BUG/MEDIUM: resume from LI_ASSIGNED in default_resume_listener()
Since fc974887c ("MEDIUM: protocol: explicitly start the receiver before
the listener"), resume from LI_ASSIGNED state does not work anymore.
This is because the binding part has been divided into 2 distinct steps
since: first bind(), then listen().
This new logic was properly implemented in startup sequence
through protocol_bind_all() but wasn't properly reported in
default_resume_listener() function.
Fixing default_resume_listener() to comply with the new logic.
This should help ABNS sockets to properly rebind in resume_listener()
after they have been stopped by pause_listener():
See Redmine:4475 for more context.
This commit depends on:
- "MINOR: listener: workaround for closing a tiny race between resume_listener() and stopping"
- "MINOR: listener: make sure we don't pause/resume bypassed listeners"
This could be backported up to 2.4 after a reasonable observation period to
make sure that this change doesn't cause unwanted side-effects.
diff --git a/src/listener.c b/src/listener.c
index 82c8631..dfd1fd9 100644
--- a/src/listener.c
+++ b/src/listener.c
@@ -427,8 +427,28 @@
if (l->state == LI_ASSIGNED) {
char msg[100];
+ char *errmsg;
int err;
+ /* first, try to bind the receiver */
+ err = l->rx.proto->fam->bind(&l->rx, &errmsg);
+ if (err != ERR_NONE) {
+ if (err & ERR_WARN)
+ ha_warning("Resuming listener: %s\n", errmsg);
+ else if (err & ERR_ALERT)
+ ha_alert("Resuming listener: %s\n", errmsg);
+ ha_free(&errmsg);
+ if (err & (ERR_FATAL | ERR_ABORT)) {
+ ret = 0;
+ goto end;
+ }
+ }
+
+ /* then, try to listen:
+ * for now there's still always a listening function
+ * (same check performed in protocol_bind_all()
+ */
+ BUG_ON(!l->rx.proto->listen);
err = l->rx.proto->listen(l, msg, sizeof(msg));
if (err & ERR_ALERT)
ha_alert("Resuming listener: %s\n", msg);