From: Gabriel Corona <gabriel.corona@loria.fr>
Date: Fri, 3 Jul 2015 11:36:53 +0000 (+0200)
Subject: Move the host-on-off fix above
X-Git-Tag: v3_12~539
X-Git-Url: http://info.iut-bm.univ-fcomte.fr/pub/gitweb/simgrid.git/commitdiff_plain/24ee99596e6ca507bf1e517d49c7e8110a8a410b

Move the host-on-off fix above

If our host is OFF, we should always get a HOST_FAILURE.

This seems to fix issues where a host was turned off but its processes
were not woken up with HOST_FAILURE.
---

diff --git a/examples/msg/masterslave/masterslave_cpu_ti_crosstraffic.tesh b/examples/msg/masterslave/masterslave_cpu_ti_crosstraffic.tesh
index c93afef039..94063dbfed 100644
--- a/examples/msg/masterslave/masterslave_cpu_ti_crosstraffic.tesh
+++ b/examples/msg/masterslave/masterslave_cpu_ti_crosstraffic.tesh
@@ -290,6 +290,7 @@ $ ${bindir:=.}/masterslave_failure$EXEEXT --log=xbt_cfg.thres:critical --log=no_
 > [  0.000000] (1:master@Tremblay) Got 20 task to process :
 > [  1.000000] (0:@) Restart processes on host: Fafard
 > [  1.000000] (1:master@Tremblay) Mmh. Something went wrong with 'Jupiter'. Nevermind. Let's keep going!
+> [  1.000000] (3:slave@Jupiter) Gloups. The cpu on which I'm running just turned off!. See you!
 > [  2.000000] (0:@) Restart processes on host: Jupiter
 > [ 11.000000] (1:master@Tremblay) Mmh. Got timeouted while speaking to 'Fafard'. Nevermind. Let's keep going!
 > [ 12.082474] (1:master@Tremblay) Send completed
diff --git a/examples/msg/masterslave/masterslave_failure.tesh b/examples/msg/masterslave/masterslave_failure.tesh
index 7b1b5f7e5b..f6596fff95 100644
--- a/examples/msg/masterslave/masterslave_failure.tesh
+++ b/examples/msg/masterslave/masterslave_failure.tesh
@@ -14,6 +14,7 @@ $ masterslave/masterslave_failure$EXEEXT --log=xbt_cfg.thres:critical --log=no_l
 > [  0.000000] (1:master@Tremblay) Got 20 task to process :
 > [  1.000000] (0:@) Restart processes on host: Fafard
 > [  1.000000] (1:master@Tremblay) Mmh. Something went wrong with 'Jupiter'. Nevermind. Let's keep going!
+> [  1.000000] (3:slave@Jupiter) Gloups. The cpu on which I'm running just turned off!. See you!
 > [  2.000000] (0:@) Restart processes on host: Jupiter
 > [ 11.000000] (1:master@Tremblay) Mmh. Got timeouted while speaking to 'Fafard'. Nevermind. Let's keep going!
 > [ 12.030928] (1:master@Tremblay) Send completed
diff --git a/examples/msg/masterslave/masterslave_failure_crosstraffic.tesh b/examples/msg/masterslave/masterslave_failure_crosstraffic.tesh
index 43feb15002..ad02729c3f 100644
--- a/examples/msg/masterslave/masterslave_failure_crosstraffic.tesh
+++ b/examples/msg/masterslave/masterslave_failure_crosstraffic.tesh
@@ -14,6 +14,7 @@ $ masterslave/masterslave_failure$EXEEXT --log=xbt_cfg.thres:critical --log=no_l
 > [  0.000000] (1:master@Tremblay) Got 20 task to process :
 > [  1.000000] (0:@) Restart processes on host: Fafard
 > [  1.000000] (1:master@Tremblay) Mmh. Something went wrong with 'Jupiter'. Nevermind. Let's keep going!
+> [  1.000000] (3:slave@Jupiter) Gloups. The cpu on which I'm running just turned off!. See you!
 > [  2.000000] (0:@) Restart processes on host: Jupiter
 > [ 11.000000] (1:master@Tremblay) Mmh. Got timeouted while speaking to 'Fafard'. Nevermind. Let's keep going!
 > [ 12.082474] (1:master@Tremblay) Send completed
diff --git a/src/simix/smx_network.c b/src/simix/smx_network.c
index 6d1b84928b..42e07bebd9 100644
--- a/src/simix/smx_network.c
+++ b/src/simix/smx_network.c
@@ -777,7 +777,6 @@ void SIMIX_comm_finish(smx_synchro_t synchro)
   unsigned int destroy_count = 0;
   smx_simcall_t simcall;
 
-
   while ((simcall = xbt_fifo_shift(synchro->simcalls))) {
 
     /* If a waitany simcall is waiting for this synchro to finish, then remove
@@ -801,6 +800,12 @@ void SIMIX_comm_finish(smx_synchro_t synchro)
 
     /* Check out for errors */
 
+    if (surf_resource_get_state(surf_workstation_resource_priv(
+          simcall->issuer->smx_host)) != SURF_RESOURCE_ON) {
+      simcall->issuer->context->iwannadie = 1;
+      SMX_EXCEPTION(simcall->issuer, host_error, 0, "Host failed");
+    } else
+
     switch (synchro->state) {
 
     case SIMIX_DONE:
@@ -836,14 +841,6 @@ void SIMIX_comm_finish(smx_synchro_t synchro)
 
     case SIMIX_LINK_FAILURE:
 
-      // There should be a cleaner way to do this.
-      // We should handle this in SIMIX_post_comm instead.
-      if (surf_resource_get_state(surf_workstation_resource_priv(
-            simcall->issuer->smx_host)) != SURF_RESOURCE_ON) {
-        SMX_EXCEPTION(simcall->issuer, host_error, 0, "Host failed");
-        break;
-      }
-
       XBT_DEBUG("Link failure in synchro %p between '%s' and '%s': posting an exception to the issuer: %s (%p) detached:%d",
                 synchro,
                 synchro->comm.src_proc ? sg_host_name(synchro->comm.src_proc->smx_host) : NULL,