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_input.c | 13 +++++++++++-- 1 file changed, 11 insertions(+), 2 deletions(-) (limited to 'net/ipv4/tcp_input.c') 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/tcp_input.c') 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 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/tcp_input.c') 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 From da34ac7626b571d262f92b93f11eb32dd58d9c4e Mon Sep 17 00:00:00 2001 From: Yuchung Cheng Date: Mon, 18 May 2015 12:31:44 -0700 Subject: tcp: only undo on partial ACKs in CA_Loss Undo based on TCP timestamps should only happen on ACKs that advance SND_UNA, according to the Eifel algorithm in RFC 3522: Section 3.2: (4) If the value of the Timestamp Echo Reply field of the acceptable ACK's Timestamps option is smaller than the value of RetransmitTS, then proceed to step (5), Section Terminology: We use the term 'acceptable ACK' as defined in [RFC793]. That is an ACK that acknowledges previously unacknowledged data. This is because upon receiving an out-of-order packet, the receiver returns the last timestamp that advances RCV_NXT, not the current timestamp of the packet in the DUPACK. Without checking the flag, the DUPACK will cause tcp_packet_delayed() to return true and tcp_try_undo_loss() will revert cwnd reduction. Note that we check the condition in CA_Recovery already by only calling tcp_try_undo_partial() if FLAG_SND_UNA_ADVANCED is set or tcp_try_undo_recovery() if snd_una crosses high_seq. Signed-off-by: Yuchung Cheng Signed-off-by: Neal Cardwell Signed-off-by: David S. Miller --- net/ipv4/tcp_input.c | 9 ++++++--- 1 file changed, 6 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 bc790ea9960f..9faf775a8c4a 100644 --- a/net/ipv4/tcp_input.c +++ b/net/ipv4/tcp_input.c @@ -2698,11 +2698,16 @@ static void tcp_process_loss(struct sock *sk, int flag, bool is_dupack) struct tcp_sock *tp = tcp_sk(sk); bool recovered = !before(tp->snd_una, tp->high_seq); + if ((flag & FLAG_SND_UNA_ADVANCED) && + tcp_try_undo_loss(sk, false)) + return; + if (tp->frto) { /* F-RTO RFC5682 sec 3.1 (sack enhanced version). */ /* Step 3.b. A timeout is spurious if not all data are * lost, i.e., never-retransmitted data are (s)acked. */ - if (tcp_try_undo_loss(sk, flag & FLAG_ORIG_SACK_ACKED)) + if ((flag & FLAG_ORIG_SACK_ACKED) && + tcp_try_undo_loss(sk, true)) return; if (after(tp->snd_nxt, tp->high_seq) && @@ -2732,8 +2737,6 @@ static void tcp_process_loss(struct sock *sk, int flag, bool is_dupack) else if (flag & FLAG_SND_UNA_ADVANCED) tcp_reset_reno_sack(tp); } - if (tcp_try_undo_loss(sk, false)) - return; tcp_xmit_retransmit_queue(sk); } -- cgit v1.2.3 From b7b0ed910cd8450db6d98cd4361c644bb1c88412 Mon Sep 17 00:00:00 2001 From: Yuchung Cheng Date: Mon, 18 May 2015 12:31:45 -0700 Subject: tcp: don't over-send F-RTO probes After sending the new data packets to probe (step 2), F-RTO may incorrectly send more probes if the next ACK advances SND_UNA and does not sack new packet. However F-RTO RFC 5682 probes at most once. This bug may cause sender to always send new data instead of repairing holes, inducing longer HoL blocking on the receiver for the application. Signed-off-by: Yuchung Cheng Signed-off-by: Neal Cardwell Signed-off-by: David S. Miller --- net/ipv4/tcp_input.c | 6 +++--- 1 file changed, 3 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 9faf775a8c4a..243d674b3ef5 100644 --- a/net/ipv4/tcp_input.c +++ b/net/ipv4/tcp_input.c @@ -2710,9 +2710,9 @@ static void tcp_process_loss(struct sock *sk, int flag, bool is_dupack) tcp_try_undo_loss(sk, true)) return; - if (after(tp->snd_nxt, tp->high_seq) && - (flag & FLAG_DATA_SACKED || is_dupack)) { - tp->frto = 0; /* Loss was real: 2nd part of step 3.a */ + if (after(tp->snd_nxt, tp->high_seq)) { + if (flag & FLAG_DATA_SACKED || is_dupack) + tp->frto = 0; /* Step 3.a. loss was real */ } else if (flag & FLAG_SND_UNA_ADVANCED && !recovered) { tp->high_seq = tp->snd_nxt; __tcp_push_pending_frames(sk, tcp_current_mss(sk), -- cgit v1.2.3 From d654976cbf852ee20612ee10dbe57cdacda9f452 Mon Sep 17 00:00:00 2001 From: Eric Dumazet Date: Thu, 21 May 2015 21:51:19 -0700 Subject: tcp: fix a potential deadlock in tcp_get_info() Taking socket spinlock in tcp_get_info() can deadlock, as inet_diag_dump_icsk() holds the &hashinfo->ehash_locks[i], while packet processing can use the reverse locking order. We could avoid this locking for TCP_LISTEN states, but lockdep would certainly get confused as all TCP sockets share same lockdep classes. [ 523.722504] ====================================================== [ 523.728706] [ INFO: possible circular locking dependency detected ] [ 523.734990] 4.1.0-dbg-DEV #1676 Not tainted [ 523.739202] ------------------------------------------------------- [ 523.745474] ss/18032 is trying to acquire lock: [ 523.750002] (slock-AF_INET){+.