From 0df48c26d8418c5c9fba63fac15b660d70ca2f1c Mon Sep 17 00:00:00 2001 From: Eric Dumazet Date: Tue, 28 Apr 2015 15:28:17 -0700 Subject: tcp: add tcpi_bytes_acked to tcp_info This patch tracks total number of bytes acked for a TCP socket. This is the sum of all changes done to tp->snd_una, and allows for precise tracking of delivered data. RFC4898 named this : tcpEStatsAppHCThruOctetsAcked This is a 64bit field, and can be fetched both from TCP_INFO getsockopt() if one has a handle on a TCP socket, or from inet_diag netlink facility (iproute2/ss patch will follow) Note that tp->bytes_acked was placed near tp->snd_una for best data locality and minimal performance impact. Signed-off-by: Eric Dumazet Acked-by: Yuchung Cheng Cc: Matt Mathis Cc: Eric Salo Cc: Martin Lau Cc: Chris Rapier Signed-off-by: David S. Miller --- net/ipv4/tcp.c | 6 +++++- net/ipv4/tcp_input.c | 13 +++++++++++-- 2 files changed, 16 insertions(+), 3 deletions(-) (limited to 'net/ipv4') diff --git a/net/ipv4/tcp.c b/net/ipv4/tcp.c index 8c5cd9efebbc..4bf0e8ca7b5b 100644 --- a/net/ipv4/tcp.c +++ b/net/ipv4/tcp.c @@ -2592,7 +2592,7 @@ EXPORT_SYMBOL(compat_tcp_setsockopt); #endif /* Return information about state of tcp endpoint in API format. */ -void tcp_get_info(const struct sock *sk, struct tcp_info *info) +void tcp_get_info(struct sock *sk, struct tcp_info *info) { const struct tcp_sock *tp = tcp_sk(sk); const struct inet_connection_sock *icsk = inet_csk(sk); @@ -2663,6 +2663,10 @@ void tcp_get_info(const struct sock *sk, struct tcp_info *info) rate = READ_ONCE(sk->sk_max_pacing_rate); info->tcpi_max_pacing_rate = rate != ~0U ? rate : ~0ULL; + + spin_lock_bh(&sk->sk_lock.slock); + info->tcpi_bytes_acked = tp->bytes_acked; + spin_unlock_bh(&sk->sk_lock.slock); } EXPORT_SYMBOL_GPL(tcp_get_info); diff --git a/net/ipv4/tcp_input.c b/net/ipv4/tcp_input.c index 3a4d9b34bed4..378d3f4d4dc3 100644 --- a/net/ipv4/tcp_input.c +++ b/net/ipv4/tcp_input.c @@ -3280,6 +3280,15 @@ static inline bool tcp_may_update_window(const struct tcp_sock *tp, (ack_seq == tp->snd_wl1 && nwin > tp->snd_wnd); } +/* If we update tp->snd_una, also update tp->bytes_acked */ +static void tcp_snd_una_update(struct tcp_sock *tp, u32 ack) +{ + u32 delta = ack - tp->snd_una; + + tp->bytes_acked += delta; + tp->snd_una = ack; +} + /* Update our send window. * * Window update algorithm, described in RFC793/RFC1122 (used in linux-2.2 @@ -3315,7 +3324,7 @@ static int tcp_ack_update_window(struct sock *sk, const struct sk_buff *skb, u32 } } - tp->snd_una = ack; + tcp_snd_una_update(tp, ack); return flag; } @@ -3497,7 +3506,7 @@ static int tcp_ack(struct sock *sk, const struct sk_buff *skb, int flag) * Note, we use the fact that SND.UNA>=SND.WL2. */ tcp_update_wl(tp, ack_seq); - tp->snd_una = ack; + tcp_snd_una_update(tp, ack); flag |= FLAG_WIN_UPDATE; tcp_in_ack_event(sk, CA_ACK_WIN_UPDATE); -- cgit v1.2.3 From bdd1f9edacb5f5835d1e6276571bbbe5b88ded48 Mon Sep 17 00:00:00 2001 From: Eric Dumazet Date: Tue, 28 Apr 2015 15:28:18 -0700 Subject: