[PATCH] Fix-up stale pointer after reconnect

[Matthew Cather]

While investigating an issue with hostapd, I seem to have stumbled
onto a bug in `hostapd_cli`. The patch below is a sugested fix but
there may be a better approach.

If our `PING` fails and we close the connection then our `ctrl`
pointer becomes invalid. After reconnect we should replace `ctrl`
with a valid pointer.

Signed-off-by: Matthew Cather <mattbob4 at gmail.com>
---
 hostapd/hostapd_cli.c | 8 +++++---
 1 file changed, 5 insertions(+), 3 deletions(-)

diff --git a/hostapd/hostapd_cli.c b/hostapd/hostapd_cli.c
index 95f7fd2d8..5efeceaf0 100644
--- a/hostapd/hostapd_cli.c
+++ b/hostapd/hostapd_cli.c
@@ -2189,9 +2189,11 @@ static void hostapd_cli_action_ping(void *eloop_ctx, void *timeout_ctx)
 			eloop_terminate();
 			return;
 		}
+		eloop_cancel_timeout(hostapd_cli_action_ping, ctrl, timeout_ctx);
+		ctrl = ctrl_conn;
 	}
 	eloop_register_timeout(ping_interval, 0, hostapd_cli_action_ping,
-			       ctrl, NULL);
+			       ctrl, timeout_ctx);
 }
 
 
@@ -2210,10 +2212,10 @@ static void hostapd_cli_action(struct wpa_ctrl *ctrl)
 
 	fd = wpa_ctrl_get_fd(ctrl);
 	eloop_register_timeout(ping_interval, 0, hostapd_cli_action_ping,
-			       ctrl, NULL);
+			       ctrl, ctrl);
 	eloop_register_read_sock(fd, hostapd_cli_action_receive, ctrl, NULL);
 	eloop_run();
-	eloop_cancel_timeout(hostapd_cli_action_ping, ctrl, NULL);
+	eloop_cancel_timeout(hostapd_cli_action_ping, ELOOP_ALL_CTX, ctrl);
 	eloop_unregister_read_sock(fd);
 }
 
-- 
2.43.0