/*
- * Copyright (c) 2000 Apple Computer, Inc. All rights reserved.
+ * Copyright (c) 2000-2011 Apple Inc. All rights reserved.
*
- * @APPLE_LICENSE_HEADER_START@
- *
- * Copyright (c) 1999-2003 Apple Computer, Inc. All Rights Reserved.
+ * @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. Please obtain a copy of the License at
- * http://www.opensource.apple.com/apsl/ and read it before using this
- * file.
+ * 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
* Please see the License for the specific language governing rights and
* limitations under the License.
*
- * @APPLE_LICENSE_HEADER_END@
+ * @APPLE_OSREFERENCE_LICENSE_HEADER_END@
*/
/*
* Copyright (c) 1982, 1986, 1988, 1993
#include <net/if.h>
#include <net/route.h>
+#include <net/ntstat.h>
#include <netinet/in.h>
#include <netinet/in_systm.h>
#include <netinet6/ipsec.h>
#endif /*IPSEC*/
+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 *);
+
/*
* TCP protocol interface to socket abstraction.
*/
extern char *tcpstates[]; /* XXX ??? */
-static int tcp_attach __P((struct socket *, struct proc *));
-static int tcp_connect __P((struct tcpcb *, struct sockaddr *,
- struct proc *));
+static int tcp_attach(struct socket *, struct proc *);
+static int tcp_connect(struct tcpcb *, struct sockaddr *, struct proc *);
#if INET6
-static int tcp6_connect __P((struct tcpcb *, struct sockaddr *,
- struct proc *));
+static int tcp6_connect(struct tcpcb *, struct sockaddr *, struct proc *);
#endif /* INET6 */
static struct tcpcb *
- tcp_disconnect __P((struct tcpcb *));
+ tcp_disconnect(struct tcpcb *);
static struct tcpcb *
- tcp_usrclosed __P((struct tcpcb *));
+ tcp_usrclosed(struct tcpcb *);
+
+__private_extern__ int tcp_win_scale = 3;
+SYSCTL_INT(_net_inet_tcp, OID_AUTO, win_scale_factor, CTLFLAG_RW | CTLFLAG_LOCKED,
+ &tcp_win_scale, 0, "Window scaling factor");
+
+static u_int32_t tcps_in_sw_cksum;
+SYSCTL_UINT(_net_inet_tcp, OID_AUTO, in_sw_cksum, CTLFLAG_RD | CTLFLAG_LOCKED,
+ &tcps_in_sw_cksum, 0,
+ "Number of received packets checksummed in software");
+
+static u_int64_t tcps_in_sw_cksum_bytes;
+SYSCTL_QUAD(_net_inet_tcp, OID_AUTO, in_sw_cksum_bytes, CTLFLAG_RD | CTLFLAG_LOCKED,
+ &tcps_in_sw_cksum_bytes,
+ "Amount of received data checksummed in software");
+
+static u_int32_t tcps_out_sw_cksum;
+SYSCTL_UINT(_net_inet_tcp, OID_AUTO, out_sw_cksum, CTLFLAG_RD | CTLFLAG_LOCKED,
+ &tcps_out_sw_cksum, 0,
+ "Number of transmitted packets checksummed in software");
+
+static u_int64_t tcps_out_sw_cksum_bytes;
+SYSCTL_QUAD(_net_inet_tcp, OID_AUTO, out_sw_cksum_bytes, CTLFLAG_RD | CTLFLAG_LOCKED,
+ &tcps_out_sw_cksum_bytes,
+ "Amount of transmitted data checksummed in software");
#if TCPDEBUG
#define TCPDEBUG0 int ostate = 0
#define TCPDEBUG2(req)
#endif
+#if CONFIG_USESOCKTHRESHOLD
+__private_extern__ unsigned int tcp_sockthreshold = 64;
+#else
+__private_extern__ unsigned int tcp_sockthreshold = 0;
+#endif
+SYSCTL_INT(_net_inet_tcp, OID_AUTO, sockthreshold, CTLFLAG_RW | CTLFLAG_LOCKED,
+ &tcp_sockthreshold , 0, "TCP Socket size increased if less than threshold");
+
+
+SYSCTL_PROC(_net_inet_tcp, OID_AUTO, info, CTLFLAG_RW | CTLFLAG_LOCKED | CTLFLAG_ANYBODY,
