[PATCH BlueZ v1 05/27] l2cap-tester: Fix race in LE Client - Close socket 2

[Luiz Augusto von Dentz <luiz.dentz@gmail.com>]

From: Luiz Augusto von Dentz <luiz.von.dentz@intel.com>

The test closes the socket once the scan is disabled, expecting the
kernel to cancel the connection attempt with LE Create Connection
Cancel, but the kernel only sends LE Create Connection after disabling
the scan, so whether there is an attempt to cancel depends on which runs
first. Close the socket once LE Create Connection was sent instead.

This shows under User Mode Linux time travel, where the test always
closed the socket too early.

Assisted-by: Claude:claude-opus-5-5
---
 tools/l2cap-tester.c | 14 +++++++++++++-
 1 file changed, 13 insertions(+), 1 deletion(-)

diff --git a/tools/l2cap-tester.c b/tools/l2cap-tester.c
index f2d3e56b8761..c2643b4182b0 100644
--- a/tools/l2cap-tester.c
+++ b/tools/l2cap-tester.c
@@ -2769,7 +2769,17 @@ static void test_close_socket_scan_disabled(void)
 
 	if (l2data == &le_client_close_socket_test_1)
 		g_idle_add(test_close_socket_1_part_3, NULL);
-	else if (l2data == &le_client_close_socket_test_2)
+}
+
+/* The kernel creates the connection once the scan is disabled, close the
+ * socket only then, so it has a connection attempt to cancel.
+ */
+static void test_close_socket_create_conn(void)
+{
+	struct test_data *data = tester_get_data();
+	const struct l2cap_data *l2data = data->test_data;
+
+	if (l2data == &le_client_close_socket_test_2)
 		g_idle_add(test_close_socket_2_part_3, NULL);
 }
 
@@ -2793,6 +2803,8 @@ static void test_close_socket_router(uint16_t opcode, const void *param,
 			test_close_socket_scan_enabled();
 		else
 			test_close_socket_scan_disabled();
+	} else if (opcode == BT_HCI_CMD_LE_CREATE_CONN) {
+		test_close_socket_create_conn();
 	} else if (opcode == BT_HCI_CMD_LE_CREATE_CONN_CANCEL) {
 		test_close_socket_conn_cancel();
 	}
-- 
2.56.0