@@ -958,6 +958,7 @@ u32 tcp_slow_start(struct tcp_sock *tp, u32 acked);
void tcp_cong_avoid_ai(struct tcp_sock *tp, u32 w, u32 acked);
u32 tcp_reno_ssthresh(struct sock *sk);
+u32 tcp_reno_undo_cwnd(struct sock *sk);
void tcp_reno_cong_avoid(struct sock *sk, u32 ack, u32 acked);
extern struct tcp_congestion_ops tcp_reno;
@@ -68,8 +68,9 @@ int tcp_register_congestion_control(struct tcp_congestion_ops *ca)
{
int ret = 0;
- /* all algorithms must implement ssthresh and cong_avoid ops */
- if (!ca->ssthresh || !(ca->cong_avoid || ca->cong_control)) {
+ /* all algorithms must implement these */
+ if (!ca->ssthresh || !ca->undo_cwnd ||
+ !(ca->cong_avoid || ca->cong_control)) {
pr_err("%s does not implement required ops\n", ca->name);
return -EINVAL;
}
@@ -443,10 +444,19 @@ u32 tcp_reno_ssthresh(struct sock *sk)
}
EXPORT_SYMBOL_GPL(tcp_reno_ssthresh);
+u32 tcp_reno_undo_cwnd(struct sock *sk)
+{
+ const struct tcp_sock *tp = tcp_sk(sk);
+
+ return max(tp->snd_cwnd, tp->snd_ssthresh << 1);
+}
+EXPORT_SYMBOL_GPL(tcp_reno_undo_cwnd);
+
struct tcp_congestion_ops tcp_reno = {
.flags = TCP_CONG_NON_RESTRICTED,
.name = "reno",
.owner = THIS_MODULE,
.ssthresh = tcp_reno_ssthresh,
.cong_avoid = tcp_reno_cong_avoid,
+ .undo_cwnd = tcp_reno_undo_cwnd,
};
@@ -342,6 +342,7 @@ static struct tcp_congestion_ops dctcp __read_mostly = {
static struct tcp_congestion_ops dctcp_reno __read_mostly = {
.ssthresh = tcp_reno_ssthresh,
.cong_avoid = tcp_reno_cong_avoid,
+ .undo_cwnd = tcp_reno_undo_cwnd,
.get_info = dctcp_get_info,
.owner = THIS_MODULE,
.name = "dctcp-reno",
@@ -166,6 +166,7 @@ static void hybla_cong_avoid(struct sock *sk, u32 ack, u32 acked)
static struct tcp_congestion_ops tcp_hybla __read_mostly = {
.init = hybla_init,
.ssthresh = tcp_reno_ssthresh,
+ .undo_cwnd = tcp_reno_undo_cwnd,
.cong_avoid = hybla_cong_avoid,
.set_state = hybla_state,
@@ -2394,10 +2394,7 @@ static void tcp_undo_cwnd_reduction(struct sock *sk, bool unmark_loss)
if (tp->prior_ssthresh) {
const struct inet_connection_sock *icsk = inet_csk(sk);
- if (icsk->icsk_ca_ops->undo_cwnd)
- tp->snd_cwnd = icsk->icsk_ca_ops->undo_cwnd(sk);
- else
- tp->snd_cwnd = max(tp->snd_cwnd, tp->snd_ssthresh << 1);
+ tp->snd_cwnd = icsk->icsk_ca_ops->undo_cwnd(sk);
if (tp->prior_ssthresh > tp->snd_ssthresh) {
tp->snd_ssthresh = tp->prior_ssthresh;
@@ -316,6 +316,7 @@ static void tcp_lp_pkts_acked(struct sock *sk, const struct ack_sample *sample)
static struct tcp_congestion_ops tcp_lp __read_mostly = {
.init = tcp_lp_init,
.ssthresh = tcp_reno_ssthresh,
+ .undo_cwnd = tcp_reno_undo_cwnd,
.cong_avoid = tcp_lp_cong_avoid,
.pkts_acked = tcp_lp_pkts_acked,
@@ -307,6 +307,7 @@ EXPORT_SYMBOL_GPL(tcp_vegas_get_info);
static struct tcp_congestion_ops tcp_vegas __read_mostly = {
.init = tcp_vegas_init,
.ssthresh = tcp_reno_ssthresh,
+ .undo_cwnd = tcp_reno_undo_cwnd,
.cong_avoid = tcp_vegas_cong_avoid,
.pkts_acked = tcp_vegas_pkts_acked,
.set_state = tcp_vegas_state,
@@ -278,6 +278,7 @@ static struct tcp_congestion_ops tcp_westwood __read_mostly = {
.init = tcp_westwood_init,
.ssthresh = tcp_reno_ssthresh,
.cong_avoid = tcp_reno_cong_avoid,
+ .undo_cwnd = tcp_reno_undo_cwnd,
.cwnd_event = tcp_westwood_event,
.in_ack_event = tcp_westwood_ack,
.get_info = tcp_westwood_info,
The undo_cwnd fallback in the stack doubles cwnd based on ssthresh, which un-does reno halving behaviour. It seems more appropriate to let congctl algorithms pair .ssthresh and .undo_cwnd properly. Add a 'tcp_reno_undo_cwnd' function and wire it up for all congestion algorithms that used to rely on the fallback. Cc: Eric Dumazet <edumazet@google.com> Cc: Yuchung Cheng <ycheng@google.com> Cc: Neal Cardwell <ncardwell@google.com> Signed-off-by: Florian Westphal <fw@strlen.de> --- Changes since v1: - only use tcp_reno_undo_cwnd in algorithms that use tcp_reno_ssthresh. The other algorithms that did not meet this requirement were changed in patch 1 of this series. include/net/tcp.h | 1 + net/ipv4/tcp_cong.c | 14 ++++++++++++-- net/ipv4/tcp_dctcp.c | 1 + net/ipv4/tcp_hybla.c | 1 + net/ipv4/tcp_input.c | 5 +---- net/ipv4/tcp_lp.c | 1 + net/ipv4/tcp_vegas.c | 1 + net/ipv4/tcp_westwood.c | 1 + 8 files changed, 19 insertions(+), 6 deletions(-)