RDMA: iWARP Core Changes.

Modifications to the existing rdma header files, core files, drivers,
and ulp files to support iWARP, including:
 - Hook iWARP CM into the build system and use it in rdma_cm.
 - Convert enum ib_node_type to enum rdma_node_type, which includes
   the possibility of RDMA_NODE_RNIC, and update everything for this.
Signed-off-by: Tom Tucker <tom@opengridcomputing.com>
Signed-off-by: Steve Wise <swise@opengridcomputing.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>

drivers/infiniband/core/Makefile

 infiniband-$(CONFIG_INFINIBAND_ADDR_TRANS)	:= ib_addr.o rdma_cm.o
 
 obj-$(CONFIG_INFINIBAND) +=		ib_core.o ib_mad.o ib_sa.o \
-					ib_cm.o $(infiniband-y)
+					ib_cm.o iw_cm.o $(infiniband-y)
 obj-$(CONFIG_INFINIBAND_USER_MAD) +=	ib_umad.o
 obj-$(CONFIG_INFINIBAND_USER_ACCESS) +=	ib_uverbs.o ib_ucm.o
 
@@ -14,6 +14,8 @@ ib_sa-y :=			sa_query.o
 
 ib_cm-y :=			cm.o
 
+iw_cm-y :=			iwcm.o
+
 rdma_cm-y :=			cma.o
 
 ib_addr-y :=			addr.o

drivers/infiniband/core/addr.c

@@ -61,12 +61,15 @@ static LIST_HEAD(req_list);
 static DECLARE_WORK(work, process_req, NULL);
 static struct workqueue_struct *addr_wq;
 
-static int copy_addr(struct rdma_dev_addr *dev_addr, struct net_device *dev,
-		     unsigned char *dst_dev_addr)
+int rdma_copy_addr(struct rdma_dev_addr *dev_addr, struct net_device *dev,
+		     const unsigned char *dst_dev_addr)
 {
 	switch (dev->type) {
 	case ARPHRD_INFINIBAND:
-		dev_addr->dev_type = IB_NODE_CA;
+		dev_addr->dev_type = RDMA_NODE_IB_CA;
+		break;
+	case ARPHRD_ETHER:
+		dev_addr->dev_type = RDMA_NODE_RNIC;
 		break;
 	default:
 		return -EADDRNOTAVAIL;
@@ -78,6 +81,7 @@ static int copy_addr(struct rdma_dev_addr *dev_addr, struct net_device *dev,
 		memcpy(dev_addr->dst_dev_addr, dst_dev_addr, MAX_ADDR_LEN);
 	return 0;
 }
+EXPORT_SYMBOL(rdma_copy_addr);
 
 int rdma_translate_ip(struct sockaddr *addr, struct rdma_dev_addr *dev_addr)
 {
@@ -89,7 +93,7 @@ int rdma_translate_ip(struct sockaddr *addr, struct rdma_dev_addr *dev_addr)
 	if (!dev)
 		return -EADDRNOTAVAIL;
 
-	ret = copy_addr(dev_addr, dev, NULL);
+	ret = rdma_copy_addr(dev_addr, dev, NULL);
 	dev_put(dev);
 	return ret;
 }
@@ -161,7 +165,7 @@ static int addr_resolve_remote(struct sockaddr_in *src_in,
 
 	/* If the device does ARP internally, return 'done' */
 	if (rt->idev->dev->flags & IFF_NOARP) {
-		copy_addr(addr, rt->idev->dev, NULL);
+		rdma_copy_addr(addr, rt->idev->dev, NULL);
 		goto put;
 	}
 
@@ -181,7 +185,7 @@ static int addr_resolve_remote(struct sockaddr_in *src_in,
 		src_in->sin_addr.s_addr = rt->rt_src;
 	}
 
-	ret = copy_addr(addr, neigh->dev, neigh->ha);
+	ret = rdma_copy_addr(addr, neigh->dev, neigh->ha);
 release:
 	neigh_release(neigh);
 put:
@@ -245,7 +249,7 @@ static int addr_resolve_local(struct sockaddr_in *src_in,
 	if (ZERONET(src_ip)) {
 		src_in->sin_family = dst_in->sin_family;
 		src_in->sin_addr.s_addr = dst_ip;
-		ret = copy_addr(addr, dev, dev->dev_addr);
+		ret = rdma_copy_addr(addr, dev, dev->dev_addr);
 	} else if (LOOPBACK(src_ip)) {
 		ret = rdma_translate_ip((struct sockaddr *)dst_in, addr);
 		if (!ret)

