)
+		 */
+	case IB_OPCODE_RC_RDMA_READ_RESPONSE_MIDDLE:
+		if (wqe->wr.opcode != IB_WR_RDMA_READ &&
+		    wqe->wr.opcode != IB_WR_RDMA_READ_WITH_INV) {
+			return COMPST_ERROR;
+		}
+		reset_retry_counters(qp);
+		return COMPST_READ;
+
+	case IB_OPCODE_RC_ATOMIC_ACKNOWLEDGE:
+		syn = aeth_syn(pkt);
+
+		if ((syn & AETH_TYPE_MASK) != AETH_ACK)
+			return COMPST_ERROR;
+
+		if (wqe->wr.opcode != IB_WR_ATOMIC_CMP_AND_SWP &&
+		    wqe->wr.opcode != IB_WR_ATOMIC_FETCH_AND_ADD)
+			return COMPST_ERROR;
+		reset_retry_counters(qp);
+		return COMPST_ATOMIC;
+
+	case IB_OPCODE_RC_ACKNOWLEDGE:
+		syn = aeth_syn(pkt);
+		switch (syn & AETH_TYPE_MASK) {
+		case AETH_ACK:
+			reset_retry_counters(qp);
+			return COMPST_WRITE_SEND;
+
+		case AETH_RNR_NAK:
+			return COMPST_RNR_RETRY;
+
+		case AETH_NAK:
+			switch (syn) {
+			case AETH_NAK_PSN_SEQ_ERROR:
+				/* a nak implicitly acks all packets with psns
+				 * before
+				 */
+				if (psn_compare(pkt->psn, qp->comp.psn) > 0) {
+					qp->comp.psn = pkt->psn;
+					if (qp->req.wait_psn) {
+						qp->req.wait_psn = 0;
+						rxe_run_task(&qp->req.task, 1);
+					}
+				}
+				return COMPST_ERROR_RETRY;
+
+			case AETH_NAK_INVALID_REQ:
+				wqe->status = IB_WC_REM_INV_REQ_ERR;
+				return COMPST_ERROR;
+
+			case AETH_NAK_REM_ACC_ERR:
+				wqe->status = IB_WC_REM_ACCESS_ERR;
+				return COMPST_ERROR;
+
+			case AETH_NAK_REM_OP_ERR:
+				wqe->status = IB_WC_REM_OP_ERR;
+				return COMPST_ERROR;
+
+			default:
+				pr_warn("unexpected nak %x\n", syn);
+				wqe->status = IB_WC_REM_OP_ERR;
+				return COMPST_ERROR;
+			}
+
+		default:
+			return COMPST_ERROR;
+		}
+		break;
+
+	default:
+		pr_warn("unexpected opcode\n");
+	}
+
+	return COMPST_ERROR;
+}
+
+static inline enum comp_state do_read(struct rxe_qp *qp,
+				      struct rxe_pkt_info *pkt,
+				      struct rxe_send_wqe *wqe)
+{
+	struct rxe_dev *rxe = to_rdev(qp->ibqp.device);
+	int ret;
+
+	ret = copy_data(rxe, qp->pd, IB_ACCESS_LOCAL_WRITE,
+			&wqe->dma, payload_addr(pkt),
+			payload_size(pkt), to_mem_obj, NULL);
+	if (ret)
+		return COMPST_ERROR;
+
+	if (wqe->dma.resid == 0 && (pkt->mask & RXE_END_MASK))
+		return COMPST_COMP_ACK;
+	else
+		return COMPST_UPDATE_COMP;
+}
+
+static inline enum comp_state do_atomic(struct rxe_qp *qp,
+					struct rxe_pkt_info *pkt,
+					struct rxe_send_wqe *wqe)
+{
+	struct rxe_dev *rxe = to_rdev(qp->ibqp.device);
+	int ret;
+
+	u64 atomic_orig = atmack_orig(pkt);
+
+	ret = copy_data(rxe, qp->pd, IB_ACCESS_LOCAL_WRITE,
+			&wqe->dma, &atomic_orig,
+			sizeof(u64), to_mem_obj, NULL);
+	if (ret)
+		return COMPST_ERROR;
+	else
+		return COMPST_COMP_ACK;
+}
+
+static void make_send_cqe(struct rxe_qp *qp, struct rxe_send_wqe *wqe,
+			  struct rxe_cqe *cqe)
+{
+	memset(cqe, 0, sizeof(*cqe));
+
+	if (!qp->is_user) {
+		struct ib_wc		*wc	= &cqe->ibwc;
+
+		wc->wr_id		= wqe->wr.wr_id;
+		wc->status		= wqe->status;
+		wc->opcode		= wr_to_wc_opcode(wqe->wr.opcode);
+		if (wqe->wr.opcode == IB_WR_RDMA_WRITE_WITH_IMM ||
+		    wqe->wr.opcode == IB_WR_SEND_WITH_IMM)
+			wc->wc_flags = IB_WC_WITH_IMM;
+		wc->byte_len		= wqe->dma.length;
+		wc->qp			= &qp->ibqp;
+	} else {
+		struct ib_uverbs_wc	*uwc	= &cqe->uibwc;
+
+		uwc->wr_id		= wqe->wr.wr_id;
+		uwc->status		= wqe->status;
+		uwc->opcode		= wr_to_wc_opcode(wqe->wr.opcode);
+		if (wqe->wr.opcode == IB_WR_RDMA_WRITE_WITH_IMM ||
+		    wqe->wr.opcode == IB_WR_SEND_WITH_IMM)
+			uwc->wc_flags = IB_WC_WITH_IMM;
+		uwc->byte_len		= wqe->dma.length;
+		uwc->qp_num		= qp->ibqp.qp_num;
+	}
+}
+
+static void do_complete(struct rxe_qp *qp, struct rxe_send_wqe *wqe)
+{
+	struct rxe_cqe cqe;
+
+	if ((qp->sq_sig_type == IB_SIGNAL_ALL_WR) ||
+	    (wqe->wr.send_flags & IB_SEND_SIGNALED) ||
+	    (qp->req.state == QP_STATE_ERROR)) {
+		make_send_cqe(qp, wqe, &cqe);
+		rxe_cq_post(qp->scq, &cqe, 0);
+	}
+
+	advance_consumer(qp->sq.queue);
+
+	/*
+	 * we completed something so let req run again
+	 * if it is trying to fence
+	 */
+	if (qp->req.wait_fence) {
+		qp->req.wait_fence = 0;
+		rxe_run_task(&qp->req.task, 1);
+	}
+}
+
+static inline enum comp_state complete_ack(struct rxe_qp *qp,
+					   struct rxe_pkt_info *pkt,
+					   struct rxe_send_wqe *wqe)
+{
+	unsigned long flags;
+
+	if (wqe->has_rd_atomic) {
+		wqe->has_rd_atomic = 0;
+		atomic_inc(&qp->req.rd_atomic);
+		if (qp->req.need_rd_atomic) {
+			qp->comp.timeout_retry = 0;
+			qp->req.need_rd_atomic = 0;
+			rxe_run_task(&qp->req.task, 1);
+		}
+	}
+
+	if (unlikely(qp->req.state == QP_STATE_DRAIN)) {
+		/* state_lock used by requester & completer */
+		spin_lock_irqsave(&qp->state_lock, flags);
+		if ((qp->req.state == QP_STATE_DRAIN) &&
+		    (qp->comp.psn == qp->req.psn)) {
+			qp->req.state = QP_STATE_DRAINED;
+			spin_unlock_irqrestore(&qp->state_lock, flags);
+
+			if (qp->ibqp.event_handler) {
+				struct ib_event ev;
+
+				ev.device = qp->ibqp.device;
+				ev.element.qp = &qp->ibqp;
+				ev.event = IB_EVENT_SQ_DRAINED;
+				qp->ibqp.event_handler(&ev,
+					qp->ibqp.qp_context);
+			}
+		} else {
+			spin_unlock_irqrestore(&qp->state_lock, flags);
+		}
+	}
+
+	do_complete(qp, wqe);
+
+	if (psn_compare(pkt->psn, qp->comp.psn) >= 0)
+		return COMPST_UPDATE_COMP;
+	else
+		return COMPST_DONE;
+}
+
+static inline enum comp_state complete_wqe(struct rxe_qp *qp,
+					   struct rxe_pkt_info *pkt,
+					   struct rxe_send_wqe *wqe)
+{
+	qp->comp.opcode = -1;
+
+	if (pkt) {
+		if (psn_compare(pkt->psn, qp->comp.psn) >= 0)
+			qp->comp.psn = (pkt->psn + 1) & BTH_PSN_MASK;
+
+		if (qp->req.wait_psn) {
+			qp->req.wait_psn = 0;
+			rxe_run_task(&qp->req.task, 1);
+		}
+	}
+
+	do_complete(qp, wqe);
+
+	return COMPST_GET_WQE;
+}
+
+int rxe_completer(void *arg)
+{
+	struct rxe_qp *qp = (struct rxe_qp *)arg;
+	struct rxe_send_wqe *wqe = wqe;
+	struct sk_buff *skb = NULL;
+	struct rxe_pkt_info *pkt = NULL;
+	enum comp_state state;
+
+	if (!qp->valid) {
+		while ((skb = skb_dequeue(&qp->resp_pkts))) {
+			rxe_drop_ref(qp);
+			kfree_skb(skb);
+		}
+		skb = NULL;
+		pkt = NULL;
+
+		while (queue_head(qp->sq.queue))
+			advance_consumer(qp->sq.queue);
+
+		goto exit;
+	}
+
+	if (qp->req.state == QP_STATE_ERROR) {
+		while ((skb = skb_dequeue(&qp->resp_pkts))) {
+			rxe_drop_ref(qp);
+			kfree_skb(skb);
+		}
+		skb = NULL;
+		pkt = NULL;
+
+		while ((wqe = queue_head(qp->sq.queue))) {
+			wqe->status = IB_WC_WR_FLUSH_ERR;
+			do_complete(qp, wqe);
+		}
+
+		goto exit;
+	}
+
+	if (qp->req.state == QP_STATE_RESET) {
+		while ((skb = skb_dequeue(&qp->resp_pkts))) {
+			rxe_drop_ref(qp);
+			kfree_skb(skb);
+		}
+		skb = NULL;
+		pkt = NULL;
+
+		while (queue_head(qp->sq.queue))
+			advance_consumer(qp->sq.queue);
+
+		goto exit;
+	}
+
+	if (qp->comp.timeout) {
+		qp->comp.timeout_retry = 1;
+		qp->comp.timeout = 0;
+	} else {
+		qp->comp.timeout_retry = 0;
+	}
+
+	if (qp->req.need_retry)
+		goto exit;
+
+	state = COMPST_GET_ACK;
+
+	while (1) {
+		pr_debug("state = %s\n", comp_state_name[state]);
+		switch (state) {
+		case COMPST_GET_ACK:
+			skb = skb_dequeue(&qp->resp_pkts);
+			if (skb) {
+				pkt = SKB_TO_PKT(skb);
+				qp->comp.timeout_retry = 0;
+			}
+			state = COMPST_GET_WQE;
+			break;
+
+		case COMPST_GET_WQE:
+			state = get_wqe(qp, pkt, &wqe);
+			break;
+
+		case COMPST_CHECK_PSN:
+			state = check_psn(qp, pkt, wqe);
+			break;
+
+		case COMPST_CHECK_ACK:
+			state = check_ack(qp, pkt, wqe);
+			break;
+
+		case COMPST_READ:
+			state = do_read(qp, pkt, wqe);
+			break;
+
+		case COMPST_ATOMIC:
+			state = do_atomic(qp, pkt, wqe);
+			break;
+
+		case COMPST_WRITE_SEND:
+			if (wqe->state == wqe_state_pending &&
+			    wqe->last_psn == pkt->psn)
+				state = COMPST_COMP_ACK;
+			else
+				state = COMPST_UPDATE_COMP;
+			break;
+
+		case COMPST_COMP_ACK:
+			state = complete_ack(qp, pkt, wqe);
+			break;
+
+		case COMPST_COMP_WQE:
+			state = complete_wqe(qp, pkt, wqe);
+			break;
+
+		case COMPST_UPDATE_COMP:
+			if (pkt->mask & RXE_END_MASK)
+				qp->comp.opcode = -1;
+			else
+				qp->comp.opcode = pkt->opcode;
+
+			if (psn_compare(pkt->psn, qp->comp.psn) >= 0)
+				qp->comp.psn = (pkt->psn + 1) & BTH_PSN_MASK;
+
+			if (qp->req.wait_psn) {
+				qp->req.wait_psn = 0;
+				rxe_run_task(&qp->req.task, 1);
+			}
+
+			state = COMPST_DONE;
+			break;
+
+		case COMPST_DONE:
+			if (pkt) {
+				rxe_drop_ref(pkt->qp);
+				kfree_skb(skb);
+			}
+			goto done;
+
+		case COMPST_EXIT:
+			if (qp->comp.timeout_retry && wqe) {
+				state = COMPST_ERROR_RETRY;
+				break;
+			}
+
+			/* re reset the timeout counter if
+			 * (1) QP is type RC
+			 * (2) the QP is alive
+			 * (3) there is a packet sent by the requester that
+			 *     might be acked (we still might get spurious
+			 *     timeouts but try to keep them as few as possible)
+			 * (4) the timeout parameter is set
+			 */
+			if ((qp_type(qp) == IB_QPT_RC) &&
+			    (qp->req.state == QP_STATE_READY) &&
+			    (psn_compare(qp->req.psn, qp->comp.psn) > 0) &&
+			    qp->qp_timeout_jiffies)
+				mod_timer(&qp->retrans_timer,
+					  jiffies + qp->qp_timeout_jiffies);
+			goto exit;
+
+		case COMPST_ERROR_RETRY:
+			/* we come here if the retry timer fired and we did
+			 * not receive a response packet. try to retry the send
+			 * queue if that makes sense and the limits have not
+			 * been exceeded. remember that some timeouts are
+			 * spurious since we do not reset the timer but kick
+			 * it down the road or let it expire
+			 */
+
+			/* there is nothing to retry in this case */
+			if (!wqe || (wqe->state == wqe_state_posted))
+				goto exit;
+
+			if (qp->comp.retry_cnt > 0) {
+				if (qp->comp.retry_cnt != 7)
+					qp->comp.retry_cnt--;
+
+				/* no point in retrying if we have already
+				 * seen the last ack that the requester could
+				 * have caused
+				 */
+				if (psn_compare(qp->req.psn,
+						qp->comp.psn) > 0) {
+					/* tell the requester to retry the
+					 * send send queue next time around
+					 */
+					qp->req.need_retry = 1;
+					rxe_run_task(&qp->req.task, 1);
+				}
+				goto exit;
+			} else {
+				wqe->status = IB_WC_RETRY_EXC_ERR;
+				state = COMPST_ERROR;
+			}
+			break;
+
+		case COMPST_RNR_RETRY:
+			if (qp->comp.rnr_retry > 0) {
+				if (qp->comp.rnr_retry != 7)
+					qp->comp.rnr_retry--;
+
+				qp->req.need_retry = 1;
+				pr_debug("set rnr nak timer\n");
+				mod_timer(&qp->rnr_nak_timer,
+					  jiffies + rnrnak_jiffies(aeth_syn(pkt)
+						& ~AETH_TYPE_MASK));
+				goto exit;
+			} else {
+				wqe->status = IB_WC_RNR_RETRY_EXC_ERR;
+				state = COMPST_ERROR;
+			}
+			break;
+
+		case COMPST_ERROR:
+			do_complete(qp, wqe);
+			rxe_qp_error(qp);
+			goto exit;
+		}
+	}
+
+exit:
+	/* we come here if we are done with processing and want the task to
+	 * exit from the loop calling us
+	 */
+	return -EAGAIN;
+
+done:
+	/* we come here if we have processed a packet we want the task to call
+	 * us again to see if there is anything else to do
+	 */
+	return 0;
+}

