osc/ucx: Fix possible win creation/destruction race condition

To avoid fully initializing the osc/ucx component for MPI application
that are not using One-Sided functionality, the initialization happens
at the first MPI window creation.

This commit ensures atomicity of global state modifications.

Signed-off-by: Artem Polyakov <[email protected]>

Author: artpol84

ompi/mca/osc/ucx/osc_ucx_component.c

@@ -24,6 +24,21 @@
     memcpy(((char*)(_dst)) + (_off), _src, _len); \
     (_off) += (_len);
 
+opal_mutex_t mca_osc_service_mutex = OPAL_MUTEX_STATIC_INIT;
+static void _osc_ucx_init_lock(void)
+{
+    if(mca_osc_ucx_component.enable_mpi_threads) {
+        opal_mutex_lock(&mca_osc_service_mutex);
+    }
+}
+static void _osc_ucx_init_unlock(void)
+{
+    if(mca_osc_ucx_component.enable_mpi_threads) {
+        opal_mutex_unlock(&mca_osc_service_mutex);
+    }
+}
+
+
 static int component_open(void);
 static int component_register(void);
 static int component_init(bool enable_progress_threads, bool enable_mpi_threads);
@@ -192,6 +207,9 @@ static void ompi_osc_ucx_unregister_progress()
 {
     int ret;
 
+    /* May be called concurrently - protect */
+    _osc_ucx_init_lock();
+
     mca_osc_ucx_component.num_modules--;
     OSC_UCX_ASSERT(mca_osc_ucx_component.num_modules >= 0);
     if (0 == mca_osc_ucx_component.num_modules) {
@@ -200,6 +218,8 @@ static void ompi_osc_ucx_unregister_progress()
             OSC_UCX_VERBOSE(1, "opal_progress_unregister failed: %d", ret);
         }
     }
+
+    _osc_ucx_init_unlock();
 }
 
 static int component_select(struct ompi_win_t *win, void **base, size_t size, int disp_unit,
@@ -226,7 +246,14 @@ static int component_select(struct ompi_win_t *win, void **base, size_t size, in
         return OMPI_ERR_NOT_SUPPORTED;
     }
 
+    /* May be called concurrently - protect */
+    _osc_ucx_init_lock();
+
     if (mca_osc_ucx_component.env_initialized == false) {
+        /* Lazy initialization of the global state.
+         * As not all of the MPI applications are using One-Sided functionality
+         * we don't want to initialize in the component_init()
+         */
 
         OBJ_CONSTRUCT(&mca_osc_ucx_component.requests, opal_free_list_t);
         ret = opal_free_list_init (&mca_osc_ucx_component.requests,
@@ -236,30 +263,52 @@ static int component_select(struct ompi_win_t *win, void **base, size_t size, in
                                    0, 0, 8, 0, 8, NULL, 0, NULL, NULL, NULL);
         if (OMPI_SUCCESS != ret) {
             OSC_UCX_VERBOSE(1, "opal_free_list_init failed: %d", ret);
-            goto error;
+            goto select_unlock;
         }
 
         ret = opal_common_ucx_wpool_init(mca_osc_ucx_component.wpool,
                                          ompi_proc_world_size(),
                                          mca_osc_ucx_component.enable_mpi_threads);
         if (OMPI_SUCCESS != ret) {
             OSC_UCX_VERBOSE(1, "opal_common_ucx_wpool_init failed: %d", ret);
-            goto error;
+            goto select_unlock;
         }
 
+        /* Make sure that all memory updates performed above are globally
+         * observable before (mca_osc_ucx_component.env_initialized = true)
+         */
         mca_osc_ucx_component.env_initialized = true;
         env_initialized = true;
     }
 
+    /* Account for the number of active "modules" = MPI windows */
+    mca_osc_ucx_component.num_modules++;
+
+    /* If this is the first window to be registered - register the progress
+     * callback
+     */
+    OSC_UCX_ASSERT(mca_osc_ucx_component.num_modules > 0);
+    if (1 == mca_osc_ucx_component.num_modules) {
+        ret = opal_progress_register(progress_callback);
+        if (OMPI_SUCCESS != ret) {
+            OSC_UCX_VERBOSE(1, "opal_progress_register failed: %d", ret);
+            goto error;
+        }
+    }
+
+select_unlock:
+    _osc_ucx_init_unlock();
+    if (ret) {
+        goto error;
+    }
+
     /* create module structure */
     module = (ompi_osc_ucx_module_t *)calloc(1, sizeof(ompi_osc_ucx_module_t));
     if (module == NULL) {
         ret = OMPI_ERR_TEMP_OUT_OF_RESOURCE;
         goto error_nomem;
     }
 
-    mca_osc_ucx_component.num_modules++;
-
     /* fill in the function pointer part */
     memcpy(module, &ompi_osc_ucx_module_template, sizeof(ompi_osc_base_module_t));
 
@@ -413,19 +462,15 @@ static int component_select(struct ompi_win_t *win, void **base, size_t size, in
         goto error;
     }
 
-    OSC_UCX_ASSERT(mca_osc_ucx_component.num_modules > 0);
-    if (1 == mca_osc_ucx_component.num_modules) {
-        ret = opal_progress_register(progress_callback);
-        if (OMPI_SUCCESS != ret) {
-            OSC_UCX_VERBOSE(1, "opal_progress_register failed: %d", ret);
-            goto error;
-        }
-    }
     return ret;
 
- error:
+error:
     if (module->disp_units) free(module->disp_units);
     if (module->comm) ompi_comm_free(&module->comm);
+    /* We update the modules count and (if need) registering a callback right
+     * prior to memory allocation for the module.
+     * So we use it as an indirect sign here
+     */
     if (module) {
         free(module);
         ompi_osc_ucx_unregister_progress();