drivers/infiniband/core/cache.c

@@ -62,12 +62,13 @@ struct ib_update_work {
 
 static inline int start_port(struct ib_device *device)
 {
-	return device->node_type == IB_NODE_SWITCH ? 0 : 1;
+	return (device->node_type == RDMA_NODE_IB_SWITCH) ? 0 : 1;
 }
 
 static inline int end_port(struct ib_device *device)
 {
-	return device->node_type == IB_NODE_SWITCH ? 0 : device->phys_port_cnt;
+	return (device->node_type == RDMA_NODE_IB_SWITCH) ?
+		0 : device->phys_port_cnt;
 }
 
 int ib_get_cached_gid(struct ib_device *device,

drivers/infiniband/core/cm.c

@@ -3280,6 +3280,9 @@ static void cm_add_one(struct ib_device *device)
 	int ret;
 	u8 i;
 
+	if (rdma_node_get_transport(device->node_type) != RDMA_TRANSPORT_IB)
+		return;
+
 	cm_dev = kmalloc(sizeof(*cm_dev) + sizeof(*port) *
 			 device->phys_port_cnt, GFP_KERNEL);
 	if (!cm_dev)

drivers/infiniband/core/device.c

@@ -505,7 +505,7 @@ int ib_query_port(struct ib_device *device,
 		  u8 port_num,
 		  struct ib_port_attr *port_attr)
 {
-	if (device->node_type == IB_NODE_SWITCH) {
+	if (device->node_type == RDMA_NODE_IB_SWITCH) {
 		if (port_num)
 			return -EINVAL;
 	} else if (port_num < 1 || port_num > device->phys_port_cnt)
@@ -580,7 +580,7 @@ int ib_modify_port(struct ib_device *device,
 		   u8 port_num, int port_modify_mask,
 		   struct ib_port_modify *port_modify)
 {
-	if (device->node_type == IB_NODE_SWITCH) {
+	if (device->node_type == RDMA_NODE_IB_SWITCH) {
 		if (port_num)
 			return -EINVAL;
 	} else if (port_num < 1 || port_num > device->phys_port_cnt)

drivers/infiniband/core/mad.c

@@ -2876,7 +2876,10 @@ static void ib_mad_init_device(struct ib_device *device)
 {
 	int start, end, i;
 
-	if (device->node_type == IB_NODE_SWITCH) {
+	if (rdma_node_get_transport(device->node_type) != RDMA_TRANSPORT_IB)
+		return;
+
+	if (device->node_type == RDMA_NODE_IB_SWITCH) {
 		start = 0;
 		end   = 0;
 	} else {
@@ -2923,7 +2926,7 @@ static void ib_mad_remove_device(struct ib_device *device)
 {
 	int i, num_ports, cur_port;
 
-	if (device->node_type == IB_NODE_SWITCH) {
+	if (device->node_type == RDMA_NODE_IB_SWITCH) {
 		num_ports = 1;
 		cur_port = 0;
 	} else {

drivers/infiniband/core/sa_query.c

@@ -919,7 +919,10 @@ static void ib_sa_add_one(struct ib_device *device)
 	struct ib_sa_device *sa_dev;
 	int s, e, i;
 
-	if (device->node_type == IB_NODE_SWITCH)
+	if (rdma_node_get_transport(device->node_type) != RDMA_TRANSPORT_IB)
+		return;
+
+	if (device->node_type == RDMA_NODE_IB_SWITCH)
 		s = e = 0;
 	else {
 		s = 1;

drivers/infiniband/core/smi.c

@@ -64,7 +64,7 @@ int smi_handle_dr_smp_send(struct ib_smp *smp,
 
 		/* C14-9:2 */
 		if (hop_ptr && hop_ptr < hop_cnt) {
-			if (node_type != IB_NODE_SWITCH)
+			if (node_type != RDMA_NODE_IB_SWITCH)
 				return 0;
 
 			/* smp->return_path set when received */
@@ -77,7 +77,7 @@ int smi_handle_dr_smp_send(struct ib_smp *smp,
 		if (hop_ptr == hop_cnt) {
 			/* smp->return_path set when received */
 			smp->hop_ptr++;
-			return (node_type == IB_NODE_SWITCH ||
+			return (node_type == RDMA_NODE_IB_SWITCH ||
 				smp->dr_dlid == IB_LID_PERMISSIVE);
 		}
 
