Skip to content

tcp: Add TCP_FASTOPEN_CONNECT sockopt #1731

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Closed
wants to merge 1 commit into from
Closed

tcp: Add TCP_FASTOPEN_CONNECT sockopt #1731

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
1 change: 1 addition & 0 deletions sys/netinet/tcp.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -224,6 +224,7 @@ __tcp_set_flags(struct tcphdr *th, uint16_t flags)
#define TCP_KEEPINTVL 512 /* L,N interval between keepalives */
#define TCP_KEEPCNT 1024 /* L,N number of keepalives before close */
#define TCP_FASTOPEN 1025 /* enable TFO / was created via TFO */
#define TCP_FASTOPEN_CONNECT 1027 /* defer connect until data; attempt TFO */
/* unused 2048 was TCP_PCAP_OUT */
/* unused 4096 was TCP_PCAP_IN */
#define TCP_FUNCTION_BLK 8192 /* Set the tcp function pointers to the specified stack */
Expand Down
141 changes: 104 additions & 37 deletions sys/netinet/tcp_usrreq.c
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -530,6 +530,8 @@ tcp_usr_connect(struct socket *so, struct sockaddr *nam, struct thread *td)
NET_EPOCH_ENTER(et);
if ((error = tcp_connect(tp, sinp, td)) != 0)
goto out_in_epoch;
if (tp->t_flags2 & TF2_TFO_CONNECT)
goto out_in_epoch;
#ifdef TCP_OFFLOAD
if (registered_toedevs > 0 &&
(so->so_options & SO_NO_OFFLOAD) == 0 &&
Expand Down Expand Up @@ -923,11 +925,11 @@ tcp_usr_send(struct socket *so, int flags, struct mbuf *m,
#ifdef INET6
struct sockaddr_in sin;
#endif
struct sockaddr_in *sinp;
struct sockaddr_in *sinp = NULL;
#endif
#ifdef INET6
struct sockaddr_in6 *sin6;
int isipv6;
struct sockaddr_in6 *sin6 = NULL;
int isipv6 = 0;
#endif
u_int8_t incflagsav;
u_char vflagsav;
Expand Down Expand Up @@ -1063,7 +1065,8 @@ tcp_usr_send(struct socket *so, int flags, struct mbuf *m,
tp->t_acktime = ticks;
sbappendstream(&so->so_snd, m, flags);
m = NULL;
if (nam && tp->t_state < TCPS_SYN_SENT) {
if ((nam || tp->t_flags2 & TF2_TFO_CONNECT)
&& tp->t_state < TCPS_SYN_SENT) {
KASSERT(tp->t_state == TCPS_CLOSED,
("%s: tp %p is listening", __func__, tp));

Expand Down Expand Up @@ -1474,18 +1477,25 @@ tcp_connect(struct tcpcb *tp, struct sockaddr_in *sin, struct thread *td)
NET_EPOCH_ASSERT();
INP_WLOCK_ASSERT(inp);

if (__predict_false((so->so_state &
(SS_ISCONNECTING | SS_ISCONNECTED | SS_ISDISCONNECTING |
SS_ISDISCONNECTED)) != 0))
return (EISCONN);
if (__predict_false((so->so_options & SO_REUSEPORT_LB) != 0))
return (EOPNOTSUPP);
/*
* If not doing the finalization of a TFO lazy-connect,
* bind the remote address.
*/
if (sin) {
if (__predict_false((so->so_state &
(SS_ISCONNECTING | SS_ISCONNECTED | SS_ISDISCONNECTING
| SS_ISDISCONNECTED)
) != 0))
return (EISCONN);
if (__predict_false((so->so_options & SO_REUSEPORT_LB) != 0))
return (EOPNOTSUPP);

INP_HASH_WLOCK(&V_tcbinfo);
error = in_pcbconnect(inp, sin, td->td_ucred);
INP_HASH_WUNLOCK(&V_tcbinfo);
if (error != 0)
return (error);
INP_HASH_WLOCK(&V_tcbinfo);
error = in_pcbconnect(inp, sin, td->td_ucred);
INP_HASH_WUNLOCK(&V_tcbinfo);
if (error != 0)
return (error);
}

/*
* Compute window scaling to request:
Expand All @@ -1497,12 +1507,22 @@ tcp_connect(struct tcpcb *tp, struct sockaddr_in *sin, struct thread *td)
tp->request_r_scale++;

soisconnecting(so);
TCPSTAT_INC(tcps_connattempt);
tcp_state_change(tp, TCPS_SYN_SENT);
tp->iss = tcp_new_isn(&inp->inp_inc);
if (tp->t_flags & TF_REQ_TSTMP)
tp->ts_offset = tcp_new_ts_offset(&inp->inp_inc);
tcp_sendseqinit(tp);
/*
* If doing the setup stage of a TFO lazy-connect set
* the socket state to connected. Otherwise (traditional
* case and finalization of TFO lazy-connect) kick off
* the TCP connect sequence.
*/
if (tp->t_flags2 & TF2_TFO_CONNECT && sin) {
soisconnected(so);
} else {
TCPSTAT_INC(tcps_connattempt);
tcp_state_change(tp, TCPS_SYN_SENT);
tp->iss = tcp_new_isn(&inp->inp_inc);
if (tp->t_flags & TF_REQ_TSTMP)
tp->ts_offset = tcp_new_ts_offset(&inp->inp_inc);
tcp_sendseqinit(tp);
}

