xnu-2422.100.13.tar.gz

[apple/xnu.git] / bsd / netinet / tcp_subr.c

diff --git a/bsd/netinet/tcp_subr.c b/bsd/netinet/tcp_subr.c
index 8cf6584826e32a5ba5fe92b27257aea281b4a90a..c9a7a6bb70af64dc57d3f7d60da5af0e13949f66 100644 (file)
--- a/bsd/netinet/tcp_subr.c
+++ b/bsd/netinet/tcp_subr.c
@@ -1,5 +1,5 @@
 /*
 /*

- * Copyright (c) 2000-2011 Apple Inc. All rights reserved.
+ * Copyright (c) 2000-2013 Apple Inc. All rights reserved.

  *
  * @APPLE_OSREFERENCE_LICENSE_HEADER_START@
  * 
  *
  * @APPLE_OSREFERENCE_LICENSE_HEADER_START@
  * 


@@ -86,6 +86,8 @@
 #include <kern/locks.h>
 #include <kern/zalloc.h>
 
 #include <kern/locks.h>
 #include <kern/zalloc.h>
 

+#include <dev/random/randomdev.h>
+

 #include <net/route.h>
 #include <net/if.h>
 
 #include <net/route.h>
 #include <net/if.h>
 


@@ -142,6 +144,8 @@
 #include <sys/kdebug.h>
 #include <mach/sdt.h>
 
 #include <sys/kdebug.h>
 #include <mach/sdt.h>
 

+#include <netinet/lro_ext.h>
+

 #define DBG_FNC_TCP_CLOSE      NETDBG_CODE(DBG_NETTCP, ((5 << 8) | 2))
 
 extern int tcp_lq_overflow;
 #define DBG_FNC_TCP_CLOSE      NETDBG_CODE(DBG_NETTCP, ((5 << 8) | 2))
 
 extern int tcp_lq_overflow;


@@ -150,6 +154,8 @@ extern int tcp_lq_overflow;
 #if IPSEC
 extern int ipsec_bypass;
 #endif
 #if IPSEC
 extern int ipsec_bypass;
 #endif

+extern struct tcptimerlist tcp_timer_list;
+extern struct tcptailq tcp_tw_tailq;

 
 int    tcp_mssdflt = TCP_MSS;
 SYSCTL_INT(_net_inet_tcp, TCPCTL_MSSDFLT, mssdflt, CTLFLAG_RW | CTLFLAG_LOCKED,
 
 int    tcp_mssdflt = TCP_MSS;
 SYSCTL_INT(_net_inet_tcp, TCPCTL_MSSDFLT, mssdflt, CTLFLAG_RW | CTLFLAG_LOCKED,


@@ -162,6 +168,8 @@ SYSCTL_INT(_net_inet_tcp, TCPCTL_V6MSSDFLT, v6mssdflt,
        "Default TCP Maximum Segment Size for IPv6");
 #endif
 
        "Default TCP Maximum Segment Size for IPv6");
 #endif
 

+extern int tcp_do_autorcvbuf;
+

 /*
  * Minimum MSS we accept and use. This prevents DoS attacks where
  * we are forced to a ridiculous low MSS like 20 and send hundreds
 /*
  * Minimum MSS we accept and use. This prevents DoS attacks where
  * we are forced to a ridiculous low MSS like 20 and send hundreds


@@ -174,23 +182,6 @@ int        tcp_minmss = TCP_MINMSS;
 SYSCTL_INT(_net_inet_tcp, OID_AUTO, minmss, CTLFLAG_RW | CTLFLAG_LOCKED,
     &tcp_minmss , 0, "Minmum TCP Maximum Segment Size");
 
 SYSCTL_INT(_net_inet_tcp, OID_AUTO, minmss, CTLFLAG_RW | CTLFLAG_LOCKED,
     &tcp_minmss , 0, "Minmum TCP Maximum Segment Size");
 

-/*
- * Number of TCP segments per second we accept from remote host
- * before we start to calculate average segment size. If average
- * segment size drops below the minimum TCP MSS we assume a DoS
- * attack and reset+drop the connection. Care has to be taken not to
- * set this value too small to not kill interactive type connections
- * (telnet, SSH) which send many small packets.
- */
-#ifdef FIX_WORKAROUND_FOR_3894301
-__private_extern__ int     tcp_minmssoverload = TCP_MINMSSOVERLOAD;
-#else
-__private_extern__ int     tcp_minmssoverload = 0;
-#endif
-SYSCTL_INT(_net_inet_tcp, OID_AUTO, minmssoverload, CTLFLAG_RW | CTLFLAG_LOCKED,
-    &tcp_minmssoverload , 0, "Number of TCP Segments per Second allowed to"
-    "be under the MINMSS Size");
-

 static int     tcp_do_rfc1323 = 1;
 SYSCTL_INT(_net_inet_tcp, TCPCTL_DO_RFC1323, rfc1323, CTLFLAG_RW | CTLFLAG_LOCKED,
     &tcp_do_rfc1323 , 0, "Enable rfc1323 (high performance TCP) extensions");
 static int     tcp_do_rfc1323 = 1;
 SYSCTL_INT(_net_inet_tcp, TCPCTL_DO_RFC1323, rfc1323, CTLFLAG_RW | CTLFLAG_LOCKED,
     &tcp_do_rfc1323 , 0, "Enable rfc1323 (high performance TCP) extensions");


@@ -207,6 +198,10 @@ SYSCTL_INT(_net_inet_tcp, OID_AUTO, do_tcpdrain, CTLFLAG_RW | CTLFLAG_LOCKED, &d
 SYSCTL_INT(_net_inet_tcp, OID_AUTO, pcbcount, CTLFLAG_RD | CTLFLAG_LOCKED, 
     &tcbinfo.ipi_count, 0, "Number of active PCBs");
 
 SYSCTL_INT(_net_inet_tcp, OID_AUTO, pcbcount, CTLFLAG_RD | CTLFLAG_LOCKED, 
     &tcbinfo.ipi_count, 0, "Number of active PCBs");
 

+SYSCTL_INT(_net_inet_tcp, OID_AUTO, tw_pcbcount, 
+    CTLFLAG_RD | CTLFLAG_LOCKED,
+    &tcbinfo.ipi_twcount, 0, "Number of pcbs in time-wait state");
+

 static int     icmp_may_rst = 1;
 SYSCTL_INT(_net_inet_tcp, OID_AUTO, icmp_may_rst, CTLFLAG_RW | CTLFLAG_LOCKED, &icmp_may_rst, 0, 
     "Certain ICMP unreachable messages may abort connections in SYN_SENT");
 static int     icmp_may_rst = 1;
 SYSCTL_INT(_net_inet_tcp, OID_AUTO, icmp_may_rst, CTLFLAG_RW | CTLFLAG_LOCKED, &icmp_may_rst, 0, 
     "Certain ICMP unreachable messages may abort connections in SYN_SENT");


@@ -242,23 +237,35 @@ extern struct tcp_cc_algo tcp_cc_ledbat;
 SYSCTL_INT(_net_inet_tcp, OID_AUTO, background_sockets, CTLFLAG_RD | CTLFLAG_LOCKED,
        &tcp_cc_ledbat.num_sockets, 0, "Number of sockets using background transport");
 
 SYSCTL_INT(_net_inet_tcp, OID_AUTO, background_sockets, CTLFLAG_RD | CTLFLAG_LOCKED,
        &tcp_cc_ledbat.num_sockets, 0, "Number of sockets using background transport");
 

+__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 void    tcp_cleartaocache(void);
 static void    tcp_notify(struct inpcb *, int);
 static void    tcp_cc_init(void);
 
 struct zone    *sack_hole_zone;
 struct zone    *tcp_reass_zone;
 static void    tcp_cleartaocache(void);
 static void    tcp_notify(struct inpcb *, int);
 static void    tcp_cc_init(void);
 
 struct zone    *sack_hole_zone;
 struct zone    *tcp_reass_zone;

-
+struct zone    *tcp_bwmeas_zone;
+#if 0
+static unsigned int tcp_mptcp_dsnm_sz;
+struct zone *tcp_mptcp_dsnm_zone;
+#endif

 /* The array containing pointers to currently implemented TCP CC algorithms */
 struct tcp_cc_algo* tcp_cc_algo_list[TCP_CC_ALGO_COUNT];
 
 /* The array containing pointers to currently implemented TCP CC algorithms */
 struct tcp_cc_algo* tcp_cc_algo_list[TCP_CC_ALGO_COUNT];
 

-extern unsigned int total_mb_cnt;
-extern unsigned int total_cl_cnt;
-extern int sbspace_factor;
-extern int tcp_sockthreshold;

 extern int slowlink_wsize;     /* window correction for slow links */
 extern int path_mtu_discovery;
 
 extern int slowlink_wsize;     /* window correction for slow links */
 extern int path_mtu_discovery;
 

+extern u_int32_t tcp_autorcvbuf_max;
+extern u_int32_t tcp_autorcvbuf_inc_shift;
+static void tcp_sbrcv_grow_rwin(struct tcpcb *tp, struct sockbuf *sb);
+
+#define TCP_BWMEAS_BURST_MINSIZE 6
+#define TCP_BWMEAS_BURST_MAXSIZE 25
+
+static uint32_t bwmeas_elm_size;

 
 /*
  * Target size of TCP PCB hash tables. Must be a power of two.
 
 /*
  * Target size of TCP PCB hash tables. Must be a power of two.