tcp: add tcpi_bytes_received to tcp_info This patch tracks total number of payload bytes received on a TCP socket. This is the sum of all changes done to tp->rcv_nxt RFC4898 named this : tcpEStatsAppHCThruOctetsReceived This is a 64bit field, and can be fetched both from TCP_INFO getsockopt() if one has a handle on a TCP socket, or from inet_diag netlink facility (iproute2/ss patch will follow) Note that tp->bytes_received was placed near tp->rcv_nxt for best data locality and minimal performance impact. Signed-off-by: Eric Dumazet Cc: Yuchung Cheng Cc: Matt Mathis Cc: Eric Salo Cc: Martin Lau Cc: Chris Rapier Acked-by: Yuchung Cheng Signed-off-by: David S. Miller --- include/linux/tcp.h | 4 ++++ include/uapi/linux/tcp.h | 1 + net/ipv4/tcp.c | 1 + net/ipv4/tcp_fastopen.c | 1 + net/ipv4/tcp_input.c | 17 +++++++++++++---- 5 files changed, 20 insertions(+), 4 deletions(-) (limited to 'net/ipv4') diff --git a/include/linux/tcp.h b/include/linux/tcp.h index 0f73b43171da..3b2911502a8c 100644 --- a/include/linux/tcp.h +++ b/include/linux/tcp.h @@ -145,6 +145,10 @@ struct tcp_sock { * read the code and the spec side by side (and laugh ...) * See RFC793 and RFC1122. The RFC writes these in capitals. */ + u64 bytes_received; /* RFC4898 tcpEStatsAppHCThruOctetsReceived + * sum(delta(rcv_nxt)), or how many bytes + * were acked. + */ u32 rcv_nxt; /* What we want to receive next */ u32 copied_seq; /* Head of yet unread data */ u32 rcv_wup; /* rcv_nxt on last window update sent */ diff --git a/include/uapi/linux/tcp.h b/include/uapi/linux/tcp.h index 6666e98a0af9..a48f93f3207b 100644 --- a/include/uapi/linux/tcp.h +++ b/include/uapi/linux/tcp.h @@ -190,6 +190,7 @@ struct tcp_info { __u64 tcpi_pacing_rate; __u64 tcpi_max_pacing_rate; __u64 tcpi_bytes_acked; /* RFC4898 tcpEStatsAppHCThruOctetsAcked */ + __u64 tcpi_bytes_received; /* RFC4898 tcpEStatsAppHCThruOctetsReceived */ }; /* for TCP_MD5SIG socket option */ diff --git a/net/ipv4/tcp.c b/net/ipv4/tcp.c index 4bf0e8ca7b5b..99fcc0b22c92 100644 --- a/net/ipv4/tcp.c +++ b/net/ipv4/tcp.c @@ -2666,6 +2666,7 @@ void tcp_get_info(struct sock *sk, struct tcp_info *info) spin_lock_bh(&sk->sk_lock.slock); info->tcpi_bytes_acked = tp->bytes_acked; + info->tcpi_bytes_received = tp->bytes_received; spin_unlock_bh(&sk->sk_lock.slock); } EXPORT_SYMBOL_GPL(tcp_get_info); diff --git a/net/ipv4/tcp_fastopen.c b/net/ipv4/tcp_fastopen.c index e3d87aca6be8..3c673d5e6cff 100644 --- a/net/ipv4/tcp_fastopen.c +++ b/net/ipv4/tcp_fastopen.c @@ -206,6 +206,7 @@ static bool tcp_fastopen_create_child(struct sock *sk, skb_set_owner_r(skb2, child); __skb_queue_tail(&child->sk_receive_queue, skb2); tp->syn_data_acked = 1; + tp->bytes_received = end_seq - TCP_SKB_CB(skb)->seq - 1; } else { end_seq = TCP_SKB_CB(skb)->seq + 1; } diff --git a/net/ipv4/tcp_input.c b/net/ipv4/tcp_input.c index 378d3f4d4dc3..7e6962bcfc30 100644 --- a/net/ipv4/tcp_input.c +++ b/net/ipv4/tcp_input.c @@ -3289,6 +3289,15 @@ static void tcp_snd_una_update(struct tcp_sock *tp, u32 