summaryrefslogtreecommitdiffstats
path: root/net/dccp/ccids
diff options
context:
space:
mode:
Diffstat (limited to 'net/dccp/ccids')
-rw-r--r--net/dccp/ccids/Kconfig30
-rw-r--r--net/dccp/ccids/ccid2.c622
-rw-r--r--net/dccp/ccids/ccid2.h63
-rw-r--r--net/dccp/ccids/ccid3.c762
-rw-r--r--net/dccp/ccids/ccid3.h153
-rw-r--r--net/dccp/ccids/lib/loss_interval.c30
-rw-r--r--net/dccp/ccids/lib/loss_interval.h4
-rw-r--r--net/dccp/ccids/lib/packet_history.c282
-rw-r--r--net/dccp/ccids/lib/packet_history.h78
-rw-r--r--net/dccp/ccids/lib/tfrc.h16
-rw-r--r--net/dccp/ccids/lib/tfrc_equation.c29
11 files changed, 953 insertions, 1116 deletions
diff --git a/net/dccp/ccids/Kconfig b/net/dccp/ccids/Kconfig
index 12275943eab..fb168be2cb4 100644
--- a/net/dccp/ccids/Kconfig
+++ b/net/dccp/ccids/Kconfig
@@ -1,10 +1,8 @@
menu "DCCP CCIDs Configuration (EXPERIMENTAL)"
- depends on EXPERIMENTAL
config IP_DCCP_CCID2
- tristate "CCID2 (TCP-Like) (EXPERIMENTAL)"
+ tristate "CCID2 (TCP-Like)"
def_tristate IP_DCCP
- select IP_DCCP_ACKVEC
---help---
CCID 2, TCP-like Congestion Control, denotes Additive Increase,
Multiplicative Decrease (AIMD) congestion control with behavior
@@ -36,7 +34,7 @@ config IP_DCCP_CCID2_DEBUG
If in doubt, say N.
config IP_DCCP_CCID3
- tristate "CCID3 (TCP-Friendly) (EXPERIMENTAL)"
+ tristate "CCID3 (TCP-Friendly)"
def_tristate IP_DCCP
select IP_DCCP_TFRC_LIB
---help---
@@ -64,9 +62,9 @@ config IP_DCCP_CCID3
If in doubt, say M.
+if IP_DCCP_CCID3
config IP_DCCP_CCID3_DEBUG
bool "CCID3 debugging messages"
- depends on IP_DCCP_CCID3
---help---
Enable CCID3-specific debugging messages.
@@ -76,10 +74,29 @@ config IP_DCCP_CCID3_DEBUG
If in doubt, say N.
+choice
+ prompt "Select method for measuring the packet size s"
+ default IP_DCCP_CCID3_MEASURE_S_AS_MPS
+
+config IP_DCCP_CCID3_MEASURE_S_AS_MPS
+ bool "Always use MPS in place of s"
+ ---help---
+ This use is recommended as it is consistent with the initialisation
+ of X and suggested when s varies (rfc3448bis, (1) in section 4.1).
+config IP_DCCP_CCID3_MEASURE_S_AS_AVG
+ bool "Use moving average"
+ ---help---
+ An alternative way of tracking s, also supported by rfc3448bis.
+ This used to be the default for CCID-3 in previous kernels.
+config IP_DCCP_CCID3_MEASURE_S_AS_MAX
+ bool "Track the maximum payload length"
+ ---help---
+ An experimental method based on tracking the maximum packet size.
+endchoice
+
config IP_DCCP_CCID3_RTO
int "Use higher bound for nofeedback timer"
default 100
- depends on IP_DCCP_CCID3 && EXPERIMENTAL
---help---
Use higher lower bound for nofeedback timer expiration.
@@ -106,6 +123,7 @@ config IP_DCCP_CCID3_RTO
The purpose of the nofeedback timer is to slow DCCP down when there
is serious network congestion: experimenting with larger values should
therefore not be performed on WANs.
+endif # IP_DCCP_CCID3
config IP_DCCP_TFRC_LIB
tristate
diff --git a/net/dccp/ccids/ccid2.c b/net/dccp/ccids/ccid2.c
index 9a430734530..fa713227c66 100644
--- a/net/dccp/ccids/ccid2.c
+++ b/net/dccp/ccids/ccid2.c
@@ -25,7 +25,7 @@
/*
* This implementation should follow RFC 4341
*/
-
+#include "../feat.h"
#include "../ccid.h"
#include "../dccp.h"
#include "ccid2.h"
@@ -34,51 +34,8 @@
#ifdef CONFIG_IP_DCCP_CCID2_DEBUG
static int ccid2_debug;
#define ccid2_pr_debug(format, a...) DCCP_PR_DEBUG(ccid2_debug, format, ##a)
-
-static void ccid2_hc_tx_check_sanity(const struct ccid2_hc_tx_sock *hctx)
-{
- int len = 0;
- int pipe = 0;
- struct ccid2_seq *seqp = hctx->ccid2hctx_seqh;
-
- /* there is data in the chain */
- if (seqp != hctx->ccid2hctx_seqt) {
- seqp = seqp->ccid2s_prev;
- len++;
- if (!seqp->ccid2s_acked)
- pipe++;
-
- while (seqp != hctx->ccid2hctx_seqt) {
- struct ccid2_seq *prev = seqp->ccid2s_prev;
-
- len++;
- if (!prev->ccid2s_acked)
- pipe++;
-
- /* packets are sent sequentially */
- BUG_ON(dccp_delta_seqno(seqp->ccid2s_seq,
- prev->ccid2s_seq ) >= 0);
- BUG_ON(time_before(seqp->ccid2s_sent,
- prev->ccid2s_sent));
-
- seqp = prev;
- }
- }
-
- BUG_ON(pipe != hctx->ccid2hctx_pipe);
- ccid2_pr_debug("len of chain=%d\n", len);
-
- do {
- seqp = seqp->ccid2s_prev;
- len++;
- } while (seqp != hctx->ccid2hctx_seqh);
-
- ccid2_pr_debug("total len=%d\n", len);
- BUG_ON(len != hctx->ccid2hctx_seqbufc * CCID2_SEQBUF_LEN);
-}
#else
#define ccid2_pr_debug(format, a...)
-#define ccid2_hc_tx_check_sanity(hctx)
#endif
static int ccid2_hc_tx_alloc_seq(struct ccid2_hc_tx_sock *hctx)
@@ -87,8 +44,7 @@ static int ccid2_hc_tx_alloc_seq(struct ccid2_hc_tx_sock *hctx)
int i;
/* check if we have space to preserve the pointer to the buffer */
- if (hctx->ccid2hctx_seqbufc >= (sizeof(hctx->ccid2hctx_seqbuf) /
- sizeof(struct ccid2_seq*)))
+ if (hctx->seqbufc >= sizeof(hctx->seqbuf) / sizeof(struct ccid2_seq *))
return -ENOMEM;
/* allocate buffer and initialize linked list */
@@ -104,38 +60,35 @@ static int ccid2_hc_tx_alloc_seq(struct ccid2_hc_tx_sock *hctx)
seqp->ccid2s_prev = &seqp[CCID2_SEQBUF_LEN - 1];
/* This is the first allocation. Initiate the head and tail. */
- if (hctx->ccid2hctx_seqbufc == 0)
- hctx->ccid2hctx_seqh = hctx->ccid2hctx_seqt = seqp;
+ if (hctx->seqbufc == 0)
+ hctx->seqh = hctx->seqt = seqp;
else {
/* link the existing list with the one we just created */
- hctx->ccid2hctx_seqh->ccid2s_next = seqp;
- seqp->ccid2s_prev = hctx->ccid2hctx_seqh;
+ hctx->seqh->ccid2s_next = seqp;
+ seqp->ccid2s_prev = hctx->seqh;
- hctx->ccid2hctx_seqt->ccid2s_prev = &seqp[CCID2_SEQBUF_LEN - 1];
- seqp[CCID2_SEQBUF_LEN - 1].ccid2s_next = hctx->ccid2hctx_seqt;
+ hctx->seqt->ccid2s_prev = &seqp[CCID2_SEQBUF_LEN - 1];
+ seqp[CCID2_SEQBUF_LEN - 1].ccid2s_next = hctx->seqt;
}
/* store the original pointer to the buffer so we can free it */
- hctx->ccid2hctx_seqbuf[hctx->ccid2hctx_seqbufc] = seqp;
- hctx->ccid2hctx_seqbufc++;
+ hctx->seqbuf[hctx->seqbufc] = seqp;
+ hctx->seqbufc++;
return 0;
}
static int ccid2_hc_tx_send_packet(struct sock *sk, struct sk_buff *skb)
{
- struct ccid2_hc_tx_sock *hctx = ccid2_hc_tx_sk(sk);
-
- if (hctx->ccid2hctx_pipe < hctx->ccid2hctx_cwnd)
- return 0;
-
- return 1; /* XXX CCID should dequeue when ready instead of polling */
+ if (ccid2_cwnd_network_limited(ccid2_hc_tx_sk(sk)))
+ return CCID_PACKET_WILL_DEQUEUE_LATER;
+ return CCID_PACKET_SEND_AT_ONCE;
}
static void ccid2_change_l_ack_ratio(struct sock *sk, u32 val)
{
struct dccp_sock *dp = dccp_sk(sk);
- u32 max_ratio = DIV_ROUND_UP(ccid2_hc_tx_sk(sk)->ccid2hctx_cwnd, 2);
+ u32 max_ratio = DIV_ROUND_UP(ccid2_hc_tx_sk(sk)->cwnd, 2);
/*
* Ensure that Ack Ratio does not exceed ceil(cwnd/2), which is (2) from
@@ -147,8 +100,8 @@ static void ccid2_change_l_ack_ratio(struct sock *sk, u32 val)
DCCP_WARN("Limiting Ack Ratio (%u) to %u\n", val, max_ratio);
val = max_ratio;
}
- if (val > 0xFFFF) /* RFC 4340, 11.3 */
- val = 0xFFFF;
+ if (val > DCCPF_ACK_RATIO_MAX)
+ val = DCCPF_ACK_RATIO_MAX;
if (val == dp->dccps_l_ack_ratio)
return;
@@ -157,99 +110,77 @@ static void ccid2_change_l_ack_ratio(struct sock *sk, u32 val)
dp->dccps_l_ack_ratio = val;
}
-static void ccid2_change_srtt(struct ccid2_hc_tx_sock *hctx, long val)
-{
- ccid2_pr_debug("change SRTT to %ld\n", val);
- hctx->ccid2hctx_srtt = val;
-}
-
-static void ccid2_start_rto_timer(struct sock *sk);
-
static void ccid2_hc_tx_rto_expire(unsigned long data)
{
struct sock *sk = (struct sock *)data;
struct ccid2_hc_tx_sock *hctx = ccid2_hc_tx_sk(sk);
- long s;
+ const bool sender_was_blocked = ccid2_cwnd_network_limited(hctx);
bh_lock_sock(sk);
if (sock_owned_by_user(sk)) {
- sk_reset_timer(sk, &hctx->ccid2hctx_rtotimer,
- jiffies + HZ / 5);
+ sk_reset_timer(sk, &hctx->rtotimer, jiffies + HZ / 5);
goto out;
}
ccid2_pr_debug("RTO_EXPIRE\n");
- ccid2_hc_tx_check_sanity(hctx);
-
/* back-off timer */
- hctx->ccid2hctx_rto <<= 1;
-
- s = hctx->ccid2hctx_rto / HZ;
- if (s > 60)
- hctx->ccid2hctx_rto = 60 * HZ;
-
- ccid2_start_rto_timer(sk);
+ hctx->rto <<= 1;
+ if (hctx->rto > DCCP_RTO_MAX)
+ hctx->rto = DCCP_RTO_MAX;
/* adjust pipe, cwnd etc */
- hctx->ccid2hctx_ssthresh = hctx->ccid2hctx_cwnd / 2;
- if (hctx->ccid2hctx_ssthresh < 2)
- hctx->ccid2hctx_ssthresh = 2;
- hctx->ccid2hctx_cwnd = 1;
- hctx->ccid2hctx_pipe = 0;
+ hctx->ssthresh = hctx->cwnd / 2;
+ if (hctx->ssthresh < 2)
+ hctx->ssthresh = 2;
+ hctx->cwnd = 1;
+ hctx->pipe = 0;
/* clear state about stuff we sent */
- hctx->ccid2hctx_seqt = hctx->ccid2hctx_seqh;
- hctx->ccid2hctx_packets_acked = 0;
+ hctx->seqt = hctx->seqh;
+ hctx->packets_acked = 0;
/* clear ack ratio state. */
- hctx->ccid2hctx_rpseq = 0;
- hctx->ccid2hctx_rpdupack = -1;
+ hctx->rpseq = 0;
+ hctx->rpdupack = -1;
ccid2_change_l_ack_ratio(sk, 1);
- ccid2_hc_tx_check_sanity(hctx);
+
+ /* if we were blocked before, we may now send cwnd=1 packet */
+ if (sender_was_blocked)
+ tasklet_schedule(&dccp_sk(sk)->dccps_xmitlet);
+ /* restart backed-off timer */
+ sk_reset_timer(sk, &hctx->rtotimer, jiffies + hctx->rto);
out:
bh_unlock_sock(sk);
sock_put(sk);
}
-static void ccid2_start_rto_timer(struct sock *sk)
-{
- struct ccid2_hc_tx_sock *hctx = ccid2_hc_tx_sk(sk);
-
- ccid2_pr_debug("setting RTO timeout=%ld\n", hctx->ccid2hctx_rto);
-
- BUG_ON(timer_pending(&hctx->ccid2hctx_rtotimer));
- sk_reset_timer(sk, &hctx->ccid2hctx_rtotimer,
- jiffies + hctx->ccid2hctx_rto);
-}
-
-static void ccid2_hc_tx_packet_sent(struct sock *sk, int more, unsigned int len)
+static void ccid2_hc_tx_packet_sent(struct sock *sk, unsigned int len)
{
struct dccp_sock *dp = dccp_sk(sk);
struct ccid2_hc_tx_sock *hctx = ccid2_hc_tx_sk(sk);
struct ccid2_seq *next;
- hctx->ccid2hctx_pipe++;
+ hctx->pipe++;
- hctx->ccid2hctx_seqh->ccid2s_seq = dp->dccps_gss;
- hctx->ccid2hctx_seqh->ccid2s_acked = 0;
- hctx->ccid2hctx_seqh->ccid2s_sent = jiffies;
+ hctx->seqh->ccid2s_seq = dp->dccps_gss;
+ hctx->seqh->ccid2s_acked = 0;
+ hctx->seqh->ccid2s_sent = jiffies;
- next = hctx->ccid2hctx_seqh->ccid2s_next;
+ next = hctx->seqh->ccid2s_next;
/* check if we need to alloc more space */
- if (next == hctx->ccid2hctx_seqt) {
+ if (next == hctx->seqt) {
if (ccid2_hc_tx_alloc_seq(hctx)) {
DCCP_CRIT("packet history - out of memory!");
/* FIXME: find a more graceful way to bail out */
return;
}
- next = hctx->ccid2hctx_seqh->ccid2s_next;
- BUG_ON(next == hctx->ccid2hctx_seqt);
+ next = hctx->seqh->ccid2s_next;
+ BUG_ON(next == hctx->seqt);
}
- hctx->ccid2hctx_seqh = next;
+ hctx->seqh = next;
- ccid2_pr_debug("cwnd=%d pipe=%d\n", hctx->ccid2hctx_cwnd,
- hctx->ccid2hctx_pipe);
+ ccid2_pr_debug("cwnd=%d pipe=%d\n", hctx->cwnd, hctx->pipe);
/*
* FIXME: The code below is broken and the variables have been removed
@@ -272,12 +203,12 @@ static void ccid2_hc_tx_packet_sent(struct sock *sk, int more, unsigned int len)
*/
#if 0
/* Ack Ratio. Need to maintain a concept of how many windows we sent */
- hctx->ccid2hctx_arsent++;
+ hctx->arsent++;
/* We had an ack loss in this window... */
- if (hctx->ccid2hctx_ackloss) {
- if (hctx->ccid2hctx_arsent >= hctx->ccid2hctx_cwnd) {
- hctx->ccid2hctx_arsent = 0;
- hctx->ccid2hctx_ackloss = 0;
+ if (hctx->ackloss) {
+ if (hctx->arsent >= hctx->cwnd) {
+ hctx->arsent = 0;
+ hctx->ackloss = 0;
}
} else {
/* No acks lost up to now... */
@@ -287,28 +218,28 @@ static void ccid2_hc_tx_packet_sent(struct sock *sk, int more, unsigned int len)
int denom = dp->dccps_l_ack_ratio * dp->dccps_l_ack_ratio -
dp->dccps_l_ack_ratio;
- denom = hctx->ccid2hctx_cwnd * hctx->ccid2hctx_cwnd / denom;
+ denom = hctx->cwnd * hctx->cwnd / denom;
- if (hctx->ccid2hctx_arsent >= denom) {
+ if (hctx->arsent >= denom) {
ccid2_change_l_ack_ratio(sk, dp->dccps_l_ack_ratio - 1);
- hctx->ccid2hctx_arsent = 0;
+ hctx->arsent = 0;
}
} else {
/* we can't increase ack ratio further [1] */
- hctx->ccid2hctx_arsent = 0; /* or maybe set it to cwnd*/
+ hctx->arsent = 0; /* or maybe set it to cwnd*/
}
}
#endif
/* setup RTO timer */
- if (!timer_pending(&hctx->ccid2hctx_rtotimer))
- ccid2_start_rto_timer(sk);
+ if (!timer_pending(&hctx->rtotimer))
+ sk_reset_timer(sk, &hctx->rtotimer, jiffies + hctx->rto);
#ifdef CONFIG_IP_DCCP_CCID2_DEBUG
do {
- struct ccid2_seq *seqp = hctx->ccid2hctx_seqt;
+ struct ccid2_seq *seqp = hctx->seqt;
- while (seqp != hctx->ccid2hctx_seqh) {
+ while (seqp != hctx->seqh) {
ccid2_pr_debug("out seq=%llu acked=%d time=%lu\n",
(unsigned long long)seqp->ccid2s_seq,
seqp->ccid2s_acked, seqp->ccid2s_sent);
@@ -316,205 +247,158 @@ static void ccid2_hc_tx_packet_sent(struct sock *sk, int more, unsigned int len)
}
} while (0);
ccid2_pr_debug("=========\n");
- ccid2_hc_tx_check_sanity(hctx);
#endif
}
-/* XXX Lame code duplication!