drivers/infiniband/sw/rxe/rxe_cq.c

+/*
+ * Copyright (c) 2016 Mellanox Technologies Ltd. All rights reserved.
+ * Copyright (c) 2015 System Fabric Works, 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.
+ */
+
+#include "rxe.h"
+#include "rxe_loc.h"
+#include "rxe_queue.h"
+
+int rxe_cq_chk_attr(struct rxe_dev *rxe, struct rxe_cq *cq,
+		    int cqe, int comp_vector, struct ib_udata *udata)
+{
+	int count;
+
+	if (cqe <= 0) {
+		pr_warn("cqe(%d) <= 0\n", cqe);
+		goto err1;
+	}
+
+	if (cqe > rxe->attr.max_cqe) {
+		pr_warn("cqe(%d) > max_cqe(%d)\n",
+			cqe, rxe->attr.max_cqe);
+		goto err1;
+	}
+
+	if (cq) {
+		count = queue_count(cq->queue);
+		if (cqe < count) {
+			pr_warn("cqe(%d) < current # elements in queue (%d)",
+				cqe, count);
+			goto err1;
+		}
+	}
+
+	return 0;
+
+err1:
+	return -EINVAL;
+}
+
+static void rxe_send_complete(unsigned long data)
+{
+	struct rxe_cq *cq = (struct rxe_cq *)data;
+
+	cq->ibcq.comp_handler(&cq->ibcq, cq->ibcq.cq_context);
+}
+
+int rxe_cq_from_init(struct rxe_dev *rxe, struct rxe_cq *cq, int cqe,
+		     int comp_vector, struct ib_ucontext *context,
+		     struct ib_udata *udata)
+{
+	int err;
+
+	cq->queue = rxe_queue_init(rxe, &cqe,
+				   sizeof(struct rxe_cqe));
+	if (!cq->queue) {
+		pr_warn("unable to create cq\n");
+		return -ENOMEM;
+	}
+
+	err = do_mmap_info(rxe, udata, false, context, cq->queue->buf,
+			   cq->queue->buf_size, &cq->queue->ip);
+	if (err) {
+		kvfree(cq->queue->buf);
+		kfree(cq->queue);
+		return err;
+	}
+
+	if (udata)
+		cq->is_user = 1;
+
+	tasklet_init(&cq->comp_task, rxe_send_complete, (unsigned long)cq);
+
+	spin_lock_init(&cq->cq_lock);
+	cq->ibcq.cqe = cqe;
+	return 0;
+}
+
+int rxe_cq_resize_queue(struct rxe_cq *cq, int cqe, struct ib_udata *udata)
+{
+	int err;
+
+	err = rxe_queue_resize(cq->queue, (unsigned int *)&cqe,
+			       sizeof(struct rxe_cqe),
+			       cq->queue->ip ? cq->queue->ip->context : NULL,
+			       udata, NULL, &cq->cq_lock);
+	if (!err)
+		cq->ibcq.cqe = cqe;
+
+	return err;
+}
+
+int rxe_cq_post(struct rxe_cq *cq, struct rxe_cqe *cqe, int solicited)
+{
+	struct ib_event ev;
+	unsigned long flags;
+
+	spin_lock_irqsave(&cq->cq_lock, flags);
+
+	if (unlikely(queue_full(cq->queue))) {
+		spin_unlock_irqrestore(&cq->cq_lock, flags);
+		if (cq->ibcq.event_handler) {
+			ev.device = cq->ibcq.device;
+			ev.element.cq = &cq->ibcq;
+			ev.event = IB_EVENT_CQ_ERR;
+			cq->ibcq.event_handler(&ev, cq->ibcq.cq_context);
+		}
+
+		return -EBUSY;
+	}
+
+	memcpy(producer_addr(cq->queue), cqe, sizeof(*cqe));
+
+	/* make sure all changes to the CQ are written before we update the
+	 * producer pointer
+	 */
+	smp_wmb();
+
+	advance_producer(cq->queue);
+	spin_unlock_irqrestore(&cq->cq_lock, flags);
+
+	if ((cq->notify == IB_CQ_NEXT_COMP) ||
+	    (cq->notify == IB_CQ_SOLICITED && solicited)) {
+		cq->notify = 0;
+		tasklet_schedule(&cq->comp_task);
+	}
+
+	return 0;
+}
+
+void rxe_cq_cleanup(void *arg)
+{
+	struct rxe_cq *cq = arg;
+
+	if (cq->queue)
+		rxe_queue_cleanup(cq->queue);
+}

drivers/infiniband/sw/rxe/rxe_dma.c

+/*
+ * Copyright (c) 2016 Mellanox Technologies Ltd. All rights reserved.
+ * Copyright (c) 2015 System Fabric Works, 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.
+ */
+
+#include "rxe.h"
+#include "rxe_loc.h"
+
+#define DMA_BAD_ADDER ((u64)0)
+
+static int rxe_mapping_error(struct ib_device *dev, u64 dma_addr)
+{
+	return dma_addr == DMA_BAD_ADDER;
+}
+
+static u64 rxe_dma_map_single(struct ib_device *dev,
+			      void *cpu_addr, size_t size,
+			      enum dma_data_direction direction)
+{
+	WARN_ON(!valid_dma_direction(direction));
+	return (uintptr_t)cpu_addr;
+}
+
+static void rxe_dma_unmap_single(struct ib_device *dev,
+				 u64 addr, size_t size,
+				 enum dma_data_direction direction)
+{
+	WARN_ON(!valid_dma_direction(direction));
+}
+
+static u64 rxe_dma_map_page(struct ib_device *dev,
+			    struct page *page,
+			    unsigned long offset,
+			    size_t size, enum dma_data_direction direction)
+{
+	u64 addr;
+
+	WARN_ON(!valid_dma_direction(direction));
+
+	if (offset + size > PAGE_SIZE) {
+		addr = DMA_BAD_ADDER;
+		goto done;
+	}
+
+	addr = (uintptr_t)page_address(page);
+	if (addr)
+		addr += offset;
+
+done:
+	return addr;
+}
+
+static void rxe_dma_unmap_page(struct ib_device *dev,
+			       u64 addr, size_t size,
+			       enum dma_data_direction direction)
+{
+	WARN_ON(!valid_dma_direction(direction));
+}
+
+static int rxe_map_sg(struct ib_device *dev, struct scatterlist *sgl,
+		      int nents, enum dma_data_direction direction)
+{
+	struct scatterlist *sg;
+	u64 addr;
+	int i;
+	int ret = nents;
+
+	WARN_ON(!valid_dma_direction(direction));
+
+	for_each_sg(sgl, sg, nents, i) {
+		addr = (uintptr_t)page_address(sg_page(sg));
+		if (!addr) {
+			ret = 0;
+			break;
+		}
+		sg->dma_address = addr + sg->offset;
+#ifdef CONFIG_NEED_SG_DMA_LENGTH
+		sg->dma_length = sg->length;
+#endif
+	}
+
+	return ret;
+}
+
+static void rxe_unmap_sg(struct ib_device *dev,
+			 struct scatterlist *sg, int nents,
+			 enum dma_data_direction direction)
+{
+	WARN_ON(!valid_dma_direction(direction));
+}
+
+static void rxe_sync_single_for_cpu(struct ib_device *dev,
+				    u64 addr,
+				    size_t size, enum dma_data_direction dir)
+{
+}
+
+static void rxe_sync_single_for_device(struct ib_device *dev,
+				       u64 addr,
+				       size_t size, enum dma_data_direction dir)
+{
+}
+
+static void *rxe_dma_alloc_coherent(struct ib_device *dev, size_t size,
+				    u64 *dma_handle, gfp_t flag)
+{
+	struct page *p;
+	void *addr = NULL;
+
+	p = alloc_pages(flag, get_order(size));
+	if (p)
+		addr = page_address(p);
+
+	if (dma_handle)
+		*dma_handle = (uintptr_t)addr;
+
+	return addr;
+}
+
+static void rxe_dma_free_coherent(struct ib_device *dev, size_t size,
+				  void *cpu_addr, u64 dma_handle)
+{
+	free_pages((unsigned long)cpu_addr, get_order(size));
+}
+
+struct ib_dma_mapping_ops rxe_dma_mapping_ops = {
+	.mapping_error		= rxe_mapping_error,
+	.map_single		= rxe_dma_map_single,
+	.unmap_single		= rxe_dma_unmap_single,
+	.map_page		= rxe_dma_map_page,
+	.unmap_page		= rxe_dma_unmap_page,
+	.map_sg			= rxe_map_sg,
+	.unmap_sg		= rxe_unmap_sg,
+	.sync_single_for_cpu	= rxe_sync_single_for_cpu,
+	.sync_single_for_device	= rxe_sync_single_for_device,
+	.alloc_coherent		= rxe_dma_alloc_coherent,
+	.free_coherent		= rxe_dma_free_coherent
+};

drivers/infiniband/sw/rxe/rxe_icrc.c

+/*
+ * Copyright (c) 2016 Mellanox Technologies Ltd. All rights reserved.
+ * Copyright (c) 2015 System Fabric Works, 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.
+ */
+
+#include "rxe.h"
+#include "rxe_loc.h"
+
+/* Compute a partial ICRC for all the IB transport headers. */
+u32 rxe_icrc_hdr(struct rxe_pkt_info *pkt, struct sk_buff *skb)
+{
+	unsigned int bth_offset = 0;
+	struct iphdr *ip4h = NULL;
+	struct ipv6hdr *ip6h = NULL;
+	struct udphdr *udph;
+	struct rxe_bth *bth;
+	int crc;
+	int length;
+	int hdr_size = sizeof(struct udphdr) +
+		(skb->protocol == htons(ETH_P_IP) ?
+		sizeof(struct iphdr) : sizeof(struct ipv6hdr));
+	/* pseudo header buffer size is calculate using ipv6 header size since
+	 * it is bigger than ipv4
+	 */
+	u8 pshdr[sizeof(struct udphdr) +
+		sizeof(struct ipv6hdr) +
+		RXE_BTH_BYTES];
+
+	/* This seed is the result of computing a CRC with a seed of
+	 * 0xfffffff and 8 bytes of 0xff representing a masked LRH.
+	 */
+	crc = 0xdebb20e3;
+
+	if (skb->protocol == htons(ETH_P_IP)) { /* IPv4 */
+		memcpy(pshdr, ip_hdr(skb), hdr_size);
+		ip4h = (struct iphdr *)pshdr;
+		udph = (struct udphdr *)(ip4h + 1);
+
+		ip4h->ttl = 0xff;
+		ip4h->check = CSUM_MANGLED_0;
+		ip4h->tos = 0xff;
+	} else {				/* IPv6 */
+		memcpy(pshdr, ipv6_hdr(skb), hdr_size);
+		ip6h = (struct ipv6hdr *)pshdr;
+		udph = (struct udphdr *)(ip6h + 1);
+
+		memset(ip6h->flow_lbl, 0xff, sizeof(ip6h->flow_lbl));
+		ip6h->priority = 0xf;
+		ip6h->hop_limit = 0xff;
+	}
+	udph->check = CSUM_MANGLED_0;
+
+	bth_offset += hdr_size;
+
+	memcpy(&pshdr[bth_offset], pkt->hdr, RXE_BTH_BYTES);
+	bth = (struct rxe_bth *)&pshdr[bth_offset];
+
+	/* exclude bth.resv8a */
+	bth->qpn |= cpu_to_be32(~BTH_QPN_MASK);
+
+	length = hdr_size + RXE_BTH_BYTES;
+	crc = crc32_le(crc, pshdr, length);
+
+	/* And finish to compute the CRC on the remainder of the headers. */
+	crc = crc32_le(crc, pkt->hdr + RXE_BTH_BYTES,
+		       rxe_opcode[pkt->opcode].length - RXE_BTH_BYTES);
+	return crc;
+}

