sys/net/gnrc/tcp: fix gnrc_tcp_open() netif handling

gnrc_tcp_open() previously would eventually fail with a timeout without
sending any data when no netif was specified and a link-local target
address was used. This fixes the behavior: Since any interface is
equally valid for a link local address, we just pick the first netif
when no netif is specified.

Especially in the case of a single netif (which is pretty common), the
new behavior will match expectations a lot better.

Co-authored-by: benpicco <[email protected]>

sys/net/gnrc/transport_layer/tcp/gnrc_tcp.c

@@ -28,6 +28,7 @@
 #include "net/af.h"
 #include "net/tcp.h"
 #include "net/gnrc.h"
+#include "net/gnrc/netif.h"
 #include "net/gnrc/tcp.h"
 #include "include/gnrc_tcp_common.h"
 #include "include/gnrc_tcp_fsm.h"
@@ -394,6 +395,20 @@ int gnrc_tcp_open(gnrc_tcp_tcb_t *tcb, const gnrc_tcp_ep_t *remote, uint16_t loc
             return -EINVAL;
         }
         tcb->ll_iface = remote->netif;
+
+        /* When the target is a link-local address and no netif is specified,
+         * any interface is equally valid. We just take whatever the first
+         * one is. Often there is only a single interface and than the
+         * behavior is as expected. If there are multiple interfaces, the
+         * caller really just has to specify the netif anyway. */
+        if (ipv6_addr_is_link_local((const ipv6_addr_t *)tcb->peer_addr)
+                && tcb->ll_iface == SOCK_ADDR_ANY_NETIF) {
+            const gnrc_netif_t *netif = gnrc_netif_iter(NULL);
+            if (!netif) {
+                return -EINVAL;
+            }
+            tcb->ll_iface = netif->pid;
+        }
     }
 #endif