- * returns -1 if none was found.
- * else returns the next offset to use in the function call.
+/**
+ * ccid2_rtt_estimator - Sample RTT and compute RTO using RFC2988 algorithm
+ * This code is almost identical with TCP's tcp_rtt_estimator(), since
+ * - it has a higher sampling frequency (recommended by RFC 1323),
+ * - the RTO does not collapse into RTT due to RTTVAR going towards zero,
+ * - it is simple (cf. more complex proposals such as Eifel timer or research
+ * which suggests that the gain should be set according to window size),
+ * - in tests it was found to work well with CCID2 [gerrit].
*/
-static int ccid2_ackvector(struct sock *sk, struct sk_buff *skb, int offset,
- unsigned char **vec, unsigned char *veclen)
+static void ccid2_rtt_estimator(struct sock *sk, const long mrtt)
{
- const struct dccp_hdr *dh = dccp_hdr(skb);
- unsigned char *options = (unsigned char *)dh + dccp_hdr_len(skb);
- unsigned char *opt_ptr;
- const unsigned char *opt_end = (unsigned char *)dh +
- (dh->dccph_doff * 4);
- unsigned char opt, len;
- unsigned char *value;
-
- BUG_ON(offset < 0);
- options += offset;
- opt_ptr = options;
- if (opt_ptr >= opt_end)
- return -1;
-
- while (opt_ptr != opt_end) {
- opt = *opt_ptr++;
- len = 0;
- value = NULL;
-
- /* Check if this isn't a single byte option */
- if (opt > DCCPO_MAX_RESERVED) {
- if (opt_ptr == opt_end)
- goto out_invalid_option;
-
- len = *opt_ptr++;
- if (len < 3)
- goto out_invalid_option;
+ struct ccid2_hc_tx_sock *hctx = ccid2_hc_tx_sk(sk);
+ long m = mrtt ? : 1;
+
+ if (hctx->srtt == 0) {
+ /* First measurement m */
+ hctx->srtt = m << 3;
+ hctx->mdev = m << 1;
+
+ hctx->mdev_max = max(TCP_RTO_MIN, hctx->mdev);
+ hctx->rttvar = hctx->mdev_max;
+ hctx->rtt_seq = dccp_sk(sk)->dccps_gss;
+ } else {
+ /* Update scaled SRTT as SRTT += 1/8 * (m - SRTT) */
+ m -= (hctx->srtt >> 3);
+ hctx->srtt += m;
+
+ /* Similarly, update scaled mdev with regard to |m| */
+ if (m < 0) {
+ m = -m;
+ m -= (hctx->mdev >> 2);
/*
- * Remove the type and len fields, leaving
- * just the value size
+ * This neutralises RTO increase when RTT < SRTT - mdev
+ * (see P. Sarolahti, A. Kuznetsov,"Congestion Control
+ * in Linux TCP", USENIX 2002, pp. 49-62).
*/
- len -= 2;
- value = opt_ptr;
- opt_ptr += len;
+ if (m > 0)
+ m >>= 3;
+ } else {
+ m -= (hctx->mdev >> 2);
+ }
+ hctx->mdev += m;
- if (opt_ptr > opt_end)
- goto out_invalid_option;
+ if (hctx->mdev > hctx->mdev_max) {
+ hctx->mdev_max = hctx->mdev;
+ if (hctx->mdev_max > hctx->rttvar)
+ hctx->rttvar = hctx->mdev_max;
}
- switch (opt) {
- case DCCPO_ACK_VECTOR_0:
- case DCCPO_ACK_VECTOR_1:
- *vec = value;
- *veclen = len;
- return offset + (opt_ptr - options);
+ /*
+ * Decay RTTVAR at most once per flight, exploiting that
+ * 1) pipe <= cwnd <= Sequence_Window = W (RFC 4340, 7.5.2)
+ * 2) AWL = GSS-W+1 <= GAR <= GSS (RFC 4340, 7.5.1)
+ * GAR is a useful bound for FlightSize = pipe, AWL is probably
+ * too low as it over-estimates pipe.
+ */
+ if (after48(dccp_sk(sk)->dccps_gar, hctx->rtt_seq)) {
+ if (hctx->mdev_max < hctx->rttvar)
+ hctx->rttvar -= (hctx->rttvar -
+ hctx->mdev_max) >> 2;
+ hctx->rtt_seq = dccp_sk(sk)->dccps_gss;
+ hctx->mdev_max = TCP_RTO_MIN;
}
}
- return -1;
-
-out_invalid_option:
- DCCP_BUG("Invalid option - this should not happen (previous parsing)!");
- return -1;
-}
-
-static void ccid2_hc_tx_kill_rto_timer(struct sock *sk)
-{
- struct ccid2_hc_tx_sock *hctx = ccid2_hc_tx_sk(sk);
+ /*
+ * Set RTO from SRTT and RTTVAR
+ * Clock granularity is ignored since the minimum error for RTTVAR is
+ * clamped to 50msec (corresponding to HZ=20). This leads to a minimum
+ * RTO of 200msec. This agrees with TCP and RFC 4341, 5.: "Because DCCP
+ * does not retransmit data, DCCP does not require TCP's recommended
+ * minimum timeout of one second".
+ */
+ hctx->rto = (hctx->srtt >> 3) + hctx->rttvar;
- sk_stop_timer(sk, &hctx->ccid2hctx_rtotimer);
- ccid2_pr_debug("deleted RTO timer\n");
+ if (hctx->rto > DCCP_RTO_MAX)
+ hctx->rto = DCCP_RTO_MAX;
}
-static inline void ccid2_new_ack(struct sock *sk,
- struct ccid2_seq *seqp,
- unsigned int *maxincr)
+static void ccid2_new_ack(struct sock *sk, struct ccid2_seq *seqp,
+ unsigned int *maxincr)
{
struct ccid2_hc_tx_sock *hctx = ccid2_hc_tx_sk(sk);
- if (hctx->ccid2hctx_cwnd < hctx->ccid2hctx_ssthresh) {
- if (*maxincr > 0 && ++hctx->ccid2hctx_packets_acked == 2) {
- hctx->ccid2hctx_cwnd += 1;
- *maxincr -= 1;
- hctx->ccid2hctx_packets_acked = 0;
+ if (hctx->cwnd < hctx->ssthresh) {
+ if (*maxincr > 0 && ++hctx->packets_acked == 2) {
+ hctx->cwnd += 1;
+ *maxincr -= 1;
+ hctx->packets_acked = 0;
}
- } else if (++hctx->ccid2hctx_packets_acked >= hctx->ccid2hctx_cwnd) {
- hctx->ccid2hctx_cwnd += 1;
- hctx->ccid2hctx_packets_acked = 0;
+ } else if (++hctx->packets_acked >= hctx->cwnd) {
+ hctx->cwnd += 1;
+ hctx->packets_acked = 0;
}
-
- /* update RTO */
- if (hctx->ccid2hctx_srtt == -1 ||
- time_after(jiffies, hctx->ccid2hctx_lastrtt + hctx->ccid2hctx_srtt)) {
- unsigned long r = (long)jiffies - (long)seqp->ccid2s_sent;
- int s;
-
- /* first measurement */
- if (hctx->ccid2hctx_srtt == -1) {
- ccid2_pr_debug("R: %lu Time=%lu seq=%llu\n",
- r, jiffies,
- (unsigned long long)seqp->ccid2s_seq);
- ccid2_change_srtt(hctx, r);
- hctx->ccid2hctx_rttvar = r >> 1;
- } else {
- /* RTTVAR */
- long tmp = hctx->ccid2hctx_srtt - r;
- long srtt;
-
- if (tmp < 0)
- tmp *= -1;
-
- tmp >>= 2;
- hctx->ccid2hctx_rttvar *= 3;
- hctx->ccid2hctx_rttvar >>= 2;
- hctx->ccid2hctx_rttvar += tmp;
-
- /* SRTT */
- srtt = hctx->ccid2hctx_srtt;
- srtt *= 7;
- srtt >>= 3;
- tmp = r >> 3;
- srtt += tmp;
- ccid2_change_srtt(hctx, srtt);
- }
- s = hctx->ccid2hctx_rttvar << 2;
- /* clock granularity is 1 when based on jiffies */
- if (!s)
- s = 1;
- hctx->ccid2hctx_rto = hctx->ccid2hctx_srtt + s;
-
- /* must be at least a second */
- s = hctx->ccid2hctx_rto / HZ;
- /* DCCP doesn't require this [but I like it cuz my code sux] */
-#if 1
- if (s < 1)
- hctx->ccid2hctx_rto = HZ;
-#endif
- /* max 60 seconds */
- if (s > 60)
- hctx->ccid2hctx_rto = HZ * 60;
-
- hctx->ccid2hctx_lastrtt = jiffies;
-
- ccid2_pr_debug("srtt: %ld rttvar: %ld rto: %ld (HZ=%d) R=%lu\n",
- hctx->ccid2hctx_srtt, hctx->ccid2hctx_rttvar,
- hctx->ccid2hctx_rto, HZ, r);
- }
-
- /* we got a new ack, so re-start RTO timer */
- ccid2_hc_tx_kill_rto_timer(sk);
- ccid2_start_rto_timer(sk);
-}
-
-static void ccid2_hc_tx_dec_pipe(struct sock *sk)
-{
- struct ccid2_hc_tx_sock *hctx = ccid2_hc_tx_sk(sk);
-
- if (hctx->ccid2hctx_pipe == 0)
- DCCP_BUG("pipe == 0");
- else
- hctx->ccid2hctx_pipe--;
-
- if (hctx->ccid2hctx_pipe == 0)
- ccid2_hc_tx_kill_rto_timer(sk);
+ /*
+ * FIXME: RTT is sampled several times per acknowledgment (for each
+ * entry in the Ack Vector), instead of once per Ack (as in TCP SACK).
+ * This causes the RTT to be over-estimated, since the older entries
+ * in the Ack Vector have earlier sending times.
+ * The cleanest solution is to not use the ccid2s_sent field at all
+ * and instead use DCCP timestamps - need to be resolved at some time.
+ */
+ ccid2_rtt_estimator(sk, jiffies - seqp->ccid2s_sent);
}
static void ccid2_congestion_event(struct sock *sk, struct ccid2_seq *seqp)
{
struct ccid2_hc_tx_sock *hctx = ccid2_hc_tx_sk(sk);
- if (time_before(seqp->ccid2s_sent, hctx->ccid2hctx_last_cong)) {
+ if (time_before(seqp->ccid2s_sent, hctx->last_cong)) {
ccid2_pr_debug("Multiple losses in an RTT---treating as one\n");
return;
}
- hctx->ccid2hctx_last_cong = jiffies;
+ hctx->last_cong = jiffies;
- hctx->ccid2hctx_cwnd = hctx->ccid2hctx_cwnd / 2 ? : 1U;
- hctx->ccid2hctx_ssthresh = max(hctx->ccid2hctx_cwnd, 2U);
+ hctx->cwnd = hctx->cwnd / 2 ? : 1U;
+ hctx->ssthresh = max(hctx->cwnd, 2U);
/* Avoid spurious timeouts resulting from Ack Ratio > cwnd */
- if (dccp_sk(sk)->dccps_l_ack_ratio > hctx->ccid2hctx_cwnd)
- ccid2_change_l_ack_ratio(sk, hctx->ccid2hctx_cwnd);
+ if (dccp_sk(sk)->dccps_l_ack_ratio > hctx->cwnd)
+ ccid2_change_l_ack_ratio(sk, hctx->cwnd);
+}
+
+static int ccid2_hc_tx_parse_options(struct sock *sk, u8 packet_type,
+ u8 option, u8 *optval, u8 optlen)
+{
+ struct ccid2_hc_tx_sock *hctx = ccid2_hc_tx_sk(sk);
+
+ switch (option) {
+ case DCCPO_ACK_VECTOR_0:
+ case DCCPO_ACK_VECTOR_1:
+ return dccp_ackvec_parsed_add(&hctx->av_chunks, optval, optlen,
+ option - DCCPO_ACK_VECTOR_0);
+ }
+ return 0;
}
static void ccid2_hc_tx_packet_recv(struct sock *sk, struct sk_buff *skb)
{
struct dccp_sock *dp = dccp_sk(sk);
struct ccid2_hc_tx_sock *hctx = ccid2_hc_tx_sk(sk);
+ const bool sender_was_blocked = ccid2_cwnd_network_limited(hctx);
+ struct dccp_ackvec_parsed *avp;
u64 ackno, seqno;
struct ccid2_seq *seqp;
- unsigned char *vector;
- unsigned char veclen;
- int offset = 0;
int done = 0;
unsigned int maxincr = 0;
- ccid2_hc_tx_check_sanity(hctx);
/* check reverse path congestion */
seqno = DCCP_SKB_CB(skb)->dccpd_seq;
@@ -523,21 +407,21 @@ static void ccid2_hc_tx_packet_recv(struct sock *sk, struct sk_buff *skb)
* -sorbo.
*/
/* need to bootstrap */
- if (hctx->ccid2hctx_rpdupack == -1) {
- hctx->ccid2hctx_rpdupack = 0;
- hctx->ccid2hctx_rpseq = seqno;
+ if (hctx->rpdupack == -1) {
+ hctx->rpdupack = 0;
+ hctx->rpseq = seqno;
} else {
/* check if packet is consecutive */
- if (dccp_delta_seqno(hctx->ccid2hctx_rpseq, seqno) == 1)
- hctx->ccid2hctx_rpseq = seqno;
+ if (dccp_delta_seqno(hctx->rpseq, seqno) == 1)
+ hctx->rpseq = seqno;
/* it's a later packet */
- else if (after48(seqno, hctx->ccid2hctx_rpseq)) {
- hctx->ccid2hctx_rpdupack++;
+ else if (after48(seqno, hctx->rpseq)) {
+ hctx->rpdupack++;
/* check if we got enough dupacks */
- if (hctx->ccid2hctx_rpdupack >= NUMDUPACK) {
- hctx->ccid2hctx_rpdupack = -1; /* XXX lame */
- hctx->ccid2hctx_rpseq = 0;
+ if (hctx->rpdupack >= NUMDUPACK) {
+ hctx->rpdupack = -1; /* XXX lame */
+ hctx->rpseq = 0;
ccid2_change_l_ack_ratio(sk, 2 * dp->dccps_l_ack_ratio);
}
@@ -545,27 +429,22 @@ static void ccid2_hc_tx_packet_recv(struct sock *sk, struct sk_buff *skb)
}
/* check forward path congestion */
- /* still didn't send out new data packets */
- if (hctx->ccid2hctx_seqh == hctx->ccid2hctx_seqt)
+ if (dccp_packet_without_ack(skb))
return;
- switch (DCCP_SKB_CB(skb)->dccpd_type) {
- case DCCP_PKT_ACK:
- case DCCP_PKT_DATAACK:
- break;
- default:
- return;
- }
+ /* still didn't send out new data packets */
+ if (hctx->seqh == hctx->seqt)
+ goto done;
ackno = DCCP_SKB_CB(skb)->dccpd_ack_seq;
- if (after48(ackno, hctx->ccid2hctx_high_ack))
- hctx->ccid2hctx_high_ack = ackno;
+ if (after48(ackno, hctx->high_ack))
+ hctx->high_ack = ackno;
- seqp = hctx->ccid2hctx_seqt;
+ seqp = hctx->seqt;
while (before48(seqp->ccid2s_seq, ackno)) {
seqp = seqp->ccid2s_next;
- if (seqp == hctx->ccid2hctx_seqh) {
- seqp = hctx->ccid2hctx_seqh->ccid2s_prev;
+ if (seqp == hctx->seqh) {
+ seqp = hctx->seqh->ccid2s_prev;
break;
}
}
@@ -575,26 +454,26 @@ static void ccid2_hc_tx_packet_recv(struct sock *sk, struct sk_buff *skb)
* packets per acknowledgement. Rounding up avoids that cwnd is not
* advanced when Ack Ratio is 1 and gives a slight edge otherwise.
*/
- if (hctx->ccid2hctx_cwnd < hctx->ccid2hctx_ssthresh)
+ if (hctx->cwnd < hctx->ssthresh)
maxincr = DIV_ROUND_UP(dp->dccps_l_ack_ratio, 2);
/* go through all ack vectors */
- while ((offset = ccid2_ackvector(sk, skb, offset,
- &vector, &veclen)) != -1) {
+ list_for_each_entry(avp, &hctx->av_chunks, node) {
/* go through this ack vector */
- while (veclen--) {
- const u8 rl = *vector & DCCP_ACKVEC_LEN_MASK;
- u64 ackno_end_rl = SUB48(ackno, rl);
+ for (; avp->len--; avp->vec++) {
+ u64 ackno_end_rl = SUB48(ackno,
+ dccp_ackvec_runlen(avp->vec));
- ccid2_pr_debug("ackvec start:%llu end:%llu\n",
+ ccid2_pr_debug("ackvec %llu |%u,%u|\n",
(unsigned long long)ackno,
- (unsigned long long)ackno_end_rl);
+ dccp_ackvec_state(avp->vec) >> 6,
+ dccp_ackvec_runlen(avp->vec));
/* if the seqno we are analyzing is larger than the
* current ackno, then move towards the tail of our
* seqnos.