return (0);
}
Expand All @@ -1519,30 +1539,46 @@ tcp6_connect(struct tcpcb *tp, struct sockaddr_in6 *sin6, struct thread *td)
NET_EPOCH_ASSERT();
INP_WLOCK_ASSERT(inp);

if (__predict_false((so->so_state &
(SS_ISCONNECTING | SS_ISCONNECTED)) != 0))
return (EISCONN);
if (__predict_false((so->so_options & SO_REUSEPORT_LB) != 0))
return (EOPNOTSUPP);
/*
* If not doing the finalization of a TFO lazy-connect,
* bind the remote address.
*/
if (sin6) {
if (__predict_false((so->so_state &
(SS_ISCONNECTING | SS_ISCONNECTED)) != 0))
return (EISCONN);
if (__predict_false((so->so_options & SO_REUSEPORT_LB) != 0))
return (EOPNOTSUPP);

INP_HASH_WLOCK(&V_tcbinfo);
error = in6_pcbconnect(inp, sin6, td->td_ucred, true);
INP_HASH_WUNLOCK(&V_tcbinfo);
if (error != 0)
return (error);
INP_HASH_WLOCK(&V_tcbinfo);
error = in6_pcbconnect(inp, sin6, td->td_ucred, true);
INP_HASH_WUNLOCK(&V_tcbinfo);
if (error != 0)
return (error);
}

/* Compute window scaling to request. */
while (tp->request_r_scale < TCP_MAX_WINSHIFT &&
(TCP_MAXWIN << tp->request_r_scale) < sb_max)
tp->request_r_scale++;

soisconnecting(so);
TCPSTAT_INC(tcps_connattempt);
tcp_state_change(tp, TCPS_SYN_SENT);
tp->iss = tcp_new_isn(&inp->inp_inc);
if (tp->t_flags & TF_REQ_TSTMP)
tp->ts_offset = tcp_new_ts_offset(&inp->inp_inc);
tcp_sendseqinit(tp);
/*
* If doing the setup stage of a TFO lazy-connect set
* the socket state to connected. Otherwise (traditional
* case and finalization of TFO lazy-connect) kick off
* the TCP connect sequence.
*/
if (tp->t_flags2 & TF2_TFO_CONNECT && sin6) {
soisconnected(so);
} else {
TCPSTAT_INC(tcps_connattempt);
tcp_state_change(tp, TCPS_SYN_SENT);
tp->iss = tcp_new_isn(&inp->inp_inc);
if (tp->t_flags & TF_REQ_TSTMP)
tp->ts_offset = tcp_new_ts_offset(&inp->inp_inc);
tcp_sendseqinit(tp);
}

return (0);
}
Expand Down Expand Up @@ -2396,6 +2432,31 @@ tcp_default_ctloutput(struct tcpcb *tp, struct sockopt *sopt)
goto unlock_and_done;
}

case TCP_FASTOPEN_CONNECT:
INP_WUNLOCK(inp);
if (!V_tcp_fastopen_client_enable &&
!V_tcp_fastopen_server_enable)
return (EPERM);

error = sooptcopyin(sopt, &optval,
sizeof(optval), sizeof(int));
if (error)
return (error);

INP_WLOCK_RECHECK(inp);
if (tp->t_state != TCPS_CLOSED) {
error = EINVAL;
goto unlock_and_done;
}
if (optval) {
tp->t_flags |= TF_FASTOPEN;
tp->t_flags2 |= TF2_TFO_CONNECT;
} else {
tp->t_flags &= ~TF_FASTOPEN;
tp->t_flags2 &= ~TF2_TFO_CONNECT;
}
goto unlock_and_done;

#ifdef TCP_BLACKBOX
case TCP_LOG:
INP_WUNLOCK(inp);
Expand Down Expand Up @@ -2580,6 +2641,12 @@ tcp_default_ctloutput(struct tcpcb *tp, struct sockopt *sopt)
INP_WUNLOCK(inp);
error = sooptcopyout(sopt, &optval, sizeof optval);
break;
case TCP_FASTOPEN_CONNECT:
optval = tp->t_flags2 & TF2_TFO_CONNECT;
INP_WUNLOCK(inp);
error = sooptcopyout(sopt, &optval, sizeof optval);
break;

#ifdef TCP_BLACKBOX
case TCP_LOG:
optval = tcp_get_bblog_state(tp);
Expand Down
1 change: 1 addition & 0 deletions sys/netinet/tcp_var.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -841,6 +841,7 @@ tcp_packets_this_ack(struct tcpcb *tp, tcp_seq ack)
#define TF2_PROC_SACK_PROHIBIT 0x00100000 /* Due to small MSS size do not process sack's */
#define TF2_IPSEC_TSO 0x00200000 /* IPSEC + TSO supported */
#define TF2_NO_ISS_CHECK 0x00400000 /* Don't check SEG.ACK against ISS */
#define TF2_TFO_CONNECT 0x00800000 /* TCP Fast Open lazy-connect */

/*
* Structure to hold TCP options that are only used during segment
Expand Down