Fix possible deadlock where worker process sends an event to the main process

while the main process sends control message to the worker process, but worker
process hasn't started control thread yet, because it waits for reply from the
main process.

The fix is to start the control thread before sending any events.

Reported and fix suggested by:	Mikolaj Golub <to.my.trociny@gmail.com>
MFC after:	3 days

1 files changed, 12 insertions, 4 deletions

diff --git a/sbin/hastd/primary.c b/sbin/hastd/primary.c
index 0b2402a..5d6896e 100644
--- a/sbin/hastd/primary.c
+++ b/sbin/hastd/primary.c
@@ -807,10 +807,20 @@ hastd_primary(struct hast_resource *res)
 	proto_send(res->hr_event, NULL, 0);
 
 	init_local(res);
-	if (real_remote(res) && init_remote(res, NULL, NULL))
-		sync_start();
 	init_ggate(res);
 	init_environment(res);
+	/*
+	 * Create the control thread before sending any event to the parent,
+	 * as we can deadlock when parent sends control request to worker,
+	 * but worker has no control thread started yet, so parent waits.
+	 * In the meantime worker sends an event to the parent, but parent
+	 * is unable to handle the event, because it waits for control
+	 * request response.
+	 */
+	error = pthread_create(&td, NULL, ctrl_thread, res);
+	assert(error == 0);
+	if (real_remote(res) && init_remote(res, NULL, NULL))
+		sync_start();
 	error = pthread_create(&td, NULL, ggate_recv_thread, res);
 	assert(error == 0);
 	error = pthread_create(&td, NULL, local_send_thread, res);
@@ -823,8 +833,6 @@ hastd_primary(struct hast_resource *res)
 	assert(error == 0);
 	error = pthread_create(&td, NULL, sync_thread, res);
 	assert(error == 0);
-	error = pthread_create(&td, NULL, ctrl_thread, res);
-	assert(error == 0);
 	(void)guard_thread(res);
 }