Since commit 605ad7f184b60cfaacbc038aa6c55ee68dee3c89 "tcp: refine TSO
autosizing",
the TSQ limit is computed as the smaller of
sysctl_tcp_limit_output_bytes and max(2 * skb->truesize, sk->sk_pacing_rate >>
10).
For low pacing rates, this approach sets a low limit, reducing throughput
dramatically.
Compute the limit as the greater of sysctl_tcp_limit_output_bytes and max(2 *
skb->truesize, sk->sk_pacing_rate >> 10).
Test:
netperf -H remote -l -2000000 -- -s 1000000
before patch:
MIGRATED TCP STREAM TEST from 0.0.0.0 () port 0 AF_INET to remote () port 0
AF_INET : demo
Recv Send Send
Socket Socket Message Elapsed
Size Size Size Time Throughput
bytes bytes bytes secs. 10^6bits/sec
87380 327680 327680 250.17 0.06
after patch:
MIGRATED TCP STREAM TEST from 0.0.0.0 () port 0 AF_INET to remote () port 0
AF_INET : demo
Recv Send Send
Socket Socket Message Elapsed
Size Size Size Time Throughput
bytes bytes bytes secs. 10^6bits/sec
87380 327680 327680 1.29 12.54
Signed-off-by: Sergej Benilov <sergej.beni...@googlemail.com>
---
net/ipv4/tcp_output.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/net/ipv4/tcp_output.c b/net/ipv4/tcp_output.c
index e625be56..71efca72 100644
--- a/net/ipv4/tcp_output.c
+++ b/net/ipv4/tcp_output.c
@@ -2054,7 +2054,7 @@ static bool tcp_write_xmit(struct sock *sk, unsigned int
mss_now, int nonagle,
* One example is wifi aggregation (802.11 AMPDU)
*/
limit = max(2 * skb->truesize, sk->sk_pacing_rate >> 10);
- limit = min_t(u32, limit, sysctl_tcp_limit_output_bytes);
+ limit = max_t(u32, limit, sysctl_tcp_limit_output_bytes);
if (atomic_read(&sk->sk_wmem_alloc) > limit) {
set_bit(TSQ_THROTTLED, &tp->tsq_flags);
--
2.17.1
