nagai/kernel
3
0
Fork 0
forked from mirrors/linux
Code Pull requests Activity

tcp: tsq: no longer use limit_output_bytes for paced flows

Browse source 
FQ pacing guarantees that paced packets queued by one flow do not
add head-of-line blocking for other flows.

After TCP GSO conversion, increasing limit_output_bytes to 1 MB is safe,
since this maps to 16 skbs at most in qdisc or device queues.
(or slightly more if some drivers lower {gso_max_segs|size})

We still can queue at most 1 ms worth of traffic (this can be scaled
by wifi drivers if they need to)

Tested:

# ethtool -c eth0 | egrep "tx-usecs:|tx-frames:" # 40 Gbit mlx4 NIC
tx-usecs: 16
tx-frames: 16
# tc qdisc replace dev eth0 root fq
# for f in {1..10};do netperf -P0 -H lpaa24,6 -o THROUGHPUT;done

Before patch:
27711
26118
27107
27377
27712
27388
27340
27117
27278
27509

After patch:
37434
36949
36658
36998
37711
37291
37605
36659
36544
37349

Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This commit is contained in:
Eric Dumazet 2018-11-11 07:34:28 -08:00 committed by David S. Miller
parent 83afb36a70
commit c73e5807e4
3 changed files with 6 additions and 5 deletions
repo.diff.show_diff_stats Download patch file Download diff file Expand all files Collapse all files

2
Documentation/networking/ip-sysctl.txt
View file

@ -759,7 +759,7 @@ tcp_limit_output_bytes - INTEGER
flows, for typical pfifo_fast qdiscs. tcp_limit_output_bytes flows, for typical pfifo_fast qdiscs. tcp_limit_output_bytes
limits the number of bytes on qdisc or device to reduce artificial limits the number of bytes on qdisc or device to reduce artificial
RTT/cwnd and reduce bufferbloat. RTT/cwnd and reduce bufferbloat.
Default: 262144 Default: 1048576 (16 * 65536)
tcp_challenge_ack_limit - INTEGER tcp_challenge_ack_limit - INTEGER
Limits number of Challenge ACK sent per second, as recommended Limits number of Challenge ACK sent per second, as recommended

4
net/ipv4/tcp_ipv4.c
View file

@ -2574,8 +2574,8 @@ static int __net_init tcp_sk_init(struct net *net)
* which are too large can cause TCP streams to be bursty. * which are too large can cause TCP streams to be bursty.
*/ */
net->ipv4.sysctl_tcp_tso_win_divisor = 3; net->ipv4.sysctl_tcp_tso_win_divisor = 3;
/* Default TSQ limit of four TSO segments */ /* Default TSQ limit of 16 TSO segments */
net->ipv4.sysctl_tcp_limit_output_bytes = 262144; net->ipv4.sysctl_tcp_limit_output_bytes = 16 * 65536;
/* rfc5961 challenge ack rate limiting */ /* rfc5961 challenge ack rate limiting */
net->ipv4.sysctl_tcp_challenge_ack_limit = 1000; net->ipv4.sysctl_tcp_challenge_ack_limit = 1000;
net->ipv4.sysctl_tcp_min_tso_segs = 2; net->ipv4.sysctl_tcp_min_tso_segs = 2;

1
net/ipv4/tcp_output.c
View file

@ -2220,6 +2220,7 @@ static bool tcp_small_queue_check(struct sock *sk, const struct sk_buff *skb,
limit = max_t(unsigned long, limit = max_t(unsigned long,
2 * skb->truesize, 2 * skb->truesize,
sk->sk_pacing_rate >> sk->sk_pacing_shift); sk->sk_pacing_rate >> sk->sk_pacing_shift);
if (sk->sk_pacing_status == SK_PACING_NONE)
limit = min_t(unsigned long, limit, limit = min_t(unsigned long, limit,
sock_net(sk)->ipv4.sysctl_tcp_limit_output_bytes); sock_net(sk)->ipv4.sysctl_tcp_limit_output_bytes);
limit <<= factor; limit <<= factor;