* [PATCH net-next 1/2] tcp: call tcp_try_undo_recovery when an RTOd TFO SYNACK is ACKed
2023-09-12 2:33 [PATCH net-next 0/2] tcp: new TCP_INFO stats for RTO events Aananth V
@ 2023-09-12 2:33 ` Aananth V
2023-09-12 2:33 ` [PATCH net-next 2/2] tcp: new TCP_INFO stats for RTO events Aananth V
1 sibling, 0 replies; 6+ messages in thread
From: Aananth V @ 2023-09-12 2:33 UTC (permalink / raw)
To: David Miller, Eric Dumazet, Jakub Kicinski
Cc: netdev, Aananth V, Neal Cardwell, Yuchung Cheng
For passive TCP Fast Open sockets that had SYN/ACK timeout and did not
send more data in SYN_RECV, upon receiving the final ACK in 3WHS, the
congestion state may awkwardly stay in CA_Loss mode unless the CA state
was undone due to TCP timestamp checks. However, if
tcp_rcv_synrecv_state_fastopen() decides not to undo, then we should
enter CA_Open, because at that point we have received an ACK covering
the retransmitted SYNACKs. Currently, the icsk_ca_state is only set to
CA_Open after we receive an ACK for a data-packet. This is because
tcp_ack does not call tcp_fastretrans_alert (and tcp_process_loss) if
!prior_packets
Note that tcp_process_loss() calls tcp_try_undo_recovery(), so having
tcp_rcv_synrecv_state_fastopen() decide that if we're in CA_Loss we
should call tcp_try_undo_recovery() is consistent with that, and
low risk.
Fixes: dad8cea7add9 ("tcp: fix TFO SYNACK undo to avoid double-timestamp-undo")
Signed-off-by: Aananth V <aananthv@google.com>
Signed-off-by: Neal Cardwell <ncardwell@google.com>
Signed-off-by: Yuchung Cheng <ycheng@google.com>
Reviewed-by: Eric Dumazet <edumazet@google.com>
---
net/ipv4/tcp_input.c | 9 +++++----
1 file changed, 5 insertions(+), 4 deletions(-)
diff --git a/net/ipv4/tcp_input.c b/net/ipv4/tcp_input.c
index 06fe1cf645d5..fe2ab0db2eb7 100644
--- a/net/ipv4/tcp_input.c
+++ b/net/ipv4/tcp_input.c
@@ -6436,22 +6436,23 @@ static int tcp_rcv_synsent_state_process(struct sock *sk, struct sk_buff *skb,
static void tcp_rcv_synrecv_state_fastopen(struct sock *sk)
{
+ struct tcp_sock *tp = tcp_sk(sk);
struct request_sock *req;
/* If we are still handling the SYNACK RTO, see if timestamp ECR allows
* undo. If peer SACKs triggered fast recovery, we can't undo here.
*/
- if (inet_csk(sk)->icsk_ca_state == TCP_CA_Loss)
- tcp_try_undo_loss(sk, false);
+ if (inet_csk(sk)->icsk_ca_state == TCP_CA_Loss && !tp->packets_out)
+ tcp_try_undo_recovery(sk);
/* Reset rtx states to prevent spurious retransmits_timed_out() */
- tcp_sk(sk)->retrans_stamp = 0;
+ tp->retrans_stamp = 0;
inet_csk(sk)->icsk_retransmits = 0;
/* Once we leave TCP_SYN_RECV or TCP_FIN_WAIT_1,
* we no longer need req so release it.
*/
- req = rcu_dereference_protected(tcp_sk(sk)->fastopen_rsk,
+ req = rcu_dereference_protected(tp->fastopen_rsk,
lockdep_sock_is_held(sk));
reqsk_fastopen_remove(sk, req, false);
--
2.42.0.283.g2d96d420d3-goog
^ permalink raw reply related [flat|nested] 6+ messages in thread* [PATCH net-next 2/2] tcp: new TCP_INFO stats for RTO events
2023-09-12 2:33 [PATCH net-next 0/2] tcp: new TCP_INFO stats for RTO events Aananth V
2023-09-12 2:33 ` [PATCH net-next 1/2] tcp: call tcp_try_undo_recovery when an RTOd TFO SYNACK is ACKed Aananth V
@ 2023-09-12 2:33 ` Aananth V
2023-09-14 9:02 ` Paolo Abeni
1 sibling, 1 reply; 6+ messages in thread
From: Aananth V @ 2023-09-12 2:33 UTC (permalink / raw)
To: David Miller, Eric Dumazet, Jakub Kicinski
Cc: netdev, Aananth V, Neal Cardwell, Yuchung Cheng
The 2023 SIGCOMM paper "Improving Network Availability with Protective
ReRoute" has indicated Linux TCP's RTO-triggered txhash rehashing can
effectively reduce application disruption during outages. To better
measure the efficacy of this feature, this patch adds three more
detailed stats during RTO recovery and exports via TCP_INFO.
