This updates the code with regard to handling idle and application-limited
periods as specified in [RFC 4342, 5.1].
Background:
----------
The current code does not conform with this, it implements TFRC as per RFC
3448, 4.4:
"If the sender has been idle since this nofeedback timer has been set and
X_recv < 4 * s/R then X_recv should not be halved in response to timer
expiration."
However, what is required from the code is that "the allowed sending rate is
never reduced less than the [RFC3390] initial sending rate as the result of
an idle period." [RFC 4342, 5.1].
Implementation:
---------------
The CCID 3 RFC leaves implementation choices open as to implement the above; I
have chosen the mechanism described in draft-ietf-dccp-rfc3448bis, since it
achieves exactly the described objective, and is likely to become standard in
some future time.
Since the expected normal behaviour is a non-idle connection (where ACKs
continually come in), the test whether the connection is idle has been wrapped
with the `unlikely' macro to favour this case.
Additional Notes.
-----------------
As a minor optimisation, the setting of the `idle' flag was shifted from
tx_packet_sent to tx_send_packet, as this happens slightly earlier.
A further but small optimisation would be possible by not updating the cached
value of X_recv when it is less than half the RFC 3390 initial sending rate.
This however will bring only marginal gain in exchange for more complicated
code, and is not compliant with 3448bis. Thus omitted.
Signed-off-by: Gerrit Renker <[EMAIL PROTECTED]>
Signed-off-by: Ian McDonald <[EMAIL PROTECTED]>
Signed-off-by: Arnaldo Carvalho de Melo <[EMAIL PROTECTED]>
---
net/dccp/ccids/ccid3.c | 84 +++++++++++++++++++++++-------------------------
1 files changed, 40 insertions(+), 44 deletions(-)
diff --git a/net/dccp/ccids/ccid3.c b/net/dccp/ccids/ccid3.c
index 8d33a09..35123c1 100644
--- a/net/dccp/ccids/ccid3.c
+++ b/net/dccp/ccids/ccid3.c
@@ -133,12 +133,23 @@ static void ccid3_hc_tx_update_x(struct sock *sk, struct
timeval *now)
{
struct ccid3_hc_tx_sock *hctx = ccid3_hc_tx_sk(sk);
+ __u64 min_rate = 2 * hctx->ccid3hctx_x_recv;
const __u64 old_x = hctx->ccid3hctx_x;
+ /*
+ * Handle IDLE periods: do not reduce below RFC3390 initial sending rate
+ * when idling [RFC 4342, 5.1]. See also draft-ietf-dccp-rfc3448bis.
+ * For consistency with X and X_recv, min_rate is also scaled by 2^6.
+ */
+ if (unlikely(hctx->ccid3hctx_idle)) {
+ min_rate = rfc3390_initial_rate(sk);
+ min_rate = max(min_rate, 2 * hctx->ccid3hctx_x_recv);
+ }
+
if (hctx->ccid3hctx_p > 0) {
hctx->ccid3hctx_x = min(((__u64)hctx->ccid3hctx_x_calc) << 6,
- hctx->ccid3hctx_x_recv * 2);
+ min_rate);
hctx->ccid3hctx_x = max(hctx->ccid3hctx_x,
(((__u64)hctx->ccid3hctx_s) << 6) /
TFRC_T_MBI);
@@ -147,7 +158,7 @@ static void ccid3_hc_tx_update_x(struct sock *sk, struct
timeval *now)
(suseconds_t)hctx->ccid3hctx_rtt >= 0) {
hctx->ccid3hctx_x =
- max(2 * min(hctx->ccid3hctx_x, hctx->ccid3hctx_x_recv),
+ max(min(2 * hctx->ccid3hctx_x, min_rate),
scaled_div(((__u64)hctx->ccid3hctx_s) << 6,
hctx->ccid3hctx_rtt));
hctx->ccid3hctx_t_ld = *now;
@@ -209,6 +220,7 @@ static void ccid3_hc_tx_no_feedback_timer(unsigned long
data)
{
struct sock *sk = (struct sock *)data;
struct ccid3_hc_tx_sock *hctx = ccid3_hc_tx_sk(sk);
+ struct timeval now;
unsigned long t_nfb = USEC_PER_SEC / 5;
bh_lock_sock(sk);
@@ -221,6 +233,8 @@ static void ccid3_hc_tx_no_feedback_timer(unsigned long
data)
