From: Nick Porter <nick@portercomputing.co.uk>
Date: Thu, 9 Oct 2025 14:17:48 +0000 (+0100)
Subject: Handle cancelling SQL requests that are already reapable
X-Git-Url: http://git.ipfire.org/cgi-bin/gitweb.cgi?a=commitdiff_plain;h=601a41f2609fd9e647bdcb828b2ee71058a75fe2;p=thirdparty%2Ffreeradius-server.git

Handle cancelling SQL requests that are already reapable

These are really complete, and don't need to go round the query
cancellation path.
---

diff --git a/src/modules/rlm_sql/sql.c b/src/modules/rlm_sql/sql.c
index 6c340b94753..9e893333697 100644
--- a/src/modules/rlm_sql/sql.c
+++ b/src/modules/rlm_sql/sql.c
@@ -212,6 +212,15 @@ static void sql_trunk_query_cancel(UNUSED request_t *request, UNUSED fr_signal_t
 
 	if (!query_ctx->treq) return;
 
+	/*
+	 *	A reapable trunk request has already completed.
+	 */
+	if (unlikely(query_ctx->treq->state == TRUNK_REQUEST_STATE_REAPABLE)) {
+		trunk_request_signal_complete(query_ctx->treq);
+		query_ctx->treq = NULL;
+		return;
+	}
+
 	/*
 	 *	The query_ctx needs to be parented by the treq so that it still exists
 	 *	when the cancel_mux callback is run.