X-Git-Url: https://git.saurik.com/apple/xnu.git/blobdiff_plain/6601e61aa18bf4f09af135ff61fc7f4771d23b06..d190cdc3f5544636abb56dc1874be391d3e1b148:/bsd/netinet/tcp_usrreq.c diff --git a/bsd/netinet/tcp_usrreq.c b/bsd/netinet/tcp_usrreq.c index 2a5ab6988..f50103ea0 100644 --- a/bsd/netinet/tcp_usrreq.c +++ b/bsd/netinet/tcp_usrreq.c @@ -1,23 +1,29 @@ /* - * Copyright (c) 2000 Apple Computer, Inc. All rights reserved. + * Copyright (c) 2000-2016 Apple Inc. All rights reserved. * - * @APPLE_LICENSE_HEADER_START@ - * - * The contents of this file constitute Original Code as defined in and - * are subject to the Apple Public Source License Version 1.1 (the - * "License"). You may not use this file except in compliance with the - * License. Please obtain a copy of the License at - * http://www.apple.com/publicsource and read it before using this file. - * - * This Original Code and all software distributed under the License are - * distributed on an "AS IS" basis, WITHOUT WARRANTY OF ANY KIND, EITHER + * @APPLE_OSREFERENCE_LICENSE_HEADER_START@ + * + * This file contains Original Code and/or Modifications of Original Code + * as defined in and that are subject to the Apple Public Source License + * Version 2.0 (the 'License'). You may not use this file except in + * compliance with the License. The rights granted to you under the License + * may not be used to create, or enable the creation or redistribution of, + * unlawful or unlicensed copies of an Apple operating system, or to + * circumvent, violate, or enable the circumvention or violation of, any + * terms of an Apple operating system software license agreement. + * + * Please obtain a copy of the License at + * http://www.opensource.apple.com/apsl/ and read it before using this file. + * + * The Original Code and all software distributed under the License are + * distributed on an 'AS IS' basis, WITHOUT WARRANTY OF ANY KIND, EITHER * EXPRESS OR IMPLIED, AND APPLE HEREBY DISCLAIMS ALL SUCH WARRANTIES, * INCLUDING WITHOUT LIMITATION, ANY WARRANTIES OF MERCHANTABILITY, - * FITNESS FOR A PARTICULAR PURPOSE OR NON-INFRINGEMENT. Please see the - * License for the specific language governing rights and limitations - * under the License. - * - * @APPLE_LICENSE_HEADER_END@ + * FITNESS FOR A PARTICULAR PURPOSE, QUIET ENJOYMENT OR NON-INFRINGEMENT. + * Please see the License for the specific language governing rights and + * limitations under the License. + * + * @APPLE_OSREFERENCE_LICENSE_HEADER_END@ */ /* * Copyright (c) 1982, 1986, 1988, 1993 @@ -64,12 +70,16 @@ #if INET6 #include #endif /* INET6 */ +#include #include #include #include +#include #include #include +#include +#include #include #include @@ -91,14 +101,29 @@ #include #include #include +#include #if TCPDEBUG #include #endif +#if MPTCP +#include +#endif /* MPTCP */ #if IPSEC #include #endif /*IPSEC*/ +#if FLOW_DIVERT +#include +#endif /* FLOW_DIVERT */ + +void tcp_fill_info(struct tcpcb *, struct tcp_info *); +errno_t tcp_fill_info_for_info_tuple(struct info_tuple *, struct tcp_info *); + +int tcp_sysctl_info(struct sysctl_oid *, void *, int , struct sysctl_req *); +static void tcp_connection_fill_info(struct tcpcb *tp, + struct tcp_connection_info *tci); + /* * TCP protocol interface to socket abstraction. */ @@ -108,11 +133,12 @@ static int tcp_attach(struct socket *, struct proc *); static int tcp_connect(struct tcpcb *, struct sockaddr *, struct proc *); #if INET6 static int tcp6_connect(struct tcpcb *, struct sockaddr *, struct proc *); +static int tcp6_usr_connect(struct socket *, struct sockaddr *, + struct proc *); #endif /* INET6 */ -static struct tcpcb * - tcp_disconnect(struct tcpcb *); -static struct tcpcb * - tcp_usrclosed(struct tcpcb *); +static struct tcpcb *tcp_disconnect(struct tcpcb *); +static struct tcpcb *tcp_usrclosed(struct tcpcb *); +extern void tcp_sbrcv_trim(struct tcpcb *tp, struct sockbuf *sb); #if TCPDEBUG #define TCPDEBUG0 int ostate = 0 @@ -125,12 +151,22 @@ static struct tcpcb * #define TCPDEBUG2(req) #endif +SYSCTL_PROC(_net_inet_tcp, OID_AUTO, info, + CTLFLAG_RW | CTLFLAG_LOCKED | CTLFLAG_ANYBODY | CTLFLAG_KERN, + 0 , 0, tcp_sysctl_info, "S", "TCP info per tuple"); + /* * TCP attaches to socket via pru_attach(), reserving space, * and an internet control block. + * + * Returns: 0 Success + * EISCONN + * tcp_attach:ENOBUFS + * tcp_attach:ENOMEM + * tcp_attach:??? [IPSEC specific] */ static int -tcp_usr_attach(struct socket *so, int proto, struct proc *p) +tcp_usr_attach(struct socket *so, __unused int proto, struct proc *p) { int error; struct inpcb *inp = sotoinpcb(so); @@ -173,34 +209,59 @@ tcp_usr_detach(struct socket *so) if (inp == 0 || (inp->inp_state == INPCB_STATE_DEAD)) { return EINVAL; /* XXX */ } -#if 1 - lck_mtx_assert(((struct inpcb *)so->so_pcb)->inpcb_mtx, LCK_MTX_ASSERT_OWNED); -#endif + lck_mtx_assert(&((struct inpcb *)so->so_pcb)->inpcb_mtx, LCK_MTX_ASSERT_OWNED); tp = intotcpcb(inp); /* In case we got disconnected from the peer */ - if (tp == 0) - goto out; + if (tp == NULL) + goto out; TCPDEBUG1(); + + calculate_tcp_clock(); + tp = tcp_disconnect(tp); out: TCPDEBUG2(PRU_DETACH); return error; } -#define COMMON_START() TCPDEBUG0; \ - do { \ - if (inp == 0 || (inp->inp_state == INPCB_STATE_DEAD)) { \ - return EINVAL; \ - } \ - tp = intotcpcb(inp); \ - TCPDEBUG1(); \ - } while(0) - +#if NECP +#define COMMON_START() TCPDEBUG0; \ +do { \ + if (inp == NULL || inp->inp_state == INPCB_STATE_DEAD) \ + return (EINVAL); \ + if (necp_socket_should_use_flow_divert(inp)) \ + return (EPROTOTYPE); \ + tp = intotcpcb(inp); \ + TCPDEBUG1(); \ + calculate_tcp_clock(); \ +} while (0) +#else /* NECP */ +#define COMMON_START() TCPDEBUG0; \ +do { \ + if (inp == NULL || inp->inp_state == INPCB_STATE_DEAD) \ + return (EINVAL); \ + tp = intotcpcb(inp); \ + TCPDEBUG1(); \ + calculate_tcp_clock(); \ +} while (0) +#endif /* !NECP */ + #define COMMON_END(req) out: TCPDEBUG2(req); return error; goto out /* * Give the socket an address. + * + * Returns: 0 Success + * EINVAL Invalid argument [COMMON_START] + * EAFNOSUPPORT Address family not supported + * in_pcbbind:EADDRNOTAVAIL Address not available. + * in_pcbbind:EINVAL Invalid argument + * in_pcbbind:EAFNOSUPPORT Address family not supported [notdef] + * in_pcbbind:EACCES Permission denied + * in_pcbbind:EADDRINUSE Address in use + * in_pcbbind:EAGAIN Resource unavailable, try again + * in_pcbbind:EPERM Operation not permitted */ static int tcp_usr_bind(struct socket *so, struct sockaddr *nam, struct proc *p) @@ -212,11 +273,16 @@ tcp_usr_bind(struct socket *so, struct sockaddr *nam, struct proc *p) COMMON_START(); + if (nam->sa_family != 0 && nam->sa_family != AF_INET) { + error = EAFNOSUPPORT; + goto out; + } + /* * Must check for multicast addresses and disallow binding * to them. */ - sinp = (struct sockaddr_in *)nam; + sinp = (struct sockaddr_in *)(void *)nam; if (sinp->sin_family == AF_INET && IN_MULTICAST(ntohl(sinp->sin_addr.s_addr))) { error = EAFNOSUPPORT; @@ -240,11 +306,16 @@ tcp6_usr_bind(struct socket *so, struct sockaddr *nam, struct proc *p) COMMON_START(); + if (nam->sa_family != 0 && nam->sa_family != AF_INET6) { + error = EAFNOSUPPORT; + goto out; + } + /* * Must check for multicast addresses and disallow binding * to them. */ - sin6p = (struct sockaddr_in6 *)nam; + sin6p = (struct sockaddr_in6 *)(void *)nam; if (sin6p->sin6_family == AF_INET6 && IN6_IS_ADDR_MULTICAST(&sin6p->sin6_addr)) { error = EAFNOSUPPORT; @@ -274,6 +345,16 @@ tcp6_usr_bind(struct socket *so, struct sockaddr *nam, struct proc *p) /* * Prepare to accept connections. + * + * Returns: 0 Success + * EINVAL [COMMON_START] + * in_pcbbind:EADDRNOTAVAIL Address not available. + * in_pcbbind:EINVAL Invalid argument + * in_pcbbind:EAFNOSUPPORT Address family not supported [notdef] + * in_pcbbind:EACCES Permission denied + * in_pcbbind:EADDRINUSE Address in use + * in_pcbbind:EAGAIN Resource unavailable, try again + * in_pcbbind:EPERM Operation not permitted */ static int tcp_usr_listen(struct socket *so, struct proc *p) @@ -284,7 +365,7 @@ tcp_usr_listen(struct socket *so, struct proc *p) COMMON_START(); if (inp->inp_lport == 0) - error = in_pcbbind(inp, (struct sockaddr *)0, p); + error = in_pcbbind(inp, NULL, p); if (error == 0) tp->t_state = TCPS_LISTEN; COMMON_END(PRU_LISTEN); @@ -303,7 +384,7 @@ tcp6_usr_listen(struct socket *so, struct proc *p) inp->inp_vflag &= ~INP_IPV4; if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0) inp->inp_vflag |= INP_IPV4; - error = in6_pcbbind(inp, (struct sockaddr *)0, p); + error = in6_pcbbind(inp, NULL, p); } if (error == 0) tp->t_state = TCPS_LISTEN; @@ -311,6 +392,27 @@ tcp6_usr_listen(struct socket *so, struct proc *p) } #endif /* INET6 */ +static int +tcp_connect_complete(struct socket *so) +{ + struct tcpcb *tp = sototcpcb(so); + int error = 0; + + /* TFO delays the tcp_output until later, when the app calls write() */ + if (so->so_flags1 & SOF1_PRECONNECT_DATA) { + if (!necp_socket_is_allowed_to_send_recv(sotoinpcb(so), NULL, NULL)) + return (EHOSTUNREACH); + + /* Initialize enough state so that we can actually send data */ + tcp_mss(tp, -1, IFSCOPE_NONE); + tp->snd_wnd = tp->t_maxseg; + } else { + error = tcp_output(tp); + } + + return (error); +} + /* * Initiate connection to peer. * Create a template for use in transmissions on this connection. @@ -326,28 +428,210 @@ tcp_usr_connect(struct socket *so, struct sockaddr *nam, struct proc *p) struct tcpcb *tp; struct sockaddr_in *sinp; - COMMON_START(); + TCPDEBUG0; + if (inp == NULL) { + return EINVAL; + } else if (inp->inp_state == INPCB_STATE_DEAD) { + if (so->so_error) { + error = so->so_error; + so->so_error = 0; + return error; + } else + return EINVAL; + } +#if NECP +#if FLOW_DIVERT + else if (necp_socket_should_use_flow_divert(inp)) { + uint32_t fd_ctl_unit = necp_socket_get_flow_divert_control_unit(inp); + if (fd_ctl_unit > 0) { + error = flow_divert_pcb_init(so, fd_ctl_unit); + if (error == 0) { + error = flow_divert_connect_out(so, nam, p); + } + } else { + error = ENETDOWN; + } + + return error; + } +#endif /* FLOW_DIVERT */ +#if CONTENT_FILTER + error = cfil_sock_attach(so); + if (error != 0) + return error; +#endif /* CONTENT_FILTER */ +#endif /* NECP */ + tp = intotcpcb(inp); + TCPDEBUG1(); + calculate_tcp_clock(); + + if (nam->sa_family != 0 && nam->sa_family != AF_INET) { + error = EAFNOSUPPORT; + goto out; + } /* * Must disallow TCP ``connections'' to multicast addresses. */ - sinp = (struct sockaddr_in *)nam; + sinp = (struct sockaddr_in *)(void *)nam; if (sinp->sin_family == AF_INET && IN_MULTICAST(ntohl(sinp->sin_addr.s_addr))) { error = EAFNOSUPPORT; goto out; } -#ifndef __APPLE__ - prison_remote_ip(p, 0, &sinp->sin_addr.s_addr); -#endif - if ((error = tcp_connect(tp, nam, p)) != 0) goto out; - error = tcp_output(tp); + + error = tcp_connect_complete(so); + COMMON_END(PRU_CONNECT); } +static int +tcp_usr_connectx_common(struct socket *so, int af, + struct sockaddr_list **src_sl, struct sockaddr_list **dst_sl, + struct proc *p, uint32_t ifscope, sae_associd_t aid, sae_connid_t *pcid, + uint32_t flags, void *arg, uint32_t arglen, struct uio *auio, + user_ssize_t *bytes_written) +{ +#pragma unused(aid) +#if !MPTCP +#pragma unused(flags, arg, arglen) +#endif /* !MPTCP */ + struct sockaddr_entry *src_se = NULL, *dst_se = NULL; + struct inpcb *inp = sotoinpcb(so); + int error; + user_ssize_t datalen = 0; + + if (inp == NULL) + return (EINVAL); + + VERIFY(dst_sl != NULL); + + /* select source (if specified) and destination addresses */ + error = in_selectaddrs(af, src_sl, &src_se, dst_sl, &dst_se); + if (error != 0) + return (error); + + VERIFY(*dst_sl != NULL && dst_se != NULL); + VERIFY(src_se == NULL || *src_sl != NULL); + VERIFY(dst_se->se_addr->sa_family == af); + VERIFY(src_se == NULL || src_se->se_addr->sa_family == af); + +#if NECP + inp_update_necp_policy(inp, src_se ? src_se->se_addr : NULL, dst_se ? dst_se->se_addr : NULL, ifscope); +#endif /* NECP */ + + if ((so->so_flags1 & SOF1_DATA_IDEMPOTENT) && + (tcp_fastopen & TCP_FASTOPEN_CLIENT)) + sototcpcb(so)->t_flagsext |= TF_FASTOPEN; + + /* + * We get here for 2 cases: + * + * a. From MPTCP, to connect a subflow. There is no need to + * bind the socket to the source address and/or interface, + * since everything has been taken care of by MPTCP. We + * simply check whether or not this is for the initial + * MPTCP connection attempt, or to join an existing one. + * + * b. From the socket layer, to connect a TCP. Perform the + * bind to source address and/or interface as necessary. + */ +#if MPTCP + if (flags & CONNREQF_MPTCP) { + struct mptsub_connreq *mpcr = arg; + + /* Check to make sure this came down from MPTCP */ + if (arg == NULL || arglen != sizeof (*mpcr)) + return (EOPNOTSUPP); + + switch (mpcr->mpcr_type) { + case MPTSUB_CONNREQ_MP_ENABLE: + break; + case MPTSUB_CONNREQ_MP_ADD: + break; + default: + return (EOPNOTSUPP); + } + } else +#endif /* MPTCP */ + { + /* bind socket to the specified interface, if requested */ + if (ifscope != IFSCOPE_NONE && + (error = inp_bindif(inp, ifscope, NULL)) != 0) + return (error); + + /* if source address and/or port is specified, bind to it */ + if (src_se != NULL) { + struct sockaddr *sa = src_se->se_addr; + error = sobindlock(so, sa, 0); /* already locked */ + if (error != 0) + return (error); + } + } + + switch (af) { + case AF_INET: + error = tcp_usr_connect(so, dst_se->se_addr, p); + break; +#if INET6 + case AF_INET6: + error = tcp6_usr_connect(so, dst_se->se_addr, p); + break; +#endif /* INET6 */ + default: + VERIFY(0); + /* NOTREACHED */ + } + + if (error != 0) + return (error); + + /* if there is data, copy it */ + if (auio != NULL) { + socket_unlock(so, 0); + + VERIFY(bytes_written != NULL); + + datalen = uio_resid(auio); + error = so->so_proto->pr_usrreqs->pru_sosend(so, NULL, + (uio_t)auio, NULL, NULL, 0); + socket_lock(so, 0); + + if (error == 0 || error == EWOULDBLOCK) + *bytes_written = datalen - uio_resid(auio); + + /* + * sosend returns EWOULDBLOCK if it's a non-blocking + * socket or a timeout occured (this allows to return + * the amount of queued data through sendit()). + * + * However, connectx() returns EINPROGRESS in case of a + * blocking socket. So we change the return value here. + */ + if (error == EWOULDBLOCK) + error = EINPROGRESS; + } + + if (error == 0 && pcid != NULL) + *pcid = 1; /* there is only one connection in regular TCP */ + + return (error); +} + +static int +tcp_usr_connectx(struct socket *so, struct sockaddr_list **src_sl, + struct sockaddr_list **dst_sl, struct proc *p, uint32_t ifscope, + sae_associd_t aid, sae_connid_t *pcid, uint32_t flags, void *arg, + uint32_t arglen, struct uio *uio, user_ssize_t *bytes_written) +{ + return (tcp_usr_connectx_common(so, AF_INET, src_sl, dst_sl, + p, ifscope, aid, pcid, flags, arg, arglen, uio, + bytes_written)); +} + #if INET6 static int tcp6_usr_connect(struct socket *so, struct sockaddr *nam, struct proc *p) @@ -357,12 +641,54 @@ tcp6_usr_connect(struct socket *so, struct sockaddr *nam, struct proc *p) struct tcpcb *tp; struct sockaddr_in6 *sin6p; - COMMON_START(); + TCPDEBUG0; + if (inp == NULL) { + return EINVAL; + } else if (inp->inp_state == INPCB_STATE_DEAD) { + if (so->so_error) { + error = so->so_error; + so->so_error = 0; + return error; + } else + return EINVAL; + } +#if NECP +#if FLOW_DIVERT + else if (necp_socket_should_use_flow_divert(inp)) { + uint32_t fd_ctl_unit = necp_socket_get_flow_divert_control_unit(inp); + if (fd_ctl_unit > 0) { + error = flow_divert_pcb_init(so, fd_ctl_unit); + if (error == 0) { + error = flow_divert_connect_out(so, nam, p); + } + } else { + error = ENETDOWN; + } + + return error; + } +#endif /* FLOW_DIVERT */ +#if CONTENT_FILTER + error = cfil_sock_attach(so); + if (error != 0) + return error; +#endif /* CONTENT_FILTER */ +#endif /* NECP */ + + tp = intotcpcb(inp); + TCPDEBUG1(); + + calculate_tcp_clock(); + + if (nam->sa_family != 0 && nam->sa_family != AF_INET6) { + error = EAFNOSUPPORT; + goto out; + } /* * Must disallow TCP ``connections'' to multicast addresses. */ - sin6p = (struct sockaddr_in6 *)nam; + sin6p = (struct sockaddr_in6 *)(void *)nam; if (sin6p->sin6_family == AF_INET6 && IN6_IS_ADDR_MULTICAST(&sin6p->sin6_addr)) { error = EAFNOSUPPORT; @@ -380,18 +706,29 @@ tcp6_usr_connect(struct socket *so, struct sockaddr *nam, struct proc *p) inp->inp_vflag &= ~INP_IPV6; if ((error = tcp_connect(tp, (struct sockaddr *)&sin, p)) != 0) goto out; - error = tcp_output(tp); + + error = tcp_connect_complete(so); goto out; } inp->inp_vflag &= ~INP_IPV4; inp->inp_vflag |= INP_IPV6; if ((error = tcp6_connect(tp, nam, p)) != 0) goto out; - error = tcp_output(tp); - if (error) - goto out; + + error = tcp_connect_complete(so); COMMON_END(PRU_CONNECT); } + +static int +tcp6_usr_connectx(struct socket *so, struct sockaddr_list **src_sl, + struct sockaddr_list **dst_sl, struct proc *p, uint32_t ifscope, + sae_associd_t aid, sae_connid_t *pcid, uint32_t flags, void *arg, + uint32_t arglen, struct uio *uio, user_ssize_t *bytes_written) +{ + return (tcp_usr_connectx_common(so, AF_INET6, src_sl, dst_sl, + p, ifscope, aid, pcid, flags, arg, arglen, uio, + bytes_written)); +} #endif /* INET6 */ /* @@ -411,18 +748,30 @@ tcp_usr_disconnect(struct socket *so) int error = 0; struct inpcb *inp = sotoinpcb(so); struct tcpcb *tp; - -#if 1 - lck_mtx_assert(((struct inpcb *)so->so_pcb)->inpcb_mtx, LCK_MTX_ASSERT_OWNED); -#endif + + lck_mtx_assert(&((struct inpcb *)so->so_pcb)->inpcb_mtx, + LCK_MTX_ASSERT_OWNED); COMMON_START(); /* In case we got disconnected from the peer */ - if (tp == 0) - goto out; + if (tp == NULL) + goto out; tp = tcp_disconnect(tp); COMMON_END(PRU_DISCONNECT); } +/* + * User-protocol pru_disconnectx callback. + */ +static int +tcp_usr_disconnectx(struct socket *so, sae_associd_t aid, sae_connid_t cid) +{ +#pragma unused(cid) + if (aid != SAE_ASSOCID_ANY && aid != SAE_ASSOCID_ALL) + return (EINVAL); + + return (tcp_usr_disconnect(so)); +} + /* * Accept a connection. Essentially all the work is * done at higher levels; just return the address @@ -436,16 +785,29 @@ tcp_usr_accept(struct socket *so, struct sockaddr **nam) struct tcpcb *tp = NULL; TCPDEBUG0; + in_getpeeraddr(so, nam); + if (so->so_state & SS_ISDISCONNECTED) { error = ECONNABORTED; goto out; } - if (inp == 0 || (inp->inp_state == INPCB_STATE_DEAD)) { + if (inp == NULL || inp->inp_state == INPCB_STATE_DEAD) return (EINVAL); - } +#if NECP + else if (necp_socket_should_use_flow_divert(inp)) + return (EPROTOTYPE); +#if CONTENT_FILTER + error = cfil_sock_attach(so); + if (error != 0) + return (error); +#endif /* CONTENT_FILTER */ +#endif /* NECP */ + tp = intotcpcb(inp); TCPDEBUG1(); - in_setpeeraddr(so, nam); + + calculate_tcp_clock(); + COMMON_END(PRU_ACCEPT); } @@ -462,17 +824,44 @@ tcp6_usr_accept(struct socket *so, struct sockaddr **nam) error = ECONNABORTED; goto out; } - if (inp == 0 || (inp->inp_state == INPCB_STATE_DEAD)) { + if (inp == NULL || inp->inp_state == INPCB_STATE_DEAD) return (EINVAL); - } +#if NECP + else if (necp_socket_should_use_flow_divert(inp)) + return (EPROTOTYPE); +#if CONTENT_FILTER + error = cfil_sock_attach(so); + if (error != 0) + return (error); +#endif /* CONTENT_FILTER */ +#endif /* NECP */ + tp = intotcpcb(inp); TCPDEBUG1(); + + calculate_tcp_clock(); + in6_mapped_peeraddr(so, nam); COMMON_END(PRU_ACCEPT); } #endif /* INET6 */ + /* * Mark the connection as being incapable of further output. + * + * Returns: 0 Success + * EINVAL [COMMON_START] + * tcp_output:EADDRNOTAVAIL + * tcp_output:ENOBUFS + * tcp_output:EMSGSIZE + * tcp_output:EHOSTUNREACH + * tcp_output:ENETUNREACH + * tcp_output:ENETDOWN + * tcp_output:ENOMEM + * tcp_output:EACCES + * tcp_output:EMSGSIZE + * tcp_output:ENOBUFS + * tcp_output:??? [ignorable: mostly IPSEC/firewall/DLIL] */ static int tcp_usr_shutdown(struct socket *so) @@ -481,12 +870,45 @@ tcp_usr_shutdown(struct socket *so) struct inpcb *inp = sotoinpcb(so); struct tcpcb *tp; - COMMON_START(); + TCPDEBUG0; + if (inp == NULL || inp->inp_state == INPCB_STATE_DEAD) + return (EINVAL); + socantsendmore(so); - /* In case we got disconnected from the peer */ - if (tp == 0) - goto out; + + /* + * In case we got disconnected from the peer, or if this is + * a socket that is to be flow-diverted (but not yet). + */ + tp = intotcpcb(inp); + TCPDEBUG1(); + + if (tp == NULL +#if NECP + || (necp_socket_should_use_flow_divert(inp)) +#endif /* NECP */ + ) { + if (tp != NULL) + error = EPROTOTYPE; + goto out; + } + + calculate_tcp_clock(); + tp = tcp_usrclosed(tp); +#if MPTCP + /* A reset has been sent but socket exists, do not send FIN */ + if ((so->so_flags & SOF_MP_SUBFLOW) && + (tp) && (tp->t_mpflags & TMPF_RESET)) { + goto out; + } +#endif +#if CONTENT_FILTER + /* Don't send a FIN yet */ + if (tp && !(so->so_state & SS_ISDISCONNECTED) && + cfil_sock_data_pending(&so->so_snd)) + goto out; +#endif /* CONTENT_FILTER */ if (tp) error = tcp_output(tp); COMMON_END(PRU_SHUTDOWN); @@ -496,7 +918,7 @@ tcp_usr_shutdown(struct socket *so) * After a receive, possibly send window update to peer. */ static int -tcp_usr_rcvd(struct socket *so, int flags) +tcp_usr_rcvd(struct socket *so, __unused int flags) { int error = 0; struct inpcb *inp = sotoinpcb(so); @@ -504,9 +926,22 @@ tcp_usr_rcvd(struct socket *so, int flags) COMMON_START(); /* In case we got disconnected from the peer */ - if (tp == 0) - goto out; - tcp_output(tp); + if (tp == NULL) + goto out; + tcp_sbrcv_trim(tp, &so->so_rcv); + + /* + * This tcp_output is solely there to trigger window-updates. + * However, we really do not want these window-updates while we + * are still in SYN_SENT or SYN_RECEIVED. + */ + if (TCPS_HAVEESTABLISHED(tp->t_state)) + tcp_output(tp); + +#if CONTENT_FILTER + cfil_sock_buf_update(&so->so_rcv); +#endif /* CONTENT_FILTER */ + COMMON_END(PRU_RCVD); } @@ -516,30 +951,65 @@ tcp_usr_rcvd(struct socket *so, int flags) * pru_*() routines, the mbuf chains are our responsibility. We * must either enqueue them or free them. The other pru_* routines * generally are caller-frees. + * + * Returns: 0 Success + * ECONNRESET + * EINVAL + * ENOBUFS + * tcp_connect:EADDRINUSE Address in use + * tcp_connect:EADDRNOTAVAIL Address not available. + * tcp_connect:EINVAL Invalid argument + * tcp_connect:EAFNOSUPPORT Address family not supported [notdef] + * tcp_connect:EACCES Permission denied + * tcp_connect:EAGAIN Resource unavailable, try again + * tcp_connect:EPERM Operation not permitted + * tcp_output:EADDRNOTAVAIL + * tcp_output:ENOBUFS + * tcp_output:EMSGSIZE + * tcp_output:EHOSTUNREACH + * tcp_output:ENETUNREACH + * tcp_output:ENETDOWN + * tcp_output:ENOMEM + * tcp_output:EACCES + * tcp_output:EMSGSIZE + * tcp_output:ENOBUFS + * tcp_output:??? [ignorable: mostly IPSEC/firewall/DLIL] + * tcp6_connect:??? [IPV6 only] */ static int -tcp_usr_send(struct socket *so, int flags, struct mbuf *m, - struct sockaddr *nam, struct mbuf *control, struct proc *p) +tcp_usr_send(struct socket *so, int flags, struct mbuf *m, + struct sockaddr *nam, struct mbuf *control, struct proc *p) { int error = 0; struct inpcb *inp = sotoinpcb(so); struct tcpcb *tp; + uint32_t msgpri = MSG_PRI_DEFAULT; #if INET6 int isipv6; #endif TCPDEBUG0; - if (inp == NULL || inp->inp_state == INPCB_STATE_DEAD) { + if (inp == NULL || inp->inp_state == INPCB_STATE_DEAD +#if NECP + || (necp_socket_should_use_flow_divert(inp)) +#endif /* NECP */ + ) { /* * OOPS! we lost a race, the TCP session got reset after * we checked SS_CANTSENDMORE, eg: while doing uiomove or a * network interrupt in the non-splnet() section of sosend(). */ - if (m) + if (m != NULL) m_freem(m); - if (control) + if (control != NULL) { m_freem(control); - error = ECONNRESET; /* XXX EPIPE? */ + control = NULL; + } + + if (inp == NULL) + error = ECONNRESET; /* XXX EPIPE? */ + else + error = EPROTOTYPE; tp = NULL; TCPDEBUG1(); goto out; @@ -549,20 +1019,56 @@ tcp_usr_send(struct socket *so, int flags, struct mbuf *m, #endif /* INET6 */ tp = intotcpcb(inp); TCPDEBUG1(); - if (control) { - /* TCP doesn't do control messages (rights, creds, etc) */ - if (control->m_len) { + + calculate_tcp_clock(); + + if (control != NULL) { + if (so->so_flags & SOF_ENABLE_MSGS) { + /* Get the msg priority from control mbufs */ + error = tcp_get_msg_priority(control, &msgpri); + if (error) { + m_freem(control); + if (m != NULL) + m_freem(m); + control = NULL; + m = NULL; + goto out; + } + m_freem(control); + control = NULL; + } else if (control->m_len) { + /* + * if not unordered, TCP should not have + * control mbufs + */ m_freem(control); - if (m) + if (m != NULL) m_freem(m); + control = NULL; + m = NULL; error = EINVAL; goto out; } - m_freem(control); /* empty control, just free it */ } - if(!(flags & PRUS_OOB)) { - sbappend(&so->so_snd, m); + + if (so->so_flags & SOF_ENABLE_MSGS) { + VERIFY(m->m_flags & M_PKTHDR); + m->m_pkthdr.msg_pri = msgpri; + } + + /* MPTCP sublow socket buffers must not be compressed */ + VERIFY(!(so->so_flags & SOF_MP_SUBFLOW) || + (so->so_snd.sb_flags & SB_NOCOMPRESS)); + + if(!(flags & PRUS_OOB) || (so->so_flags1 & SOF1_PRECONNECT_DATA)) { + /* Call msg send if message delivery is enabled */ + if (so->so_flags & SOF_ENABLE_MSGS) + sbappendmsg_snd(&so->so_snd, m); + else + sbappendstream(&so->so_snd, m); + if (nam && tp->t_state < TCPS_SYN_SENT) { + /* * Do implied connect if not yet connected, * initialize window to default value, and @@ -574,11 +1080,11 @@ tcp_usr_send(struct socket *so, int flags, struct mbuf *m, error = tcp6_connect(tp, nam, p); else #endif /* INET6 */ - error = tcp_connect(tp, nam, p); + error = tcp_connect(tp, nam, p); if (error) goto out; tp->snd_wnd = TTCP_CLIENT_SND_WND; - tcp_mss(tp, -1); + tcp_mss(tp, -1, IFSCOPE_NONE); } if (flags & PRUS_EOF) { @@ -597,7 +1103,9 @@ tcp_usr_send(struct socket *so, int flags, struct mbuf *m, tp->t_flags &= ~TF_MORETOCOME; } } else { - if (sbspace(&so->so_snd) < -512) { + if (sbspace(&so->so_snd) == 0) { + /* if no space is left in sockbuf, + * do not try to squeeze in OOB traffic */ m_freem(m); error = ENOBUFS; goto out; @@ -610,7 +1118,7 @@ tcp_usr_send(struct socket *so, int flags, struct mbuf *m, * of data past the urgent section. * Otherwise, snd_up should be one lower. */ - sbappend(&so->so_snd, m); + sbappendstream(&so->so_snd, m); if (nam && tp->t_state < TCPS_SYN_SENT) { /* * Do implied connect if not yet connected, @@ -627,13 +1135,26 @@ tcp_usr_send(struct socket *so, int flags, struct mbuf *m, if (error) goto out; tp->snd_wnd = TTCP_CLIENT_SND_WND; - tcp_mss(tp, -1); + tcp_mss(tp, -1, IFSCOPE_NONE); } tp->snd_up = tp->snd_una + so->so_snd.sb_cc; - tp->t_force = 1; + tp->t_flagsext |= TF_FORCE; error = tcp_output(tp); - tp->t_force = 0; + tp->t_flagsext &= ~TF_FORCE; + } + + + /* + * We wait for the socket to successfully connect before returning. + * This allows us to signal a timeout to the application. + */ + if (so->so_state & SS_ISCONNECTING) { + if (so->so_state & SS_NBIO) + error = EWOULDBLOCK; + else + error = sbwait(&so->so_snd); } + COMMON_END((flags & PRUS_OOB) ? PRU_SENDOOB : ((flags & PRUS_EOF) ? PRU_SEND_EOF : PRU_SEND)); } @@ -650,15 +1171,21 @@ tcp_usr_abort(struct socket *so) COMMON_START(); /* In case we got disconnected from the peer */ - if (tp == 0) - goto out; + if (tp == NULL) + goto out; tp = tcp_drop(tp, ECONNABORTED); + VERIFY(so->so_usecount > 0); so->so_usecount--; COMMON_END(PRU_ABORT); } /* * Receive out-of-band data. + * + * Returns: 0 Success + * EINVAL [COMMON_START] + * EINVAL + * EWOULDBLOCK */ static int tcp_usr_rcvoob(struct socket *so, struct mbuf *m, int