@@ -289,9 +296,6 @@ struct      inp_tp {
 };
 #undef ALIGNMENT
 
 };
 #undef ALIGNMENT
 

-extern struct  inpcbhead       time_wait_slots[];
-extern struct tcptimerlist tcp_timer_list;
-

 int  get_inpcb_str_size(void);
 int  get_tcp_str_size(void);
 
 int  get_inpcb_str_size(void);
 int  get_tcp_str_size(void);
 


@@ -300,6 +304,7 @@ static void tcpcb_to_otcpcb(struct tcpcb *, struct otcpcb *);
 static lck_attr_t *tcp_uptime_mtx_attr = NULL;         /* mutex attributes */
 static lck_grp_t *tcp_uptime_mtx_grp = NULL;           /* mutex group definition */
 static lck_grp_attr_t *tcp_uptime_mtx_grp_attr = NULL; /* mutex group attributes */
 static lck_attr_t *tcp_uptime_mtx_attr = NULL;         /* mutex attributes */
 static lck_grp_t *tcp_uptime_mtx_grp = NULL;           /* mutex group definition */
 static lck_grp_attr_t *tcp_uptime_mtx_grp_attr = NULL; /* mutex group attributes */

+int tcp_notsent_lowat_check(struct socket *so);

 
 
 int  get_inpcb_str_size(void)
 
 
 int  get_inpcb_str_size(void)


@@ -331,18 +336,26 @@ tcp_cc_init(void)
  * Tcp initialization
  */
 void
  * Tcp initialization
  */
 void

-tcp_init()
+tcp_init(struct protosw *pp, struct domain *dp)

 {
 {

+#pragma unused(dp)
+       static int tcp_initialized = 0;

        vm_size_t       str_size;
        vm_size_t       str_size;

-       int i;
-       struct inpcbinfo *pcbinfo;
-       
+       struct inpcbinfo *pcbinfo;
+
+       VERIFY((pp->pr_flags & (PR_INITIALIZED|PR_ATTACHED)) == PR_ATTACHED);
+
+       if (tcp_initialized)
+               return;
+       tcp_initialized = 1;
+

        tcp_ccgen = 1;
        tcp_cleartaocache();
 
        tcp_keepinit = TCPTV_KEEP_INIT;
        tcp_keepidle = TCPTV_KEEP_IDLE;
        tcp_keepintvl = TCPTV_KEEPINTVL;
        tcp_ccgen = 1;
        tcp_cleartaocache();
 
        tcp_keepinit = TCPTV_KEEP_INIT;
        tcp_keepidle = TCPTV_KEEP_IDLE;
        tcp_keepintvl = TCPTV_KEEPINTVL;

+       tcp_keepcnt = TCPTV_KEEPCNT;

        tcp_maxpersistidle = TCPTV_KEEP_IDLE;
        tcp_msl = TCPTV_MSL;
 
        tcp_maxpersistidle = TCPTV_KEEP_IDLE;
        tcp_msl = TCPTV_MSL;
 


@@ -351,21 +364,41 @@ tcp_init()
        tcp_now = tcp_now & 0x3fffffff; /* Starts tcp internal clock at a random value */
 
        LIST_INIT(&tcb);
        tcp_now = tcp_now & 0x3fffffff; /* Starts tcp internal clock at a random value */
 
        LIST_INIT(&tcb);

-       tcbinfo.listhead = &tcb;
+       tcbinfo.ipi_listhead = &tcb;
+

        pcbinfo = &tcbinfo;
        pcbinfo = &tcbinfo;

+       /*
+        * allocate lock group attribute and group for tcp pcb mutexes
+        */
+       pcbinfo->ipi_lock_grp_attr = lck_grp_attr_alloc_init();
+       pcbinfo->ipi_lock_grp = lck_grp_alloc_init("tcppcb", pcbinfo->ipi_lock_grp_attr);
+
+       /*
+        * allocate the lock attribute for tcp pcb mutexes
+        */
+       pcbinfo->ipi_lock_attr = lck_attr_alloc_init();
+
+       if ((pcbinfo->ipi_lock = lck_rw_alloc_init(pcbinfo->ipi_lock_grp,
+           pcbinfo->ipi_lock_attr)) == NULL) {
+               panic("%s: unable to allocate PCB lock\n", __func__);
+               /* NOTREACHED */
+       }
+

        if (!powerof2(tcp_tcbhashsize)) {
                printf("WARNING: TCB hash size not a power of 2\n");
                tcp_tcbhashsize = 512; /* safe default */
        }
        if (!powerof2(tcp_tcbhashsize)) {
                printf("WARNING: TCB hash size not a power of 2\n");
                tcp_tcbhashsize = 512; /* safe default */
        }

-       tcbinfo.hashsize = tcp_tcbhashsize;
-       tcbinfo.hashbase = hashinit(tcp_tcbhashsize, M_PCB, &tcbinfo.hashmask);
-       tcbinfo.porthashbase = hashinit(tcp_tcbhashsize, M_PCB,
-                                       &tcbinfo.porthashmask);
+       tcbinfo.ipi_hashbase = hashinit(tcp_tcbhashsize, M_PCB, &tcbinfo.ipi_hashmask);
+       tcbinfo.ipi_porthashbase = hashinit(tcp_tcbhashsize, M_PCB,
+                                       &tcbinfo.ipi_porthashmask);

        str_size = P2ROUNDUP(sizeof(struct inp_tp), sizeof(u_int64_t));
        str_size = P2ROUNDUP(sizeof(struct inp_tp), sizeof(u_int64_t));

-       tcbinfo.ipi_zone = (void *) zinit(str_size, 120000*str_size, 8192, "tcpcb");
+       tcbinfo.ipi_zone = zinit(str_size, 120000*str_size, 8192, "tcpcb");

        zone_change(tcbinfo.ipi_zone, Z_CALLERACCT, FALSE);
        zone_change(tcbinfo.ipi_zone, Z_EXPAND, TRUE);
 
        zone_change(tcbinfo.ipi_zone, Z_CALLERACCT, FALSE);
        zone_change(tcbinfo.ipi_zone, Z_EXPAND, TRUE);
 

+       tcbinfo.ipi_gc = tcp_gc;
+       in_pcbinfo_attach(&tcbinfo);
+

        str_size = P2ROUNDUP(sizeof(struct sackhole), sizeof(u_int64_t));
        sack_hole_zone = zinit(str_size, 120000*str_size, 8192, "sack_hole zone");
        zone_change(sack_hole_zone, Z_CALLERACCT, FALSE);
        str_size = P2ROUNDUP(sizeof(struct sackhole), sizeof(u_int64_t));
        sack_hole_zone = zinit(str_size, 120000*str_size, 8192, "sack_hole zone");
        zone_change(sack_hole_zone, Z_CALLERACCT, FALSE);


@@ -382,36 +415,30 @@ tcp_init()
        zone_change(tcp_reass_zone, Z_CALLERACCT, FALSE);
        zone_change(tcp_reass_zone, Z_EXPAND, TRUE);
 
        zone_change(tcp_reass_zone, Z_CALLERACCT, FALSE);
        zone_change(tcp_reass_zone, Z_EXPAND, TRUE);
 

+       bwmeas_elm_size = P2ROUNDUP(sizeof(struct bwmeas), sizeof(u_int64_t));
+       tcp_bwmeas_zone = zinit(bwmeas_elm_size, (100 * bwmeas_elm_size), 0, "tcp_bwmeas_zone");
+       if (tcp_bwmeas_zone == NULL) {
+               panic("%s: failed allocating tcp_bwmeas_zone", __func__);
+               /* NOTREACHED */
+       }
+       zone_change(tcp_bwmeas_zone, Z_CALLERACCT, FALSE);
+       zone_change(tcp_bwmeas_zone, Z_EXPAND, TRUE);
+

 #if INET6
 #define TCP_MINPROTOHDR (sizeof(struct ip6_hdr) + sizeof(struct tcphdr))
 #else /* INET6 */
 #define TCP_MINPROTOHDR (sizeof(struct tcpiphdr))
 #endif /* INET6 */
 #if INET6
 #define TCP_MINPROTOHDR (sizeof(struct ip6_hdr) + sizeof(struct tcphdr))
 #else /* INET6 */
 #define TCP_MINPROTOHDR (sizeof(struct tcpiphdr))
 #endif /* INET6 */

-       if (max_protohdr < TCP_MINPROTOHDR)
-               max_protohdr = TCP_MINPROTOHDR;
-       if (max_linkhdr + TCP_MINPROTOHDR > MHLEN)
+       if (max_protohdr < TCP_MINPROTOHDR) {
+               _max_protohdr = TCP_MINPROTOHDR;
+               _max_protohdr = max_protohdr;   /* round it up */
+       }
+       if (max_linkhdr + max_protohdr > MCLBYTES)

                panic("tcp_init");
 #undef TCP_MINPROTOHDR
 
                panic("tcp_init");
 #undef TCP_MINPROTOHDR
 

-       /*
-        * allocate lock group attribute and group for tcp pcb mutexes
-        */
-       pcbinfo->mtx_grp_attr = lck_grp_attr_alloc_init();
-       pcbinfo->mtx_grp = lck_grp_alloc_init("tcppcb", pcbinfo->mtx_grp_attr);
-               
-       /*
-        * allocate the lock attribute for tcp pcb mutexes
-        */
-       pcbinfo->mtx_attr = lck_attr_alloc_init();
-
-       if ((pcbinfo->mtx = lck_rw_alloc_init(pcbinfo->mtx_grp, pcbinfo->mtx_attr)) == NULL) {
-               printf("tcp_init: mutex not alloced!\n");
-               return; /* pretty much dead if this fails... */
-       }
-
-       for (i=0; i < N_TIME_WAIT_SLOTS; i++) {
-            LIST_INIT(&time_wait_slots[i]);
-       }
+       /* Initialize time wait and timer lists */
+       TAILQ_INIT(&tcp_tw_tailq);

 
        bzero(&tcp_timer_list, sizeof(tcp_timer_list));
        LIST_INIT(&tcp_timer_list.lhead);
 
        bzero(&tcp_timer_list, sizeof(tcp_timer_list));
        LIST_INIT(&tcp_timer_list.lhead);


@@ -440,6 +467,9 @@ tcp_init()
 
        /* Initialize TCP congestion control algorithms list */
        tcp_cc_init();
 
        /* Initialize TCP congestion control algorithms list */
        tcp_cc_init();

+
+       /* Initialize TCP LRO data structures */
+       tcp_lro_init();

 }
 
 /*
 }
 
 /*


@@ -462,16 +492,15 @@ tcp_fillheaders(tp, ip_ptr, tcp_ptr)
 
                ip6 = (struct ip6_hdr *)ip_ptr;
                ip6->ip6_flow = (ip6->ip6_flow & ~IPV6_FLOWINFO_MASK) |
 
                ip6 = (struct ip6_hdr *)ip_ptr;
                ip6->ip6_flow = (ip6->ip6_flow & ~IPV6_FLOWINFO_MASK) |

-                       (inp->in6p_flowinfo & IPV6_FLOWINFO_MASK);
+                       (inp->inp_flow & IPV6_FLOWINFO_MASK);

                ip6->ip6_vfc = (ip6->ip6_vfc & ~IPV6_VERSION_MASK) |
                        (IPV6_VERSION & IPV6_VERSION_MASK);
                ip6->ip6_nxt = IPPROTO_TCP;
                ip6->ip6_plen = sizeof(struct tcphdr);
                ip6->ip6_src = inp->in6p_laddr;
                ip6->ip6_dst = inp->in6p_faddr;
                ip6->ip6_vfc = (ip6->ip6_vfc & ~IPV6_VERSION_MASK) |
                        (IPV6_VERSION & IPV6_VERSION_MASK);
                ip6->ip6_nxt = IPPROTO_TCP;
                ip6->ip6_plen = sizeof(struct tcphdr);
                ip6->ip6_src = inp->in6p_laddr;
                ip6->ip6_dst = inp->in6p_faddr;

-               tcp_hdr->th_sum = in6_cksum_phdr(&inp->in6p_laddr,
-                   &inp->in6p_faddr, htonl(sizeof(struct tcphdr)),
-                   htonl(IPPROTO_TCP));
+               tcp_hdr->th_sum = in6_pseudo(&inp->in6p_laddr, &inp->in6p_faddr,
+                   htonl(sizeof (struct tcphdr) + IPPROTO_TCP));