ack) tp->snd_una = ack; } +/* If we update tp->rcv_nxt, also update tp->bytes_received */ +static void tcp_rcv_nxt_update(struct tcp_sock *tp, u32 seq) +{ + u32 delta = seq - tp->rcv_nxt; + + tp->bytes_received += delta; + tp->rcv_nxt = seq; +} + /* Update our send window. * * Window update algorithm, described in RFC793/RFC1122 (used in linux-2.2 @@ -4245,7 +4254,7 @@ static void tcp_ofo_queue(struct sock *sk) tail = skb_peek_tail(&sk->sk_receive_queue); eaten = tail && tcp_try_coalesce(sk, tail, skb, &fragstolen); - tp->rcv_nxt = TCP_SKB_CB(skb)->end_seq; + tcp_rcv_nxt_update(tp, TCP_SKB_CB(skb)->end_seq); if (!eaten) __skb_queue_tail(&sk->sk_receive_queue, skb); if (TCP_SKB_CB(skb)->tcp_flags & TCPHDR_FIN) @@ -4413,7 +4422,7 @@ static int __must_check tcp_queue_rcv(struct sock *sk, struct sk_buff *skb, int __skb_pull(skb, hdrlen); eaten = (tail && tcp_try_coalesce(sk, tail, skb, fragstolen)) ? 1 : 0; - tcp_sk(sk)->rcv_nxt = TCP_SKB_CB(skb)->end_seq; + tcp_rcv_nxt_update(tcp_sk(sk), TCP_SKB_CB(skb)->end_seq); if (!eaten) { __skb_queue_tail(&sk->sk_receive_queue, skb); skb_set_owner_r(skb, sk); @@ -4506,7 +4515,7 @@ queue_and_out: eaten = tcp_queue_rcv(sk, skb, 0, &fragstolen); } - tp->rcv_nxt = TCP_SKB_CB(skb)->end_seq; + tcp_rcv_nxt_update(tp, TCP_SKB_CB(skb)->end_seq); if (skb->len) tcp_event_data_recv(sk, skb); if (TCP_SKB_CB(skb)->tcp_flags & TCPHDR_FIN) @@ -5254,7 +5263,7 @@ void tcp_rcv_established(struct sock *sk, struct sk_buff *skb, tcp_rcv_rtt_measure_ts(sk, skb); __skb_pull(skb, tcp_header_len); - tp->rcv_nxt = TCP_SKB_CB(skb)->end_seq; + tcp_rcv_nxt_update(tp, TCP_SKB_CB(skb)->end_seq); NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_TCPHPHITSTOUSER); eaten = 1; } -- cgit v1.2.3 From 64f40ff5bbdb1b679fb3c4dbc8230d6517d2b8dc Mon Sep 17 00:00:00 2001 From: Eric Dumazet Date: Tue, 28 Apr 2015 16:23:48 -0700 Subject: tcp: prepare CC get_info() access from getsockopt() We would like that optional info provided by Congestion Control modules using netlink can also be read using getsockopt() This patch changes get_info() to put this information in a buffer, instead of skb, like tcp_get_info(), so that following patch can reuse this common infrastructure. Signed-off-by: Eric Dumazet Cc: Yuchung Cheng Cc: Neal Cardwell Acked-by: Neal Cardwell Acked-by: Daniel Borkmann Acked-by: Yuchung Cheng Signed-off-by: David S. Miller --- include/net/tcp.h | 5 ++++- include/uapi/linux/inet_diag.h | 4 ++++ net/ipv4/inet_diag.c | 8 +++++--- net/ipv4/tcp_dctcp.c | 20 ++++++++++---------- net/ipv4/tcp_illinois.c | 21 +++++++++++---------- net/ipv4/tcp_vegas.c | 19 ++++++++++--------- net/ipv4/tcp_vegas.h | 3 ++- 7 files changed, 46 insertions(+), 34 deletions(-) (limited to 'net/ipv4') diff --git a/include/net/tcp.h b/include/net/tcp.h index dd7b4ea6a10c..6d204f3f9df8 100644 --- a/include/net/tcp.h +++ b/include/net/tcp.h @@ -804,6 +804,8 @@ enum tcp_ca_ack_event_flags { /* Requires ECN/ECT set on all packets */ #define TCP_CONG_NEEDS_ECN 0x2 +union tcp_cc_info; + struct