ccid3_pr_debug("%s(%p, state=%s) - entry \n", dccp_role(sk), sk,
ccid3_tx_state_name(hctx->ccid3hctx_state));
+ hctx->ccid3hctx_idle = 1;
+
switch (hctx->ccid3hctx_state) {
case TFRC_SSTATE_NO_FBACK:
/* RFC 3448, 4.4: Halve send rate directly */
@@ -239,49 +253,33 @@ static void ccid3_hc_tx_no_feedback_timer(unsigned long
data)
break;
case TFRC_SSTATE_FBACK:
/*
- * Check if IDLE since last timeout and recv rate is less than
- * 4 packets (in units of 64*bytes/sec) per RTT
+ * Modify the cached value of X_recv [RFC 3448, 4.4]
+ *
+ * If (p == 0 || X_calc > 2 * X_recv)
+ * X_recv = max(X_recv / 2, s / (2 * t_mbi));
+ * Else
+ * X_recv = X_calc / 4;
+ *
+ * Note that X_recv is scaled by 2^6 while X_calc is not
*/
- if (!hctx->ccid3hctx_idle ||
- (hctx->ccid3hctx_x_recv >= 4 *
- scaled_div(((__u64)hctx->ccid3hctx_s) << 6,
- hctx->ccid3hctx_rtt))) {
- struct timeval now;
+ BUG_ON(hctx->ccid3hctx_p && !hctx->ccid3hctx_x_calc);
- ccid3_pr_debug("%s(%p, state=%s), not idle\n",
- dccp_role(sk), sk,
- ccid3_tx_state_name(hctx->ccid3hctx_state));
+ if (hctx->ccid3hctx_p == 0 ||
+ (hctx->ccid3hctx_x_calc > (hctx->ccid3hctx_x_recv >> 5))) {
- /*
- * Modify the cached value of X_recv [RFC 3448, 4.4]
- *
- * If (p == 0 || X_calc > 2 * X_recv)
- * X_recv = max(X_recv / 2, s / (2 * t_mbi));
- * Else
- * X_recv = X_calc / 4;
- *
- * Note that X_recv is scaled by 2^6 while X_calc is
not
- */
- BUG_ON(hctx->ccid3hctx_p && !hctx->ccid3hctx_x_calc);
-
- if (hctx->ccid3hctx_p == 0 ||
- (hctx->ccid3hctx_x_calc >
- (hctx->ccid3hctx_x_recv >> 5))) {
-
- hctx->ccid3hctx_x_recv =
- max(hctx->ccid3hctx_x_recv / 2,
- (((__u64)hctx->ccid3hctx_s) << 6) /
- (2 * TFRC_T_MBI));
-
- if (hctx->ccid3hctx_p == 0)
- dccp_timestamp(sk, &now);
- } else {
- hctx->ccid3hctx_x_recv = hctx->ccid3hctx_x_calc;
- hctx->ccid3hctx_x_recv <<= 4;
- }
- /* Now recalculate X [RFC 3448, 4.3, step (4)] */
- ccid3_hc_tx_update_x(sk, &now);
+ hctx->ccid3hctx_x_recv =
+ max(hctx->ccid3hctx_x_recv / 2,
+ (((__u64)hctx->ccid3hctx_s) << 6) /
+ (2 * TFRC_T_MBI));
+
+ if (hctx->ccid3hctx_p == 0)
+ dccp_timestamp(sk, &now);
+ } else {
+ hctx->ccid3hctx_x_recv = hctx->ccid3hctx_x_calc;
+ hctx->ccid3hctx_x_recv <<= 4;
}
+ /* Now recalculate X [RFC 3448, 4.3, step (4)] */
+ ccid3_hc_tx_update_x(sk, &now);
/*
* Schedule no feedback timer to expire in
* max(t_RTO, 2 * s/X) = max(t_RTO, 2 * t_ipi)
@@ -296,8 +294,6 @@ static void ccid3_hc_tx_no_feedback_timer(unsigned long
data)
goto out;
}
- hctx->ccid3hctx_idle = 1;
-
restart_timer:
sk_reset_timer(sk, &hctx->ccid3hctx_no_feedback_timer,
jiffies + usecs_to_jiffies(t_nfb));
@@ -377,6 +373,7 @@ static int ccid3_hc_tx_send_packet(struct sock *sk, struct
sk_buff *skb)
/* prepare to send now (add options etc.) */
dp->dccps_hc_tx_insert_options = 1;
DCCP_SKB_CB(skb)->dccpd_ccval = hctx->ccid3hctx_last_win_count;
+ hctx->ccid3hctx_idle = 0;
/* set the nominal send time for the next following packet */
timeval_add_usecs(&hctx->ccid3hctx_t_nom, hctx->ccid3hctx_t_ipi);
@@ -407,7 +404,6 @@ static void ccid3_hc_tx_packet_sent(struct sock *sk, int
more,
packet->dccphtx_seqno = dccp_sk(sk)->dccps_gss;
packet->dccphtx_rtt = hctx->ccid3hctx_rtt;
packet->dccphtx_sent = 1;
- hctx->ccid3hctx_idle = 0;
}
static void ccid3_hc_tx_packet_recv(struct sock *sk, struct sk_buff *skb)
--
1.5.0.3
-
To unsubscribe from this list: send the line "unsubscribe netdev" in
the body of a message to [EMAIL PROTECTED]
More majordomo info at http://vger.kernel.org/majordomo-info.html
