Fix: hash table cleanup call_rcu deadlock

[lttng-tools.git] / src / bin / lttng-sessiond / lttng-sessiond.h

diff --git a/src/bin/lttng-sessiond/lttng-sessiond.h b/src/bin/lttng-sessiond/lttng-sessiond.h
index 9258f38a7c1018d27ee91de0f3817f92efe90bd6..6090e086c4f257311c57f068aa75ef71c585c393 100644 (file)
--- a/src/bin/lttng-sessiond/lttng-sessiond.h
+++ b/src/bin/lttng-sessiond/lttng-sessiond.h
@@ -71,7 +71,23 @@ struct ust_cmd_queue {
  */
 extern int apps_cmd_notify_pipe[2];
 
+/*
+ * Used to notify that a hash table needs to be destroyed by dedicated
+ * thread. Required by design because we don't want to move destroy
+ * paths outside of large RCU read-side lock paths, and destroy cannot
+ * be called by call_rcu thread, because it may hang (waiting for
+ * call_rcu completion).
+ */
+extern int ht_cleanup_pipe[2];
+
+/*
+ * Populated when the daemon starts with the current page size of the system.
+ */
+extern long page_size;
+
 int sessiond_set_thread_pollset(struct lttng_poll_event *events, size_t size);
 int sessiond_check_thread_quit_pipe(int fd, uint32_t events);
 
+void *thread_ht_cleanup(void *data);
+
 #endif /* _LTT_SESSIOND_H */