drivers/infiniband/sw/rxe/rxe_loc.h

+/*
+ * Copyright (c) 2016 Mellanox Technologies Ltd. All rights reserved.
+ * Copyright (c) 2015 System Fabric Works, 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 RXE_LOC_H
+#define RXE_LOC_H
+
+/* rxe_av.c */
+
+int rxe_av_chk_attr(struct rxe_dev *rxe, struct ib_ah_attr *attr);
+
+int rxe_av_from_attr(struct rxe_dev *rxe, u8 port_num,
+		     struct rxe_av *av, struct ib_ah_attr *attr);
+
+int rxe_av_to_attr(struct rxe_dev *rxe, struct rxe_av *av,
+		   struct ib_ah_attr *attr);
+
+int rxe_av_fill_ip_info(struct rxe_dev *rxe,
+			struct rxe_av *av,
+			struct ib_ah_attr *attr,
+			struct ib_gid_attr *sgid_attr,
+			union ib_gid *sgid);
+
+struct rxe_av *rxe_get_av(struct rxe_pkt_info *pkt);
+
+/* rxe_cq.c */
+int rxe_cq_chk_attr(struct rxe_dev *rxe, struct rxe_cq *cq,
+		    int cqe, int comp_vector, struct ib_udata *udata);
+
+int rxe_cq_from_init(struct rxe_dev *rxe, struct rxe_cq *cq, int cqe,
+		     int comp_vector, struct ib_ucontext *context,
+		     struct ib_udata *udata);
+
+int rxe_cq_resize_queue(struct rxe_cq *cq, int new_cqe, struct ib_udata *udata);
+
+int rxe_cq_post(struct rxe_cq *cq, struct rxe_cqe *cqe, int solicited);
+
+void rxe_cq_cleanup(void *arg);
+
+/* rxe_mcast.c */
+int rxe_mcast_get_grp(struct rxe_dev *rxe, union ib_gid *mgid,
+		      struct rxe_mc_grp **grp_p);
+
+int rxe_mcast_add_grp_elem(struct rxe_dev *rxe, struct rxe_qp *qp,
+			   struct rxe_mc_grp *grp);
+
+int rxe_mcast_drop_grp_elem(struct rxe_dev *rxe, struct rxe_qp *qp,
+			    union ib_gid *mgid);
+
+void rxe_drop_all_mcast_groups(struct rxe_qp *qp);
+
+void rxe_mc_cleanup(void *arg);
+
+/* rxe_mmap.c */
+struct rxe_mmap_info {
+	struct list_head	pending_mmaps;
+	struct ib_ucontext	*context;
+	struct kref		ref;
+	void			*obj;
+
+	struct mminfo info;
+};
+
+void rxe_mmap_release(struct kref *ref);
+
+struct rxe_mmap_info *rxe_create_mmap_info(struct rxe_dev *dev,
+					   u32 size,
+					   struct ib_ucontext *context,
+					   void *obj);
+
+int rxe_mmap(struct ib_ucontext *context, struct vm_area_struct *vma);
+
+/* rxe_mr.c */
+enum copy_direction {
+	to_mem_obj,
+	from_mem_obj,
+};
+
+int rxe_mem_init_dma(struct rxe_dev *rxe, struct rxe_pd *pd,
+		     int access, struct rxe_mem *mem);
+
+int rxe_mem_init_user(struct rxe_dev *rxe, struct rxe_pd *pd, u64 start,
+		      u64 length, u64 iova, int access, struct ib_udata *udata,
+		      struct rxe_mem *mr);
+
+int rxe_mem_init_fast(struct rxe_dev *rxe, struct rxe_pd *pd,
+		      int max_pages, struct rxe_mem *mem);
+
+int rxe_mem_copy(struct rxe_mem *mem, u64 iova, void *addr,
+		 int length, enum copy_direction dir, u32 *crcp);
+
+int copy_data(struct rxe_dev *rxe, struct rxe_pd *pd, int access,
+	      struct rxe_dma_info *dma, void *addr, int length,
+	      enum copy_direction dir, u32 *crcp);
+
+void *iova_to_vaddr(struct rxe_mem *mem, u64 iova, int length);
+
+enum lookup_type {
+	lookup_local,
+	lookup_remote,
+};
+
+struct rxe_mem *lookup_mem(struct rxe_pd *pd, int access, u32 key,
+			   enum lookup_type type);
+
+int mem_check_range(struct rxe_mem *mem, u64 iova, size_t length);
+
+int rxe_mem_map_pages(struct rxe_dev *rxe, struct rxe_mem *mem,
+		      u64 *page, int num_pages, u64 iova);
+
+void rxe_mem_cleanup(void *arg);
+
+int advance_dma_data(struct rxe_dma_info *dma, unsigned int length);
+
+/* rxe_qp.c */
+int rxe_qp_chk_init(struct rxe_dev *rxe, struct ib_qp_init_attr *init);
+
+int rxe_qp_from_init(struct rxe_dev *rxe, struct rxe_qp *qp, struct rxe_pd *pd,
+		     struct ib_qp_init_attr *init, struct ib_udata *udata,
+		     struct ib_pd *ibpd);
+
+int rxe_qp_to_init(struct rxe_qp *qp, struct ib_qp_init_attr *init);
+
+int rxe_qp_chk_attr(struct rxe_dev *rxe, struct rxe_qp *qp,
+		    struct ib_qp_attr *attr, int mask);
+
+int rxe_qp_from_attr(struct rxe_qp *qp, struct ib_qp_attr *attr,
+		     int mask, struct ib_udata *udata);
+
+int rxe_qp_to_attr(struct rxe_qp *qp, struct ib_qp_attr *attr, int mask);
+
+void rxe_qp_error(struct rxe_qp *qp);
+
+void rxe_qp_destroy(struct rxe_qp *qp);
+
+void rxe_qp_cleanup(void *arg);
+
+static inline int qp_num(struct rxe_qp *qp)
+{
+	return qp->ibqp.qp_num;
+}
+
+static inline enum ib_qp_type qp_type(struct rxe_qp *qp)
+{
+	return qp->ibqp.qp_type;
+}
+
+static inline enum ib_qp_state qp_state(struct rxe_qp *qp)
+{
+	return qp->attr.qp_state;
+}
+
+static inline int qp_mtu(struct rxe_qp *qp)
+{
+	if (qp->ibqp.qp_type == IB_QPT_RC || qp->ibqp.qp_type == IB_QPT_UC)
+		return qp->attr.path_mtu;
+	else
+		return RXE_PORT_MAX_MTU;
+}
+
+static inline int rcv_wqe_size(int max_sge)
+{
+	return sizeof(struct rxe_recv_wqe) +
+		max_sge * sizeof(struct ib_sge);
+}
+
+void free_rd_atomic_resource(struct rxe_qp *qp, struct resp_res *res);
+
+static inline void rxe_advance_resp_resource(struct rxe_qp *qp)
+{
+	qp->resp.res_head++;
+	if (unlikely(qp->resp.res_head == qp->attr.max_rd_atomic))
+		qp->resp.res_head = 0;
+}
+
+void retransmit_timer(unsigned long data);
+void rnr_nak_timer(unsigned long data);
+
+void dump_qp(struct rxe_qp *qp);
+
+/* rxe_srq.c */
+#define IB_SRQ_INIT_MASK (~IB_SRQ_LIMIT)
+
+int rxe_srq_chk_attr(struct rxe_dev *rxe, struct rxe_srq *srq,
+		     struct ib_srq_attr *attr, enum ib_srq_attr_mask mask);
+
+int rxe_srq_from_init(struct rxe_dev *rxe, struct rxe_srq *srq,
+		      struct ib_srq_init_attr *init,
+		      struct ib_ucontext *context, struct ib_udata *udata);
+
+int rxe_srq_from_attr(struct rxe_dev *rxe, struct rxe_srq *srq,
+		      struct ib_srq_attr *attr, enum ib_srq_attr_mask mask,
+		      struct ib_udata *udata);
+
+extern struct ib_dma_mapping_ops rxe_dma_mapping_ops;
+
+void rxe_release(struct kref *kref);
+
+int rxe_completer(void *arg);
+int rxe_requester(void *arg);
+int rxe_responder(void *arg);
+
+u32 rxe_icrc_hdr(struct rxe_pkt_info *pkt, struct sk_buff *skb);
+
+void rxe_resp_queue_pkt(struct rxe_dev *rxe,
+			struct rxe_qp *qp, struct sk_buff *skb);
+
+void rxe_comp_queue_pkt(struct rxe_dev *rxe,
+			struct rxe_qp *qp, struct sk_buff *skb);
+
+static inline unsigned wr_opcode_mask(int opcode, struct rxe_qp *qp)
+{
+	return rxe_wr_opcode_info[opcode].mask[qp->ibqp.qp_type];
+}
+
+static inline int rxe_xmit_packet(struct rxe_dev *rxe, struct rxe_qp *qp,
+				  struct rxe_pkt_info *pkt, struct sk_buff *skb)
+{
+	int err;
+	int is_request = pkt->mask & RXE_REQ_MASK;
+
+	if ((is_request && (qp->req.state != QP_STATE_READY)) ||
+	    (!is_request && (qp->resp.state != QP_STATE_READY))) {
+		pr_info("Packet dropped. QP is not in ready state\n");
+		goto drop;
+	}
+
+	if (pkt->mask & RXE_LOOPBACK_MASK) {
+		memcpy(SKB_TO_PKT(skb), pkt, sizeof(*pkt));
+		err = rxe->ifc_ops->loopback(skb);
+	} else {
+		err = rxe->ifc_ops->send(rxe, pkt, skb);
+	}
+
+	if (err) {
+		rxe->xmit_errors++;
+		return err;
+	}
+
+	atomic_inc(&qp->skb_out);
+
+	if ((qp_type(qp) != IB_QPT_RC) &&
+	    (pkt->mask & RXE_END_MASK)) {
+		pkt->wqe->state = wqe_state_done;
+		rxe_run_task(&qp->comp.task, 1);
+	}
+
+	goto done;
+
+drop:
+	kfree_skb(skb);
+	err = 0;
+done:
+	return err;
+}
+
+#endif /* RXE_LOC_H */