flags) @@ -681,27 +1208,82 @@ tcp_usr_rcvoob(struct socket *so, struct mbuf *m, int flags) } m->m_len = 1; *mtod(m, caddr_t) = tp->t_iobc; + so->so_state &= ~SS_RCVATMARK; if ((flags & MSG_PEEK) == 0) tp->t_oobflags ^= (TCPOOB_HAVEDATA | TCPOOB_HADDATA); COMMON_END(PRU_RCVOOB); } +static int +tcp_usr_preconnect(struct socket *so) +{ + struct inpcb *inp = sotoinpcb(so); + int error = 0; + +#if NECP + if (necp_socket_should_use_flow_divert(inp)) { + /* May happen, if in tcp_usr_connect we did not had a chance + * to set the usrreqs (due to some error). So, let's get out + * of here. + */ + goto out; + } +#endif /* NECP */ + + error = tcp_output(sototcpcb(so)); + + /* One read has been done. This was enough. Get back to "normal" behavior. */ + so->so_flags1 &= ~SOF1_PRECONNECT_DATA; + + COMMON_END(PRU_PRECONNECT); +} + /* xxx - should be const */ struct pr_usrreqs tcp_usrreqs = { - tcp_usr_abort, tcp_usr_accept, tcp_usr_attach, tcp_usr_bind, - tcp_usr_connect, pru_connect2_notsupp, in_control, tcp_usr_detach, - tcp_usr_disconnect, tcp_usr_listen, in_setpeeraddr, tcp_usr_rcvd, - tcp_usr_rcvoob, tcp_usr_send, pru_sense_null, tcp_usr_shutdown, - in_setsockaddr, sosend, soreceive, pru_sopoll_notsupp + .pru_abort = tcp_usr_abort, + .pru_accept = tcp_usr_accept, + .pru_attach = tcp_usr_attach, + .pru_bind = tcp_usr_bind, + .pru_connect = tcp_usr_connect, + .pru_connectx = tcp_usr_connectx, + .pru_control = in_control, + .pru_detach = tcp_usr_detach, + .pru_disconnect = tcp_usr_disconnect, + .pru_disconnectx = tcp_usr_disconnectx, + .pru_listen = tcp_usr_listen, + .pru_peeraddr = in_getpeeraddr, + .pru_rcvd = tcp_usr_rcvd, + .pru_rcvoob = tcp_usr_rcvoob, + .pru_send = tcp_usr_send, + .pru_shutdown = tcp_usr_shutdown, + .pru_sockaddr = in_getsockaddr, + .pru_sosend = sosend, + .pru_soreceive = soreceive, + .pru_preconnect = tcp_usr_preconnect, }; #if INET6 struct pr_usrreqs tcp6_usrreqs = { - tcp_usr_abort, tcp6_usr_accept, tcp_usr_attach, tcp6_usr_bind, - tcp6_usr_connect, pru_connect2_notsupp, in6_control, tcp_usr_detach, - tcp_usr_disconnect, tcp6_usr_listen, in6_mapped_peeraddr, tcp_usr_rcvd, - tcp_usr_rcvoob, tcp_usr_send, pru_sense_null, tcp_usr_shutdown, - in6_mapped_sockaddr, sosend, soreceive, pru_sopoll_notsupp + .pru_abort = tcp_usr_abort, + .pru_accept = tcp6_usr_accept, + .pru_attach = tcp_usr_attach, + .pru_bind = tcp6_usr_bind, + .pru_connect = tcp6_usr_connect, + .pru_connectx = tcp6_usr_connectx, + .pru_control = in6_control, + .pru_detach = tcp_usr_detach, + .pru_disconnect = tcp_usr_disconnect, + .pru_disconnectx = tcp_usr_disconnectx, + .pru_listen = tcp6_usr_listen, + .pru_peeraddr = in6_mapped_peeraddr, + .pru_rcvd = tcp_usr_rcvd, + .pru_rcvoob = tcp_usr_rcvoob, + .pru_send = tcp_usr_send, + .pru_shutdown = tcp_usr_shutdown, + .pru_sockaddr = in6_mapped_sockaddr, + .pru_sosend = sosend, + .pru_soreceive = soreceive, + .pru_preconnect = tcp_usr_preconnect, }; #endif /* INET6 */ @@ -714,26 +1296,36 @@ struct pr_usrreqs tcp6_usrreqs = { * sending CC options and if the connection duration was < MSL, then * truncate the previous TIME-WAIT state and proceed. * Initialize connection parameters and enter SYN-SENT state. + * + * Returns: 0 Success + * EADDRINUSE + * EINVAL + * in_pcbbind:EADDRNOTAVAIL Address not available. + * in_pcbbind:EINVAL Invalid argument + * in_pcbbind:EAFNOSUPPORT Address family not supported [notdef] + * in_pcbbind:EACCES Permission denied + * in_pcbbind:EADDRINUSE Address in use + * in_pcbbind:EAGAIN Resource unavailable, try again + * in_pcbbind:EPERM Operation not permitted + * in_pcbladdr:EINVAL Invalid argument + * in_pcbladdr:EAFNOSUPPORT Address family not supported + * in_pcbladdr:EADDRNOTAVAIL Address not available */ static int -tcp_connect(tp, nam, p) - register struct tcpcb *tp; - struct sockaddr *nam; - struct proc *p; +tcp_connect(struct tcpcb *tp, struct sockaddr *nam, struct proc *p) { struct inpcb *inp = tp->t_inpcb, *oinp; struct socket *so = inp->inp_socket; struct tcpcb *otp; - struct sockaddr_in *sin = (struct sockaddr_in *)nam; - struct sockaddr_in *ifaddr; - struct rmxp_tao *taop; - struct rmxp_tao tao_noncached; - int error; + struct sockaddr_in *sin = (struct sockaddr_in *)(void *)nam; + struct in_addr laddr; + int error = 0; + struct ifnet *outif = NULL; if (inp->inp_lport == 0) { - error = in_pcbbind(inp, (struct sockaddr *)0, p); + error = in_pcbbind(inp, NULL, p); if (error) - return error; + goto done; } /* @@ -741,15 +1333,14 @@ tcp_connect(tp, nam, p) * earlier incarnation of this same connection still in * TIME_WAIT state, creating an ADDRINUSE error. */ - error = in_pcbladdr(inp, nam, &ifaddr); + error = in_pcbladdr(inp, nam, &laddr, IFSCOPE_NONE, &outif, 0); if (error) - return error; + goto done; tcp_unlock(inp->inp_socket, 0, 0); oinp = in_pcblookup_hash(inp->inp_pcbinfo, sin->sin_addr, sin->sin_port, - inp->inp_laddr.s_addr != INADDR_ANY ? inp->inp_laddr - : ifaddr->sin_addr, + inp->inp_laddr.s_addr != INADDR_ANY ? inp->inp_laddr : laddr, inp->inp_lport, 0, NULL); tcp_lock(inp->inp_socket, 0, 0); @@ -763,100 +1354,96 @@ tcp_connect(tp, nam, p) } if (oinp != inp && (otp = intotcpcb(oinp)) != NULL && - otp->t_state == TCPS_TIME_WAIT && - otp->t_starttime < tcp_msl && - (otp->t_flags & TF_RCVD_CC)) + otp->t_state == TCPS_TIME_WAIT && + ((int)(tcp_now - otp->t_starttime)) < tcp_msl && + (otp->t_flags & TF_RCVD_CC)) { otp = tcp_close(otp); - else { - printf("tcp_connect: inp=%x err=EADDRINUSE\n", inp); + } else { + printf("tcp_connect: inp=0x%llx err=EADDRINUSE\n", + (uint64_t)VM_KERNEL_ADDRPERM(inp)); if (oinp != inp) tcp_unlock(oinp->inp_socket, 1, 0); - return EADDRINUSE; + error = EADDRINUSE; + goto done; } if (oinp != inp) tcp_unlock(oinp->inp_socket, 1, 0); } skip_oinp: - if ((inp->inp_laddr.s_addr == INADDR_ANY ? ifaddr->sin_addr.s_addr : - inp->inp_laddr.s_addr) == sin->sin_addr.s_addr && - inp->inp_lport == sin->sin_port) - return EINVAL; - if (!lck_rw_try_lock_exclusive(inp->inp_pcbinfo->mtx)) { + if ((inp->inp_laddr.s_addr == INADDR_ANY ? laddr.s_addr : + inp->inp_laddr.s_addr) == sin->sin_addr.s_addr && + inp->inp_lport == sin->sin_port) { + error = EINVAL; + goto done; + } + if (!lck_rw_try_lock_exclusive(inp->inp_pcbinfo->ipi_lock)) { /*lock inversion issue, mostly with udp multicast packets */ socket_unlock(inp->inp_socket, 0); - lck_rw_lock_exclusive(inp->inp_pcbinfo->mtx); + lck_rw_lock_exclusive(inp->inp_pcbinfo->ipi_lock); socket_lock(inp->inp_socket, 0); } - if (inp->inp_laddr.s_addr == INADDR_ANY) - inp->inp_laddr = ifaddr->sin_addr; + if (inp->inp_laddr.s_addr == INADDR_ANY) { + inp->inp_laddr = laddr; + /* no reference needed */ + inp->inp_last_outifp = outif; + inp->inp_flags |= INP_INADDR_ANY; + } inp->inp_faddr = sin->sin_addr; inp->inp_fport = sin->sin_port; in_pcbrehash(inp); - lck_rw_done(inp->inp_pcbinfo->mtx); + lck_rw_done(inp->inp_pcbinfo->ipi_lock); + + if (inp->inp_flowhash == 0) + inp->inp_flowhash = inp_calc_flowhash(inp); - /* Compute window scaling to request. */ - while (tp->request_r_scale < TCP_MAX_WINSHIFT && - (TCP_MAXWIN << tp->request_r_scale) < so->so_rcv.sb_hiwat) - tp->request_r_scale++; + tcp_set_max_rwinscale(tp, so, TCP_AUTORCVBUF_MAX(outif)); soisconnecting(so); tcpstat.tcps_connattempt++; tp->t_state = TCPS_SYN_SENT; - tp->t_timer[TCPT_KEEP] = tcp_keepinit; + tp->t_timer[TCPT_KEEP] = OFFSET_FROM_START(tp, TCP_CONN_KEEPINIT(tp)); tp->iss = tcp_new_isn(tp); tcp_sendseqinit(tp); + if (nstat_collect) + nstat_route_connect_attempt(inp->inp_route.ro_rt); - /* - * Generate a CC value for this connection and - * check whether CC or CCnew should be used. - */ - if ((taop = tcp_gettaocache(tp->t_inpcb)) == NULL) { - taop = &tao_noncached; - bzero(taop, sizeof(*taop)); - } - - tp->cc_send = CC_INC(tcp_ccgen); - if (taop->tao_ccsent != 0 && - CC_GEQ(tp->cc_send, taop->tao_ccsent)) { - taop->tao_ccsent = tp->cc_send; - } else { - taop->tao_ccsent = 0; - tp->t_flags |= TF_SENDCCNEW; - } +done: + if (outif != NULL) + ifnet_release(outif); - return 0; + return (error); } #if INET6 static int -tcp6_connect(tp, nam, p) - register struct tcpcb *tp; - struct sockaddr *nam; - struct proc *p; +tcp6_connect(struct tcpcb *tp, struct sockaddr *nam, struct proc *p) { struct inpcb *inp = tp->t_inpcb, *oinp; struct socket *so = inp->inp_socket; struct tcpcb *otp; - struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)nam; + struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)(void *)nam; struct in6_addr addr6; - struct rmxp_tao *taop; - struct rmxp_tao tao_noncached; - int error; + int error = 0; + struct ifnet *outif = NULL; if (inp->inp_lport == 0) { - error = in6_pcbbind(inp, (struct sockaddr *)0, p); + error = in6_pcbbind(inp, NULL, p); if (error) - return error; + goto done; } /* * Cannot simply call in_pcbconnect, because there might be an * earlier incarnation of this same connection still in * TIME_WAIT state, creating an ADDRINUSE error. + * + * in6_pcbladdr() might return an ifp with its reference held + * even in the error case, so make sure that it's released + * whenever it's non-NULL. */ - error = in6_pcbladdr(inp, nam, &addr6); + error = in6_pcbladdr(inp, nam, &addr6, &outif); if (error) - return error; + goto