tcp: Namespace-ify sysctl_tcp_pacing_ss_ratio
Also remove an obsolete comment about TCP pacing. Signed-off-by: Eric Dumazet <edumazet@google.com> Signed-off-by: David S. Miller <davem@davemloft.net>
This commit is contained in:

committed by
David S. Miller

parent
4170ba6b58
commit
23a7102a2d
@@ -767,13 +767,6 @@ static void tcp_rtt_estimator(struct sock *sk, long mrtt_us)
|
||||
tp->srtt_us = max(1U, srtt);
|
||||
}
|
||||
|
||||
/* Set the sk_pacing_rate to allow proper sizing of TSO packets.
|
||||
* Note: TCP stack does not yet implement pacing.
|
||||
* FQ packet scheduler can be used to implement cheap but effective
|
||||
* TCP pacing, to smooth the burst on large writes when packets
|
||||
* in flight is significantly lower than cwnd (or rwin)
|
||||
*/
|
||||
int sysctl_tcp_pacing_ss_ratio __read_mostly = 200;
|
||||
int sysctl_tcp_pacing_ca_ratio __read_mostly = 120;
|
||||
|
||||
static void tcp_update_pacing_rate(struct sock *sk)
|
||||
@@ -793,7 +786,7 @@ static void tcp_update_pacing_rate(struct sock *sk)
|
||||
* end of slow start and should slow down.
|
||||
*/
|
||||
if (tp->snd_cwnd < tp->snd_ssthresh / 2)
|
||||
rate *= sysctl_tcp_pacing_ss_ratio;
|
||||
rate *= sock_net(sk)->ipv4.sysctl_tcp_pacing_ss_ratio;
|
||||
else
|
||||
rate *= sysctl_tcp_pacing_ca_ratio;
|
||||
|
||||
|
Reference in New Issue
Block a user