From ec63c0d10b8d33642c3741a1bcfa889ccc3a5f9e Mon Sep 17 00:00:00 2001
From: Rusty Russell <rusty@rustcorp.com.au>
Date: Tue, 12 Sep 2017 14:25:54 +0930
Subject: [PATCH] lightningd: give option to crash if a subdaemon fails.

Either when it exits with a signal, or sends an error status message.
Then we make test_lightningd.py use it.

Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
---
 lightningd/lightningd.h  |  3 +++
 lightningd/options.c     |  2 ++
 lightningd/subd.c        | 13 +++++++++++++
 tests/test_lightningd.py |  1 +
 4 files changed, 19 insertions(+)

diff --git a/lightningd/lightningd.h b/lightningd/lightningd.h
index e2aaa3d99..93464d953 100644
--- a/lightningd/lightningd.h
+++ b/lightningd/lightningd.h
@@ -112,6 +112,9 @@ struct lightningd {
 	/* If we have a --dev-disconnect file */
 	int dev_disconnect_fd;
 
+	/* If we have --dev-fail-on-subdaemon-fail */
+	bool dev_subdaemon_fail;
+
 	/* HTLCs in flight. */
 	struct htlc_in_map htlcs_in;
 	struct htlc_out_map htlcs_out;
diff --git a/lightningd/options.c b/lightningd/options.c
index 45f7422e1..fb171108c 100644
--- a/lightningd/options.c
+++ b/lightningd/options.c
@@ -239,6 +239,8 @@ static void dev_register_opts(struct lightningd *ld)
 {
 	opt_register_noarg("--dev-no-broadcast", opt_set_bool,
 			   &ld->topology->dev_no_broadcast, opt_hidden);
+	opt_register_noarg("--dev-fail-on-subdaemon-fail", opt_set_bool,
+			   &ld->dev_subdaemon_fail, opt_hidden);
 }
 
 static const struct config testnet_config = {
diff --git a/lightningd/subd.c b/lightningd/subd.c
index 695938658..f902e0a0d 100644
--- a/lightningd/subd.c
+++ b/lightningd/subd.c
@@ -318,6 +318,9 @@ static void subdaemon_malformed_msg(struct subd *sd, const u8 *msg)
 		   tal_hexstr(msg,
 			      msg + sizeof(be16),
 			      tal_count(msg) - sizeof(be16)));
+
+	if (sd->ld->dev_subdaemon_fail)
+		fatal("Subdaemon %s sent malformed message", sd->name);
 }
 
 /* Returns true if logged, false if malformed. */
@@ -361,6 +364,10 @@ log_str_peer:
 /* Shouldn't happen. */
 log_str_broken:
 	log_broken(sd->log, "%s: %.*s", name, str_len, str);
+
+	if (sd->ld->dev_subdaemon_fail)
+		fatal("Subdaemon %s hit error", sd->name);
+
 	return true;
 }
 
@@ -444,12 +451,14 @@ next:
 static void destroy_subd(struct subd *sd)
 {
 	int status;
+	bool fail_if_subd_fails = sd->ld->dev_subdaemon_fail;
 
 	switch (waitpid(sd->pid, &status, WNOHANG)) {
 	case 0:
 		log_debug(sd->log, "Status closed, but not exited. Killing");
 		kill(sd->pid, SIGKILL);
 		waitpid(sd->pid, &status, 0);
+		fail_if_subd_fails = false;
 		break;
 	case -1:
 		log_unusual(sd->log, "Status closed, but waitpid %i says %s",
@@ -458,6 +467,10 @@ static void destroy_subd(struct subd *sd)
 		break;
 	}
 
+	if (fail_if_subd_fails && WIFSIGNALED(status))
+		fatal("Subdaemon %s killed with signal %i",
+		      sd->name, WTERMSIG(status));
+
 	/* In case we're freed manually, such as peer_fail_permanent */
 	if (sd->conn)
 		sd->conn = tal_free(sd->conn);
diff --git a/tests/test_lightningd.py b/tests/test_lightningd.py
index 0cac522eb..f23cbbca2 100644
--- a/tests/test_lightningd.py
+++ b/tests/test_lightningd.py
@@ -101,6 +101,7 @@ class NodeFactory(object):
             with open(os.path.join(lightning_dir, "dev_disconnect"), "w") as f:
                 f.write("\n".join(disconnect))
             daemon.cmd_line.append("--dev-disconnect=dev_disconnect")
+        daemon.cmd_line.append("--dev-fail-on-subdaemon-fail")
         rpc = LightningRpc(socket_path, self.executor)
 
         node = utils.LightningNode(daemon, rpc, bitcoind, self.executor)