xnu-1504.15.3.tar.gz

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

diff --git a/bsd/netinet/tcp_timer.c b/bsd/netinet/tcp_timer.c
index 3d92a0d8203ff8a04d31262e820ab52a182055ab..fd66419d09b60ab58fd586bdb5dac36e931a2044 100644 (file)
--- a/bsd/netinet/tcp_timer.c
+++ b/bsd/netinet/tcp_timer.c
@@ -1,16 +1,19 @@
 /*
 /*

- * Copyright (c) 2000 Apple Computer, Inc. All rights reserved.
+ * Copyright (c) 2000-2008 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
  * 
  * 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
  * 
  * The Original Code and all software distributed under the License are
  * distributed on an 'AS IS' basis, WITHOUT WARRANTY OF ANY KIND, EITHER


@@ -20,7 +23,7 @@
  * Please see the License for the specific language governing rights and
  * limitations under the License.
  * 
  * 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, 1990, 1993, 1995
  */
 /*
  * Copyright (c) 1982, 1986, 1988, 1990, 1993, 1995


@@ -67,6 +70,8 @@
 #include <sys/socket.h>
 #include <sys/socketvar.h>
 #include <sys/protosw.h>
 #include <sys/socket.h>
 #include <sys/socketvar.h>
 #include <sys/protosw.h>

+#include <sys/domain.h>
+#include <kern/locks.h>

 
 #include <kern/cpu_number.h>   /* before tcp_seq.h, for tcp_random18() */
 
 
 #include <kern/cpu_number.h>   /* before tcp_seq.h, for tcp_random18() */
 


@@ -84,28 +89,45 @@
 #include <netinet/tcp_seq.h>
 #include <netinet/tcp_timer.h>
 #include <netinet/tcp_var.h>
 #include <netinet/tcp_seq.h>
 #include <netinet/tcp_timer.h>
 #include <netinet/tcp_var.h>

+#if INET6
+#include <netinet6/tcp6_var.h>
+#endif

 #include <netinet/tcpip.h>
 #if TCPDEBUG
 #include <netinet/tcp_debug.h>
 #endif
 #include <sys/kdebug.h>
 
 #include <netinet/tcpip.h>
 #if TCPDEBUG
 #include <netinet/tcp_debug.h>
 #endif
 #include <sys/kdebug.h>
 

+extern void postevent(struct socket *, struct sockbuf *,
+                                               int);

 #define DBG_FNC_TCP_FAST       NETDBG_CODE(DBG_NETTCP, (5 << 8))
 #define DBG_FNC_TCP_SLOW       NETDBG_CODE(DBG_NETTCP, (5 << 8) | 1)
 
 #define DBG_FNC_TCP_FAST       NETDBG_CODE(DBG_NETTCP, (5 << 8))
 #define DBG_FNC_TCP_SLOW       NETDBG_CODE(DBG_NETTCP, (5 << 8) | 1)
 

+static int     background_io_trigger = 5;
+SYSCTL_INT(_net_inet_tcp, OID_AUTO, background_io_trigger, CTLFLAG_RW,
+    &background_io_trigger, 0, "Background IO Trigger Setting");
+
+/*
+ * NOTE - WARNING
+ *
+ *
+ * 
+ *
+ */

 static int
 sysctl_msec_to_ticks SYSCTL_HANDLER_ARGS
 {
 static int
 sysctl_msec_to_ticks SYSCTL_HANDLER_ARGS
 {

+#pragma unused(arg1, arg2)

        int error, s, tt;
 
        tt = *(int *)oidp->oid_arg1;
        int error, s, tt;
 
        tt = *(int *)oidp->oid_arg1;

-       s = tt * 1000 / hz;
+       s = tt * 1000 / TCP_RETRANSHZ;;

 
        error = sysctl_handle_int(oidp, &s, 0, req);
        if (error || !req->newptr)
                return (error);
 
 
        error = sysctl_handle_int(oidp, &s, 0, req);
        if (error || !req->newptr)
                return (error);
 

-       tt = s * hz / 1000;
+       tt = s * TCP_RETRANSHZ / 1000;

        if (tt < 1)
                return (EINVAL);
 
        if (tt < 1)
                return (EINVAL);
 


@@ -125,11 +147,6 @@ int        tcp_keepintvl;
 SYSCTL_PROC(_net_inet_tcp, TCPCTL_KEEPINTVL, keepintvl, CTLTYPE_INT|CTLFLAG_RW,
     &tcp_keepintvl, 0, sysctl_msec_to_ticks, "I", "");
 
 SYSCTL_PROC(_net_inet_tcp, TCPCTL_KEEPINTVL, keepintvl, CTLTYPE_INT|CTLFLAG_RW,
     &tcp_keepintvl, 0, sysctl_msec_to_ticks, "I", "");
 

-int    tcp_delacktime;
-SYSCTL_PROC(_net_inet_tcp, TCPCTL_DELACKTIME, delacktime,
-    CTLTYPE_INT|CTLFLAG_RW, &tcp_delacktime, 0, sysctl_msec_to_ticks, "I",
-    "Time before a delayed ACK is sent");
- 

 int    tcp_msl;
 SYSCTL_PROC(_net_inet_tcp, OID_AUTO, msl, CTLTYPE_INT|CTLFLAG_RW,
     &tcp_msl, 0, sysctl_msec_to_ticks, "I", "Maximum segment lifetime");
 int    tcp_msl;
 SYSCTL_PROC(_net_inet_tcp, OID_AUTO, msl, CTLTYPE_INT|CTLFLAG_RW,
     &tcp_msl, 0, sysctl_msec_to_ticks, "I", "Maximum segment lifetime");


