More fixes for using IPv6 addresses with RDMA:

- Added check that the SCOPE ID is only restored for IPv6 linklocal
  addresses.

- Changes made by r237263 in the "cma_bind_addr()" function did not
  check if the socket address was of type IPv6 and used the IPv4
  socket address for IPv6 addresses. This caused the function to
  fail. Fixed this.

- In the "rdma_gid2ip()" function and some other places the "sin6_len"
  and "sin6_scope_id" fields were not set for IPv6 socket
  addresses. Fixed this.

- The scope ID is not stored as part of the GID entries and must be
  passed as an argument to "rdma_gid2ip()".

- Added new method to "struct ib_device" which returns a pointer to
  the network interface which belongs to the given infiniband
  device. This is needed to be able to get the scope ID for IPv6
  addresses via the associated ethernet interface.

- Added convenience function, "rdma_get_ipv6_scope_id()", to get the
  scope ID for IPv6 addresses.

- Implemented new "get_netdev" method for mlx4ib. Other IB controller
  drivers which want to support IPv6 addresses needs to implement this
  aswell.

- Bumped the FreeBSD version due to changing "struct ib_device".

Sponsored by:	Mellanox Technologies
MFC after:	1 week

sys/ofed/drivers/infiniband/core/addr.c

@@ -43,6 +43,7 @@
 #include <net/netevent.h>
 #include <rdma/ib_addr.h>
 #include <netinet/if_ether.h>
+#include <netinet6/scope6_var.h>
 
 
 MODULE_AUTHOR("Sean Hefty");
@@ -417,7 +418,8 @@ static int addr_resolve(struct sockaddr *src_in,
 #ifdef INET6
 	if (scope_id < 256) {
 		sin6 = (struct sockaddr_in6 *)src_in;
-		SCOPE_ID_RESTORE(scope_id, sin6);
+		if (IN6_IS_SCOPE_LINKLOCAL(&sin6->sin6_addr))
+			SCOPE_ID_RESTORE(scope_id, sin6);
 		sin6 = (struct sockaddr_in6 *)dst_in;
 		SCOPE_ID_RESTORE(scope_id, sin6);
 	}
@@ -554,7 +556,7 @@ static void resolve_cb(int status, struct sockaddr *src_addr,
 }
 
 int rdma_addr_find_dmac_by_grh(union ib_gid *sgid, union ib_gid *dgid, u8 *dmac,
-			       u16 *vlan_id)
+			       u16 *vlan_id, u32 scope_id)
 {
 	int ret = 0;
 	struct rdma_dev_addr dev_addr;
@@ -568,11 +570,11 @@ int rdma_addr_find_dmac_by_grh(union ib_gid *sgid, union ib_gid *dgid, u8 *dmac,
 	} sgid_addr, dgid_addr;
 
 
-	ret = rdma_gid2ip(&sgid_addr._sockaddr, sgid);
+	ret = rdma_gid2ip(&sgid_addr._sockaddr, sgid, scope_id);
 	if (ret)
 		return ret;
 
-	ret = rdma_gid2ip(&dgid_addr._sockaddr, dgid);
+	ret = rdma_gid2ip(&dgid_addr._sockaddr, dgid, scope_id);
 	if (ret)
 		return ret;
 
@@ -598,7 +600,23 @@ int rdma_addr_find_dmac_by_grh(union ib_gid *sgid, union ib_gid *dgid, u8 *dmac,
 }
 EXPORT_SYMBOL(rdma_addr_find_dmac_by_grh);
 
-int rdma_addr_find_smac_by_sgid(union ib_gid *sgid, u8 *smac, u16 *vlan_id)
+u32 rdma_get_ipv6_scope_id(struct ib_device *ib, u8 port_num)
+{
+#ifdef INET6
+	struct ifnet *ifp;
+	if (ib->get_netdev == NULL)
+		return (-1U);
+	ifp = ib->get_netdev(ib, port_num);
+	if (ifp == NULL)
+		return (-1U);
+	return (in6_getscopezone(ifp, IPV6_ADDR_SCOPE_LINKLOCAL));
+#else
+	return (-1U);
+#endif
+}
+
+int rdma_addr_find_smac_by_sgid(union ib_gid *sgid, u8 *smac, u16 *vlan_id,
+    u32 scope_id)
 {
 	int ret = 0;
 	struct rdma_dev_addr dev_addr;
@@ -608,8 +626,7 @@ int rdma_addr_find_smac_by_sgid(union ib_gid *sgid, u8 *smac, u16 *vlan_id)
 		struct sockaddr_in6 _sockaddr_in6;
 	} gid_addr;
 