*/
while (after48(seqp->ccid2s_seq, ackno)) {
- if (seqp == hctx->ccid2hctx_seqt) {
+ if (seqp == hctx->seqt) {
done = 1;
break;
}
@@ -607,26 +486,24 @@ static void ccid2_hc_tx_packet_recv(struct sock *sk, struct sk_buff *skb)
* run length
*/
while (between48(seqp->ccid2s_seq,ackno_end_rl,ackno)) {
- const u8 state = *vector &
- DCCP_ACKVEC_STATE_MASK;
+ const u8 state = dccp_ackvec_state(avp->vec);
/* new packet received or marked */
- if (state != DCCP_ACKVEC_STATE_NOT_RECEIVED &&
+ if (state != DCCPAV_NOT_RECEIVED &&
!seqp->ccid2s_acked) {
- if (state ==
- DCCP_ACKVEC_STATE_ECN_MARKED) {
+ if (state == DCCPAV_ECN_MARKED)
ccid2_congestion_event(sk,
seqp);
- } else
+ else
ccid2_new_ack(sk, seqp,
&maxincr);
seqp->ccid2s_acked = 1;
ccid2_pr_debug("Got ack for %llu\n",
(unsigned long long)seqp->ccid2s_seq);
- ccid2_hc_tx_dec_pipe(sk);
+ hctx->pipe--;
}
- if (seqp == hctx->ccid2hctx_seqt) {
+ if (seqp == hctx->seqt) {
done = 1;
break;
}
@@ -636,7 +513,6 @@ static void ccid2_hc_tx_packet_recv(struct sock *sk, struct sk_buff *skb)
break;
ackno = SUB48(ackno_end_rl, 1);
- vector++;
}
if (done)
break;
@@ -645,11 +521,11 @@ static void ccid2_hc_tx_packet_recv(struct sock *sk, struct sk_buff *skb)
/* The state about what is acked should be correct now
* Check for NUMDUPACK
*/
- seqp = hctx->ccid2hctx_seqt;
- while (before48(seqp->ccid2s_seq, hctx->ccid2hctx_high_ack)) {
+ seqp = hctx->seqt;
+ while (before48(seqp->ccid2s_seq, hctx->high_ack)) {
seqp = seqp->ccid2s_next;
- if (seqp == hctx->ccid2hctx_seqh) {
- seqp = hctx->ccid2hctx_seqh->ccid2s_prev;
+ if (seqp == hctx->seqh) {
+ seqp = hctx->seqh->ccid2s_prev;
break;
}
}
@@ -660,7 +536,7 @@ static void ccid2_hc_tx_packet_recv(struct sock *sk, struct sk_buff *skb)
if (done == NUMDUPACK)
break;
}
- if (seqp == hctx->ccid2hctx_seqt)
+ if (seqp == hctx->seqt)
break;
seqp = seqp->ccid2s_prev;
}
@@ -681,25 +557,34 @@ static void ccid2_hc_tx_packet_recv(struct sock *sk, struct sk_buff *skb)
* one ack vector.
*/
ccid2_congestion_event(sk, seqp);
- ccid2_hc_tx_dec_pipe(sk);
+ hctx->pipe--;
}
- if (seqp == hctx->ccid2hctx_seqt)
+ if (seqp == hctx->seqt)
break;
seqp = seqp->ccid2s_prev;
}
- hctx->ccid2hctx_seqt = last_acked;
+ hctx->seqt = last_acked;
}
/* trim acked packets in tail */
- while (hctx->ccid2hctx_seqt != hctx->ccid2hctx_seqh) {
- if (!hctx->ccid2hctx_seqt->ccid2s_acked)
+ while (hctx->seqt != hctx->seqh) {
+ if (!hctx->seqt->ccid2s_acked)
break;
- hctx->ccid2hctx_seqt = hctx->ccid2hctx_seqt->ccid2s_next;
+ hctx->seqt = hctx->seqt->ccid2s_next;
}
- ccid2_hc_tx_check_sanity(hctx);
+ /* restart RTO timer if not all outstanding data has been acked */
+ if (hctx->pipe == 0)
+ sk_stop_timer(sk, &hctx->rtotimer);
+ else
+ sk_reset_timer(sk, &hctx->rtotimer, jiffies + hctx->rto);
+done:
+ /* check if incoming Acks allow pending packets to be sent */
+ if (sender_was_blocked && !ccid2_cwnd_network_limited(hctx))
+ tasklet_schedule(&dccp_sk(sk)->dccps_xmitlet);
+ dccp_ackvec_parsed_cleanup(&hctx->av_chunks);
}
static int ccid2_hc_tx_init(struct ccid *ccid, struct sock *sk)
@@ -709,17 +594,13 @@ static int ccid2_hc_tx_init(struct ccid *ccid, struct sock *sk)
u32 max_ratio;
/* RFC 4341, 5: initialise ssthresh to arbitrarily high (max) value */
- hctx->ccid2hctx_ssthresh = ~0U;
+ hctx->ssthresh = ~0U;
- /*
- * RFC 4341, 5: "The cwnd parameter is initialized to at most four
- * packets for new connections, following the rules from [RFC3390]".
- * We need to convert the bytes of RFC3390 into the packets of RFC 4341.
- */
- hctx->ccid2hctx_cwnd = clamp(4380U / dp->dccps_mss_cache, 2U, 4U);
+ /* Use larger initial windows (RFC 3390, rfc2581bis) */
+ hctx->cwnd = rfc3390_bytes_to_packets(dp->dccps_mss_cache);
/* Make sure that Ack Ratio is enabled and within bounds. */
- max_ratio = DIV_ROUND_UP(hctx->ccid2hctx_cwnd, 2);
+ max_ratio = DIV_ROUND_UP(hctx->cwnd, 2);
if (dp->dccps_l_ack_ratio == 0 || dp->dccps_l_ack_ratio > max_ratio)
dp->dccps_l_ack_ratio = max_ratio;
@@ -727,15 +608,11 @@ static int ccid2_hc_tx_init(struct ccid *ccid, struct sock *sk)
if (ccid2_hc_tx_alloc_seq(hctx))
return -ENOMEM;
- hctx->ccid2hctx_rto = 3 * HZ;
- ccid2_change_srtt(hctx, -1);
- hctx->ccid2hctx_rttvar = -1;
- hctx->ccid2hctx_rpdupack = -1;
- hctx->ccid2hctx_last_cong = jiffies;
- setup_timer(&hctx->ccid2hctx_rtotimer, ccid2_hc_tx_rto_expire,
- (unsigned long)sk);
-
- ccid2_hc_tx_check_sanity(hctx);
+ hctx->rto = DCCP_TIMEOUT_INIT;
+ hctx->rpdupack = -1;
+ hctx->last_cong = jiffies;
+ setup_timer(&hctx->rtotimer, ccid2_hc_tx_rto_expire, (unsigned long)sk);
+ INIT_LIST_HEAD(&hctx->av_chunks);
return 0;
}
@@ -744,11 +621,11 @@ static void ccid2_hc_tx_exit(struct sock *sk)
struct ccid2_hc_tx_sock *hctx = ccid2_hc_tx_sk(sk);
int i;
- ccid2_hc_tx_kill_rto_timer(sk);
+ sk_stop_timer(sk, &hctx->rtotimer);
- for (i = 0; i < hctx->ccid2hctx_seqbufc; i++)
- kfree(hctx->ccid2hctx_seqbuf[i]);
- hctx->ccid2hctx_seqbufc = 0;
+ for (i = 0; i < hctx->seqbufc; i++)
+ kfree(hctx->seqbuf[i]);
+ hctx->seqbufc = 0;
}
static void ccid2_hc_rx_packet_recv(struct sock *sk, struct sk_buff *skb)
@@ -759,27 +636,28 @@ static void ccid2_hc_rx_packet_recv(struct sock *sk, struct sk_buff *skb)
switch (DCCP_SKB_CB(skb)->dccpd_type) {
case DCCP_PKT_DATA:
case DCCP_PKT_DATAACK:
- hcrx->ccid2hcrx_data++;
- if (hcrx->ccid2hcrx_data >= dp->dccps_r_ack_ratio) {
+ hcrx->data++;
+ if (hcrx->data >= dp->dccps_r_ack_ratio) {
dccp_send_ack(sk);
- hcrx->ccid2hcrx_data = 0;
+ hcrx->data = 0;
}
break;
}
}
static struct ccid_operations ccid2 = {
- .ccid_id = DCCPC_CCID2,
- .ccid_name = "TCP-like",
- .ccid_owner = THIS_MODULE,
- .ccid_hc_tx_obj_size = sizeof(struct ccid2_hc_tx_sock),
- .ccid_hc_tx_init = ccid2_hc_tx_init,
- .ccid_hc_tx_exit = ccid2_hc_tx_exit,
- .ccid_hc_tx_send_packet = ccid2_hc_tx_send_packet,
- .ccid_hc_tx_packet_sent = ccid2_hc_tx_packet_sent,
- .ccid_hc_tx_packet_recv = ccid2_hc_tx_packet_recv,
- .ccid_hc_rx_obj_size = sizeof(struct ccid2_hc_rx_sock),
- .ccid_hc_rx_packet_recv = ccid2_hc_rx_packet_recv,
+ .ccid_id = DCCPC_CCID2,
+ .ccid_name = "TCP-like",
+ .ccid_owner = THIS_MODULE,
+ .ccid_hc_tx_obj_size = sizeof(struct ccid2_hc_tx_sock),
+ .ccid_hc_tx_init = ccid2_hc_tx_init,
+ .ccid_hc_tx_exit = ccid2_hc_tx_exit,
+ .ccid_hc_tx_send_packet = ccid2_hc_tx_send_packet,
+ .ccid_hc_tx_packet_sent = ccid2_hc_tx_packet_sent,
+ .ccid_hc_tx_parse_options = ccid2_hc_tx_parse_options,
+ .ccid_hc_tx_packet_recv = ccid2_hc_tx_packet_recv,
+ .ccid_hc_rx_obj_size = sizeof(struct ccid2_hc_rx_sock),
+ .ccid_hc_rx_packet_recv = ccid2_hc_rx_packet_recv,
};
#ifdef CONFIG_IP_DCCP_CCID2_DEBUG
diff --git a/net/dccp/ccids/ccid2.h b/net/dccp/ccids/ccid2.h
index 2c94ca02901..8b7a2dee2f6 100644
--- a/net/dccp/ccids/ccid2.h
+++ b/net/dccp/ccids/ccid2.h
@@ -42,34 +42,49 @@ struct ccid2_seq {
/** struct ccid2_hc_tx_sock - CCID2 TX half connection
*
- * @ccid2hctx_{cwnd,ssthresh,pipe}: as per RFC 4341, section 5
- * @ccid2hctx_packets_acked - Ack counter for deriving cwnd growth (RFC 3465)
- * @ccid2hctx_lastrtt -time RTT was last measured
- * @ccid2hctx_rpseq - last consecutive seqno
- * @ccid2hctx_rpdupack - dupacks since rpseq
-*/
+ * @{cwnd,ssthresh,pipe}: as per RFC 4341, section 5
+ * @packets_acked: Ack counter for deriving cwnd growth (RFC 3465)
+ * @srtt: smoothed RTT estimate, scaled by 2^3
+ * @mdev: smoothed RTT variation, scaled by 2^2
+ * @mdev_max: maximum of @mdev during one flight
+ * @rttvar: moving average/maximum of @mdev_max
+ * @rto: RTO value deriving from SRTT and RTTVAR (RFC 2988)
+ * @rtt_seq: to decay RTTVAR at most once per flight
+ * @rpseq: last consecutive seqno
+ * @rpdupack: dupacks since rpseq
+ * @av_chunks: list of Ack Vectors received on current skb
+ */
struct ccid2_hc_tx_sock {
- u32 ccid2hctx_cwnd;
- u32 ccid2hctx_ssthresh;
- u32 ccid2hctx_pipe;
- u32 ccid2hctx_packets_acked;
- struct ccid2_seq *ccid2hctx_seqbuf[CCID2_SEQBUF_MAX];
- int ccid2hctx_seqbufc;
- struct ccid2_seq *ccid2hctx_seqh;
- struct ccid2_seq *ccid2hctx_seqt;
- long ccid2hctx_rto;
- long ccid2hctx_srtt;
- long ccid2hctx_rttvar;
- unsigned long ccid2hctx_lastrtt;
- struct timer_list ccid2hctx_rtotimer;
- u64 ccid2hctx_rpseq;
- int ccid2hctx_rpdupack;
- unsigned long ccid2hctx_last_cong;
- u64 ccid2hctx_high_ack;
+ u32 cwnd;
+ u32 ssthresh;
+ u32 pipe;
+ u32 packets_acked;
+ struct ccid2_seq *seqbuf[CCID2_SEQBUF_MAX];
+ int seqbufc;
+ struct ccid2_seq *seqh;
+ struct ccid2_seq *seqt;
+ /* RTT measurement: variables/principles are the same as in TCP */
+ u32 srtt,
+ mdev,
+ mdev_max,
+ rttvar,
+ rto;
+ u64 rtt_seq:48;
+ struct timer_list rtotimer;
+ u64 rpseq;
+ int rpdupack;
+ unsigned long last_cong;
+ u64 high_ack;
+ struct list_head av_chunks;
};
+static inline bool ccid2_cwnd_network_limited(struct ccid2_hc_tx_sock *hctx)
+{
+ return (hctx->pipe >= hctx->cwnd);
+}
+
struct ccid2_hc_rx_sock {
- int ccid2hcrx_data;
+ int data;
};
static inline struct ccid2_hc_tx_sock *ccid2_hc_tx_sk(const struct sock *sk)
diff --git a/net/dccp/ccids/ccid3.c b/net/dccp/ccids/ccid3.c
index 3b8bd7ca676..06cfdad84a6 100644
--- a/net/dccp/ccids/ccid3.c
+++ b/net/dccp/ccids/ccid3.c
@@ -49,75 +49,41 @@ static int ccid3_debug;
/*
* Transmitter Half-Connection Routines
*/
-#ifdef CONFIG_IP_DCCP_CCID3_DEBUG
-static const char *ccid3_tx_state_name(enum ccid3_hc_tx_states state)
-{
- static char *ccid3_state_names[] = {
- [TFRC_SSTATE_NO_SENT] = "NO_SENT",
- [TFRC_SSTATE_NO_FBACK] = "NO_FBACK",
- [TFRC_SSTATE_FBACK] = "FBACK",
- [TFRC_SSTATE_TERM] = "TERM",
- };
-
- return ccid3_state_names[state];
-}
-#endif
-
-static void ccid3_hc_tx_set_state(struct sock *sk,
- enum ccid3_hc_tx_states state)
-{
- struct ccid3_hc_tx_sock *hctx = ccid3_hc_tx_sk(sk);
- enum ccid3_hc_tx_states oldstate = hctx->ccid3hctx_state;
-
- ccid3_pr_debug("%s(%p) %-8.8s -> %s\n",
- dccp_role(sk), sk, ccid3_tx_state_name(oldstate),
- ccid3_tx_state_name(state));
- WARN_ON(state == oldstate);
- hctx->ccid3hctx_state = state;
-}
+/* Oscillation Prevention/Reduction: recommended by rfc3448bis, on by default */
+static int do_osc_prev = true;
/*
* Compute the initial sending rate X_init in the manner of RFC 3390:
*
- * X_init = min(4 * s, max(2 * s, 4380 bytes)) / RTT
+ * X_init = min(4 * MPS, max(2 * MPS, 4380 bytes)) / RTT
*
- * Note that RFC 3390 uses MSS, RFC 4342 refers to RFC 3390, and rfc3448bis
- * (rev-02) clarifies the use of RFC 3390 with regard to the above formula.
* For consistency with other parts of the code, X_init is scaled by 2^6.
*/
static inline u64 rfc3390_initial_rate(struct sock *sk)
{
- const struct ccid3_hc_tx_sock *hctx = ccid3_hc_tx_sk(sk);
- const __u32 w_init = clamp_t(__u32, 4380U,
- 2 * hctx->ccid3hctx_s, 4 * hctx->ccid3hctx_s);
+ const u32 mps = dccp_sk(sk)->dccps_mss_cache,
+ w_init = clamp(4380U, 2 * mps, 4 * mps);
- return scaled_div(w_init << 6, hctx->ccid3hctx_rtt);
+ return scaled_div(w_init << 6, ccid3_hc_tx_sk(sk)->rtt);
}
-/*
- * Recalculate t_ipi and delta (should be called whenever X changes)
+/**
+ * ccid3_update_send_interval - Calculate new t_ipi = s / X
+ * This respects the granularity of X (64 * bytes/second) and enforces the
+ * scaled minimum of s * 64 / t_mbi = `s' bytes/second as per RFC 3448/4342.
