[PATCH net-next 2/2] tcp: add exponential backoff in __tcp_send_ack()
From: Eric Dumazet <hidden>
Date: 2020-09-30 12:55:10
Subsystem:
networking [general], networking [tcp], the rest · Maintainers:
"David S. Miller", Eric Dumazet, Jakub Kicinski, Paolo Abeni, Neal Cardwell, Linus Torvalds
From: Eric Dumazet <edumazet@google.com> Whenever host is under very high memory pressure, __tcp_send_ack() skb allocation fails, and we setup a 200 ms (TCP_DELACK_MAX) timer before retrying. On hosts with high number of TCP sockets, we can spend considerable amount of cpu cycles in these attempts, add high pressure on various spinlocks in mm-layer, ultimately blocking threads attempting to free space from making any progress. This patch adds standard exponential backoff to avoid adding fuel to the fire. Signed-off-by: Eric Dumazet <edumazet@google.com> --- include/net/inet_connection_sock.h | 3 ++- net/ipv4/tcp_output.c | 11 ++++++++--- 2 files changed, 10 insertions(+), 4 deletions(-)
diff --git a/include/net/inet_connection_sock.h b/include/net/inet_connection_sock.h
index 79875f976190750819948425e63dd0309c699050..7338b3865a2a3d278dc27c0167bba1b966bbda9f 100644
--- a/include/net/inet_connection_sock.h
+++ b/include/net/inet_connection_sock.h@@ -110,7 +110,7 @@ struct inet_connection_sock { __u8 pending; /* ACK is pending */ __u8 quick; /* Scheduled number of quick acks */ __u8 pingpong; /* The session is interactive */ - /* one byte hole. */ + __u8 retry; /* Number of attempts */ __u32 ato; /* Predicted tick of soft clock */ unsigned long timeout; /* Currently scheduled timeout */ __u32 lrcvtime; /* timestamp of last received data packet */
@@ -199,6 +199,7 @@ static inline void inet_csk_clear_xmit_timer(struct sock *sk, const int what) #endif } else if (what == ICSK_TIME_DACK) { icsk->icsk_ack.pending = 0; + icsk->icsk_ack.retry = 0; #ifdef INET_CSK_CLEAR_TIMERS sk_stop_timer(sk, &icsk->icsk_delack_timer); #endif
diff --git a/net/ipv4/tcp_output.c b/net/ipv4/tcp_output.c
index 6bd4e383030ea20441332a30e98fbda8cd90f84a..bf48cd73e96787a14b6f9af8beddb1067a7cb8dc 100644
--- a/net/ipv4/tcp_output.c
+++ b/net/ipv4/tcp_output.c@@ -3941,10 +3941,15 @@ void __tcp_send_ack(struct sock *sk, u32 rcv_nxt) buff = alloc_skb(MAX_TCP_HEADER, sk_gfp_mask(sk, GFP_ATOMIC | __GFP_NOWARN)); if (unlikely(!buff)) { + struct inet_connection_sock *icsk = inet_csk(sk); + unsigned long delay; + + delay = TCP_DELACK_MAX << icsk->icsk_ack.retry; + if (delay < TCP_RTO_MAX) + icsk->icsk_ack.retry++; inet_csk_schedule_ack(sk); - inet_csk(sk)->icsk_ack.ato = TCP_ATO_MIN; - inet_csk_reset_xmit_timer(sk, ICSK_TIME_DACK, - TCP_DELACK_MAX, TCP_RTO_MAX); + icsk->icsk_ack.ato = TCP_ATO_MIN; + inet_csk_reset_xmit_timer(sk, ICSK_TIME_DACK, delay, TCP_RTO_MAX); return; }
--
2.28.0.806.g8561365e88-goog