        } else
 #endif
        {
        } else
 #endif
        {


@@ -563,7 +592,7 @@ tcp_respond(
        struct ip6_hdr *ip6;
        int isipv6;
 #endif /* INET6 */
        struct ip6_hdr *ip6;
        int isipv6;
 #endif /* INET6 */

-       unsigned int outif;
+       struct ifnet *outif;

 
 #if INET6
        isipv6 = IP_VHL_V(((struct ip *)ipgen)->ip_vhl) == 6;
 
 #if INET6
        isipv6 = IP_VHL_V(((struct ip *)ipgen)->ip_vhl) == 6;


@@ -603,18 +632,27 @@ tcp_respond(
                m->m_data += max_linkhdr;
 #if INET6
                if (isipv6) {
                m->m_data += max_linkhdr;
 #if INET6
                if (isipv6) {

+                       VERIFY((MHLEN - max_linkhdr) >=
+                           (sizeof (*ip6) + sizeof (*nth)));

                        bcopy((caddr_t)ip6, mtod(m, caddr_t), 
                              sizeof(struct ip6_hdr));
                        ip6 = mtod(m, struct ip6_hdr *);
                        bcopy((caddr_t)ip6, mtod(m, caddr_t), 
                              sizeof(struct ip6_hdr));
                        ip6 = mtod(m, struct ip6_hdr *);

-                       nth = (struct tcphdr *)(ip6 + 1);
+                       nth = (struct tcphdr *)(void *)(ip6 + 1);

                } else
 #endif /* INET6 */
                {
                } else
 #endif /* INET6 */
                {

+                       VERIFY((MHLEN - max_linkhdr) >=
+                           (sizeof (*ip) + sizeof (*nth)));

                        bcopy((caddr_t)ip, mtod(m, caddr_t), sizeof(struct ip));
                        ip = mtod(m, struct ip *);
                        bcopy((caddr_t)ip, mtod(m, caddr_t), sizeof(struct ip));
                        ip = mtod(m, struct ip *);

-                       nth = (struct tcphdr *)(ip + 1);
+                       nth = (struct tcphdr *)(void *)(ip + 1);

                }
                bcopy((caddr_t)th, (caddr_t)nth, sizeof(struct tcphdr));
                }
                bcopy((caddr_t)th, (caddr_t)nth, sizeof(struct tcphdr));

+#if MPTCP
+               if ((tp) && (tp->t_mpflags & TMPF_RESET))
+                       flags = (TH_RST | TH_ACK);
+               else
+#endif

                flags = TH_ACK;
        } else {
                m_freem(m->m_next);
                flags = TH_ACK;
        } else {
                m_freem(m->m_next);


@@ -625,13 +663,17 @@ tcp_respond(
 #define xchg(a,b,type) { type t; t=a; a=b; b=t; }
 #if INET6
                if (isipv6) {
 #define xchg(a,b,type) { type t; t=a; a=b; b=t; }
 #if INET6
                if (isipv6) {

+                       /* Expect 32-bit aligned IP on strict-align platforms */
+                       IP6_HDR_STRICT_ALIGNMENT_CHECK(ip6);

                        xchg(ip6->ip6_dst, ip6->ip6_src, struct in6_addr);
                        xchg(ip6->ip6_dst, ip6->ip6_src, struct in6_addr);

-                       nth = (struct tcphdr *)(ip6 + 1);
+                       nth = (struct tcphdr *)(void *)(ip6 + 1);

                } else
 #endif /* INET6 */
              {
                } else
 #endif /* INET6 */
              {

+               /* Expect 32-bit aligned IP on strict-align platforms */
+               IP_HDR_STRICT_ALIGNMENT_CHECK(ip);

                xchg(ip->ip_dst.s_addr, ip->ip_src.s_addr, n_long);
                xchg(ip->ip_dst.s_addr, ip->ip_src.s_addr, n_long);

-               nth = (struct tcphdr *)(ip + 1);
+               nth = (struct tcphdr *)(void *)(ip + 1);

              }
                if (th != nth) {
                        /*
              }
                if (th != nth) {
                        /*


@@ -689,9 +731,8 @@ tcp_respond(
 #if INET6
        if (isipv6) {
                nth->th_sum = 0;
 #if INET6
        if (isipv6) {
                nth->th_sum = 0;

-               nth->th_sum = in6_cksum_phdr(&ip6->ip6_src,
-                   &ip6->ip6_dst, htons((u_short)(tlen - sizeof(struct ip6_hdr))),
-                               htonl(IPPROTO_TCP));
+               nth->th_sum = in6_pseudo(&ip6->ip6_src, &ip6->ip6_dst,
+                   htonl((tlen - sizeof (struct ip6_hdr)) + IPPROTO_TCP));

                m->m_pkthdr.csum_flags = CSUM_TCPIPV6;
                m->m_pkthdr.csum_data = offsetof(struct tcphdr, th_sum);
                ip6->ip6_hlim = in6_selecthlim(tp ? tp->t_inpcb : NULL,
                m->m_pkthdr.csum_flags = CSUM_TCPIPV6;
                m->m_pkthdr.csum_data = offsetof(struct tcphdr, th_sum);
                ip6->ip6_hlim = in6_selecthlim(tp ? tp->t_inpcb : NULL,


@@ -717,28 +758,56 @@ tcp_respond(
        }
 #endif
 
        }
 #endif
 

-       if (tp != NULL)
-               set_packet_tclass(m, tp->t_inpcb->inp_socket, MBUF_TC_UNSPEC, isipv6);
+       if (tp != NULL) {
+               u_int32_t svc_flags = 0;
+               if (isipv6) {
+                       svc_flags |= PKT_SCF_IPV6;
+               }
+               set_packet_service_class(m, tp->t_inpcb->inp_socket,
+                   MBUF_SC_UNSPEC, svc_flags);
+
+               /* Embed flowhash and flow control flags */
+               m->m_pkthdr.pkt_flowsrc = FLOWSRC_INPCB;
+               m->m_pkthdr.pkt_flowid = tp->t_inpcb->inp_flowhash;
+               m->m_pkthdr.pkt_flags |= PKTF_FLOW_ID | PKTF_FLOW_LOCALSRC;
+#if MPTCP
+               /* Disable flow advisory when using MPTCP. */
+               if (!(tp->t_mpflags & TMPF_MPTCP_TRUE))
+#endif /* MPTCP */
+                       m->m_pkthdr.pkt_flags |= PKTF_FLOW_ADV;
+               m->m_pkthdr.pkt_proto = IPPROTO_TCP;
+       }

 
 #if INET6
        if (isipv6) {
 
 #if INET6
        if (isipv6) {

-               struct ip6_out_args ip6oa = { ifscope, nocell };
+               struct ip6_out_args ip6oa = { ifscope, { 0 },
+                   IP6OAF_SELECT_SRCIF | IP6OAF_BOUND_SRCADDR, 0 };
+
+               if (ifscope != IFSCOPE_NONE)
+                       ip6oa.ip6oa_flags |= IP6OAF_BOUND_IF;
+               if (nocell)
+                       ip6oa.ip6oa_flags |= IP6OAF_NO_CELLULAR;

 
                (void) ip6_output(m, NULL, ro6, IPV6_OUTARGS, NULL,
                    NULL, &ip6oa);
 
                (void) ip6_output(m, NULL, ro6, IPV6_OUTARGS, NULL,
                    NULL, &ip6oa);

-               if (ro6->ro_rt != NULL) {
-                       if (ro6 == &sro6) {
-                               rtfree(ro6->ro_rt);
-                               ro6->ro_rt = NULL;
-                       } else if ((outif = ro6->ro_rt->rt_ifp->if_index) !=
-                           tp->t_inpcb->in6p_last_outif) {
-                               tp->t_inpcb->in6p_last_outif = outif;
-                       }
-               }
+
+               if (tp != NULL && ro6 != NULL && ro6->ro_rt != NULL &&
+                   (outif = ro6->ro_rt->rt_ifp) !=
+                   tp->t_inpcb->in6p_last_outifp)
+                       tp->t_inpcb->in6p_last_outifp = outif;
+
+               if (ro6 == &sro6)
+                       ROUTE_RELEASE(ro6);

        } else
 #endif /* INET6 */
        {
        } else
 #endif /* INET6 */
        {

-               struct ip_out_args ipoa = { ifscope, nocell };
+               struct ip_out_args ipoa = { ifscope, { 0 },
+                   IPOAF_SELECT_SRCIF | IPOAF_BOUND_SRCADDR, 0 };
+
+               if (ifscope != IFSCOPE_NONE)
+                       ipoa.ipoa_flags |= IPOAF_BOUND_IF;
+               if (nocell)
+                       ipoa.ipoa_flags |= IPOAF_NO_CELLULAR;

 
                if (ro != &sro) {
                        /* Copy the cached route and take an extra reference */
 
                if (ro != &sro) {
                        /* Copy the cached route and take an extra reference */


@@ -749,15 +818,16 @@ tcp_respond(
                 */
                (void) ip_output(m, NULL, &sro, IP_OUTARGS, NULL, &ipoa);
 
                 */
                (void) ip_output(m, NULL, &sro, IP_OUTARGS, NULL, &ipoa);
 

+               if (tp != NULL && sro.ro_rt != NULL &&
+                   (outif = sro.ro_rt->rt_ifp) !=
+                   tp->t_inpcb->inp_last_outifp)
+                       tp->t_inpcb->inp_last_outifp = outif;
+

                if (ro != &sro) {
                if (ro != &sro) {

-                       if (sro.ro_rt != NULL &&
-                           (outif = sro.ro_rt->rt_ifp->if_index) !=
-                           tp->t_inpcb->inp_last_outif)
-                               tp->t_inpcb->inp_last_outif = outif;

                        /* Synchronize cached PCB route */
                        inp_route_copyin(tp->t_inpcb, &sro);
                        /* Synchronize cached PCB route */
                        inp_route_copyin(tp->t_inpcb, &sro);

-               } else if (sro.ro_rt != NULL) {
-                       rtfree(sro.ro_rt);
+               } else {
+                       ROUTE_RELEASE(&sro);

                }
        }
 }
                }
        }
 }


@@ -781,12 +851,12 @@ tcp_newtcpcb(inp)
 
        calculate_tcp_clock();
 
 
        calculate_tcp_clock();
 

-       if (so->cached_in_sock_layer == 0) {
-            it = (struct inp_tp *)inp;
+       if (!so->cached_in_sock_layer) {
+            it = (struct inp_tp *)(void *)inp;

             tp = &it->tcb;
             tp = &it->tcb;

+       } else {
+            tp = (struct tcpcb *)(void *)inp->inp_saved_ppcb;

        }
        }

-       else
-            tp = (struct tcpcb *) inp->inp_saved_ppcb;

        
        bzero((char *) tp, sizeof(struct tcpcb));
        LIST_INIT(&tp->t_segq);
        
        bzero((char *) tp, sizeof(struct tcpcb));
        LIST_INIT(&tp->t_segq);


