Btl tcp: BTL_ERROR to show_help & update func behaviour

mohanasudhan · mohanasudhan · commit 0741fad479a7 · 2017-08-17T16:45:14.000-07:00
As part of improvement towards tcp debugging
we are moving few BTL_ERROR to show_help and also
update the function behaviour of
mca_btl_tcp_endpoint_complete_connect to return
SUCCESS and ERROR cases.

Signed-off-by: Mohan Gandhi &lt;mohgan@amazon.com&gt;
diff --git a/opal/mca/btl/tcp/btl_tcp_component.c b/opal/mca/btl/tcp/btl_tcp_component.c
@@ -729,7 +729,9 @@ static int mca_btl_tcp_component_create_instances(void)
         char* if_name = *argv;
         int if_index = opal_ifnametokindex(if_name);
         if(if_index < 0) {
-            BTL_ERROR(("invalid interface \"%s\"", if_name));
+            opal_show_help("help-mpi-btl-tcp.txt", "invalid if_inexclude",
+                           true, "include", opal_process_info.nodename,
+                           if_name, "Unknown interface name");
             ret = OPAL_ERR_NOT_FOUND;
             goto cleanup;
         }
@@ -960,15 +962,20 @@ static int mca_btl_tcp_component_create_listen(uint16_t af_family)
 
     /* set socket up to be non-blocking, otherwise accept could block */
     if((flags = fcntl(sd, F_GETFL, 0)) < 0) {
-        BTL_ERROR(("fcntl(F_GETFL) failed: %s (%d)",
-                   strerror(opal_socket_errno), opal_socket_errno));
+        opal_show_help("help-mpi-btl-tcp.txt", "socket flag fail",
+                       true, opal_process_info.nodename,
+                       getpid(), "fcntl(sd, F_GETFL, 0)",
+                       strerror(opal_socket_errno), opal_socket_errno);
         CLOSE_THE_SOCKET(sd);
         return OPAL_ERROR;
     } else {
         flags |= O_NONBLOCK;
         if(fcntl(sd, F_SETFL, flags) < 0) {
-            BTL_ERROR(("fcntl(F_SETFL) failed: %s (%d)",
-                       strerror(opal_socket_errno), opal_socket_errno));
+            opal_show_help("help-mpi-btl-tcp.txt", "socket flag fail",
+                           true, opal_process_info.nodename,
+                           getpid(),
+                           "fcntl(sd, F_SETFL, flags & O_NONBLOCK)",
+                           strerror(opal_socket_errno), opal_socket_errno);
             CLOSE_THE_SOCKET(sd);
             return OPAL_ERROR;
         }
diff --git a/opal/mca/btl/tcp/btl_tcp_endpoint.c b/opal/mca/btl/tcp/btl_tcp_endpoint.c
@@ -721,13 +721,23 @@ static int mca_btl_tcp_endpoint_start_connect(mca_btl_base_endpoint_t* btl_endpo
 
     /* setup the socket as non-blocking */
     if((flags = fcntl(btl_endpoint->endpoint_sd, F_GETFL, 0)) < 0) {
-        BTL_ERROR(("fcntl(F_GETFL) failed: %s (%d)",
-                   strerror(opal_socket_errno), opal_socket_errno));
+        opal_show_help("help-mpi-btl-tcp.txt", "socket flag fail",
+                       true, opal_process_info.nodename,
+                       getpid(), "fcntl(sd, F_GETFL, 0)",
+                       strerror(opal_socket_errno), opal_socket_errno);
+        /* Upper layer will handler the error */
+        return OPAL_ERR_UNREACH;
     } else {
         flags |= O_NONBLOCK;
-        if(fcntl(btl_endpoint->endpoint_sd, F_SETFL, flags) < 0)
-            BTL_ERROR(("fcntl(F_SETFL) failed: %s (%d)",
-                       strerror(opal_socket_errno), opal_socket_errno));
+        if(fcntl(btl_endpoint->endpoint_sd, F_SETFL, flags) < 0) {
+            opal_show_help("help-mpi-btl-tcp.txt", "socket flag fail",
+                           true, opal_process_info.nodename,
+                           getpid(),
+                           "fcntl(sd, F_SETFL, flags & O_NONBLOCK)",
+                           strerror(opal_socket_errno), opal_socket_errno);
+            /* Upper layer will handler the error */
+            return OPAL_ERR_UNREACH;
+        }
     }
 
     /* start the connect - will likely fail with EINPROGRESS */
@@ -778,7 +788,7 @@ static int mca_btl_tcp_endpoint_start_connect(mca_btl_base_endpoint_t* btl_endpo
  * later. Otherwise, send this processes identifier to the endpoint on the
  * newly connected socket.
  */
-static void mca_btl_tcp_endpoint_complete_connect(mca_btl_base_endpoint_t* btl_endpoint)
+static int mca_btl_tcp_endpoint_complete_connect(mca_btl_base_endpoint_t* btl_endpoint)
 {
     int so_error = 0;
     opal_socklen_t so_length = sizeof(so_error);
@@ -794,32 +804,49 @@ static void mca_btl_tcp_endpoint_complete_connect(mca_btl_base_endpoint_t* btl_e
 
