IB/core: Validate route when we init ah

In order to make sure API users don't try to use SGIDs which don't
conform to the routing table, validate the route before searching
the RoCE GID table.

Signed-off-by: Matan Barak <matanb@mellanox.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>

1 files changed, 67 insertions, 8 deletions

diff --git a/drivers/infiniband/core/sa_query.c b/drivers/infiniband/core/sa_query.c
index 270faaa2810f..e364a42db14d 100644
--- a/drivers/infiniband/core/sa_query.c
+++ b/drivers/infiniband/core/sa_query.c
@@ -49,7 +49,9 @@
 #include <net/netlink.h>
 #include <uapi/rdma/ib_user_sa.h>
 #include <rdma/ib_marshall.h>
+#include <rdma/ib_addr.h>
 #include "sa.h"
+#include "core_priv.h"
 
 MODULE_AUTHOR("Roland Dreier");
 MODULE_DESCRIPTION("InfiniBand subnet administration query support");
@@ -996,7 +998,8 @@ int ib_init_ah_from_path(struct ib_device *device, u8 port_num,
 {
 	int ret;
 	u16 gid_index;
-	int force_grh;
+	int use_roce;
+	struct net_device *ndev = NULL;
 
 	memset(ah_attr, 0, sizeof *ah_attr);
 	ah_attr->dlid = be16_to_cpu(rec->dlid);
@@ -1006,16 +1009,71 @@ int ib_init_ah_from_path(struct ib_device *device, u8 port_num,
 	ah_attr->port_num = port_num;
 	ah_attr->static_rate = rec->rate;
 
-	force_grh = rdma_cap_eth_ah(device, port_num);
+	use_roce = rdma_cap_eth_ah(device, port_num);
+
+	if (use_roce) {
+		struct net_device *idev;
+		struct net_device *resolved_dev;
+		struct rdma_dev_addr dev_addr = {.bound_dev_if = rec->ifindex,
+						 .net = rec->net ? rec->net :
+							 &init_net};
+		union {
+			struct sockaddr     _sockaddr;
+			struct sockaddr_in  _sockaddr_in;
+			struct sockaddr_in6 _sockaddr_in6;
+		} sgid_addr, dgid_addr;
+
+		if (!device->get_netdev)
+			return -EOPNOTSUPP;
+
+		rdma_gid2ip(&sgid_addr._sockaddr, &rec->sgid);
+		rdma_gid2ip(&dgid_addr._sockaddr, &rec->dgid);
+
+		/* validate the route */
+		ret = rdma_resolve_ip_route(&sgid_addr._sockaddr,
+					    &dgid_addr._sockaddr, &dev_addr);
+		if (ret)
+			return ret;
 
-	if (rec->hop_limit > 1 || force_grh) {
-		struct net_device *ndev = ib_get_ndev_from_path(rec);
+		if ((dev_addr.network == RDMA_NETWORK_IPV4 ||
+		     dev_addr.network == RDMA_NETWORK_IPV6) &&
+		    rec->gid_type != IB_GID_TYPE_ROCE_UDP_ENCAP)
+			return -EINVAL;
+
+		idev = device->get_netdev(device, port_num);
+		if (!idev)
+			return -ENODEV;
+
+		resolved_dev = dev_get_by_index(dev_addr.net,
+						dev_addr.bound_dev_if);
+		if (resolved_dev->flags & IFF_LOOPBACK) {
+			dev_put(resolved_dev);
+			resolved_dev = idev;
+			dev_hold(resolved_dev);
+		}
+		ndev = ib_get_ndev_from_path(rec);
+		rcu_read_lock();
+		if ((ndev && ndev != resolved_dev) ||
+		    (resolved_dev != idev &&
+		     !rdma_is_upper_dev_rcu(idev, resolved_dev)))
+			ret = -EHOSTUNREACH;
+		rcu_read_unlock();
+		dev_put(idev);
+		dev_put(resolved_dev);
+		if (ret) {
+			if (ndev)
+				dev_put(ndev);
+			return ret;
+		}
+	}
 
+	if (rec->hop_limit > 1 || use_roce) {
 		ah_attr->ah_flags = IB_AH_GRH;
 		ah_attr->grh.dgid = rec->dgid;
 
-		ret = ib_find_cached_gid(device, &rec->sgid, rec->gid_type, ndev,
-					 &port_num, &gid_index);
+		ret = ib_find_cached_gid_by_port(device, &rec->sgid,
+						 rec->gid_type, port_num, ndev,
+						 &gid_index);
 		if (ret) {
 			if (ndev)
 				dev_put(ndev);
@@ -1029,9 +1087,10 @@ int ib_init_ah_from_path(struct ib_device *device, u8 port_num,
 		if (ndev)
 			dev_put(ndev);
 	}
-	if (force_grh) {
+
+	if (use_roce)
 		memcpy(ah_attr->dmac, rec->dmac, ETH_ALEN);
-	}
+
 	return 0;
 }
 EXPORT_SYMBOL(ib_init_ah_from_path);