@@ -798,7 +868,9 @@ tcp_newtcpcb(inp)
 
        if (tcp_do_rfc1323)
                tp->t_flags = (TF_REQ_SCALE|TF_REQ_TSTMP);
 
        if (tcp_do_rfc1323)
                tp->t_flags = (TF_REQ_SCALE|TF_REQ_TSTMP);

-       tp->sack_enable = tcp_do_sack;
+       if (tcp_do_sack)
+               tp->t_flagsext |= TF_SACK_ENABLE;
+

        TAILQ_INIT(&tp->snd_holes);
        tp->t_inpcb = inp;      /* XXX */
        /*
        TAILQ_INIT(&tp->snd_holes);
        tp->t_inpcb = inp;      /* XXX */
        /*


@@ -820,15 +892,19 @@ tcp_newtcpcb(inp)
        }
 
        tp->snd_cwnd = TCP_MAXWIN << TCP_MAX_WINSHIFT;
        }
 
        tp->snd_cwnd = TCP_MAXWIN << TCP_MAX_WINSHIFT;

-       tp->snd_bwnd = TCP_MAXWIN << TCP_MAX_WINSHIFT;

        tp->snd_ssthresh = TCP_MAXWIN << TCP_MAX_WINSHIFT;
        tp->snd_ssthresh_prev = TCP_MAXWIN << TCP_MAX_WINSHIFT;
        tp->t_rcvtime = tcp_now;
        tp->snd_ssthresh = TCP_MAXWIN << TCP_MAX_WINSHIFT;
        tp->snd_ssthresh_prev = TCP_MAXWIN << TCP_MAX_WINSHIFT;
        tp->t_rcvtime = tcp_now;

-       tp->t_bw_rtttime = 0;

        tp->tentry.timer_start = tcp_now;
        tp->t_persist_timeout = tcp_max_persist_timeout;
        tp->t_persist_stop = 0;
        tp->t_flagsext |= TF_RCVUNACK_WAITSS;
        tp->tentry.timer_start = tcp_now;
        tp->t_persist_timeout = tcp_max_persist_timeout;
        tp->t_persist_stop = 0;
        tp->t_flagsext |= TF_RCVUNACK_WAITSS;

+       tp->t_rexmtthresh = tcprexmtthresh;
+
+       /* Clear time wait tailq entry */
+       tp->t_twentry.tqe_next = NULL;
+       tp->t_twentry.tqe_prev = NULL;
+

        /*
         * IPv4 TTL initialization is necessary for an IPv6 socket as well,
         * because the socket may be bound to an IPv6 wildcard address,
        /*
         * IPv4 TTL initialization is necessary for an IPv6 socket as well,
         * because the socket may be bound to an IPv6 wildcard address,


@@ -852,7 +928,7 @@ tcp_drop(tp, errno)
        struct socket *so = tp->t_inpcb->inp_socket;
 #if CONFIG_DTRACE
        struct inpcb *inp = tp->t_inpcb;
        struct socket *so = tp->t_inpcb->inp_socket;
 #if CONFIG_DTRACE
        struct inpcb *inp = tp->t_inpcb;

-#endif /* CONFIG_DTRACE */
+#endif

 
        if (TCPS_HAVERCVDSYN(tp->t_state)) {
                DTRACE_TCP4(state__change, void, NULL, struct inpcb *, inp,
 
        if (TCPS_HAVERCVDSYN(tp->t_state)) {
                DTRACE_TCP4(state__change, void, NULL, struct inpcb *, inp,


@@ -868,6 +944,39 @@ tcp_drop(tp, errno)
        return (tcp_close(tp));
 }
 
        return (tcp_close(tp));
 }
 

+void
+tcp_getrt_rtt(struct tcpcb *tp, struct rtentry *rt) 
+{
+       u_int32_t rtt = rt->rt_rmx.rmx_rtt;
+       int isnetlocal = (tp->t_flags & TF_LOCAL);
+
+       if (rtt != 0) {
+               /*
+                * XXX the lock bit for RTT indicates that the value
+                * is also a minimum value; this is subject to time.
+                */
+               if (rt->rt_rmx.rmx_locks & RTV_RTT)
+                       tp->t_rttmin = rtt / (RTM_RTTUNIT / TCP_RETRANSHZ);
+               else
+                       tp->t_rttmin = isnetlocal ? tcp_TCPTV_MIN : TCPTV_REXMTMIN;
+               tp->t_srtt = rtt / (RTM_RTTUNIT / (TCP_RETRANSHZ * TCP_RTT_SCALE));
+               tcpstat.tcps_usedrtt++;
+               if (rt->rt_rmx.rmx_rttvar) {
+                       tp->t_rttvar = rt->rt_rmx.rmx_rttvar /
+                               (RTM_RTTUNIT / (TCP_RETRANSHZ * TCP_RTTVAR_SCALE));
+                       tcpstat.tcps_usedrttvar++;
+               } else {
+                       /* default variation is +- 1 rtt */
+                       tp->t_rttvar =
+                               tp->t_srtt * TCP_RTTVAR_SCALE / TCP_RTT_SCALE;
+               }
+               TCPT_RANGESET(tp->t_rxtcur,
+                       ((tp->t_srtt >> 2) + tp->t_rttvar) >> 1,
+                       tp->t_rttmin, TCPTV_REXMTMAX,
+                       TCP_ADD_REXMTSLOP(tp));
+       }
+}
+

 /*
  * Close a TCP control block:
  *     discard all space held by the tcp
 /*
  * Close a TCP control block:
  *     discard all space held by the tcp


@@ -883,10 +992,12 @@ tcp_close(tp)
 #if INET6
        int isipv6 = (inp->inp_vflag & INP_IPV6) != 0;
 #endif /* INET6 */
 #if INET6
        int isipv6 = (inp->inp_vflag & INP_IPV6) != 0;
 #endif /* INET6 */

+       struct route *ro;

        struct rtentry *rt;
        int dosavessthresh;
 
        struct rtentry *rt;
        int dosavessthresh;
 

-       if ( inp->inp_ppcb == NULL) /* tcp_close was called previously, bail */
+       /* tcp_close was called previously, bail */
+       if ( inp->inp_ppcb == NULL) 

                return(NULL);
 
        tcp_canceltimers(tp);
                return(NULL);
 
        tcp_canceltimers(tp);


@@ -904,20 +1015,25 @@ tcp_close(tp)
         * point both flags should be cleared and we can proceed further
         * with the cleanup.
         */
         * point both flags should be cleared and we can proceed further
         * with the cleanup.
         */

-       if (tp->t_flags & (TF_CLOSING|TF_SENDINPROG)) {
+       if ((tp->t_flags & TF_CLOSING) ||
+               inp->inp_sndinprog_cnt > 0) {

                tp->t_flags |= TF_CLOSING;
                return (NULL);
        }
 
                tp->t_flags |= TF_CLOSING;
                return (NULL);
        }
 

+       DTRACE_TCP4(state__change, void, NULL, struct inpcb *, inp,
+               struct tcpcb *, tp, int32_t, TCPS_CLOSED);
+

        if (CC_ALGO(tp)->cleanup != NULL) {
                CC_ALGO(tp)->cleanup(tp);
        }
 
 #if INET6
        if (CC_ALGO(tp)->cleanup != NULL) {
                CC_ALGO(tp)->cleanup(tp);
        }
 
 #if INET6

-       rt = isipv6 ? inp->in6p_route.ro_rt : inp->inp_route.ro_rt;
+       ro = (isipv6 ? (struct route *)&inp->in6p_route : &inp->inp_route);

 #else
 #else

-       rt = inp->inp_route.ro_rt;
+       ro = &inp->inp_route;

 #endif
 #endif

+       rt = ro->ro_rt;

        if (rt != NULL)
                RT_LOCK_SPIN(rt);
 
        if (rt != NULL)
                RT_LOCK_SPIN(rt);
 


@@ -941,18 +1057,19 @@ tcp_close(tp)
 
                        if (rt == NULL)
                                goto no_valid_rt;
 
                        if (rt == NULL)
                                goto no_valid_rt;

-                       sin6 = (struct sockaddr_in6 *)rt_key(rt);
+                       sin6 = (struct sockaddr_in6 *)(void *)rt_key(rt);

                        if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr))
                                goto no_valid_rt;
                }
                else
 #endif /* INET6 */
                        if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr))
                                goto no_valid_rt;
                }
                else
 #endif /* INET6 */

-               if (rt == NULL || !(rt->rt_flags & RTF_UP) ||
-                   ((struct sockaddr_in *)rt_key(rt))->sin_addr.s_addr ==
-                   INADDR_ANY || rt->generation_id != route_generation) {
+               if (ROUTE_UNUSABLE(ro) ||
+                   SIN(rt_key(rt))->sin_addr.s_addr == INADDR_ANY) {

                        if (tp->t_state >= TCPS_CLOSE_WAIT) {
                        if (tp->t_state >= TCPS_CLOSE_WAIT) {

-                               DTRACE_TCP4(state__change, void, NULL, struct inpcb *, inp,
-                                       struct tcpcb *, tp, int32_t, TCPS_CLOSING);
+                               DTRACE_TCP4(state__change, 
+                                   void, NULL, struct inpcb *, inp,
+                                   struct tcpcb *, tp, int32_t, 
+                                   TCPS_CLOSING);

                                tp->t_state = TCPS_CLOSING;
                        }
                        goto no_valid_rt;
                                tp->t_state = TCPS_CLOSING;
                        }
                        goto no_valid_rt;


@@ -1051,23 +1168,40 @@ no_valid_rt:
        (void) tcp_freeq(tp);
 
        tcp_free_sackholes(tp);
        (void) tcp_freeq(tp);
 
        tcp_free_sackholes(tp);

+       if (tp->t_bwmeas != NULL) {
+               tcp_bwmeas_free(tp);
+       }

 
        /* Free the packet list */
        if (tp->t_pktlist_head != NULL)
                m_freem_list(tp->t_pktlist_head);
        TCP_PKTLIST_CLEAR(tp);
 
 
        /* Free the packet list */
        if (tp->t_pktlist_head != NULL)
                m_freem_list(tp->t_pktlist_head);
        TCP_PKTLIST_CLEAR(tp);
 

-#ifdef __APPLE__
+#if MPTCP
+       /* Clear MPTCP state */
+       tp->t_mpflags = 0;
+#endif /* MPTCP */
+

        if (so->cached_in_sock_layer)
            inp->inp_saved_ppcb = (caddr_t) tp;
        if (so->cached_in_sock_layer)
            inp->inp_saved_ppcb = (caddr_t) tp;

-#endif
+

        /* Issue a wakeup before detach so that we don't miss
         * a wakeup
         */
        sodisconnectwakeup(so);
 
        /* Issue a wakeup before detach so that we don't miss
         * a wakeup
         */
        sodisconnectwakeup(so);
 

+       /* 
+        * Clean up any LRO state 
+        */
+       if (tp->t_flagsext & TF_LRO_OFFLOADED) {
+               tcp_lro_remove_state(inp->inp_laddr, inp->inp_faddr, 
+                       inp->inp_lport,
+                       inp->inp_fport);
+               tp->t_flagsext &= ~TF_LRO_OFFLOADED;
+       }
+       tp->t_state = TCPS_CLOSED;

 #if INET6
 #if INET6

-       if (INP_CHECK_SOCKAF(so, AF_INET6))
+       if (SOCK_CHECK_DOM(so, PF_INET6))

                in6_pcbdetach(inp);
        else
 #endif /* INET6 */
                in6_pcbdetach(inp);
        else
 #endif /* INET6 */


@@ -1103,34 +1237,32 @@ tcp_drain()
 {
        if (do_tcpdrain)
        {
 {
        if (do_tcpdrain)
        {

-               struct inpcb *inpb;
-               struct tcpcb *tcpb;
-               struct tseg_qent *te;
-
+               struct inpcb *inp;
+               struct tcpcb *tp;

        /*
         * Walk the tcpbs, if existing, and flush the reassembly queue,
         * if there is one...
        /*
         * Walk the tcpbs, if existing, and flush the reassembly queue,
         * if there is one...

