Merge pull request #4346 from matcabral/psm2_mtl_mq_thread_fix

MTL PSM2: add a thread lock while peeking and completing the psm2 requests.

ompi/mca/mtl/psm2/mtl_psm2.c

@@ -406,20 +406,25 @@ int ompi_mtl_psm2_progress( void ) {
     int completed = 1;
 
     do {
+        OPAL_THREAD_LOCK(&mtl_psm2_mq_mutex);
         err = psm2_mq_ipeek2(ompi_mtl_psm2.mq, &req, NULL);
         if (err == PSM2_MQ_INCOMPLETE) {
+            OPAL_THREAD_UNLOCK(&mtl_psm2_mq_mutex);
             return completed;
-	} else if (err != PSM2_OK) {
+        } else if (OPAL_UNLIKELY(err != PSM2_OK)) {
+            OPAL_THREAD_UNLOCK(&mtl_psm2_mq_mutex);
+            goto error;
+        }
+
+        err = psm2_mq_test2(&req, &psm2_status);
+        OPAL_THREAD_UNLOCK(&mtl_psm2_mq_mutex);
+
+        if (OPAL_UNLIKELY (err != PSM2_OK)) {
             goto error;
         }
 
         completed++;
 
-	err = psm2_mq_test2(&req, &psm2_status);
-	if (err != PSM2_OK) {
-	    goto error;
-	}
-
         mtl_psm2_request = (mca_mtl_psm2_request_t*) psm2_status.context;
 
         if (mtl_psm2_request->type == OMPI_mtl_psm2_IRECV) {
@@ -457,7 +462,6 @@ int ompi_mtl_psm2_progress( void ) {
         }
 
         mtl_psm2_request->super.completion_callback(&mtl_psm2_request->super);
-
     }
     while (1);
 

ompi/mca/mtl/psm2/mtl_psm2.h

@@ -34,6 +34,8 @@
 
 BEGIN_C_DECLS
 
+/* MPI_THREAD_MULTIPLE_SUPPORT */
+extern opal_mutex_t mtl_psm2_mq_mutex;
 
 /* MTL interface functions */
 extern int ompi_mtl_psm2_add_procs(struct mca_mtl_base_module_t* mtl,

ompi/mca/mtl/psm2/mtl_psm2_component.c

@@ -42,6 +42,8 @@
 #include <glob.h>
 
 static int param_priority;
+/* MPI_THREAD_MULTIPLE_SUPPORT */
+opal_mutex_t mtl_psm2_mq_mutex = OPAL_MUTEX_STATIC_INIT;
 
 static int ompi_mtl_psm2_component_open(void);
 static int ompi_mtl_psm2_component_close(void);