@@ -95,7 +95,7 @@ int smi_handle_dr_smp_send(struct ib_smp *smp,
 
 		/* C14-13:2 */
 		if (2 <= hop_ptr && hop_ptr <= hop_cnt) {
-			if (node_type != IB_NODE_SWITCH)
+			if (node_type != RDMA_NODE_IB_SWITCH)
 				return 0;
 
 			smp->hop_ptr--;
@@ -107,7 +107,7 @@ int smi_handle_dr_smp_send(struct ib_smp *smp,
 		if (hop_ptr == 1) {
 			smp->hop_ptr--;
 			/* C14-13:3 -- SMPs destined for SM shouldn't be here */
-			return (node_type == IB_NODE_SWITCH ||
+			return (node_type == RDMA_NODE_IB_SWITCH ||
 				smp->dr_slid == IB_LID_PERMISSIVE);
 		}
 
@@ -142,7 +142,7 @@ int smi_handle_dr_smp_recv(struct ib_smp *smp,
 
 		/* C14-9:2 -- intermediate hop */
 		if (hop_ptr && hop_ptr < hop_cnt) {
-			if (node_type != IB_NODE_SWITCH)
+			if (node_type != RDMA_NODE_IB_SWITCH)
 				return 0;
 
 			smp->return_path[hop_ptr] = port_num;
@@ -156,7 +156,7 @@ int smi_handle_dr_smp_recv(struct ib_smp *smp,
 				smp->return_path[hop_ptr] = port_num;
 			/* smp->hop_ptr updated when sending */
 
-			return (node_type == IB_NODE_SWITCH ||
+			return (node_type == RDMA_NODE_IB_SWITCH ||
 				smp->dr_dlid == IB_LID_PERMISSIVE);
 		}
 
@@ -175,7 +175,7 @@ int smi_handle_dr_smp_recv(struct ib_smp *smp,
 
 		/* C14-13:2 */
 		if (2 <= hop_ptr && hop_ptr <= hop_cnt) {
-			if (node_type != IB_NODE_SWITCH)
+			if (node_type != RDMA_NODE_IB_SWITCH)
 				return 0;
 
 			/* smp->hop_ptr updated when sending */
@@ -190,7 +190,7 @@ int smi_handle_dr_smp_recv(struct ib_smp *smp,
 				return 1;
 			}
 			/* smp->hop_ptr updated when sending */
-			return (node_type == IB_NODE_SWITCH);
+			return (node_type == RDMA_NODE_IB_SWITCH);
 		}
 
 		/* C14-13:4 -- hop_ptr = 0 -> give to SM */

drivers/infiniband/core/sysfs.c

@@ -589,9 +589,10 @@ static ssize_t show_node_type(struct class_device *cdev, char *buf)
 		return -ENODEV;
 
 	switch (dev->node_type) {
-	case IB_NODE_CA:     return sprintf(buf, "%d: CA\n", dev->node_type);
-	case IB_NODE_SWITCH: return sprintf(buf, "%d: switch\n", dev->node_type);
-	case IB_NODE_ROUTER: return sprintf(buf, "%d: router\n", dev->node_type);
+	case RDMA_NODE_IB_CA:	  return sprintf(buf, "%d: CA\n", dev->node_type);
+	case RDMA_NODE_RNIC:	  return sprintf(buf, "%d: RNIC\n", dev->node_type);
+	case RDMA_NODE_IB_SWITCH: return sprintf(buf, "%d: switch\n", dev->node_type);
+	case RDMA_NODE_IB_ROUTER: return sprintf(buf, "%d: router\n", dev->node_type);
 	default:		  return sprintf(buf, "%d: <unknown>\n", dev->node_type);
 	}
 }
@@ -708,7 +709,7 @@ int ib_device_register_sysfs(struct ib_device *device)
 	if (ret)
 		goto err_put;
 
-	if (device->node_type == IB_NODE_SWITCH) {
+	if (device->node_type == RDMA_NODE_IB_SWITCH) {
 		ret = add_port(device, 0);
 		if (ret)
 			goto err_put;

drivers/infiniband/core/ucm.c

@@ -1247,7 +1247,8 @@ static void ib_ucm_add_one(struct ib_device *device)
 {
 	struct ib_ucm_device *ucm_dev;
 
-	if (!device->alloc_ucontext)
+	if (!device->alloc_ucontext ||
+	    rdma_node_get_transport(device->node_type) != RDMA_TRANSPORT_IB)
 		return;
 
 	ucm_dev = kzalloc(sizeof *ucm_dev, GFP_KERNEL);

drivers/infiniband/core/user_mad.c

@@ -1032,7 +1032,10 @@ static void ib_umad_add_one(struct ib_device *device)
 	struct ib_umad_device *umad_dev;
 	int s, e, i;
 