tcp_congestion_ops { struct list_head list; u32 key; @@ -829,7 +831,8 @@ struct tcp_congestion_ops { /* hook for packet ack accounting (optional) */ void (*pkts_acked)(struct sock *sk, u32 num_acked, s32 rtt_us); /* get info for inet_diag (optional) */ - int (*get_info)(struct sock *sk, u32 ext, struct sk_buff *skb); + size_t (*get_info)(struct sock *sk, u32 ext, int *attr, + union tcp_cc_info *info); char name[TCP_CA_NAME_MAX]; struct module *owner; diff --git a/include/uapi/linux/inet_diag.h b/include/uapi/linux/inet_diag.h index d65c0a09efd3..c7093c75bdd6 100644 --- a/include/uapi/linux/inet_diag.h +++ b/include/uapi/linux/inet_diag.h @@ -143,4 +143,8 @@ struct tcp_dctcp_info { __u32 dctcp_ab_tot; }; +union tcp_cc_info { + struct tcpvegas_info vegas; + struct tcp_dctcp_info dctcp; +}; #endif /* _UAPI_INET_DIAG_H_ */ diff --git a/net/ipv4/inet_diag.c b/net/ipv4/inet_diag.c index bb77ebdae3b3..4d32262c7502 100644 --- a/net/ipv4/inet_diag.c +++ b/net/ipv4/inet_diag.c @@ -224,14 +224,16 @@ int inet_sk_diag_fill(struct sock *sk, struct inet_connection_sock *icsk, handler->idiag_get_info(sk, r, info); if (sk->sk_state < TCP_TIME_WAIT) { - int err = 0; + union tcp_cc_info info; + size_t sz = 0; + int attr; rcu_read_lock(); ca_ops = READ_ONCE(icsk->icsk_ca_ops); if (ca_ops && ca_ops->get_info) - err = ca_ops->get_info(sk, ext, skb); + sz = ca_ops->get_info(sk, ext, &attr, &info); rcu_read_unlock(); - if (err < 0) + if (sz && nla_put(skb, attr, sz, &info) < 0) goto errout; } diff --git a/net/ipv4/tcp_dctcp.c b/net/ipv4/tcp_dctcp.c index 4376016f7fa5..4c41c1287197 100644 --- a/net/ipv4/tcp_dctcp.c +++ b/net/ipv4/tcp_dctcp.c @@ -277,7 +277,8 @@ static void dctcp_cwnd_event(struct sock *sk, enum tcp_ca_event ev) } } -static int dctcp_get_info(struct sock *sk, u32 ext, struct sk_buff *skb) +static size_t dctcp_get_info(struct sock *sk, u32 ext, int *attr, + union tcp_cc_info *info) { const struct dctcp *ca = inet_csk_ca(sk); @@ -286,18 +287,17 @@ static int dctcp_get_info(struct sock *sk, u32 ext, struct sk_buff *skb) */ if (ext & (1 << (INET_DIAG_DCTCPINFO - 1)) || ext & (1 << (INET_DIAG_VEGASINFO - 1))) { - struct tcp_dctcp_info info; - - memset(&info, 0, sizeof(info)); + memset(info, 0, sizeof(struct tcp_dctcp_info)); if (inet_csk(sk)->icsk_ca_ops != &dctcp_reno) { - info.dctcp_enabled = 1; - info.dctcp_ce_state = (u16) ca->ce_state; - info.dctcp_alpha = ca->dctcp_alpha; - info.dctcp_ab_ecn = ca->acked_bytes_ecn; - info.dctcp_ab_tot = ca->acked_bytes_total; + info->dctcp.dctcp_enabled = 1; + info->dctcp.dctcp_ce_state = (u16) ca->ce_state; + info->dctcp.dctcp_alpha = ca->dctcp_alpha; + info->dctcp.dctcp_ab_ecn = ca->acked_bytes_ecn; + info->dctcp.dctcp_ab_tot = ca->acked_bytes_total; } - return nla_put(skb, INET_DIAG_DCTCPINFO, sizeof(info), &info); + *attr = INET_DIAG_DCTCPINFO; + return sizeof(*info); } return 0; } diff --git a/net/ipv4/tcp_illinois.c b/net/ipv4/tcp_illinois.c index 67476f085e48..f71002e4db0b 100644 --- a/net/ipv4/tcp_illinois.c +++ b/net/ipv4/tcp_illinois.c @@ -300,24 +300,25 @@ static u32 tcp_illinois_ssthresh(struct sock *sk) } /* Extract info for Tcp socket info provided via netlink. */ -static int tcp_illinois_info(struct sock *sk, u32 ext, struct sk_buff *skb) +static size_t tcp_illinois_info(struct sock *sk, u32 ext, int *attr, + union tcp_cc_info *info) { const struct illinois *ca = inet_csk_ca(sk); if (ext & (1 << (INET_DIAG_VEGASINFO - 1))) { - struct tcpvegas_info info = { - .tcpv_enabled = 1, - .tcpv_rttcnt = ca->cnt_rtt, - .tcpv_minrtt = ca->base_rtt, - }; + info->vegas.tcpv_enabled = 1; + info->vegas.tcpv_rttcnt = ca->cnt_rtt; + info->vegas.tcpv_minrtt = ca->base_rtt; + info->vegas.tcpv_rtt = 0; - if (info.tcpv_rttcnt > 0) { + if (info->vegas.tcpv_rttcnt > 0) { u64 t = ca->sum_rtt; - do_div(t, info.tcpv_rttcnt); - info.tcpv_rtt = t; + do_div(t, info->vegas.tcpv_rttcnt); + info->vegas.tcpv_rtt = t; } - return nla_put(skb, INET_DIAG_VEGASINFO, sizeof(info), &info); + *attr = INET_DIAG_VEGASINFO; + return sizeof(struct tcpvegas_info); } return 0; } diff --git a/net/ipv4/tcp_vegas.c b/net/ipv4/tcp_vegas.c index c71a1b8f7bde..a6cea1d5e20d 100644 --- a/net/ipv4/tcp_vegas.c +++ b/net/ipv4/tcp_vegas.c @@ -286,18 +286,19 @@ static void tcp_vegas_cong_avoid(struct sock *sk, u32 ack, u32 acked) } /* Extract info for Tcp socket info provided via netlink. */ -int tcp_vegas_get_info(struct sock *sk, u32 ext, struct sk_buff *skb) +size_t tcp_vegas_get_info(struct sock *sk, u32 ext, int *attr, + union tcp_cc_info *info) { const struct vegas *ca = inet_csk_ca(sk); + if (ext & (1 << (INET_DIAG_VEGASINFO - 1))) { - struct tcpvegas_info info = { - .tcpv_enabled = ca->doing_vegas_now, - .tcpv_rttcnt = ca->cntRTT, - .tcpv_rtt = ca->baseRTT, - .tcpv_minrtt = ca->minRTT, - }; - - return nla_put(skb, INET_DIAG_VEGASINFO, sizeof(info), &info); + info->vegas.tcpv_enabled = ca->doing_vegas_now, + info->vegas.tcpv_rttcnt = ca->cntRTT, + info->vegas.tcpv_rtt = ca->baseRTT, + info->vegas.tcpv_minrtt = ca->minRTT, + + *attr = INET_DIAG_VEGASINFO; + return sizeof(struct tcpvegas_info); } return 0; } diff --git a/net/ipv4/tcp_vegas.h b/net/ipv4/tcp_vegas.h index e8a6b33cc61d..ef9da5306c68 100644 --- a/net/ipv4/tcp_vegas.h +++ b/net/ipv4/tcp_vegas.h @@ -19,6 +19,7 @@ void tcp_vegas_init(struct sock *sk); void tcp_vegas_state(struct sock *sk, u8 ca_state); void tcp_vegas_pkts_acked(struct sock *sk, u32 cnt, s32 rtt_us); void tcp_vegas_cwnd_event(struct sock *sk, enum tcp_ca_event event); -int tcp_vegas_get_info(struct sock *sk, u32 ext, struct sk_buff *skb); +size_t tcp_vegas_get_info(struct sock *sk, u32 ext, int *attr, + union tcp_cc_info *info); #endif /* __TCP_VEGAS_H */ -- cgit v1.2.3 From 6e9250f59ef9efb932c84850cd221f22c2a03c4a Mon Sep 17 00:00:00 2001 From: Eric Dumazet Date: Tue, 28 Apr 2015 16:23:49 -0700 Subject: tcp: add TCP_CC_INFO socket option Some Congestion Control modules can provide per flow information, but current way to get this information is to use netlink. Like TCP_INFO, let's add TCP_CC_INFO so that