ike: Use optional jitter to calculate retransmission timeouts

Also adds an optional limit to avoid very high retransmission timeouts
with high numbers of retries.

conf/options/charon.opt

@@ -311,6 +311,13 @@ charon.retransmit_timeout = 4.0
 charon.retransmit_tries = 5
 	Number of times to retransmit a packet before giving up.
 
+charon.retransmit_jitter = 0
+	Maximum jitter in percent to apply randomly to calculated retransmission
+	timeout (0 to disable).
+
+charon.retransmit_limit = 0
+	Upper limit in seconds for calculated retransmission timeout (0 to disable).
+
 charon.retry_initiate_interval = 0
 	Interval in seconds to use when retrying to initiate an IKE_SA (e.g. if DNS
 	resolution failed), 0 to disable retries.

conf/strongswan.conf.5.tail.in

@@ -408,6 +408,8 @@ using the three keys listed below:
 .BR charon.retransmit_base " [1.8]"
 .BR charon.retransmit_timeout " [4.0]"
 .BR charon.retransmit_tries " [5]"
+.BR charon.retransmit_jitter " [0]"
+.BR charon.retransmit_limit " [0]"
 .fi
 .RE
 .PP
@@ -419,7 +421,15 @@ The following algorithm is used to calculate the timeout:
 .PP
 Where
 .I n
-is the current retransmission count.
+is the current retransmission count. The calculated timeout can't exceed the
+configured retransmit_limit (if any), which is useful if the number of retries
+is high.
+.PP
+If a jitter in percent is configured, the timeout is modified as follows:
+.PP
+.EX
+	relative timeout -= random(0, retransmit_jitter * relative timeout)
+.EE
 .PP
 Using the default values, packets are retransmitted in:
 

src/libcharon/sa/ikev1/task_manager_v1.c

@@ -209,6 +209,16 @@ struct private_task_manager_t {
 	 */
 	double retransmit_base;
 
+	/**
+	 * Jitter to apply to calculated retransmit timeout (in percent)
+	 */
+	u_int retransmit_jitter;
+
+	/**
+	 * Limit retransmit timeout to this value
+	 */
+	uint32_t retransmit_limit;
+
 	/**
 	 * Sequence number for sending DPD requests
 	 */
@@ -345,7 +355,7 @@ static status_t retransmit_packet(private_task_manager_t *this, uint32_t seqnr,
 							u_int mid, u_int retransmitted, array_t *packets)
 {
 	packet_t *packet;
-	uint32_t t;
+	uint32_t t, max_jitter;
 
 	array_get(packets, 0, &packet);
 	if (retransmitted > this->retransmit_tries)
@@ -356,6 +366,15 @@ static status_t retransmit_packet(private_task_manager_t *this, uint32_t seqnr,
 	}
 	t = (uint32_t)(this->retransmit_timeout * 1000.0 *
 					pow(this->retransmit_base, retransmitted));
+	if (this->retransmit_jitter)
+	{
+		max_jitter = (t / 100.0) * this->retransmit_jitter;
+		t -= max_jitter * (random() / (RAND_MAX + 1.0));
+	}
+	if (this->retransmit_limit)
+	{
+		t = min(t, this->retransmit_limit);
+	}
 	if (retransmitted)
 	{
 		DBG1(DBG_IKE, "sending retransmit %u of %s message ID %u, seq %u",
@@ -2034,11 +2053,15 @@ task_manager_v1_t *task_manager_v1_create(ike_sa_t *ike_sa)
 		.active_tasks = linked_list_create(),
 		.passive_tasks = linked_list_create(),
 		.retransmit_tries = lib->settings->get_int(lib->settings,
-						"%s.retransmit_tries", RETRANSMIT_TRIES, lib->ns),
+					"%s.retransmit_tries", RETRANSMIT_TRIES, lib->ns),
 		.retransmit_timeout = lib->settings->get_double(lib->settings,
-						"%s.retransmit_timeout", RETRANSMIT_TIMEOUT, lib->ns),
+					"%s.retransmit_timeout", RETRANSMIT_TIMEOUT, lib->ns),
 		.retransmit_base = lib->settings->get_double(lib->settings,
-						"%s.retransmit_base", RETRANSMIT_BASE, lib->ns),
+					"%s.retransmit_base", RETRANSMIT_BASE, lib->ns),
+		.retransmit_jitter = min(lib->settings->get_int(lib->settings,
+					"%s.retransmit_jitter", 0, lib->ns), RETRANSMIT_JITTER_MAX),
+		.retransmit_limit = lib->settings->get_int(lib->settings,
+					"%s.retransmit_limit", 0, lib->ns) * 1000,
 	);
 
 	if (!this->rng)

src/libcharon/sa/ikev2/task_manager_v2.c

@@ -160,6 +160,16 @@ struct private_task_manager_t {
 	 */
 	double retransmit_base;
 
+	/**
+	 * Jitter to apply to calculated retransmit timeout (in percent)
+	 */
+	u_int retransmit_jitter;
+
+	/**
+	 * Limit retransmit timeout to this value
+	 */
+	uint32_t retransmit_limit;
+
 	/**
 	 * Use make-before-break instead of break-before-make reauth?
 	 */
@@ -321,7 +331,7 @@ METHOD(task_manager_t, retransmit, status_t,
 	if (message_id == this->initiating.mid &&
 		array_count(this->initiating.packets))
 	{
-		uint32_t timeout;
+		uint32_t timeout, max_jitter;
 		job_t *job;
 		enumerator_t *enumerator;
 		packet_t *packet;
@@ -351,6 +361,16 @@ METHOD(task_manager_t, retransmit, status_t,
 			{
 				timeout = (uint32_t)(this->retransmit_timeout * 1000.0 *
 					pow(this->retransmit_base, this->initiating.retransmitted));
+
+				if (this->retransmit_jitter)
+				{
+					max_jitter = (timeout / 100.0) * this->retransmit_jitter;
+					timeout -= max_jitter * (random() / (RAND_MAX + 1.0));
+				}
+				if (this->retransmit_limit)
+				{
+					timeout = min(timeout, this->retransmit_limit);
+				}
 			}
 			else
 			{
@@ -2151,6 +2171,10 @@ task_manager_v2_t *task_manager_v2_create(ike_sa_t *ike_sa)
 					"%s.retransmit_timeout", RETRANSMIT_TIMEOUT, lib->ns),
 		.retransmit_base = lib->settings->get_double(lib->settings,
 					"%s.retransmit_base", RETRANSMIT_BASE, lib->ns),
+		.retransmit_jitter = min(lib->settings->get_int(lib->settings,
+					"%s.retransmit_jitter", 0, lib->ns), RETRANSMIT_JITTER_MAX),
+		.retransmit_limit = lib->settings->get_int(lib->settings,
+					"%s.retransmit_limit", 0, lib->ns) * 1000,
 		.make_before_break = lib->settings->get_bool(lib->settings,
 					"%s.make_before_break", FALSE, lib->ns),
 	);

src/libcharon/sa/task_manager.h

@@ -47,6 +47,11 @@ typedef enum task_queue_t task_queue_t;
  */
 #define RETRANSMIT_TRIES 5
 
+/**
+ * Maximum jitter in percent.
+ */
+#define RETRANSMIT_JITTER_MAX 20
+
 /**
  * Interval for mobike routability checks in ms.
  */