[CCID 3]: Larger initial windows
This implements the larger-initial-windows feature for CCID 3, as described
in section 5 of RFC 4342. When the first feedback packet arrives, the sender
can send up to 2..4 packets per RTT, instead of just one.
The patch further
* reduces the number of timestamping calls by passing the timestamp value
(which is computed in one of the calling functions anyway) as argument
* renames one constant with a very long name into one which is shorter and
resembles the one in RFC 3448 (t_mbi)
* simplifies some of the min_t/max_t cases where both `x', `y' have the same
type
Signed-off-by: Gerrit Renker <gerrit@xxxxxxxxxxxxxx>
---
net/dccp/ccids/ccid3.c | 75 +++++++++++++++++++++++++++----------------------
net/dccp/ccids/ccid3.h | 4 +-
2 files changed, 44 insertions(+), 35 deletions(-)
--- a/net/dccp/ccids/ccid3.c
+++ b/net/dccp/ccids/ccid3.c
@@ -129,35 +129,42 @@ static inline void ccid3_update_send_tim
* If X has changed, we also update the scheduled send time t_now,
* the inter-packet interval t_ipi, and the delta value.
*/
-static void ccid3_hc_tx_update_x(struct sock *sk)
+static void ccid3_hc_tx_update_x(struct sock *sk, struct timeval *now)
+
{
struct ccid3_hc_tx_sock *hctx = ccid3_hc_tx_sk(sk);
const __u32 old_x = hctx->ccid3hctx_x;
DCCP_BUG_ON(hctx->ccid3hctx_rtt == 0);
- /* To avoid large error in calcX */
- if (hctx->ccid3hctx_p >= TFRC_SMALLEST_P) {
- hctx->ccid3hctx_x_calc = tfrc_calc_x(hctx->ccid3hctx_s,
- hctx->ccid3hctx_rtt,
- hctx->ccid3hctx_p);
- hctx->ccid3hctx_x = max_t(u32, min_t(u32, hctx->ccid3hctx_x_calc,
- 2 * hctx->ccid3hctx_x_recv),
- (hctx->ccid3hctx_s /
- TFRC_MAX_BACK_OFF_TIME));
- } else {
- struct timeval now;
+ /* Compute X after the first feedback packet has arrived */
+ if (hctx->ccid3hctx_state == TFRC_SSTATE_NO_FBACK) {
+ /* This implements `Larger Initial Windows' [RFC 4342, sec. 5]
+ * We deviate in that we use `s' instead of `MSS'. */
+ u16 w_init = max( 4 * hctx->ccid3hctx_s,
+ max( 2 * hctx->ccid3hctx_s, 4380));
+ hctx->ccid3hctx_x = usecs_div(w_init, hctx->ccid3hctx_rtt);
+ hctx->ccid3hctx_t_ld = *now;
+
+ /* To avoid large error in calcX */
+ } else if (hctx->ccid3hctx_p >= TFRC_SMALLEST_P) {
+ hctx->ccid3hctx_x_calc = tfrc_calc_x(hctx->ccid3hctx_s,
+ hctx->ccid3hctx_rtt,
+ hctx->ccid3hctx_p);
+ hctx->ccid3hctx_x = max_t(u32, min(hctx->ccid3hctx_x_calc,
+ hctx->ccid3hctx_x_recv * 2),
+ hctx->ccid3hctx_s / TFRC_t_mbi );
+
+ } else if (timeval_delta(now, &hctx->ccid3hctx_t_ld) >=
+ hctx->ccid3hctx_rtt) {
+ hctx->ccid3hctx_x = max(min(hctx->ccid3hctx_x_recv,
+ hctx->ccid3hctx_x ) * 2,
+ usecs_div(hctx->ccid3hctx_s,
+ hctx->ccid3hctx_rtt) );
+ hctx->ccid3hctx_t_ld = *now;
+ } else
+ ccid3_pr_debug("Not changing X\n");
- dccp_timestamp(sk, &now);
- if (timeval_delta(&now, &hctx->ccid3hctx_t_ld) >=
- hctx->ccid3hctx_rtt) {
- hctx->ccid3hctx_x = max_t(u32, min_t(u32, hctx->ccid3hctx_x_recv,
- hctx->ccid3hctx_x) * 2,
- usecs_div(hctx->ccid3hctx_s,
- hctx->ccid3hctx_rtt));