drivers/infiniband/sw/rxe/rxe_mcast.c

+/*
+ * Copyright (c) 2016 Mellanox Technologies Ltd. All rights reserved.
+ * Copyright (c) 2015 System Fabric Works, 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.
+ */
+
+#include "rxe.h"
+#include "rxe_loc.h"
+
+int rxe_mcast_get_grp(struct rxe_dev *rxe, union ib_gid *mgid,
+		      struct rxe_mc_grp **grp_p)
+{
+	int err;
+	struct rxe_mc_grp *grp;
+
+	if (rxe->attr.max_mcast_qp_attach == 0) {
+		err = -EINVAL;
+		goto err1;
+	}
+
+	grp = rxe_pool_get_key(&rxe->mc_grp_pool, mgid);
+	if (grp)
+		goto done;
+
+	grp = rxe_alloc(&rxe->mc_grp_pool);
+	if (!grp) {
+		err = -ENOMEM;
+		goto err1;
+	}
+
+	INIT_LIST_HEAD(&grp->qp_list);
+	spin_lock_init(&grp->mcg_lock);
+	grp->rxe = rxe;
+
+	rxe_add_key(grp, mgid);
+
+	err = rxe->ifc_ops->mcast_add(rxe, mgid);
+	if (err)
+		goto err2;
+
+done:
+	*grp_p = grp;
+	return 0;
+
+err2:
+	rxe_drop_ref(grp);
+err1:
+	return err;
+}
+
+int rxe_mcast_add_grp_elem(struct rxe_dev *rxe, struct rxe_qp *qp,
+			   struct rxe_mc_grp *grp)
+{
+	int err;
+	struct rxe_mc_elem *elem;
+
+	/* check to see of the qp is already a member of the group */
+	spin_lock_bh(&qp->grp_lock);
+	spin_lock_bh(&grp->mcg_lock);
+	list_for_each_entry(elem, &grp->qp_list, qp_list) {
+		if (elem->qp == qp) {
+			err = 0;
+			goto out;
+		}
+	}
+
+	if (grp->num_qp >= rxe->attr.max_mcast_qp_attach) {
+		err = -ENOMEM;
+		goto out;
+	}
+
+	elem = rxe_alloc(&rxe->mc_elem_pool);
+	if (!elem) {
+		err = -ENOMEM;
+		goto out;
+	}
+
+	/* each qp holds a ref on the grp */
+	rxe_add_ref(grp);
+
+	grp->num_qp++;
+	elem->qp = qp;
+	elem->grp = grp;
+
+	list_add(&elem->qp_list, &grp->qp_list);
+	list_add(&elem->grp_list, &qp->grp_list);
+
+	err = 0;
+out:
+	spin_unlock_bh(&grp->mcg_lock);
+	spin_unlock_bh(&qp->grp_lock);
+	return err;
+}
+
+int rxe_mcast_drop_grp_elem(struct rxe_dev *rxe, struct rxe_qp *qp,
+			    union ib_gid *mgid)
+{
+	struct rxe_mc_grp *grp;
+	struct rxe_mc_elem *elem, *tmp;
+
+	grp = rxe_pool_get_key(&rxe->mc_grp_pool, mgid);
+	if (!grp)
+		goto err1;
+
+	spin_lock_bh(&qp->grp_lock);
+	spin_lock_bh(&grp->mcg_lock);
+
+	list_for_each_entry_safe(elem, tmp, &grp->qp_list, qp_list) {
+		if (elem->qp == qp) {
+			list_del(&elem->qp_list);
+			list_del(&elem->grp_list);
+			grp->num_qp--;
+
+			spin_unlock_bh(&grp->mcg_lock);
+			spin_unlock_bh(&qp->grp_lock);
+			rxe_drop_ref(elem);
+			rxe_drop_ref(grp);	/* ref held by QP */
+			rxe_drop_ref(grp);	/* ref from get_key */
+			return 0;
+		}
+	}
+
+	spin_unlock_bh(&grp->mcg_lock);
+	spin_unlock_bh(&qp->grp_lock);
+	rxe_drop_ref(grp);			/* ref from get_key */
+err1:
+	return -EINVAL;
+}
+
+void rxe_drop_all_mcast_groups(struct rxe_qp *qp)
+{
+	struct rxe_mc_grp *grp;
+	struct rxe_mc_elem *elem;
+
+	while (1) {
+		spin_lock_bh(&qp->grp_lock);
+		if (list_empty(&qp->grp_list)) {
+			spin_unlock_bh(&qp->grp_lock);
+			break;
+		}
+		elem = list_first_entry(&qp->grp_list, struct rxe_mc_elem,
+					grp_list);
+		list_del(&elem->grp_list);
+		spin_unlock_bh(&qp->grp_lock);
+
+		grp = elem->grp;
+		spin_lock_bh(&grp->mcg_lock);
+		list_del(&elem->qp_list);
+		grp->num_qp--;
+		spin_unlock_bh(&grp->mcg_lock);
+		rxe_drop_ref(grp);
+		rxe_drop_ref(elem);
+	}
+}
+
+void rxe_mc_cleanup(void *arg)
+{
+	struct rxe_mc_grp *grp = arg;
+	struct rxe_dev *rxe = grp->rxe;
+
+	rxe_drop_key(grp);
+	rxe->ifc_ops->mcast_delete(rxe, &grp->mgid);
+}