done; tcp_unlock(inp->inp_socket, 0, 0); oinp = in6_pcblookup_hash(inp->inp_pcbinfo, &sin6->sin6_addr, sin6->sin6_port, @@ -868,60 +1455,421 @@ tcp6_connect(tp, nam, p) if (oinp) { if (oinp != inp && (otp = intotcpcb(oinp)) != NULL && otp->t_state == TCPS_TIME_WAIT && - otp->t_starttime < tcp_msl && - (otp->t_flags & TF_RCVD_CC)) + ((int)(tcp_now - otp->t_starttime)) < tcp_msl && + (otp->t_flags & TF_RCVD_CC)) { otp = tcp_close(otp); - else - return EADDRINUSE; + } else { + error = EADDRINUSE; + goto done; + } } - if (!lck_rw_try_lock_exclusive(inp->inp_pcbinfo->mtx)) { + if (!lck_rw_try_lock_exclusive(inp->inp_pcbinfo->ipi_lock)) { /*lock inversion issue, mostly with udp multicast packets */ socket_unlock(inp->inp_socket, 0); - lck_rw_lock_exclusive(inp->inp_pcbinfo->mtx); + lck_rw_lock_exclusive(inp->inp_pcbinfo->ipi_lock); socket_lock(inp->inp_socket, 0); } - if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)) + if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)) { inp->in6p_laddr = addr6; + inp->in6p_last_outifp = outif; /* no reference needed */ + inp->in6p_flags |= INP_IN6ADDR_ANY; + } inp->in6p_faddr = sin6->sin6_addr; inp->inp_fport = sin6->sin6_port; - if ((sin6->sin6_flowinfo & IPV6_FLOWINFO_MASK) != NULL) - inp->in6p_flowinfo = sin6->sin6_flowinfo; + if ((sin6->sin6_flowinfo & IPV6_FLOWINFO_MASK) != 0) + inp->inp_flow = sin6->sin6_flowinfo; in_pcbrehash(inp); - lck_rw_done(inp->inp_pcbinfo->mtx); + lck_rw_done(inp->inp_pcbinfo->ipi_lock); + + if (inp->inp_flowhash == 0) + inp->inp_flowhash = inp_calc_flowhash(inp); + /* update flowinfo - RFC 6437 */ + if (inp->inp_flow == 0 && inp->in6p_flags & IN6P_AUTOFLOWLABEL) { + inp->inp_flow &= ~IPV6_FLOWLABEL_MASK; + inp->inp_flow |= + (htonl(inp->inp_flowhash) & IPV6_FLOWLABEL_MASK); + } - /* Compute window scaling to request. */ - while (tp->request_r_scale < TCP_MAX_WINSHIFT && - (TCP_MAXWIN << tp->request_r_scale) < so->so_rcv.sb_hiwat) - tp->request_r_scale++; + tcp_set_max_rwinscale(tp, so, TCP_AUTORCVBUF_MAX(outif)); soisconnecting(so); tcpstat.tcps_connattempt++; tp->t_state = TCPS_SYN_SENT; - tp->t_timer[TCPT_KEEP] = tcp_keepinit; + tp->t_timer[TCPT_KEEP] = OFFSET_FROM_START(tp, + TCP_CONN_KEEPINIT(tp)); tp->iss = tcp_new_isn(tp); tcp_sendseqinit(tp); + if (nstat_collect) + nstat_route_connect_attempt(inp->inp_route.ro_rt); - /* - * Generate a CC value for this connection and - * check whether CC or CCnew should be used. - */ - if ((taop = tcp_gettaocache(tp->t_inpcb)) == NULL) { - taop = &tao_noncached; - bzero(taop, sizeof(*taop)); +done: + if (outif != NULL) + ifnet_release(outif); + + return (error); +} +#endif /* INET6 */ + +/* + * Export TCP internal state information via a struct tcp_info + */ +__private_extern__ void +tcp_fill_info(struct tcpcb *tp, struct tcp_info *ti) +{ + struct inpcb *inp = tp->t_inpcb; + + bzero(ti, sizeof(*ti)); + + ti->tcpi_state = tp->t_state; + ti->tcpi_flowhash = inp->inp_flowhash; + + if (tp->t_state > TCPS_LISTEN) { + if (TSTMP_SUPPORTED(tp)) + ti->tcpi_options |= TCPI_OPT_TIMESTAMPS; + if (SACK_ENABLED(tp)) + ti->tcpi_options |= TCPI_OPT_SACK; + if (TCP_WINDOW_SCALE_ENABLED(tp)) { + ti->tcpi_options |= TCPI_OPT_WSCALE; + ti->tcpi_snd_wscale = tp->snd_scale; + ti->tcpi_rcv_wscale = tp->rcv_scale; + } + if (TCP_ECN_ENABLED(tp)) + ti->tcpi_options |= TCPI_OPT_ECN; + + /* Are we in retranmission episode */ + if (IN_FASTRECOVERY(tp) || tp->t_rxtshift > 0) + ti->tcpi_flags |= TCPI_FLAG_LOSSRECOVERY; + + if (tp->t_flags & TF_STREAMING_ON) + ti->tcpi_flags |= TCPI_FLAG_STREAMING_ON; + + ti->tcpi_rto = tp->t_timer[TCPT_REXMT] ? tp->t_rxtcur : 0; + ti->tcpi_snd_mss = tp->t_maxseg; + ti->tcpi_rcv_mss = tp->t_maxseg; + + ti->tcpi_rttcur = tp->t_rttcur; + ti->tcpi_srtt = tp->t_srtt >> TCP_RTT_SHIFT; + ti->tcpi_rttvar = tp->t_rttvar >> TCP_RTTVAR_SHIFT; + ti->tcpi_rttbest = tp->t_rttbest >> TCP_RTT_SHIFT; + + ti->tcpi_snd_ssthresh = tp->snd_ssthresh; + ti->tcpi_snd_cwnd = tp->snd_cwnd; + ti->tcpi_snd_sbbytes = inp->inp_socket->so_snd.sb_cc; + + ti->tcpi_rcv_space = tp->rcv_wnd; + + ti->tcpi_snd_wnd = tp->snd_wnd; + ti->tcpi_snd_nxt = tp->snd_nxt; + ti->tcpi_rcv_nxt = tp->rcv_nxt; + + /* convert bytes/msec to bits/sec */ + if ((tp->t_flagsext & TF_MEASURESNDBW) != 0 && + tp->t_bwmeas != NULL) { + ti->tcpi_snd_bw = (tp->t_bwmeas->bw_sndbw * 8000); + } + + ti->tcpi_last_outif = (tp->t_inpcb->inp_last_outifp == NULL) ? 0 : + tp->t_inpcb->inp_last_outifp->if_index; + + //atomic_get_64(ti->tcpi_txbytes, &inp->inp_stat->txbytes); + ti->tcpi_txpackets = inp->inp_stat->txpackets; + ti->tcpi_txbytes = inp->inp_stat->txbytes; + ti->tcpi_txretransmitbytes = tp->t_stat.txretransmitbytes; + ti->tcpi_txunacked = tp->snd_max - tp->snd_una; + + //atomic_get_64(ti->tcpi_rxbytes, &inp->inp_stat->rxbytes); + ti->tcpi_rxpackets = inp->inp_stat->rxpackets; + ti->tcpi_rxbytes = inp->inp_stat->rxbytes; + ti->tcpi_rxduplicatebytes = tp->t_stat.rxduplicatebytes; + ti->tcpi_rxoutoforderbytes = tp->t_stat.rxoutoforderbytes; + + if (tp->t_state > TCPS_LISTEN) { + ti->tcpi_synrexmits = tp->t_stat.synrxtshift; + } + ti->tcpi_cell_rxpackets = inp->inp_cstat->rxpackets; + ti->tcpi_cell_rxbytes = inp->inp_cstat->rxbytes; + ti->tcpi_cell_txpackets = inp->inp_cstat->txpackets; + ti->tcpi_cell_txbytes = inp->inp_cstat->txbytes; + + ti->tcpi_wifi_rxpackets = inp->inp_wstat->rxpackets; + ti->tcpi_wifi_rxbytes = inp->inp_wstat->rxbytes; + ti->tcpi_wifi_txpackets = inp->inp_wstat->txpackets; + ti->tcpi_wifi_txbytes = inp->inp_wstat->txbytes; + + ti->tcpi_wired_rxpackets = inp->inp_Wstat->rxpackets; + ti->tcpi_wired_rxbytes = inp->inp_Wstat->rxbytes; + ti->tcpi_wired_txpackets = inp->inp_Wstat->txpackets; + ti->tcpi_wired_txbytes = inp->inp_Wstat->txbytes; + tcp_get_connectivity_status(tp, &ti->tcpi_connstatus); + + ti->tcpi_tfo_syn_data_rcv = !!(tp->t_tfo_stats & TFO_S_SYNDATA_RCV); + ti->tcpi_tfo_cookie_req_rcv = !!(tp->t_tfo_stats & TFO_S_COOKIEREQ_RECV); + ti->tcpi_tfo_cookie_sent = !!(tp->t_tfo_stats & TFO_S_COOKIE_SENT); + ti->tcpi_tfo_cookie_invalid = !!(tp->t_tfo_stats & TFO_S_COOKIE_INVALID); + + ti->tcpi_tfo_cookie_req = !!(tp->t_tfo_stats & TFO_S_COOKIE_REQ); + ti->tcpi_tfo_cookie_rcv = !!(tp->t_tfo_stats & TFO_S_COOKIE_RCV); + ti->tcpi_tfo_syn_data_sent = !!(tp->t_tfo_stats & TFO_S_SYN_DATA_SENT); + ti->tcpi_tfo_syn_data_acked = !!(tp->t_tfo_stats & TFO_S_SYN_DATA_ACKED); + ti->tcpi_tfo_syn_loss = !!(tp->t_tfo_stats & TFO_S_SYN_LOSS); + ti->tcpi_tfo_cookie_wrong = !!(tp->t_tfo_stats & TFO_S_COOKIE_WRONG); + ti->tcpi_tfo_no_cookie_rcv = !!(tp->t_tfo_stats & TFO_S_NO_COOKIE_RCV); + ti->tcpi_tfo_heuristics_disable = !!(tp->t_tfo_stats & TFO_S_HEURISTICS_DISABLE); + ti->tcpi_tfo_send_blackhole = !!(tp->t_tfo_stats & TFO_S_SEND_BLACKHOLE); + ti->tcpi_tfo_recv_blackhole = !!(tp->t_tfo_stats & TFO_S_RECV_BLACKHOLE); + + ti->tcpi_ecn_client_setup = !!(tp->ecn_flags & TE_SETUPSENT); + ti->tcpi_ecn_server_setup = !!(tp->ecn_flags & TE_SETUPRECEIVED); + ti->tcpi_ecn_success = (tp->ecn_flags & TE_ECN_ON) == TE_ECN_ON ? 1 : 0; + ti->tcpi_ecn_lost_syn = !!(tp->ecn_flags & TE_LOST_SYN); + ti->tcpi_ecn_lost_synack = !!(tp->ecn_flags & TE_LOST_SYNACK); + + ti->tcpi_local_peer = !!(tp->t_flags & TF_LOCAL); + + if (tp->t_inpcb->inp_last_outifp != NULL) { + if (IFNET_IS_CELLULAR(tp->t_inpcb->inp_last_outifp)) + ti->tcpi_if_cell = 1; + else if (IFNET_IS_WIFI(tp->t_inpcb->inp_last_outifp)) + ti->tcpi_if_wifi = 1; + } + + ti->tcpi_ecn_recv_ce = tp->t_ecn_recv_ce; + ti->tcpi_ecn_recv_cwr = tp->t_ecn_recv_cwr; + + ti->tcpi_rcvoopack = tp->t_rcvoopack; + ti->tcpi_pawsdrop = tp->t_pawsdrop; + ti->tcpi_sack_recovery_episode = tp->t_sack_recovery_episode; + ti->tcpi_reordered_pkts = tp->t_reordered_pkts; + ti->tcpi_dsack_sent = tp->t_dsack_sent; + ti->tcpi_dsack_recvd = tp->t_dsack_recvd; } +} - tp->cc_send = CC_INC(tcp_ccgen); - if (taop->tao_ccsent != 0 && - CC_GEQ(tp->cc_send, taop->tao_ccsent)) { - taop->tao_ccsent = tp->cc_send; +__private_extern__ errno_t +tcp_fill_info_for_info_tuple(struct info_tuple *itpl, struct tcp_info *ti) +{ + struct inpcbinfo *pcbinfo = NULL; + struct inpcb *inp = NULL; + struct socket *so; + struct tcpcb *tp; + + if (itpl->itpl_proto == IPPROTO_TCP) + pcbinfo = &tcbinfo; + else + return EINVAL; + + if (itpl->itpl_local_sa.sa_family == AF_INET && + itpl->itpl_remote_sa.sa_family == AF_INET) { + inp = in_pcblookup_hash(pcbinfo, + itpl->itpl_remote_sin.sin_addr, + itpl->itpl_remote_sin.sin_port, + itpl->itpl_local_sin.sin_addr, + itpl->itpl_local_sin.sin_port, + 0, NULL); + } else if (itpl->itpl_local_sa.sa_family == AF_INET6 && + itpl->itpl_remote_sa.sa_family == AF_INET6) { + struct in6_addr ina6_local; + struct in6_addr ina6_remote; + + ina6_local = itpl->itpl_local_sin6.sin6_addr; + if (IN6_IS_SCOPE_LINKLOCAL(&ina6_local) && + itpl->itpl_local_sin6.sin6_scope_id) + ina6_local.s6_addr16[1] = htons(itpl->itpl_local_sin6.sin6_scope_id); + + ina6_remote = itpl->itpl_remote_sin6.sin6_addr; + if (IN6_IS_SCOPE_LINKLOCAL(&ina6_remote) && + itpl->itpl_remote_sin6.sin6_scope_id) + ina6_remote.s6_addr16[1] = htons(itpl->itpl_remote_sin6.sin6_scope_id); + + inp = in6_pcblookup_hash(pcbinfo, + &ina6_remote, + itpl->itpl_remote_sin6.sin6_port, + &ina6_local, + itpl->itpl_local_sin6.sin6_port, + 