*/
static void ccid3_update_send_interval(struct ccid3_hc_tx_sock *hctx)
{
- /* Calculate new t_ipi = s / X_inst (X_inst is in 64 * bytes/second) */
- hctx->ccid3hctx_t_ipi = scaled_div32(((u64)hctx->ccid3hctx_s) << 6,
- hctx->ccid3hctx_x);
-
- /* Calculate new delta by delta = min(t_ipi / 2, t_gran / 2) */
- hctx->ccid3hctx_delta = min_t(u32, hctx->ccid3hctx_t_ipi / 2,
- TFRC_OPSYS_HALF_TIME_GRAN);
-
- ccid3_pr_debug("t_ipi=%u, delta=%u, s=%u, X=%u\n",
- hctx->ccid3hctx_t_ipi, hctx->ccid3hctx_delta,
- hctx->ccid3hctx_s, (unsigned)(hctx->ccid3hctx_x >> 6));
-
+ if (unlikely(hctx->x <= hctx->s))
+ hctx->x = hctx->s;
+ hctx->t_ipi = scaled_div32(((u64)hctx->s) << 6, hctx->x);
}
static u32 ccid3_hc_tx_idle_rtt(struct ccid3_hc_tx_sock *hctx, ktime_t now)
{
- u32 delta = ktime_us_delta(now, hctx->ccid3hctx_t_last_win_count);
+ u32 delta = ktime_us_delta(now, hctx->t_last_win_count);
- return delta / hctx->ccid3hctx_rtt;
+ return delta / hctx->rtt;
}
/**
@@ -133,8 +99,8 @@ static u32 ccid3_hc_tx_idle_rtt(struct ccid3_hc_tx_sock *hctx, ktime_t now)
static void ccid3_hc_tx_update_x(struct sock *sk, ktime_t *stamp)
{
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;
+ u64 min_rate = 2 * hctx->x_recv;
+ const u64 old_x = hctx->x;
ktime_t now = stamp ? *stamp : ktime_get_real();
/*
@@ -145,50 +111,44 @@ static void ccid3_hc_tx_update_x(struct sock *sk, ktime_t *stamp)
*/
if (ccid3_hc_tx_idle_rtt(hctx, now) >= 2) {
min_rate = rfc3390_initial_rate(sk);
- min_rate = max(min_rate, 2 * hctx->ccid3hctx_x_recv);
+ min_rate = max(min_rate, 2 * hctx->x_recv);
}
- if (hctx->ccid3hctx_p > 0) {
+ if (hctx->p > 0) {
- hctx->ccid3hctx_x = min(((__u64)hctx->ccid3hctx_x_calc) << 6,
- min_rate);
- hctx->ccid3hctx_x = max(hctx->ccid3hctx_x,
- (((__u64)hctx->ccid3hctx_s) << 6) /
- TFRC_T_MBI);
+ hctx->x = min(((u64)hctx->x_calc) << 6, min_rate);
- } else if (ktime_us_delta(now, hctx->ccid3hctx_t_ld)
- - (s64)hctx->ccid3hctx_rtt >= 0) {
+ } else if (ktime_us_delta(now, hctx->t_ld) - (s64)hctx->rtt >= 0) {
- hctx->ccid3hctx_x = min(2 * hctx->ccid3hctx_x, min_rate);
- hctx->ccid3hctx_x = max(hctx->ccid3hctx_x,
- scaled_div(((__u64)hctx->ccid3hctx_s) << 6,
- hctx->ccid3hctx_rtt));
- hctx->ccid3hctx_t_ld = now;
+ hctx->x = min(2 * hctx->x, min_rate);
+ hctx->x = max(hctx->x,
+ scaled_div(((u64)hctx->s) << 6, hctx->rtt));
+ hctx->t_ld = now;
}
- if (hctx->ccid3hctx_x != old_x) {
+ if (hctx->x != old_x) {
ccid3_pr_debug("X_prev=%u, X_now=%u, X_calc=%u, "
"X_recv=%u\n", (unsigned)(old_x >> 6),
- (unsigned)(hctx->ccid3hctx_x >> 6),
- hctx->ccid3hctx_x_calc,
- (unsigned)(hctx->ccid3hctx_x_recv >> 6));
+ (unsigned)(hctx->x >> 6), hctx->x_calc,
+ (unsigned)(hctx->x_recv >> 6));
ccid3_update_send_interval(hctx);
}
}
/*
- * Track the mean packet size `s' (cf. RFC 4342, 5.3 and RFC 3448, 4.1)
- * @len: DCCP packet payload size in bytes
+ * ccid3_hc_tx_measure_packet_size - Measuring the packet size `s' (sec 4.1)
+ * @new_len: DCCP payload size in bytes (not used by all methods)
*/
-static inline void ccid3_hc_tx_update_s(struct ccid3_hc_tx_sock *hctx, int len)
+static u32 ccid3_hc_tx_measure_packet_size(struct sock *sk, const u16 new_len)
{
- const u16 old_s = hctx->ccid3hctx_s;
-
- hctx->ccid3hctx_s = tfrc_ewma(hctx->ccid3hctx_s, len, 9);
-
- if (hctx->ccid3hctx_s != old_s)
- ccid3_update_send_interval(hctx);
+#if defined(CONFIG_IP_DCCP_CCID3_MEASURE_S_AS_AVG)
+ return tfrc_ewma(ccid3_hc_tx_sk(sk)->s, new_len, 9);
+#elif defined(CONFIG_IP_DCCP_CCID3_MEASURE_S_AS_MAX)
+ return max(ccid3_hc_tx_sk(sk)->s, new_len);
+#else /* CONFIG_IP_DCCP_CCID3_MEASURE_S_AS_MPS */
+ return dccp_sk(sk)->dccps_mss_cache;
+#endif
}
/*
@@ -198,13 +158,13 @@ static inline void ccid3_hc_tx_update_s(struct ccid3_hc_tx_sock *hctx, int len)
static inline void ccid3_hc_tx_update_win_count(struct ccid3_hc_tx_sock *hctx,
ktime_t now)
{
- u32 delta = ktime_us_delta(now, hctx->ccid3hctx_t_last_win_count),
- quarter_rtts = (4 * delta) / hctx->ccid3hctx_rtt;
+ u32 delta = ktime_us_delta(now, hctx->t_last_win_count),
+ quarter_rtts = (4 * delta) / hctx->rtt;
if (quarter_rtts > 0) {
- hctx->ccid3hctx_t_last_win_count = now;
- hctx->ccid3hctx_last_win_count += min(quarter_rtts, 5U);
- hctx->ccid3hctx_last_win_count &= 0xF; /* mod 16 */
+ hctx->t_last_win_count = now;
+ hctx->last_win_count += min(quarter_rtts, 5U);
+ hctx->last_win_count &= 0xF; /* mod 16 */
}
}
@@ -221,25 +181,26 @@ static void ccid3_hc_tx_no_feedback_timer(unsigned long data)
goto restart_timer;
}
- ccid3_pr_debug("%s(%p, state=%s) - entry \n", dccp_role(sk), sk,
- ccid3_tx_state_name(hctx->ccid3hctx_state));
+ ccid3_pr_debug("%s(%p) entry with%s feedback\n", dccp_role(sk), sk,
+ hctx->feedback ? "" : "out");
- if (hctx->ccid3hctx_state == TFRC_SSTATE_FBACK)
- ccid3_hc_tx_set_state(sk, TFRC_SSTATE_NO_FBACK);
- else if (hctx->ccid3hctx_state != TFRC_SSTATE_NO_FBACK)
+ /* Ignore and do not restart after leaving the established state */
+ if ((1 << sk->sk_state) & ~(DCCPF_OPEN | DCCPF_PARTOPEN))
goto out;
+ /* Reset feedback state to "no feedback received" */
+ hctx->feedback = false;
+
/*
* Determine new allowed sending rate X as per draft rfc3448bis-00, 4.4
+ * RTO is 0 if and only if no feedback has been received yet.
*/
- if (hctx->ccid3hctx_t_rto == 0 || /* no feedback received yet */
- hctx->ccid3hctx_p == 0) {
+ if (hctx->t_rto == 0 || hctx->p == 0) {
/* halve send rate directly */
- hctx->ccid3hctx_x = max(hctx->ccid3hctx_x / 2,
- (((__u64)hctx->ccid3hctx_s) << 6) /
- TFRC_T_MBI);
+ hctx->x /= 2;
ccid3_update_send_interval(hctx);
+
} else {
/*
* Modify the cached value of X_recv
@@ -251,44 +212,41 @@ static void ccid3_hc_tx_no_feedback_timer(unsigned long data)
*
* Note that X_recv is scaled by 2^6 while X_calc is not
*/
- BUG_ON(hctx->ccid3hctx_p && !hctx->ccid3hctx_x_calc);
+ BUG_ON(hctx->p && !hctx->x_calc);
- if (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->x_calc > (hctx->x_recv >> 5))
+ hctx->x_recv /= 2;
else {
- hctx->ccid3hctx_x_recv = hctx->ccid3hctx_x_calc;
- hctx->ccid3hctx_x_recv <<= 4;
+ hctx->x_recv = hctx->x_calc;
+ hctx->x_recv <<= 4;
}
ccid3_hc_tx_update_x(sk, NULL);
}
ccid3_pr_debug("Reduced X to %llu/64 bytes/sec\n",
- (unsigned long long)hctx->ccid3hctx_x);
+ (unsigned long long)hctx->x);
/*
* Set new timeout for the nofeedback timer.
* See comments in packet_recv() regarding the value of t_RTO.
*/
- if (unlikely(hctx->ccid3hctx_t_rto == 0)) /* no feedback yet */
+ if (unlikely(hctx->t_rto == 0)) /* no feedback received yet */
t_nfb = TFRC_INITIAL_TIMEOUT;
else
- t_nfb = max(hctx->ccid3hctx_t_rto, 2 * hctx->ccid3hctx_t_ipi);
+ t_nfb = max(hctx->t_rto, 2 * hctx->t_ipi);
restart_timer:
- sk_reset_timer(sk, &hctx->ccid3hctx_no_feedback_timer,
+ sk_reset_timer(sk, &hctx->no_feedback_timer,
jiffies + usecs_to_jiffies(t_nfb));
out:
bh_unlock_sock(sk);
sock_put(sk);
}
-/*
- * returns
- * > 0: delay (in msecs) that should pass before actually sending
- * = 0: can send immediately
- * < 0: error condition; do not send packet
+/**
+ * ccid3_hc_tx_send_packet - Delay-based dequeueing of TX packets
+ * @skb: next packet candidate to send on @sk
+ * This function uses the convention of ccid_packet_dequeue_eval() and
+ * returns a millisecond-delay value between 0 and t_mbi = 64000 msec.
*/
static int ccid3_hc_tx_send_packet(struct sock *sk, struct sk_buff *skb)
{
@@ -305,18 +263,14 @@ static int ccid3_hc_tx_send_packet(struct sock *sk, struct sk_buff *skb)
if (unlikely(skb->len == 0))
return -EBADMSG;
- switch (hctx->ccid3hctx_state) {
- case TFRC_SSTATE_NO_SENT:
- sk_reset_timer(sk, &hctx->ccid3hctx_no_feedback_timer,
- (jiffies +
+ if (hctx->s == 0) {
+ sk_reset_timer(sk, &hctx->no_feedback_timer, (jiffies +
usecs_to_jiffies(TFRC_INITIAL_TIMEOUT)));
- hctx->ccid3hctx_last_win_count = 0;
- hctx->ccid3hctx_t_last_win_count = now;
+ hctx->last_win_count = 0;
+ hctx->t_last_win_count = now;
/* Set t_0 for initial packet */
- hctx->ccid3hctx_t_nom = now;
-
- hctx->ccid3hctx_s = skb->len;
+ hctx->t_nom = now;
/*
* Use initial RTT sample when available: recommended by erratum
@@ -325,9 +279,9 @@ static int ccid3_hc_tx_send_packet(struct sock *sk, struct sk_buff *skb)
*/
if (dp->dccps_syn_rtt) {
ccid3_pr_debug("SYN RTT = %uus\n", dp->dccps_syn_rtt);
- hctx->ccid3hctx_rtt = dp->dccps_syn_rtt;
- hctx->ccid3hctx_x = rfc3390_initial_rate(sk);
- hctx->ccid3hctx_t_ld = now;
+ hctx->rtt = dp->dccps_syn_rtt;
+ hctx->x = rfc3390_initial_rate(sk);
+ hctx->t_ld = now;
} else {
/*
* Sender does not have RTT sample:
@@ -335,17 +289,20 @@ static int ccid3_hc_tx_send_packet(struct sock *sk, struct sk_buff *skb)
* is needed in several parts (e.g. window counter);
* - set sending rate X_pps = 1pps as per RFC 3448, 4.2.
*/
- hctx->ccid3hctx_rtt = DCCP_FALLBACK_RTT;
- hctx->ccid3hctx_x = hctx->ccid3hctx_s;
- hctx->ccid3hctx_x <<= 6;
+ hctx->rtt = DCCP_FALLBACK_RTT;
+ hctx->x = dp->dccps_mss_cache;
+ hctx->x <<= 6;
}
+
+ /* Compute t_ipi = s / X */
+ hctx->s = ccid3_hc_tx_measure_packet_size(sk, skb->len);
ccid3_update_send_interval(hctx);
- ccid3_hc_tx_set_state(sk, TFRC_SSTATE_NO_FBACK);
- break;
- case TFRC_SSTATE_NO_FBACK:
- case TFRC_SSTATE_FBACK:
- delay = ktime_us_delta(hctx->ccid3hctx_t_nom, now);
+ /* Seed value for Oscillation Prevention (sec. 4.5) */
+ hctx->r_sqmean = tfrc_scaled_sqrt(hctx->rtt);
+
+ } else {
+ delay = ktime_us_delta(hctx->t_nom, now);
ccid3_pr_debug("delay=%ld\n", (long)delay);
/*
* Scheduling of packet transmissions [RFC 3448, 4.6]
@@ -355,99 +312,80 @@ static int ccid3_hc_tx_send_packet(struct sock *sk, struct sk_buff *skb)
* else
* // send the packet in (t_nom - t_now) milliseconds.
*/
- if (delay - (s64)hctx->ccid3hctx_delta >= 1000)
- return (u32)delay / 1000L;
+ if (delay >= TFRC_T_DELTA)
+ return (u32)delay / USEC_PER_MSEC;
ccid3_hc_tx_update_win_count(hctx, now);
- break;
- case TFRC_SSTATE_TERM:
- DCCP_BUG("%s(%p) - Illegal state TERM", dccp_role(sk), sk);
- return -EINVAL;
}
/* 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;
+ DCCP_SKB_CB(skb)->dccpd_ccval = hctx->last_win_count;
/* set the nominal send time for the next following packet */
- hctx->ccid3hctx_t_nom = ktime_add_us(hctx->ccid3hctx_t_nom,
- hctx->ccid3hctx_t_ipi);
- return 0;
+ hctx->t_nom = ktime_add_us(hctx->t_nom, hctx->t_ipi);
+ return CCID_PACKET_SEND_AT_ONCE;
}
-static void ccid3_hc_tx_packet_sent(struct sock *sk, int more,
- unsigned int len)
+static void ccid3_hc_tx_packet_sent(struct sock *sk, unsigned int len)
{
struct ccid3_hc_tx_sock *hctx = ccid3_hc_tx_sk(sk);
- ccid3_hc_tx_update_s(hctx, len);
+ /* Changes to s will become effective the next time X is computed */
+ hctx->s = ccid3_hc_tx_measure_packet_size(sk, len);
- if (tfrc_tx_hist_add(&hctx->ccid3hctx_hist, dccp_sk(sk)->dccps_gss))
+ if (tfrc_tx_hist_add(&hctx->hist, dccp_sk(sk)->dccps_gss))
DCCP_CRIT("packet history - out of memory!");
}
static void ccid3_hc_tx_packet_recv(struct sock *sk, struct sk_buff *skb)
{
struct ccid3_hc_tx_sock *hctx = ccid3_hc_tx_sk(sk);
- struct ccid3_options_received *opt_recv;
+ struct tfrc_tx_hist_entry *acked;
ktime_t now;
unsigned long t_nfb;
- u32 pinv, r_sample;
+ u32 r_sample;
/* we are only interested in ACKs */
if (!(DCCP_SKB_CB(skb)->dccpd_type == DCCP_PKT_ACK ||
DCCP_SKB_CB(skb)->dccpd_type == DCCP_PKT_DATAACK))
return;
- /* ... and only in the established state */
- if (hctx->ccid3hctx_state != TFRC_SSTATE_FBACK &&
- hctx->ccid3hctx_state != TFRC_SSTATE_NO_FBACK)
- return;
-
- opt_recv = &hctx->ccid3hctx_options_received;
- now = ktime_get_real();
-
- /* Estimate RTT from history if ACK number is valid */
- r_sample = tfrc_tx_hist_rtt(hctx->ccid3hctx_hist,
- DCCP_SKB_CB(skb)->dccpd_ack_seq, now);
- if (r_sample == 0) {
- DCCP_WARN("%s(%p): %s with bogus ACK-%llu\n", dccp_role(sk), sk,
- dccp_packet_name(DCCP_SKB_CB(skb)->dccpd_type),
- (unsigned long long)DCCP_SKB_CB(skb)->dccpd_ack_seq);
+ /*
+ * Locate the acknowledged packet in the TX history.
+ *
+ * Returning "entry not found" here can for instance happen when
+ * - the host has not sent out anything (e.g. a passive server),
+ * - the Ack is outdated (packet with higher Ack number was received),
+ * - it is a bogus Ack (for a packet not sent on this connection).
+ */
+ acked = tfrc_tx_hist_find_entry(hctx->hist, dccp_hdr_ack_seq(skb));
+ if (acked == NULL)
return;
- }
+ /* For the sake of RTT sampling, ignore/remove all older entries */
+ tfrc_tx_hist_purge(&acked->next);
- /* Update receive rate in units of 64 * bytes/second */
- hctx->ccid3hctx_x_recv = opt_recv->ccid3or_receive_rate;
- hctx->ccid3hctx_x_recv <<= 6;
+ /* Update the moving average for the RTT estimate (RFC 3448, 4.3) */
+ now = ktime_get_real();
+ r_sample = dccp_sample_rtt(sk, ktime_us_delta(now, acked->stamp));
+ hctx->rtt = tfrc_ewma(hctx->rtt, r_sample, 9);
- /* Update loss event rate (which is scaled by 1e6) */
- pinv = opt_recv->ccid3or_loss_event_rate;
- if (pinv == ~0U || pinv == 0) /* see RFC 4342, 8.5 */
- hctx->ccid3hctx_p = 0;
- else /* can not exceed 100% */
- hctx->ccid3hctx_p = scaled_div(1, pinv);
- /*
- * Validate new RTT sample and update moving average
- */
- r_sample = dccp_sample_rtt(sk, r_sample);
- hctx->ccid3hctx_rtt = tfrc_ewma(hctx->ccid3hctx_rtt, r_sample, 9);
/*
* Update allowed sending rate X as per draft rfc3448bis-00, 4.2/3
*/
- if (hctx->ccid3hctx_state == TFRC_SSTATE_NO_FBACK) {
- ccid3_hc_tx_set_state(sk, TFRC_SSTATE_FBACK);
+ if (!hctx->feedback) {
+ hctx->feedback = true;
- if (hctx->ccid3hctx_t_rto == 0) {
+ if (hctx->t_rto == 0) {
/*
* Initial feedback packet: Larger Initial Windows (4.2)
*/
- hctx->ccid3hctx_x = rfc3390_initial_rate(sk);
- hctx->ccid3hctx_t_ld = now;
+ hctx->x = rfc3390_initial_rate(sk);
+ hctx->t_ld = now;
ccid3_update_send_interval(hctx);
goto done_computing_x;
- } else if (hctx->ccid3hctx_p == 0) {
+ } else if (hctx->p == 0) {
/*
* First feedback after nofeedback timer expiry (4.3)
*/
@@ -456,25 +394,52 @@ static void ccid3_hc_tx_packet_recv(struct sock *sk, struct sk_buff *skb)
}
/* Update sending rate (step 4 of [RFC 3448, 4.3]) */
- if (hctx->ccid3hctx_p > 0)
- hctx->ccid3hctx_x_calc =
- tfrc_calc_x(hctx->ccid3hctx_s,
- hctx->ccid3hctx_rtt,
- hctx->ccid3hctx_p);
+ if (hctx->p > 0)
+ hctx->x_calc = tfrc_calc_x(hctx->s, hctx->rtt, hctx->p);
ccid3_hc_tx_update_x(sk, &now);
done_computing_x:
ccid3_pr_debug("%s(%p), RTT=%uus (sample=%uus), s=%u, "
"p=%u, X_calc=%u, X_recv=%u, X=%u\n",
- dccp_role(sk),
- sk, hctx->ccid3hctx_rtt, r_sample,
- hctx->ccid3hctx_s, hctx->ccid3hctx_p,
- hctx->ccid3hctx_x_calc,
- (unsigned)(hctx->ccid3hctx_x_recv >> 6),
- (unsigned)(hctx->ccid3hctx_x >> 6));
+ dccp_role(sk), sk, hctx->rtt, r_sample,
+ hctx->s, hctx->p, hctx->x_calc,
+ (unsigned)(hctx->x_recv >> 6),
+ (unsigned)(hctx->x >> 6));
+ /*
+ * Oscillation Reduction (RFC 3448, 4.5) - modifying t_ipi according to
+ * RTT changes, multiplying by X/X_inst = sqrt(R_sample)/R_sqmean. This
+ * can be useful if few connections share a link, avoiding that buffer
+ * fill levels (RTT) oscillate as a result of frequent adjustments to X.
