Linux kernel mirror (for testing)
git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git
kernel
os
linux
Merge branch 'support-tcp_rto_min_us-and-tcp_delack_max_us-for-set-getsockopt'
Jason Xing says:
====================
support TCP_RTO_MIN_US and TCP_DELACK_MAX_US for set/getsockopt
Add set/getsockopt supports for TCP_RTO_MIN_US and TCP_DELACK_MAX_US.
====================
Link: https://patch.msgid.link/20250317120314.41404-1-kerneljasonxing@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
+30
-6
+2
-2
Documentation/networking/ip-sysctl.rst
+2
-2
Documentation/networking/ip-sysctl.rst
···
1229
1229
tcp_rto_min_us - INTEGER
1230
1230
Minimal TCP retransmission timeout (in microseconds). Note that the
1231
1231
rto_min route option has the highest precedence for configuring this
1232
-
setting, followed by the TCP_BPF_RTO_MIN socket option, followed by
1233
-
this tcp_rto_min_us sysctl.
1232
+
setting, followed by the TCP_BPF_RTO_MIN and TCP_RTO_MIN_US socket
1233
+
options, followed by this tcp_rto_min_us sysctl.
1234
1234
1235
1235
The recommended practice is to use a value less or equal to 200000
1236
1236
microseconds.
+1
-1
include/net/tcp.h
+1
-1
include/net/tcp.h
···
844
844
static inline u32 tcp_rto_min(const struct sock *sk)
845
845
{
846
846
const struct dst_entry *dst = __sk_dst_get(sk);
847
-
u32 rto_min = inet_csk(sk)->icsk_rto_min;
847
+
u32 rto_min = READ_ONCE(inet_csk(sk)->icsk_rto_min);
848
848
849
849
if (dst && dst_metric_locked(dst, RTAX_RTO_MIN))
850
850
rto_min = dst_metric_rtt(dst, RTAX_RTO_MIN);
+2
include/uapi/linux/tcp.h
+2
include/uapi/linux/tcp.h
···
140
140
141
141
#define TCP_IS_MPTCP 43 /* Is MPTCP being used? */
142
142
#define TCP_RTO_MAX_MS 44 /* max rto time in ms */
143
+
#define TCP_RTO_MIN_US 45 /* min rto time in us */
144
+
#define TCP_DELACK_MAX_US 46 /* max delayed ack time in us */
143
145
144
146
#define TCP_REPAIR_ON 1
145
147
#define TCP_REPAIR_OFF 0
+24
-2
net/ipv4/tcp.c
+24
-2
net/ipv4/tcp.c
···
3352
3352
icsk->icsk_probes_out = 0;
3353
3353
icsk->icsk_probes_tstamp = 0;
3354
3354
icsk->icsk_rto = TCP_TIMEOUT_INIT;
3355
-
icsk->icsk_rto_min = TCP_RTO_MIN;
3356
-
icsk->icsk_delack_max = TCP_DELACK_MAX;
3355
+
WRITE_ONCE(icsk->icsk_rto_min, TCP_RTO_MIN);
3356
+
WRITE_ONCE(icsk->icsk_delack_max, TCP_DELACK_MAX);
3357
3357
tp->snd_ssthresh = TCP_INFINITE_SSTHRESH;
3358
3358
tcp_snd_cwnd_set(tp, TCP_INIT_CWND);
3359
3359
tp->snd_cwnd_cnt = 0;
···
3833
3833
return -EINVAL;
3834
3834
WRITE_ONCE(inet_csk(sk)->icsk_rto_max, msecs_to_jiffies(val));
3835
3835
return 0;
3836
+
case TCP_RTO_MIN_US: {
3837
+
int rto_min = usecs_to_jiffies(val);
3838
+
3839
+
if (rto_min > TCP_RTO_MIN || rto_min < TCP_TIMEOUT_MIN)
3840
+
return -EINVAL;
3841
+
WRITE_ONCE(inet_csk(sk)->icsk_rto_min, rto_min);
3842
+
return 0;
3843
+
}
3844
+
case TCP_DELACK_MAX_US: {
3845
+
int delack_max = usecs_to_jiffies(val);
3846
+
3847
+
if (delack_max > TCP_DELACK_MAX || delack_max < TCP_TIMEOUT_MIN)
3848
+
return -EINVAL;
3849
+
WRITE_ONCE(inet_csk(sk)->icsk_delack_max, delack_max);
3850
+
return 0;
3851
+
}
3836
3852
}
3837
3853
3838
3854
sockopt_lock_sock(sk);
···
4687
4671
break;
4688
4672
case TCP_RTO_MAX_MS:
4689
4673
val = jiffies_to_msecs(tcp_rto_max(sk));
4674
+
break;
4675
+
case TCP_RTO_MIN_US:
4676
+
val = jiffies_to_usecs(READ_ONCE(inet_csk(sk)->icsk_rto_min));
4677
+
break;
4678
+
case TCP_DELACK_MAX_US:
4679
+
val = jiffies_to_usecs(READ_ONCE(inet_csk(sk)->icsk_delack_max));
4690
4680
break;
4691
4681
default:
4692
4682
return -ENOPROTOOPT;
+1
-1
net/ipv4/tcp_output.c
+1
-1
net/ipv4/tcp_output.c
···
4179
4179
{
4180
4180
u32 delack_from_rto_min = max(tcp_rto_min(sk), 2) - 1;
4181
4181
4182
-
return min(inet_csk(sk)->icsk_delack_max, delack_from_rto_min);
4182
+
return min(READ_ONCE(inet_csk(sk)->icsk_delack_max), delack_from_rto_min);
4183
4183
}
4184
4184
4185
4185
/* Send out a delayed ack, the caller does the policy checking