+ 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 s = splnet();
int error;
struct inpcb *inp = sotoinpcb(so);
struct tcpcb *tp = 0;
error = EISCONN;
goto out;
}
-
+
error = tcp_attach(so, p);
if (error)
goto out;
tp = sototcpcb(so);
out:
TCPDEBUG2(PRU_ATTACH);
- splx(s);
return error;
}
static int
tcp_usr_detach(struct socket *so)
{
- int s = splnet();
int error = 0;
struct inpcb *inp = sotoinpcb(so);
struct tcpcb *tp;
TCPDEBUG0;
- if (inp == 0) {
- splx(s);
+ if (inp == 0 || (inp->inp_state == INPCB_STATE_DEAD)) {
return EINVAL; /* XXX */
}
+ 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;
TCPDEBUG1();
+
+ calculate_tcp_clock();
+
tp = tcp_disconnect(tp);
out:
TCPDEBUG2(PRU_DETACH);
- splx(s);
return error;
}
#define COMMON_START() TCPDEBUG0; \
do { \
- if (inp == 0) { \
- splx(s); \
+ if (inp == 0 || (inp->inp_state == INPCB_STATE_DEAD)) { \
return EINVAL; \
} \
tp = intotcpcb(inp); \
TCPDEBUG1(); \
+ calculate_tcp_clock(); \
} while(0)
-#define COMMON_END(req) out: TCPDEBUG2(req); splx(s); return error; goto out
+#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)
{
- int s = splnet();
int error = 0;
struct inpcb *inp = sotoinpcb(so);
struct tcpcb *tp;
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.
static int
tcp6_usr_bind(struct socket *so, struct sockaddr *nam, struct proc *p)
{
- int s = splnet();
int error = 0;
struct inpcb *inp = sotoinpcb(so);
struct tcpcb *tp;
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.
}
inp->inp_vflag &= ~INP_IPV4;
inp->inp_vflag |= INP_IPV6;
- if (ip6_mapped_addr_on && (inp->inp_flags & IN6P_IPV6_V6ONLY) == 0) {
+ if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0) {
if (IN6_IS_ADDR_UNSPECIFIED(&sin6p->sin6_addr))
inp->inp_vflag |= INP_IPV4;
else if (IN6_IS_ADDR_V4MAPPED(&sin6p->sin6_addr)) {
/*
* 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)
{
- int s = splnet();
int error = 0;
struct inpcb *inp = sotoinpcb(so);
struct tcpcb *tp;
static int
tcp6_usr_listen(struct socket *so, struct proc *p)
{
- int s = splnet();
int error = 0;
struct inpcb *inp = sotoinpcb(so);
struct tcpcb *tp;
COMMON_START();
if (inp->inp_lport == 0) {
inp->inp_vflag &= ~INP_IPV4;
- if (ip6_mapped_addr_on &&
- (inp->inp_flags & IN6P_IPV6_V6ONLY) == 0)
+ if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0)
inp->inp_vflag |= INP_IPV4;
error = in6_pcbbind(inp, (struct sockaddr *)0, p);
}
static int
tcp_usr_connect(struct socket *so, struct sockaddr *nam, struct proc *p)
{
- int s = splnet();
int error = 0;
struct inpcb *inp = sotoinpcb(so);
struct tcpcb *tp;
struct sockaddr_in *sinp;
- COMMON_START();
+ TCPDEBUG0;
+ if (inp == 0)
+ 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;
+ }
+ 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.
*/
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;
static int
tcp6_usr_connect(struct socket *so, struct sockaddr *nam, struct proc *p)
{
- int s = splnet();
int error = 0;
struct inpcb *inp = sotoinpcb(so);
struct tcpcb *tp;
COMMON_START();
+ if (nam->sa_family != 0 && nam->sa_family != AF_INET6) {
+ error = EAFNOSUPPORT;
+ goto out;
+ }
+
/*
* Must disallow TCP ``connections'' to multicast addresses.