@@ -138,7 +155,28 @@ static int always_keepalive = 0;
 SYSCTL_INT(_net_inet_tcp, OID_AUTO, always_keepalive, CTLFLAG_RW, 
     &always_keepalive , 0, "Assume SO_KEEPALIVE on all TCP connections");
 
 SYSCTL_INT(_net_inet_tcp, OID_AUTO, always_keepalive, CTLFLAG_RW, 
     &always_keepalive , 0, "Assume SO_KEEPALIVE on all TCP connections");
 

+/*
+ * See tcp_syn_backoff[] for interval values between SYN retransmits;
+ * the value set below defines the number of retransmits, before we
+ * disable the timestamp and window scaling options during subsequent
+ * SYN retransmits.  Setting it to 0 disables the dropping off of those
+ * two options.
+ */
+static int tcp_broken_peer_syn_rxmit_thres = 7;
+SYSCTL_INT(_net_inet_tcp, OID_AUTO, broken_peer_syn_rxmit_thres, CTLFLAG_RW,
+    &tcp_broken_peer_syn_rxmit_thres, 0, "Number of retransmitted SYNs before "
+    "TCP disables rfc1323 and rfc1644 during the rest of attempts");
+
+int    tcp_pmtud_black_hole_detect = 1 ;
+SYSCTL_INT(_net_inet_tcp, OID_AUTO, pmtud_blackhole_detection, CTLFLAG_RW,
+    &tcp_pmtud_black_hole_detect, 0, "Path MTU Discovery Black Hole Detection");
+
+int    tcp_pmtud_black_hole_mss = 1200 ;
+SYSCTL_INT(_net_inet_tcp, OID_AUTO, pmtud_blackhole_mss, CTLFLAG_RW,
+    &tcp_pmtud_black_hole_mss, 0, "Path MTU Discovery Black Hole Detection lowered MSS");
+

 static int     tcp_keepcnt = TCPTV_KEEPCNT;
 static int     tcp_keepcnt = TCPTV_KEEPCNT;

+static int     tcp_gc_done = FALSE;    /* perfromed garbage collection of "used" sockets */

        /* max idle probes */
 int    tcp_maxpersistidle;
        /* max idle time in persist */
        /* max idle probes */
 int    tcp_maxpersistidle;
        /* max idle time in persist */


@@ -147,20 +185,37 @@ int       tcp_maxidle;
 struct inpcbhead       time_wait_slots[N_TIME_WAIT_SLOTS];
 int            cur_tw_slot = 0;
 
 struct inpcbhead       time_wait_slots[N_TIME_WAIT_SLOTS];
 int            cur_tw_slot = 0;
 

-u_long         *delack_bitmask;
+u_int32_t              *delack_bitmask;
+
+void   add_to_time_wait_locked(struct tcpcb *tp);
+void   add_to_time_wait(struct tcpcb *tp) ;

 
 

+static void tcp_garbage_collect(struct inpcb *, int);

 
 

-void   add_to_time_wait(tp) 
-       struct tcpcb    *tp;
+void   add_to_time_wait_locked(struct tcpcb *tp) 

 {
        int             tw_slot;
 {
        int             tw_slot;

+    struct inpcbinfo *pcbinfo  = &tcbinfo;
+
+       /* pcb list should be locked when we get here */        
+       lck_rw_assert(pcbinfo->mtx, LCK_RW_ASSERT_EXCLUSIVE);

 
        LIST_REMOVE(tp->t_inpcb, inp_list);
 
 
        LIST_REMOVE(tp->t_inpcb, inp_list);
 

-       if (tp->t_timer[TCPT_2MSL] == 0) 
+       if (tp->t_timer[TCPT_2MSL] <= 0) 

            tp->t_timer[TCPT_2MSL] = 1;
 
            tp->t_timer[TCPT_2MSL] = 1;
 

-       tp->t_rcvtime += tp->t_timer[TCPT_2MSL] & (N_TIME_WAIT_SLOTS - 1);
+       /*
+        * Because we're pulling this pcb out of the main TCP pcb list,
+        * we need to recalculate the TCPT_2MSL timer value for tcp_slowtimo
+        * higher timer granularity.
+        */
+
+       tp->t_timer[TCPT_2MSL] = (tp->t_timer[TCPT_2MSL] / TCP_RETRANSHZ) * PR_SLOWHZ;
+       tp->t_rcvtime = (tp->t_rcvtime / TCP_RETRANSHZ) * PR_SLOWHZ;
+
+       tp->t_rcvtime += tp->t_timer[TCPT_2MSL] & (N_TIME_WAIT_SLOTS - 1); 
+

        tw_slot = (tp->t_timer[TCPT_2MSL] & (N_TIME_WAIT_SLOTS - 1)) + cur_tw_slot; 
        if (tw_slot >= N_TIME_WAIT_SLOTS)
            tw_slot -= N_TIME_WAIT_SLOTS;
        tw_slot = (tp->t_timer[TCPT_2MSL] & (N_TIME_WAIT_SLOTS - 1)) + cur_tw_slot; 
        if (tw_slot >= N_TIME_WAIT_SLOTS)
            tw_slot -= N_TIME_WAIT_SLOTS;


@@ -168,6 +223,18 @@ void       add_to_time_wait(tp)
        LIST_INSERT_HEAD(&time_wait_slots[tw_slot], tp->t_inpcb, inp_list);
 }
 