+ * A useful presentation with background information is in
+ * Joerg Widmer, "Equation-Based Congestion Control",
+ * MSc Thesis, University of Mannheim, Germany, 2000
+ * (sec. 3.6.4), who calls this ISM ("Inter-packet Space Modulation").
+ */
+ if (do_osc_prev) {
+ r_sample = tfrc_scaled_sqrt(r_sample);
+ /*
+ * The modulation can work in both ways: increase/decrease t_ipi
+ * according to long-term increases/decreases of the RTT. The
+ * former is a useful measure, since it works against queue
+ * build-up. The latter temporarily increases the sending rate,
+ * so that buffers fill up more quickly. This in turn causes
+ * the RTT to increase, so that either later reduction becomes
+ * necessary or the RTT stays at a very high level. Decreasing
+ * t_ipi is therefore not supported.
+ * Furthermore, during the initial slow-start phase the RTT
+ * naturally increases, where using the algorithm would cause
+ * delays. Hence it is disabled during the initial slow-start.
+ */
+ if (r_sample > hctx->r_sqmean && hctx->p > 0)
+ hctx->t_ipi = div_u64((u64)hctx->t_ipi * (u64)r_sample,
+ hctx->r_sqmean);
+ hctx->t_ipi = min_t(u32, hctx->t_ipi, TFRC_T_MBI);
+ /* update R_sqmean _after_ computing the modulation factor */
+ hctx->r_sqmean = tfrc_ewma(hctx->r_sqmean, r_sample, 9);
+ }
/* unschedule no feedback timer */
- sk_stop_timer(sk, &hctx->ccid3hctx_no_feedback_timer);
+ sk_stop_timer(sk, &hctx->no_feedback_timer);
/*
* As we have calculated new ipi, delta, t_nom it is possible
@@ -488,95 +453,66 @@ done_computing_x:
* This can help avoid triggering the nofeedback timer too
* often ('spinning') on LANs with small RTTs.
*/
- hctx->ccid3hctx_t_rto = max_t(u32, 4 * hctx->ccid3hctx_rtt,
- (CONFIG_IP_DCCP_CCID3_RTO *
- (USEC_PER_SEC / 1000)));
+ hctx->t_rto = max_t(u32, 4 * hctx->rtt, (CONFIG_IP_DCCP_CCID3_RTO *
+ (USEC_PER_SEC / 1000)));
/*
* Schedule no feedback timer to expire in
* max(t_RTO, 2 * s/X) = max(t_RTO, 2 * t_ipi)
*/
- t_nfb = max(hctx->ccid3hctx_t_rto, 2 * hctx->ccid3hctx_t_ipi);
+ t_nfb = max(hctx->t_rto, 2 * hctx->t_ipi);
ccid3_pr_debug("%s(%p), Scheduled no feedback timer to "
"expire in %lu jiffies (%luus)\n",
- dccp_role(sk),
- sk, usecs_to_jiffies(t_nfb), t_nfb);
+ dccp_role(sk), sk, usecs_to_jiffies(t_nfb), t_nfb);
- sk_reset_timer(sk, &hctx->ccid3hctx_no_feedback_timer,
+ sk_reset_timer(sk, &hctx->no_feedback_timer,
jiffies + usecs_to_jiffies(t_nfb));
}
-static int ccid3_hc_tx_parse_options(struct sock *sk, unsigned char option,
- unsigned char len, u16 idx,
- unsigned char *value)
+static int ccid3_hc_tx_parse_options(struct sock *sk, u8 packet_type,
+ u8 option, u8 *optval, u8 optlen)
{
- int rc = 0;
- const struct dccp_sock *dp = dccp_sk(sk);
struct ccid3_hc_tx_sock *hctx = ccid3_hc_tx_sk(sk);
- struct ccid3_options_received *opt_recv;
__be32 opt_val;
- opt_recv = &hctx->ccid3hctx_options_received;
-
- if (opt_recv->ccid3or_seqno != dp->dccps_gsr) {
- opt_recv->ccid3or_seqno = dp->dccps_gsr;
- opt_recv->ccid3or_loss_event_rate = ~0;
- opt_recv->ccid3or_loss_intervals_idx = 0;
- opt_recv->ccid3or_loss_intervals_len = 0;
- opt_recv->ccid3or_receive_rate = 0;
- }
-
switch (option) {
+ case TFRC_OPT_RECEIVE_RATE:
case TFRC_OPT_LOSS_EVENT_RATE:
- if (unlikely(len != 4)) {
- DCCP_WARN("%s(%p), invalid len %d "
- "for TFRC_OPT_LOSS_EVENT_RATE\n",
- dccp_role(sk), sk, len);
- rc = -EINVAL;
- } else {
- opt_val = get_unaligned((__be32 *)value);
- opt_recv->ccid3or_loss_event_rate = ntohl(opt_val);
- ccid3_pr_debug("%s(%p), LOSS_EVENT_RATE=%u\n",
- dccp_role(sk), sk,
- opt_recv->ccid3or_loss_event_rate);
+ /* Must be ignored on Data packets, cf. RFC 4342 8.3 and 8.5 */
+ if (packet_type == DCCP_PKT_DATA)
+ break;
+ if (unlikely(optlen != 4)) {
+ DCCP_WARN("%s(%p), invalid len %d for %u\n",
+ dccp_role(sk), sk, optlen, option);
+ return -EINVAL;
}
- break;
- case TFRC_OPT_LOSS_INTERVALS:
- opt_recv->ccid3or_loss_intervals_idx = idx;
- opt_recv->ccid3or_loss_intervals_len = len;
- ccid3_pr_debug("%s(%p), LOSS_INTERVALS=(%u, %u)\n",
- dccp_role(sk), sk,
- opt_recv->ccid3or_loss_intervals_idx,
- opt_recv->ccid3or_loss_intervals_len);
- break;
- case TFRC_OPT_RECEIVE_RATE:
- if (unlikely(len != 4)) {
- DCCP_WARN("%s(%p), invalid len %d "
- "for TFRC_OPT_RECEIVE_RATE\n",
- dccp_role(sk), sk, len);
- rc = -EINVAL;
- } else {
- opt_val = get_unaligned((__be32 *)value);
- opt_recv->ccid3or_receive_rate = ntohl(opt_val);
+ opt_val = ntohl(get_unaligned((__be32 *)optval));
+
+ if (option == TFRC_OPT_RECEIVE_RATE) {
+ /* Receive Rate is kept in units of 64 bytes/second */
+ hctx->x_recv = opt_val;
+ hctx->x_recv <<= 6;
+
ccid3_pr_debug("%s(%p), RECEIVE_RATE=%u\n",
- dccp_role(sk), sk,
- opt_recv->ccid3or_receive_rate);
+ dccp_role(sk), sk, opt_val);
+ } else {
+ /* Update the fixpoint Loss Event Rate fraction */
+ hctx->p = tfrc_invert_loss_event_rate(opt_val);
+
+ ccid3_pr_debug("%s(%p), LOSS_EVENT_RATE=%u\n",
+ dccp_role(sk), sk, opt_val);
}
- break;
}
-
- return rc;
+ return 0;
}
static int ccid3_hc_tx_init(struct ccid *ccid, struct sock *sk)
{
struct ccid3_hc_tx_sock *hctx = ccid_priv(ccid);
- hctx->ccid3hctx_state = TFRC_SSTATE_NO_SENT;
- hctx->ccid3hctx_hist = NULL;
- setup_timer(&hctx->ccid3hctx_no_feedback_timer,
- ccid3_hc_tx_no_feedback_timer, (unsigned long)sk);
-
+ hctx->hist = NULL;
+ setup_timer(&hctx->no_feedback_timer,
+ ccid3_hc_tx_no_feedback_timer, (unsigned long)sk);
return 0;
}
@@ -584,42 +520,36 @@ static void ccid3_hc_tx_exit(struct sock *sk)
{
struct ccid3_hc_tx_sock *hctx = ccid3_hc_tx_sk(sk);
- ccid3_hc_tx_set_state(sk, TFRC_SSTATE_TERM);
- sk_stop_timer(sk, &hctx->ccid3hctx_no_feedback_timer);
-
- tfrc_tx_hist_purge(&hctx->ccid3hctx_hist);
+ sk_stop_timer(sk, &hctx->no_feedback_timer);
+ tfrc_tx_hist_purge(&hctx->hist);
}
static void ccid3_hc_tx_get_info(struct sock *sk, struct tcp_info *info)
{
- struct ccid3_hc_tx_sock *hctx;
-
- /* Listen socks doesn't have a private CCID block */
- if (sk->sk_state == DCCP_LISTEN)
- return;
-
- hctx = ccid3_hc_tx_sk(sk);
- info->tcpi_rto = hctx->ccid3hctx_t_rto;
- info->tcpi_rtt = hctx->ccid3hctx_rtt;
+ info->tcpi_rto = ccid3_hc_tx_sk(sk)->t_rto;
+ info->tcpi_rtt = ccid3_hc_tx_sk(sk)->rtt;
}
static int ccid3_hc_tx_getsockopt(struct sock *sk, const int optname, int len,
u32 __user *optval, int __user *optlen)
{
- const struct ccid3_hc_tx_sock *hctx;
+ const struct ccid3_hc_tx_sock *hctx = ccid3_hc_tx_sk(sk);
+ struct tfrc_tx_info tfrc;
const void *val;
- /* Listen socks doesn't have a private CCID block */
- if (sk->sk_state == DCCP_LISTEN)
- return -EINVAL;
-
- hctx = ccid3_hc_tx_sk(sk);
switch (optname) {
case DCCP_SOCKOPT_CCID_TX_INFO:
- if (len < sizeof(hctx->ccid3hctx_tfrc))
+ if (len < sizeof(tfrc))
return -EINVAL;
- len = sizeof(hctx->ccid3hctx_tfrc);
- val = &hctx->ccid3hctx_tfrc;
+ tfrc.tfrctx_x = hctx->x;
+ tfrc.tfrctx_x_recv = hctx->x_recv;
+ tfrc.tfrctx_x_calc = hctx->x_calc;
+ tfrc.tfrctx_rtt = hctx->rtt;
+ tfrc.tfrctx_p = hctx->p;
+ tfrc.tfrctx_rto = hctx->t_rto;
+ tfrc.tfrctx_ipi = hctx->t_ipi;
+ len = sizeof(tfrc);
+ val = &tfrc;
break;
default:
return -ENOPROTOOPT;
@@ -634,112 +564,82 @@ static int ccid3_hc_tx_getsockopt(struct sock *sk, const int optname, int len,
/*
* Receiver Half-Connection Routines
*/
-
-/* CCID3 feedback types */
-enum ccid3_fback_type {
- CCID3_FBACK_NONE = 0,
- CCID3_FBACK_INITIAL,
- CCID3_FBACK_PERIODIC,
- CCID3_FBACK_PARAM_CHANGE
-};
-
-#ifdef CONFIG_IP_DCCP_CCID3_DEBUG
-static const char *ccid3_rx_state_name(enum ccid3_hc_rx_states state)
-{
- static char *ccid3_rx_state_names[] = {
- [TFRC_RSTATE_NO_DATA] = "NO_DATA",
- [TFRC_RSTATE_DATA] = "DATA",
- [TFRC_RSTATE_TERM] = "TERM",
- };
-
- return ccid3_rx_state_names[state];
-}
-#endif
-
-static void ccid3_hc_rx_set_state(struct sock *sk,
- enum ccid3_hc_rx_states state)
-{
- struct ccid3_hc_rx_sock *hcrx = ccid3_hc_rx_sk(sk);
- enum ccid3_hc_rx_states oldstate = hcrx->ccid3hcrx_state;
-
- ccid3_pr_debug("%s(%p) %-8.8s -> %s\n",
- dccp_role(sk), sk, ccid3_rx_state_name(oldstate),
- ccid3_rx_state_name(state));
- WARN_ON(state == oldstate);
- hcrx->ccid3hcrx_state = state;
-}
-
static void ccid3_hc_rx_send_feedback(struct sock *sk,
const struct sk_buff *skb,
enum ccid3_fback_type fbtype)
{
struct ccid3_hc_rx_sock *hcrx = ccid3_hc_rx_sk(sk);
- struct dccp_sock *dp = dccp_sk(sk);
- ktime_t now;
- s64 delta = 0;
-
- if (unlikely(hcrx->ccid3hcrx_state == TFRC_RSTATE_TERM))
- return;
-
- now = ktime_get_real();
switch (fbtype) {
case CCID3_FBACK_INITIAL:
- hcrx->ccid3hcrx_x_recv = 0;
- hcrx->ccid3hcrx_pinv = ~0U; /* see RFC 4342, 8.5 */
+ hcrx->x_recv = 0;
+ hcrx->p_inverse = ~0U; /* see RFC 4342, 8.5 */
break;
case CCID3_FBACK_PARAM_CHANGE:
+ if (unlikely(hcrx->feedback == CCID3_FBACK_NONE)) {
+ /*
+ * rfc3448bis-06, 6.3.1: First packet(s) lost or marked
+ * FIXME: in rfc3448bis the receiver returns X_recv=0
+ * here as it normally would in the first feedback packet.
+ * However this is not possible yet, since the code still
+ * uses RFC 3448, i.e.
+ * If (p > 0)
+ * Calculate X_calc using the TCP throughput equation.
+ * X = max(min(X_calc, 2*X_recv), s/t_mbi);
+ * would bring X down to s/t_mbi. That is why we return
+ * X_recv according to rfc3448bis-06 for the moment.
+ */
+ u32 s = tfrc_rx_hist_packet_size(&hcrx->hist),
+ rtt = tfrc_rx_hist_rtt(&hcrx->hist);
+
+ hcrx->x_recv = scaled_div32(s, 2 * rtt);
+ break;
+ }
/*
* When parameters change (new loss or p > p_prev), we do not
* have a reliable estimate for R_m of [RFC 3448, 6.2] and so
- * need to reuse the previous value of X_recv. However, when
- * X_recv was 0 (due to early loss), this would kill X down to
- * s/t_mbi (i.e. one packet in 64 seconds).
- * To avoid such drastic reduction, we approximate X_recv as
- * the number of bytes since last feedback.
- * This is a safe fallback, since X is bounded above by X_calc.
+ * always check whether at least RTT time units were covered.
*/
- if (hcrx->ccid3hcrx_x_recv > 0)
- break;
- /* fall through */
+ hcrx->x_recv = tfrc_rx_hist_x_recv(&hcrx->hist, hcrx->x_recv);
+ break;
case CCID3_FBACK_PERIODIC:
- delta = ktime_us_delta(now, hcrx->ccid3hcrx_tstamp_last_feedback);
- if (delta <= 0)
- DCCP_BUG("delta (%ld) <= 0", (long)delta);
- else
- hcrx->ccid3hcrx_x_recv =
- scaled_div32(hcrx->ccid3hcrx_bytes_recv, delta);
+ /*
+ * Step (2) of rfc3448bis-06, 6.2:
+ * - if no data packets have been received, just restart timer
+ * - if data packets have been received, re-compute X_recv
+ */
+ if (hcrx->hist.bytes_recvd == 0)
+ goto prepare_for_next_time;
+ hcrx->x_recv = tfrc_rx_hist_x_recv(&hcrx->hist, hcrx->x_recv);
break;
default:
return;
}
- ccid3_pr_debug("Interval %ldusec, X_recv=%u, 1/p=%u\n", (long)delta,
- hcrx->ccid3hcrx_x_recv, hcrx->ccid3hcrx_pinv);
-
- hcrx->ccid3hcrx_tstamp_last_feedback = now;
- hcrx->ccid3hcrx_last_counter = dccp_hdr(skb)->dccph_ccval;
- hcrx->ccid3hcrx_bytes_recv = 0;
+ ccid3_pr_debug("X_recv=%u, 1/p=%u\n", hcrx->x_recv, hcrx->p_inverse);
- dp->dccps_hc_rx_insert_options = 1;
+ dccp_sk(sk)->dccps_hc_rx_insert_options = 1;
dccp_send_ack(sk);
+
+prepare_for_next_time:
+ tfrc_rx_hist_restart_byte_counter(&hcrx->hist);
+ hcrx->last_counter = dccp_hdr(skb)->dccph_ccval;
+ hcrx->feedback = fbtype;
}
static int ccid3_hc_rx_insert_options(struct sock *sk, struct sk_buff *skb)
{
- const struct ccid3_hc_rx_sock *hcrx;
+ const struct ccid3_hc_rx_sock *hcrx = ccid3_hc_rx_sk(sk);
__be32 x_recv, pinv;
if (!(sk->sk_state == DCCP_OPEN || sk->sk_state == DCCP_PARTOPEN))
return 0;
- hcrx = ccid3_hc_rx_sk(sk);
-
if (dccp_packet_without_ack(skb))
return 0;
- x_recv = htonl(hcrx->ccid3hcrx_x_recv);
- pinv = htonl(hcrx->ccid3hcrx_pinv);
+ x_recv = htonl(hcrx->x_recv);
+ pinv = htonl(hcrx->p_inverse);
if (dccp_insert_option(sk, skb, TFRC_OPT_LOSS_EVENT_RATE,
&pinv, sizeof(pinv)) ||
@@ -762,171 +662,95 @@ static int ccid3_hc_rx_insert_options(struct sock *sk, struct sk_buff *skb)
static u32 ccid3_first_li(struct sock *sk)
{
struct ccid3_hc_rx_sock *hcrx = ccid3_hc_rx_sk(sk);
- u32 x_recv, p, delta;
+ u32 s = tfrc_rx_hist_packet_size(&hcrx->hist),
+ rtt = tfrc_rx_hist_rtt(&hcrx->hist), x_recv, p;
u64 fval;
- if (hcrx->ccid3hcrx_rtt == 0) {
- DCCP_WARN("No RTT estimate available, using fallback RTT\n");
- hcrx->ccid3hcrx_rtt = DCCP_FALLBACK_RTT;
- }
+ /*
+ * rfc3448bis-06, 6.3.1: First data packet(s) are marked or lost. Set p
+ * to give the equivalent of X_target = s/(2*R). Thus fval = 2 and so p
+ * is about 20.64%. This yields an interval length of 4.84 (rounded up).
