From 30c6f0bf9579debce27e45fac34fdc97e46acacc Mon Sep 17 00:00:00 2001 From: fuyuanli Date: Wed, 31 May 2023 16:01:50 +0800 Subject: tcp: fix mishandling when the sack compression is deferred. In this patch, we mainly try to handle sending a compressed ack correctly if it's deferred. Here are more details in the old logic: When sack compression is triggered in the tcp_compressed_ack_kick(), if the sock is owned by user, it will set TCP_DELACK_TIMER_DEFERRED and then defer to the release cb phrase. Later once user releases the sock, tcp_delack_timer_handler() should send a ack as expected, which, however, cannot happen due to lack of ICSK_ACK_TIMER flag. Therefore, the receiver would not sent an ack until the sender's retransmission timeout. It definitely increases unnecessary latency. Fixes: 5d9f4262b7ea ("tcp: add SACK compression") Suggested-by: Eric Dumazet Signed-off-by: fuyuanli Signed-off-by: Jason Xing Link: https://lore.kernel.org/netdev/20230529113804.GA20300@didi-ThinkCentre-M920t-N000/ Reviewed-by: Eric Dumazet Link: https://lore.kernel.org/r/20230531080150.GA20424@didi-ThinkCentre-M920t-N000 Signed-off-by: Paolo Abeni --- net/ipv4/tcp_input.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'net/ipv4/tcp_input.c') diff --git a/net/ipv4/tcp_input.c b/net/ipv4/tcp_input.c index 61b6710f337a..bf8b22218dd4 100644 --- a/net/ipv4/tcp_input.c +++ b/net/ipv4/tcp_input.c @@ -4530,7 +4530,7 @@ static void tcp_sack_maybe_coalesce(struct tcp_sock *tp) } } -static void tcp_sack_compress_send_ack(struct sock *sk) +void tcp_sack_compress_send_ack(struct sock *sk) { struct tcp_sock *tp = tcp_sk(sk); -- cgit v1.2.3 From 304b1875ba02022f2504952a32e5e90482bbdb39 Mon Sep 17 00:00:00 2001 From: Yueh-Shun Li Date: Thu, 22 Jun 2023 01:26:31 +0000 Subject: tcp: fix comment typo Spell "transmissions" properly. Found by searching for keyword "tranm". Signed-off-by: Yueh-Shun Li Link: https://lore.kernel.org/r/20230622012627.15050-6-shamrocklee@posteo.net Signed-off-by: Jakub Kicinski --- net/ipv4/tcp_input.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'net/ipv4/tcp_input.c') diff --git a/net/ipv4/tcp_input.c b/net/ipv4/tcp_input.c index bf8b22218dd4..6f072095211e 100644 --- a/net/ipv4/tcp_input.c +++ b/net/ipv4/tcp_input.c @@ -2867,7 +2867,7 @@ static void tcp_process_loss(struct sock *sk, int flag, int num_dupack, } if (tcp_is_reno(tp)) { /* A Reno DUPACK means new data in F-RTO step 2.b above are - * delivered. Lower inflight to clock out (re)tranmissions. + * delivered. Lower inflight to clock out (re)transmissions. */ if (after(tp->snd_nxt, tp->high_seq) && num_dupack) tcp_add_reno_sack(sk, num_dupack, flag & FLAG_ECE); -- cgit v1.2.3 From 998127cdb4699b9d470a9348ffe9f1154346be5f Mon Sep 17 00:00:00 2001 From: Eric Dumazet Date: Thu, 29 Jun 2023 16:41:50 +0000 Subject: tcp: annotate data races in __tcp_oow_rate_limited() request sockets are lockless, __tcp_oow_rate_limited() could be called on the same object from different cpus. This is harmless. Add READ_ONCE()/WRITE_ONCE() annotations to avoid a KCSAN report. Fixes: 4ce7e93cb3fe ("tcp: rate limit ACK sent by SYN_RECV request sockets") Signed-off-by: Eric Dumazet Signed-off-by: David S. Miller --- net/ipv4/tcp_input.c | 12 +++++++++--- 1 file changed, 9 insertions(+), 3 deletions(-) (limited to 'net/ipv4/tcp_input.c') diff --git a/net/ipv4/tcp_input.c b/net/ipv4/tcp_input.c index 6f072095211e..57c8af1859c1 100644 --- a/net/ipv4/tcp_input.c +++ b/net/ipv4/tcp_input.c @@ -3590,8 +3590,11 @@ static int tcp_ack_update_window(struct sock *sk, const struct sk_buff *skb, u32 static bool __tcp_oow_rate_limited(struct net *net, int mib_idx, u32 *last_oow_ack_time) { - if (*last_oow_ack_time) { - s32 elapsed = (s32)(tcp_jiffies32 - *last_oow_ack_time); + /* Paired with the WRITE_ONCE() in this function. */ + u32 val = READ_ONCE(*last_oow_ack_time); + + if (val) { + s32 elapsed = (s32)(tcp_jiffies32 - val); if (0 <= elapsed && elapsed < READ_ONCE(net->ipv4.sysctl_tcp_invalid_ratelimit)) { @@ -3600,7 +3603,10 @@ static bool __tcp_oow_rate_limited(struct net *net, int mib_idx, } } - *last_oow_ack_time = tcp_jiffies32; + /* Paired with the prior READ_ONCE() and with itself, + * as we might be lockless. + */ + WRITE_ONCE(*last_oow_ack_time, tcp_jiffies32); return false; /* not rate-limited: go ahead, send dupack now! */ } -- cgit v1.2.3