From 95b9a87c6a6b708cccda1f9b7baf9920b80cdabf Mon Sep 17 00:00:00 2001 From: David Morley Date: Fri, 6 Oct 2023 01:18:40 +0000 Subject: [PATCH 1/2] tcp: record last received ipv6 flowlabel In order to better estimate whether a data packet has been retransmitted or is the result of a TLP, we save the last received ipv6 flowlabel. To make space for this field we resize the "ato" field in inet_connection_sock as the current value of TCP_DELACK_MAX can be fully contained in 8 bits and add a compile_time_assert ensuring this field is the required size. v2: addressed kernel bot feedback about dccp_delack_timer() v3: addressed build error introduced by commit bbf80d713fe7 ("tcp: derive delack_max from rto_min") Signed-off-by: David Morley Signed-off-by: Neal Cardwell Signed-off-by: Yuchung Cheng Tested-by: David Morley Reviewed-by: Eric Dumazet Signed-off-by: Paolo Abeni --- include/net/inet_connection_sock.h | 5 ++++- include/net/tcp.h | 2 ++ net/dccp/timer.c | 4 ++-- net/ipv4/tcp.c | 4 ++-- net/ipv4/tcp_input.c | 15 +++++++++++++++ net/ipv4/tcp_timer.c | 2 +- 6 files changed, 26 insertions(+), 6 deletions(-) diff --git a/include/net/inet_connection_sock.h b/include/net/inet_connection_sock.h index 5d2fcc137b88..d6d9d1c1985a 100644 --- a/include/net/inet_connection_sock.h +++ b/include/net/inet_connection_sock.h @@ -114,7 +114,10 @@ struct inet_connection_sock { __u8 quick; /* Scheduled number of quick acks */ __u8 pingpong; /* The session is interactive */ __u8 retry; /* Number of attempts */ - __u32 ato; /* Predicted tick of soft clock */ + #define ATO_BITS 8 + __u32 ato:ATO_BITS, /* Predicted tick of soft clock */ + lrcv_flowlabel:20, /* last received ipv6 flowlabel */ + unused:4; unsigned long timeout; /* Currently scheduled timeout */ __u32 lrcvtime; /* timestamp of last received data packet */ __u16 last_seg_size; /* Size of last incoming segment */ diff --git a/include/net/tcp.h b/include/net/tcp.h index 9eb0a2855311..7fdedf5c71f0 100644 --- a/include/net/tcp.h +++ b/include/net/tcp.h @@ -131,6 +131,8 @@ void tcp_time_wait(struct sock *sk, int state, int timeo); #define TCP_FIN_TIMEOUT_MAX (120 * HZ) /* max TCP_LINGER2 value (two minutes) */ #define TCP_DELACK_MAX ((unsigned)(HZ/5)) /* maximal time to delay before sending an ACK */ +static_assert((1 << ATO_BITS) > TCP_DELACK_MAX); + #if HZ >= 100 #define TCP_DELACK_MIN ((unsigned)(HZ/25)) /* minimal time to delay before sending an ACK */ #define TCP_ATO_MIN ((unsigned)(HZ/25)) diff --git a/net/dccp/timer.c b/net/dccp/timer.c index b3255e87cc7e..a4cfb47b60e5 100644 --- a/net/dccp/timer.c +++ b/net/dccp/timer.c @@ -196,8 +196,8 @@ static void dccp_delack_timer(struct timer_list *t) if (inet_csk_ack_scheduled(sk)) { if (!inet_csk_in_pingpong_mode(sk)) { /* Delayed ACK missed: inflate ATO. */ - icsk->icsk_ack.ato = min(icsk->icsk_ack.ato << 1, - icsk->icsk_rto); + icsk->icsk_ack.ato = min_t(u32, icsk->icsk_ack.ato << 1, + icsk->icsk_rto); } else { /* Delayed ACK missed: leave pingpong mode and * deflate ATO. diff --git a/net/ipv4/tcp.c b/net/ipv4/tcp.c index 9a8b134d8ada..faabb5a4a378 100644 --- a/net/ipv4/tcp.c +++ b/net/ipv4/tcp.c @@ -3756,8 +3756,8 @@ void tcp_get_info(struct sock *sk, struct tcp_info *info) info->tcpi_options |= TCPI_OPT_SYN_DATA; info->tcpi_rto = jiffies_to_usecs(icsk->icsk_rto); - info->tcpi_ato = jiffies_to_usecs(min(icsk->icsk_ack.ato, - tcp_delack_max(sk))); + info->tcpi_ato = jiffies_to_usecs(min_t(u32, icsk->icsk_ack.ato, + tcp_delack_max(sk))); info->tcpi_snd_mss = tp->mss_cache; info->tcpi_rcv_mss = icsk->icsk_ack.rcv_mss; diff --git a/net/ipv4/tcp_input.c b/net/ipv4/tcp_input.c index 4b8f2e74d71d..2ae4b211c125 100644 --- a/net/ipv4/tcp_input.c +++ b/net/ipv4/tcp_input.c @@ -778,6 +778,16 @@ new_measure: tp->rcvq_space.time = tp->tcp_mstamp; } +static void tcp_save_lrcv_flowlabel(struct sock *sk, const struct sk_buff *skb) +{ +#if IS_ENABLED(CONFIG_IPV6) + struct inet_connection_sock *icsk = inet_csk(sk); + + if (skb->protocol == htons(ETH_P_IPV6)) + icsk->icsk_ack.lrcv_flowlabel = ntohl(ip6_flowlabel(ipv6_hdr(skb))); +#endif +} + /* There is something which you must keep in mind when you analyze the * behavior of the tp->ato delayed ack timeout interval. When a * connection starts up, we want to ack as quickly as possible. The @@ -826,6 +836,7 @@ static void tcp_event_data_recv(struct sock *sk, struct sk_buff *skb) } } icsk->icsk_ack.lrcvtime = now; + tcp_save_lrcv_flowlabel(sk, skb); tcp_ecn_check_ce(sk, skb); @@ -4519,6 +4530,9 @@ static void tcp_rcv_spurious_retrans(struct sock *sk, const struct sk_buff *skb) if (TCP_SKB_CB(skb)->seq == tcp_sk(sk)->duplicate_sack[0].start_seq && sk_rethink_txhash(sk)) NET_INC_STATS(sock_net(sk), LINUX_MIB_TCPDUPLICATEDATAREHASH); + + /* Save last flowlabel after a spurious retrans. */ + tcp_save_lrcv_flowlabel(sk, skb); } static void tcp_send_dupack(struct sock *sk, const struct sk_buff *skb) @@ -4835,6 +4849,7 @@ static void tcp_data_queue_ofo(struct sock *sk, struct sk_buff *skb) u32 seq, end_seq; bool fragstolen; + tcp_save_lrcv_flowlabel(sk, skb); tcp_ecn_check_ce(sk, skb); if (unlikely(tcp_try_rmem_schedule(sk, skb, skb->truesize))) { diff --git a/net/ipv4/tcp_timer.c b/net/ipv4/tcp_timer.c index 3f61c6a70a1f..0862b73dd3b5 100644 --- a/net/ipv4/tcp_timer.c +++ b/net/ipv4/tcp_timer.c @@ -322,7 +322,7 @@ void tcp_delack_timer_handler(struct sock *sk) if (inet_csk_ack_scheduled(sk)) { if (!inet_csk_in_pingpong_mode(sk)) { /* Delayed ACK missed: inflate ATO. */ - icsk->icsk_ack.ato = min(icsk->icsk_ack.ato << 1, icsk->icsk_rto); + icsk->icsk_ack.ato = min_t(u32, icsk->icsk_ack.ato << 1, icsk->icsk_rto); } else { /* Delayed ACK missed: leave pingpong mode and * deflate ATO. From 939463016b7a869d8b407cfcda4d6545de399698 Mon Sep 17 00:00:00 2001 From: David Morley Date: Fri, 6 Oct 2023 01:18:41 +0000 Subject: [PATCH 2/2] tcp: change data receiver flowlabel after one dup This commit changes the data receiver repath behavior to occur after receiving a single duplicate. This can help recover ACK connectivity quicker if a TLP was sent along a nonworking path. For instance, consider the case where we have an initially nonworking forward path and reverse path and subsequently switch to only working forward paths. Before this patch we would have the following behavior. +---------+--------+--------+----------+----------+----------+ | Event | For FL | Rev FL | FP Works | RP Works | Data Del | +---------+--------+--------+----------+----------+----------+ | Initial | A | 1 | N | N | 0 | +---------+--------+--------+----------+----------+----------+ | TLP | A | 1 | N | N | 0 | +---------+--------+--------+----------+----------+----------+ | RTO 1 | B | 1 | Y | N | 1 | +---------+--------+--------+----------+----------+----------+ | RTO 2 | C | 1 | Y | N | 2 | +---------+--------+--------+----------+----------+----------+ | RTO 3 | D | 2 | Y | Y | 3 | +---------+--------+--------+----------+----------+----------+ This patch gets rid of at least RTO 3, avoiding additional unnecessary repaths of a working forward path to a (potentially) nonworking one. In addition, this commit changes the behavior to avoid repathing upon rx of duplicate data if the local endpoint is in CA_Loss (in which case the RTOs will already be changing the outgoing flowlabel). Signed-off-by: David Morley Signed-off-by: Neal Cardwell Signed-off-by: Yuchung Cheng Tested-by: David Morley Reviewed-by: Eric Dumazet Signed-off-by: Paolo Abeni --- net/ipv4/tcp_input.c | 14 +++++++++++--- 1 file changed, 11 insertions(+), 3 deletions(-) diff --git a/net/ipv4/tcp_input.c b/net/ipv4/tcp_input.c index 2ae4b211c125..ab87f0285b72 100644 --- a/net/ipv4/tcp_input.c +++ b/net/ipv4/tcp_input.c @@ -4524,15 +4524,23 @@ static void tcp_rcv_spurious_retrans(struct sock *sk, const struct sk_buff *skb) { /* When the ACK path fails or drops most ACKs, the sender would * timeout and spuriously retransmit the same segment repeatedly. - * The receiver remembers and reflects via DSACKs. Leverage the - * DSACK state and change the txhash to re-route speculatively. + * If it seems our ACKs are not reaching the other side, + * based on receiving a duplicate data segment with new flowlabel + * (suggesting the sender suffered an RTO), and we are not already + * repathing due to our own RTO, then rehash the socket to repath our + * packets. */ - if (TCP_SKB_CB(skb)->seq == tcp_sk(sk)->duplicate_sack[0].start_seq && +#if IS_ENABLED(CONFIG_IPV6) + if (inet_csk(sk)->icsk_ca_state != TCP_CA_Loss && + skb->protocol == htons(ETH_P_IPV6) && + (tcp_sk(sk)->inet_conn.icsk_ack.lrcv_flowlabel != + ntohl(ip6_flowlabel(ipv6_hdr(skb)))) && sk_rethink_txhash(sk)) NET_INC_STATS(sock_net(sk), LINUX_MIB_TCPDUPLICATEDATAREHASH); /* Save last flowlabel after a spurious retrans. */ tcp_save_lrcv_flowlabel(sk, skb); +#endif } static void tcp_send_dupack(struct sock *sk, const struct sk_buff *skb)