*/
if (IN6_IS_ADDR_V4MAPPED(&sin6p->sin6_addr)) {
struct sockaddr_in sin;
- if (!ip6_mapped_addr_on ||
- (inp->inp_flags & IN6P_IPV6_V6ONLY))
- return(EINVAL);
+ if ((inp->inp_flags & IN6P_IPV6_V6ONLY) != 0)
+ return (EINVAL);
in6_sin6_2_sin(&sin, sin6p);
inp->inp_vflag |= INP_IPV4;
static int
tcp_usr_disconnect(struct socket *so)
{
- int s = splnet();
int error = 0;
struct inpcb *inp = sotoinpcb(so);
struct tcpcb *tp;
-
+
+ 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)
static int
tcp_usr_accept(struct socket *so, struct sockaddr **nam)
{
- int s = splnet();
int error = 0;
struct inpcb *inp = sotoinpcb(so);
struct tcpcb *tp = NULL;
TCPDEBUG0;
+ in_setpeeraddr(so, nam);
+
if (so->so_state & SS_ISDISCONNECTED) {
error = ECONNABORTED;
goto out;
}
- if (inp == 0) {
- splx(s);
+ if (inp == 0 || (inp->inp_state == INPCB_STATE_DEAD)) {
return (EINVAL);
}
tp = intotcpcb(inp);
TCPDEBUG1();
- in_setpeeraddr(so, nam);
+
+ calculate_tcp_clock();
+
COMMON_END(PRU_ACCEPT);
}
static int
tcp6_usr_accept(struct socket *so, struct sockaddr **nam)
{
- int s = splnet();
int error = 0;
struct inpcb *inp = sotoinpcb(so);
struct tcpcb *tp = NULL;
error = ECONNABORTED;
goto out;
}
- if (inp == 0) {
- splx(s);
+ if (inp == 0 || (inp->inp_state == INPCB_STATE_DEAD)) {
return (EINVAL);
}
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)
{
- int s = splnet();
int error = 0;
struct inpcb *inp = sotoinpcb(so);
struct tcpcb *tp;
* 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 s = splnet();
int error = 0;
struct inpcb *inp = sotoinpcb(so);
struct tcpcb *tp;
* 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)
{
- int s = splnet();
int error = 0;
struct inpcb *inp = sotoinpcb(so);
struct tcpcb *tp;
#endif
TCPDEBUG0;
- if (inp == NULL) {
+ if (inp == NULL || inp->inp_state == INPCB_STATE_DEAD) {
/*
* OOPS! we lost a race, the TCP session got reset after
* we checked SS_CANTSENDMORE, eg: while doing uiomove or a
#endif /* INET6 */
tp = intotcpcb(inp);
TCPDEBUG1();
+
+ calculate_tcp_clock();
+
if (control) {
/* TCP doesn't do control messages (rights, creds, etc) */
if (control->m_len) {
m_freem(control); /* empty control, just free it */
}
if(!(flags & PRUS_OOB)) {
- 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,
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) {
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;
* 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,
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;
static int
tcp_usr_abort(struct socket *so)
{
- int s = splnet();
int error = 0;
struct inpcb *inp = sotoinpcb(so);
struct tcpcb *tp;
if (tp == 0)
goto out;
tp = tcp_drop(tp, ECONNABORTED);
+ 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)
{
- int s = splnet();
int error = 0;
struct inpcb *inp = sotoinpcb(so);
struct tcpcb *tp;
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, sopoll
+ in_setsockaddr, sosend, soreceive, pru_sopoll_notsupp
};
#if INET6
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, sopoll
+ in6_mapped_sockaddr, sosend, soreceive, pru_sopoll_notsupp
};
#endif /* INET6 */
* 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)
struct socket *so = inp->inp_socket;
struct tcpcb *otp;
struct sockaddr_in *sin = (struct sockaddr_in *)nam;
