X-Git-Url: https://git.lttng.org/?a=blobdiff_plain;f=src%2Fbin%2Flttng-sessiond%2Fust-app.c;h=e1d0c40e9cf03e34b935a171b1ca8d67686f55f5;hb=bdd88757eedddd53dea46645574f218094573c74;hp=3accde269509e0e8b66c3b25c35b86a90289fb5b;hpb=d42f20df7f4ef656485ebb19c94a35152cb386f7;p=lttng-tools.git

diff --git a/src/bin/lttng-sessiond/ust-app.c b/src/bin/lttng-sessiond/ust-app.c
index 3accde269..e1d0c40e9 100644
--- a/src/bin/lttng-sessiond/ust-app.c
+++ b/src/bin/lttng-sessiond/ust-app.c
@@ -1212,11 +1212,12 @@ error:
 }
 
 /*
- * Create UST app channel and create it on the tracer.
+ * Create UST app channel and create it on the tracer. Set ua_chanp of the
+ * newly created channel if not NULL.
  */
-static struct ust_app_channel *create_ust_app_channel(
-		struct ust_app_session *ua_sess, struct ltt_ust_channel *uchan,
-		struct ust_app *app)
+static int create_ust_app_channel(struct ust_app_session *ua_sess,
+		struct ltt_ust_channel *uchan, struct ust_app *app,
+		struct ust_app_channel **ua_chanp)
 {
 	int ret = 0;
 	struct lttng_ht_iter iter;
@@ -1234,6 +1235,7 @@ static struct ust_app_channel *create_ust_app_channel(
 	ua_chan = alloc_ust_app_channel(uchan->name, &uchan->attr);
 	if (ua_chan == NULL) {
 		/* Only malloc can fail here */
+		ret = -ENOMEM;
 		goto error;
 	}
 	shadow_copy_channel(ua_chan, uchan);
@@ -1245,17 +1247,23 @@ static struct ust_app_channel *create_ust_app_channel(
 		goto error;
 	}
 
+	/* Only add the channel if successful on the tracer side. */
 	lttng_ht_add_unique_str(ua_sess->channels, &ua_chan->node);
 
 	DBG2("UST app create channel %s for PID %d completed", ua_chan->name,
 			app->pid);
 
 end:
-	return ua_chan;
+	if (ua_chanp) {
+		*ua_chanp = ua_chan;
+	}
+
+	/* Everything went well. */
+	return 0;
 
 error:
 	delete_ust_app_channel(-1, ua_chan);
-	return NULL;
+	return ret;
 }
 
 /*
@@ -2021,7 +2029,6 @@ int ust_app_create_channel_glb(struct ltt_ust_session *usess,
 	struct lttng_ht_iter iter;
 	struct ust_app *app;
 	struct ust_app_session *ua_sess;
-	struct ust_app_channel *ua_chan;
 
 	/* Very wrong code flow */
 	assert(usess);
@@ -2049,26 +2056,27 @@ int ust_app_create_channel_glb(struct ltt_ust_session *usess,
 		ua_sess = create_ust_app_session(usess, app);
 		if (ua_sess == NULL) {
 			/* The malloc() failed. */
-			ret = -1;
-			goto error;
+			ret = -ENOMEM;
+			goto error_rcu_unlock;
 		} else if (ua_sess == (void *) -1UL) {
-			/* The application's socket is not valid. Contiuing */
-			ret = -1;
+			/*
+			 * The application's socket is not valid. Either a bad socket or a
+			 * timeout on it. We can't inform yet the caller that for a
+			 * specific app, the session failed so we continue here.
+			 */
 			continue;
 		}
 
-		/* Create channel onto application */
-		ua_chan = create_ust_app_channel(ua_sess, uchan, app);
-		if (ua_chan == NULL) {
-			/* Major problem here and it's maybe the tracer or malloc() */
-			ret = -1;
-			goto error;
+		/* Create channel onto application. We don't need the chan ref. */
+		ret = create_ust_app_channel(ua_sess, uchan, app, NULL);
+		if (ret < 0 && ret == -ENOMEM) {
+			/* No more memory is a fatal error. Stop right now. */
+			goto error_rcu_unlock;
 		}
 	}
 
+error_rcu_unlock:
 	rcu_read_unlock();
-
-error:
 	return ret;
 }
 
@@ -2244,9 +2252,6 @@ int ust_app_start_trace(struct ltt_ust_session *usess, struct ust_app *app)
 		}
 	}
 
-	/* Indicate that the session has been started once */
-	ua_sess->started = 1;
-
 	ret = create_ust_app_metadata(ua_sess, usess->pathname, app);
 	if (ret < 0) {
 		ret = LTTNG_ERR_UST_META_FAIL;
@@ -2344,6 +2349,9 @@ skip_setup:
 		goto error_rcu_unlock;
 	}
 
+	/* Indicate that the session has been started once */
+	ua_sess->started = 1;
+
 	health_code_update(&health_thread_cmd);
 
 	/* Quiescent wait after starting trace */
@@ -2385,10 +2393,13 @@ int ust_app_stop_trace(struct ltt_ust_session *usess, struct ust_app *app)
 
 	/*
 	 * If started = 0, it means that stop trace has been called for a session
-	 * that was never started. This is a code flow error and should never
-	 * happen.
+	 * that was never started. It's possible since we can have a fail start
+	 * from either the application manager thread or the command thread. Simply
+	 * indicate that this is a stop error.
 	 */
-	assert(ua_sess->started == 1);
+	if (ua_sess->started == 1) {
+		goto error_rcu_unlock;
+	}
 
 	health_code_update(&health_thread_cmd);
 
@@ -2442,7 +2453,7 @@ error_rcu_unlock:
 /*
  * Destroy a specific UST session in apps.
  */
-int ust_app_destroy_trace(struct ltt_ust_session *usess, struct ust_app *app)
+static int destroy_trace(struct ltt_ust_session *usess, struct ust_app *app)
 {
 	struct ust_app_session *ua_sess;
 	struct lttng_ust_object_data obj;
@@ -2556,7 +2567,7 @@ int ust_app_destroy_trace_all(struct ltt_ust_session *usess)
 	rcu_read_lock();
 
 	cds_lfht_for_each_entry(ust_app_ht->ht, &iter.iter, app, pid_n.node) {
-		ret = ust_app_destroy_trace(usess, app);
+		ret = destroy_trace(usess, app);
 		if (ret < 0) {
 			/* Continue to next apps even on error */
 			continue;
@@ -2581,10 +2592,7 @@ void ust_app_global_update(struct ltt_ust_session *usess, int sock)
 	struct ust_app_event *ua_event;
 	struct ust_app_ctx *ua_ctx;
 
-	if (usess == NULL) {
-		ERR("No UST session on global update. Returning");
-		goto error;
-	}
+	assert(usess);
 
 	DBG2("UST app global update for app sock %d for session id %d", sock,
 			usess->id);