-        * XXX: The "Net/3" implementation doesn't imply that the TCP
-        *      reassembly queue should be flushed, but in a situation
-        *      where we're really low on mbufs, this is potentially
-        *      usefull.        
+        * Do it next time if the pcbinfo lock is in use

         */
         */

-               if (!lck_rw_try_lock_exclusive(tcbinfo.mtx)) /* do it next time if the lock is in use */
+               if (!lck_rw_try_lock_exclusive(tcbinfo.ipi_lock)) 

                        return;
 
                        return;
 

-               for (inpb = LIST_FIRST(tcbinfo.listhead); inpb;
-                       inpb = LIST_NEXT(inpb, inp_list)) {
-                               if ((tcpb = intotcpcb(inpb))) {
-                                       while ((te = LIST_FIRST(&tcpb->t_segq))
-                                              != NULL) {
-                                       LIST_REMOVE(te, tqe_q);
-                                       m_freem(te->tqe_m);
-                                       zfree(tcp_reass_zone, te);
-                                       tcp_reass_qsize--;
-                               }
+               LIST_FOREACH(inp, tcbinfo.ipi_listhead, inp_list) {
+                       if (in_pcb_checkstate(inp, WNT_ACQUIRE, 0) !=
+                               WNT_STOPUSING) {
+                               tcp_lock(inp->inp_socket, 1, 0);
+                               if (in_pcb_checkstate(inp, WNT_RELEASE, 1)
+                                       == WNT_STOPUSING) {
+                                       /* lost a race, try the next one */
+                                       tcp_unlock(inp->inp_socket, 1, 0);
+                                       continue;
+                               } 
+                               tp = intotcpcb(inp);
+                               tcp_freeq(tp);
+                               tcp_unlock(inp->inp_socket, 1, 0);

                        }
                }
                        }
                }

-               lck_rw_done(tcbinfo.mtx);
+               lck_rw_done(tcbinfo.ipi_lock);

 
        }
 }
 
        }
 }


@@ -1178,6 +1310,30 @@ tcp_notify(inp, error)
 #endif
 }
 
 #endif
 }
 

+struct bwmeas*
+tcp_bwmeas_alloc(struct tcpcb *tp)
+{
+       struct bwmeas *elm;
+       elm = zalloc(tcp_bwmeas_zone);
+       if (elm == NULL)
+               return(elm);
+
+       bzero(elm, bwmeas_elm_size);
+       elm->bw_minsizepkts = TCP_BWMEAS_BURST_MINSIZE;
+       elm->bw_maxsizepkts = TCP_BWMEAS_BURST_MAXSIZE;
+       elm->bw_minsize = elm->bw_minsizepkts * tp->t_maxseg;
+       elm->bw_maxsize = elm->bw_maxsizepkts * tp->t_maxseg;
+       return(elm);
+}
+
+void
+tcp_bwmeas_free(struct tcpcb* tp)
+{
+       zfree(tcp_bwmeas_zone, tp->t_bwmeas);
+       tp->t_bwmeas = NULL;
+       tp->t_flagsext &= ~(TF_MEASURESNDBW);
+}
+

 /*
  * tcpcb_to_otcpcb copies specific bits of a tcpcb to a otcpcb format.
  * The otcpcb data structure is passed to user space and must not change.
 /*
  * tcpcb_to_otcpcb copies specific bits of a tcpcb to a otcpcb format.
  * The otcpcb data structure is passed to user space and must not change.


@@ -1238,34 +1394,34 @@ tcpcb_to_otcpcb(struct tcpcb *tp, struct otcpcb *otp)
        otp->snd_recover = tp->snd_recover;
        otp->snd_cwnd_prev = tp->snd_cwnd_prev;
        otp->snd_ssthresh_prev = tp->snd_ssthresh_prev;
        otp->snd_recover = tp->snd_recover;
        otp->snd_cwnd_prev = tp->snd_cwnd_prev;
        otp->snd_ssthresh_prev = tp->snd_ssthresh_prev;

-       otp->t_badrxtwin = tp->t_badrxtwin;
+       otp->t_badrxtwin = 0;

 }
 
 static int
 tcp_pcblist SYSCTL_HANDLER_ARGS
 {
 #pragma unused(oidp, arg1, arg2)
 }
 
 static int
 tcp_pcblist SYSCTL_HANDLER_ARGS
 {
 #pragma unused(oidp, arg1, arg2)

-       int error, i, n;
+       int error, i = 0, n;

        struct inpcb *inp, **inp_list;
        struct inpcb *inp, **inp_list;

+       struct tcpcb *tp;

        inp_gen_t gencnt;
        struct xinpgen xig;
        inp_gen_t gencnt;
        struct xinpgen xig;

-       int slot;

 
        /*
         * The process of preparing the TCB list is too time-consuming and
         * resource-intensive to repeat twice on every request.
         */
 
        /*
         * The process of preparing the TCB list is too time-consuming and
         * resource-intensive to repeat twice on every request.
         */

-       lck_rw_lock_shared(tcbinfo.mtx);
+       lck_rw_lock_shared(tcbinfo.ipi_lock);

        if (req->oldptr == USER_ADDR_NULL) {
                n = tcbinfo.ipi_count;
                req->oldidx = 2 * (sizeof xig)
                        + (n + n/8) * sizeof(struct xtcpcb);
        if (req->oldptr == USER_ADDR_NULL) {
                n = tcbinfo.ipi_count;
                req->oldidx = 2 * (sizeof xig)
                        + (n + n/8) * sizeof(struct xtcpcb);

-               lck_rw_done(tcbinfo.mtx);
+               lck_rw_done(tcbinfo.ipi_lock);

                return 0;
        }
 
        if (req->newptr != USER_ADDR_NULL) {
                return 0;
        }
 
        if (req->newptr != USER_ADDR_NULL) {

-               lck_rw_done(tcbinfo.mtx);
+               lck_rw_done(tcbinfo.ipi_lock);

                return EPERM;
        }
 
                return EPERM;
        }
 


@@ -1282,41 +1438,36 @@ tcp_pcblist SYSCTL_HANDLER_ARGS
        xig.xig_sogen = so_gencnt;
        error = SYSCTL_OUT(req, &xig, sizeof xig);
        if (error) {
        xig.xig_sogen = so_gencnt;
        error = SYSCTL_OUT(req, &xig, sizeof xig);
        if (error) {

-               lck_rw_done(tcbinfo.mtx);
+               lck_rw_done(tcbinfo.ipi_lock);

                return error;
        }
        /*
         * We are done if there is no pcb
         */
        if (n == 0) {
                return error;
        }
        /*
         * We are done if there is no pcb
         */
        if (n == 0) {

-               lck_rw_done(tcbinfo.mtx);
+               lck_rw_done(tcbinfo.ipi_lock);

                return 0; 
        }
 
        inp_list = _MALLOC(n * sizeof *inp_list, M_TEMP, M_WAITOK);
        if (inp_list == 0) {
                return 0; 
        }
 
        inp_list = _MALLOC(n * sizeof *inp_list, M_TEMP, M_WAITOK);
        if (inp_list == 0) {

-               lck_rw_done(tcbinfo.mtx);
+               lck_rw_done(tcbinfo.ipi_lock);

                return ENOMEM;
        }
        
                return ENOMEM;
        }
        

-       for (inp = LIST_FIRST(tcbinfo.listhead), i = 0; inp && i < n;
-            inp = LIST_NEXT(inp, inp_list)) {
-#ifdef __APPLE__
-               if (inp->inp_gencnt <= gencnt && inp->inp_state != INPCB_STATE_DEAD)
-#else
-               if (inp->inp_gencnt <= gencnt && !prison_xinpcb(req->p, inp))
-#endif
+       LIST_FOREACH(inp, tcbinfo.ipi_listhead, inp_list) {
+               if (inp->inp_gencnt <= gencnt && 
+                       inp->inp_state != INPCB_STATE_DEAD)

                        inp_list[i++] = inp;
                        inp_list[i++] = inp;

+               if (i >= n) break;

        }
 
        }
 

-       for (slot = 0; slot < N_TIME_WAIT_SLOTS; slot++) {
-               struct inpcb *inpnxt;
-
-               for (inp = time_wait_slots[slot].lh_first; inp && i < n; inp = inpnxt) {
-                       inpnxt = inp->inp_list.le_next;
-                       if (inp->inp_gencnt <= gencnt && inp->inp_state != INPCB_STATE_DEAD)
-                               inp_list[i++] = inp;
-               }
+       TAILQ_FOREACH(tp, &tcp_tw_tailq, t_twentry) {
+               inp = tp->t_inpcb;
+               if (inp->inp_gencnt <= gencnt && 
+                       inp->inp_state != INPCB_STATE_DEAD)
+                       inp_list[i++] = inp;
+               if (i >= n) break;

        }
 
        n = i;
        }
 
        n = i;


@@ -1324,7 +1475,8 @@ tcp_pcblist SYSCTL_HANDLER_ARGS
        error = 0;
        for (i = 0; i < n; i++) {
                inp = inp_list[i];
        error = 0;
        for (i = 0; i < n; i++) {
                inp = inp_list[i];

-               if (inp->inp_gencnt <= gencnt && inp->inp_state != INPCB_STATE_DEAD) {
+               if (inp->inp_gencnt <= gencnt && 
+                       inp->inp_state != INPCB_STATE_DEAD) {

                        struct xtcpcb xt;
                        caddr_t inp_ppcb;
 
                        struct xtcpcb xt;
                        caddr_t inp_ppcb;
 