+ */
+ if (unlikely(hcrx->feedback == CCID3_FBACK_NONE))
+ return 5;
- delta = ktime_to_us(net_timedelta(hcrx->ccid3hcrx_tstamp_last_feedback));
- x_recv = scaled_div32(hcrx->ccid3hcrx_bytes_recv, delta);
- if (x_recv == 0) { /* would also trigger divide-by-zero */
- DCCP_WARN("X_recv==0\n");
- if ((x_recv = hcrx->ccid3hcrx_x_recv) == 0) {
- DCCP_BUG("stored value of X_recv is zero");
- return ~0U;
- }
- }
+ x_recv = tfrc_rx_hist_x_recv(&hcrx->hist, hcrx->x_recv);
+ if (x_recv == 0)
+ goto failed;
- fval = scaled_div(hcrx->ccid3hcrx_s, hcrx->ccid3hcrx_rtt);
- fval = scaled_div32(fval, x_recv);
+ fval = scaled_div32(scaled_div(s, rtt), x_recv);
p = tfrc_calc_x_reverse_lookup(fval);
ccid3_pr_debug("%s(%p), receive rate=%u bytes/s, implied "
"loss rate=%u\n", dccp_role(sk), sk, x_recv, p);
- return p == 0 ? ~0U : scaled_div(1, p);
+ if (p > 0)
+ return scaled_div(1, p);
+failed:
+ return UINT_MAX;
}
static void ccid3_hc_rx_packet_recv(struct sock *sk, struct sk_buff *skb)
{
struct ccid3_hc_rx_sock *hcrx = ccid3_hc_rx_sk(sk);
- enum ccid3_fback_type do_feedback = CCID3_FBACK_NONE;
const u64 ndp = dccp_sk(sk)->dccps_options_received.dccpor_ndp;
const bool is_data_packet = dccp_data_packet(skb);
- if (unlikely(hcrx->ccid3hcrx_state == TFRC_RSTATE_NO_DATA)) {
- if (is_data_packet) {
- const u32 payload = skb->len - dccp_hdr(skb)->dccph_doff * 4;
- do_feedback = CCID3_FBACK_INITIAL;
- ccid3_hc_rx_set_state(sk, TFRC_RSTATE_DATA);
- hcrx->ccid3hcrx_s = payload;
- /*
- * Not necessary to update ccid3hcrx_bytes_recv here,
- * since X_recv = 0 for the first feedback packet (cf.
- * RFC 3448, 6.3) -- gerrit
- */
- }
- goto update_records;
- }
-
- if (tfrc_rx_hist_duplicate(&hcrx->ccid3hcrx_hist, skb))
- return; /* done receiving */
-
- if (is_data_packet) {
- const u32 payload = skb->len - dccp_hdr(skb)->dccph_doff * 4;
- /*
- * Update moving-average of s and the sum of received payload bytes
- */
- hcrx->ccid3hcrx_s = tfrc_ewma(hcrx->ccid3hcrx_s, payload, 9);
- hcrx->ccid3hcrx_bytes_recv += payload;
- }
-
/*
* Perform loss detection and handle pending losses
*/
- if (tfrc_rx_handle_loss(&hcrx->ccid3hcrx_hist, &hcrx->ccid3hcrx_li_hist,
- skb, ndp, ccid3_first_li, sk)) {
- do_feedback = CCID3_FBACK_PARAM_CHANGE;
- goto done_receiving;
- }
-
- if (tfrc_rx_hist_loss_pending(&hcrx->ccid3hcrx_hist))
- return; /* done receiving */
-
+ if (tfrc_rx_congestion_event(&hcrx->hist, &hcrx->li_hist,
+ skb, ndp, ccid3_first_li, sk))
+ ccid3_hc_rx_send_feedback(sk, skb, CCID3_FBACK_PARAM_CHANGE);
/*
- * Handle data packets: RTT sampling and monitoring p
+ * Feedback for first non-empty data packet (RFC 3448, 6.3)
*/
- if (unlikely(!is_data_packet))
- goto update_records;
-
- if (!tfrc_lh_is_initialised(&hcrx->ccid3hcrx_li_hist)) {
- const u32 sample = tfrc_rx_hist_sample_rtt(&hcrx->ccid3hcrx_hist, skb);
- /*
- * Empty loss history: no loss so far, hence p stays 0.
- * Sample RTT values, since an RTT estimate is required for the
- * computation of p when the first loss occurs; RFC 3448, 6.3.1.
- */
- if (sample != 0)
- hcrx->ccid3hcrx_rtt = tfrc_ewma(hcrx->ccid3hcrx_rtt, sample, 9);
-
- } else if (tfrc_lh_update_i_mean(&hcrx->ccid3hcrx_li_hist, skb)) {
- /*
- * Step (3) of [RFC 3448, 6.1]: Recompute I_mean and, if I_mean
- * has decreased (resp. p has increased), send feedback now.
- */
- do_feedback = CCID3_FBACK_PARAM_CHANGE;
- }
-
+ else if (unlikely(hcrx->feedback == CCID3_FBACK_NONE && is_data_packet))
+ ccid3_hc_rx_send_feedback(sk, skb, CCID3_FBACK_INITIAL);
/*
* Check if the periodic once-per-RTT feedback is due; RFC 4342, 10.3
*/
- if (SUB16(dccp_hdr(skb)->dccph_ccval, hcrx->ccid3hcrx_last_counter) > 3)
- do_feedback = CCID3_FBACK_PERIODIC;
-
-update_records:
- tfrc_rx_hist_add_packet(&hcrx->ccid3hcrx_hist, skb, ndp);
-
-done_receiving:
- if (do_feedback)
- ccid3_hc_rx_send_feedback(sk, skb, do_feedback);
+ else if (!tfrc_rx_hist_loss_pending(&hcrx->hist) && is_data_packet &&
+ SUB16(dccp_hdr(skb)->dccph_ccval, hcrx->last_counter) > 3)
+ ccid3_hc_rx_send_feedback(sk, skb, CCID3_FBACK_PERIODIC);
}
static int ccid3_hc_rx_init(struct ccid *ccid, struct sock *sk)
{
struct ccid3_hc_rx_sock *hcrx = ccid_priv(ccid);
- hcrx->ccid3hcrx_state = TFRC_RSTATE_NO_DATA;
- tfrc_lh_init(&hcrx->ccid3hcrx_li_hist);
- return tfrc_rx_hist_alloc(&hcrx->ccid3hcrx_hist);
+ tfrc_lh_init(&hcrx->li_hist);
+ return tfrc_rx_hist_init(&hcrx->hist, sk);
}
static void ccid3_hc_rx_exit(struct sock *sk)
{
struct ccid3_hc_rx_sock *hcrx = ccid3_hc_rx_sk(sk);
- ccid3_hc_rx_set_state(sk, TFRC_RSTATE_TERM);
-
- tfrc_rx_hist_purge(&hcrx->ccid3hcrx_hist);
- tfrc_lh_cleanup(&hcrx->ccid3hcrx_li_hist);
+ tfrc_rx_hist_purge(&hcrx->hist);
+ tfrc_lh_cleanup(&hcrx->li_hist);
}
static void ccid3_hc_rx_get_info(struct sock *sk, struct tcp_info *info)
{
- const struct ccid3_hc_rx_sock *hcrx;
-
- /* Listen socks doesn't have a private CCID block */
- if (sk->sk_state == DCCP_LISTEN)
- return;
-
- hcrx = ccid3_hc_rx_sk(sk);
- info->tcpi_ca_state = hcrx->ccid3hcrx_state;
info->tcpi_options |= TCPI_OPT_TIMESTAMPS;
- info->tcpi_rcv_rtt = hcrx->ccid3hcrx_rtt;
+ info->tcpi_rcv_rtt = tfrc_rx_hist_rtt(&ccid3_hc_rx_sk(sk)->hist);
}
static int ccid3_hc_rx_getsockopt(struct sock *sk, const int optname, int len,
u32 __user *optval, int __user *optlen)
{
- const struct ccid3_hc_rx_sock *hcrx;
+ const struct ccid3_hc_rx_sock *hcrx = ccid3_hc_rx_sk(sk);
struct tfrc_rx_info rx_info;
const void *val;
- /* Listen socks doesn't have a private CCID block */
- if (sk->sk_state == DCCP_LISTEN)
- return -EINVAL;
-
- hcrx = ccid3_hc_rx_sk(sk);
switch (optname) {
case DCCP_SOCKOPT_CCID_RX_INFO:
if (len < sizeof(rx_info))
return -EINVAL;
- rx_info.tfrcrx_x_recv = hcrx->ccid3hcrx_x_recv;
- rx_info.tfrcrx_rtt = hcrx->ccid3hcrx_rtt;
- rx_info.tfrcrx_p = hcrx->ccid3hcrx_pinv == 0 ? ~0U :
- scaled_div(1, hcrx->ccid3hcrx_pinv);
+ rx_info.tfrcrx_x_recv = hcrx->x_recv;
+ rx_info.tfrcrx_rtt = tfrc_rx_hist_rtt(&hcrx->hist);
+ rx_info.tfrcrx_p = tfrc_invert_loss_event_rate(hcrx->p_inverse);
len = sizeof(rx_info);
val = &rx_info;
break;
@@ -962,6 +786,9 @@ static struct ccid_operations ccid3 = {
.ccid_hc_tx_getsockopt = ccid3_hc_tx_getsockopt,
};
+module_param(do_osc_prev, bool, 0644);
+MODULE_PARM_DESC(do_osc_prev, "Use Oscillation Prevention (RFC 3448, 4.5)");
+
#ifdef CONFIG_IP_DCCP_CCID3_DEBUG
module_param(ccid3_debug, bool, 0644);
MODULE_PARM_DESC(ccid3_debug, "Enable debug messages");
@@ -969,6 +796,19 @@ MODULE_PARM_DESC(ccid3_debug, "Enable debug messages");
static __init int ccid3_module_init(void)
{
+ struct timespec tp;
+
+ /*
+ * Without a fine-grained clock resolution, RTTs/X_recv are not sampled
+ * correctly and feedback is sent either too early or too late.
+ */
+ hrtimer_get_res(CLOCK_MONOTONIC, &tp);
+ if (tp.tv_sec || tp.tv_nsec > DCCP_TIME_RESOLUTION * NSEC_PER_USEC) {
+ printk(KERN_ERR "%s: Timer too coarse (%ld usec), need %u-usec"
+ " resolution - check your clocksource.\n", __func__,
+ tp.tv_nsec/NSEC_PER_USEC, DCCP_TIME_RESOLUTION);
+ return -ESOCKTNOSUPPORT;
+ }
return ccid_register(&ccid3);
}
module_init(ccid3_module_init);
diff --git a/net/dccp/ccids/ccid3.h b/net/dccp/ccids/ccid3.h
index 49ca32bd7e7..af6e1bf937d 100644
--- a/net/dccp/ccids/ccid3.h
+++ b/net/dccp/ccids/ccid3.h
@@ -47,11 +47,22 @@
/* Two seconds as per RFC 3448 4.2 */
#define TFRC_INITIAL_TIMEOUT (2 * USEC_PER_SEC)
-/* In usecs - half the scheduling granularity as per RFC3448 4.6 */
-#define TFRC_OPSYS_HALF_TIME_GRAN (USEC_PER_SEC / (2 * HZ))
+/* Maximum backoff interval t_mbi (RFC 3448, 4.3) */
+#define TFRC_T_MBI (64 * USEC_PER_SEC)
-/* Parameter t_mbi from [RFC 3448, 4.3]: backoff interval in seconds */
-#define TFRC_T_MBI 64
+/*
+ * The t_delta parameter (RFC 3448, 4.6): delays of less than %USEC_PER_MSEC are
+ * rounded down to 0, since sk_reset_timer() here uses millisecond granularity.
+ * Hence we can use a constant t_delta = %USEC_PER_MSEC when HZ >= 500. A coarse
+ * resolution of HZ < 500 means that the error is below one timer tick (t_gran)
+ * when using the constant t_delta = t_gran / 2 = %USEC_PER_SEC / (2 * HZ).
+ */
+#if (HZ >= 500)
+# define TFRC_T_DELTA USEC_PER_MSEC
+#else
+# define TFRC_T_DELTA (USEC_PER_SEC / (2 * HZ))
+#warning Coarse CONFIG_HZ resolution -- higher value recommended for TFRC.
+#endif
enum ccid3_options {
TFRC_OPT_LOSS_EVENT_RATE = 192,
@@ -59,62 +70,43 @@ enum ccid3_options {
TFRC_OPT_RECEIVE_RATE = 194,
};
-struct ccid3_options_received {
- u64 ccid3or_seqno:48,
- ccid3or_loss_intervals_idx:16;
- u16 ccid3or_loss_intervals_len;
- u32 ccid3or_loss_event_rate;
- u32 ccid3or_receive_rate;
-};
-
-/* TFRC sender states */
-enum ccid3_hc_tx_states {
- TFRC_SSTATE_NO_SENT = 1,
- TFRC_SSTATE_NO_FBACK,
- TFRC_SSTATE_FBACK,
- TFRC_SSTATE_TERM,
-};
-
/** struct ccid3_hc_tx_sock - CCID3 sender half-connection socket
*
- * @ccid3hctx_x - Current sending rate in 64 * bytes per second
- * @ccid3hctx_x_recv - Receive rate in 64 * bytes per second
- * @ccid3hctx_x_calc - Calculated rate in bytes per second
- * @ccid3hctx_rtt - Estimate of current round trip time in usecs
- * @ccid3hctx_p - Current loss event rate (0-1) scaled by 1000000
- * @ccid3hctx_s - Packet size in bytes
- * @ccid3hctx_t_rto - Nofeedback Timer setting in usecs
- * @ccid3hctx_t_ipi - Interpacket (send) interval (RFC 3448, 4.6) in usecs
- * @ccid3hctx_state - Sender state, one of %ccid3_hc_tx_states
- * @ccid3hctx_last_win_count - Last window counter sent
- * @ccid3hctx_t_last_win_count - Timestamp of earliest packet
- * with last_win_count value sent
- * @ccid3hctx_no_feedback_timer - Handle to no feedback timer
- * @ccid3hctx_t_ld - Time last doubled during slow start
- * @ccid3hctx_t_nom - Nominal send time of next packet
- * @ccid3hctx_delta - Send timer delta (RFC 3448, 4.6) in usecs
- * @ccid3hctx_hist - Packet history
- * @ccid3hctx_options_received - Parsed set of retrieved options
+ * @x - Current sending rate in 64 * bytes per second
+ * @x_recv - Receive rate in 64 * bytes per second
+ * @x_calc - Calculated rate in bytes per second
+ * @rtt - Estimate of current round trip time in usecs
+ * @r_sqmean - Estimate of long-term RTT (RFC 3448, 4.5)
+ * @p - Current loss event rate (0-1) scaled by 1000000
+ * @s - Packet size in bytes
+ * @t_rto - Nofeedback Timer setting in usecs
+ * @t_ipi - Interpacket (send) interval (RFC 3448, 4.6) in usecs
+ * @feedback - Whether feedback has been received or not
+ * @last_win_count - Last window counter sent
+ * @t_last_win_count - Timestamp of earliest packet with
+ * last_win_count value sent
+ * @no_feedback_timer - Handle to no feedback timer
+ * @t_ld - Time last doubled during slow start
+ * @t_nom - Nominal send time of next packet
+ * @hist - Packet history
*/
struct ccid3_hc_tx_sock {
- struct tfrc_tx_info ccid3hctx_tfrc;
-#define ccid3hctx_x ccid3hctx_tfrc.tfrctx_x
-#define ccid3hctx_x_recv ccid3hctx_tfrc.tfrctx_x_recv
-#define ccid3hctx_x_calc ccid3hctx_tfrc.tfrctx_x_calc
-#define ccid3hctx_rtt ccid3hctx_tfrc.tfrctx_rtt
-#define ccid3hctx_p ccid3hctx_tfrc.tfrctx_p
-#define ccid3hctx_t_rto ccid3hctx_tfrc.tfrctx_rto
-#define ccid3hctx_t_ipi ccid3hctx_tfrc.tfrctx_ipi
- u16 ccid3hctx_s;
- enum ccid3_hc_tx_states ccid3hctx_state:8;
- u8 ccid3hctx_last_win_count;
- ktime_t ccid3hctx_t_last_win_count;
- struct timer_list ccid3hctx_no_feedback_timer;
- ktime_t ccid3hctx_t_ld;
- ktime_t ccid3hctx_t_nom;
- u32 ccid3hctx_delta;
- struct tfrc_tx_hist_entry *ccid3hctx_hist;
- struct ccid3_options_received ccid3hctx_options_received;
+ u64 x;
+ u64 x_recv;
+ u32 x_calc;
+ u32 rtt;
+ u16 r_sqmean;
+ u32 p;
+ u32 t_rto;
+ u32 t_ipi;
+ u16 s;
+ bool feedback:1;
+ u8 last_win_count;
+ ktime_t t_last_win_count;
+ struct timer_list no_feedback_timer;
+ ktime_t t_ld;
+ ktime_t t_nom;
+ struct tfrc_tx_hist_entry *hist;
};
static inline struct ccid3_hc_tx_sock *ccid3_hc_tx_sk(const struct sock *sk)
@@ -124,41 +116,32 @@ static inline struct ccid3_hc_tx_sock *ccid3_hc_tx_sk(const struct sock *sk)
return hctx;
}
-/* TFRC receiver states */
-enum ccid3_hc_rx_states {
- TFRC_RSTATE_NO_DATA = 1,
- TFRC_RSTATE_DATA,
- TFRC_RSTATE_TERM = 127,
+
+enum ccid3_fback_type {
+ CCID3_FBACK_NONE = 0,
+ CCID3_FBACK_INITIAL,
+ CCID3_FBACK_PERIODIC,
+ CCID3_FBACK_PARAM_CHANGE
};
/** struct ccid3_hc_rx_sock - CCID3 receiver half-connection socket
*
- * @ccid3hcrx_x_recv - Receiver estimate of send rate (RFC 3448 4.3)
- * @ccid3hcrx_rtt - Receiver estimate of rtt (non-standard)
- * @ccid3hcrx_p - Current loss event rate (RFC 3448 5.4)
- * @ccid3hcrx_last_counter - Tracks window counter (RFC 4342, 8.1)
- * @ccid3hcrx_state - Receiver state, one of %ccid3_hc_rx_states
- * @ccid3hcrx_bytes_recv - Total sum of DCCP payload bytes
- * @ccid3hcrx_x_recv - Receiver estimate of send rate (RFC 3448, sec. 4.3)
- * @ccid3hcrx_rtt - Receiver estimate of RTT
- * @ccid3hcrx_tstamp_last_feedback - Time at which last feedback was sent
- * @ccid3hcrx_tstamp_last_ack - Time at which last feedback was sent
- * @ccid3hcrx_hist - Packet history (loss detection + RTT sampling)
- * @ccid3hcrx_li_hist - Loss Interval database
- * @ccid3hcrx_s - Received packet size in bytes
- * @ccid3hcrx_pinv - Inverse of Loss Event Rate (RFC 4342, sec. 8.5)
+ * @last_counter - Tracks window counter (RFC 4342, 8.1)
+ * @feedback - The type of the feedback last sent
+ * @x_recv - Receiver estimate of send rate (RFC 3448, sec. 4.3)
+ * @tstamp_last_feedback - Time at which last feedback was sent
+ * @hist - Packet history (loss detection + RTT sampling)
+ * @li_hist - Loss Interval database
+ * @p_inverse - Inverse of Loss Event Rate (RFC 4342, sec. 8.5)
*/
struct ccid3_hc_rx_sock {
- u8 ccid3hcrx_last_counter:4;
- enum ccid3_hc_rx_states ccid3hcrx_state:8;
- u32 ccid3hcrx_bytes_recv;
- u32 ccid3hcrx_x_recv;
- u32 ccid3hcrx_rtt;
- ktime_t ccid3hcrx_tstamp_last_feedback;
- struct tfrc_rx_hist ccid3hcrx_hist;
- struct tfrc_loss_hist ccid3hcrx_li_hist;
- u16 ccid3hcrx_s;
-#define ccid3hcrx_pinv ccid3hcrx_li_hist.i_mean
+ u8 last_counter:4;
+ enum ccid3_fback_type feedback:4;
+ u32 x_recv;
+ ktime_t tstamp_last_feedback;
+ struct tfrc_rx_hist hist;
+ struct tfrc_loss_hist li_hist;
+#define p_inverse li_hist.i_mean
};
static inline struct ccid3_hc_rx_sock *ccid3_hc_rx_sk(const struct sock *sk)
diff --git a/net/dccp/ccids/lib/loss_interval.c b/net/dccp/ccids/lib/loss_interval.c
index 5b3ce0688c5..b1ae8f8259e 100644
--- a/net/dccp/ccids/lib/loss_interval.c
+++ b/net/dccp/ccids/lib/loss_interval.c
@@ -86,21 +86,26 @@ static void tfrc_lh_calc_i_mean(struct tfrc_loss_hist *lh)
/**
* tfrc_lh_update_i_mean - Update the `open' loss interval I_0
- * For recomputing p: returns `true' if p > p_prev <=> 1/p < 1/p_prev
+ * This updates I_mean as the sequence numbers increase. As a consequence, the
+ * open loss interval I_0 increases, hence p = W_tot/max(I_tot0, I_tot1)
+ * decreases, and thus there is no need to send renewed feedback.