     /* check connect completion status */
     if(getsockopt(btl_endpoint->endpoint_sd, SOL_SOCKET, SO_ERROR, (char *)&so_error, &so_length) < 0) {
-        BTL_ERROR(("getsockopt() to %s failed: %s (%d)",
+        opal_show_help("help-mpi-btl-tcp.txt", "socket flag fail",
+                       true, opal_process_info.nodename,
+                       getpid(), "fcntl(sd, F_GETFL, 0)",
+                       strerror(opal_socket_errno), opal_socket_errno);
+        BTL_ERROR(("getsockopt() to %s:%d failed: %s (%d)",
                    opal_net_get_hostname((struct sockaddr*) &endpoint_addr),
+                   ((struct sockaddr_in*) &endpoint_addr)->sin_port,
                    strerror(opal_socket_errno), opal_socket_errno));
         mca_btl_tcp_endpoint_close(btl_endpoint);
-        return;
+        return OPAL_ERROR;
     }
     if(so_error == EINPROGRESS || so_error == EWOULDBLOCK) {
-        return;
+        return OPAL_SUCCESS;
     }
     if(so_error != 0) {
-        BTL_ERROR(("connect() to %s failed: %s (%d)",
-                   opal_net_get_hostname((struct sockaddr*) &endpoint_addr),
-                   strerror(so_error), so_error));
+        char *msg;
+        asprintf(&msg, "connect() to %s:%d failed",
+                 opal_net_get_hostname((struct sockaddr*) &endpoint_addr),
+                 ntohs(((struct sockaddr_in*) &endpoint_addr)->sin_port));
+        opal_show_help("help-mpi-btl-tcp.txt", "client connect fail",
+                       true, opal_process_info.nodename,
+                       getpid(), msg,
+                       strerror(opal_socket_errno), opal_socket_errno);
+        free(msg);
         mca_btl_tcp_endpoint_close(btl_endpoint);
-        return;
+        return OPAL_ERROR;
     }
 
+    opal_output_verbose(10, opal_btl_base_framework.framework_output,
+                        "btl:tcp: connect() to %s:%d completed (complete_connect), sending connect ACK",
+                        opal_net_get_hostname((struct sockaddr*) &endpoint_addr),
+                        ntohs(((struct sockaddr_in*) &endpoint_addr)->sin_port));
+
     if(mca_btl_tcp_endpoint_send_connect_ack(btl_endpoint) == OPAL_SUCCESS) {
         btl_endpoint->endpoint_state = MCA_BTL_TCP_CONNECT_ACK;
         opal_event_add(&btl_endpoint->endpoint_recv_event, 0);
         MCA_BTL_TCP_ENDPOINT_DUMP(10, btl_endpoint, false, "event_add(recv) [complete_connect]");
-        return;
+        return OPAL_SUCCESS;
     }
     MCA_BTL_TCP_ENDPOINT_DUMP(1, btl_endpoint, false, " [complete_connect]");
     btl_endpoint->endpoint_state = MCA_BTL_TCP_FAILED;
     mca_btl_tcp_endpoint_close(btl_endpoint);
+    return OPAL_ERROR;
 }
 
 
diff --git a/opal/mca/btl/tcp/help-mpi-btl-tcp.txt b/opal/mca/btl/tcp/help-mpi-btl-tcp.txt
@@ -100,3 +100,68 @@ hopefully be able to continue).
   Peer hostname:       %s (%s)
   Source IP of socket: %s
   Known IPs of peer:   %s
+#
+[socket flag fail]
+WARNING: Open MPI failed to set flags on a TCP socket.  This should
+not happen.  It is likely that your MPI job will now fail.
+
+  Local host: %s
+  PID:        %d
+  Flag:       %s
+  Error:      %s (%d)
+#
+[server did not get guid]
+WARNING: Open MPI accepted a TCP connection from what appears to be a
+another Open MPI process but the peer process did not complete the
+initial handshake properly.  This should not happen.
+
+This attempted connection will be ignored; your MPI job may or may not
+continue properly.
+
+  Local host: %s
+  PID:        %d
+#
+[server accept cannot find guid]
+WARNING: Open MPI accepted a TCP connection from what appears to be a
+another Open MPI process but cannot find a corresponding process
+entry for that peer.
+
+This attempted connection will be ignored; your MPI job may or may not
+continue properly.
+
+  Local host: %s
+  PID:        %d
+#
+[server getpeername failed]
+WARNING: Open MPI failed to look up the peer IP address information of
+a TCP connection that it just accepted.  This should not happen.
+
+This attempted connection will be ignored; your MPI job may or may not
+continue properly.
+
+  Local host: %s
+  PID:        %d
+  Error:      %s (%d)
+#
+[server cannot find endpoint]
+WARNING: Open MPI accepted a TCP connection from what appears to be a
+valid peer Open MPI process but cannot find a corresponding endpoint
+entry for that peer.  This should not happen.
+
+This attempted connection will be ignored; your MPI job may or may not
+continue properly.
+
+  Local host: %s
+  PID:        %d
+#
+[client connect fail]
+WARNING: Open MPI failed to TCP connect to a peer MPI process via
+TCP.  This should not happen.
+
+Your Open MPI job may now fail.
+
+  Local host: %s
+  PID:        %d
+  Message:    %s
+  Error:      %s (%d)
+#