@@ -1334,7 +1486,8 @@ tcp_pcblist SYSCTL_HANDLER_ARGS
                        inpcb_to_compat(inp, &xt.xt_inp);
                        inp_ppcb = inp->inp_ppcb;
                        if (inp_ppcb != NULL) {
                        inpcb_to_compat(inp, &xt.xt_inp);
                        inp_ppcb = inp->inp_ppcb;
                        if (inp_ppcb != NULL) {

-                               tcpcb_to_otcpcb((struct tcpcb *)inp_ppcb,
+                               tcpcb_to_otcpcb(
+                                   (struct tcpcb *)(void *)inp_ppcb,

                                    &xt.xt_tp);
                        } else {
                                bzero((char *) &xt.xt_tp, sizeof xt.xt_tp);
                                    &xt.xt_tp);
                        } else {
                                bzero((char *) &xt.xt_tp, sizeof xt.xt_tp);


@@ -1360,14 +1513,13 @@ tcp_pcblist SYSCTL_HANDLER_ARGS
                error = SYSCTL_OUT(req, &xig, sizeof xig);
        }
        FREE(inp_list, M_TEMP);
                error = SYSCTL_OUT(req, &xig, sizeof xig);
        }
        FREE(inp_list, M_TEMP);

-       lck_rw_done(tcbinfo.mtx);
+       lck_rw_done(tcbinfo.ipi_lock);

        return error;
 }
 
 SYSCTL_PROC(_net_inet_tcp, TCPCTL_PCBLIST, pcblist, CTLFLAG_RD | CTLFLAG_LOCKED, 0, 0,
            tcp_pcblist, "S,xtcpcb", "List of active TCP connections");
 
        return error;
 }
 
 SYSCTL_PROC(_net_inet_tcp, TCPCTL_PCBLIST, pcblist, CTLFLAG_RD | CTLFLAG_LOCKED, 0, 0,
            tcp_pcblist, "S,xtcpcb", "List of active TCP connections");
 

-#if !CONFIG_EMBEDDED

 
 static void
 tcpcb_to_xtcpcb64(struct tcpcb *tp, struct xtcpcb64 *otp)
 
 static void
 tcpcb_to_xtcpcb64(struct tcpcb *tp, struct xtcpcb64 *otp)


@@ -1424,7 +1576,7 @@ tcpcb_to_xtcpcb64(struct tcpcb *tp, struct xtcpcb64 *otp)
         otp->snd_recover = tp->snd_recover;
         otp->snd_cwnd_prev = tp->snd_cwnd_prev;
         otp->snd_ssthresh_prev = tp->snd_ssthresh_prev;
         otp->snd_recover = tp->snd_recover;
         otp->snd_cwnd_prev = tp->snd_cwnd_prev;
         otp->snd_ssthresh_prev = tp->snd_ssthresh_prev;

-        otp->t_badrxtwin = tp->t_badrxtwin;
+        otp->t_badrxtwin = 0;

 }
 
 
 }
 
 


@@ -1432,27 +1584,27 @@ static int
 tcp_pcblist64 SYSCTL_HANDLER_ARGS
 {
 #pragma unused(oidp, arg1, arg2)
 tcp_pcblist64 SYSCTL_HANDLER_ARGS
 {
 #pragma unused(oidp, arg1, arg2)

-        int error, i, n;
+        int error, i = 0, n;

         struct inpcb *inp, **inp_list;
         struct inpcb *inp, **inp_list;

+       struct tcpcb *tp;

         inp_gen_t gencnt;
         struct xinpgen xig;
         inp_gen_t gencnt;
         struct xinpgen xig;

-        int slot;

 
         /*
          * The process of preparing the TCB list is too time-consuming and
          * resource-intensive to repeat twice on every request.
          */
 
         /*
          * The process of preparing the TCB list is too time-consuming and
          * resource-intensive to repeat twice on every request.
          */

-        lck_rw_lock_shared(tcbinfo.mtx);
+        lck_rw_lock_shared(tcbinfo.ipi_lock);

         if (req->oldptr == USER_ADDR_NULL) {
                 n = tcbinfo.ipi_count;
                 req->oldidx = 2 * (sizeof xig)
                         + (n + n/8) * sizeof(struct xtcpcb64);
         if (req->oldptr == USER_ADDR_NULL) {
                 n = tcbinfo.ipi_count;
                 req->oldidx = 2 * (sizeof xig)
                         + (n + n/8) * sizeof(struct xtcpcb64);

-                lck_rw_done(tcbinfo.mtx);
+                lck_rw_done(tcbinfo.ipi_lock);

                 return 0;
         }
 
         if (req->newptr != USER_ADDR_NULL) {
                 return 0;
         }
 
         if (req->newptr != USER_ADDR_NULL) {

-                lck_rw_done(tcbinfo.mtx);
+                lck_rw_done(tcbinfo.ipi_lock);

                 return EPERM;
         }
 
                 return EPERM;
         }
 


@@ -1469,41 +1621,36 @@ tcp_pcblist64 SYSCTL_HANDLER_ARGS
         xig.xig_sogen = so_gencnt;
         error = SYSCTL_OUT(req, &xig, sizeof xig);
         if (error) {
         xig.xig_sogen = so_gencnt;
         error = SYSCTL_OUT(req, &xig, sizeof xig);
         if (error) {

-                lck_rw_done(tcbinfo.mtx);
+                lck_rw_done(tcbinfo.ipi_lock);

                 return error;
         }
         /*
          * We are done if there is no pcb
          */
         if (n == 0) {
                 return error;
         }
         /*
          * We are done if there is no pcb
          */
         if (n == 0) {

-                lck_rw_done(tcbinfo.mtx);
+                lck_rw_done(tcbinfo.ipi_lock);

                 return 0;
         }
 
         inp_list = _MALLOC(n * sizeof *inp_list, M_TEMP, M_WAITOK);
         if (inp_list == 0) {
                 return 0;
         }
 
         inp_list = _MALLOC(n * sizeof *inp_list, M_TEMP, M_WAITOK);
         if (inp_list == 0) {

-                lck_rw_done(tcbinfo.mtx);
+                lck_rw_done(tcbinfo.ipi_lock);

                 return ENOMEM;
         }
 
                 return ENOMEM;
         }
 

-        for (inp = LIST_FIRST(tcbinfo.listhead), i = 0; inp && i < n;
-             inp = LIST_NEXT(inp, inp_list)) {
-#ifdef __APPLE__
-                if (inp->inp_gencnt <= gencnt && inp->inp_state != INPCB_STATE_DEAD)
-#else
-                if (inp->inp_gencnt <= gencnt && !prison_xinpcb(req->p, inp))
-#endif
+       LIST_FOREACH(inp, tcbinfo.ipi_listhead, inp_list) {
+                if (inp->inp_gencnt <= gencnt && 
+                       inp->inp_state != INPCB_STATE_DEAD)

                         inp_list[i++] = inp;
                         inp_list[i++] = inp;

+               if (i >= n) break;

         }
 
         }
 

-        for (slot = 0; slot < N_TIME_WAIT_SLOTS; slot++) {
-                struct inpcb *inpnxt;
-
-                for (inp = time_wait_slots[slot].lh_first; inp && i < n; inp = inpnxt) {
-                        inpnxt = inp->inp_list.le_next;
-                        if (inp->inp_gencnt <= gencnt && inp->inp_state != INPCB_STATE_DEAD)
-                                inp_list[i++] = inp;
-                }
+       TAILQ_FOREACH(tp, &tcp_tw_tailq, t_twentry) {
+               inp = tp->t_inpcb;
+               if (inp->inp_gencnt <= gencnt && 
+                       inp->inp_state != INPCB_STATE_DEAD)
+                       inp_list[i++] = inp;
+               if (i >= n) break;

         }
 
         n = i;
         }
 
         n = i;


@@ -1541,14 +1688,13 @@ tcp_pcblist64 SYSCTL_HANDLER_ARGS
                        error = SYSCTL_OUT(req, &xig, sizeof xig);
         }
         FREE(inp_list, M_TEMP);
                        error = SYSCTL_OUT(req, &xig, sizeof xig);
         }
         FREE(inp_list, M_TEMP);

-        lck_rw_done(tcbinfo.mtx);
+        lck_rw_done(tcbinfo.ipi_lock);

         return error;
 }
 
 SYSCTL_PROC(_net_inet_tcp, OID_AUTO, pcblist64, CTLFLAG_RD | CTLFLAG_LOCKED, 0, 0,
             tcp_pcblist64, "S,xtcpcb64", "List of active TCP connections");
 
         return error;
 }
 
 SYSCTL_PROC(_net_inet_tcp, OID_AUTO, pcblist64, CTLFLAG_RD | CTLFLAG_LOCKED, 0, 0,
             tcp_pcblist64, "S,xtcpcb64", "List of active TCP connections");
 

-#endif /* !CONFIG_EMBEDDED */

 
 static int
 tcp_pcblist_n SYSCTL_HANDLER_ARGS
 
 static int
 tcp_pcblist_n SYSCTL_HANDLER_ARGS


@@ -1566,6 +1712,25 @@ SYSCTL_PROC(_net_inet_tcp, OID_AUTO, pcblist_n, CTLFLAG_RD | CTLFLAG_LOCKED, 0,
             tcp_pcblist_n, "S,xtcpcb_n", "List of active TCP connections");
 
 
             tcp_pcblist_n, "S,xtcpcb_n", "List of active TCP connections");
 
 

+__private_extern__ void
+tcp_get_ports_used(uint32_t ifindex, int protocol, uint32_t wildcardok,
+    bitstr_t *bitfield)
+{
+       inpcb_get_ports_used(ifindex, protocol, wildcardok, bitfield, &tcbinfo);
+}
+
+__private_extern__ uint32_t
+tcp_count_opportunistic(unsigned int ifindex, u_int32_t flags)
+{
+       return inpcb_count_opportunistic(ifindex, &tcbinfo, flags);
+}
+
+__private_extern__ uint32_t
+tcp_find_anypcb_byaddr(struct ifaddr *ifa)
+{
+       return inpcb_find_anypcb_byaddr(ifa, &tcbinfo);
+}
+

 void
 tcp_ctlinput(cmd, sa, vip)
        int cmd;
 void
 tcp_ctlinput(cmd, sa, vip)
        int cmd;


@@ -1574,16 +1739,13 @@ tcp_ctlinput(cmd, sa, vip)
 {
        tcp_seq icmp_tcp_seq;
        struct ip *ip = vip;
 {
        tcp_seq icmp_tcp_seq;
        struct ip *ip = vip;

-       struct tcphdr *th;

        struct in_addr faddr;
        struct inpcb *inp;
        struct tcpcb *tp;
        struct in_addr faddr;
        struct inpcb *inp;
        struct tcpcb *tp;

-       
-       void (*notify)(struct inpcb *, int) = tcp_notify;

 
 

-       struct icmp *icp;
+       void (*notify)(struct inpcb *, int) = tcp_notify;

 
 