0, NULL); } else { - taop->tao_ccsent = 0; - tp->t_flags |= TF_SENDCCNEW; + return EINVAL; } + if (inp == NULL || (so = inp->inp_socket) == NULL) + return ENOENT; + + socket_lock(so, 0); + if (in_pcb_checkstate(inp, WNT_RELEASE, 1) == WNT_STOPUSING) { + socket_unlock(so, 0); + return ENOENT; + } + tp = intotcpcb(inp); + + tcp_fill_info(tp, ti); + socket_unlock(so, 0); return 0; } -#endif /* INET6 */ + +static void +tcp_connection_fill_info(struct tcpcb *tp, struct tcp_connection_info *tci) +{ + struct inpcb *inp = tp->t_inpcb; + + bzero(tci, sizeof(*tci)); + tci->tcpi_state = tp->t_state; + if (tp->t_state > TCPS_LISTEN) { + if (TSTMP_SUPPORTED(tp)) + tci->tcpi_options |= TCPCI_OPT_TIMESTAMPS; + if (SACK_ENABLED(tp)) + tci->tcpi_options |= TCPCI_OPT_SACK; + if (TCP_WINDOW_SCALE_ENABLED(tp)) { + tci->tcpi_options |= TCPCI_OPT_WSCALE; + tci->tcpi_snd_wscale = tp->snd_scale; + tci->tcpi_rcv_wscale = tp->rcv_scale; + } + if (TCP_ECN_ENABLED(tp)) + tci->tcpi_options |= TCPCI_OPT_ECN; + if (IN_FASTRECOVERY(tp) || tp->t_rxtshift > 0) + tci->tcpi_flags |= TCPCI_FLAG_LOSSRECOVERY; + if (tp->t_flagsext & TF_PKTS_REORDERED) + tci->tcpi_flags |= TCPCI_FLAG_REORDERING_DETECTED; + tci->tcpi_rto = (tp->t_timer[TCPT_REXMT] > 0) ? + tp->t_rxtcur : 0; + tci->tcpi_maxseg = tp->t_maxseg; + tci->tcpi_snd_ssthresh = tp->snd_ssthresh; + tci->tcpi_snd_cwnd = tp->snd_cwnd; + tci->tcpi_snd_wnd = tp->snd_wnd; + tci->tcpi_snd_sbbytes = inp->inp_socket->so_snd.sb_cc; + tci->tcpi_rcv_wnd = tp->rcv_wnd; + tci->tcpi_rttcur = tp->t_rttcur; + tci->tcpi_srtt = (tp->t_srtt >> TCP_RTT_SHIFT); + tci->tcpi_rttvar = (tp->t_rttvar >> TCP_RTTVAR_SHIFT); + tci->tcpi_txpackets = inp->inp_stat->txpackets; + tci->tcpi_txbytes = inp->inp_stat->txbytes; + tci->tcpi_txretransmitbytes = tp->t_stat.txretransmitbytes; + tci->tcpi_rxpackets = inp->inp_stat->rxpackets; + tci->tcpi_rxbytes = inp->inp_stat->rxbytes; + tci->tcpi_rxoutoforderbytes = tp->t_stat.rxoutoforderbytes; + + tci->tcpi_tfo_syn_data_rcv = !!(tp->t_tfo_stats & TFO_S_SYNDATA_RCV); + tci->tcpi_tfo_cookie_req_rcv = !!(tp->t_tfo_stats & TFO_S_COOKIEREQ_RECV); + tci->tcpi_tfo_cookie_sent = !!(tp->t_tfo_stats & TFO_S_COOKIE_SENT); + tci->tcpi_tfo_cookie_invalid = !!(tp->t_tfo_stats & TFO_S_COOKIE_INVALID); + tci->tcpi_tfo_cookie_req = !!(tp->t_tfo_stats & TFO_S_COOKIE_REQ); + tci->tcpi_tfo_cookie_rcv = !!(tp->t_tfo_stats & TFO_S_COOKIE_RCV); + tci->tcpi_tfo_syn_data_sent = !!(tp->t_tfo_stats & TFO_S_SYN_DATA_SENT); + tci->tcpi_tfo_syn_data_acked = !!(tp->t_tfo_stats & TFO_S_SYN_DATA_ACKED); + tci->tcpi_tfo_syn_loss = !!(tp->t_tfo_stats & TFO_S_SYN_LOSS); + tci->tcpi_tfo_cookie_wrong = !!(tp->t_tfo_stats & TFO_S_COOKIE_WRONG); + tci->tcpi_tfo_no_cookie_rcv = !!(tp->t_tfo_stats & TFO_S_NO_COOKIE_RCV); + tci->tcpi_tfo_heuristics_disable = !!(tp->t_tfo_stats & TFO_S_HEURISTICS_DISABLE); + tci->tcpi_tfo_send_blackhole = !!(tp->t_tfo_stats & TFO_S_SEND_BLACKHOLE); + tci->tcpi_tfo_recv_blackhole = !!(tp->t_tfo_stats & TFO_S_RECV_BLACKHOLE); + } +} + + +__private_extern__ int +tcp_sysctl_info(__unused struct sysctl_oid *oidp, __unused void *arg1, __unused int arg2, struct sysctl_req *req) +{ + int error; + struct tcp_info ti; + struct info_tuple itpl; + proc_t caller = PROC_NULL; + proc_t caller_parent = PROC_NULL; + char command_name[MAXCOMLEN + 1] = ""; + char parent_name[MAXCOMLEN + 1] = ""; + + if ((caller = proc_self()) != PROC_NULL) { + /* get process name */ + strlcpy(command_name, caller->p_comm, sizeof(command_name)); + + /* get parent process name if possible */ + if ((caller_parent = proc_find(caller->p_ppid)) != PROC_NULL) { + strlcpy(parent_name, caller_parent->p_comm, + sizeof(parent_name)); + proc_rele(caller_parent); + } + + if ((escape_str(command_name, strlen(command_name), + sizeof(command_name)) == 0) && + (escape_str(parent_name, strlen(parent_name), + sizeof(parent_name)) == 0)) { + kern_asl_msg(LOG_DEBUG, "messagetracer", + 5, + "com.apple.message.domain", + "com.apple.kernel.tcpstat", /* 1 */ + "com.apple.message.signature", + "tcpinfo", /* 2 */ + "com.apple.message.signature2", command_name, /* 3 */ + "com.apple.message.signature3", parent_name, /* 4 */ + "com.apple.message.summarize", "YES", /* 5 */ + NULL); + } + } + + if (caller != PROC_NULL) + proc_rele(caller); + + if (req->newptr == USER_ADDR_NULL) { + return EINVAL; + } + if (req->newlen < sizeof(struct info_tuple)) { + return EINVAL; + } + error = SYSCTL_IN(req, &itpl, sizeof(struct info_tuple)); + if (error != 0) { + return error; + } + error = tcp_fill_info_for_info_tuple(&itpl, &ti); + if (error != 0) { + return error; + } + error = SYSCTL_OUT(req, &ti, sizeof(struct tcp_info)); + if (error != 0) { + return error; + } + + return 0; +} + +static int +tcp_lookup_peer_pid_locked(struct socket *so, pid_t *out_pid) +{ + int error = EHOSTUNREACH; + *out_pid = -1; + if ((so->so_state & SS_ISCONNECTED) == 0) return ENOTCONN; + + struct inpcb *inp = (struct inpcb*)so->so_pcb; + uint16_t lport = inp->inp_lport; + uint16_t fport = inp->inp_fport; + struct inpcb *finp = NULL; + + if (inp->inp_vflag & INP_IPV6) { + struct in6_addr laddr6 = inp->in6p_laddr; + struct in6_addr faddr6 = inp->in6p_faddr; + socket_unlock(so, 0); + finp = in6_pcblookup_hash(&tcbinfo, &laddr6, lport, &faddr6, fport, 0, NULL); + socket_lock(so, 0); + } else if (inp->inp_vflag & INP_IPV4) { + struct in_addr laddr4 = inp->inp_laddr; + struct in_addr faddr4 = inp->inp_faddr; + socket_unlock(so, 0); + finp = in_pcblookup_hash(&tcbinfo, laddr4, lport, faddr4, fport, 0, NULL); + socket_lock(so, 0); + } + + if (finp) { + *out_pid = finp->inp_socket->last_pid; + error = 0; + in_pcb_checkstate(finp, WNT_RELEASE, 0); + } + + return error; +} + +void +tcp_getconninfo(struct socket *so, struct conninfo_tcp *tcp_ci) +{ + (void) tcp_lookup_peer_pid_locked(so, &tcp_ci->tcpci_peer_pid); + tcp_fill_info(sototcpcb(so), &tcp_ci->tcpci_tcp_info); +} /* * The new sockopt interface makes it possible for us to block in the @@ -931,9 +1879,7 @@ tcp6_connect(tp, nam, p) * splnet() any more. This needs more examination.) */ int -tcp_ctloutput(so, sopt) - struct socket *so; - struct sockopt *sopt; +tcp_ctloutput(struct socket *so, struct sockopt *sopt) { int error, opt, optval; struct inpcb *inp; @@ -944,9 +1890,12 @@ tcp_ctloutput(so, sopt) if (inp == NULL) { return (ECONNRESET); } - if (sopt->sopt_level != IPPROTO_TCP) { + /* Allow at this level */ + if (sopt->sopt_level != IPPROTO_TCP && + !(sopt->sopt_level == SOL_SOCKET && (sopt->sopt_name == SO_FLUSH || + sopt->sopt_name == SO_TRAFFIC_MGT_BACKGROUND))) { #if INET6 - if (INP_CHECK_SOCKAF(so, AF_INET6)) + if (SOCK_CHECK_DOM(so, PF_INET6)) error = ip6_ctloutput(so, sopt); else #endif /* INET6 */ @@ -954,9 +1903,11 @@ tcp_ctloutput(so, sopt) return (error); } tp = intotcpcb(inp); - if (tp == NULL) { - return (ECONNRESET); - } + if (tp == NULL) { + return (ECONNRESET); + } + + calculate_tcp_clock(); switch (sopt->sopt_dir) { case SOPT_SET: @@ -989,7 +1940,80 @@ tcp_ctloutput(so, sopt) else tp->t_flags &= ~opt; break; + case TCP_RXT_FINDROP: + case TCP_NOTIMEWAIT: + error = sooptcopyin(sopt, &optval, sizeof optval, + sizeof optval); + if (error) + break; + switch (sopt->sopt_name) { + case TCP_RXT_FINDROP: + opt = TF_RXTFINDROP; + break; + case TCP_NOTIMEWAIT: + opt = TF_NOTIMEWAIT; + break; + default: + opt = 0; + break; + } + if (optval) + tp->t_flagsext |= opt; + else + tp->t_flagsext &= ~opt; + break; + case TCP_MEASURE_SND_BW: + error = sooptcopyin(sopt, &optval, sizeof optval, + sizeof optval); + if (error) + break; + opt = TF_MEASURESNDBW; + if (optval) { + if (tp->t_bwmeas == NULL) { + tp->t_bwmeas = tcp_bwmeas_alloc(tp); + if (tp->t_bwmeas == NULL) { + error = ENOMEM; + break; + } + } + tp->t_flagsext |= opt; + } else { + tp->t_flagsext &= ~opt; + /* Reset snd bw measurement state */ + tp->t_flagsext &= ~(TF_BWMEAS_INPROGRESS); + if (tp->t_bwmeas != NULL) { + tcp_bwmeas_free(tp); + } + } + break; + case TCP_MEASURE_BW_BURST: { + struct tcp_measure_bw_burst in; + uint32_t minpkts, maxpkts; + bzero(&in, sizeof(in)); + error = sooptcopyin(sopt, &in, sizeof(in), + sizeof(in)); + if (error) + break; + if ((tp->t_flagsext & TF_MEASURESNDBW) == 0 || + tp->t_bwmeas == NULL) { + error = EINVAL; + break; + } + minpkts = (in.min_burst_size != 0) ? in.min_burst_size : + tp->t_bwmeas->bw_minsizepkts; + maxpkts = (in.max_burst_size != 0) ? in.max_burst_size : + tp->t_bwmeas->bw_maxsizepkts; + if (minpkts > maxpkts) { + error = EINVAL; + break; + } + tp->t_bwmeas->bw_minsizepkts = minpkts; + tp->t_bwmeas->bw_maxsizepkts = maxpkts; + tp->t_bwmeas->bw_minsize = (minpkts * tp->t_maxseg); + tp->t_bwmeas->bw_maxsize = (maxpkts * tp->t_maxseg); + break; + } case TCP_MAXSEG: error = sooptcopyin(sopt, &optval, sizeof optval, sizeof optval); @@ -1003,17 +2027,328 @@ tcp_ctloutput(so, sopt) error = EINVAL; break; - case TCP_KEEPALIVE: - error = sooptcopyin(sopt, &optval, sizeof optval, - sizeof optval); - if (error) - break; - if (optval < 0) - error = EINVAL; - else - tp->t_keepidle = optval * PR_SLOWHZ; + case TCP_KEEPALIVE: + error = sooptcopyin(sopt, &optval, sizeof optval, + sizeof optval); + if (error) + break; + if (optval < 0 || optval > UINT32_MAX/TCP_RETRANSHZ) { + error = EINVAL; + } else { + tp->t_keepidle = optval * TCP_RETRANSHZ; + /* reset the timer to new value */ + tp->t_timer[TCPT_KEEP] = OFFSET_FROM_START(tp, + TCP_CONN_KEEPIDLE(tp)); + tcp_check_timer_state(tp); + } break; - + + case TCP_CONNECTIONTIMEOUT: + error = sooptcopyin(sopt, &optval, sizeof optval, + sizeof optval); + if (error) + break; + if (optval < 0 || optval > UINT32_MAX/TCP_RETRANSHZ) { + error = EINVAL; + } else { + tp->t_keepinit = optval * TCP_RETRANSHZ; + if (tp->t_state == TCPS_SYN_RECEIVED || + tp->t_state == TCPS_SYN_SENT) { + tp->t_timer[TCPT_KEEP] = OFFSET_FROM_START(tp, + TCP_CONN_KEEPINIT(tp)); + tcp_check_timer_state(tp); + } + } + break; + + case TCP_KEEPINTVL: + error = sooptcopyin(sopt, &optval, sizeof(optval), + sizeof(optval)); + if (error) + break; + if (optval < 0 || optval > UINT32_MAX/TCP_RETRANSHZ) { + error = EINVAL; + } else { + tp->t_keepintvl = optval * TCP_RETRANSHZ; + if (tp->t_state == TCPS_FIN_WAIT_2 && + TCP_CONN_MAXIDLE(tp) > 0) { + tp->t_timer[TCPT_2MSL] = OFFSET_FROM_START(tp, + TCP_CONN_MAXIDLE(tp)); + tcp_check_timer_state(tp); + } + } + break; + + case TCP_KEEPCNT: + error = sooptcopyin(sopt, &optval, sizeof(optval), + sizeof(optval)); + if (error) + break; + if (optval < 0 || optval > INT32_MAX) { + error = EINVAL; + } else { + tp->t_keepcnt = optval; + if (tp->t_state == TCPS_FIN_WAIT_2 && + TCP_CONN_MAXIDLE(tp) > 0) { + tp->t_timer[TCPT_2MSL] = OFFSET_FROM_START(tp, + TCP_CONN_MAXIDLE(tp)); + tcp_check_timer_state(tp); + } + } + break; + + case TCP_KEEPALIVE_OFFLOAD: + error = sooptcopyin(sopt, &optval, sizeof(optval), + sizeof(optval)); + if (error) + break; + if (optval < 0 || optval > INT32_MAX) { + error = EINVAL; + break; + } + if (optval != 0) + inp->inp_flags2 |= INP2_KEEPALIVE_OFFLOAD; + else + inp->inp_flags2 &= ~INP2_KEEPALIVE_OFFLOAD; + break; + + case PERSIST_TIMEOUT: + error = sooptcopyin(sopt, &optval, sizeof optval, + sizeof optval); + if (error) + break; + if (optval < 0) + error = EINVAL; + else + tp->t_persist_timeout = optval * TCP_RETRANSHZ; + break; + case TCP_RXT_CONNDROPTIME: + error = sooptcopyin(sopt, &optval, sizeof(optval), + sizeof(optval)); + if (error) + break; + if (optval < 0) + error = EINVAL; + else + tp->t_rxt_conndroptime = optval * TCP_RETRANSHZ; + break; + case TCP_NOTSENT_LOWAT: + error = sooptcopyin(sopt, &optval, sizeof(optval), + sizeof(optval)); + if (error) + break; + if (optval < 0) { + error = EINVAL; + break; + } else { + if (optval == 0) { + so->so_flags &= ~(SOF_NOTSENT_LOWAT); + tp->t_notsent_lowat = 0; + } else { + so->so_flags |= SOF_NOTSENT_LOWAT; + tp->t_notsent_lowat = optval; + } + } + break; + case TCP_ADAPTIVE_READ_TIMEOUT: + error = sooptcopyin(sopt, &optval, sizeof (optval), + sizeof(optval)); + if (error) + break; + if (optval < 0 || + optval > TCP_ADAPTIVE_TIMEOUT_MAX) { + error = EINVAL; + break; + } else if (optval == 0) { + tp->t_adaptive_rtimo = 0; + tcp_keepalive_reset(tp); + } else { + tp->t_adaptive_rtimo = optval; + } + break; + case TCP_ADAPTIVE_WRITE_TIMEOUT: + error = sooptcopyin(sopt, &optval, sizeof (optval), + sizeof (optval)); + if (error) + break; + if (optval < 0 || + optval > TCP_ADAPTIVE_TIMEOUT_MAX) { + error = EINVAL; + break; + } else { + tp->t_adaptive_wtimo = optval; + } + break; + case TCP_ENABLE_MSGS: + error = sooptcopyin(sopt, &optval, sizeof(optval), + sizeof(optval)); + if (error) + break; + if (optval < 0 || optval > 1) { + error = EINVAL; + } else if (optval == 1) { + /* + * Check if messages option is already + * enabled, if so return. + */ + if (so->so_flags & SOF_ENABLE_MSGS) { + VERIFY(so->so_msg_state != NULL); + break; + } + + /* + * allocate memory for storing message + * related state + */ + VERIFY(so->so_msg_state == NULL); + MALLOC(so->so_msg_state, + struct msg_state *, + sizeof(struct msg_state), + M_TEMP, M_WAITOK | M_ZERO); + if (so->so_msg_state == NULL) { + error = ENOMEM; + break; + } + + /* Enable message delivery */ + so->so_flags |= SOF_ENABLE_MSGS; + } else { + /* + * Can't disable message delivery on socket + * because of restrictions imposed by + * encoding/decoding + */ + error = EINVAL; + } + break; + case TCP_SENDMOREACKS: + error = sooptcopyin(sopt, &optval, sizeof(optval), + sizeof(optval)); + if (error) + break; + if (optval < 0 || optval > 1) { + error = EINVAL; + } else if (optval == 0) { + tp->t_flagsext &= ~(TF_NOSTRETCHACK); + } else { + tp->t_flagsext |= TF_NOSTRETCHACK; + } + break; + case TCP_DISABLE_BLACKHOLE_DETECTION: + error = sooptcopyin(sopt, &optval, sizeof(optval), + sizeof(optval)); + if (error) + break; + if (optval < 0 || optval > 1) { + error = EINVAL; + } else if (optval == 0) { + tp->t_flagsext &= ~TF_NOBLACKHOLE_DETECTION; + } else { + tp->t_flagsext |= TF_NOBLACKHOLE_DETECTION; + if ((tp->t_flags & TF_BLACKHOLE) && + tp->t_pmtud_saved_maxopd > 0) + tcp_pmtud_revert_segment_size(tp); + } + break; + case TCP_FASTOPEN: + if (!(tcp_fastopen & TCP_FASTOPEN_SERVER)) { + error = ENOTSUP; + break; + } + + error = sooptcopyin(sopt, &optval, sizeof(optval), + sizeof(optval)); + if (error) + break; + if (optval < 0 || optval > 1) { + error = EINVAL; + break; + } + if (tp->t_state != TCPS_LISTEN) { + error = EINVAL; + break; + } + if (optval) + tp->t_flagsext |= TF_FASTOPEN; + else + tcp_disable_tfo(tp); + break; + case TCP_ENABLE_ECN: + error = sooptcopyin(sopt, &optval, sizeof optval, + sizeof optval); + if (error) + break; + if (optval) { + tp->ecn_flags |= TE_ECN_MODE_ENABLE; + tp->ecn_flags &= ~TE_ECN_MODE_DISABLE; + } else { + tp->ecn_flags &= ~TE_ECN_MODE_ENABLE; + } + break; + case TCP_ECN_MODE: + error = sooptcopyin(sopt, &optval, sizeof optval, + sizeof optval); + if (error) + break; + if (optval == ECN_MODE_DEFAULT) { + tp->ecn_flags &= ~TE_ECN_MODE_ENABLE; + tp->ecn_flags &= ~TE_ECN_MODE_DISABLE; + } else if (optval == ECN_MODE_ENABLE) { + tp->ecn_flags |= TE_ECN_MODE_ENABLE; + tp->ecn_flags &= ~TE_ECN_MODE_DISABLE; + } else if (optval == ECN_MODE_DISABLE) { + tp->ecn_flags &= ~TE_ECN_MODE_ENABLE; + tp->ecn_flags |= TE_ECN_MODE_DISABLE; + } else { + error = EINVAL; + } + break; + case TCP_NOTIFY_ACKNOWLEDGEMENT: + error = sooptcopyin(sopt, &optval, + sizeof(optval), sizeof(optval)); + if (error) + break; + if (optval <= 0) { + error = EINVAL; + break; + } + if (tp->t_notify_ack_count >= TCP_MAX_NOTIFY_ACK) { + error = ETOOMANYREFS; + break; + } + + /* + * validate that the given marker id is not + * a duplicate to avoid ambiguity + */ + if ((error = tcp_notify_ack_id_valid(tp, so, + optval)) != 0) { + break; + } + error = tcp_add_notify_ack_marker(tp, optval); + break; + case SO_FLUSH: + if ((error = sooptcopyin(sopt, &optval, sizeof (optval), + sizeof (optval))) != 0) + break; + + error = inp_flush(inp, optval); + break; + + case SO_TRAFFIC_MGT_BACKGROUND: + if ((error = sooptcopyin(sopt, &optval, sizeof (optval), + sizeof (optval))) != 0) + break; + + if (optval) { + socket_set_traffic_mgt_flags_locked(so, + TRAFFIC_MGT_SO_BACKGROUND); + } else { + socket_clear_traffic_mgt_flags_locked(so, + TRAFFIC_MGT_SO_BACKGROUND); + } + break; + default: error = ENOPROTOOPT; break; @@ -1029,7 +2364,25 @@ tcp_ctloutput(so, sopt) optval = tp->t_maxseg; break; case TCP_KEEPALIVE: - optval = tp->t_keepidle / PR_SLOWHZ; + if (tp->t_keepidle > 0) + optval = tp->t_keepidle / TCP_RETRANSHZ; + else + optval = tcp_keepidle / TCP_RETRANSHZ; + break; + case TCP_KEEPINTVL: + if (tp->t_keepintvl > 0) + optval = tp->t_keepintvl / TCP_RETRANSHZ; + else + optval = tcp_keepintvl / TCP_RETRANSHZ; + break; + case TCP_KEEPCNT: + if (tp->t_keepcnt > 0) + optval = tp->t_keepcnt; + else + optval = tcp_keepcnt; + break; + case TCP_KEEPALIVE_OFFLOAD: + optval = !!(inp->inp_flags2 & INP2_KEEPALIVE_OFFLOAD); break; case TCP_NOOPT: optval = tp->t_flags & TF_NOOPT; @@ -1037,6 +2390,129 @@ tcp_ctloutput(so, sopt) case TCP_NOPUSH: optval = tp->t_flags & TF_NOPUSH; break; + case TCP_ENABLE_ECN: + optval = (tp->ecn_flags & TE_ECN_MODE_ENABLE) ? 1 : 0; + break; + case TCP_ECN_MODE: + if (tp->ecn_flags & TE_ECN_MODE_ENABLE) + optval = ECN_MODE_ENABLE; + else if (tp->ecn_flags & TE_ECN_MODE_DISABLE) + optval = ECN_MODE_DISABLE; + else + optval = ECN_MODE_DEFAULT; + break; + case TCP_CONNECTIONTIMEOUT: + optval = tp->t_keepinit / TCP_RETRANSHZ; + break; + case PERSIST_TIMEOUT: + optval = tp->t_persist_timeout / TCP_RETRANSHZ; + break; + case TCP_RXT_CONNDROPTIME: + optval = tp->t_rxt_conndroptime / TCP_RETRANSHZ; + break; + case TCP_RXT_FINDROP: + optval = tp->t_flagsext & TF_RXTFINDROP; + break; + case TCP_NOTIMEWAIT: + optval = (tp->t_flagsext & TF_NOTIMEWAIT) ? 1 : 0; + break; + case TCP_FASTOPEN: + if (tp->t_state != TCPS_LISTEN || + !