#include "smx_context_java.h"
#include "jxbt_utilities.h"
#include "xbt/dynar.h"
+#include "../../simix/smx_private.h"
extern JavaVM *__java_vm;
XBT_LOG_NEW_DEFAULT_SUBCATEGORY(jmsg, bindings, "MSG for Java(TM)");
if (context->iwannadie) {
context->iwannadie = 0;
JNIEnv *env = get_current_thread_env();
- XBT_DEBUG("Gonnal launch Killed Error");
+ XBT_DEBUG("Gonna launch Killed Error");
// TODO Adrien, if the process has not been created at the java layer, why should we raise the exception/error at the java level (this happens
// for instance during the migration process that creates at the C level two processes: one on the SRC node and one on the DST node, if the DST process is killed.
// it is not required to raise an exception at the JAVA level, the low level should be able to manage such an issue correctly but this is not the case right now unfortunately ...
// TODO it will be nice to have the name of the process to help the end-user to know which Process has been killed
- jxbt_throw_by_name(env, "org/simgrid/msg/ProcessKilledError", bprintf("Process %s killed :) (file smx_context_java.c)", MSG_process_get_name( (msg_process_t)context) ));
+ // jxbt_throw_by_name(env, "org/simgrid/msg/ProcessKilledError", bprintf("Process %s killed :) (file smx_context_java.c)", MSG_process_get_name( (msg_process_t)context) ));
+ jxbt_throw_by_name(env, "org/simgrid/msg/ProcessKilledError", bprintf("Process %s killed :) (file smx_context_java.c)", simcall_process_get_name((smx_process_t) SIMIX_context_get_process(context))) );
XBT_DEBUG("Trigger a cancel error at the C level");
THROWF(cancel_error, 0, "process cancelled");
} else {
//hostfailure (if you want to know whether this is the SRC or the DST please check directly in send_migration_data code)
// Stop the dirty page tracking an return (there is no memory space to release)
stop_dirty_page_tracking(vm);
+ XBT_INFO ("Process tx migration catches an expection and thus return");
return 0;
}
remaining_size -= ramsize;
}CATCH_ANONYMOUS{
//hostfailure (if you want to know whether this is the SRC or the DST please check directly in send_migration_data code)
// Stop the dirty page tracking an return (there is no memory space to release)
+ XBT_INFO ("Process tx migration catches an expection and thus return");
stop_dirty_page_tracking(vm);
return 0;
}
}CATCH_ANONYMOUS{
//hostfailure (if you want to know whether this is the SRC or the DST please check directly in send_migration_data code)
// Stop the dirty page tracking an return (there is no memory space to release)
+ XBT_INFO ("Process tx migration catches an expection and thus return");
simcall_vm_resume(vm);
return 0;
}
// At that point the Migration is considered valid for the SRC node but remind that the DST side should relocate effectively the VM on the DST node.
- XBT_DEBUG("mig: tx_done");
+ XBT_INFO("mig: tx_done");
return 0;
}
char *pr_rx_name = get_mig_process_rx_name(vm, src_pm, dst_pm);
char *pr_tx_name = get_mig_process_tx_name(vm, src_pm, dst_pm);
+ msg_process_t tx_process, rx_process;
// MSG_process_create(pr_rx_name, migration_rx_fun, ms, dst_pm);
// MSG_process_create(pr_tx_name, migration_tx_fun, ms, src_pm);
#if 1
char **argv = xbt_new(char *, 2);
argv[0] = pr_rx_name;
argv[1] = NULL;
- MSG_process_create_with_arguments(pr_rx_name, migration_rx_fun, ms, dst_pm, 1, argv);
+ rx_process = MSG_process_create_with_arguments(pr_rx_name, migration_rx_fun, ms, dst_pm, 1, argv);
}
{
char **argv = xbt_new(char *, 2);
argv[0] = pr_tx_name;
argv[1] = NULL;
- MSG_process_create_with_arguments(pr_tx_name, migration_tx_fun, ms, src_pm, 1, argv);
+ tx_process = MSG_process_create_with_arguments(pr_tx_name, migration_tx_fun, ms, src_pm, 1, argv);
}
#endif
XBT_DEBUG("wait for reception of the final ACK (i.e. migration has been correctly performed");
msg_task_t task = NULL;
msg_error_t ret = MSG_TIMEOUT;
- while (ret == MSG_TIMEOUT && MSG_host_is_on(dst_pm)) //Wait while you receive the message o
- ret = MSG_task_receive_with_timeout(&task, ms->mbox_ctl, 10);
+ while (ret == MSG_TIMEOUT && MSG_host_is_on(dst_pm)) //Wait while you receive the message ok
+ // Active waiting, evaluation is performed every one second.
+ ret = MSG_task_receive_with_timeout(&task, ms->mbox_ctl, 1);
xbt_free(ms->mbox_ctl);
xbt_free(ms->mbox);
if(ret == MSG_HOST_FAILURE){
// Note that since the communication failed, the owner did not change and the task should be destroyed on the other side.
// Hence, just throw the execption
- //XBT_INFO("SRC crashes, throw an exception (m-control)");
+ XBT_INFO("SRC crashes, throw an exception (m-control)");
+ MSG_process_kill(tx_process);
return -1;
}
else if((ret == MSG_TRANSFER_FAILURE) || (ret == MSG_TIMEOUT)){ // MSG_TIMEOUT here means that MSG_host_is_avail() returned false.
- //XBT_INFO("DST crashes, throw an exception (m-control)");
+ XBT_INFO("DST crashes, throw an exception (m-control)");
return -2;
}