From 78b3971782dd726e45cf619f15c75602c1bca59c Mon Sep 17 00:00:00 2001
From: d00f <8052722+DrDoof@users.noreply.github.com>
Date: Sun, 16 Aug 2026 11:46:02 +0200
Subject: [PATCH] stp: pass a received topology change through the switch

The tree structure already crossed the switch by regeneration, but the
topology-change information did not: a received TC flag was ignored and
a TCN only acknowledged, so bridges behind this one kept stale entries
until normal aging.

A TC flag in a received BPDU now flushes the other non-edge ports once
and arms the transmit window our BPDUs already copy the flag from,
refreshed to hello+1 seconds by every further flagged frame so it ends
one hello after the neighbour stops, without shortening the longer
window a local change arms. A TCN is acknowledged as before and then
treated like a local change on that port.
---
 doc/stp.md    | 13 +++++++++++++
 rtl837x_stp.c | 28 +++++++++++++++++++++++++---
 2 files changed, 38 insertions(+), 3 deletions(-)

diff --git a/doc/stp.md b/doc/stp.md
index 61bb5e5..0ccca7c 100644
--- a/doc/stp.md
+++ b/doc/stp.md
@@ -80,6 +80,16 @@ A port entering the tree spends `fwd` seconds in blocking before it forwards
 (an edge port skips the wait). Root information is discarded after `maxage`
 seconds without a BPDU, and the switch then reclaims the root role.
 
+## Topology changes
+
+A change on a local non-edge port (the link coming or going, a port promoted
+to forwarding) flushes the addresses learned on it and sets the TC flag in
+our BPDUs for `maxage + fwd` seconds. A TC flag received in a BPDU is passed
+on: the switch flushes the other non-edge ports once and keeps the flag in
+its own BPDUs until one hello after the last flagged frame, so the
+notification crosses the switch instead of dying at it. A legacy TCN is
+acknowledged with TCA and then treated like a local change.
+
 ## Bridge settings
 
 ```
@@ -139,3 +149,6 @@ The `stp status` command prints the same view on the serial console.
   converge, but through the timers rather than the fast transition.
 * Port roles are approximated: the root port and designated ports are
   distinguished, alternate/backup are not.
+* Topology changes propagate away from the root only: nothing is announced
+  on the root port (no TCN and no BPDUs at all), so bridges upstream rely on
+  their own detection.
diff --git a/rtl837x_stp.c b/rtl837x_stp.c
index 87bfb22..5bafe17 100644
--- a/rtl837x_stp.c
+++ b/rtl837x_stp.c
@@ -501,12 +501,14 @@ void stp_in(void) __banked
 
 	if (STP_I->bpdu_type == 0x80) {
 		/* TCN: a downstream bridge reports a topology change. Acknowledge it
-		 * on this port so the sender stops repeating; the change itself is
-		 * counted (and, once implemented, propagated rootward). */
+		 * on this port so the sender stops repeating, then treat it like a
+		 * change of our own: flush the port and carry the TC flag on the
+		 * designated ports for the full window. No TCN goes towards the
+		 * root, so a legacy root further up keeps its normal aging. */
 		stp_tx_flags_extra = 0x80;	/* Topology Change Acknowledgment */
 		stp_cnf_send(port);		/* transmits internally */
 		uip_len = 0;			/* ...so handle_rx must not TX again */
-		stp_tc_count++;
+		stp_topology_change(port);
 		return;
 	}
 
@@ -563,6 +565,26 @@ void stp_in(void) __banked
 		return;
 	}
 
+	/* Topology Change in transit. The flag arms a short window that our
+	 * own BPDUs copy downstream (the TX side already sends TC while
+	 * stp_tc_while runs) and that each further flagged BPDU refreshes, so
+	 * it expires one hello after the neighbour stops - without shortening
+	 * the long window a local change may have armed. The flush runs once,
+	 * on the arming edge: everything learned on the other non-edge ports
+	 * may sit behind the moved link and must be relearned. */
+	if (STP_I->flags & 0x01) {
+		if (!stp_tc_while) {
+			uint8_t i;
+			stp_tc_count++;
+			for (i = machine.min_port; i <= machine.max_port; i++)
+				if (i != port && (stp_pflags[i] & STP_PF_ENABLED)
+				    && !(stp_pflags[i] & STP_PF_OPEREDGE))
+					port_l2_forget_port(i);
+		}
+		if (stp_tc_while < ((uint16_t)stp_hello_s + 1) * STP_HZ)
+			stp_tc_while = ((uint16_t)stp_hello_s + 1) * STP_HZ;
+	}
+
 	stp_record_designated(port);
 
 	/* Better root than the one we know? The identifier is priority, system