*/
-u8 tfrc_lh_update_i_mean(struct tfrc_loss_hist *lh, struct sk_buff *skb)
+void tfrc_lh_update_i_mean(struct tfrc_loss_hist *lh, struct sk_buff *skb)
{
struct tfrc_loss_interval *cur = tfrc_lh_peek(lh);
- u32 old_i_mean = lh->i_mean;
s64 len;
if (cur == NULL) /* not initialised */
- return 0;
+ return;
+
+ /* FIXME: should probably also count non-data packets (RFC 4342, 6.1) */
+ if (!dccp_data_packet(skb))
+ return;
len = dccp_delta_seqno(cur->li_seqno, DCCP_SKB_CB(skb)->dccpd_seq) + 1;
if (len - (s64)cur->li_length <= 0) /* duplicate or reordered */
- return 0;
+ return;
if (SUB16(dccp_hdr(skb)->dccph_ccval, cur->li_ccval) > 4)
/*
@@ -114,14 +119,11 @@ u8 tfrc_lh_update_i_mean(struct tfrc_loss_hist *lh, struct sk_buff *skb)
cur->li_is_closed = 1;
if (tfrc_lh_length(lh) == 1) /* due to RFC 3448, 6.3.1 */
- return 0;
+ return;
cur->li_length = len;
tfrc_lh_calc_i_mean(lh);
-
- return (lh->i_mean < old_i_mean);
}
-EXPORT_SYMBOL_GPL(tfrc_lh_update_i_mean);
/* Determine if `new_loss' does begin a new loss interval [RFC 4342, 10.2] */
static inline u8 tfrc_lh_is_new_loss(struct tfrc_loss_interval *cur,
@@ -138,18 +140,18 @@ static inline u8 tfrc_lh_is_new_loss(struct tfrc_loss_interval *cur,
* @sk: Used by @calc_first_li in caller-specific way (subtyping)
* Updates I_mean and returns 1 if a new interval has in fact been added to @lh.
*/
-int tfrc_lh_interval_add(struct tfrc_loss_hist *lh, struct tfrc_rx_hist *rh,
- u32 (*calc_first_li)(struct sock *), struct sock *sk)
+bool tfrc_lh_interval_add(struct tfrc_loss_hist *lh, struct tfrc_rx_hist *rh,
+ u32 (*calc_first_li)(struct sock *), struct sock *sk)
{
struct tfrc_loss_interval *cur = tfrc_lh_peek(lh), *new;
if (cur != NULL && !tfrc_lh_is_new_loss(cur, tfrc_rx_hist_loss_prev(rh)))
- return 0;
+ return false;
new = tfrc_lh_demand_next(lh);
if (unlikely(new == NULL)) {
DCCP_CRIT("Cannot allocate/add loss record.");
- return 0;
+ return false;
}
new->li_seqno = tfrc_rx_hist_loss_prev(rh)->tfrchrx_seqno;
@@ -167,7 +169,7 @@ int tfrc_lh_interval_add(struct tfrc_loss_hist *lh, struct tfrc_rx_hist *rh,
tfrc_lh_calc_i_mean(lh);
}
- return 1;
+ return true;
}
EXPORT_SYMBOL_GPL(tfrc_lh_interval_add);
diff --git a/net/dccp/ccids/lib/loss_interval.h b/net/dccp/ccids/lib/loss_interval.h
index 246018a3b26..d08a226db43 100644
--- a/net/dccp/ccids/lib/loss_interval.h
+++ b/net/dccp/ccids/lib/loss_interval.h
@@ -67,9 +67,9 @@ static inline u8 tfrc_lh_length(struct tfrc_loss_hist *lh)
struct tfrc_rx_hist;
-extern int tfrc_lh_interval_add(struct tfrc_loss_hist *, struct tfrc_rx_hist *,
+extern bool tfrc_lh_interval_add(struct tfrc_loss_hist *, struct tfrc_rx_hist *,
u32 (*first_li)(struct sock *), struct sock *);
-extern u8 tfrc_lh_update_i_mean(struct tfrc_loss_hist *lh, struct sk_buff *);
+extern void tfrc_lh_update_i_mean(struct tfrc_loss_hist *lh, struct sk_buff *);
extern void tfrc_lh_cleanup(struct tfrc_loss_hist *lh);
#endif /* _DCCP_LI_HIST_ */
diff --git a/net/dccp/ccids/lib/packet_history.c b/net/dccp/ccids/lib/packet_history.c
index 6cc108afdc3..cce9f03bda3 100644
--- a/net/dccp/ccids/lib/packet_history.c
+++ b/net/dccp/ccids/lib/packet_history.c
@@ -40,18 +40,6 @@
#include "packet_history.h"
#include "../../dccp.h"
-/**
- * tfrc_tx_hist_entry - Simple singly-linked TX history list
- * @next: next oldest entry (LIFO order)
- * @seqno: sequence number of this entry
- * @stamp: send time of packet with sequence number @seqno
- */
-struct tfrc_tx_hist_entry {
- struct tfrc_tx_hist_entry *next;
- u64 seqno;
- ktime_t stamp;
-};
-
/*
* Transmitter History Routines
*/
@@ -73,15 +61,6 @@ void tfrc_tx_packet_history_exit(void)
}
}
-static struct tfrc_tx_hist_entry *
- tfrc_tx_hist_find_entry(struct tfrc_tx_hist_entry *head, u64 seqno)
-{
- while (head != NULL && head->seqno != seqno)
- head = head->next;
-
- return head;
-}
-
int tfrc_tx_hist_add(struct tfrc_tx_hist_entry **headp, u64 seqno)
{
struct tfrc_tx_hist_entry *entry = kmem_cache_alloc(tfrc_tx_hist_slab, gfp_any());
@@ -111,25 +90,6 @@ void tfrc_tx_hist_purge(struct tfrc_tx_hist_entry **headp)
}
EXPORT_SYMBOL_GPL(tfrc_tx_hist_purge);
-u32 tfrc_tx_hist_rtt(struct tfrc_tx_hist_entry *head, const u64 seqno,
- const ktime_t now)
-{
- u32 rtt = 0;
- struct tfrc_tx_hist_entry *packet = tfrc_tx_hist_find_entry(head, seqno);
-
- if (packet != NULL) {
- rtt = ktime_us_delta(now, packet->stamp);
- /*
- * Garbage-collect older (irrelevant) entries:
- */
- tfrc_tx_hist_purge(&packet->next);
- }
-
- return rtt;
-}
-EXPORT_SYMBOL_GPL(tfrc_tx_hist_rtt);
-
-
/*
* Receiver History Routines
*/
@@ -191,14 +151,31 @@ int tfrc_rx_hist_duplicate(struct tfrc_rx_hist *h, struct sk_buff *skb)
}
EXPORT_SYMBOL_GPL(tfrc_rx_hist_duplicate);
+
+static void __tfrc_rx_hist_swap(struct tfrc_rx_hist *h, const u8 a, const u8 b)
+{
+ struct tfrc_rx_hist_entry *tmp = h->ring[a];
+
+ h->ring[a] = h->ring[b];
+ h->ring[b] = tmp;
+}
+
static void tfrc_rx_hist_swap(struct tfrc_rx_hist *h, const u8 a, const u8 b)
{
- const u8 idx_a = tfrc_rx_hist_index(h, a),
- idx_b = tfrc_rx_hist_index(h, b);
- struct tfrc_rx_hist_entry *tmp = h->ring[idx_a];
+ __tfrc_rx_hist_swap(h, tfrc_rx_hist_index(h, a),
+ tfrc_rx_hist_index(h, b));
+}
- h->ring[idx_a] = h->ring[idx_b];
- h->ring[idx_b] = tmp;
+/**
+ * tfrc_rx_hist_resume_rtt_sampling - Prepare RX history for RTT sampling
+ * This is called after loss detection has finished, when the history entry
+ * with the index of `loss_count' holds the highest-received sequence number.
+ * RTT sampling requires this information at ring[0] (tfrc_rx_hist_sample_rtt).
+ */
+static inline void tfrc_rx_hist_resume_rtt_sampling(struct tfrc_rx_hist *h)
+{
+ __tfrc_rx_hist_swap(h, 0, tfrc_rx_hist_index(h, h->loss_count));
+ h->loss_count = h->loss_start = 0;
}
/*
@@ -215,10 +192,8 @@ static void __do_track_loss(struct tfrc_rx_hist *h, struct sk_buff *skb, u64 n1)
u64 s0 = tfrc_rx_hist_loss_prev(h)->tfrchrx_seqno,
s1 = DCCP_SKB_CB(skb)->dccpd_seq;
- if (!dccp_loss_free(s0, s1, n1)) { /* gap between S0 and S1 */
+ if (!dccp_loss_free(s0, s1, n1)) /* gap between S0 and S1 */
h->loss_count = 1;
- tfrc_rx_hist_entry_from_skb(tfrc_rx_hist_entry(h, 1), skb, n1);
- }
}
static void __one_after_loss(struct tfrc_rx_hist *h, struct sk_buff *skb, u32 n2)
@@ -240,8 +215,7 @@ static void __one_after_loss(struct tfrc_rx_hist *h, struct sk_buff *skb, u32 n2
if (dccp_loss_free(s2, s1, n1)) {
/* hole is filled: S0, S2, and S1 are consecutive */
- h->loss_count = 0;
- h->loss_start = tfrc_rx_hist_index(h, 1);
+ tfrc_rx_hist_resume_rtt_sampling(h);
} else
/* gap between S2 and S1: just update loss_prev */
tfrc_rx_hist_entry_from_skb(tfrc_rx_hist_loss_prev(h), skb, n2);
@@ -294,8 +268,7 @@ static int __two_after_loss(struct tfrc_rx_hist *h, struct sk_buff *skb, u32 n3)
if (dccp_loss_free(s1, s2, n2)) {
/* entire hole filled by S0, S3, S1, S2 */
- h->loss_start = tfrc_rx_hist_index(h, 2);
- h->loss_count = 0;
+ tfrc_rx_hist_resume_rtt_sampling(h);
} else {
/* gap remains between S1 and S2 */
h->loss_start = tfrc_rx_hist_index(h, 1);
@@ -339,8 +312,7 @@ static void __three_after_loss(struct tfrc_rx_hist *h)
if (dccp_loss_free(s2, s3, n3)) {
/* no gap between S2 and S3: entire hole is filled */
- h->loss_start = tfrc_rx_hist_index(h, 3);
- h->loss_count = 0;
+ tfrc_rx_hist_resume_rtt_sampling(h);
} else {
/* gap between S2 and S3 */
h->loss_start = tfrc_rx_hist_index(h, 2);
@@ -354,13 +326,13 @@ static void __three_after_loss(struct tfrc_rx_hist *h)
}
/**
- * tfrc_rx_handle_loss - Loss detection and further processing
- * @h: The non-empty RX history object
- * @lh: Loss Intervals database to update
- * @skb: Currently received packet
- * @ndp: The NDP count belonging to @skb
- * @calc_first_li: Caller-dependent computation of first loss interval in @lh
- * @sk: Used by @calc_first_li (see tfrc_lh_interval_add)
+ * tfrc_rx_congestion_event - Loss detection and further processing
+ * @h: The non-empty RX history object
+ * @lh: Loss Intervals database to update
+ * @skb: Currently received packet
+ * @ndp: The NDP count belonging to @skb
+ * @first_li: Caller-dependent computation of first loss interval in @lh
+ * @sk: Used by @calc_first_li (see tfrc_lh_interval_add)
* Chooses action according to pending loss, updates LI database when a new
* loss was detected, and does required post-processing. Returns 1 when caller
* should send feedback, 0 otherwise.
@@ -368,15 +340,20 @@ static void __three_after_loss(struct tfrc_rx_hist *h)
* records accordingly, the caller should not perform any more RX history
* operations when loss_count is greater than 0 after calling this function.
*/
-int tfrc_rx_handle_loss(struct tfrc_rx_hist *h,
- struct tfrc_loss_hist *lh,
- struct sk_buff *skb, const u64 ndp,
- u32 (*calc_first_li)(struct sock *), struct sock *sk)
+bool tfrc_rx_congestion_event(struct tfrc_rx_hist *h,
+ struct tfrc_loss_hist *lh,
+ struct sk_buff *skb, const u64 ndp,
+ u32 (*first_li)(struct sock *), struct sock *sk)
{
- int is_new_loss = 0;
+ bool new_event = false;
+
+ if (tfrc_rx_hist_duplicate(h, skb))
+ return 0;
if (h->loss_count == 0) {
__do_track_loss(h, skb, ndp);
+ tfrc_rx_hist_sample_rtt(h, skb);
+ tfrc_rx_hist_add_packet(h, skb, ndp);
} else if (h->loss_count == 1) {
__one_after_loss(h, skb, ndp);
} else if (h->loss_count != 2) {
@@ -385,34 +362,57 @@ int tfrc_rx_handle_loss(struct tfrc_rx_hist *h,
/*
* Update Loss Interval database and recycle RX records
*/
- is_new_loss = tfrc_lh_interval_add(lh, h, calc_first_li, sk);
+ new_event = tfrc_lh_interval_add(lh, h, first_li, sk);
__three_after_loss(h);
}
- return is_new_loss;
+
+ /*
+ * Update moving-average of `s' and the sum of received payload bytes.