        LIST_INSERT_HEAD(&time_wait_slots[tw_slot], tp->t_inpcb, inp_list);
 }
 

+void   add_to_time_wait(struct tcpcb *tp) 
+{
+       struct inpcbinfo *pcbinfo               = &tcbinfo;
+       
+       if (!lck_rw_try_lock_exclusive(pcbinfo->mtx)) {
+               tcp_unlock(tp->t_inpcb->inp_socket, 0, 0);
+               lck_rw_lock_exclusive(pcbinfo->mtx);
+               tcp_lock(tp->t_inpcb->inp_socket, 0, 0);
+       }
+       add_to_time_wait_locked(tp);
+       lck_rw_done(pcbinfo->mtx);
+}

 
 
 
 
 
 


@@ -176,139 +243,337 @@ void    add_to_time_wait(tp)
  * Fast timeout routine for processing delayed acks
  */
 void
  * Fast timeout routine for processing delayed acks
  */
 void

-tcp_fasttimo()
+tcp_fasttimo(void *arg)

 {
 {

-    register struct inpcb *inp;
+#pragma unused(arg)
+    struct inpcb *inp;

     register struct tcpcb *tp;
     register struct tcpcb *tp;

+    struct socket *so;
+#if TCPDEBUG
+    int ostate;
+#endif

 
 
 
 

-    register u_long                    i,j;
-    register u_long                    temp_mask;
-    register u_long                    elem_base = 0;
-    struct inpcbhead                   *head;
-    int s = splnet();
+    struct inpcbinfo *pcbinfo  = &tcbinfo;

 
 

-    static
-    int delack_checked = 0;
+    int delack_done = 0;

 
     KERNEL_DEBUG(DBG_FNC_TCP_FAST | DBG_FUNC_START, 0,0,0,0,0);
 
 
     KERNEL_DEBUG(DBG_FNC_TCP_FAST | DBG_FUNC_START, 0,0,0,0,0);
 

-    if (!tcp_delack_enabled) 
-       return;
-
-    for (i=0; i < (tcbinfo.hashsize / 32); i++) {
-           if (delack_bitmask[i]) {
-               temp_mask = 1;
-               for (j=0; j < 32; j++) {
-                   if (temp_mask & delack_bitmask[i]) {
-                       head = &tcbinfo.hashbase[elem_base + j];
-                       for (inp=head->lh_first; inp != 0; inp = inp->inp_hash.le_next) {
-                           delack_checked++;
-                           if ((tp = (struct tcpcb *)inp->inp_ppcb) && (tp->t_flags & TF_DELACK)) {
-                               tp->t_flags &= ~TF_DELACK;
-                               tp->t_flags |= TF_ACKNOW;
-                               tcpstat.tcps_delack++;
-                               (void) tcp_output(tp);
-                           }
-                       }
-                   }
-                   temp_mask <<=  1;
-               }
-               delack_bitmask[i] = 0;
-           }
-           elem_base += 32;
+
+    lck_rw_lock_shared(pcbinfo->mtx);
+
+    /* Walk the list of valid tcpcbs and send ACKS on the ones with DELACK bit set */
+
+    LIST_FOREACH(inp, &tcb, inp_list) {
+
+       so = inp->inp_socket;
+
+       if (in_pcb_checkstate(inp, WNT_ACQUIRE, 0) == WNT_STOPUSING) 
+               continue;
+
+       tcp_lock(so, 1, 0);
+
+       if ((in_pcb_checkstate(inp, WNT_RELEASE,1) == WNT_STOPUSING)  && so->so_usecount == 1) {
+               tcp_unlock(so, 1, 0);
+               continue;
+       }
+
+       tp = intotcpcb(inp);
+
+       if (tp == 0 || tp->t_state == TCPS_LISTEN) {
+               tcp_unlock(so, 1, 0);
+               continue; 
+       }
+
+
+       /* Only run the retransmit timer in that case */
+       if (tp->t_timer[0] && --tp->t_timer[0] == 0) {
+               tp = tcp_timers(tp, 0);
+               if (tp == NULL)
+                       goto tpgone;
+       }
+
+       /* TCP pcb  timers following the tcp_now clock rate */
+
+       tp->t_rcvtime++;
+       tp->t_starttime++;
+       if (tp->t_rtttime)
+               tp->t_rtttime++;        
+
+       /*
+        * Process delayed acks (if enabled) according to PR_FASTHZ, not the retrans timer
+        */
+
+       if (tcp_delack_enabled && (tcp_now % (TCP_RETRANSHZ/PR_FASTHZ)) && tp->t_flags & TF_DELACK) {
+               delack_done++;
+               tp->t_flags &= ~TF_DELACK;
+               tp->t_flags |= TF_ACKNOW;
+               tcpstat.tcps_delack++;
+               tp->t_unacksegs = 0;
+               (void) tcp_output(tp);
+       }
+tpgone:
+       tcp_unlock(so, 1, 0);

     }
     }

-    KERNEL_DEBUG(DBG_FNC_TCP_FAST | DBG_FUNC_END, delack_checked,tcpstat.tcps_delack,0,0,0);
-    splx(s);
+    KERNEL_DEBUG(DBG_FNC_TCP_FAST | DBG_FUNC_END, delack_done, 0, tcpstat.tcps_delack,0,0);
+    lck_rw_done(pcbinfo->mtx);

 
 

+    tcp_now++;
+    timeout(tcp_fasttimo, 0, hz/TCP_RETRANSHZ);