applications can issue a getsockopt() if they have a socket file descriptor, instead of playing complex netlink games. Sample usage would be : union tcp_cc_info info; socklen_t len = sizeof(info); if (getsockopt(fd, SOL_TCP, TCP_CC_INFO, &info, &len) == -1) Signed-off-by: Eric Dumazet Cc: Yuchung Cheng Cc: Neal Cardwell Acked-by: Neal Cardwell Acked-by: Daniel Borkmann Acked-by: Yuchung Cheng Signed-off-by: David S. Miller --- include/uapi/linux/tcp.h | 1 + net/ipv4/tcp.c | 21 +++++++++++++++++++++ 2 files changed, 22 insertions(+) (limited to 'net/ipv4') diff --git a/include/uapi/linux/tcp.h b/include/uapi/linux/tcp.h index a48f93f3207b..faa72f4fa547 100644 --- a/include/uapi/linux/tcp.h +++ b/include/uapi/linux/tcp.h @@ -112,6 +112,7 @@ enum { #define TCP_FASTOPEN 23 /* Enable FastOpen on listeners */ #define TCP_TIMESTAMP 24 #define TCP_NOTSENT_LOWAT 25 /* limit number of unsent bytes in write queue */ +#define TCP_CC_INFO 26 /* Get Congestion Control (optional) info */ struct tcp_repair_opt { __u32 opt_code; diff --git a/net/ipv4/tcp.c b/net/ipv4/tcp.c index 99fcc0b22c92..46efa03d2b11 100644 --- a/net/ipv4/tcp.c +++ b/net/ipv4/tcp.c @@ -252,6 +252,7 @@ #include #include #include +#include #include #include #include @@ -2739,6 +2740,26 @@ static int do_tcp_getsockopt(struct sock *sk, int level, return -EFAULT; return 0; } + case TCP_CC_INFO: { + const struct tcp_congestion_ops *ca_ops; + union tcp_cc_info info; + size_t sz = 0; + int attr; + + if (get_user(len, optlen)) + return -EFAULT; + + ca_ops = icsk->icsk_ca_ops; + if (ca_ops && ca_ops->get_info) + sz = ca_ops->get_info(sk, ~0U, &attr, &info); + + len = min_t(unsigned int, len, sz); + if (put_user(len, optlen)) + return -EFAULT; + if (copy_to_user(optval, &info, len)) + return -EFAULT; + return 0; + } case TCP_QUICKACK: val = !icsk->icsk_ack.pingpong; break; -- cgit v1.2.3 From 9dac8835440622c2f84591673969d510ce198c11 Mon Sep 17 00:00:00 2001 From: Yuchung Cheng Date: Wed, 29 Apr 2015 11:28:30 -0700 Subject: tcp: update reordering first before detecting loss tcp_mark_lost_retrans is not used when FACK is disabled. Since tcp_update_reordering may disable FACK, it should be called first before tcp_mark_lost_retrans. Signed-off-by: Yuchung Cheng Signed-off-by: Nandita Dukkipati Signed-off-by: Neal Cardwell Signed-off-by: David S. Miller --- net/ipv4/tcp_input.c | 6 ++---- 1 file changed, 2 insertions(+), 4 deletions(-) (limited to 'net/ipv4') diff --git a/net/ipv4/tcp_input.c b/net/ipv4/tcp_input.c index 7e6962bcfc30..bc790ea9960f 100644 --- a/net/ipv4/tcp_input.c +++ b/net/ipv4/tcp_input.c @@ -1820,14 +1820,12 @@ advance_sp: for (j = 0; j < used_sacks; j++) tp->recv_sack_cache[i++] = sp[j]; - tcp_mark_lost_retrans(sk); - - tcp_verify_left_out(tp); - if ((state.reord < tp->fackets_out) && ((inet_csk(sk)->icsk_ca_state != TCP_CA_Loss) || tp->undo_marker)) tcp_update_reordering(sk, tp->fackets_out - state.reord, 0); + tcp_mark_lost_retrans(sk); + tcp_verify_left_out(tp); out: #if FASTRETRANS_DEBUG > 0 -- cgit v1.2.3