-	ret = rdma_gid2ip(&gid_addr._sockaddr, sgid);
-
+	ret = rdma_gid2ip(&gid_addr._sockaddr, sgid, scope_id);
 	if (ret)
 		return ret;
 	memset(&dev_addr, 0, sizeof(dev_addr));

sys/ofed/drivers/infiniband/core/cma.c

@@ -689,8 +689,11 @@ static int cma_modify_qp_rtr(struct rdma_id_private *id_priv,
 	    == RDMA_TRANSPORT_IB &&
 	    rdma_port_get_link_layer(id_priv->id.device, id_priv->id.port_num)
 	    == IB_LINK_LAYER_ETHERNET) {
-		ret = rdma_addr_find_smac_by_sgid(&sgid, qp_attr.smac, NULL);
+		u32 scope_id = rdma_get_ipv6_scope_id(id_priv->id.device,
+		    id_priv->id.port_num);
 
+		ret = rdma_addr_find_smac_by_sgid(&sgid, qp_attr.smac, NULL,
+		    scope_id);
 		if (ret)
 			goto out;
 	}
@@ -931,11 +934,13 @@ static void cma_save_net_info(struct rdma_addr *addr,
 		ip4->sin_family = listen4->sin_family;
 		ip4->sin_addr.s_addr = dst->ip4.addr;
 		ip4->sin_port = listen4->sin_port;
+		ip4->sin_len = sizeof(struct sockaddr_in);
 
 		ip4 = (struct sockaddr_in *) &addr->dst_addr;
 		ip4->sin_family = listen4->sin_family;
 		ip4->sin_addr.s_addr = src->ip4.addr;
 		ip4->sin_port = port;
+		ip4->sin_len = sizeof(struct sockaddr_in);
 		break;
 	case 6:
 		listen6 = (struct sockaddr_in6 *) &listen_addr->src_addr;
@@ -943,11 +948,15 @@ static void cma_save_net_info(struct rdma_addr *addr,
 		ip6->sin6_family = listen6->sin6_family;
 		ip6->sin6_addr = dst->ip6;
 		ip6->sin6_port = listen6->sin6_port;
+		ip6->sin6_len = sizeof(struct sockaddr_in6);
+		ip6->sin6_scope_id = listen6->sin6_scope_id;
 
 		ip6 = (struct sockaddr_in6 *) &addr->dst_addr;
 		ip6->sin6_family = listen6->sin6_family;
 		ip6->sin6_addr = src->ip6;
 		ip6->sin6_port = port;
+		ip6->sin6_len = sizeof(struct sockaddr_in6);
+		ip6->sin6_scope_id = listen6->sin6_scope_id;
 		break;
 	default:
 		break;
@@ -1431,16 +1440,19 @@ static int cma_req_handler(struct ib_cm_id *cm_id, struct ib_cm_event *ib_event)
 		goto err3;
 