 }
 
 }
 

-/*
- * Tcp protocol timeout routine called every 500 ms.
- * Updates the timers in all active tcb's and
- * causes finite state machine actions if timers expire.
- */
+static void
+tcp_garbage_collect(struct inpcb *inp, int istimewait)
+{
+       struct socket *so;
+       struct tcpcb *tp;
+
+       so = inp->inp_socket;
+       tp = intotcpcb(inp);
+
+       /*
+        * Skip if still in use or busy; it would have been more efficient
+        * if we were to test so_usecount against 0, but this isn't possible
+        * due to the current implementation of tcp_dropdropablreq() where
+        * overflow sockets that are eligible for garbage collection have
+        * their usecounts set to 1.
+        */
+       if (so->so_usecount > 1 || !lck_mtx_try_lock_spin(inp->inpcb_mtx))
+               return;
+
+       /* Check again under the lock */
+       if (so->so_usecount > 1) {
+               lck_mtx_unlock(inp->inpcb_mtx);
+               return;
+       }
+
+       /*
+        * Overflowed socket dropped from the listening queue?  Do this
+        * only if we are called to clean up the time wait slots, since
+        * tcp_dropdropablreq() considers a socket to have been fully
+        * dropped after add_to_time_wait() is finished.
+        * Also handle the case of connections getting closed by the peer while in the queue as
+        * seen with rdar://6422317
+        * 
+        */
+       if (so->so_usecount == 1 && 
+           ((istimewait && (so->so_flags & SOF_OVERFLOW)) ||
+           ((tp != NULL) && (tp->t_state == TCPS_CLOSED) && (so->so_head != NULL)
+                && ((so->so_state & (SS_INCOMP|SS_CANTSENDMORE|SS_CANTRCVMORE)) ==
+                        (SS_INCOMP|SS_CANTSENDMORE|SS_CANTRCVMORE))))) {
+
+               if (inp->inp_state != INPCB_STATE_DEAD) {
+                       /* Become a regular mutex */
+                       lck_mtx_convert_spin(inp->inpcb_mtx);
+#if INET6
+                       if (INP_CHECK_SOCKAF(so, AF_INET6))
+                               in6_pcbdetach(inp);
+                       else
+#endif /* INET6 */
+                       in_pcbdetach(inp);
+               }
+               so->so_usecount--;
+               lck_mtx_unlock(inp->inpcb_mtx);
+               return;
+       } else if (inp->inp_wantcnt != WNT_STOPUSING) {
+               lck_mtx_unlock(inp->inpcb_mtx);
+               return;
+       }
+
+       /*
+        * We get here because the PCB is no longer searchable (WNT_STOPUSING);
+        * detach (if needed) and dispose if it is dead (usecount is 0).  This
+        * covers all cases, including overflow sockets and those that are
+        * considered as "embryonic", i.e. created by sonewconn() in TCP input
+        * path, and have not yet been committed.  For the former, we reduce
+        * the usecount to 0 as done by the code above.  For the latter, the
+        * usecount would have reduced to 0 as part calling soabort() when the
+        * socket is dropped at the end of tcp_input().
+        */
+       if (so->so_usecount == 0) {
+               /* Become a regular mutex */
+               lck_mtx_convert_spin(inp->inpcb_mtx);
+               if (inp->inp_state != INPCB_STATE_DEAD) {
+#if INET6
+                       if (INP_CHECK_SOCKAF(so, AF_INET6))
+                               in6_pcbdetach(inp);
+                       else
+#endif /* INET6 */
+                       in_pcbdetach(inp);
+               }
+               in_pcbdispose(inp);
+       } else {
+               lck_mtx_unlock(inp->inpcb_mtx);
+       }
+}
+
+static int bg_cnt = 0;
+#define BG_COUNTER_MAX 3
+

 void
 void

-tcp_slowtimo()
+tcp_slowtimo(void)

 {
 {

-       register struct inpcb *ip, *ipnxt;
-       register struct tcpcb *tp;
-       register int i;
-       int s;
+       struct inpcb *inp, *nxt;
+       struct tcpcb *tp;
+       struct socket *so;
+       int i;

 #if TCPDEBUG
        int ostate;
 #endif
 #if TCPDEBUG
        int ostate;
 #endif

-#if KDEBUG
-       static int tws_checked;
+
+#if  KDEBUG
+       static int tws_checked = 0;

 #endif
 
 #endif
 

+       struct inpcbinfo *pcbinfo               = &tcbinfo;
+

        KERNEL_DEBUG(DBG_FNC_TCP_SLOW | DBG_FUNC_START, 0,0,0,0,0);
        KERNEL_DEBUG(DBG_FNC_TCP_SLOW | DBG_FUNC_START, 0,0,0,0,0);

-       s = splnet();

 
        tcp_maxidle = tcp_keepcnt * tcp_keepintvl;
 
 
        tcp_maxidle = tcp_keepcnt * tcp_keepintvl;
 

-       ip = tcb.lh_first;
-       if (ip == NULL) {
-               splx(s);
-               return;
-       }
-       /*
-        * Search through tcb's and update active timers.
-        */
-       for (; ip != NULL; ip = ipnxt) {
-               ipnxt = ip->inp_list.le_next;
-               tp = intotcpcb(ip);
-               if (tp == 0 || tp->t_state == TCPS_LISTEN)
+       lck_rw_lock_shared(pcbinfo->mtx);
+
+       bg_cnt++;
+
+       LIST_FOREACH(inp, &tcb, inp_list) {
+
+               so = inp->inp_socket;
+
+               if (in_pcb_checkstate(inp, WNT_ACQUIRE, 0) == WNT_STOPUSING) 

                        continue;
                        continue;

-               /*
-                * Bogus state when port owned by SharedIP with loopback as the
-                * only configured interface: BlueBox does not filters loopback
-                */
-               if (tp->t_state == TCP_NSTATES)
-                       continue;      

 
 