- hctx->ccid3hctx_t_ld = now;
- }
- }
if(hctx->ccid3hctx_x != old_x)
ccid3_update_send_time(hctx);
}
@@ -185,6 +192,7 @@ static void ccid3_hc_tx_no_feedback_time
struct sock *sk = (struct sock *)data;
struct ccid3_hc_tx_sock *hctx = ccid3_hc_tx_sk(sk);
unsigned long next_tmout = USEC_PER_SEC / 5;
+ struct timeval now;
bh_lock_sock(sk);
if (sock_owned_by_user(sk)) {
@@ -200,10 +208,8 @@ static void ccid3_hc_tx_no_feedback_time
case TFRC_SSTATE_NO_FBACK:
/* Halve send rate */
hctx->ccid3hctx_x /= 2;
- if (hctx->ccid3hctx_x < (hctx->ccid3hctx_s /
- TFRC_MAX_BACK_OFF_TIME))
- hctx->ccid3hctx_x = (hctx->ccid3hctx_s /
- TFRC_MAX_BACK_OFF_TIME);
+ if (hctx->ccid3hctx_x < (hctx->ccid3hctx_s / TFRC_t_mbi))
+ hctx->ccid3hctx_x = hctx->ccid3hctx_s / TFRC_t_mbi;
ccid3_pr_debug("%s, sk=%p, state=%s, updated tx rate to %d "
"bytes/s\n",
@@ -240,12 +246,13 @@ static void ccid3_hc_tx_no_feedback_time
if (hctx->ccid3hctx_p < TFRC_SMALLEST_P ||
hctx->ccid3hctx_x_calc > 2 * hctx->ccid3hctx_x_recv)
hctx->ccid3hctx_x_recv = max_t(u32, hctx->ccid3hctx_x_recv / 2,
- hctx->ccid3hctx_s / (2 * TFRC_MAX_BACK_OFF_TIME));
+ hctx->ccid3hctx_s / (2 * TFRC_t_mbi));
else
hctx->ccid3hctx_x_recv = hctx->ccid3hctx_x_calc / 4;
/* Update sending rate */
- ccid3_hc_tx_update_x(sk);
+ dccp_timestamp(sk, &now);
+ ccid3_hc_tx_update_x(sk, &now);
}
/*
* Schedule no feedback timer to expire in
@@ -488,10 +495,9 @@ static void ccid3_hc_tx_packet_recv(stru
*
* q is a constant, RFC 3448 recomments 0.9
*/
- if (hctx->ccid3hctx_state == TFRC_SSTATE_NO_FBACK) {
- ccid3_hc_tx_set_state(sk, TFRC_SSTATE_FBACK);
+ if (hctx->ccid3hctx_state == TFRC_SSTATE_NO_FBACK)
hctx->ccid3hctx_rtt = r_sample;
- } else
+ else
hctx->ccid3hctx_rtt = (hctx->ccid3hctx_rtt * 9) / 10 +
r_sample / 10;
@@ -508,7 +514,11 @@ static void ccid3_hc_tx_packet_recv(stru
sk_stop_timer(sk, &hctx->ccid3hctx_no_feedback_timer);
/* Update sending rate (and likely t_ipi, t_nom, and delta) */
- ccid3_hc_tx_update_x(sk);
+ ccid3_hc_tx_update_x(sk, &now);
+
+ /* Update the state if necessary */
+ if (hctx->ccid3hctx_state == TFRC_SSTATE_NO_FBACK)
+ ccid3_hc_tx_set_state(sk, TFRC_SSTATE_FBACK);
/* remove all packets older than the one acked from history */
dccp_tx_hist_purge_older(ccid3_tx_hist,
@@ -519,7 +529,6 @@ static void ccid3_hc_tx_packet_recv(stru
*/
sk->sk_write_space(sk);
-
/* Update timeout interval. We use the alternative variant of
* [RFC 3448, 3.1] which sets the upper bound of t_rto to one
* second, as it is suggested for TCP (see RFC 2988, 2.4). */
--- a/net/dccp/ccids/ccid3.h
+++ b/net/dccp/ccids/ccid3.h
@@ -48,8 +48,8 @@
/* In usecs - half the scheduling granularity as per RFC3448 4.6 */
#define TFRC_OPSYS_HALF_TIME_GRAN (USEC_PER_SEC / (2 * HZ))
-/* In seconds */
-#define TFRC_MAX_BACK_OFF_TIME 64
+/* Parameter t_mbi from [RFC 3448, 4.3]: In seconds */
+#define TFRC_t_mbi 64
#define TFRC_SMALLEST_P 40
-
To unsubscribe from this list: send the line "unsubscribe dccp" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
