Make changes to the transition mechanism

First, reply with the most recently received LACP PDU version. This is
so that if one side is sending both the old and new versions, then the
peer side will send both versions to keep both listeners happy. This
also allows for using the new version packets even when the retry count
is 3 on both sides. This can only be triggered remotely; there's no
local command to force a particular version to be used.

Second, reset the 60-second timer preventing the retry count from being
reset back to 3 every time a new version PDU is received. This keeps
that setting persistent for longer with the downside of a delayed
response to the retry count being reset to 3.

Signed-off-by: Saikrishna Arcot <sarcot@microsoft.com>

src/libteam/patch/0016-block-retry-count-changes.patch

@@ -5,13 +5,13 @@ Date: 2023-01-18 14:26:36 -0800
 
 After setting the retry count to some custom value, if a normal LACP
 packet comes in without a custom retry count, don't reset it back to
-that custom retry count for the first 60 seconds.
+that custom retry count for 60 seconds since the last new packet.
 ---
- teamd/teamd_runner_lacp.c |   40 ++++++++++++++++++++++++++++++----------
- 1 file changed, 30 insertions(+), 10 deletions(-)
+ teamd/teamd_runner_lacp.c |   44 +++++++++++++++++++++++++++++++++++---------
+ 1 file changed, 35 insertions(+), 9 deletions(-)
 
 diff --git a/teamd/teamd_runner_lacp.c b/teamd/teamd_runner_lacp.c
-index 4b77692..5a55bd5 100644
+index 5c7e7b4..4f971d9 100644
 --- a/teamd/teamd_runner_lacp.c
 +++ b/teamd/teamd_runner_lacp.c
 @@ -180,6 +180,7 @@ struct lacp {
@@ -22,7 +22,36 @@ index 4b77692..5a55bd5 100644
  	struct {
  		bool active;
  #define		LACP_CFG_DFLT_ACTIVE true
-@@ -1457,9 +1458,12 @@ static int lacpdu_send(struct lacp_port *lacp_port)
+@@ -232,6 +233,7 @@ struct lacp_port {
+ 	struct lacp_port *agg_lead; /* leading port of aggregator.
+ 				     * NULL in case this port is not selected */
+ 	enum lacp_port_state state;
++	int last_received_lacpdu_version;
+ 	bool lacpdu_saved;
+ 	struct lacpdu last_pdu;
+ 	struct {
+@@ -1331,6 +1333,7 @@ static int lacp_port_set_state(struct lacp_port *lacp_port,
+ 		if (err)
+ 			return err;
+ 		lacp_port->lacp->cfg.retry_count = LACP_CFG_DFLT_RETRY_COUNT;
++		lacp_port->last_received_lacpdu_version = 0x01;
+ 		teamd_loop_callback_disable(lacp_port->ctx,
+ 					    LACP_RETRY_COUNT_TIMEOUT_CB_NAME, lacp_port->lacp);
+ 		lacp_port_timeout_set(lacp_port, true);
+@@ -1439,10 +1442,10 @@ static int lacpdu_send(struct lacp_port *lacp_port)
+ 	if (hwaddr_len != ETH_ALEN)
+ 		return 0;
+
+-	if (lacp_port->lacp->cfg.retry_count != LACP_CFG_DFLT_RETRY_COUNT || lacp_port->partner_retry_count != LACP_CFG_DFLT_RETRY_COUNT)
++	if (lacp_port->lacp->cfg.retry_count != LACP_CFG_DFLT_RETRY_COUNT)
+ 		lacpdu_init(&lacpdu, 0xf1);
+ 	else
+-		lacpdu_init(&lacpdu, 0x01);
++		lacpdu_init(&lacpdu, lacp_port->last_received_lacpdu_version);
+ 	lacpdu.actor = lacp_port->actor;
+ 	lacpdu.partner = lacp_port->partner;
+ 	memcpy(lacpdu.hdr.ether_shost, hwaddr, hwaddr_len);
+@@ -1457,9 +1460,12 @@ static int lacpdu_send(struct lacp_port *lacp_port)
  	return err;
  }
 
