Make the request lock recursive.

If during the request completion callback we post another request that
completes right away (such a small send or a match for an unexpected
short message) we will try to complete the second request while holding
the lock for the completion of the first. For performance reasons
(mainly to avoid unlocking and locking the request mutex several times)
we have made the request lock recursive.

(cherry picked from commit open-mpi/ompi@bf19067)

Signed-off-by: Nathan Hjelm <[email protected]>

Author: bosilca

ompi/mca/pml/ob1/pml_ob1_recvreq.h

@@ -123,8 +123,8 @@ do {                                                                \
  */
 #define MCA_PML_OB1_RECV_REQUEST_MPI_COMPLETE( recvreq )                              \
     do {                                                                              \
-       PERUSE_TRACE_COMM_EVENT( PERUSE_COMM_REQ_COMPLETE,                             \
-                                &(recvreq->req_recv.req_base), PERUSE_RECV );         \
+        PERUSE_TRACE_COMM_EVENT( PERUSE_COMM_REQ_COMPLETE,                            \
+                                 &(recvreq->req_recv.req_base), PERUSE_RECV );        \
         ompi_request_complete( &(recvreq->req_recv.req_base.req_ompi), true );        \
     } while (0)
 

ompi/mca/pml/ob1/pml_ob1_sendreq.h

@@ -209,10 +209,10 @@ do {
    (sendreq)->req_send.req_base.req_ompi.req_status.MPI_ERROR = OMPI_SUCCESS;        \
    (sendreq)->req_send.req_base.req_ompi.req_status._ucount =                        \
         (sendreq)->req_send.req_bytes_packed;                                        \
-   ompi_request_complete( &((sendreq)->req_send.req_base.req_ompi), (with_signal) ); \
-                                                                                     \
    PERUSE_TRACE_COMM_EVENT( PERUSE_COMM_REQ_COMPLETE,                                \
                             &(sendreq->req_send.req_base), PERUSE_SEND);             \
+                                                                                     \
+   ompi_request_complete( &((sendreq)->req_send.req_base.req_ompi), (with_signal) ); \
 } while(0)
 
 static inline void mca_pml_ob1_send_request_fini (mca_pml_ob1_send_request_t *sendreq)

ompi/request/request.c

@@ -36,7 +36,7 @@ opal_pointer_array_t             ompi_request_f_to_c_table = {{0}};
 size_t                           ompi_request_waiting = 0;
 size_t                           ompi_request_completed = 0;
 size_t                           ompi_request_failed = 0;
-opal_mutex_t                     ompi_request_lock = {{0}};
+opal_recursive_mutex_t           ompi_request_lock = {{0}};
 opal_condition_t                 ompi_request_cond = {{0}};
 ompi_predefined_request_t        ompi_request_null = {{{{{0}}}}};
 ompi_predefined_request_t        *ompi_request_null_addr = &ompi_request_null;
@@ -109,7 +109,7 @@ OBJ_CLASS_INSTANCE(
 
 int ompi_request_init(void)
 {
-    OBJ_CONSTRUCT(&ompi_request_lock, opal_mutex_t);
+    OBJ_CONSTRUCT(&ompi_request_lock, opal_recursive_mutex_t);
     OBJ_CONSTRUCT(&ompi_request_cond, opal_condition_t);
 
     OBJ_CONSTRUCT(&ompi_request_null, ompi_request_t);

ompi/request/request.h

@@ -306,18 +306,18 @@ typedef struct ompi_request_fns_t {
 /**
  * Globals used for tracking requests and request completion.
  */
-OMPI_DECLSPEC extern opal_pointer_array_t  ompi_request_f_to_c_table;
-OMPI_DECLSPEC extern size_t                ompi_request_waiting;
-OMPI_DECLSPEC extern size_t                ompi_request_completed;
-OMPI_DECLSPEC extern size_t                ompi_request_failed;
-OMPI_DECLSPEC extern int32_t               ompi_request_poll;
-OMPI_DECLSPEC extern opal_mutex_t          ompi_request_lock;
-OMPI_DECLSPEC extern opal_condition_t      ompi_request_cond;
+OMPI_DECLSPEC extern opal_pointer_array_t   ompi_request_f_to_c_table;
+OMPI_DECLSPEC extern size_t                 ompi_request_waiting;
+OMPI_DECLSPEC extern size_t                 ompi_request_completed;
+OMPI_DECLSPEC extern size_t                 ompi_request_failed;
+OMPI_DECLSPEC extern int32_t                ompi_request_poll;
+OMPI_DECLSPEC extern opal_recursive_mutex_t ompi_request_lock;
+OMPI_DECLSPEC extern opal_condition_t       ompi_request_cond;
 OMPI_DECLSPEC extern ompi_predefined_request_t        ompi_request_null;
 OMPI_DECLSPEC extern ompi_predefined_request_t        *ompi_request_null_addr;
-OMPI_DECLSPEC extern ompi_request_t        ompi_request_empty;
-OMPI_DECLSPEC extern ompi_status_public_t  ompi_status_empty;
-OMPI_DECLSPEC extern ompi_request_fns_t    ompi_request_functions;
+OMPI_DECLSPEC extern ompi_request_t         ompi_request_empty;
+OMPI_DECLSPEC extern ompi_status_public_t   ompi_status_empty;
+OMPI_DECLSPEC extern ompi_request_fns_t     ompi_request_functions;
 
 /**
  * Initialize the MPI_Request subsystem; invoked during MPI_INIT.