-	if (device->node_type == IB_NODE_SWITCH)
+	if (rdma_node_get_transport(device->node_type) != RDMA_TRANSPORT_IB)
+		return;
+
+	if (device->node_type == RDMA_NODE_IB_SWITCH)
 		s = e = 0;
 	else {
 		s = 1;

drivers/infiniband/core/verbs.c

@@ -79,6 +79,23 @@ enum ib_rate mult_to_ib_rate(int mult)
 }
 EXPORT_SYMBOL(mult_to_ib_rate);
 
+enum rdma_transport_type
+rdma_node_get_transport(enum rdma_node_type node_type)
+{
+	switch (node_type) {
+	case RDMA_NODE_IB_CA:
+	case RDMA_NODE_IB_SWITCH:
+	case RDMA_NODE_IB_ROUTER:
+		return RDMA_TRANSPORT_IB;
+	case RDMA_NODE_RNIC:
+		return RDMA_TRANSPORT_IWARP;
+	default:
+		BUG();
+		return 0;
+	}
+}
+EXPORT_SYMBOL(rdma_node_get_transport);
+
 /* Protection domains */
 
 struct ib_pd *ib_alloc_pd(struct ib_device *device)

drivers/infiniband/hw/ehca/ehca_main.c

@@ -269,7 +269,7 @@ int ehca_register_device(struct ehca_shca *shca)
 		(1ull << IB_USER_VERBS_CMD_ATTACH_MCAST)	|
 		(1ull << IB_USER_VERBS_CMD_DETACH_MCAST);
 
-	shca->ib_device.node_type           = IB_NODE_CA;
+	shca->ib_device.node_type           = RDMA_NODE_IB_CA;
 	shca->ib_device.phys_port_cnt       = shca->num_ports;
 	shca->ib_device.dma_device          = &shca->ibmebus_dev->ofdev.dev;
 	shca->ib_device.query_device        = ehca_query_device;

drivers/infiniband/hw/ipath/ipath_verbs.c

@@ -1538,7 +1538,7 @@ int ipath_register_ib_device(struct ipath_devdata *dd)
 		(1ull << IB_USER_VERBS_CMD_QUERY_SRQ)		|
 		(1ull << IB_USER_VERBS_CMD_DESTROY_SRQ)		|
 		(1ull << IB_USER_VERBS_CMD_POST_SRQ_RECV);
-	dev->node_type = IB_NODE_CA;
+	dev->node_type = RDMA_NODE_IB_CA;
 	dev->phys_port_cnt = 1;
 	dev->dma_device = &dd->pcidev->dev;
 	dev->class_dev.dev = dev->dma_device;

drivers/infiniband/hw/mthca/mthca_provider.c

@@ -1288,7 +1288,7 @@ int mthca_register_device(struct mthca_dev *dev)
 		(1ull << IB_USER_VERBS_CMD_DESTROY_QP)		|
 		(1ull << IB_USER_VERBS_CMD_ATTACH_MCAST)	|
 		(1ull << IB_USER_VERBS_CMD_DETACH_MCAST);
-	dev->ib_dev.node_type            = IB_NODE_CA;
+	dev->ib_dev.node_type            = RDMA_NODE_IB_CA;
 	dev->ib_dev.phys_port_cnt        = dev->limits.num_ports;
 	dev->ib_dev.dma_device           = &dev->pdev->dev;
 	dev->ib_dev.class_dev.dev        = &dev->pdev->dev;

drivers/infiniband/ulp/ipoib/ipoib_main.c

@@ -1111,13 +1111,16 @@ static void ipoib_add_one(struct ib_device *device)
 	struct ipoib_dev_priv *priv;
 	int s, e, p;
 
+	if (rdma_node_get_transport(device->node_type) != RDMA_TRANSPORT_IB)
+		return;
+
 	dev_list = kmalloc(sizeof *dev_list, GFP_KERNEL);
 	if (!dev_list)
 		return;
 
 	INIT_LIST_HEAD(dev_list);
 
-	if (device->node_type == IB_NODE_SWITCH) {
+	if (device->node_type == RDMA_NODE_IB_SWITCH) {
 		s = 0;
 		e = 0;
 	} else {
@@ -1141,6 +1144,9 @@ static void ipoib_remove_one(struct ib_device *device)
 	struct ipoib_dev_priv *priv, *tmp;
 	struct list_head *dev_list;
 
