From: mquinson <mquinson@48e7efb5-ca39-0410-a469-dd3cf9ba447f>
Date: Tue, 4 May 2010 13:10:31 +0000 (+0000)
Subject: Fix a bug occuring when a host involved in a communication fails
X-Git-Tag: SVN~68
X-Git-Url: http://info.iut-bm.univ-fcomte.fr/pub/gitweb/simgrid.git/commitdiff_plain/deba0f7f0441a96f742af845cfceabf340d0f27f

Fix a bug occuring when a host involved in a communication fails

git-svn-id: svn+ssh://scm.gforge.inria.fr/svn/simgrid/simgrid/trunk@7676 48e7efb5-ca39-0410-a469-dd3cf9ba447f
---

diff --git a/ChangeLog b/ChangeLog
index d60ccb7ae4..b6ea39124d 100644
--- a/ChangeLog
+++ b/ChangeLog
@@ -1,12 +1,17 @@
 SimGrid (3.4.1-svn) unstable; urgency=low
 
+ Java Bindings
+ * Fix a bug preventing the tasks from begin garbage collected.
+
+ MSG
+ * Fix a bug occuring when a host involved in a communication fails.
+   This was not detected properly by the other peer involved in the
+   communication. Now, it's reported as a network error.
+
  SimDag
  * Warn the user about loop dependencies in data flow of DAX files
  * Obey the control-flow dependencies of DAX files
 
- Java Bindings
- * Fix a bug preventing the tasks from begin garbage collected.
-
  Cmake
  * Add option "enable_smpi" for fix mac problem with smpi
  * Add verification of lua5.1 and ruby1.8
diff --git a/src/simix/smx_network.c b/src/simix/smx_network.c
index 0b48fa6d3c..4987f8f20e 100644
--- a/src/simix/smx_network.c
+++ b/src/simix/smx_network.c
@@ -237,9 +237,9 @@ static inline void SIMIX_communication_start(smx_comm_t comm)
  *  \param timeout The max amount of time to wait for the communication to finish
  *
  *  Throws:
- *   - host_error if peer failed
- *   - timeout_error if communication reached the timeout specified
- *   - network_error if network failed or peer issued a timeout
+ *   - host_error if local peer failed
+ *   - timeout_error if communication reached the timeout specified (either because of local peer or remote peer)
+ *   - network_error if network failed or remote peer failed
  */
 static inline void SIMIX_communication_wait_for_completion(smx_comm_t comm, double timeout)
 {
@@ -290,7 +290,7 @@ static inline void SIMIX_communication_wait_for_completion(smx_comm_t comm, doub
   SIMIX_sem_release_forever(comm->sem);
   
   /* Check for errors other than timeouts (they are catched above) */
-  if(!SIMIX_host_get_state(SIMIX_host_self())){
+  if (!SIMIX_host_get_state(SIMIX_host_self())){
     if(comm->rdv)
       SIMIX_rdv_remove(comm->rdv, comm);
     SIMIX_communication_destroy(comm);
@@ -298,6 +298,16 @@ static inline void SIMIX_communication_wait_for_completion(smx_comm_t comm, doub
   } else if (SIMIX_action_get_state(comm->act) == SURF_ACTION_FAILED){
     SIMIX_communication_destroy(comm);
     THROW0(network_error, 0, "Link failure");
+  } else if (!SIMIX_host_get_state(SIMIX_process_get_host(comm->dst_proc)) ||
+      !SIMIX_host_get_state(SIMIX_process_get_host(comm->src_proc))) {
+    /* We test both src&dst because we dunno who we are today, and we already tested myself above.
+     *    So, at the end, we test the remote peer only
+     * Moreover, we have to test it because if the remote peer fails, the action comm->act is not done nor failed.
+     *    In that case, we got awaken by the little endless actions created in the SIMIX_sem_acquire(comm->sem)
+     *    at the beginning of this function. */
+    SIMIX_communication_destroy(comm);
+    THROW0(network_error, 0, "Remote peer failed");
+
   }
   SIMIX_communication_destroy(comm);
 }