X-Git-Url: https://git.saurik.com/apple/xnu.git/blobdiff_plain/43866e378188c25dd1e2208016ab3cbeb086ae6c..13fec9890cf095cc781fdf7b8917cb03bf32dd4c:/bsd/netinet/tcp_usrreq.c

diff --git a/bsd/netinet/tcp_usrreq.c b/bsd/netinet/tcp_usrreq.c
index cc9202e54..2a5ab6988 100644
--- a/bsd/netinet/tcp_usrreq.c
+++ b/bsd/netinet/tcp_usrreq.c
@@ -3,22 +3,19 @@
  *
  * @APPLE_LICENSE_HEADER_START@
  * 
- * Copyright (c) 1999-2003 Apple Computer, Inc.  All Rights Reserved.
+ * 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 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.
- * 
- * The Original Code and all software distributed under the License are
- * distributed on an 'AS IS' basis, WITHOUT WARRANTY OF ANY KIND, EITHER
+ * This 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, QUIET ENJOYMENT OR NON-INFRINGEMENT.
- * Please see the License for the specific language governing rights and
- * limitations under the License.
+ * 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@
  */
@@ -107,17 +104,15 @@
  */
 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 *);
 
 #if TCPDEBUG
 #define	TCPDEBUG0	int ostate = 0
@@ -137,7 +132,6 @@ static struct tcpcb *
 static int
 tcp_usr_attach(struct socket *so, int proto, struct proc *p)
 {
-	int s = splnet();
 	int error;
 	struct inpcb *inp = sotoinpcb(so);
 	struct tcpcb *tp = 0;
@@ -158,7 +152,6 @@ tcp_usr_attach(struct socket *so, int proto, struct proc *p)
 	tp = sototcpcb(so);
 out:
 	TCPDEBUG2(PRU_ATTACH);
-	splx(s);
 	return error;
 }
 
@@ -172,16 +165,17 @@ out:
 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 */
 	}
+#if 1
+	lck_mtx_assert(((struct inpcb *)so->so_pcb)->inpcb_mtx, LCK_MTX_ASSERT_OWNED);
+#endif
 	tp = intotcpcb(inp);
 	/* In case we got disconnected from the peer */
         if (tp == 0) 
@@ -190,21 +184,19 @@ tcp_usr_detach(struct socket *so)
 	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(); \
 		     } 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
 
 
 /*
@@ -213,7 +205,6 @@ out:
 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;
@@ -242,7 +233,6 @@ tcp_usr_bind(struct socket *so, struct sockaddr *nam, struct proc *p)
 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;
@@ -262,7 +252,7 @@ tcp6_usr_bind(struct socket *so, struct sockaddr *nam, struct proc *p)
 	}
 	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)) {
@@ -288,7 +278,6 @@ tcp6_usr_bind(struct socket *so, struct sockaddr *nam, struct proc *p)
 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;
@@ -305,7 +294,6 @@ tcp_usr_listen(struct socket *so, struct proc *p)
 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;
@@ -313,8 +301,7 @@ tcp6_usr_listen(struct socket *so, struct proc *p)
 	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);
 	}
@@ -334,7 +321,6 @@ tcp6_usr_listen(struct socket *so, struct proc *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;
@@ -366,7 +352,6 @@ tcp_usr_connect(struct socket *so, struct sockaddr *nam, struct proc *p)
 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;
@@ -387,9 +372,8 @@ tcp6_usr_connect(struct socket *so, struct sockaddr *nam, struct proc *p)
 	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;
@@ -424,11 +408,13 @@ tcp6_usr_connect(struct socket *so, struct sockaddr *nam, struct proc *p)
 static int
 tcp_usr_disconnect(struct socket *so)
 {
-	int s = splnet();
 	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
 	COMMON_START();
         /* In case we got disconnected from the peer */
         if (tp == 0)
@@ -445,7 +431,6 @@ tcp_usr_disconnect(struct socket *so)
 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;
@@ -455,8 +440,7 @@ tcp_usr_accept(struct socket *so, struct sockaddr **nam)
 		error = ECONNABORTED;
 		goto out;
 	}
-	if (inp == 0) {
-		splx(s);
+	if (inp == 0 || (inp->inp_state == INPCB_STATE_DEAD)) {
 		return (EINVAL);
 	}
 	tp = intotcpcb(inp);
@@ -469,7 +453,6 @@ tcp_usr_accept(struct socket *so, struct sockaddr **nam)
 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;
@@ -479,8 +462,7 @@ tcp6_usr_accept(struct socket *so, struct sockaddr **nam)
 		error = ECONNABORTED;
 		goto out;
 	}
-	if (inp == 0) {
-		splx(s);
+	if (inp == 0 || (inp->inp_state == INPCB_STATE_DEAD)) {
 		return (EINVAL);
 	}
 	tp = intotcpcb(inp);
@@ -495,7 +477,6 @@ tcp6_usr_accept(struct socket *so, struct sockaddr **nam)
 static int
 tcp_usr_shutdown(struct socket *so)
 {
-	int s = splnet();
 	int error = 0;
 	struct inpcb *inp = sotoinpcb(so);
 	struct tcpcb *tp;
@@ -517,7 +498,6 @@ tcp_usr_shutdown(struct socket *so)
 static int
 tcp_usr_rcvd(struct socket *so, int flags)
 {
-	int s = splnet();
 	int error = 0;
 	struct inpcb *inp = sotoinpcb(so);
 	struct tcpcb *tp;
@@ -541,7 +521,6 @@ 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;
@@ -550,7 +529,7 @@ tcp_usr_send(struct socket *so, int flags, struct mbuf *m,
 #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
@@ -665,7 +644,6 @@ tcp_usr_send(struct socket *so, int flags, struct mbuf *m,
 static int
 tcp_usr_abort(struct socket *so)
 {
-	int s = splnet();
 	int error = 0;
 	struct inpcb *inp = sotoinpcb(so);
 	struct tcpcb *tp;
@@ -675,6 +653,7 @@ tcp_usr_abort(struct socket *so)
         if (tp == 0)
             goto out;
 	tp = tcp_drop(tp, ECONNABORTED);
+	so->so_usecount--;
 	COMMON_END(PRU_ABORT);
 }
 