+ */
+ if (dccp_data_packet(skb)) {
+ const u32 payload = skb->len - dccp_hdr(skb)->dccph_doff * 4;
+
+ h->packet_size = tfrc_ewma(h->packet_size, payload, 9);
+ h->bytes_recvd += payload;
+ }
+
+ /* RFC 3448, 6.1: update I_0, whose growth implies p <= p_prev */
+ if (!new_event)
+ tfrc_lh_update_i_mean(lh, skb);
+
+ return new_event;
}
-EXPORT_SYMBOL_GPL(tfrc_rx_handle_loss);
+EXPORT_SYMBOL_GPL(tfrc_rx_congestion_event);
-int tfrc_rx_hist_alloc(struct tfrc_rx_hist *h)
+/* Compute the sending rate X_recv measured between feedback intervals */
+u32 tfrc_rx_hist_x_recv(struct tfrc_rx_hist *h, const u32 last_x_recv)
{
- int i;
+ u64 bytes = h->bytes_recvd, last_rtt = h->rtt_estimate;
+ s64 delta = ktime_to_us(net_timedelta(h->bytes_start));
- for (i = 0; i <= TFRC_NDUPACK; i++) {
- h->ring[i] = kmem_cache_alloc(tfrc_rx_hist_slab, GFP_ATOMIC);
- if (h->ring[i] == NULL)
- goto out_free;
- }
+ WARN_ON(delta <= 0);
+ /*
+ * Ensure that the sampling interval for X_recv is at least one RTT,
+ * by extending the sampling interval backwards in time, over the last
+ * R_(m-1) seconds, as per rfc3448bis-06, 6.2.
+ * To reduce noise (e.g. when the RTT changes often), this is only
+ * done when delta is smaller than RTT/2.
+ */
+ if (last_x_recv > 0 && delta < last_rtt/2) {
+ tfrc_pr_debug("delta < RTT ==> %ld us < %u us\n",
+ (long)delta, (unsigned)last_rtt);
- h->loss_count = h->loss_start = 0;
- return 0;
+ delta = (bytes ? delta : 0) + last_rtt;
+ bytes += div_u64((u64)last_x_recv * last_rtt, USEC_PER_SEC);
+ }
-out_free:
- while (i-- != 0) {
- kmem_cache_free(tfrc_rx_hist_slab, h->ring[i]);
- h->ring[i] = NULL;
+ if (unlikely(bytes == 0)) {
+ DCCP_WARN("X_recv == 0, using old value of %u\n", last_x_recv);
+ return last_x_recv;
}
- return -ENOBUFS;
+ return scaled_div32(bytes, delta);
}
-EXPORT_SYMBOL_GPL(tfrc_rx_hist_alloc);
+EXPORT_SYMBOL_GPL(tfrc_rx_hist_x_recv);
void tfrc_rx_hist_purge(struct tfrc_rx_hist *h)
{
@@ -426,73 +426,81 @@ void tfrc_rx_hist_purge(struct tfrc_rx_hist *h)
}
EXPORT_SYMBOL_GPL(tfrc_rx_hist_purge);
-/**
- * tfrc_rx_hist_rtt_last_s - reference entry to compute RTT samples against
- */
-static inline struct tfrc_rx_hist_entry *
- tfrc_rx_hist_rtt_last_s(const struct tfrc_rx_hist *h)
+static int tfrc_rx_hist_alloc(struct tfrc_rx_hist *h)
{
- return h->ring[0];
+ int i;
+
+ memset(h, 0, sizeof(*h));
+
+ for (i = 0; i <= TFRC_NDUPACK; i++) {
+ h->ring[i] = kmem_cache_alloc(tfrc_rx_hist_slab, GFP_ATOMIC);
+ if (h->ring[i] == NULL) {
+ tfrc_rx_hist_purge(h);
+ return -ENOBUFS;
+ }
+ }
+ return 0;
}
-/**
- * tfrc_rx_hist_rtt_prev_s: previously suitable (wrt rtt_last_s) RTT-sampling entry
- */
-static inline struct tfrc_rx_hist_entry *
- tfrc_rx_hist_rtt_prev_s(const struct tfrc_rx_hist *h)
+int tfrc_rx_hist_init(struct tfrc_rx_hist *h, struct sock *sk)
{
- return h->ring[h->rtt_sample_prev];
+ if (tfrc_rx_hist_alloc(h))
+ return -ENOBUFS;
+ /*
+ * Initialise first entry with GSR to start loss detection as early as
+ * possible. Code using this must not use any other fields. The entry
+ * will be overwritten once the CCID updates its received packets.
+ */
+ tfrc_rx_hist_loss_prev(h)->tfrchrx_seqno = dccp_sk(sk)->dccps_gsr;
+ return 0;
}
+EXPORT_SYMBOL_GPL(tfrc_rx_hist_init);
/**
* tfrc_rx_hist_sample_rtt - Sample RTT from timestamp / CCVal
- * Based on ideas presented in RFC 4342, 8.1. Returns 0 if it was not able
- * to compute a sample with given data - calling function should check this.
+ * Based on ideas presented in RFC 4342, 8.1. This function expects that no loss
+ * is pending and uses the following history entries (via rtt_sample_prev):
+ * - h->ring[0] contains the most recent history entry prior to @skb;
+ * - h->ring[1] is an unused `dummy' entry when the current difference is 0;
*/
-u32 tfrc_rx_hist_sample_rtt(struct tfrc_rx_hist *h, const struct sk_buff *skb)
+void tfrc_rx_hist_sample_rtt(struct tfrc_rx_hist *h, const struct sk_buff *skb)
{
- u32 sample = 0,
- delta_v = SUB16(dccp_hdr(skb)->dccph_ccval,
- tfrc_rx_hist_rtt_last_s(h)->tfrchrx_ccval);
-
- if (delta_v < 1 || delta_v > 4) { /* unsuitable CCVal delta */
- if (h->rtt_sample_prev == 2) { /* previous candidate stored */
- sample = SUB16(tfrc_rx_hist_rtt_prev_s(h)->tfrchrx_ccval,
- tfrc_rx_hist_rtt_last_s(h)->tfrchrx_ccval);
- if (sample)
- sample = 4 / sample *
- ktime_us_delta(tfrc_rx_hist_rtt_prev_s(h)->tfrchrx_tstamp,
- tfrc_rx_hist_rtt_last_s(h)->tfrchrx_tstamp);
- else /*
- * FIXME: This condition is in principle not
- * possible but occurs when CCID is used for
- * two-way data traffic. I have tried to trace
- * it, but the cause does not seem to be here.
- */
- DCCP_BUG("please report to dccp@vger.kernel.org"
- " => prev = %u, last = %u",
- tfrc_rx_hist_rtt_prev_s(h)->tfrchrx_ccval,
- tfrc_rx_hist_rtt_last_s(h)->tfrchrx_ccval);
- } else if (delta_v < 1) {
- h->rtt_sample_prev = 1;
- goto keep_ref_for_next_time;
- }
+ struct tfrc_rx_hist_entry *last = h->ring[0];
+ u32 sample, delta_v;
- } else if (delta_v == 4) /* optimal match */
- sample = ktime_to_us(net_timedelta(tfrc_rx_hist_rtt_last_s(h)->tfrchrx_tstamp));
- else { /* suboptimal match */
- h->rtt_sample_prev = 2;
- goto keep_ref_for_next_time;
- }
+ /*
+ * When not to sample:
+ * - on non-data packets
+ * (RFC 4342, 8.1: CCVal only fully defined for data packets);
+ * - when no data packets have been received yet
+ * (FIXME: using sampled packet size as indicator here);
+ * - as long as there are gaps in the sequence space (pending loss).
+ */
+ if (!dccp_data_packet(skb) || h->packet_size == 0 ||
+ tfrc_rx_hist_loss_pending(h))
+ return;
- if (unlikely(sample > DCCP_SANE_RTT_MAX)) {
- DCCP_WARN("RTT sample %u too large, using max\n", sample);
- sample = DCCP_SANE_RTT_MAX;
+ h->rtt_sample_prev = 0; /* reset previous candidate */
+
+ delta_v = SUB16(dccp_hdr(skb)->dccph_ccval, last->tfrchrx_ccval);
+ if (delta_v == 0) { /* less than RTT/4 difference */
+ h->rtt_sample_prev = 1;
+ return;
}
+ sample = dccp_sane_rtt(ktime_to_us(net_timedelta(last->tfrchrx_tstamp)));
- h->rtt_sample_prev = 0; /* use current entry as next reference */
-keep_ref_for_next_time:
+ if (delta_v <= 4) /* between RTT/4 and RTT */
+ sample *= 4 / delta_v;
+ else if (!(sample < h->rtt_estimate && sample > h->rtt_estimate/2))
+ /*
+ * Optimisation: CCVal difference is greater than 1 RTT, yet the
+ * sample is less than the local RTT estimate; which means that
+ * the RTT estimate is too high.
+ * To avoid noise, it is not done if the sample is below RTT/2.
+ */
+ return;
- return sample;
+ /* Use a lower weight than usual to increase responsiveness */
+ h->rtt_estimate = tfrc_ewma(h->rtt_estimate, sample, 5);
}
EXPORT_SYMBOL_GPL(tfrc_rx_hist_sample_rtt);
diff --git a/net/dccp/ccids/lib/packet_history.h b/net/dccp/ccids/lib/packet_history.h
index 461cc91cce8..555e65cd73a 100644
--- a/net/dccp/ccids/lib/packet_history.h
+++ b/net/dccp/ccids/lib/packet_history.h
@@ -40,12 +40,28 @@
#include <linux/slab.h>
#include "tfrc.h"
-struct tfrc_tx_hist_entry;
+/**
+ * tfrc_tx_hist_entry - Simple singly-linked TX history list
+ * @next: next oldest entry (LIFO order)
+ * @seqno: sequence number of this entry
+ * @stamp: send time of packet with sequence number @seqno
+ */
+struct tfrc_tx_hist_entry {
+ struct tfrc_tx_hist_entry *next;
+ u64 seqno;
+ ktime_t stamp;
+};
+
+static inline struct tfrc_tx_hist_entry *
+ tfrc_tx_hist_find_entry(struct tfrc_tx_hist_entry *head, u64 seqno)
+{
+ while (head != NULL && head->seqno != seqno)
+ head = head->next;
+ return head;
+}
extern int tfrc_tx_hist_add(struct tfrc_tx_hist_entry **headp, u64 seqno);
extern void tfrc_tx_hist_purge(struct tfrc_tx_hist_entry **headp);
-extern u32 tfrc_tx_hist_rtt(struct tfrc_tx_hist_entry *head,
- const u64 seqno, const ktime_t now);
/* Subtraction a-b modulo-16, respects circular wrap-around */
#define SUB16(a, b) (((a) + 16 - (b)) & 0xF)
@@ -75,12 +91,22 @@ struct tfrc_rx_hist_entry {
* @loss_count: Number of entries in circular history
* @loss_start: Movable index (for loss detection)
* @rtt_sample_prev: Used during RTT sampling, points to candidate entry
+ * @rtt_estimate: Receiver RTT estimate
+ * @packet_size: Packet size in bytes (as per RFC 3448, 3.1)
+ * @bytes_recvd: Number of bytes received since @bytes_start
+ * @bytes_start: Start time for counting @bytes_recvd
*/
struct tfrc_rx_hist {
struct tfrc_rx_hist_entry *ring[TFRC_NDUPACK + 1];
u8 loss_count:2,
loss_start:2;
+ /* Receiver RTT sampling */
#define rtt_sample_prev loss_start
+ u32 rtt_estimate;
+ /* Receiver sampling of application payload lengths */
+ u32 packet_size,
+ bytes_recvd;
+ ktime_t bytes_start;
};
/**
@@ -124,20 +150,50 @@ static inline bool tfrc_rx_hist_loss_pending(const struct tfrc_rx_hist *h)
return h->loss_count > 0;
}
+/*
+ * Accessor functions to retrieve parameters sampled by the RX history
+ */
+static inline u32 tfrc_rx_hist_packet_size(const struct tfrc_rx_hist *h)
+{
+ if (h->packet_size == 0) {
+ DCCP_WARN("No sample for s, using fallback\n");
+ return TCP_MIN_RCVMSS;
+ }
+ return h->packet_size;
+
+}
+static inline u32 tfrc_rx_hist_rtt(const struct tfrc_rx_hist *h)
+{
+ if (h->rtt_estimate == 0) {
+ DCCP_WARN("No RTT estimate available, using fallback RTT\n");
+ return DCCP_FALLBACK_RTT;
+ }
+ return h->rtt_estimate;
+}
+
+static inline void tfrc_rx_hist_restart_byte_counter(struct tfrc_rx_hist *h)
+{
+ h->bytes_recvd = 0;
+ h->bytes_start = ktime_get_real();
+}
+
+extern u32 tfrc_rx_hist_x_recv(struct tfrc_rx_hist *h, const u32 last_x_recv);
+
+
extern void tfrc_rx_hist_add_packet(struct tfrc_rx_hist *h,
const struct sk_buff *skb, const u64 ndp);
extern int tfrc_rx_hist_duplicate(struct tfrc_rx_hist *h, struct sk_buff *skb);
struct tfrc_loss_hist;
-extern int tfrc_rx_handle_loss(struct tfrc_rx_hist *h,
- struct tfrc_loss_hist *lh,
- struct sk_buff *skb, const u64 ndp,
- u32 (*first_li)(struct sock *sk),
- struct sock *sk);
-extern u32 tfrc_rx_hist_sample_rtt(struct tfrc_rx_hist *h,
- const struct sk_buff *skb);
-extern int tfrc_rx_hist_alloc(struct tfrc_rx_hist *h);
+extern bool tfrc_rx_congestion_event(struct tfrc_rx_hist *h,
+ struct tfrc_loss_hist *lh,
+ struct sk_buff *skb, const u64 ndp,
+ u32 (*first_li)(struct sock *sk),
+ struct sock *sk);
+extern void tfrc_rx_hist_sample_rtt(struct tfrc_rx_hist *h,
+ const struct sk_buff *skb);
+extern int tfrc_rx_hist_init(struct tfrc_rx_hist *h, struct sock *sk);
extern void tfrc_rx_hist_purge(struct tfrc_rx_hist *h);
#endif /* _DCCP_PKT_HIST_ */
diff --git a/net/dccp/ccids/lib/tfrc.h b/net/dccp/ccids/lib/tfrc.h
index ed9857527ac..ede12f53de5 100644
--- a/net/dccp/ccids/lib/tfrc.h
+++ b/net/dccp/ccids/lib/tfrc.h
@@ -48,6 +48,21 @@ static inline u32 scaled_div32(u64 a, u64 b)
}
/**
+ * tfrc_scaled_sqrt - Compute scaled integer sqrt(x) for 0 < x < 2^22-1
+ * Uses scaling to improve accuracy of the integer approximation of sqrt(). The
+ * scaling factor of 2^10 limits the maximum @sample to 4e6; this is okay for
+ * clamped RTT samples (dccp_sample_rtt).
+ * Should best be used for expressions of type sqrt(x)/sqrt(y), since then the
+ * scaling factor is neutralised. For this purpose, it avoids returning zero.
+ */
+static inline u16 tfrc_scaled_sqrt(const u32 sample)
+{
+ const unsigned long non_zero_sample = sample ? : 1;
+
+ return int_sqrt(non_zero_sample << 10);
+}
+
+/**
* tfrc_ewma - Exponentially weighted moving average
* @weight: Weight to be used as damping factor, in units of 1/10
*/
@@ -58,6 +73,7 @@ static inline u32 tfrc_ewma(const u32 avg, const u32 newval, const u8 weight)
extern u32 tfrc_calc_x(u16 s, u32 R, u32 p);
extern u32 tfrc_calc_x_reverse_lookup(u32 fvalue);
+extern u32 tfrc_invert_loss_event_rate(u32 loss_event_rate);
extern int tfrc_tx_packet_history_init(void);
extern void tfrc_tx_packet_history_exit(void);
diff --git a/net/dccp/ccids/lib/tfrc_equation.c b/net/dccp/ccids/lib/tfrc_equation.c
index 2f20a29cffe..38239c4d5e1 100644
--- a/net/dccp/ccids/lib/tfrc_equation.c
+++ b/net/dccp/ccids/lib/tfrc_equation.c
@@ -632,8 +632,16 @@ u32 tfrc_calc_x(u16 s, u32 R, u32 p)
if (p <= TFRC_CALC_X_SPLIT) { /* 0.0000 < p <= 0.05 */
if (p < TFRC_SMALLEST_P) { /* 0.0000 < p < 0.0001 */
- DCCP_WARN("Value of p (%d) below resolution. "
- "Substituting %d\n", p, TFRC_SMALLEST_P);
+ /*
+ * In the congestion-avoidance phase p decays towards 0
+ * when there are no further losses, so this case is
+ * natural. Truncating to p_min = 0.01% means that the
+ * maximum achievable throughput is limited to about
+ * X_calc_max = 122.4 * s/RTT (see RFC 3448, 3.1); e.g.
+ * with s=1500 bytes, RTT=0.01 s: X_calc_max = 147 Mbps.
+ */
+ tfrc_pr_debug("Value of p (%d) below resolution. "
+ "Substituting %d\n", p, TFRC_SMALLEST_P);
index = 0;
} else /* 0.0001 <= p <= 0.05 */
index = p/TFRC_SMALLEST_P - 1;
@@ -658,7 +666,6 @@ u32 tfrc_calc_x(u16 s, u32 R, u32 p)
result = scaled_div(s, R);
return scaled_div32(result, f);
}
-
EXPORT_SYMBOL_GPL(tfrc_calc_x);
/**
@@ -693,5 +700,19 @@ u32 tfrc_calc_x_reverse_lookup(u32 fvalue)
index = tfrc_binsearch(fvalue, 0);
return (index + 1) * 1000000 / TFRC_CALC_X_ARRSIZE;
}
-
EXPORT_SYMBOL_GPL(tfrc_calc_x_reverse_lookup);
+
+/**
+ * tfrc_invert_loss_event_rate - Compute p so that 10^6 corresponds to 100%
+ * When @loss_event_rate is large, there is a chance that p is truncated to 0.
+ * To avoid re-entering slow-start in that case, we set p = TFRC_SMALLEST_P > 0.
+ */
+u32 tfrc_invert_loss_event_rate(u32 loss_event_rate)
+{
+ if (loss_event_rate == UINT_MAX) /* see RFC 4342, 8.5 */
+ return 0;
+ if (unlikely(loss_event_rate == 0)) /* map 1/0 into 100% */
+ return 1000000;
+ return max_t(u32, scaled_div(1, loss_event_rate), TFRC_SMALLEST_P);
+}
+EXPORT_SYMBOL_GPL(tfrc_invert_loss_event_rate);
generated by cgit v1.2.3-70-g09d2 (git 2.46.0) at 2024-12-16 14:20:53 +0000