Merge pull request #9844 from hjelmn/btl_sm_should_not_crash_when_using_rdma_on_self_ever

btl/sm: fix crash on put/get to self

Author: hjelmn

opal/mca/btl/sm/Makefile.am

@@ -29,7 +29,6 @@ libmca_btl_sm_la_sources = \
     btl_sm_module.c \
     btl_sm.h \
     btl_sm_component.c \
-    btl_sm_endpoint.h \
     btl_sm_fifo.h \
     btl_sm_frag.c \
     btl_sm_frag.h \

opal/mca/btl/sm/btl_sm.h

@@ -157,6 +157,10 @@ mca_btl_base_descriptor_t *mca_btl_sm_alloc(struct mca_btl_base_module_t *btl,
  */
 int mca_btl_sm_free(struct mca_btl_base_module_t *btl, mca_btl_base_descriptor_t *des);
 
+static inline bool mca_btl_is_self_endpoint(mca_btl_base_endpoint_t *endpoint) {
+    return endpoint->peer_smp_rank == MCA_BTL_SM_LOCAL_RANK;
+}
+
 END_C_DECLS
 
 #endif

opal/mca/btl/sm/btl_sm_endpoint.h

@@ -4,7 +4,7 @@
  *                         reserved.
  * Copyright (c) 2018      Research Organization for Information Science
  *                         and Technology (RIST).  All rights reserved.
- * Copyright (c) 2019-2021 Google, Inc. All rights reserved.
+ * Copyright (c) 2019-2022 Google, Inc. All rights reserved.
  * Copyright (c) 2020      The University of Tennessee and The University
  *                         of Tennessee Research Foundation.  All rights
  *                         reserved.
@@ -18,7 +18,6 @@
 #include "opal_config.h"
 
 #include "opal/mca/btl/sm/btl_sm.h"
-#include "opal/mca/btl/sm/btl_sm_endpoint.h"
 #include "opal/mca/btl/sm/btl_sm_frag.h"
 #include "opal/mca/smsc/smsc.h"
 
@@ -37,10 +36,14 @@ int mca_btl_sm_get(mca_btl_base_module_t *btl, mca_btl_base_endpoint_t *endpoint
                    int order, mca_btl_base_rdma_completion_fn_t cbfunc, void *cbcontext,
                    void *cbdata)
 {
-    int ret = MCA_SMSC_CALL(copy_from, endpoint->smsc_endpoint, local_address,
-                            (void *) (intptr_t) remote_address, size, remote_handle);
-    if (OPAL_UNLIKELY(OPAL_SUCCESS != ret)) {
-        return ret;
+    if (!mca_btl_is_self_endpoint(endpoint)) {
+        int ret = MCA_SMSC_CALL(copy_from, endpoint->smsc_endpoint, local_address,
+                                (void *) (intptr_t) remote_address, size, remote_handle);
+        if (OPAL_UNLIKELY(OPAL_SUCCESS != ret)) {
+            return ret;
+        }
+    } else {
+        memcpy(local_address, (void *)(uintptr_t) remote_address, size);
     }
 
     /* always call the callback function */

opal/mca/btl/sm/btl_sm_get.c

@@ -19,7 +19,7 @@
  *                         and Technology (RIST).  All rights reserved.
  * Copyright (c) 2018-2019 Triad National Security, LLC. All rights
  *                         reserved.
- * Copyright (c) 2020-2021 Google, LLC. All rights reserved.
+ * Copyright (c) 2020-2022 Google, LLC. All rights reserved.
  * Copyright (c) 2021      Nanook Consulting.  All rights reserved.
  * $COPYRIGHT$
  *
@@ -32,7 +32,6 @@
 #include "opal/util/show_help.h"
 
 #include "opal/mca/btl/sm/btl_sm.h"
-#include "opal/mca/btl/sm/btl_sm_endpoint.h"
 #include "opal/mca/btl/sm/btl_sm_fbox.h"
 #include "opal/mca/btl/sm/btl_sm_fifo.h"
 #include "opal/mca/btl/sm/btl_sm_frag.h"
@@ -168,7 +167,7 @@ static int init_sm_endpoint(struct mca_btl_base_endpoint_t **ep_out, struct opal
 
     ep->peer_smp_rank = peer_local_rank;
 
-    if (peer_local_rank != MCA_BTL_SM_LOCAL_RANK) {
+    if (!mca_btl_is_self_endpoint(ep)) {
         OPAL_MODEX_RECV_IMMEDIATE(rc, &component->super.btl_version, &proc->proc_name,
                                   (void **) &modex, &msg_size);
         if (OPAL_SUCCESS != rc) {

opal/mca/btl/sm/btl_sm_module.c

@@ -4,7 +4,7 @@
  *                         reserved.
  * Copyright (c) 2014-2018 Research Organization for Information Science
  *                         and Technology (RIST).  All rights reserved.
- * Copyright (c) 2019-2021 Google, Inc. All rights reserved.
+ * Copyright (c) 2019-2022 Google, Inc. All rights reserved.
  * Copyright (c) 2020      The University of Tennessee and The University
  *                         of Tennessee Research Foundation.  All rights
  *                         reserved.
@@ -18,7 +18,6 @@
 #include "opal_config.h"
 
 #include "opal/mca/btl/sm/btl_sm.h"
-#include "opal/mca/btl/sm/btl_sm_endpoint.h"
 #include "opal/mca/btl/sm/btl_sm_frag.h"
 #include "opal/mca/smsc/smsc.h"
 
@@ -37,10 +36,14 @@ int mca_btl_sm_put(mca_btl_base_module_t *btl, mca_btl_base_endpoint_t *endpoint
                    int order, mca_btl_base_rdma_completion_fn_t cbfunc, void *cbcontext,
                    void *cbdata)
 {
-    int ret = MCA_SMSC_CALL(copy_to, endpoint->smsc_endpoint, local_address,
-                            (void *) (intptr_t) remote_address, size, remote_handle);
-    if (OPAL_UNLIKELY(OPAL_SUCCESS != ret)) {
-        return ret;
+    if (!mca_btl_is_self_endpoint(endpoint)) {
+        int ret = MCA_SMSC_CALL(copy_to, endpoint->smsc_endpoint, local_address,
+                                (void *) (intptr_t) remote_address, size, remote_handle);
+        if (OPAL_UNLIKELY(OPAL_SUCCESS != ret)) {
+            return ret;
+        }
+    } else {
+        memcpy((void *)(uintptr_t) remote_address, local_address, size);
     }
 
     /* always call the callback function */