@@ -35,26 +64,20 @@ index 4b77692..5a55bd5 100644
 
  	if (!lacpdu_check(lacpdu)) {
  		teamd_log_warn("malformed LACP PDU came.");
-@@ -1489,18 +1493,34 @@ static int lacpdu_process(struct lacp_port *lacp_port, struct lacpdu* lacpdu)
+@@ -1493,17 +1499,38 @@ static int lacpdu_process(struct lacp_port *lacp_port, struct lacpdu* lacpdu)
+ 					lacp_port->partner_retry_count,
+ 					lacpdu->v2.actor_retry_count);
+ 			lacp_port->partner_retry_count = lacpdu->v2.actor_retry_count;
++			if (clock_gettime(CLOCK_MONOTONIC, &monotonic_time)) {
++				err = errno;
++				teamd_log_err("%s: unable to get current time: %s", lacp_port->tdport->ifname, strerror(err));
++				return -err;
++			}
++		}
++		if (lacp_port->partner_retry_count != LACP_CFG_DFLT_RETRY_COUNT) {
++			// Reset the change time every time a 0xf1 packet comes in
++			lacp_port->lacp->next_retry_count_change_time = monotonic_time.tv_sec + LACP_NEXT_RETRY_COUNT_CHANGE_TIMEOUT_SECONDS;
  		}
- 		if (lacp_port->partner_retry_count != lacpdu->v2.actor_retry_count) {
- 			teamd_log_dbg(lacp_port->ctx, "%s: retry count from partner changed from %u to %u",
--					lacp_port->tdport->ifname,
--					lacp_port->partner_retry_count,
--					lacpdu->v2.actor_retry_count);
--			lacp_port->partner_retry_count = lacpdu->v2.actor_retry_count;
--		}
-+					   lacp_port->tdport->ifname,
-+					   lacp_port->partner_retry_count,
-+					   lacpdu->v2.actor_retry_count);
-+            lacp_port->partner_retry_count = lacpdu->v2.actor_retry_count;
-+            if (clock_gettime(CLOCK_MONOTONIC, &monotonic_time)) {
-+                err = errno;
-+                teamd_log_err("%s: unable to get current time: %s", lacp_port->tdport->ifname, strerror(err));
-+                return -err;
-+            }
-+            lacp_port->lacp->next_retry_count_change_time = monotonic_time.tv_sec + LACP_NEXT_RETRY_COUNT_CHANGE_TIMEOUT_SECONDS;
-+        }
  	} else {
  		if (lacp_port->partner_retry_count != LACP_CFG_DFLT_RETRY_COUNT) {
 -			teamd_log_dbg(lacp_port->ctx, "%s: retry count from partner changed from %u to %u",
@@ -80,3 +103,18 @@ index 4b77692..5a55bd5 100644
  		}
  	}
 
++	lacp_port->last_received_lacpdu_version = lacpdu->version_number;
++
+ 	err = lacp_port_set_state(lacp_port, PORT_STATE_CURRENT);
+ 	if (err)
+ 		return err;
+@@ -1512,8 +1539,7 @@ static int lacpdu_process(struct lacp_port *lacp_port, struct lacpdu* lacpdu)
+
+ 	/* Check if the other side has correct info about us */
+ 	if (memcmp(&lacpdu->partner, &lacp_port->actor, sizeof(struct lacpdu_info))
+-            || (lacpdu->version_number == 0xf1 && lacp_port->lacp->cfg.retry_count != lacpdu->v2.partner_retry_count)
+-            || (lacpdu->version_number != 0xf1 && lacp_port->lacp->cfg.retry_count != LACP_CFG_DFLT_RETRY_COUNT)) {
++            || (lacpdu->version_number == 0xf1 && lacp_port->lacp->cfg.retry_count != lacpdu->v2.partner_retry_count)) {
+ 		err = lacpdu_send(lacp_port);
+ 		if (err)
+ 			return err;