Fix: consumer: dangling chunk on buffer allocation failure

[lttng-tools.git] / src / bin / lttng-sessiond / ust-app.c

diff --git a/src/bin/lttng-sessiond/ust-app.c b/src/bin/lttng-sessiond/ust-app.c
index d297cf0a30cb68f34b123576aacd6893edd9b5a7..41f201c48bb4ef8ea31b8c318c1ae95f2c4b434a 100644 (file)
--- a/src/bin/lttng-sessiond/ust-app.c
+++ b/src/bin/lttng-sessiond/ust-app.c
@@ -27,7 +27,6 @@
 #include <sys/types.h>
 #include <unistd.h>
 #include <urcu/compiler.h>
-#include <lttng/ust-error.h>
 #include <signal.h>
 
 #include <common/common.h>
@@ -38,13 +37,18 @@
 #include "health-sessiond.h"
 #include "ust-app.h"
 #include "ust-consumer.h"
-#include "ust-ctl.h"
+#include "lttng-ust-ctl.h"
+#include "lttng-ust-error.h"
 #include "utils.h"
 #include "session.h"
 #include "lttng-sessiond.h"
 #include "notification-thread-commands.h"
 #include "rotate.h"
 
+struct lttng_ht *ust_app_ht;
+struct lttng_ht *ust_app_ht_by_sock;
+struct lttng_ht *ust_app_ht_by_notify_sock;
+
 static
 int ust_app_flush_app_session(struct ust_app *app, struct ust_app_session *ua_sess);
 
@@ -741,6 +745,10 @@ error:
  * nullified. The session lock MUST be held unless the application is
  * in the destroy path.
  *
+ * Do not hold the registry lock while communicating with the consumerd, because
+ * doing so causes inter-process deadlocks between consumerd and sessiond with
+ * the metadata request notification.
+ *
  * Return 0 on success else a negative value.
  */
 static int close_metadata(struct ust_registry_session *registry,
@@ -748,6 +756,8 @@ static int close_metadata(struct ust_registry_session *registry,
 {
        int ret;
        struct consumer_socket *socket;
+       uint64_t metadata_key;
+       bool registry_was_already_closed;
 
        assert(registry);
        assert(consumer);
@@ -755,8 +765,19 @@ static int close_metadata(struct ust_registry_session *registry,
        rcu_read_lock();
 
        pthread_mutex_lock(&registry->lock);
+       metadata_key = registry->metadata_key;
+       registry_was_already_closed = registry->metadata_closed;
+       if (metadata_key != 0) {
+               /*
+                * Metadata closed. Even on error this means that the consumer
+                * is not responding or not found so either way a second close
+                * should NOT be emit for this registry.
+                */
+               registry->metadata_closed = 1;
+       }
+       pthread_mutex_unlock(&registry->lock);
 
-       if (!registry->metadata_key || registry->metadata_closed) {
+       if (metadata_key == 0 || registry_was_already_closed) {
                ret = 0;
                goto end;
        }
@@ -766,23 +787,15 @@ static int close_metadata(struct ust_registry_session *registry,
                        consumer);
        if (!socket) {
                ret = -1;
-               goto error;
+               goto end;
        }
 
-       ret = consumer_close_metadata(socket, registry->metadata_key);
+       ret = consumer_close_metadata(socket, metadata_key);
        if (ret < 0) {
-               goto error;
+               goto end;
        }
 
-error:
-       /*
-        * Metadata closed. Even on error this means that the consumer is not
-        * responding or not found so either way a second close should NOT be emit
-        * for this registry.
-        */
-       registry->metadata_closed = 1;
 end:
-       pthread_mutex_unlock(&registry->lock);
        rcu_read_unlock();
        return ret;
 }
@@ -1100,7 +1113,7 @@ struct ust_app_event *alloc_ust_app_event(char *name,
        /* Init most of the default value by allocating and zeroing */
        ua_event = zmalloc(sizeof(struct ust_app_event));
        if (ua_event == NULL) {
-               PERROR("malloc");
+               PERROR("Failed to allocate ust_app_event structure");
                goto error;
        }
 
@@ -3133,7 +3146,7 @@ int create_ust_app_event(struct ust_app_session *ua_sess,
 
        ua_event = alloc_ust_app_event(uevent->attr.name, &uevent->attr);
        if (ua_event == NULL) {
-               /* Only malloc can failed so something is really wrong */
+               /* Only failure mode of alloc_ust_app_event(). */
                ret = -ENOMEM;
                goto end;
        }
@@ -4109,11 +4122,14 @@ int ust_app_channel_create(struct ltt_ust_session *usess,
                ret = ust_app_channel_allocate(ua_sess, uchan,
                        LTTNG_UST_CHAN_PER_CPU, usess,
                        &ua_chan);
-               if (ret == 0) {
-                       ret = ust_app_channel_send(app, usess,
-                               ua_sess, ua_chan);
-               } else {
-                       goto end;
+               if (ret < 0) {
+                       goto error;
+               }
+
+               ret = ust_app_channel_send(app, usess,
+                       ua_sess, ua_chan);
+               if (ret) {
+                       goto error;
                }
 
                /* Add contexts. */
@@ -4121,10 +4137,12 @@ int ust_app_channel_create(struct ltt_ust_session *usess,
                        ret = create_ust_app_channel_context(ua_chan,
                                &uctx->ctx, app);
                        if (ret) {
-                               goto end;
+                               goto error;
                        }
                }
        }
+
+error:
        if (ret < 0) {
                switch (ret) {
                case -ENOTCONN:
@@ -4140,7 +4158,7 @@ int ust_app_channel_create(struct ltt_ust_session *usess,
                        break;
                }
        }
-end:
+
        if (ret == 0 && _ua_chan) {
                /*
                 * Only return the application's channel on success. Note
@@ -4212,7 +4230,7 @@ int ust_app_create_channel_glb(struct ltt_ust_session *usess,
                        if (session_was_created) {
                                destroy_app_session(app, ua_sess);
                        }
-                       goto error_rcu_unlock;
+                       /* Continue to the next application. */
                }
        }
 
@@ -6477,6 +6495,17 @@ enum lttng_error_code ust_app_create_channel_subdirectories(
        {
                struct ust_app *app;
 
+               /*
+                * Create the toplevel ust/ directory in case no apps are running.
+                */
+               chunk_status = lttng_trace_chunk_create_subdirectory(
+                               usess->current_trace_chunk,
+                               DEFAULT_UST_TRACE_DIR);
+               if (chunk_status != LTTNG_TRACE_CHUNK_STATUS_OK) {
+                       ret = LTTNG_ERR_CREATE_DIR_FAIL;
+                       goto error;
+               }
+
                cds_lfht_for_each_entry(ust_app_ht->ht, &iter.iter, app,
                                pid_n.node) {
                        struct ust_app_session *ua_sess;