third_party/linux: fix LACP issues
This fixes two major issues with the Linux LACP implementation:
First, the bond interface indicates carrier availability before any port
is even in aggregation state. It pretty much only cares about underlying
port carrier state which is not meaningful in LACP-controlled
aggregation.
Second, individual ports are added to the list of transmitting ports
immediately after coming up. This causes packets to be transmitted
before the LACP state indicates that this should happen.
Fix both of these issues by only enabling ports when the LACP state
machine places them in collecting/distributing state and making the bond
carrier state dependent on ports being enabled. This makes the interface
also behave logically consistent, i.e. it can transmit packets when its
carrier is reported up and not when its carrier is reported down.
While in there, fix some timer-related annoyances which make convergence
unnecessarily slow.
This also comes with a ktest which can be used for testing and
verification of these changes.
Change-Id: I60d0ed483f4f4ccea4d582b80e2bb29ff741783d
Reviewed-on: https://review.monogon.dev/c/monogon/+/3073
Reviewed-by: Serge Bazanski <serge@monogon.tech>
Tested-by: Jenkins CI
diff --git a/third_party/linux/external/lacp_fix.patch b/third_party/linux/external/lacp_fix.patch
new file mode 100644
index 0000000..b1eecfb
--- /dev/null
+++ b/third_party/linux/external/lacp_fix.patch
@@ -0,0 +1,102 @@
+From dd369885e1c0e3a6eff97db6acc5c765ee5dd421 Mon Sep 17 00:00:00 2001
+From: Lorenz Brun <lorenz@monogon.tech>
+Date: Thu, 2 May 2024 10:17:14 +0200
+Subject: [PATCH] bonding: 3ad: fix carrier and tx with no active ports
+
+bond carrier state should not be be up when no ports are in
+aggregation. Ports should only be enabled when they are in aggregation.
+
+While in there, set the default template to fast mode to quickly recover
+links and shorten the non-standard aggregator selection timer.
+---
+ drivers/net/bonding/bond_3ad.c | 39 ++++++++++++++++++----------------
+ 1 file changed, 21 insertions(+), 18 deletions(-)
+
+diff --git a/drivers/net/bonding/bond_3ad.c b/drivers/net/bonding/bond_3ad.c
+index c99ffe6c683a..a310c27ea659 100644
+--- a/drivers/net/bonding/bond_3ad.c
++++ b/drivers/net/bonding/bond_3ad.c
+@@ -699,6 +699,22 @@ static int __agg_active_ports(struct aggregator *agg)
+ return active;
+ }
+
++// __agg_enabled_ports counts the number of ports which are currently in
++// aggregation, different from __agg_active_ports which counts ports which
++// are up (but not in aggregation because of LACP).
++static int __agg_enabled_ports(struct aggregator *agg) {
++ struct port *port;
++ int en_count = 0;
++
++ for (port = agg->lag_ports; port;
++ port = port->next_port_in_aggregator) {
++ if (__port_is_enabled(port))
++ en_count++;
++ }
++
++ return en_count;
++}
++
+ /**
+ * __get_agg_bandwidth - get the total bandwidth of an aggregator
+ * @aggregator: the aggregator we're looking at
+@@ -1086,6 +1102,8 @@ static void ad_mux_machine(struct port *port, bool *update_slave_arr)
+ break;
+ }
+ }
++ // Update carrier state as we might have enabled/disabled ports
++ bond_3ad_set_carrier(__get_bond_by_port(port));
+ }
+
+ /**
+@@ -1780,21 +1798,6 @@ static void ad_agg_selection_logic(struct aggregator *agg,
+ *update_slave_arr = true;
+ }
+
+- /* if the selected aggregator is of join individuals
+- * (partner_system is NULL), enable their ports
+- */
+- active = __get_active_agg(origin);
+-
+- if (active) {
+- if (!__agg_has_partner(active)) {
+- for (port = active->lag_ports; port;
+- port = port->next_port_in_aggregator) {
+- __enable_port(port);
+- }
+- *update_slave_arr = true;
+- }
+- }
+-
+ rcu_read_unlock();
+
+ bond_3ad_set_carrier(bond);
+@@ -1852,7 +1855,7 @@ static void ad_initialize_port(struct port *port, int lacp_fast)
+ .key = 1,
+ .port_number = 1,
+ .port_priority = 0xff,
+- .port_state = 1,
++ .port_state = LACP_STATE_LACP_ACTIVITY | LACP_STATE_LACP_TIMEOUT,
+ };
+ static const struct lacpdu lacpdu = {
+ .subtype = 0x01,
+@@ -1994,7 +1997,7 @@ static void ad_marker_response_received(struct bond_marker *marker,
+ /* ========= AD exported functions to the main bonding code ========= */
+
+ /* Check aggregators status in team every T seconds */
+-#define AD_AGGREGATOR_SELECTION_TIMER 8
++#define AD_AGGREGATOR_SELECTION_TIMER 3
+
+ /**
+ * bond_3ad_initiate_agg_selection - initate aggregator selection
+@@ -2619,7 +2622,7 @@ int bond_3ad_set_carrier(struct bonding *bond)
+ active = __get_active_agg(&(SLAVE_AD_INFO(first_slave)->aggregator));
+ if (active) {
+ /* are enough slaves available to consider link up? */
+- if (__agg_active_ports(active) < bond->params.min_links) {
++ if (__agg_enabled_ports(active) < bond->params.min_links) {
+ if (netif_carrier_ok(bond->dev)) {
+ netif_carrier_off(bond->dev);
+ goto out;
+--
+2.42.0
+