- struct sockaddr_in *ifaddr;
+ struct sockaddr_in ifaddr;
struct rmxp_tao *taop;
struct rmxp_tao tao_noncached;
int error;
+ unsigned int outif = 0;
if (inp->inp_lport == 0) {
error = in_pcbbind(inp, (struct sockaddr *)0, 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, &ifaddr, &outif);
if (error)
return error;
+
+ 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,
+ : ifaddr.sin_addr,
inp->inp_lport, 0, NULL);
+
+ tcp_lock(inp->inp_socket, 0, 0);
if (oinp) {
+ if (oinp != inp) /* 4143933: avoid deadlock if inp == oinp */
+ tcp_lock(oinp->inp_socket, 1, 0);
+ if (in_pcb_checkstate(oinp, WNT_RELEASE, 1) == WNT_STOPUSING) {
+ if (oinp != inp)
+ tcp_unlock(oinp->inp_socket, 1, 0);
+ goto skip_oinp;
+ }
+
if (oinp != inp && (otp = intotcpcb(oinp)) != NULL &&
otp->t_state == TCPS_TIME_WAIT &&
- otp->t_starttime < tcp_msl &&
+ ((int)(tcp_now - otp->t_starttime)) < tcp_msl &&
(otp->t_flags & TF_RCVD_CC))
otp = tcp_close(otp);
- else
+ else {
+ printf("tcp_connect: inp=%p err=EADDRINUSE\n", inp);
+ if (oinp != inp)
+ tcp_unlock(oinp->inp_socket, 1, 0);
return EADDRINUSE;
+ }
+ if (oinp != inp)
+ tcp_unlock(oinp->inp_socket, 1, 0);
}
- if ((inp->inp_laddr.s_addr == INADDR_ANY ? ifaddr->sin_addr.s_addr :
+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 (inp->inp_laddr.s_addr == INADDR_ANY)
- inp->inp_laddr = ifaddr->sin_addr;
+ if (!lck_rw_try_lock_exclusive(inp->inp_pcbinfo->mtx)) {
+ /*lock inversion issue, mostly with udp multicast packets */
+ socket_unlock(inp->inp_socket, 0);
+ lck_rw_lock_exclusive(inp->inp_pcbinfo->mtx);
+ socket_lock(inp->inp_socket, 0);
+ }
+ if (inp->inp_laddr.s_addr == INADDR_ANY) {
+ inp->inp_laddr = ifaddr.sin_addr;
+ inp->inp_last_outif = outif;
+ }
inp->inp_faddr = sin->sin_addr;
inp->inp_fport = sin->sin_port;
in_pcbrehash(inp);
+ lck_rw_done(inp->inp_pcbinfo->mtx);
- /* Compute window scaling to request. */
+ /* Compute window scaling to requesti according to sb_hiwat
+ * or leave us some room to increase potentially increase the window size depending
+ * on the default win scale
+ */
while (tp->request_r_scale < TCP_MAX_WINSHIFT &&
- (TCP_MAXWIN << tp->request_r_scale) < so->so_rcv.sb_hiwat)
+ (TCP_MAXWIN << tp->request_r_scale) < so->so_rcv.sb_hiwat)
tp->request_r_scale++;
+ /*
+ * Inflate window size only if no setsockopt was performed on the recv sockbuf and
+ * if we're not over our number of active pcbs.
+ */
+
+ if (((so->so_rcv.sb_flags & SB_USRSIZE) == 0) && (inp->inp_pcbinfo->ipi_count < tcp_sockthreshold)) {
+ tp->request_r_scale = max(tcp_win_scale, tp->request_r_scale);
+ so->so_rcv.sb_hiwat = min(TCP_MAXWIN << tp->request_r_scale, (sb_max / (MSIZE+MCLBYTES)) * MCLBYTES);
+ }
+
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,
+ tp->t_keepinit ? tp->t_keepinit : tcp_keepinit);
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
struct socket *so = inp->inp_socket;
struct tcpcb *otp;
struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)nam;
- struct in6_addr *addr6;
+ struct in6_addr addr6;
struct rmxp_tao *taop;
struct rmxp_tao tao_noncached;
int error;
+ unsigned int outif = 0;
if (inp->inp_lport == 0) {
error = in6_pcbbind(inp, (struct sockaddr *)0, p);
* earlier incarnation of this same connection still in
* TIME_WAIT state, creating an ADDRINUSE error.