-...}, at: [] tcp_get_info+0x2c4/0x360 [ 523.758129] [ 523.758129] but task is already holding lock: [ 523.763968] (&(&hashinfo->ehash_locks[i])->rlock){+.-...}, at: [] inet_diag_dump_icsk+0x1d5/0x6c0 [ 523.774661] [ 523.774661] which lock already depends on the new lock. [ 523.774661] [ 523.782850] [ 523.782850] the existing dependency chain (in reverse order) is: [ 523.790326] -> #1 (&(&hashinfo->ehash_locks[i])->rlock){+.-...}: [ 523.796599] [] lock_acquire+0xbb/0x270 [ 523.802565] [] _raw_spin_lock+0x38/0x50 [ 523.808628] [] __inet_hash_nolisten+0x78/0x110 [ 523.815273] [] tcp_v4_syn_recv_sock+0x24b/0x350 [ 523.822067] [] tcp_check_req+0x3c1/0x500 [ 523.828199] [] tcp_v4_do_rcv+0x239/0x3d0 [ 523.834331] [] tcp_v4_rcv+0xa8e/0xc10 [ 523.840202] [] ip_local_deliver_finish+0x133/0x3e0 [ 523.847214] [] ip_local_deliver+0xaa/0xc0 [ 523.853440] [] ip_rcv_finish+0x168/0x5c0 [ 523.859624] [] ip_rcv+0x307/0x420 Lets use u64_sync infrastructure instead. As a bonus, 64bit arches get optimized, as these are nop for them. Fixes: 0df48c26d841 ("tcp: add tcpi_bytes_acked to tcp_info") Signed-off-by: Eric Dumazet Signed-off-by: David S. Miller --- include/linux/tcp.h | 2 ++ net/ipv4/tcp.c | 11 +++++++---- net/ipv4/tcp_fastopen.c | 4 ++++ net/ipv4/tcp_input.c | 4 ++++ 4 files changed, 17 insertions(+), 4 deletions(-) (limited to 'net/ipv4/tcp_input.c') diff --git a/include/linux/tcp.h b/include/linux/tcp.h index 3b2911502a8c..e8bbf403618f 100644 --- a/include/linux/tcp.h +++ b/include/linux/tcp.h @@ -158,6 +158,8 @@ struct tcp_sock { * sum(delta(snd_una)), or how many bytes * were acked. */ + struct u64_stats_sync syncp; /* protects 64bit vars (cf tcp_get_info()) */ + u32 snd_una; /* First byte we want an ack for */ u32 snd_sml; /* Last byte of the most recently transmitted small packet */ u32 rcv_tstamp; /* timestamp of last received ACK (for keepalives) */ diff --git a/net/ipv4/tcp.c b/net/ipv4/tcp.c index 46efa03d2b11..f1377f2a0472 100644 --- a/net/ipv4/tcp.c +++ b/net/ipv4/tcp.c @@ -402,6 +402,7 @@ void tcp_init_sock(struct sock *sk) tp->snd_ssthresh = TCP_INFINITE_SSTHRESH; tp->snd_cwnd_clamp = ~0; tp->mss_cache = TCP_MSS_DEFAULT; + u64_stats_init(&tp->syncp); tp->reordering = sysctl_tcp_reordering; tcp_enable_early_retrans(tp); @@ -2598,6 +2599,7 @@ 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); u32 now = tcp_time_stamp; + unsigned int start; u32 rate; memset(info, 0, sizeof(*info)); @@ -2665,10 +2667,11 @@ void tcp_get_info(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; - info->tcpi_bytes_received = tp->bytes_received; - spin_unlock_bh(&sk->sk_lock.slock); + do { + start = u64_stats_fetch_begin_irq(&tp->syncp); + info->tcpi_bytes_acked = tp->bytes_acked; + info->tcpi_bytes_received = tp->bytes_received; + } while (u64_stats_fetch_retry_irq(&tp->syncp, start)); } EXPORT_SYMBOL_GPL(tcp_get_info); diff --git a/net/ipv4/tcp_fastopen.c b/net/ipv4/tcp_fastopen.c index 3c673d5e6cff..46b087a27503 100644 --- a/net/ipv4/tcp_fastopen.c +++ b/net/ipv4/tcp_fastopen.c @@ -206,6 +206,10 @@ 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; + + /* u64_stats_update_begin(&tp->syncp) not needed here, + * as we certainly are not changing upper 32bit value (0) + */ 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 243d674b3ef5..c9ab964189a0 100644 --- a/net/ipv4/tcp_input.c +++ b/net/ipv4/tcp_input.c @@ -3286,7 +3286,9 @@ static void tcp_snd_una_update(struct tcp_sock *tp, u32 ack) { u32 delta = ack - tp->snd_una; + u64_stats_update_begin(&tp->syncp); tp->bytes_acked += delta; + u64_stats_update_end(&tp->syncp); tp->snd_una = ack; } @@ -3295,7 +3297,9 @@ static void tcp_rcv_nxt_update(struct tcp_sock *tp, u32 seq) { u32 delta = seq - tp->rcv_nxt; + u64_stats_update_begin(&tp->syncp); tp->bytes_received += delta; + u64_stats_update_end(&tp->syncp); tp->rcv_nxt = seq; } -- cgit v1.2.3