X-Git-Url: http://bilbo.iut-bm.univ-fcomte.fr/pub/gitweb/simgrid.git/blobdiff_plain/fc727903530d55d5d3f903a8213c7cb8a3273c7a..b4852fd3bb330d081de77ad7092fb2f11ee0b7a4:/src/simix/smx_network.c

diff --git a/src/simix/smx_network.c b/src/simix/smx_network.c
index 82a372fced..c72e0af016 100644
--- a/src/simix/smx_network.c
+++ b/src/simix/smx_network.c
@@ -20,8 +20,8 @@ static void SIMIX_comm_copy_data(smx_action_t comm);
 static smx_action_t SIMIX_comm_new(e_smx_comm_type_t type);
 static XBT_INLINE void SIMIX_rdv_push(smx_rdv_t rdv, smx_action_t comm);
 static smx_action_t SIMIX_rdv_get_comm(smx_rdv_t rdv, e_smx_comm_type_t type,
-					  int (*match_fun)(void *, void *,smx_action_t),
-					  void *user_data, smx_action_t my_action);
+            int (*match_fun)(void *, void *,smx_action_t),
+            void *user_data, smx_action_t my_action);
 static void SIMIX_rdv_free(void *data);
 
 void SIMIX_network_init(void)
@@ -205,9 +205,9 @@ void SIMIX_comm_destroy(smx_action_t action)
             action, action->comm.refcount, (int)action->state);
 
   if (action->comm.refcount <= 0) {
-	xbt_backtrace_display_current();
-    xbt_die("the refcount of comm %p is already 0 before decreasing it. "
-            "That's a bug!", action);
+    xbt_backtrace_display_current();
+    xbt_die("The refcount of comm %p is already 0 before decreasing it. "
+            "That's a bug! If you didn't test and/or wait the same communication twice in your code, then the bug is SimGrid's...", action);
   }
   action->comm.refcount--;
   if (action->comm.refcount > 0)
@@ -518,7 +518,7 @@ XBT_INLINE void SIMIX_comm_start(smx_action_t action)
     /* If a link is failed, detect it immediately */
     if (surf_workstation_model->action_state_get(action->comm.surf_comm) == SURF_ACTION_FAILED) {
       XBT_DEBUG("Communication from '%s' to '%s' failed to start because of a link failure",
-	  SIMIX_host_get_name(sender), SIMIX_host_get_name(receiver));
+    SIMIX_host_get_name(sender), SIMIX_host_get_name(receiver));
       action->state = SIMIX_LINK_FAILURE;
       SIMIX_comm_destroy_internal_actions(action);
     }
@@ -528,7 +528,16 @@ XBT_INLINE void SIMIX_comm_start(smx_action_t action)
     if (SIMIX_process_is_suspended(action->comm.src_proc) ||
         SIMIX_process_is_suspended(action->comm.dst_proc)) {
       /* FIXME: check what should happen with the action state */
+
+      if (SIMIX_process_is_suspended(action->comm.src_proc))
+        XBT_DEBUG("The communication is suspended on startup because src (%s:%s) were suspended since it initiated the communication",
+            SIMIX_host_get_name(action->comm.src_proc->smx_host), action->comm.src_proc->name);
+      else
+        XBT_DEBUG("The communication is suspended on startup because dst (%s:%s) were suspended since it initiated the communication",
+            SIMIX_host_get_name(action->comm.dst_proc->smx_host), action->comm.dst_proc->name);
+
       surf_workstation_model->suspend(action->comm.surf_comm);
+
     }
   }
 }
@@ -580,14 +589,16 @@ void SIMIX_comm_finish(smx_action_t action)
 
       case SIMIX_SRC_HOST_FAILURE:
         if (simcall->issuer == action->comm.src_proc)
-          SMX_EXCEPTION(simcall->issuer, host_error, 0, "Host failed");
+          simcall->issuer->context->iwannadie = 1;
+//          SMX_EXCEPTION(simcall->issuer, host_error, 0, "Host failed");
         else
           SMX_EXCEPTION(simcall->issuer, network_error, 0, "Remote peer failed");
         break;
 
       case SIMIX_DST_HOST_FAILURE:
         if (simcall->issuer == action->comm.dst_proc)
-          SMX_EXCEPTION(simcall->issuer, host_error, 0, "Host failed");
+          simcall->issuer->context->iwannadie = 1;
+//          SMX_EXCEPTION(simcall->issuer, host_error, 0, "Host failed");
         else
           SMX_EXCEPTION(simcall->issuer, network_error, 0, "Remote peer failed");
         break;
@@ -631,6 +642,11 @@ void SIMIX_comm_finish(smx_action_t action)
       }
     }
 
+    if (surf_workstation_model->extension.
+        workstation.get_state(simcall->issuer->smx_host->host) != SURF_RESOURCE_ON) {
+      simcall->issuer->context->iwannadie = 1;
+    }
+
     simcall->issuer->waiting_action = NULL;
     xbt_fifo_remove(simcall->issuer->comms, action);
     SIMIX_simcall_answer(simcall);
@@ -662,7 +678,7 @@ void SIMIX_post_comm(smx_action_t action)
      action->state = SIMIX_DST_HOST_FAILURE;
   else if (action->comm.surf_comm &&
           surf_workstation_model->action_state_get(action->comm.surf_comm) == SURF_ACTION_FAILED) {
-	  XBT_DEBUG("Puta madre. Surf says that the link broke");
+    XBT_DEBUG("Puta madre. Surf says that the link broke");
      action->state = SIMIX_LINK_FAILURE;
   } else
     action->state = SIMIX_DONE;
@@ -706,13 +722,17 @@ void SIMIX_comm_cancel(smx_action_t action)
 void SIMIX_comm_suspend(smx_action_t action)
 {
   /*FIXME: shall we suspend also the timeout actions? */
-  surf_workstation_model->suspend(action->comm.surf_comm);
+  if (action->comm.surf_comm)
+    surf_workstation_model->suspend(action->comm.surf_comm);
+  /* in the other case, the action will be suspended on creation, in SIMIX_comm_start() */
 }
 
 void SIMIX_comm_resume(smx_action_t action)
 {
   /*FIXME: check what happen with the timeouts */
-  surf_workstation_model->resume(action->comm.surf_comm);
+  if (action->comm.surf_comm)
+    surf_workstation_model->resume(action->comm.surf_comm);
+  /* in the other case, the action were not really suspended yet, see SIMIX_comm_suspend() and SIMIX_comm_start() */
 }
 
 
@@ -845,7 +865,7 @@ void SIMIX_comm_copy_data(smx_action_t comm)
 {
   size_t buff_size = comm->comm.src_buff_size;
   /* If there is no data to be copy then return */
-  if (!comm->comm.src_buff || !comm->comm.dst_buff || comm->comm.copied == 1)
+  if (!comm->comm.src_buff || !comm->comm.dst_buff || comm->comm.copied)
     return;
 
   XBT_DEBUG("Copying comm %p data from %s (%p) -> %s (%p) (%zu bytes)",