usnic: use fi_getname in newer libfabric

When using an external libfabric (or really any libfabric newer than
libfabric commit 607e863), we must use fi_getname to determine the local
port of our endpoint.  Without this fix, OMPI will hang endlessly
while retransmitting packets to port 0 on the remote host.

opal/mca/btl/usnic/btl_usnic_module.c

@@ -1393,7 +1393,9 @@ static int create_ep(opal_btl_usnic_module_t* module,
                             struct opal_btl_usnic_channel_t *channel)
 {
     int rc;
+    struct sockaddr *sa;
     struct sockaddr_in *sin;
+    size_t addrlen;
     struct fi_info *hint;
 
     hint = fi_dupinfo(module->fabric_info);
@@ -1436,6 +1438,21 @@ static int create_ep(opal_btl_usnic_module_t* module,
         channel->info->caps &= ~(1ULL << 63);
     }
 
+    /* all of the OMPI code assumes IPv4, but some versions of libfabric will
+     * return FI_SOCKADDR instead of FI_SOCKADDR_IN, so we need to do a little
+     * bit of sanity checking */
+    assert(FI_SOCKADDR_IN == channel->info->addr_format ||
+           FI_SOCKADDR == channel->info->addr_format);
+    if (FI_SOCKADDR == channel->info->addr_format) {
+        sa = (struct sockaddr *)channel->info->src_addr;
+        assert(AF_INET == sa->sa_family);
+    }
+    sin = (struct sockaddr_in *)channel->info->src_addr;
+    assert(sizeof(struct sockaddr_in) == channel->info->src_addrlen);
+
+    /* no matter the version of libfabric, this should hold */
+    assert(0 == sin->sin_port);
+
     rc = fi_endpoint(module->domain, channel->info, &channel->ep, NULL);
     if (0 != rc || NULL == channel->ep) {
         opal_show_help("help-mpi-btl-usnic.txt",
@@ -1496,6 +1513,30 @@ static int create_ep(opal_btl_usnic_module_t* module,
         return OPAL_ERR_OUT_OF_RESOURCE;
     }
 
+    /* Immediately after libfabric v1.0 was released, we implemented support
+     * for fi_getname and changed the behavior of fi_endpoint w.r.t. setting
+     * the src_addr field of the fi_info struct passed in.  Before the change
+     * fi_endpoint would set the src_addr field, including the sin_port field
+     * but calling fi_getname would return -FI_ENOSYS.  Afterwards the address
+     * would not be touched relative to whatever was set by fi_getinfo.  So we
+     * must call fi_getname in that case.
+     */
+    if (0 == sin->sin_port) {
+        addrlen = sizeof(struct sockaddr_in);
+        rc = fi_getname(&channel->ep->fid, channel->info->src_addr, &addrlen);
+        if (0 != rc) {
+            opal_show_help("help-mpi-btl-usnic.txt",
+                           "internal error during init",
+                           true,
+                           opal_process_info.nodename,
+                           module->fabric_info->fabric_attr->name,
+                           "fi_getname() failed", __FILE__, __LINE__,
+                           rc, fi_strerror(-rc));
+            return OPAL_ERR_OUT_OF_RESOURCE;
+        }
+        assert(0 != sin->sin_port);
+    }
+
     /* actual sizes */
     channel->chan_rd_num = channel->info->rx_attr->size;
     channel->chan_sd_num = channel->info->tx_attr->size;

opal/mca/btl/usnic/btl_usnic_module.h

@@ -25,6 +25,7 @@
 #define OPAL_BTL_USNIC_MODULE_H
 
 #include <rdma/fabric.h>
+#include <rdma/fi_cm.h>
 #include <rdma/fi_eq.h>
 #include <rdma/fi_endpoint.h>
 #include <rdma/fi_errno.h>