-               for (i = 0; i < TCPT_NTIMERS; i++) {
-                       if (tp->t_timer[i] && --tp->t_timer[i] == 0) {
+               tcp_lock(so, 1, 0);
+
+               if ((in_pcb_checkstate(inp, WNT_RELEASE,1) == WNT_STOPUSING)  && so->so_usecount == 1) {
+                       tcp_unlock(so, 1, 0);
+                       continue;
+               }
+               tp = intotcpcb(inp);
+               if (tp == 0 || tp->t_state == TCPS_LISTEN) {
+                       tcp_unlock(so, 1, 0);
+                       continue; 
+               }
+
+               tp = intotcpcb(inp);
+
+               if (tp == 0 || tp->t_state == TCPS_LISTEN) 
+                       goto tpgone;
+
+#if TRAFFIC_MGT
+               if (so->so_traffic_mgt_flags & TRAFFIC_MGT_SO_BG_REGULATE && 
+                       bg_cnt > BG_COUNTER_MAX) {
+                       u_int32_t       curr_recvtotal = tcpstat.tcps_rcvtotal;
+                       u_int32_t       curr_bg_recvtotal = tcpstat.tcps_bg_rcvtotal;
+                       u_int32_t       bg_recvdiff = curr_bg_recvtotal - tp->bg_recv_snapshot;
+                       u_int32_t       tot_recvdiff = curr_recvtotal - tp->tot_recv_snapshot;
+                       u_int32_t       fg_recv_change = tot_recvdiff - bg_recvdiff;
+                       u_int32_t       recv_change;
+                       
+                       if (!(so->so_traffic_mgt_flags & TRAFFIC_MGT_SO_BG_SUPPRESSED)) {
+                               if (tot_recvdiff) 
+                                       recv_change = (fg_recv_change * 100) / tot_recvdiff;
+                               else 
+                                       recv_change = 0;
+
+                               if (recv_change > background_io_trigger) {
+                                       socket_set_traffic_mgt_flags(so, TRAFFIC_MGT_SO_BG_SUPPRESSED);
+                               }
+                               
+                               tp->tot_recv_snapshot = curr_recvtotal;
+                               tp->bg_recv_snapshot = curr_bg_recvtotal;
+                       }
+                       else {  // SUPPRESSED
+                               // this allows for bg traffic to subside before we start measuring total traffic change
+                               if (tot_recvdiff)
+                                       recv_change = (bg_recvdiff * 100) / tot_recvdiff;
+                               else
+                                       recv_change = 0;
+                                       
+                               if (recv_change < background_io_trigger) {
+                                       // Draconian for now: if there is any change at all, keep suppressed
+                                       if (!tot_recvdiff) {
+                                               socket_clear_traffic_mgt_flags(so, TRAFFIC_MGT_SO_BG_SUPPRESSED);
+                                               tp->t_unacksegs = 0;
+                                               (void) tcp_output(tp);  // open window
+                                       }
+                               }
+
+                               tp->tot_recv_snapshot = curr_recvtotal;
+                               tp->bg_recv_snapshot = curr_bg_recvtotal;
+                       }
+               }
+#endif /* TRAFFIC_MGT */
+
+               for (i = 1; i < TCPT_NTIMERS; i++) {
+                       if (tp->t_timer[i] != 0) {
+                               tp->t_timer[i] -= TCP_RETRANSHZ/PR_SLOWHZ;
+                               if (tp->t_timer[i] <=  0) {

 #if TCPDEBUG
 #if TCPDEBUG

-                               ostate = tp->t_state;
+                                       ostate = tp->t_state;

 #endif
 #endif

-                               tp = tcp_timers(tp, i);
-                               if (tp == NULL)
-                                       goto tpgone;
+
+                                       tp->t_timer[i] = 0; /* account for granularity change between tcp_now and slowtimo */
+                                       tp = tcp_timers(tp, i);
+                                       if (tp == NULL)
+                                               goto tpgone;

 #if TCPDEBUG
 #if TCPDEBUG

-                               if (tp->t_inpcb->inp_socket->so_options
-                                   & SO_DEBUG)
-                                       tcp_trace(TA_USER, ostate, tp,
-                                                 (void *)0,
-                                                 (struct tcphdr *)0,
-                                                 PRU_SLOWTIMO);
+                                       if (tp->t_inpcb->inp_socket->so_options
+                                           & SO_DEBUG)
+                                               tcp_trace(TA_USER, ostate, tp,
+                                                         (void *)0,
+                                                         (struct tcphdr *)0,
+                                                         PRU_SLOWTIMO);

 #endif
 #endif

+                               }

                        }
                }
                        }
                }

-               tp->t_rcvtime++;
-               tp->t_starttime++;
-               if (tp->t_rtttime)
-                       tp->t_rtttime++;

 tpgone:
 tpgone:

-               ;
+               tcp_unlock(so, 1, 0);

        }
        }

+       
+       if (bg_cnt > 3) 
+               bg_cnt = 0;

 
 

-#if KDEBUG
-       tws_checked = 0;
-#endif
-       KERNEL_DEBUG(DBG_FNC_TCP_SLOW | DBG_FUNC_NONE, tws_checked,0,0,0,0);
+       /* Second part of tcp_slowtimo: garbage collect socket/tcpcb
+        * We need to acquire the list lock exclusively to do this
+        */
+
+       if (lck_rw_lock_shared_to_exclusive(pcbinfo->mtx) == FALSE) {
+               if (tcp_gc_done == TRUE) {      /* don't sweat it this time. cleanup was done last time */
+                       tcp_gc_done = FALSE;
+                       KERNEL_DEBUG(DBG_FNC_TCP_SLOW | DBG_FUNC_END, tws_checked, cur_tw_slot,0,0,0);
+                       return; /* Upgrade failed and lost lock - give up this time. */
+               }
+               lck_rw_lock_exclusive(pcbinfo->mtx);    /* Upgrade failed, lost lock now take it again exclusive */
+       }
+       tcp_gc_done = TRUE;