drivers/infiniband/sw/rxe/rxe_mmap.c

+/*
+ * Copyright (c) 2016 Mellanox Technologies Ltd. All rights reserved.
+ * Copyright (c) 2015 System Fabric Works, 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.
+ */
+
+#include <linux/module.h>
+#include <linux/vmalloc.h>
+#include <linux/mm.h>
+#include <linux/errno.h>
+#include <asm/pgtable.h>
+
+#include "rxe.h"
+#include "rxe_loc.h"
+#include "rxe_queue.h"
+
+void rxe_mmap_release(struct kref *ref)
+{
+	struct rxe_mmap_info *ip = container_of(ref,
+					struct rxe_mmap_info, ref);
+	struct rxe_dev *rxe = to_rdev(ip->context->device);
+
+	spin_lock_bh(&rxe->pending_lock);
+
+	if (!list_empty(&ip->pending_mmaps))
+		list_del(&ip->pending_mmaps);
+
+	spin_unlock_bh(&rxe->pending_lock);
+
+	vfree(ip->obj);		/* buf */
+	kfree(ip);
+}
+
+/*
+ * open and close keep track of how many times the memory region is mapped,
+ * to avoid releasing it.
+ */
+static void rxe_vma_open(struct vm_area_struct *vma)
+{
+	struct rxe_mmap_info *ip = vma->vm_private_data;
+
+	kref_get(&ip->ref);
+}
+
+static void rxe_vma_close(struct vm_area_struct *vma)
+{
+	struct rxe_mmap_info *ip = vma->vm_private_data;
+
+	kref_put(&ip->ref, rxe_mmap_release);
+}
+
+static struct vm_operations_struct rxe_vm_ops = {
+	.open = rxe_vma_open,
+	.close = rxe_vma_close,
+};
+
+/**
+ * rxe_mmap - create a new mmap region
+ * @context: the IB user context of the process making the mmap() call
+ * @vma: the VMA to be initialized
+ * Return zero if the mmap is OK. Otherwise, return an errno.
+ */
+int rxe_mmap(struct ib_ucontext *context, struct vm_area_struct *vma)
+{
+	struct rxe_dev *rxe = to_rdev(context->device);
+	unsigned long offset = vma->vm_pgoff << PAGE_SHIFT;
+	unsigned long size = vma->vm_end - vma->vm_start;
+	struct rxe_mmap_info *ip, *pp;
+	int ret;
+
+	/*
+	 * Search the device's list of objects waiting for a mmap call.
+	 * Normally, this list is very short since a call to create a
+	 * CQ, QP, or SRQ is soon followed by a call to mmap().
+	 */
+	spin_lock_bh(&rxe->pending_lock);
+	list_for_each_entry_safe(ip, pp, &rxe->pending_mmaps, pending_mmaps) {
+		if (context != ip->context || (__u64)offset != ip->info.offset)
+			continue;
+
+		/* Don't allow a mmap larger than the object. */
+		if (size > ip->info.size) {
+			pr_err("mmap region is larger than the object!\n");
+			spin_unlock_bh(&rxe->pending_lock);
+			ret = -EINVAL;
+			goto done;
+		}
+
+		goto found_it;
+	}
+	pr_warn("unable to find pending mmap info\n");
+	spin_unlock_bh(&rxe->pending_lock);
+	ret = -EINVAL;
+	goto done;
+
+found_it:
+	list_del_init(&ip->pending_mmaps);
+	spin_unlock_bh(&rxe->pending_lock);
+
+	ret = remap_vmalloc_range(vma, ip->obj, 0);
+	if (ret) {
+		pr_err("rxe: err %d from remap_vmalloc_range\n", ret);
+		goto done;
+	}
+
+	vma->vm_ops = &rxe_vm_ops;
+	vma->vm_private_data = ip;
+	rxe_vma_open(vma);
+done:
+	return ret;
+}
+
+/*
+ * Allocate information for rxe_mmap
+ */
+struct rxe_mmap_info *rxe_create_mmap_info(struct rxe_dev *rxe,
+					   u32 size,
+					   struct ib_ucontext *context,
+					   void *obj)
+{
+	struct rxe_mmap_info *ip;
+
+	ip = kmalloc(sizeof(*ip), GFP_KERNEL);
+	if (!ip)
+		return NULL;
+
+	size = PAGE_ALIGN(size);
+
+	spin_lock_bh(&rxe->mmap_offset_lock);
+
+	if (rxe->mmap_offset == 0)
+		rxe->mmap_offset = PAGE_SIZE;
+
+	ip->info.offset = rxe->mmap_offset;
+	rxe->mmap_offset += size;
+
+	spin_unlock_bh(&rxe->mmap_offset_lock);
+
+	INIT_LIST_HEAD(&ip->pending_mmaps);
+	ip->info.size = size;
+	ip->context = context;
+	ip->obj = obj;
+	kref_init(&ip->ref);
+
+	return ip;
+}