*/
- error = in6_pcbladdr(inp, nam, &addr6);
+ error = in6_pcbladdr(inp, nam, &addr6, &outif);
if (error)
return error;
+ tcp_unlock(inp->inp_socket, 0, 0);
oinp = in6_pcblookup_hash(inp->inp_pcbinfo,
&sin6->sin6_addr, sin6->sin6_port,
IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)
- ? addr6
+ ? &addr6
: &inp->in6p_laddr,
inp->inp_lport, 0, NULL);
+ tcp_lock(inp->inp_socket, 0, 0);
if (oinp) {
if (oinp != inp && (otp = intotcpcb(oinp)) != NULL &&
otp->t_state == TCPS_TIME_WAIT &&
- otp->t_starttime < tcp_msl &&
+ ((int)(tcp_now - otp->t_starttime)) < tcp_msl &&
(otp->t_flags & TF_RCVD_CC))
otp = tcp_close(otp);
else
return EADDRINUSE;
}
- if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr))
- inp->in6p_laddr = *addr6;
+ if (!lck_rw_try_lock_exclusive(inp->inp_pcbinfo->mtx)) {
+ /*lock inversion issue, mostly with udp multicast packets */
+ socket_unlock(inp->inp_socket, 0);
+ lck_rw_lock_exclusive(inp->inp_pcbinfo->mtx);
+ socket_lock(inp->inp_socket, 0);
+ }
+ if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)) {
+ inp->in6p_laddr = addr6;
+ inp->in6p_last_outif = outif;
+ }
inp->in6p_faddr = sin6->sin6_addr;
inp->inp_fport = sin6->sin6_port;
- if ((sin6->sin6_flowinfo & IPV6_FLOWINFO_MASK) != NULL)
+ if ((sin6->sin6_flowinfo & IPV6_FLOWINFO_MASK) != 0)
inp->in6p_flowinfo = sin6->sin6_flowinfo;
in_pcbrehash(inp);
+ lck_rw_done(inp->inp_pcbinfo->mtx);
/* Compute window scaling to request. */
while (tp->request_r_scale < TCP_MAX_WINSHIFT &&
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,
+ tp->t_keepinit ? tp->t_keepinit : tcp_keepinit);
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
}
#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)
+{
+ bzero(ti, sizeof(*ti));
+
+ ti->tcpi_state = tp->t_state;
+
+ if (tp->t_state > TCPS_LISTEN) {
+ if ((tp->t_flags & TF_REQ_TSTMP) && (tp->t_flags & TF_RCVD_TSTMP))
+ ti->tcpi_options |= TCPI_OPT_TIMESTAMPS;
+ if (tp->t_flags & TF_SACK_PERMIT)
+ ti->tcpi_options |= TCPI_OPT_SACK;
+ if ((tp->t_flags & TF_REQ_SCALE) && (tp->t_flags & TF_RCVD_SCALE)) {
+ ti->tcpi_options |= TCPI_OPT_WSCALE;
+ ti->tcpi_snd_wscale = tp->snd_scale;
+ ti->tcpi_rcv_wscale = tp->rcv_scale;
+ }
+
+ ti->tcpi_snd_mss = tp->t_maxseg;
+ ti->tcpi_rcv_mss = tp->t_maxseg;
+
+ ti->tcpi_snd_ssthresh = tp->snd_ssthresh;
+ ti->tcpi_snd_cwnd = tp->snd_cwnd;
+
+ ti->tcpi_rcv_space = tp->rcv_wnd;
+
+ ti->tcpi_snd_wnd = tp->snd_wnd;
+ ti->tcpi_snd_bwnd = tp->snd_bwnd;
+ ti->tcpi_snd_nxt = tp->snd_nxt;
+ ti->tcpi_rcv_nxt = tp->rcv_nxt;
+
+ ti->tcpi_last_outif = tp->t_inpcb->inp_last_outif;
+ }
+}
+
+__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
+ 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;
+}
+
+
+__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;
+
+ 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;
+}
+
/*
* The new sockopt interface makes it possible for us to block in the
* copyin/out step (if we take a page fault). Taking a page fault at
struct socket *so;
struct sockopt *sopt;
{
- int error, opt, optval, s;
+ int error, opt, optval;
struct inpcb *inp;
struct tcpcb *tp;
error = 0;
- s = splnet(); /* XXX */
inp = sotoinpcb(so);
if (inp == NULL) {
- splx(s);
return (ECONNRESET);
}
if (sopt->sopt_level != IPPROTO_TCP) {
else
#endif /* INET6 */
error = ip_ctloutput(so, sopt);
- splx(s);
return (error);
}
tp = intotcpcb(inp);
if (tp == NULL) {
- splx(s);
return (ECONNRESET);
}
+ calculate_tcp_clock();
+
switch (sopt->sopt_dir) {
case SOPT_SET:
switch (sopt->sopt_name) {
else
tp->t_flags &= ~opt;
break;
-
+ case TCP_RXT_FINDROP:
+ error = sooptcopyin(sopt, &optval, sizeof optval,
+ sizeof optval);
+ if (error)
+ break;
+ opt = TF_RXTFINDROP;
+ if (optval)
+ tp->t_flagsext |= opt;
+ else
+ tp->t_flagsext &= ~opt;
+ break;
