prov/efa: fix efa multi recv setopt segfault

efa uses the util srx and sets the minimum multi receive size through the srx. However, the srx code doesn't get initialized until the endpoint gets enabled. So if the application calls setopt (before FI_ENABLE), this will segfault because the srx has not been initialized. Instead, we need to save the multi recv size in the efa endpoint to be valid during setopt and then pass that into the util_srx creation to set the multi recv size Signed-off-by: Alexia Ingerson <[email protected]> Signed-off-by: Shi Jin <[email protected]>
HewlettPackard · Dec 10, 2024 · 38adf06 · 38adf06
1 parent 95a7e84
commit 38adf06
Show file tree

Hide file tree

Showing 2 changed files with 5 additions and 10 deletions.
diff --git a/prov/efa/src/rdm/efa_rdm_ep_fiops.c b/prov/efa/src/rdm/efa_rdm_ep_fiops.c
@@ -1650,7 +1650,6 @@ static int efa_rdm_ep_setopt(fid_t fid, int level, int optname,
 {
 	struct efa_rdm_ep *efa_rdm_ep;
 	int intval, ret;
-	struct util_srx_ctx *srx;
 
 	efa_rdm_ep = container_of(fid, struct efa_rdm_ep, base_ep.util_ep.ep_fid.fid);
 
@@ -1663,8 +1662,6 @@ static int efa_rdm_ep_setopt(fid_t fid, int level, int optname,
 			return -FI_EINVAL;
 
 		efa_rdm_ep->min_multi_recv_size = *(size_t *)optval;
-		srx = util_get_peer_srx(efa_rdm_ep->peer_srx_ep)->ep_fid.fid.context;
-		srx->min_multi_recv_size = *(size_t *)optval;
 		break;
 	case FI_OPT_EFA_RNR_RETRY:
 		if (optlen != sizeof(size_t))

diff --git a/prov/efa/test/efa_unit_test_srx.c b/prov/efa/test/efa_unit_test_srx.c
@@ -18,21 +18,19 @@ void test_efa_srx_min_multi_recv_size(struct efa_resource **state)
         struct util_srx_ctx *srx_ctx;
         size_t min_multi_recv_size_new;
 
-        efa_unit_test_resource_construct(resource, FI_EP_RDM);
+        efa_unit_test_resource_construct_ep_not_enabled(resource, FI_EP_RDM);
 
         efa_rdm_ep = container_of(resource->ep, struct efa_rdm_ep, base_ep.util_ep.ep_fid);
-        srx_ctx = efa_rdm_ep_get_peer_srx_ctx(efa_rdm_ep);
-        /*
-         * After ep is enabled, the srx->min_multi_recv_size should be
-         * exactly the same with ep->min_multi_recv_size
-         */
-        assert_true(efa_rdm_ep->min_multi_recv_size == srx_ctx->min_multi_recv_size);
         /* Set a new min_multi_recv_size via setopt*/
         min_multi_recv_size_new = 1024;
         assert_int_equal(fi_setopt(&resource->ep->fid, FI_OPT_ENDPOINT, FI_OPT_MIN_MULTI_RECV,
 			&min_multi_recv_size_new, sizeof(min_multi_recv_size_new)), 0);
 
+        /* Enable EP */
+        assert_int_equal(fi_enable(resource->ep), FI_SUCCESS);
+
         /* Check whether srx->min_multi_recv_size is set correctly */
+        srx_ctx = efa_rdm_ep_get_peer_srx_ctx(efa_rdm_ep);
         assert_true(srx_ctx->min_multi_recv_size == min_multi_recv_size_new);
 }