 	if (is_iboe && !is_sidr) {
+		u32 scope_id = rdma_get_ipv6_scope_id(cm_id->device,
+		    ib_event->param.req_rcvd.port);
+
 		if (ib_event->param.req_rcvd.primary_path != NULL)
 			rdma_addr_find_smac_by_sgid(
 				&ib_event->param.req_rcvd.primary_path->sgid,
-				psmac, NULL);
+				psmac, NULL, scope_id);
 		else
 			psmac = NULL;
 		if (ib_event->param.req_rcvd.alternate_path != NULL)
 			rdma_addr_find_smac_by_sgid(
 				&ib_event->param.req_rcvd.alternate_path->sgid,
-				palt_smac, NULL);
+				palt_smac, NULL, scope_id);
 		else
 			palt_smac = NULL;
 	}
@@ -2296,22 +2308,47 @@ static int cma_bind_addr(struct rdma_cm_id *id, struct sockaddr *src_addr,
 {
 	if (!src_addr || !src_addr->sa_family) {
 		src_addr = (struct sockaddr *) &id->route.addr.src_addr;
-		if ((src_addr->sa_family = dst_addr->sa_family) == AF_INET6) {
+		src_addr->sa_family = dst_addr->sa_family;
+#ifdef INET6
+		if (dst_addr->sa_family == AF_INET6) {
 			((struct sockaddr_in6 *) src_addr)->sin6_scope_id =
 				((struct sockaddr_in6 *) dst_addr)->sin6_scope_id;
 		}
+#endif
 	}
 	if (!cma_any_addr(src_addr))
 		return rdma_bind_addr(id, src_addr);
 	else {
-		struct sockaddr_in addr_in;
+		union {
+#ifdef INET
+			struct sockaddr_in in;
+#endif
+#ifdef INET6
+			struct sockaddr_in6 in6;
+#endif
+		} addr;
 
-		memset(&addr_in, 0, sizeof addr_in);
-		addr_in.sin_family = dst_addr->sa_family;
-		addr_in.sin_len = sizeof addr_in;
-		return rdma_bind_addr(id, (struct sockaddr *) &addr_in);
+		switch(dst_addr->sa_family) {
+#ifdef INET
+		case AF_INET:
+			memset(&addr.in, 0, sizeof(addr.in));
+			addr.in.sin_family = dst_addr->sa_family;
+			addr.in.sin_len = sizeof(addr.in);
+			return rdma_bind_addr(id, (struct sockaddr *)&addr.in);
+#endif
+#ifdef INET6
+		case AF_INET6:
+			memset(&addr.in6, 0, sizeof(addr.in6));
+			addr.in6.sin6_family = dst_addr->sa_family;
+			addr.in6.sin6_len = sizeof(addr.in6);
+			addr.in6.sin6_scope_id =
+			    ((struct sockaddr_in6 *)dst_addr)->sin6_scope_id;
+			return rdma_bind_addr(id, (struct sockaddr *)&addr.in6);
+#endif
+		default:
+			return -EINVAL;
+		}
 	}
-
 }
 
 int rdma_resolve_addr(struct rdma_cm_id *id, struct sockaddr *src_addr,

sys/ofed/drivers/infiniband/core/uverbs_cmd.c

@@ -2094,11 +2094,11 @@ static ssize_t __uverbs_modify_qp(struct ib_uverbs_file *file,
 		} else {
 			ret = rdma_addr_find_dmac_by_grh(&sgid, dgid,
 							 attr->ah_attr.dmac,
-							 &attr->vlan_id);
+							 &attr->vlan_id, -1U);
 			if (ret)
 				goto out;
 			ret = rdma_addr_find_smac_by_sgid(&sgid, attr->smac,
-							  NULL);
+							  NULL, -1U);
 			if (ret)
 				goto out;
 		}

sys/ofed/drivers/infiniband/core/verbs.c

@@ -207,8 +207,10 @@ int ib_init_ah_from_wc(struct ib_device *device, u8 port_num, struct ib_wc *wc,
 			memcpy(ah_attr->dmac, wc->smac, ETH_ALEN);
 			ah_attr->vlan_id = wc->vlan_id;
 		} else {
+			u32 scope_id = rdma_get_ipv6_scope_id(device, port_num);
 			ret = rdma_addr_find_dmac_by_grh(&grh->dgid, &grh->sgid,
-					ah_attr->dmac, &ah_attr->vlan_id);
+					 ah_attr->dmac, &ah_attr->vlan_id,
+					 scope_id);
 			if (ret)
 				return ret;
 		}

sys/ofed/drivers/infiniband/hw/mlx4/main.c

@@ -1620,6 +1620,12 @@ static void update_gids_task(struct work_struct *work)
 	kfree(gw);
 }
 