case TCP_MAXSEG:
error = sooptcopyin(sopt, &optval, sizeof optval,
sizeof optval);
if (error)
break;
- if (optval > 0 && optval <= tp->t_maxseg)
+ if (optval > 0 && optval <= tp->t_maxseg &&
+ optval + 40 >= tcp_minmss)
tp->t_maxseg = optval;
else
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 * TCP_RETRANSHZ;
+ tp->t_timer[TCPT_KEEP] = OFFSET_FROM_START(tp,
+ TCP_KEEPIDLE(tp)); /* reset the timer to new value */
+ tcp_check_timer_state(tp);
+ }
+ break;
+
+ case TCP_CONNECTIONTIMEOUT:
+ error = sooptcopyin(sopt, &optval, sizeof optval,
+ sizeof optval);
+ if (error)
+ break;
+ if (optval < 0)
+ error = EINVAL;
+ else
+ tp->t_keepinit = optval * TCP_RETRANSHZ;
+ 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->rxt_conndroptime = optval * TCP_RETRANSHZ;
+ break;
default:
error = ENOPROTOOPT;
break;
case TCP_MAXSEG:
optval = tp->t_maxseg;
break;
+ case TCP_KEEPALIVE:
+ optval = tp->t_keepidle / TCP_RETRANSHZ;
+ break;
case TCP_NOOPT:
optval = tp->t_flags & TF_NOOPT;
break;
case TCP_NOPUSH:
optval = tp->t_flags & TF_NOPUSH;
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->rxt_conndroptime / TCP_RETRANSHZ;
+ break;
+ case TCP_RXT_FINDROP:
+ optval = tp->t_flagsext & TF_RXTFINDROP;
+ break;
+ case TCP_INFO: {
+ struct tcp_info ti;
+
+ tcp_fill_info(tp, &ti);
+ error = sooptcopyout(sopt, &ti, sizeof(struct tcp_info));
+ goto done;
+ }
default:
error = ENOPROTOOPT;
break;
error = sooptcopyout(sopt, &optval, sizeof optval);
break;
}
- splx(s);
+done:
return (error);
}
* 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");
+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)
register struct tcpcb *tp;
struct inpcb *inp;
int error;
+ u_long sb_effective_max;
#if INET6
- int isipv6 = INP_CHECK_SOCKAF(so, AF_INET6) != NULL;
+ int isipv6 = INP_CHECK_SOCKAF(so, AF_INET6) != 0;
#endif
- if (so->so_snd.sb_hiwat == 0 || so->so_rcv.sb_hiwat == 0) {
- error = soreserve(so, tcp_sendspace, tcp_recvspace);
- if (error)
- return (error);
- }
error = in_pcballoc(so, &tcbinfo, p);
if (error)
return (error);
+
inp = sotoinpcb(so);
+
+ if (so->so_snd.sb_hiwat == 0 || so->so_rcv.sb_hiwat == 0) {
+ /*
+ * The goal is to let clients machines use large send/rcv default windows to compensate for link
+ * latency and make sure the receiver is not constraining the sender window.
+ * But we doon't want to have a few connections use all our mbuf space for servers.
+ * This is done by watching a threshold of tcpcbs in use and bumping the default send and rcvspace
+ * only if that threshold isn't reached.
+ * We're also advertising a much bigger window size (tuneable by sysctl) in correlation with * the max socket buffer size if
+ * we consider that we have enough ressources for it. This window will be adjusted depending on the
+ * global socket layer buffer use with the use of tcp_sbpace
+ */
+
+ if (inp->inp_pcbinfo->ipi_count < tcp_sockthreshold) {
+ sb_effective_max = (sb_max / (MSIZE+MCLBYTES)) * MCLBYTES;
+ error = soreserve(so, max(min((TCP_MAXWIN << tcp_win_scale)/4, sb_effective_max), tcp_sendspace),
+ max(min((TCP_MAXWIN << tcp_win_scale)/2, sb_effective_max), tcp_recvspace));
+ }
+ else
+ error = soreserve(so, tcp_sendspace, tcp_recvspace);
+ if (error)
+ return (error);
+ }
+
#if INET6
if (isipv6) {
inp->inp_vflag |= INP_IPV6;
so->so_state |= nofd;
return (ENOBUFS);
}
+ if (nstat_collect) {
+ nstat_tcp_new_pcb(inp);
+ }
tp->t_state = TCPS_CLOSED;
return (0);
}
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_maxidle);
}
return (tp);
}
+void
+tcp_in_cksum_stats(u_int32_t len)
+{
+ tcps_in_sw_cksum++;
+ tcps_in_sw_cksum_bytes += len;
+}
+
+void
+tcp_out_cksum_stats(u_int32_t len)
+{
+ tcps_out_sw_cksum++;
+ tcps_out_sw_cksum_bytes += len;
+}
projects / apple / xnu.git / blobdiff