flux-resource: direct status to rank 0

Problem: if flux resource status is invoked on a rank other
than zero, a segfault occurs.

Explicitly set nodeid=0 in the command's resource.status RPC call.

In additon, check the rank in the resource module's request handler
and return an error response if invoked on a rank other than zero,
rather than segfault.

Add tests.

Fixes #3484

Author: garlick

src/cmd/flux-resource.py

@@ -274,7 +274,7 @@ def status(args):
     if args.from_stdin:
         resp = sys.stdin.read()
     else:
-        resp = RPC(flux.Flux(), "resource.status").get()
+        resp = RPC(flux.Flux(), "resource.status", nodeid=0).get()
 
     rstat = ResourceStatus.from_status_response(resp, fmt)
 

src/modules/resource/resource.c

@@ -156,15 +156,18 @@ static void status_cb (flux_t *h,
                        void *arg)
 {
     struct resource_ctx *ctx = arg;
-    const struct idset *online = monitor_get_up (ctx->monitor);
-    const struct idset *offline = monitor_get_down (ctx->monitor);
-    const struct idset *exclude = exclude_get (ctx->exclude);
     json_t *drain;
     const json_t *R;
     json_t *o = NULL;
+    const char *errstr = NULL;
 
     if (flux_request_decode (msg, NULL, NULL) < 0)
         goto error;
+    if (ctx->rank != 0) {
+        errno = EPROTO;
+        errstr = "this RPC only works on rank 0";
+        goto error;
+    }
     if (!(R = inventory_get (ctx->inventory)))
         goto error;
     if (!(drain = drain_get_info (ctx->drain)))
@@ -174,19 +177,25 @@ static void status_cb (flux_t *h,
         errno = ENOMEM;
         goto error;
     }
-    if (rutil_set_json_idset (o, "online", online) < 0)
+    if (rutil_set_json_idset (o,
+                              "online",
+                              monitor_get_up (ctx->monitor)) < 0)
         goto error;
-    if (rutil_set_json_idset (o, "offline", offline) < 0)
+    if (rutil_set_json_idset (o,
+                              "offline",
+                              monitor_get_down (ctx->monitor)) < 0)
         goto error;
-    if (rutil_set_json_idset (o, "exclude", exclude) < 0)
+    if (rutil_set_json_idset (o,
+                              "exclude",
+                              exclude_get (ctx->exclude)) < 0)
         goto error;
     if (flux_respond_pack (h, msg, "o", o) < 0) {
         flux_log_error (h, "error responding to resource.status request");
         json_decref (o);
     }
     return;
 error:
-    if (flux_respond_error (h, msg, errno, NULL) < 0)
+    if (flux_respond_error (h, msg, errno, errstr) < 0)
         flux_log_error (h, "error responding to resource.status request");
     json_decref (o);
 }

t/t2310-resource-module.t

@@ -195,6 +195,19 @@ test_expect_success 'resource.get-xml blocks until all ranks are up' '
 	flux python ${SHARNESS_TEST_SRCDIR}/resource/get-xml-test.py
 '
 
+test_expect_success 'flux resource status works on rank > 0' '
+	flux exec -r 1 flux resource status
+'
+
+status_onrank() {
+	flux python -c "import flux; print(flux.Flux().rpc(\"resource.status\",nodeid=$1).get())"
+}
+
+test_expect_success 'resource.status RPC fails on rank > 0' '
+	test_must_fail status_onrank 1 2>status.err &&
+	grep "only works on rank 0" status.err
+'
+
 test_expect_success 'unload resource module' '
 	flux exec -r all flux module remove resource
 '