-       faddr = ((struct sockaddr_in *)sa)->sin_addr;
+       faddr = ((struct sockaddr_in *)(void *)sa)->sin_addr;

        if (sa->sa_family != AF_INET || faddr.s_addr == INADDR_ANY)
                return;
 
        if (sa->sa_family != AF_INET || faddr.s_addr == INADDR_ANY)
                return;
 


@@ -1603,19 +1765,22 @@ tcp_ctlinput(cmd, sa, vip)
        else if ((unsigned)cmd > PRC_NCMDS || inetctlerrmap[cmd] == 0)
                return;
        if (ip) {
        else if ((unsigned)cmd > PRC_NCMDS || inetctlerrmap[cmd] == 0)
                return;
        if (ip) {

-               icp = (struct icmp *)((caddr_t)ip
-                               - offsetof(struct icmp, icmp_ip));
-               th = (struct tcphdr *)((caddr_t)ip 
-                              + (IP_VHL_HL(ip->ip_vhl) << 2));
-               inp = in_pcblookup_hash(&tcbinfo, faddr, th->th_dport,
-                   ip->ip_src, th->th_sport, 0, NULL);
+               struct tcphdr th;
+               struct icmp *icp;
+
+               icp = (struct icmp *)(void *)
+                   ((caddr_t)ip - offsetof(struct icmp, icmp_ip));
+               bcopy(((caddr_t)ip + (IP_VHL_HL(ip->ip_vhl) << 2)),
+                   &th, sizeof (th));
+               inp = in_pcblookup_hash(&tcbinfo, faddr, th.th_dport,
+                   ip->ip_src, th.th_sport, 0, NULL);

                if (inp != NULL && inp->inp_socket != NULL) {
                        tcp_lock(inp->inp_socket, 1, 0);
                        if (in_pcb_checkstate(inp, WNT_RELEASE, 1) == WNT_STOPUSING) {
                                tcp_unlock(inp->inp_socket, 1, 0);
                                return;
                        }
                if (inp != NULL && inp->inp_socket != NULL) {
                        tcp_lock(inp->inp_socket, 1, 0);
                        if (in_pcb_checkstate(inp, WNT_RELEASE, 1) == WNT_STOPUSING) {
                                tcp_unlock(inp->inp_socket, 1, 0);
                                return;
                        }

-                       icmp_tcp_seq = htonl(th->th_seq);
+                       icmp_tcp_seq = htonl(th.th_seq);

                        tp = intotcpcb(inp);
                        if (SEQ_GEQ(icmp_tcp_seq, tp->snd_una) &&
                            SEQ_LT(icmp_tcp_seq, tp->snd_max)) {
                        tp = intotcpcb(inp);
                        if (SEQ_GEQ(icmp_tcp_seq, tp->snd_una) &&
                            SEQ_LT(icmp_tcp_seq, tp->snd_max)) {


@@ -1809,7 +1974,7 @@ tcp_new_isn(tp)
        if (((tp->t_state == TCPS_LISTEN) || (tp->t_state == TCPS_TIME_WAIT))
           && tcp_strict_rfc1948 == 0)
 #ifdef __APPLE__
        if (((tp->t_state == TCPS_LISTEN) || (tp->t_state == TCPS_TIME_WAIT))
           && tcp_strict_rfc1948 == 0)
 #ifdef __APPLE__

-               return random();
+               return RandomULong();

 #else
                return arc4random();
 #endif
 #else
                return arc4random();
 #endif


@@ -1954,6 +2119,9 @@ tcp_mtudisc(
                    (tp->t_flags & TF_RCVD_TSTMP) == TF_RCVD_TSTMP)
                        mss -= TCPOLEN_TSTAMP_APPA;
 
                    (tp->t_flags & TF_RCVD_TSTMP) == TF_RCVD_TSTMP)
                        mss -= TCPOLEN_TSTAMP_APPA;
 

+#if MPTCP
+               mss -= mptcp_adj_mss(tp, TRUE);
+#endif

                if (so->so_snd.sb_hiwat < mss)
                        mss = so->so_snd.sb_hiwat;
 
                if (so->so_snd.sb_hiwat < mss)
                        mss = so->so_snd.sb_hiwat;
 


@@ -1993,15 +2161,19 @@ tcp_rtlookup(inp, input_ifscope)
        if ((rt = ro->ro_rt) != NULL)
                RT_LOCK(rt);
 
        if ((rt = ro->ro_rt) != NULL)
                RT_LOCK(rt);
 

-       if (rt == NULL || !(rt->rt_flags & RTF_UP) ||
-           rt->generation_id != route_generation) {
+       if (ROUTE_UNUSABLE(ro)) {
+               if (rt != NULL) {
+                       RT_UNLOCK(rt);
+                       rt = NULL;
+               }
+               ROUTE_RELEASE(ro);

                /* No route yet, so try to acquire one */
                if (inp->inp_faddr.s_addr != INADDR_ANY) {
                        unsigned int ifscope;
 
                        ro->ro_dst.sa_family = AF_INET;
                        ro->ro_dst.sa_len = sizeof(struct sockaddr_in);
                /* No route yet, so try to acquire one */
                if (inp->inp_faddr.s_addr != INADDR_ANY) {
                        unsigned int ifscope;
 
                        ro->ro_dst.sa_family = AF_INET;
                        ro->ro_dst.sa_len = sizeof(struct sockaddr_in);

-                       ((struct sockaddr_in *) &ro->ro_dst)->sin_addr =
+                       ((struct sockaddr_in *)(void *)&ro->ro_dst)->sin_addr =

                                inp->inp_faddr;
 
                        /*
                                inp->inp_faddr;
 
                        /*


@@ -2012,15 +2184,15 @@ tcp_rtlookup(inp, input_ifscope)
                         * input_ifscope is IFSCOPE_NONE).
                         */
                        ifscope = (inp->inp_flags & INP_BOUND_IF) ?
                         * input_ifscope is IFSCOPE_NONE).
                         */
                        ifscope = (inp->inp_flags & INP_BOUND_IF) ?

-                           inp->inp_boundif : input_ifscope;
+                           inp->inp_boundifp->if_index : input_ifscope;

 
 

-                       if (rt != NULL)
-                               RT_UNLOCK(rt);

                        rtalloc_scoped(ro, ifscope);
                        if ((rt = ro->ro_rt) != NULL)
                                RT_LOCK(rt);
                }
        }
                        rtalloc_scoped(ro, ifscope);
                        if ((rt = ro->ro_rt) != NULL)
                                RT_LOCK(rt);
                }
        }

+       if (rt != NULL)
+               RT_LOCK_ASSERT_HELD(rt);

 
        /*
         * Update MTU discovery determination. Don't do it if:
 
        /*
         * Update MTU discovery determination. Don't do it if:


@@ -2053,6 +2225,14 @@ tcp_rtlookup(inp, input_ifscope)
                tcp_set_tso(tp, rt->rt_ifp);
        }
 
                tcp_set_tso(tp, rt->rt_ifp);
        }
 

+       /* Note if the peer is local */
+       if (rt != NULL &&
+               (rt->rt_gateway->sa_family == AF_LINK ||
+               rt->rt_ifp->if_flags & IFF_LOOPBACK ||
+               in_localaddr(inp->inp_faddr))) {
+               tp->t_flags |= TF_LOCAL;
+       }
+    

        /*
         * Caller needs to call RT_UNLOCK(rt).
         */
        /*
         * Caller needs to call RT_UNLOCK(rt).
         */


@@ -2075,8 +2255,12 @@ tcp_rtlookup6(inp, input_ifscope)
        if ((rt = ro6->ro_rt) != NULL)
                RT_LOCK(rt);
 
        if ((rt = ro6->ro_rt) != NULL)
                RT_LOCK(rt);
 

-       if (rt == NULL || !(rt->rt_flags & RTF_UP) ||
-           rt->generation_id != route_generation) {
+       if (ROUTE_UNUSABLE(ro6)) {
+               if (rt != NULL) {
+                       RT_UNLOCK(rt);
+                       rt = NULL;
+               }
+               ROUTE_RELEASE(ro6);

                /* No route yet, so try to acquire one */
                if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
                        struct sockaddr_in6 *dst6;
                /* No route yet, so try to acquire one */
                if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
                        struct sockaddr_in6 *dst6;


@@ -2095,15 +2279,16 @@ tcp_rtlookup6(inp, input_ifscope)
                         * input_ifscope is IFSCOPE_NONE).
                         */
                        ifscope = (inp->inp_flags & INP_BOUND_IF) ?
                         * input_ifscope is IFSCOPE_NONE).
                         */
                        ifscope = (inp->inp_flags & INP_BOUND_IF) ?

-                           inp->inp_boundif : input_ifscope;
+                           inp->inp_boundifp->if_index : input_ifscope;

 
 

-                       if (rt != NULL)
-                               RT_UNLOCK(rt);

                        rtalloc_scoped((struct route *)ro6, ifscope);
                        if ((rt = ro6->ro_rt) != NULL)
                                RT_LOCK(rt);
                }
        }
                        rtalloc_scoped((struct route *)ro6, ifscope);
                        if ((rt = ro6->ro_rt) != NULL)
                                RT_LOCK(rt);
                }
        }

+       if (rt != NULL)
+               RT_LOCK_ASSERT_HELD(rt);
+

        /*
         * Update path MTU Discovery determination
         * while looking up the route:
        /*
         * Update path MTU Discovery determination
         * while looking up the route:


@@ -2144,6 +2329,15 @@ tcp_rtlookup6(inp, input_ifscope)
                tcp_set_tso(tp, rt->rt_ifp);
        }
 
                tcp_set_tso(tp, rt->rt_ifp);
        }
 

+       /* Note if the peer is local */
+       if (rt != NULL &&
+               (IN6_IS_ADDR_LOOPBACK(&inp->in6p_faddr) ||
+               IN6_IS_ADDR_LINKLOCAL(&inp->in6p_faddr) ||
+               rt->rt_gateway->sa_family == AF_LINK ||
+               in6_localaddr(&inp->in6p_faddr))) {
+               tp->t_flags |= TF_LOCAL;
+       }
+

        /*
         * Caller needs to call RT_UNLOCK(rt).
         */
        /*
         * Caller needs to call RT_UNLOCK(rt).
         */


@@ -2175,7 +2369,7 @@ ipsec_hdrsiz_tcp(tp)
 #if INET6
        if ((inp->inp_vflag & INP_IPV6) != 0) {
                ip6 = mtod(m, struct ip6_hdr *);
 #if INET6
        if ((inp->inp_vflag & INP_IPV6) != 0) {
                ip6 = mtod(m, struct ip6_hdr *);

-               th = (struct tcphdr *)(ip6 + 1);
+               th = (struct tcphdr *)(void *)(ip6 + 1);

                m->m_pkthdr.len = m->m_len =
                        sizeof(struct ip6_hdr) + sizeof(struct tcphdr);
                tcp_fillheaders(tp, ip6, th);
                m->m_pkthdr.len = m->m_len =
                        sizeof(struct ip6_hdr) + sizeof(struct tcphdr);
                tcp_fillheaders(tp, ip6, th);