Applications and monitoring systems can leverage this data to measure
the network path diversity and end-to-end repair latency during network
outages to improve their network infrastructure.
The following counters are added to tcp_sock in order to track RTO
events over the lifetime of a TCP socket.
1. u16 total_rto - Counts the total number of RTO timeouts.
2. u16 total_rto_recoveries - Counts the total number of RTO recoveries.
3. u32 total_rto_time - Counts the total time spent (ms) in RTO
recoveries. (time spent in CA_Loss and
CA_Recovery states)
To compute total_rto_time, we add a new u32 rto_stamp field to
tcp_sock. rto_stamp records the start timestamp (ms) of the last RTO
recovery (CA_Loss).
Corresponding fields are also added to the tcp_info struct.
Signed-off-by: Aananth V <aananthv@google.com>
Signed-off-by: Neal Cardwell <ncardwell@google.com>
Signed-off-by: Yuchung Cheng <ycheng@google.com>
Reviewed-by: Eric Dumazet <edumazet@google.com>
---
include/linux/tcp.h | 8 ++++++++
include/uapi/linux/tcp.h | 12 ++++++++++++
net/ipv4/tcp.c | 9 +++++++++
net/ipv4/tcp_input.c | 15 +++++++++++++++
net/ipv4/tcp_minisocks.c | 4 ++++
net/ipv4/tcp_timer.c | 17 +++++++++++++++--
6 files changed, 63 insertions(+), 2 deletions(-)
diff --git a/include/linux/tcp.h b/include/linux/tcp.h
index 3c5efeeb024f..91731cd3bc3d 100644
--- a/include/linux/tcp.h
+++ b/include/linux/tcp.h
@@ -377,6 +377,14 @@ struct tcp_sock {
* Total data bytes retransmitted
*/
u32 total_retrans; /* Total retransmits for entire connection */
+ u32 rto_stamp; /* Start time (ms) of the last RTO recovery (CA_Loss) */
+ u16 total_rto; /* Total number of RTO timeouts, including
+ * SYN/SYN-ACK and recurring timeouts.
+ */
+ u16 total_rto_recoveries; /* Total number of RTO recoveries,
+ * including any unfinished recovery.
+ */
+ u32 total_rto_time; /* ms spent in (completed) RTO recoveries. */
u32 urg_seq; /* Seq of received urgent pointer */
unsigned int keepalive_time; /* time before keep alive takes place */
diff --git a/include/uapi/linux/tcp.h b/include/uapi/linux/tcp.h
index 879eeb0a084b..d1d08da6331a 100644
--- a/include/uapi/linux/tcp.h
+++ b/include/uapi/linux/tcp.h
@@ -289,6 +289,18 @@ struct tcp_info {
*/
__u32 tcpi_rehash; /* PLB or timeout triggered rehash attempts */
+
+ __u16 tcpi_total_rto; /* Total number of RTO timeouts, including
+ * SYN/SYN-ACK and recurring timeouts.
+ */
+ __u16 tcpi_total_rto_recoveries; /* Total number of RTO
+ * recoveries, including any
+ * unfinished recovery.
+ */
+ __u32 tcpi_total_rto_time; /* Total time spent in RTO recoveries
+ * in milliseconds, including any
+ * unfinished recovery.
+ */
};
/* netlink attributes types for SCM_TIMESTAMPING_OPT_STATS */
diff --git a/net/ipv4/tcp.c b/net/ipv4/tcp.c
index 0c3040a63ebd..69b8d7073708 100644
--- a/net/ipv4/tcp.c
+++ b/net/ipv4/tcp.c
@@ -3818,6 +3818,15 @@ void tcp_get_info(struct sock *sk, struct tcp_info *info)
info->tcpi_rcv_wnd = tp->rcv_wnd;
info->tcpi_rehash = tp->plb_rehash + tp->timeout_rehash;
info->tcpi_fastopen_client_fail = tp->fastopen_client_fail;
+
+ info->tcpi_total_rto = tp->total_rto;
+ info->tcpi_total_rto_recoveries = tp->total_rto_recoveries;
+ info->tcpi_total_rto_time = tp->total_rto_time;
+ if (tp->rto_stamp) {
+ info->tcpi_total_rto_time += tcp_time_stamp_raw() -
+ tp->rto_stamp;
+ }
+
unlock_sock_fast(sk, slow);
}
EXPORT_SYMBOL_GPL(tcp_get_info);
diff --git a/net/ipv4/tcp_input.c b/net/ipv4/tcp_input.c
index fe2ab0db2eb7..d04f501a7590 100644
--- a/net/ipv4/tcp_input.c
+++ b/net/ipv4/tcp_input.c
@@ -2088,6 +2088,10 @@ void tcp_clear_retrans(struct tcp_sock *tp)
tp->undo_marker = 0;
tp->undo_retrans = -1;
tp->sacked_out = 0;
+ tp->rto_stamp = 0;
+ tp->total_rto = 0;
+ tp->total_rto_recoveries = 0;
+ tp->total_rto_time = 0;
}
static inline void tcp_init_undo(struct tcp_sock *tp)
@@ -2825,6 +2829,14 @@ void tcp_enter_recovery(struct sock *sk, bool ece_ack)
tcp_set_ca_state(sk, TCP_CA_Recovery);
}
+static inline void tcp_update_rto_time(struct tcp_sock *tp)
+{
+ if (tp->rto_stamp) {
+ tp->total_rto_time += tcp_time_stamp(tp) - tp->rto_stamp;
+ tp->rto_stamp = 0;
+ }
+}
+
/* Process an ACK in CA_Loss state. Move to CA_Open if lost data are
* recovered or spurious. Otherwise retransmits more on partial ACKs.