 
        /*
         * Process the items in the current time-wait slot
         */
 
        /*
         * Process the items in the current time-wait slot
         */

+#if  KDEBUG
+       tws_checked = 0;
+#endif
+       KERNEL_DEBUG(DBG_FNC_TCP_SLOW | DBG_FUNC_NONE, tws_checked,0,0,0,0);

 
 

-       for (ip = time_wait_slots[cur_tw_slot].lh_first; ip; ip = ipnxt)
-       {
+       LIST_FOREACH(inp, &time_wait_slots[cur_tw_slot], inp_list) {

 #if KDEBUG
                tws_checked++;
 #endif
 #if KDEBUG
                tws_checked++;
 #endif

-               ipnxt = ip->inp_list.le_next;
-               tp = intotcpcb(ip);
+
+               if (in_pcb_checkstate(inp, WNT_ACQUIRE, 0) == WNT_STOPUSING) 
+                       continue;
+
+               tcp_lock(inp->inp_socket, 1, 0);
+
+               if (in_pcb_checkstate(inp, WNT_RELEASE, 1) == WNT_STOPUSING) 
+                       goto twunlock;
+
+               tp = intotcpcb(inp);
+               if (tp == NULL)  /* tp already closed, remove from list */
+                       goto twunlock;
+

                if (tp->t_timer[TCPT_2MSL] >= N_TIME_WAIT_SLOTS) {
                    tp->t_timer[TCPT_2MSL] -= N_TIME_WAIT_SLOTS;
                    tp->t_rcvtime += N_TIME_WAIT_SLOTS;
                if (tp->t_timer[TCPT_2MSL] >= N_TIME_WAIT_SLOTS) {
                    tp->t_timer[TCPT_2MSL] -= N_TIME_WAIT_SLOTS;
                    tp->t_rcvtime += N_TIME_WAIT_SLOTS;


@@ -316,14 +581,30 @@ tpgone:
                else
                    tp->t_timer[TCPT_2MSL] = 0;
 
                else
                    tp->t_timer[TCPT_2MSL] = 0;
 

-               if (tp->t_timer[TCPT_2MSL] == 0) 
-                   tp = tcp_timers(tp, TCPT_2MSL);
+               if (tp->t_timer[TCPT_2MSL] == 0)  {
+
+                       /* That pcb is ready for a close */     
+                       tcp_free_sackholes(tp);
+                       tp = tcp_close(tp);
+               }
+twunlock:
+               tcp_unlock(inp->inp_socket, 1, 0);
+       }
+
+
+       LIST_FOREACH_SAFE(inp, &tcb, inp_list, nxt) {
+               tcp_garbage_collect(inp, 0);
+       }
+
+       /* Now cleanup the time wait ones */
+       LIST_FOREACH_SAFE(inp, &time_wait_slots[cur_tw_slot], inp_list, nxt) {
+               tcp_garbage_collect(inp, 1);

        }
 
        if (++cur_tw_slot >= N_TIME_WAIT_SLOTS)
                cur_tw_slot = 0;
        }
 
        if (++cur_tw_slot >= N_TIME_WAIT_SLOTS)
                cur_tw_slot = 0;

-       tcp_now++;                                      /* for timestamps */
-       splx(s);
+       
+       lck_rw_done(pcbinfo->mtx);

        KERNEL_DEBUG(DBG_FNC_TCP_SLOW | DBG_FUNC_END, tws_checked, cur_tw_slot,0,0,0);
 }
 
        KERNEL_DEBUG(DBG_FNC_TCP_SLOW | DBG_FUNC_END, tws_checked, cur_tw_slot,0,0,0);
 }
 


@@ -359,28 +640,39 @@ tcp_timers(tp, timer)
        register int rexmt;
        struct socket *so_tmp;
        struct tcptemp *t_template;
        register int rexmt;
        struct socket *so_tmp;
        struct tcptemp *t_template;

+       int optlen = 0;
+
+#if TCPDEBUG
+       int ostate;
+#endif

 
 #if INET6
        int isipv6 = (tp->t_inpcb->inp_vflag & INP_IPV4) == 0;
 #endif /* INET6 */
 
 
 #if INET6
        int isipv6 = (tp->t_inpcb->inp_vflag & INP_IPV4) == 0;
 #endif /* INET6 */
 

+       so_tmp = tp->t_inpcb->inp_socket;

 
        switch (timer) {
 
        /*
         * 2 MSL timeout in shutdown went off.  If we're closed but
         * still waiting for peer to close and connection has been idle
 
        switch (timer) {
 
        /*
         * 2 MSL timeout in shutdown went off.  If we're closed but
         * still waiting for peer to close and connection has been idle

-        * too long, or if 2MSL time is up from TIME_WAIT, delete connection
-        * control block.  Otherwise, check again in a bit.
+        * too long, or if 2MSL time is up from TIME_WAIT or FIN_WAIT_2,
+        * delete connection control block.
+        * Otherwise, (this case shouldn't happen) check again in a bit
+        * we keep the socket in the main list in that case.