+	if (rdma_node_get_transport(device->node_type) != RDMA_TRANSPORT_IB)
+		return;
+
 	dev_list = ib_get_client_data(device, &ipoib_client);
 
 	list_for_each_entry_safe(priv, tmp, dev_list, list) {

drivers/infiniband/ulp/srp/ib_srp.c

@@ -1899,7 +1899,7 @@ static void srp_add_one(struct ib_device *device)
 	if (IS_ERR(srp_dev->fmr_pool))
 		srp_dev->fmr_pool = NULL;
 
-	if (device->node_type == IB_NODE_SWITCH) {
+	if (device->node_type == RDMA_NODE_IB_SWITCH) {
 		s = 0;
 		e = 0;
 	} else {

include/rdma/ib_addr.h

@@ -40,7 +40,7 @@ struct rdma_dev_addr {
 	unsigned char src_dev_addr[MAX_ADDR_LEN];
 	unsigned char dst_dev_addr[MAX_ADDR_LEN];
 	unsigned char broadcast[MAX_ADDR_LEN];
-	enum ib_node_type dev_type;
+	enum rdma_node_type dev_type;
 };
 
 /**
@@ -72,6 +72,9 @@ int rdma_resolve_ip(struct sockaddr *src_addr, struct sockaddr *dst_addr,
 
 void rdma_addr_cancel(struct rdma_dev_addr *addr);
 
+int rdma_copy_addr(struct rdma_dev_addr *dev_addr, struct net_device *dev,
+	      const unsigned char *dst_dev_addr);
+
 static inline int ip_addr_size(struct sockaddr *addr)
 {
 	return addr->sa_family == AF_INET6 ?
@@ -113,4 +116,16 @@ static inline void ib_addr_set_dgid(struct rdma_dev_addr *dev_addr,
 	memcpy(dev_addr->dst_dev_addr + 4, gid, sizeof *gid);
 }
 
+static inline void iw_addr_get_sgid(struct rdma_dev_addr *dev_addr,
+				    union ib_gid *gid)
+{
+	memcpy(gid, dev_addr->src_dev_addr, sizeof *gid);
+}
+
+static inline void iw_addr_get_dgid(struct rdma_dev_addr *dev_addr,
+				    union ib_gid *gid)
+{
+	memcpy(gid, dev_addr->dst_dev_addr, sizeof *gid);
+}
+
 #endif /* IB_ADDR_H */

include/rdma/ib_verbs.h

@@ -56,12 +56,22 @@ union ib_gid {
 	} global;
 };
 
-enum ib_node_type {
-	IB_NODE_CA 	= 1,
-	IB_NODE_SWITCH,
-	IB_NODE_ROUTER
+enum rdma_node_type {
+	/* IB values map to NodeInfo:NodeType. */
+	RDMA_NODE_IB_CA 	= 1,
+	RDMA_NODE_IB_SWITCH,
+	RDMA_NODE_IB_ROUTER,
+	RDMA_NODE_RNIC
 };
 
+enum rdma_transport_type {
+	RDMA_TRANSPORT_IB,
+	RDMA_TRANSPORT_IWARP
+};
+
+enum rdma_transport_type
+rdma_node_get_transport(enum rdma_node_type node_type) __attribute_const__;
+
 enum ib_device_cap_flags {
 	IB_DEVICE_RESIZE_MAX_WR		= 1,
 	IB_DEVICE_BAD_PKEY_CNTR		= (1<<1),
@@ -78,6 +88,9 @@ enum ib_device_cap_flags {
 	IB_DEVICE_RC_RNR_NAK_GEN	= (1<<12),
 	IB_DEVICE_SRQ_RESIZE		= (1<<13),
 	IB_DEVICE_N_NOTIFY_CQ		= (1<<14),
+	IB_DEVICE_ZERO_STAG		= (1<<15),
+	IB_DEVICE_SEND_W_INV		= (1<<16),
+	IB_DEVICE_MEM_WINDOW		= (1<<17)
 };
 
 enum ib_atomic_cap {
@@ -835,6 +848,8 @@ struct ib_cache {
 	u8                     *lmc_cache;
 };
 
+struct iw_cm_verbs;
+
 struct ib_device {
 	struct device                *dma_device;
 
@@ -851,6 +866,8 @@ struct ib_device {
 
 	u32                           flags;
 
+	struct iw_cm_verbs	     *iwcm;
+
 	int		           (*query_device)(struct ib_device *device,
 						   struct ib_device_attr *device_attr);
 	int		           (*query_port)(struct ib_device *device,