@@ -2282,9 +2476,11 @@ tcp_unlock(struct socket *so, int refcount, void *lr)
                lr_saved = lr;
 
 #ifdef MORE_TCPLOCK_DEBUG
                lr_saved = lr;
 
 #ifdef MORE_TCPLOCK_DEBUG

-       printf("tcp_unlock: so=%p sopcb=%p lock=%p ref=%x lr=%p\n",
-           so, so->so_pcb, &((struct inpcb *)so->so_pcb)->inpcb_mtx,
-           so->so_usecount, lr_saved);
+       printf("tcp_unlock: so=0x%llx sopcb=0x%llx lock=0x%llx ref=%x "
+           "lr=0x%llx\n", (uint64_t)VM_KERNEL_ADDRPERM(so),
+           (uint64_t)VM_KERNEL_ADDRPERM(so->so_pcb),
+           (uint64_t)VM_KERNEL_ADDRPERM(&(sotoinpcb(so)->inpcb_mtx)),
+           so->so_usecount, (uint64_t)VM_KERNEL_ADDRPERM(lr_saved));

 #endif
        if (refcount)
                so->so_usecount--;
 #endif
        if (refcount)
                so->so_usecount--;


@@ -2328,13 +2524,55 @@ tcp_getlock(
        }
 }
 
        }
 }
 

+/* Determine if we can grow the recieve socket buffer to avoid sending
+ * a zero window update to the peer. We allow even socket buffers that 
+ * have fixed size (set by the application) to grow if the resource
+ * constraints are met. They will also be trimmed after the application
+ * reads data.
+ */
+static void
+tcp_sbrcv_grow_rwin(struct tcpcb *tp, struct sockbuf *sb) {
+       u_int32_t rcvbufinc = tp->t_maxseg << tcp_autorcvbuf_inc_shift;
+       u_int32_t rcvbuf = sb->sb_hiwat;
+       struct socket *so = tp->t_inpcb->inp_socket;
+
+       /*
+        * If message delivery is enabled, do not count
+        * unordered bytes in receive buffer towards hiwat
+        */
+       if (so->so_flags & SOF_ENABLE_MSGS)
+               rcvbuf = rcvbuf - so->so_msg_state->msg_uno_bytes;
+
+       if (tcp_do_autorcvbuf == 1 &&
+               tcp_cansbgrow(sb) &&
+               (tp->t_flags & TF_SLOWLINK) == 0 &&
+               (rcvbuf - sb->sb_cc) < rcvbufinc &&
+               (rcvbuf < tcp_autorcvbuf_max)) {
+               sbreserve(sb, (sb->sb_hiwat + rcvbufinc));
+       }
+}
+

 int32_t
 tcp_sbspace(struct tcpcb *tp)
 {
        struct sockbuf *sb = &tp->t_inpcb->inp_socket->so_rcv;
 int32_t
 tcp_sbspace(struct tcpcb *tp)
 {
        struct sockbuf *sb = &tp->t_inpcb->inp_socket->so_rcv;

-       int32_t space, newspace;
+       u_int32_t rcvbuf = sb->sb_hiwat;
+       int32_t space;
+       struct socket *so = tp->t_inpcb->inp_socket;

 
 

-       space =  ((int32_t) imin((sb->sb_hiwat - sb->sb_cc), 
+       /*
+        * If message delivery is enabled, do not count
+        * unordered bytes in receive buffer towards hiwat mark.
+        * This value is used to return correct rwnd that does
+        * not reflect the extra unordered bytes added to the
+        * receive socket buffer.
+        */
+       if (so->so_flags & SOF_ENABLE_MSGS)
+               rcvbuf = rcvbuf - so->so_msg_state->msg_uno_bytes;
+
+       tcp_sbrcv_grow_rwin(tp, sb);
+
+       space =  ((int32_t) imin((rcvbuf - sb->sb_cc),

                (sb->sb_mbmax - sb->sb_mbcnt)));
        if (space < 0) 
                space = 0;
                (sb->sb_mbmax - sb->sb_mbcnt)));
        if (space < 0) 
                space = 0;


@@ -2352,37 +2590,29 @@ tcp_sbspace(struct tcpcb *tp)
        if (((tp->t_flags & TF_SLOWLINK) != 0) && slowlink_wsize > 0 )  
                return imin(space, slowlink_wsize);
 
        if (((tp->t_flags & TF_SLOWLINK) != 0) && slowlink_wsize > 0 )  
                return imin(space, slowlink_wsize);
 

-       /*
-        * Check for ressources constraints before over-ajusting the amount of space we can
-        * advertise in the TCP window size updates.
-        */
-
-       if (sbspace_factor && (tp->t_inpcb->inp_pcbinfo->ipi_count < tcp_sockthreshold) &&
-                   (total_mb_cnt / 8) < (mbstat.m_clusters / sbspace_factor)) {
-               if (space < (int32_t)(sb->sb_maxused - sb->sb_cc)) {/* make sure we don't constrain the window if we have enough ressources */
-                       space = (int32_t) imax((sb->sb_maxused - sb->sb_cc), tp->rcv_maxbyps);
-               }
-               newspace = (int32_t) imax(((int32_t)sb->sb_maxused - sb->sb_cc), (int32_t)tp->rcv_maxbyps);
-
-               if (newspace > space)
-                       space = newspace;
-       }

        return space;
 }
 /*
  * Checks TCP Segment Offloading capability for a given connection and interface pair.
  */
 void
        return space;
 }
 /*
  * Checks TCP Segment Offloading capability for a given connection and interface pair.
  */
 void

-tcp_set_tso(tp, ifp)
-       struct tcpcb *tp;
-       struct ifnet *ifp;
+tcp_set_tso(struct tcpcb *tp, struct ifnet *ifp)

 {
 {

+#if MPTCP
+       /*
+        * We can't use TSO if this tcpcb belongs to an MPTCP session.
+        */
+       if (tp->t_mpflags & TMPF_MPTCP_TRUE) {
+               tp->t_flags &= ~TF_TSO;
+               return;
+       }
+#endif

 #if INET6
        struct inpcb *inp = tp->t_inpcb;
        int isipv6 = (inp->inp_vflag & INP_IPV6) != 0;
 
        if (isipv6) {
 #if INET6
        struct inpcb *inp = tp->t_inpcb;
        int isipv6 = (inp->inp_vflag & INP_IPV6) != 0;
 
        if (isipv6) {

-               if (ifp && ifp->if_hwassist & IFNET_TSO_IPV6) {
+               if (ifp && (ifp->if_hwassist & IFNET_TSO_IPV6)) {

                        tp->t_flags |= TF_TSO;
                        if (ifp->if_tso_v6_mtu != 0) 
                                tp->tso_max_segment_size = ifp->if_tso_v6_mtu;
                        tp->t_flags |= TF_TSO;
                        if (ifp->if_tso_v6_mtu != 0) 
                                tp->tso_max_segment_size = ifp->if_tso_v6_mtu;


@@ -2395,7 +2625,7 @@ tcp_set_tso(tp, ifp)
 #endif /* INET6 */
 
        {
 #endif /* INET6 */
 
        {

-               if (ifp && ifp->if_hwassist & IFNET_TSO_IPV4) {
+               if (ifp && (ifp->if_hwassist & IFNET_TSO_IPV4)) {

                        tp->t_flags |= TF_TSO;
                        if (ifp->if_tso_v4_mtu != 0) 
                                tp->tso_max_segment_size = ifp->if_tso_v4_mtu;
                        tp->t_flags |= TF_TSO;
                        if (ifp->if_tso_v4_mtu != 0) 
                                tp->tso_max_segment_size = ifp->if_tso_v4_mtu;


@@ -2424,8 +2654,16 @@ calculate_tcp_clock()
        struct timeval now, hold_now;
        uint32_t incr = 0;
 
        struct timeval now, hold_now;
        uint32_t incr = 0;
 

-       timevaladd(&tv, &interval);

        microuptime(&now);
        microuptime(&now);

+
+       /*
+        * Update coarse-grained networking timestamp (in sec.); the idea
+        * is to update the counter returnable via net_uptime() when
+        * we read time.
+        */
+       net_update_uptime_secs(now.tv_sec);
+
+       timevaladd(&tv, &interval);

        if (timevalcmp(&now, &tv, >)) {
                /* time to update the clock */
                lck_spin_lock(tcp_uptime_lock);
        if (timevalcmp(&now, &tv, >)) {
                /* time to update the clock */
                lck_spin_lock(tcp_uptime_lock);


@@ -2451,4 +2689,57 @@ calculate_tcp_clock()
         return;
 }
 
         return;
 }
 

+/* Compute receive window scaling that we are going to request 
+ * for this connection based on  sb_hiwat. Try to leave some 
+ * room to potentially increase the window size upto a maximum 
+ * defined by the constant tcp_autorcvbuf_max.
+ */
+void
+tcp_set_max_rwinscale(struct tcpcb *tp, struct socket *so) {
+       u_int32_t maxsockbufsize;
+
+       tp->request_r_scale = max(tcp_win_scale, tp->request_r_scale);
+       maxsockbufsize = ((so->so_rcv.sb_flags & SB_USRSIZE) != 0) ?
+               so->so_rcv.sb_hiwat : tcp_autorcvbuf_max;
+
+       while (tp->request_r_scale < TCP_MAX_WINSHIFT &&
+               (TCP_MAXWIN << tp->request_r_scale) < maxsockbufsize)
+               tp->request_r_scale++;
+       tp->request_r_scale = min(tp->request_r_scale, TCP_MAX_WINSHIFT);
+
+}
+
+int
+tcp_notsent_lowat_check(struct socket *so) {
+       struct inpcb *inp = sotoinpcb(so);
+       struct tcpcb *tp = NULL;
+       int notsent = 0;
+       if (inp != NULL) {
+               tp = intotcpcb(inp);
+       }
+
+       notsent = so->so_snd.sb_cc - 
+               (tp->snd_nxt - tp->snd_una);
+
+       /* When we send a FIN or SYN, not_sent can be negative.
+        * In that case also we need to send a write event to the 
+        * process if it is waiting. In the FIN case, it will
+        * get an error from send because cantsendmore will be set.
+        */
+       if (notsent <= tp->t_notsent_lowat) {
+               return(1);
+       }
+
+       /* When Nagle's algorithm is not disabled, it is better
+        * to wakeup the client until there is atleast one
+        * maxseg of data to write.
+        */
+       if ((tp->t_flags & TF_NODELAY) == 0 && 
+               notsent > 0 && notsent < tp->t_maxseg) {
+               return(1);
+       }
+       return(0);
+}
+
+

 /* DSEP Review Done pl-20051213-v02 @3253,@3391,@3400 */
 /* DSEP Review Done pl-20051213-v02 @3253,@3391,@3400 */