         */
        case TCPT_2MSL:
         */
        case TCPT_2MSL:

+               tcp_free_sackholes(tp);

                if (tp->t_state != TCPS_TIME_WAIT &&
                if (tp->t_state != TCPS_TIME_WAIT &&

-                   tp->t_rcvtime <= tcp_maxidle) {
-                       tp->t_timer[TCPT_2MSL] = tcp_keepintvl;
-                       add_to_time_wait(tp);
+                   tp->t_state != TCPS_FIN_WAIT_2 &&
+                   tp->t_rcvtime < tcp_maxidle) {
+                       tp->t_timer[TCPT_2MSL] = (u_int32_t)tcp_keepintvl;

                }
                }

-               else
+               else {

                        tp = tcp_close(tp);
                        tp = tcp_close(tp);

+                       return(tp);
+               }

                break;
 
        /*
                break;
 
        /*


@@ -389,10 +681,10 @@ tcp_timers(tp, timer)
         * to a longer retransmit interval and retransmit one segment.
         */
        case TCPT_REXMT:
         * to a longer retransmit interval and retransmit one segment.
         */
        case TCPT_REXMT:

+               tcp_free_sackholes(tp);

                if (++tp->t_rxtshift > TCP_MAXRXTSHIFT) {
                        tp->t_rxtshift = TCP_MAXRXTSHIFT;
                        tcpstat.tcps_timeoutdrop++;
                if (++tp->t_rxtshift > TCP_MAXRXTSHIFT) {
                        tp->t_rxtshift = TCP_MAXRXTSHIFT;
                        tcpstat.tcps_timeoutdrop++;

-                       so_tmp = tp->t_inpcb->inp_socket;

                        tp = tcp_drop(tp, tp->t_softerror ?
                            tp->t_softerror : ETIMEDOUT);
                        postevent(so_tmp, 0, EV_TIMEOUT);                       
                        tp = tcp_drop(tp, tp->t_softerror ?
                            tp->t_softerror : ETIMEDOUT);
                        postevent(so_tmp, 0, EV_TIMEOUT);                       


@@ -411,7 +703,12 @@ tcp_timers(tp, timer)
                         */
                        tp->snd_cwnd_prev = tp->snd_cwnd;
                        tp->snd_ssthresh_prev = tp->snd_ssthresh;
                         */
                        tp->snd_cwnd_prev = tp->snd_cwnd;
                        tp->snd_ssthresh_prev = tp->snd_ssthresh;

-                       tp->t_badrxtwin = tcp_now + (tp->t_srtt >> (TCP_RTT_SHIFT + 1));
+                       tp->snd_recover_prev = tp->snd_recover;
+                       if (IN_FASTRECOVERY(tp))
+                                 tp->t_flags |= TF_WASFRECOVERY;
+                       else
+                                 tp->t_flags &= ~TF_WASFRECOVERY;
+                       tp->t_badrxtwin = tcp_now  + (tp->t_srtt >> (TCP_RTT_SHIFT)); 

                }
                tcpstat.tcps_rexmttimeo++;
                if (tp->t_state == TCPS_SYN_SENT)
                }
                tcpstat.tcps_rexmttimeo++;
                if (tp->t_state == TCPS_SYN_SENT)


@@ -423,14 +720,60 @@ tcp_timers(tp, timer)
                tp->t_timer[TCPT_REXMT] = tp->t_rxtcur;
 
                /*
                tp->t_timer[TCPT_REXMT] = tp->t_rxtcur;
 
                /*

-                * Disable rfc1323 and rfc1644 if we havn't got any response to
-                * our third SYN to work-around some broken terminal servers 
-                * (most of which have hopefully been retired) that have bad VJ 
-                * header compression code which trashes TCP segments containing 
-                * unknown-to-them TCP options.
+                * Check for potential Path MTU Discovery Black Hole 
+                */
+
+               if (tcp_pmtud_black_hole_detect && (tp->t_state == TCPS_ESTABLISHED)) {
+                       if (((tp->t_flags & (TF_PMTUD|TF_MAXSEGSNT)) == (TF_PMTUD|TF_MAXSEGSNT)) && (tp->t_rxtshift == 2)) {
+                               /* 
+                                * Enter Path MTU Black-hole Detection mechanism:
+                                * - Disable Path MTU Discovery (IP "DF" bit).
+                                * - Reduce MTU to lower value than what we negociated with peer.
+                                */
+
+                               tp->t_flags &= ~TF_PMTUD; /* Disable Path MTU Discovery for now */
+                               tp->t_flags |= TF_BLACKHOLE; /* Record that we may have found a black hole */
+                               optlen = tp->t_maxopd - tp->t_maxseg;
+                               tp->t_pmtud_saved_maxopd = tp->t_maxopd; /* Keep track of previous MSS */
+                               if (tp->t_maxopd > tcp_pmtud_black_hole_mss)
+                                       tp->t_maxopd = tcp_pmtud_black_hole_mss; /* Reduce the MSS to intermediary value */
+                               else {
+                                       tp->t_maxopd =  /* use the default MSS */
+#if INET6
+                                               isipv6 ? tcp_v6mssdflt :
+#endif /* INET6 */
+                                                       tcp_mssdflt;
+                               }
+                               tp->t_maxseg = tp->t_maxopd - optlen;
+                       }
+                       /*
+                        * If further retransmissions are still unsuccessful with a lowered MTU,
+                        * maybe this isn't a Black Hole and we restore the previous MSS and
+                        * blackhole detection flags.
+                        */
+                       else {
+       
+                               if ((tp->t_flags & TF_BLACKHOLE) && (tp->t_rxtshift > 4)) {
+                                       tp->t_flags |= TF_PMTUD; 
+                                       tp->t_flags &= ~TF_BLACKHOLE; 
+                                       optlen = tp->t_maxopd - tp->t_maxseg;
+                                       tp->t_maxopd = tp->t_pmtud_saved_maxopd;
+                                       tp->t_maxseg = tp->t_maxopd - optlen;
+                               }
+                       }
+               }
+
+
+               /*
+                * Disable rfc1323 and rfc1644 if we haven't got any response to
+                * our SYN (after we reach the threshold) to work-around some
+                * broken terminal servers (most of which have hopefully been
+                * retired) that have bad VJ header compression code which
+                * trashes TCP segments containing unknown-to-them TCP options.

