prov/efa: Fix bugs when reposting queued read and long cts pkts

When efa_rdm_ope_post_read has error, we should still continue after writing the txe/rxe error. When ctsdata pkt post returns FI_EAGAIN, we should still continue instead of break because opes may come from different eps. Signed-off-by: Shi Jin <[email protected]>
ofiwg · Jun 27, 2024 · 71e0ef3 · 71e0ef3
1 parent e72f856
commit 71e0ef3
Showing 1 changed file with 2 additions and 2 deletions.
diff --git a/prov/efa/src/efa_domain.c b/prov/efa/src/efa_domain.c
@@ -566,7 +566,7 @@ void efa_domain_progress_rdm_peers_and_queues(struct efa_domain *domain)
 					efa_rdm_txe_handle_error(ope, -ret, FI_EFA_ERR_READ_POST);
 				else
 					efa_rdm_rxe_handle_error(ope, -ret, FI_EFA_ERR_READ_POST);
-				return;
+				continue;
 			}
 
 			ope->internal_flags &= ~EFA_RDM_OPE_QUEUED_READ;
@@ -611,7 +611,7 @@ void efa_domain_progress_rdm_peers_and_queues(struct efa_domain *domain)
 			ret = efa_rdm_ope_post_send(ope, EFA_RDM_CTSDATA_PKT);
 			if (OFI_UNLIKELY(ret)) {
 				if (ret == -FI_EAGAIN)
-					break;
+					continue;
 
 				efa_rdm_txe_handle_error(ope, -ret, FI_EFA_ERR_PKT_POST);
 				continue;