@@ -684,7 +663,6 @@ tcp_usr_abort(struct socket *so)
 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;
@@ -714,7 +692,7 @@ struct pr_usrreqs tcp_usrreqs = {
 	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
@@ -723,7 +701,7 @@ struct pr_usrreqs tcp6_usrreqs = {
 	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 */
 
@@ -766,29 +744,55 @@ tcp_connect(tp, nam, p)
 	error = in_pcbladdr(inp, nam, &ifaddr);
 	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,
 	    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 &&
 		    (otp->t_flags & TF_RCVD_CC))
 			otp = tcp_close(otp);
-		else
+		else {
+			printf("tcp_connect: inp=%x 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);
 	}
+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)) {
+		/*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_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.  */
 	while (tp->request_r_scale < TCP_MAX_WINSHIFT &&
@@ -834,7 +838,7 @@ tcp6_connect(tp, nam, p)
 	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;
@@ -853,12 +857,14 @@ tcp6_connect(tp, nam, p)
 	error = in6_pcbladdr(inp, nam, &addr6);
 	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 &&
@@ -868,13 +874,20 @@ tcp6_connect(tp, nam, p)
 		else
 			return EADDRINUSE;
 	}
+	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_laddr = addr6;
 	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;
 	in_pcbrehash(inp);
+	lck_rw_done(inp->inp_pcbinfo->mtx);
 
 	/* Compute window scaling to request.  */
 	while (tp->request_r_scale < TCP_MAX_WINSHIFT &&
@@ -922,15 +935,13 @@ tcp_ctloutput(so, sopt)
 	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) {
@@ -940,12 +951,10 @@ tcp_ctloutput(so, sopt)
 		else
 #endif /* INET6 */
 		error = ip_ctloutput(so, sopt);
-		splx(s);
 		return (error);
 	}
 	tp = intotcpcb(inp);
         if (tp == NULL) {
-                splx(s);
                 return (ECONNRESET);
         }
 
@@ -987,12 +996,24 @@ tcp_ctloutput(so, sopt)
 			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 * PR_SLOWHZ;
+                        break;
+		
 		default:
 			error = ENOPROTOOPT;
 			break;
@@ -1007,6 +1028,9 @@ tcp_ctloutput(so, sopt)
 		case TCP_MAXSEG:
 			optval = tp->t_maxseg;
 			break;
+		case TCP_KEEPALIVE:
+			optval = tp->t_keepidle / PR_SLOWHZ;
+			break;
 		case TCP_NOOPT:
 			optval = tp->t_flags & TF_NOOPT;
 			break;
@@ -1021,7 +1045,6 @@ tcp_ctloutput(so, sopt)
 			error = sooptcopyout(sopt, &optval, sizeof optval);
 		break;
 	}
-	splx(s);
 	return (error);
 }
 
@@ -1037,6 +1060,11 @@ 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 */
 /*
  * Attach TCP protocol to socket, allocating
  * internet protocol control block, tcp control block,
@@ -1054,15 +1082,28 @@ tcp_attach(so, p)
 	int isipv6 = INP_CHECK_SOCKAF(so, AF_INET6) != NULL;
 #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 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);
+		if (error)
+			return (error);
+	}
+
 #if INET6
 	if (isipv6) {
 		inp->inp_vflag |= INP_IPV6;