DAOS-17135 cart: not exclude self when create corpc for IV

Otherwise, it may misguide subsequent crt_get_filtered_grp_rank_list()
and regard the IV root to be as non-exist in the ranks list, then fail
related IV operation with -DER_NONEXIST.

It maybe not a perfected solution for current cart IV logic. But to be
some temporary option, it makes CR to be workable when some ranks dead.

Add new test case to verfiy such corner case.

Signed-off-by: Fan Yong <fan.yong@hpe.com>

Author: Nasf-Fan

src/cart/crt_iv.c

@@ -2217,8 +2217,6 @@ crt_ivsync_rpc_issue(struct crt_ivns_internal *ivns_internal, uint32_t class_id,
 	struct iv_sync_cb_info	*iv_sync_cb = NULL;
 	struct crt_iv_ops	*iv_ops;
 	crt_bulk_t		local_bulk = CRT_BULK_NULL;
-	d_rank_list_t		excluded_list;
-	d_rank_t		excluded_ranks[1]; /* Excluding self */
 
 	iv_ops = crt_iv_ops_get(ivns_internal, class_id);
 	D_ASSERT(iv_ops != NULL);
@@ -2243,10 +2241,6 @@ crt_ivsync_rpc_issue(struct crt_ivns_internal *ivns_internal, uint32_t class_id,
 		D_GOTO(exit, rc = -DER_INVAL);
 	}
 
-	/* Exclude self from corpc */
-	excluded_list.rl_nr = 1;
-	excluded_list.rl_ranks = excluded_ranks;
-	excluded_ranks[0] = ivns_internal->cii_grp_priv->gp_self;
 	/* Perform refresh on local node */
 	if (sync_type->ivs_event == CRT_IV_SYNC_EVENT_UPDATE)
 		rc = iv_ops->ivo_on_refresh(ivns_internal, iv_key, 0,
@@ -2273,13 +2267,9 @@ crt_ivsync_rpc_issue(struct crt_ivns_internal *ivns_internal, uint32_t class_id,
 		}
 	}
 
-	rc = crt_corpc_req_create(ivns_internal->cii_ctx,
-				  &ivns_internal->cii_grp_priv->gp_pub,
-				  &excluded_list,
-				  CRT_OPC_IV_SYNC,
-				  local_bulk, NULL, 0,
-				  ivns_internal->cii_gns.gn_tree_topo,
-				  &corpc_req);
+	rc = crt_corpc_req_create(ivns_internal->cii_ctx, &ivns_internal->cii_grp_priv->gp_pub,
+				  NULL, CRT_OPC_IV_SYNC, local_bulk, NULL, 0,
+				  ivns_internal->cii_gns.gn_tree_topo, &corpc_req);
 	if (rc != 0) {
 		D_ERROR("crt_corpc_req_create(): "DF_RC"\n", DP_RC(rc));
 		D_GOTO(exit, rc);

src/tests/suite/daos_cr.c

@@ -3753,6 +3753,63 @@ cr_maintenance_mode(void **state)
 	cr_cleanup(arg, &pool, 1);
 }
 
+/*
+ * 1. Exclude rank 0.
+ * 2. Create pool without inconsistency.
+ * 3. Start checker without options.
+ * 4. Query checker, it should be completed instead of being blocked.
+ * 5. Switch to normal mode and cleanup.
+ */
+static void
+cr_lost_rank0(void **state)
+{
+	test_arg_t		*arg = *state;
+	struct test_pool	 pool = { 0 };
+	struct daos_check_info	 dci = { 0 };
+	int			 rc;
+
+	FAULT_INJECTION_REQUIRED();
+
+	print_message("CR29: run CR with rank 0 excluded at the beginning\n");
+
+	print_message("CR: excluding the rank 0 ...\n");
+	rc = dmg_system_exclude_rank(dmg_config_file, 0);
+	assert_rc_equal(rc, 0);
+
+	rc = cr_pool_create(state, &pool, false, TCC_NONE);
+	assert_rc_equal(rc, 0);
+
+	rc = cr_system_stop(false);
+	assert_rc_equal(rc, 0);
+
+	rc = cr_mode_switch(true);
+	assert_rc_equal(rc, 0);
+
+	rc = cr_check_start(TCSF_RESET, 0, NULL, NULL);
+	assert_rc_equal(rc, 0);
+
+	cr_ins_wait(1, &pool.pool_uuid, &dci);
+
+	rc = cr_ins_verify(&dci, TCIS_COMPLETED);
+	assert_rc_equal(rc, 0);
+
+	rc = cr_pool_verify(&dci, pool.pool_uuid, TCPS_CHECKED, 0, NULL, NULL, NULL);
+	assert_rc_equal(rc, 0);
+
+	/* Reint the rank for subsequent test. */
+	rc = cr_rank_reint(0);
+	assert_rc_equal(rc, 0);
+
+	rc = cr_mode_switch(false);
+	assert_rc_equal(rc, 0);
+
+	rc = cr_system_start();
+	assert_rc_equal(rc, 0);
+
+	cr_dci_fini(&dci);
+	cr_cleanup(arg, &pool, 1);
+}
+
 static const struct CMUnitTest cr_tests[] = {
 	{ "CR1: start checker for specified pools",
 	  cr_start_specified, async_disable, test_case_teardown},
@@ -3810,6 +3867,8 @@ static const struct CMUnitTest cr_tests[] = {
 	  cr_handle_fail_pool2, async_disable, test_case_teardown},
 	{ "CR28: maintenance mode after dry-run check",
 	  cr_maintenance_mode, async_disable, test_case_teardown},
+	{ "CR29: run CR with rank 0 excluded at the beginning",
+	  cr_lost_rank0, async_disable, test_case_teardown},
 };
 
 static int