(tcp_fastopen & TCP_FASTOPEN_SERVER)) { + error = ENOTSUP; + break; + } + optval = tfo_enabled(tp); + break; + case TCP_MEASURE_SND_BW: + optval = tp->t_flagsext & TF_MEASURESNDBW; + break; + case TCP_INFO: { + struct tcp_info ti; + + tcp_fill_info(tp, &ti); + error = sooptcopyout(sopt, &ti, sizeof(struct tcp_info)); + goto done; + /* NOT REACHED */ + } + case TCP_CONNECTION_INFO: { + struct tcp_connection_info tci; + tcp_connection_fill_info(tp, &tci); + error = sooptcopyout(sopt, &tci, + sizeof(struct tcp_connection_info)); + goto done; + } + case TCP_MEASURE_BW_BURST: { + struct tcp_measure_bw_burst out; + if ((tp->t_flagsext & TF_MEASURESNDBW) == 0 || + tp->t_bwmeas == NULL) { + error = EINVAL; + break; + } + out.min_burst_size = tp->t_bwmeas->bw_minsizepkts; + out.max_burst_size = tp->t_bwmeas->bw_maxsizepkts; + error = sooptcopyout(sopt, &out, sizeof(out)); + goto done; + } + case TCP_NOTSENT_LOWAT: + if ((so->so_flags & SOF_NOTSENT_LOWAT) != 0) { + optval = tp->t_notsent_lowat; + } else { + optval = 0; + } + break; + + case TCP_ENABLE_MSGS: + if (so->so_flags & SOF_ENABLE_MSGS) { + optval = 1; + } else { + optval = 0; + } + break; + case TCP_SENDMOREACKS: + if (tp->t_flagsext & TF_NOSTRETCHACK) + optval = 1; + else + optval = 0; + break; + case TCP_DISABLE_BLACKHOLE_DETECTION: + if (tp->t_flagsext & TF_NOBLACKHOLE_DETECTION) + optval = 1; + else + optval = 0; + break; + case TCP_PEER_PID: { + pid_t pid; + error = tcp_lookup_peer_pid_locked(so, &pid); + if (error == 0) + error = sooptcopyout(sopt, &pid, sizeof(pid)); + goto done; + } + case TCP_ADAPTIVE_READ_TIMEOUT: + optval = tp->t_adaptive_rtimo; + break; + case TCP_ADAPTIVE_WRITE_TIMEOUT: + optval = tp->t_adaptive_wtimo; + break; + case SO_TRAFFIC_MGT_BACKGROUND: + optval = (so->so_flags1 & + SOF1_TRAFFIC_MGT_SO_BACKGROUND) ? 1 : 0; + break; + case TCP_NOTIFY_ACKNOWLEDGEMENT: { + struct tcp_notify_ack_complete retid; + + if (sopt->sopt_valsize != sizeof (retid)) { + error = EINVAL; + break; + } + bzero(&retid, sizeof (retid)); + tcp_get_notify_ack_count(tp, &retid); + if (retid.notify_complete_count > 0) + tcp_get_notify_ack_ids(tp, &retid); + + error = sooptcopyout(sopt, &retid, sizeof (retid)); + goto done; + } default: error = ENOPROTOOPT; break; @@ -1045,6 +2521,7 @@ tcp_ctloutput(so, sopt) error = sooptcopyout(sopt, &optval, sizeof optval); break; } +done: return (error); } @@ -1053,33 +2530,67 @@ tcp_ctloutput(so, sopt) * sizes, respectively. These are obsolescent (this information should * be set by the route). */ -u_long tcp_sendspace = 1024*16; -SYSCTL_INT(_net_inet_tcp, TCPCTL_SENDSPACE, sendspace, CTLFLAG_RW, - &tcp_sendspace , 0, "Maximum outgoing TCP datagram size"); -u_long tcp_recvspace = 1024*16; -SYSCTL_INT(_net_inet_tcp, TCPCTL_RECVSPACE, recvspace, CTLFLAG_RW, - &tcp_recvspace , 0, "Maximum incoming TCP datagram size"); - -__private_extern__ int tcp_sockthreshold = 256; -SYSCTL_INT(_net_inet_tcp, OID_AUTO, sockthreshold, CTLFLAG_RW, - &tcp_sockthreshold , 0, "TCP Socket size increased if less than threshold"); - -#define TCP_INCREASED_SPACE 65535 /* Automatically increase tcp send/rcv space to this value */ +u_int32_t tcp_sendspace = 1448*256; +u_int32_t tcp_recvspace = 1448*384; + +/* During attach, the size of socket buffer allocated is limited to + * sb_max in sbreserve. Disallow setting the tcp send and recv space + * to be more than sb_max because that will cause tcp_attach to fail + * (see radar 5713060) + */ +static int +sysctl_tcp_sospace(struct sysctl_oid *oidp, __unused void *arg1, + __unused int arg2, struct sysctl_req *req) { + u_int32_t new_value = 0, *space_p = NULL; + int changed = 0, error = 0; + u_quad_t sb_effective_max = (sb_max / (MSIZE+MCLBYTES)) * MCLBYTES; + + switch (oidp->oid_number) { + case TCPCTL_SENDSPACE: + space_p = &tcp_sendspace; + break; + case TCPCTL_RECVSPACE: + space_p = &tcp_recvspace; + break; + default: + return EINVAL; + } + error = sysctl_io_number(req, *space_p, sizeof(u_int32_t), + &new_value, &changed); + if (changed) { + if (new_value > 0 && new_value <= sb_effective_max) { + *space_p = new_value; + } else { + error = ERANGE; + } + } + return error; +} + +SYSCTL_PROC(_net_inet_tcp, TCPCTL_SENDSPACE, sendspace, CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_LOCKED, + &tcp_sendspace , 0, &sysctl_tcp_sospace, "IU", "Maximum outgoing TCP datagram size"); +SYSCTL_PROC(_net_inet_tcp, TCPCTL_RECVSPACE, recvspace, CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_LOCKED, + &tcp_recvspace , 0, &sysctl_tcp_sospace, "IU", "Maximum incoming TCP datagram size"); + /* * Attach TCP protocol to socket, allocating * internet protocol control block, tcp control block, * bufer space, and entering LISTEN state if to accept connections. + * + * Returns: 0 Success + * in_pcballoc:ENOBUFS + * in_pcballoc:ENOMEM + * in_pcballoc:??? [IPSEC specific] + * soreserve:ENOBUFS */ static int -tcp_attach(so, p) - struct socket *so; - struct proc *p; +tcp_attach(struct socket *so, struct proc *p) { - register struct tcpcb *tp; + struct tcpcb *tp; struct inpcb *inp; int error; #if INET6 - int isipv6 = INP_CHECK_SOCKAF(so, AF_INET6) != NULL; + int isipv6 = SOCK_CHECK_DOM(so, PF_INET6) != 0; #endif error = in_pcballoc(so, &tcbinfo, p); @@ -1089,21 +2600,20 @@ tcp_attach(so, p) inp = sotoinpcb(so); if (so->so_snd.sb_hiwat == 0 || so->so_rcv.sb_hiwat == 0) { - /* - * The goal is to let clients have large send/rcv default windows (TCP_INCREASED_SPACE) - * while not hogging mbuf space for servers. This is done by watching a threshold - * of tcpcbs in use and bumping the default send and rcvspace only if under that threshold. - * The theory being that busy servers have a lot more active tcpcbs and don't want the potential - * memory penalty of having much larger sockbuffs. The sysctl allows to fine tune that threshold value. */ - - if (inp->inp_pcbinfo->ipi_count < tcp_sockthreshold) - error = soreserve(so, MAX(TCP_INCREASED_SPACE, tcp_sendspace), MAX(TCP_INCREASED_SPACE,tcp_recvspace)); - else - error = soreserve(so, tcp_sendspace, tcp_recvspace); + error = soreserve(so, tcp_sendspace, tcp_recvspace); if (error) return (error); } + if (so->so_snd.sb_preconn_hiwat == 0) { + soreserve_preconnect(so, 2048); + } + + if ((so->so_rcv.sb_flags & SB_USRSIZE) == 0) + so->so_rcv.sb_flags |= SB_AUTOSIZE; + if ((so->so_snd.sb_flags & SB_USRSIZE) == 0) + so->so_snd.sb_flags |= SB_AUTOSIZE; + #if INET6 if (isipv6) { inp->inp_vflag |= INP_IPV6; @@ -1113,7 +2623,7 @@ tcp_attach(so, p) #endif /* INET6 */ inp->inp_vflag |= INP_IPV4; tp = tcp_newtcpcb(inp); - if (tp == 0) { + if (tp == NULL) { int nofd = so->so_state & SS_NOFDREF; /* XXX */ so->so_state &= ~SS_NOFDREF; /* don't free the socket yet */ @@ -1126,6 +2636,8 @@ tcp_attach(so, p) so->so_state |= nofd; return (ENOBUFS); } + if (nstat_collect) + nstat_tcp_new_pcb(inp); tp->t_state = TCPS_CLOSED; return (0); } @@ -1139,8 +2651,7 @@ tcp_attach(so, p) * send segment to peer (with FIN). */ static struct tcpcb * -tcp_disconnect(tp) - register struct tcpcb *tp; +tcp_disconnect(struct tcpcb *tp) { struct socket *so = tp->t_inpcb->inp_socket; @@ -1152,6 +2663,12 @@ tcp_disconnect(tp) soisdisconnecting(so); sbflush(&so->so_rcv); tp = tcp_usrclosed(tp); +#if MPTCP + /* A reset has been sent but socket exists, do not send FIN */ + if ((so->so_flags & SOF_MP_SUBFLOW) && + (tp) && (tp->t_mpflags & TMPF_RESET)) + return (tp); +#endif if (tp) (void) tcp_output(tp); } @@ -1169,15 +2686,12 @@ tcp_disconnect(tp) * We can let the user exit from the close as soon as the FIN is acked. */ static struct tcpcb * -tcp_usrclosed(tp) - register struct tcpcb *tp; +tcp_usrclosed(struct tcpcb *tp) { - switch (tp->t_state) { case TCPS_CLOSED: case TCPS_LISTEN: - tp->t_state = TCPS_CLOSED; tp = tcp_close(tp); break; @@ -1187,10 +2701,18 @@ tcp_usrclosed(tp) break; case TCPS_ESTABLISHED: + DTRACE_TCP4(state__change, void, NULL, + struct inpcb *, tp->t_inpcb, + struct tcpcb *, tp, + int32_t, TCPS_FIN_WAIT_1); tp->t_state = TCPS_FIN_WAIT_1; break; case TCPS_CLOSE_WAIT: + DTRACE_TCP4(state__change, void, NULL, + struct inpcb *, tp->t_inpcb, + struct tcpcb *, tp, + int32_t, TCPS_LAST_ACK); tp->t_state = TCPS_LAST_ACK; break; } @@ -1198,8 +2720,66 @@ tcp_usrclosed(tp) soisdisconnected(tp->t_inpcb->inp_socket); /* To prevent the connection hanging in FIN_WAIT_2 forever. */ if (tp->t_state == TCPS_FIN_WAIT_2) - tp->t_timer[TCPT_2MSL] = tcp_maxidle; + tp->t_timer[TCPT_2MSL] = OFFSET_FROM_START(tp, + TCP_CONN_MAXIDLE(tp)); } return (tp); } +void +tcp_in_cksum_stats(u_int32_t len) +{ + tcpstat.tcps_rcv_swcsum++; + tcpstat.tcps_rcv_swcsum_bytes += len; +} + +void +tcp_out_cksum_stats(u_int32_t len) +{ + tcpstat.tcps_snd_swcsum++; + tcpstat.tcps_snd_swcsum_bytes += len; +} + +#if INET6 +void +tcp_in6_cksum_stats(u_int32_t len) +{ + tcpstat.tcps_rcv6_swcsum++; + tcpstat.tcps_rcv6_swcsum_bytes += len; +} + +void +tcp_out6_cksum_stats(u_int32_t len) +{ + tcpstat.tcps_snd6_swcsum++; + tcpstat.tcps_snd6_swcsum_bytes += len; +} + +/* + * When messages are enabled on a TCP socket, the message priority + * is sent as a control message. This function will extract it. + */ +int +tcp_get_msg_priority(struct mbuf *control, uint32_t *msgpri) +{ + struct cmsghdr *cm; + if (control == NULL) + return(EINVAL); + + for (cm = M_FIRST_CMSGHDR(control); cm; + cm = M_NXT_CMSGHDR(control, cm)) { + if (cm->cmsg_len < sizeof(struct cmsghdr) || + cm->cmsg_len > control->m_len) { + return (EINVAL); + } + if (cm->cmsg_level == SOL_SOCKET && + cm->cmsg_type == SCM_MSG_PRIORITY) { + *msgpri = *(unsigned int *)(void *)CMSG_DATA(cm); + break; + } + } + + VERIFY(*msgpri >= MSG_PRI_MIN && *msgpri <= MSG_PRI_MAX); + return (0); +} +#endif /* INET6 */