drivers/infiniband/sw/rxe/rxe_mr.c

+/*
+ * Copyright (c) 2016 Mellanox Technologies Ltd. All rights reserved.
+ * Copyright (c) 2015 System Fabric Works, 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.
+ */
+
+#include "rxe.h"
+#include "rxe_loc.h"
+
+/*
+ * lfsr (linear feedback shift register) with period 255
+ */
+static u8 rxe_get_key(void)
+{
+	static unsigned key = 1;
+
+	key = key << 1;
+
+	key |= (0 != (key & 0x100)) ^ (0 != (key & 0x10))
+		^ (0 != (key & 0x80)) ^ (0 != (key & 0x40));
+
+	key &= 0xff;
+
+	return key;
+}
+
+int mem_check_range(struct rxe_mem *mem, u64 iova, size_t length)
+{
+	switch (mem->type) {
+	case RXE_MEM_TYPE_DMA:
+		return 0;
+
+	case RXE_MEM_TYPE_MR:
+	case RXE_MEM_TYPE_FMR:
+		return ((iova < mem->iova) ||
+			((iova + length) > (mem->iova + mem->length))) ?
+			-EFAULT : 0;
+
+	default:
+		return -EFAULT;
+	}
+}
+
+#define IB_ACCESS_REMOTE	(IB_ACCESS_REMOTE_READ		\
+				| IB_ACCESS_REMOTE_WRITE	\
+				| IB_ACCESS_REMOTE_ATOMIC)
+
+static void rxe_mem_init(int access, struct rxe_mem *mem)
+{
+	u32 lkey = mem->pelem.index << 8 | rxe_get_key();
+	u32 rkey = (access & IB_ACCESS_REMOTE) ? lkey : 0;
+
+	if (mem->pelem.pool->type == RXE_TYPE_MR) {
+		mem->ibmr.lkey		= lkey;
+		mem->ibmr.rkey		= rkey;
+	}
+
+	mem->lkey		= lkey;
+	mem->rkey		= rkey;
+	mem->state		= RXE_MEM_STATE_INVALID;
+	mem->type		= RXE_MEM_TYPE_NONE;
+	mem->map_shift		= ilog2(RXE_BUF_PER_MAP);
+}
+
+void rxe_mem_cleanup(void *arg)
+{
+	struct rxe_mem *mem = arg;
+	int i;
+
+	if (mem->umem)
+		ib_umem_release(mem->umem);
+
+	if (mem->map) {
+		for (i = 0; i < mem->num_map; i++)
+			kfree(mem->map[i]);
+
+		kfree(mem->map);
+	}
+}
+
+static int rxe_mem_alloc(struct rxe_dev *rxe, struct rxe_mem *mem, int num_buf)
+{
+	int i;
+	int num_map;
+	struct rxe_map **map = mem->map;
+
+	num_map = (num_buf + RXE_BUF_PER_MAP - 1) / RXE_BUF_PER_MAP;
+
+	mem->map = kmalloc_array(num_map, sizeof(*map), GFP_KERNEL);
+	if (!mem->map)
+		goto err1;
+
+	for (i = 0; i < num_map; i++) {
+		mem->map[i] = kmalloc(sizeof(**map), GFP_KERNEL);
+		if (!mem->map[i])
+			goto err2;
+	}
+
+	WARN_ON(!is_power_of_2(RXE_BUF_PER_MAP));
+
+	mem->map_shift	= ilog2(RXE_BUF_PER_MAP);
+	mem->map_mask	= RXE_BUF_PER_MAP - 1;
+
+	mem->num_buf = num_buf;
+	mem->num_map = num_map;
+	mem->max_buf = num_map * RXE_BUF_PER_MAP;
+
+	return 0;
+
+err2:
+	for (i--; i >= 0; i--)
+		kfree(mem->map[i]);
+
+	kfree(mem->map);
+err1:
+	return -ENOMEM;
+}
+
+int rxe_mem_init_dma(struct rxe_dev *rxe, struct rxe_pd *pd,
+		     int access, struct rxe_mem *mem)
+{
+	rxe_mem_init(access, mem);
+
+	mem->pd			= pd;
+	mem->access		= access;
+	mem->state		= RXE_MEM_STATE_VALID;
+	mem->type		= RXE_MEM_TYPE_DMA;
+
+	return 0;
+}
+
+int rxe_mem_init_user(struct rxe_dev *rxe, struct rxe_pd *pd, u64 start,
+		      u64 length, u64 iova, int access, struct ib_udata *udata,
+		      struct rxe_mem *mem)
+{
+	int			entry;
+	struct rxe_map		**map;
+	struct rxe_phys_buf	*buf = NULL;
+	struct ib_umem		*umem;
+	struct scatterlist	*sg;
+	int			num_buf;
+	void			*vaddr;
+	int err;
+
+	umem = ib_umem_get(pd->ibpd.uobject->context, start, length, access, 0);
+	if (IS_ERR(umem)) {
+		pr_warn("err %d from rxe_umem_get\n",
+			(int)PTR_ERR(umem));
+		err = -EINVAL;
+		goto err1;
+	}
+
+	mem->umem = umem;
+	num_buf = umem->nmap;
+
+	rxe_mem_init(access, mem);
+
+	err = rxe_mem_alloc(rxe, mem, num_buf);
+	if (err) {
+		pr_warn("err %d from rxe_mem_alloc\n", err);
+		ib_umem_release(umem);
+		goto err1;
+	}
+
+	WARN_ON(!is_power_of_2(umem->page_size));
+
+	mem->page_shift		= ilog2(umem->page_size);
+	mem->page_mask		= umem->page_size - 1;
+
+	num_buf			= 0;
+	map			= mem->map;
+	if (length > 0) {
+		buf = map[0]->buf;
+
+		for_each_sg(umem->sg_head.sgl, sg, umem->nmap, entry) {
+			vaddr = page_address(sg_page(sg));
+			if (!vaddr) {
+				pr_warn("null vaddr\n");
+				err = -ENOMEM;
+				goto err1;
+			}
+
+			buf->addr = (uintptr_t)vaddr;
+			buf->size = umem->page_size;
+			num_buf++;
+			buf++;
+
+			if (num_buf >= RXE_BUF_PER_MAP) {
+				map++;
+				buf = map[0]->buf;
+				num_buf = 0;
+			}
+		}
+	}
+
+	mem->pd			= pd;
+	mem->umem		= umem;
+	mem->access		= access;
+	mem->length		= length;
+	mem->iova		= iova;
+	mem->va			= start;
+	mem->offset		= ib_umem_offset(umem);
+	mem->state		= RXE_MEM_STATE_VALID;
+	mem->type		= RXE_MEM_TYPE_MR;
+
+	return 0;
+
+err1:
+	return err;
+}
+
+int rxe_mem_init_fast(struct rxe_dev *rxe, struct rxe_pd *pd,
+		      int max_pages, struct rxe_mem *mem)
+{
+	int err;
+
+	rxe_mem_init(0, mem);
+
+	/* In fastreg, we also set the rkey */
+	mem->ibmr.rkey = mem->ibmr.lkey;
+
+	err = rxe_mem_alloc(rxe, mem, max_pages);
+	if (err)
+		goto err1;
+
+	mem->pd			= pd;
+	mem->max_buf		= max_pages;
+	mem->state		= RXE_MEM_STATE_FREE;
+	mem->type		= RXE_MEM_TYPE_MR;
+
+	return 0;
+
+err1:
+	return err;
+}
+
+static void lookup_iova(
+	struct rxe_mem	*mem,
+	u64			iova,
+	int			*m_out,
+	int			*n_out,
+	size_t			*offset_out)
+{
+	size_t			offset = iova - mem->iova + mem->offset;
+	int			map_index;
+	int			buf_index;
+	u64			length;
+
+	if (likely(mem->page_shift)) {
+		*offset_out = offset & mem->page_mask;
+		offset >>= mem->page_shift;
+		*n_out = offset & mem->map_mask;
+		*m_out = offset >> mem->map_shift;
+	} else {
+		map_index = 0;
+		buf_index = 0;
+
+		length = mem->map[map_index]->buf[buf_index].size;
+
+		while (offset >= length) {
+			offset -= length;
+			buf_index++;
+
+			if (buf_index == RXE_BUF_PER_MAP) {
+				map_index++;
+				buf_index = 0;
+			}
+			length = mem->map[map_index]->buf[buf_index].size;
+		}
+
+		*m_out = map_index;
+		*n_out = buf_index;
+		*offset_out = offset;
+	}
+}
+
+void *iova_to_vaddr(struct rxe_mem *mem, u64 iova, int length)
+{
+	size_t offset;
+	int m, n;
+	void *addr;
+
+	if (mem->state != RXE_MEM_STATE_VALID) {
+		pr_warn("mem not in valid state\n");
+		addr = NULL;
+		goto out;
+	}
+
+	if (!mem->map) {
+		addr = (void *)(uintptr_t)iova;
+		goto out;
