ch4/am: check is_local in matching posted_list

This is necessary when both shmmod and netmod are using active messages and an anysource recv pair are posted in the same posted_list. Without it, a netmod send will get matched to a shmmod recv, resulting in errors.
hzhou · Feb 25, 2020 · 21870d1 · 21870d1
1 parent baad166
commit 21870d1
Show file tree

Hide file tree

Showing 2 changed files with 12 additions and 7 deletions.
diff --git a/src/mpid/ch4/src/ch4r_callbacks.c b/src/mpid/ch4/src/ch4r_callbacks.c
@@ -381,7 +381,7 @@ int MPIDIG_send_target_msg_cb(int handler_id, void *am_hdr, void **data, size_t
         /* MPIDI_CS_ENTER(); */
         while (TRUE) {
             rreq = MPIDIG_dequeue_posted(hdr->src_rank, hdr->tag, hdr->context_id,
-                                         &MPIDIG_COMM(root_comm, posted_list));
+                                         is_local, &MPIDIG_COMM(root_comm, posted_list));
 #ifndef MPIDI_CH4_DIRECT_NETMOD
             if (rreq) {
                 int is_cancelled;
@@ -488,7 +488,7 @@ int MPIDIG_send_long_req_target_msg_cb(int handler_id, void *am_hdr, void **data
         /* MPIDI_CS_ENTER(); */
         while (TRUE) {
             rreq = MPIDIG_dequeue_posted(hdr->src_rank, hdr->tag, hdr->context_id,
-                                         &MPIDIG_COMM(root_comm, posted_list));
+                                         is_local, &MPIDIG_COMM(root_comm, posted_list));
 #ifndef MPIDI_CH4_DIRECT_NETMOD
             if (rreq) {
                 int is_cancelled;

diff --git a/src/mpid/ch4/src/ch4r_recvq.h b/src/mpid/ch4/src/ch4r_recvq.h
@@ -148,7 +148,7 @@ MPL_STATIC_INLINE_PREFIX MPIR_Request *MPIDIG_find_unexp(int rank, int tag,
 
 MPL_STATIC_INLINE_PREFIX MPIR_Request *MPIDIG_dequeue_posted(int rank, int tag,
                                                              MPIR_Context_id_t context_id,
-                                                             MPIDIG_rreq_t ** list)
+                                                             int is_local, MPIDIG_rreq_t ** list)
 {
     MPIR_Request *req = NULL;
     MPIDIG_rreq_t *curr, *tmp;
@@ -158,13 +158,18 @@ MPL_STATIC_INLINE_PREFIX MPIR_Request *MPIDIG_dequeue_posted(int rank, int tag,
     MPIR_T_PVAR_TIMER_START(RECVQ, time_failed_matching_postedq);
     DL_FOREACH_SAFE(*list, curr, tmp) {
         MPIR_T_PVAR_COUNTER_INC(RECVQ, posted_recvq_match_attempts, 1);
-        req = curr->request;
-        if (MPIDIG_match_posted(rank, tag, context_id, req)) {
+#ifndef MPIDI_CH4_DIRECT_NETMOD
+        /* NOTE: extra negation to force logical comparisons */
+        if (!MPIDI_REQUEST(curr->request, is_local) != !is_local) {
+            continue;
+        }
+#endif
+        if (MPIDIG_match_posted(rank, tag, context_id, curr->request)) {
+            req = curr->request;
             DL_DELETE(*list, curr);
             MPIR_T_PVAR_LEVEL_DEC(RECVQ, posted_recvq_length, 1);
             break;
         }
-        req = NULL;
     }
     if (!req)
         MPIR_T_PVAR_TIMER_END(RECVQ, time_failed_matching_postedq);
@@ -305,7 +310,7 @@ MPL_STATIC_INLINE_PREFIX MPIR_Request *MPIDIG_find_unexp(int rank, int tag,
 
 MPL_STATIC_INLINE_PREFIX MPIR_Request *MPIDIG_dequeue_posted(int rank, int tag,
                                                              MPIR_Context_id_t context_id,
-                                                             MPIDIG_rreq_t ** list)
+                                                             int is_local, MPIDIG_rreq_t ** list)
 {
     MPIR_Request *req = NULL;
     MPIDIG_rreq_t *curr, *tmp;