+static struct net_device *mlx4_ib_get_netdev(struct ib_device *device, u8 port_num)
+{
+	struct mlx4_ib_dev *ibdev = to_mdev(device);
+	return mlx4_get_protocol_dev(ibdev->dev, MLX4_PROT_ETH, port_num);
+}
+
 static void reset_gids_task(struct work_struct *work)
 {
 	struct update_gid_work *gw =
@@ -2353,6 +2359,7 @@ static void *mlx4_ib_add(struct mlx4_dev *dev)
 	ibdev->ib_dev.attach_mcast	= mlx4_ib_mcg_attach;
 	ibdev->ib_dev.detach_mcast	= mlx4_ib_mcg_detach;
 	ibdev->ib_dev.process_mad	= mlx4_ib_process_mad;
+	ibdev->ib_dev.get_netdev	= mlx4_ib_get_netdev;
 	ibdev->ib_dev.ioctl		= mlx4_ib_ioctl;
 	ibdev->ib_dev.query_values	= mlx4_ib_query_values;
 

sys/ofed/include/rdma/ib_addr.h

@@ -105,9 +105,10 @@ 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);
-int rdma_addr_find_smac_by_sgid(union ib_gid *sgid, u8 *smac, u16 *vlan_id);
+int rdma_addr_find_smac_by_sgid(union ib_gid *sgid, u8 *smac, u16 *vlan_id,
+				u32 scope_id);
 int rdma_addr_find_dmac_by_grh(union ib_gid *sgid, union ib_gid *dgid, u8 *smac,
-			       u16 *vlan_id);
+			       u16 *vlan_id, u32 scope_id);
 
 static inline int ip_addr_size(struct sockaddr *addr)
 {
@@ -164,7 +165,8 @@ static inline int rdma_ip2gid(struct sockaddr *addr, union ib_gid *gid)
 }
 
 /* Important - sockaddr should be a union of sockaddr_in and sockaddr_in6 */
-static inline int rdma_gid2ip(struct sockaddr *out, union ib_gid *gid)
+static inline int rdma_gid2ip(struct sockaddr *out, union ib_gid *gid,
+    uint32_t scope_id)
 {
 	if (ipv6_addr_v4mapped((struct in6_addr *)gid)) {
 		struct sockaddr_in *out_in = (struct sockaddr_in *)out;
@@ -175,12 +177,18 @@ static inline int rdma_gid2ip(struct sockaddr *out, union ib_gid *gid)
 	} else {
 		struct sockaddr_in6 *out_in = (struct sockaddr_in6 *)out;
 		memset(out_in, 0, sizeof(*out_in));
+		out_in->sin6_len = sizeof(*out_in);
 		out_in->sin6_family = AF_INET6;
 		memcpy(&out_in->sin6_addr.s6_addr, gid->raw, 16);
+		if (scope_id < 256 &&
+		    IN6_IS_SCOPE_LINKLOCAL(&out_in->sin6_addr))
+			out_in->sin6_scope_id = scope_id;
 	}
 	return 0;
 }
 
+u32 rdma_get_ipv6_scope_id(struct ib_device *ib, u8 port_num);
+
 /* This func is called only in loopback ip address (127.0.0.1)
  * case in which sgid is not relevant
  */

sys/ofed/include/rdma/ib_verbs.h

@@ -1655,6 +1655,14 @@ struct ib_device {
 						 struct ib_port_attr *port_attr);
 	enum rdma_link_layer	   (*get_link_layer)(struct ib_device *device,
 						     u8 port_num);
+	/* When calling get_netdev, the HW vendor's driver should return the
+	 * net device of device @device at port @port_num. The function
+	 * is called in rtnl_lock. The HW vendor's device driver must guarantee
+	 * to return NULL before the net device has reached
+	 * NETDEV_UNREGISTER_FINAL state.
+	 */
+	struct net_device	*(*get_netdev)(struct ib_device *device,
+					       u8 port_num);
 	int		           (*query_gid)(struct ib_device *device,
 						u8 port_num, int index,
 						union ib_gid *gid);

sys/sys/param.h

@@ -58,7 +58,7 @@
  *		in the range 5 to 9.
  */
 #undef __FreeBSD_version
-#define __FreeBSD_version 1100105	/* Master, propagated to newvers */
+#define __FreeBSD_version 1100106	/* Master, propagated to newvers */
 
 /*
  * __FreeBSD_kernel__ indicates that this system uses the kernel of FreeBSD,