*/
@@ -3029,6 +3041,8 @@ static void tcp_fastretrans_alert(struct sock *sk, const u32 prior_snd_una,
break;
case TCP_CA_Loss:
tcp_process_loss(sk, flag, num_dupack, rexmit);
+ if (icsk->icsk_ca_state != TCP_CA_Loss)
+ tcp_update_rto_time(tp);
tcp_identify_packet_loss(sk, ack_flag);
if (!(icsk->icsk_ca_state == TCP_CA_Open ||
(*ack_flag & FLAG_LOST_RETRANS)))
@@ -6446,6 +6460,7 @@ static void tcp_rcv_synrecv_state_fastopen(struct sock *sk)
tcp_try_undo_recovery(sk);
/* Reset rtx states to prevent spurious retransmits_timed_out() */
+ tcp_update_rto_time(tp);
tp->retrans_stamp = 0;
inet_csk(sk)->icsk_retransmits = 0;
diff --git a/net/ipv4/tcp_minisocks.c b/net/ipv4/tcp_minisocks.c
index b98d476f1594..eee8ab1bfa0e 100644
--- a/net/ipv4/tcp_minisocks.c
+++ b/net/ipv4/tcp_minisocks.c
@@ -565,6 +565,10 @@ struct sock *tcp_create_openreq_child(const struct sock *sk,
newtp->undo_marker = treq->snt_isn;
newtp->retrans_stamp = div_u64(treq->snt_synack,
USEC_PER_SEC / TCP_TS_HZ);
+ newtp->total_rto = req->num_timeout;
+ newtp->total_rto_recoveries = 1;
+ newtp->total_rto_time = tcp_time_stamp_raw() -
+ newtp->retrans_stamp;
}
newtp->tsoffset = treq->ts_off;
#ifdef CONFIG_TCP_MD5SIG
diff --git a/net/ipv4/tcp_timer.c b/net/ipv4/tcp_timer.c
index 984ab4a0421e..d630db3d899a 100644
--- a/net/ipv4/tcp_timer.c
+++ b/net/ipv4/tcp_timer.c
@@ -415,6 +415,19 @@ abort: tcp_write_err(sk);
}
}
+static inline void tcp_update_rto_stats(struct sock *sk)
+{
+ struct inet_connection_sock *icsk = inet_csk(sk);
+ struct tcp_sock *tp = tcp_sk(sk);
+
+ if (!icsk->icsk_retransmits) {
+ tp->total_rto_recoveries++;
+ tp->rto_stamp = tcp_time_stamp(tp);
+ }
+ icsk->icsk_retransmits++;
+ tp->total_rto++;
+}
+
/*
* Timer for Fast Open socket to retransmit SYNACK. Note that the
* sk here is the child socket, not the parent (listener) socket.
@@ -447,7 +460,7 @@ static void tcp_fastopen_synack_timer(struct sock *sk, struct request_sock *req)
*/
inet_rtx_syn_ack(sk, req);
req->num_timeout++;
- icsk->icsk_retransmits++;
+ tcp_update_rto_stats(sk);
if (!tp->retrans_stamp)
tp->retrans_stamp = tcp_time_stamp(tp);
inet_csk_reset_xmit_timer(sk, ICSK_TIME_RETRANS,
@@ -575,7 +588,7 @@ void tcp_retransmit_timer(struct sock *sk)
tcp_enter_loss(sk);
- icsk->icsk_retransmits++;
+ tcp_update_rto_stats(sk);
if (tcp_retransmit_skb(sk, tcp_rtx_queue_head(sk), 1) > 0) {
/* Retransmission failed because of local congestion,
* Let senders fight for local resources conservatively.
--
2.42.0.283.g2d96d420d3-goog
^ permalink raw reply related [flat|nested] 6+ messages in thread