+	}
+
+	if (mem_check_range(mem, iova, length)) {
+		pr_warn("range violation\n");
+		addr = NULL;
+		goto out;
+	}
+
+	lookup_iova(mem, iova, &m, &n, &offset);
+
+	if (offset + length > mem->map[m]->buf[n].size) {
+		pr_warn("crosses page boundary\n");
+		addr = NULL;
+		goto out;
+	}
+
+	addr = (void *)(uintptr_t)mem->map[m]->buf[n].addr + offset;
+
+out:
+	return addr;
+}
+
+/* copy data from a range (vaddr, vaddr+length-1) to or from
+ * a mem object starting at iova. Compute incremental value of
+ * crc32 if crcp is not zero. caller must hold a reference to mem
+ */
+int rxe_mem_copy(struct rxe_mem *mem, u64 iova, void *addr, int length,
+		 enum copy_direction dir, u32 *crcp)
+{
+	int			err;
+	int			bytes;
+	u8			*va;
+	struct rxe_map		**map;
+	struct rxe_phys_buf	*buf;
+	int			m;
+	int			i;
+	size_t			offset;
+	u32			crc = crcp ? (*crcp) : 0;
+
+	if (mem->type == RXE_MEM_TYPE_DMA) {
+		u8 *src, *dest;
+
+		src  = (dir == to_mem_obj) ?
+			addr : ((void *)(uintptr_t)iova);
+
+		dest = (dir == to_mem_obj) ?
+			((void *)(uintptr_t)iova) : addr;
+
+		if (crcp)
+			*crcp = crc32_le(*crcp, src, length);
+
+		memcpy(dest, src, length);
+
+		return 0;
+	}
+
+	WARN_ON(!mem->map);
+
+	err = mem_check_range(mem, iova, length);
+	if (err) {
+		err = -EFAULT;
+		goto err1;
+	}
+
+	lookup_iova(mem, iova, &m, &i, &offset);
+
+	map	= mem->map + m;
+	buf	= map[0]->buf + i;
+
+	while (length > 0) {
+		u8 *src, *dest;
+
+		va	= (u8 *)(uintptr_t)buf->addr + offset;
+		src  = (dir == to_mem_obj) ? addr : va;
+		dest = (dir == to_mem_obj) ? va : addr;
+
+		bytes	= buf->size - offset;
+
+		if (bytes > length)
+			bytes = length;
+
+		if (crcp)
+			crc = crc32_le(crc, src, bytes);
+
+		memcpy(dest, src, bytes);
+
+		length	-= bytes;
+		addr	+= bytes;
+
+		offset	= 0;
+		buf++;
+		i++;
+
+		if (i == RXE_BUF_PER_MAP) {
+			i = 0;
+			map++;
+			buf = map[0]->buf;
+		}
+	}
+
+	if (crcp)
+		*crcp = crc;
+
+	return 0;
+
+err1:
+	return err;
+}
+
+/* copy data in or out of a wqe, i.e. sg list
+ * under the control of a dma descriptor
+ */
+int copy_data(
+	struct rxe_dev		*rxe,
+	struct rxe_pd		*pd,
+	int			access,
+	struct rxe_dma_info	*dma,
+	void			*addr,
+	int			length,
+	enum copy_direction	dir,
+	u32			*crcp)
+{
+	int			bytes;
+	struct rxe_sge		*sge	= &dma->sge[dma->cur_sge];
+	int			offset	= dma->sge_offset;
+	int			resid	= dma->resid;
+	struct rxe_mem		*mem	= NULL;
+	u64			iova;
+	int			err;
+
+	if (length == 0)
+		return 0;
+
+	if (length > resid) {
+		err = -EINVAL;
+		goto err2;
+	}
+
+	if (sge->length && (offset < sge->length)) {
+		mem = lookup_mem(pd, access, sge->lkey, lookup_local);
+		if (!mem) {
+			err = -EINVAL;
+			goto err1;
+		}
+	}
+
+	while (length > 0) {
+		bytes = length;
+
+		if (offset >= sge->length) {
+			if (mem) {
+				rxe_drop_ref(mem);
+				mem = NULL;
+			}
+			sge++;
+			dma->cur_sge++;
+			offset = 0;
+
+			if (dma->cur_sge >= dma->num_sge) {
+				err = -ENOSPC;
+				goto err2;
+			}
+
+			if (sge->length) {
+				mem = lookup_mem(pd, access, sge->lkey,
+						 lookup_local);
+				if (!mem) {
+					err = -EINVAL;
+					goto err1;
+				}
+			} else {
+				continue;
+			}
+		}
+
+		if (bytes > sge->length - offset)
+			bytes = sge->length - offset;
+
+		if (bytes > 0) {
+			iova = sge->addr + offset;
+
+			err = rxe_mem_copy(mem, iova, addr, bytes, dir, crcp);
+			if (err)
+				goto err2;
+
+			offset	+= bytes;
+			resid	-= bytes;
+			length	-= bytes;
+			addr	+= bytes;
+		}
+	}
+
+	dma->sge_offset = offset;
+	dma->resid	= resid;
+
+	if (mem)
+		rxe_drop_ref(mem);
+
+	return 0;
+
+err2:
+	if (mem)
+		rxe_drop_ref(mem);
+err1:
+	return err;
+}
+
+int advance_dma_data(struct rxe_dma_info *dma, unsigned int length)
+{
+	struct rxe_sge		*sge	= &dma->sge[dma->cur_sge];
+	int			offset	= dma->sge_offset;
+	int			resid	= dma->resid;
+
+	while (length) {
+		unsigned int bytes;
+
+		if (offset >= sge->length) {
+			sge++;
+			dma->cur_sge++;
+			offset = 0;
+			if (dma->cur_sge >= dma->num_sge)
+				return -ENOSPC;
+		}
+
+		bytes = length;
+
+		if (bytes > sge->length - offset)
+			bytes = sge->length - offset;
+
+		offset	+= bytes;
+		resid	-= bytes;
+		length	-= bytes;
+	}
+
+	dma->sge_offset = offset;
+	dma->resid	= resid;
+
+	return 0;
+}
+
+/* (1) find the mem (mr or mw) corresponding to lkey/rkey
+ *     depending on lookup_type
+ * (2) verify that the (qp) pd matches the mem pd
+ * (3) verify that the mem can support the requested access
+ * (4) verify that mem state is valid
+ */
+struct rxe_mem *lookup_mem(struct rxe_pd *pd, int access, u32 key,
+			   enum lookup_type type)
+{
+	struct rxe_mem *mem;
+	struct rxe_dev *rxe = to_rdev(pd->ibpd.device);
+	int index = key >> 8;
+
+	if (index >= RXE_MIN_MR_INDEX && index <= RXE_MAX_MR_INDEX) {
+		mem = rxe_pool_get_index(&rxe->mr_pool, index);
+		if (!mem)
+			goto err1;
+	} else {
+		goto err1;
+	}
+
+	if ((type == lookup_local && mem->lkey != key) ||
+	    (type == lookup_remote && mem->rkey != key))
+		goto err2;
+
+	if (mem->pd != pd)
+		goto err2;
+
+	if (access && !(access & mem->access))
+		goto err2;
+
+	if (mem->state != RXE_MEM_STATE_VALID)
+		goto err2;
+
+	return mem;
+
+err2:
+	rxe_drop_ref(mem);
+err1:
+	return NULL;
+}
+
+int rxe_mem_map_pages(struct rxe_dev *rxe, struct rxe_mem *mem,
+		      u64 *page, int num_pages, u64 iova)
+{
+	int i;
+	int num_buf;
+	int err;
+	struct rxe_map **map;
+	struct rxe_phys_buf *buf;
+	int page_size;
+
+	if (num_pages > mem->max_buf) {
+		err = -EINVAL;
+		goto err1;
+	}
+
+	num_buf		= 0;
+	page_size	= 1 << mem->page_shift;
+	map		= mem->map;
+	buf		= map[0]->buf;
+
+	for (i = 0; i < num_pages; i++) {
+		buf->addr = *page++;
+		buf->size = page_size;
+		buf++;
+		num_buf++;
+
+		if (num_buf == RXE_BUF_PER_MAP) {
+			map++;
+			buf = map[0]->buf;
+			num_buf = 0;
+		}
+	}
+
+	mem->iova	= iova;
+	mem->va		= iova;
+	mem->length	= num_pages << mem->page_shift;
+	mem->state	= RXE_MEM_STATE_VALID;
+
+	return 0;
+
+err1:
+	return err;
+}