                 */
                 */

-               if ((tp->t_state == TCPS_SYN_SENT) && (tp->t_rxtshift == 3))
-                               tp->t_flags &= ~(TF_REQ_SCALE|TF_REQ_TSTMP|TF_REQ_CC);
+               if ((tp->t_state == TCPS_SYN_SENT) &&
+                   (tp->t_rxtshift == tcp_broken_peer_syn_rxmit_thres))
+                       tp->t_flags &= ~(TF_REQ_SCALE|TF_REQ_TSTMP|TF_REQ_CC);

                /*
                 * If losing, let the lower level know and try for
                 * a better route.  Also, if we backed off this far,
                /*
                 * If losing, let the lower level know and try for
                 * a better route.  Also, if we backed off this far,


@@ -487,14 +830,17 @@ tcp_timers(tp, timer)
                 * growth is 2 mss.  We don't allow the threshhold
                 * to go below this.)
                 */
                 * growth is 2 mss.  We don't allow the threshhold
                 * to go below this.)
                 */

-               {
-               u_int win = min(tp->snd_wnd, tp->snd_cwnd) / 2 / tp->t_maxseg;
-               if (win < 2)
-                       win = 2;
-               tp->snd_cwnd = tp->t_maxseg;
-               tp->snd_ssthresh = win * tp->t_maxseg;
-               tp->t_dupacks = 0;
+               if (tp->t_state >=  TCPS_ESTABLISHED) {
+                       u_int win = min(tp->snd_wnd, tp->snd_cwnd) / 2 / tp->t_maxseg;
+                       if (win < 2)
+                               win = 2;
+                       tp->snd_cwnd = tp->t_maxseg;
+                       tp->snd_ssthresh = win * tp->t_maxseg;
+                       tp->t_bytes_acked = 0;
+                       tp->t_dupacks = 0;
+                       tp->t_unacksegs = 0;

                }
                }

+               EXIT_FASTRECOVERY(tp);

                (void) tcp_output(tp);
                break;
 
                (void) tcp_output(tp);
                break;
 


@@ -522,6 +868,7 @@ tcp_timers(tp, timer)
                }
                tcp_setpersist(tp);
                tp->t_force = 1;
                }
                tcp_setpersist(tp);
                tp->t_force = 1;

+               tp->t_unacksegs = 0;

                (void) tcp_output(tp);
                tp->t_force = 0;
                break;
                (void) tcp_output(tp);
                tp->t_force = 0;
                break;


@@ -536,8 +883,8 @@ tcp_timers(tp, timer)
                        goto dropit;
                if ((always_keepalive ||
                    tp->t_inpcb->inp_socket->so_options & SO_KEEPALIVE) &&
                        goto dropit;
                if ((always_keepalive ||
                    tp->t_inpcb->inp_socket->so_options & SO_KEEPALIVE) &&

-                   tp->t_state <= TCPS_CLOSING) {
-                       if (tp->t_rcvtime >= tcp_keepidle + tcp_maxidle)
+                   (tp->t_state <= TCPS_CLOSING || tp->t_state == TCPS_FIN_WAIT_2)) {
+                       if (tp->t_rcvtime >= TCP_KEEPIDLE(tp) + (u_int32_t)tcp_maxidle)

                                goto dropit;
                        /*
                         * Send a packet designed to force a response
                                goto dropit;
                        /*
                         * Send a packet designed to force a response


@@ -554,14 +901,21 @@ tcp_timers(tp, timer)
                        tcpstat.tcps_keepprobe++;
                        t_template = tcp_maketemplate(tp);
                        if (t_template) {
                        tcpstat.tcps_keepprobe++;
                        t_template = tcp_maketemplate(tp);
                        if (t_template) {

+                               unsigned int ifscope;
+
+                               if (tp->t_inpcb->inp_flags & INP_BOUND_IF)
+                                       ifscope = tp->t_inpcb->inp_boundif;
+                               else
+                                       ifscope = IFSCOPE_NONE;
+

                                tcp_respond(tp, t_template->tt_ipgen,
                                    &t_template->tt_t, (struct mbuf *)NULL,
                                tcp_respond(tp, t_template->tt_ipgen,
                                    &t_template->tt_t, (struct mbuf *)NULL,

-                                   tp->rcv_nxt, tp->snd_una - 1, 0);
+                                   tp->rcv_nxt, tp->snd_una - 1, 0, ifscope);

                                (void) m_free(dtom(t_template));
                        }
                        tp->t_timer[TCPT_KEEP] = tcp_keepintvl;
                } else
                                (void) m_free(dtom(t_template));
                        }
                        tp->t_timer[TCPT_KEEP] = tcp_keepintvl;
                } else

-                       tp->t_timer[TCPT_KEEP] = tcp_keepidle;
+                       tp->t_timer[TCPT_KEEP] = TCP_KEEPIDLE(tp);

                break;
 
 #if TCPDEBUG
                break;
 
 #if TCPDEBUG


@@ -571,7 +925,6 @@ tcp_timers(tp, timer)
 #endif
        dropit:
                tcpstat.tcps_keepdrops++;
 #endif
        dropit:
                tcpstat.tcps_keepdrops++;

-               so_tmp = tp->t_inpcb->inp_socket;

                tp = tcp_drop(tp, ETIMEDOUT);
                postevent(so_tmp, 0, EV_TIMEOUT);
                break;
                tp = tcp_drop(tp, ETIMEDOUT);
                postevent(so_tmp, 0, EV_TIMEOUT);
                break;