drivers/infiniband/sw/rxe/rxe_net.h

+/*
+ * Copyright (c) 2016 Mellanox Technologies Ltd. All rights reserved.
+ * Copyright (c) 2015 System Fabric Works, 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 RXE_NET_H
+#define RXE_NET_H
+
+#include <net/sock.h>
+#include <net/if_inet6.h>
+#include <linux/module.h>
+
+struct rxe_recv_sockets {
+	struct socket *sk4;
+	struct socket *sk6;
+};
+
+extern struct rxe_recv_sockets recv_sockets;
+
+struct rxe_dev *rxe_net_add(struct net_device *ndev);
+
+int rxe_net_init(void);
+void rxe_net_exit(void);
+
+#endif /* RXE_NET_H */

drivers/infiniband/ulp/ipoib/ipoib_main.c

@@ -1967,8 +1967,7 @@ int ipoib_set_dev_features(struct ipoib_dev_priv *priv, struct ib_device *hca)
 	priv->hca_caps = hca->attrs.device_cap_flags;
 
 	if (priv->hca_caps & IB_DEVICE_UD_IP_CSUM) {
-		priv->dev->hw_features = NETIF_F_SG |
-			NETIF_F_IP_CSUM | NETIF_F_RXCSUM;
+		priv->dev->hw_features = NETIF_F_IP_CSUM | NETIF_F_RXCSUM;
 
 		if (priv->hca_caps & IB_DEVICE_UD_TSO)
 			priv->dev->hw_features |= NETIF_F_TSO;

drivers/infiniband/ulp/ipoib/ipoib_verbs.c

@@ -135,7 +135,8 @@ int ipoib_transport_dev_init(struct net_device *dev, struct ib_device *ca)
 		.cap = {
 			.max_send_wr  = ipoib_sendq_size,
 			.max_recv_wr  = ipoib_recvq_size,
-			.max_send_sge = 1,
+			.max_send_sge = min_t(u32, priv->ca->attrs.max_sge,
+					      MAX_SKB_FRAGS + 1),
 			.max_recv_sge = IPOIB_UD_RX_SG
 		},
 		.sq_sig_type = IB_SIGNAL_ALL_WR,
@@ -205,10 +206,6 @@ int ipoib_transport_dev_init(struct net_device *dev, struct ib_device *ca)
 	if (priv->hca_caps & IB_DEVICE_MANAGED_FLOW_STEERING)
 		init_attr.create_flags |= IB_QP_CREATE_NETIF_QP;
 
-	if (dev->features & NETIF_F_SG)
-		init_attr.cap.max_send_sge =
-			min_t(u32, priv->ca->attrs.max_sge, MAX_SKB_FRAGS + 1);
-
 	priv->qp = ib_create_qp(priv->pd, &init_attr);
 	if (IS_ERR(priv->qp)) {
 		printk(KERN_WARNING "%s: failed to create QP\n", ca->name);
@@ -234,6 +231,9 @@ int ipoib_transport_dev_init(struct net_device *dev, struct ib_device *ca)
 	priv->rx_wr.next = NULL;
 	priv->rx_wr.sg_list = priv->rx_sge;
 
+	if (init_attr.cap.max_send_sge > 1)
+		dev->features |= NETIF_F_SG;
+
 	priv->max_send_sge = init_attr.cap.max_send_sge;
 
 	return 0;

include/linux/mlx4/device.h

@@ -220,6 +220,7 @@ enum {
 	MLX4_DEV_CAP_FLAG2_LB_SRC_CHK           = 1ULL << 32,
 	MLX4_DEV_CAP_FLAG2_ROCE_V1_V2		= 1ULL <<  33,
 	MLX4_DEV_CAP_FLAG2_DMFS_UC_MC_SNIFFER   = 1ULL <<  34,
+	MLX4_DEV_CAP_FLAG2_DIAG_PER_PORT	= 1ULL <<  35,
 };
 
 enum {
@@ -1340,6 +1341,9 @@ enum {
 	VXLAN_STEER_BY_INNER_VLAN	= 1 << 4,
 };
 
+enum {
+	MLX4_OP_MOD_QUERY_TRANSPORT_CI_ERRORS = 0x2,
+};
 
 int mlx4_flow_steer_promisc_add(struct mlx4_dev *dev, u8 port, u32 qpn,
 				enum mlx4_net_trans_promisc_mode mode);
@@ -1380,6 +1384,9 @@ void mlx4_fmr_unmap(struct mlx4_dev *dev, struct mlx4_fmr *fmr,
 int mlx4_fmr_free(struct mlx4_dev *dev, struct mlx4_fmr *fmr);
 int mlx4_SYNC_TPT(struct mlx4_dev *dev);
 int mlx4_test_interrupts(struct mlx4_dev *dev);
+int mlx4_query_diag_counters(struct mlx4_dev *dev, u8 op_modifier,
+			     const u32 offset[], u32 value[],
+			     size_t array_len, u8 port);
 u32 mlx4_get_eqs_per_port(struct mlx4_dev *dev, u8 port);
 bool mlx4_is_eq_vector_valid(struct mlx4_dev *dev, u8 port, int vector);
 struct cpu_rmap *mlx4_get_cpu_rmap(struct mlx4_dev *dev, int port);

include/rdma/rdma_cm.h

@@ -333,11 +333,13 @@ int rdma_disconnect(struct rdma_cm_id *id);
  *   address.
  * @id: Communication identifier associated with the request.
  * @addr: Multicast address identifying the group to join.
+ * @join_state: Multicast JoinState bitmap requested by port.
+ *		Bitmap is based on IB_SA_MCMEMBER_REC_JOIN_STATE bits.
  * @context: User-defined context associated with the join request, returned
  * to the user through the private_data pointer in multicast events.
  */
 int rdma_join_multicast(struct rdma_cm_id *id, struct sockaddr *addr,
-			void *context);
+			u8 join_state, void *context);
 
 /**
  * rdma_leave_multicast - Leave the multicast group specified by the given