]> git.saurik.com Git - apple/xnu.git/blobdiff - bsd/netinet/ip_divert.c
xnu-6153.81.5.tar.gz
[apple/xnu.git] / bsd / netinet / ip_divert.c
index 8e1e7f3dfcd95e9302c6aa5eb792f75fa2302611..aef9c2c11fa2a21a8de1d20b057002dbd5823518 100644 (file)
@@ -1,8 +1,8 @@
 /*
- * Copyright (c) 2000 Apple Computer, Inc. All rights reserved.
+ * Copyright (c) 2000-2019 Apple 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
  * unlawful or unlicensed copies of an Apple operating system, or to
  * circumvent, violate, or enable the circumvention or violation of, any
  * terms of an Apple operating system software license agreement.
- * 
+ *
  * Please obtain a copy of the License at
  * http://www.opensource.apple.com/apsl/ and read it before using this file.
- * 
+ *
  * The Original Code and all software distributed under the License are
  * distributed on an 'AS IS' basis, WITHOUT WARRANTY OF ANY KIND, EITHER
  * EXPRESS OR IMPLIED, AND APPLE HEREBY DISCLAIMS ALL SUCH WARRANTIES,
@@ -22,7 +22,7 @@
  * FITNESS FOR A PARTICULAR PURPOSE, QUIET ENJOYMENT OR NON-INFRINGEMENT.
  * Please see the License for the specific language governing rights and
  * limitations under the License.
- * 
+ *
  * @APPLE_OSREFERENCE_LICENSE_HEADER_END@
  */
 /*
@@ -60,7 +60,7 @@
  * $FreeBSD: src/sys/netinet/ip_divert.c,v 1.98 2004/08/17 22:05:54 andre Exp $
  */
 
-#ifndef INET
+#if !INET
 #error "IPDIVERT requires INET."
 #endif
 
 #include <sys/systm.h>
 #include <sys/proc.h>
 
+#include <machine/endian.h>
 
 #include <net/if.h>
 #include <net/route.h>
+#include <net/kpi_protocol.h>
 
 #include <netinet/in.h>
 #include <netinet/in_systm.h>
@@ -90,6 +92,7 @@
 #include <netinet/ip_divert.h>
 
 #include <kern/zalloc.h>
+#include <libkern/OSAtomic.h>
 
 /*
  * Divert sockets
 /*
  * Allocate enough space to hold a full IP packet
  */
-#define        DIVSNDQ         (65536 + 100)
-#define        DIVRCVQ         (65536 + 100)
+#define DIVSNDQ         (65536 + 100)
+#define DIVRCVQ         (65536 + 100)
 
 /*
  * Divert sockets work in conjunction with ipfw, see the divert(4)
 static struct inpcbhead divcb;
 static struct inpcbinfo divcbinfo;
 
-static u_long  div_sendspace = DIVSNDQ;        /* XXX sysctl ? */
-static u_long  div_recvspace = DIVRCVQ;        /* XXX sysctl ? */
+static u_int32_t        div_sendspace = DIVSNDQ;        /* XXX sysctl ? */
+static u_int32_t        div_recvspace = DIVRCVQ;        /* XXX sysctl ? */
 
 /* Optimization: have this preinitialized */
-static struct sockaddr_in divsrc = { sizeof(divsrc), AF_INET,  };
+static struct sockaddr_in divsrc = {
+       .sin_len = sizeof(divsrc),
+       .sin_family = AF_INET,
+       .sin_port = 0,
+       .sin_addr = { .s_addr = 0 },
+       .sin_zero = { 0, 0, 0, 0, 0, 0, 0, 0 }
+};
 
 /* Internal functions */
 static int div_output(struct socket *so,
-               struct mbuf *m, struct sockaddr *addr, struct mbuf *control);
+    struct mbuf *m, struct sockaddr_in *addr, struct mbuf *control);
 
 extern int load_ipfw(void);
 /*
  * Initialize divert connection block queue.
  */
 void
-div_init(void)
+div_init(struct protosw *pp, struct domain *dp)
 {
+#pragma unused(dp)
+       static int div_initialized = 0;
        struct inpcbinfo *pcbinfo;
+
+       VERIFY((pp->pr_flags & (PR_INITIALIZED | PR_ATTACHED)) == PR_ATTACHED);
+
+       if (div_initialized) {
+               return;
+       }
+       div_initialized = 1;
+
        LIST_INIT(&divcb);
-       divcbinfo.listhead = &divcb;
+       divcbinfo.ipi_listhead = &divcb;
        /*
         * XXX We don't use the hash list for divert IP, but it's easier
         * to allocate a one entry hash list than it is to check all
-        * over the place for hashbase == NULL.
+        * over the place for ipi_hashbase == NULL.
         */
-       divcbinfo.hashbase = hashinit(1, M_PCB, &divcbinfo.hashmask);
-       divcbinfo.porthashbase = hashinit(1, M_PCB, &divcbinfo.porthashmask);
-       divcbinfo.ipi_zone = (void *) zinit(sizeof(struct inpcb),(maxsockets * sizeof(struct inpcb)),
-                                  4096, "divzone");
+       divcbinfo.ipi_hashbase = hashinit(1, M_PCB, &divcbinfo.ipi_hashmask);
+       divcbinfo.ipi_porthashbase = hashinit(1, M_PCB, &divcbinfo.ipi_porthashmask);
+       divcbinfo.ipi_zone = zinit(sizeof(struct inpcb), (512 * sizeof(struct inpcb)),
+           4096, "divzone");
        pcbinfo = &divcbinfo;
-        /*
+       /*
         * allocate lock group attribute and group for udp pcb mutexes
         */
-       pcbinfo->mtx_grp_attr = lck_grp_attr_alloc_init();
+       pcbinfo->ipi_lock_grp_attr = lck_grp_attr_alloc_init();
+
+       pcbinfo->ipi_lock_grp = lck_grp_alloc_init("divcb", pcbinfo->ipi_lock_grp_attr);
 
-       pcbinfo->mtx_grp = lck_grp_alloc_init("divcb", pcbinfo->mtx_grp_attr);
-               
        /*
         * allocate the lock attribute for divert pcb mutexes
         */
-       pcbinfo->mtx_attr = lck_attr_alloc_init();
-       lck_attr_setdefault(pcbinfo->mtx_attr);
+       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 ((pcbinfo->mtx = lck_rw_alloc_init(pcbinfo->mtx_grp, pcbinfo->mtx_attr)) == NULL)
-               return; /* pretty much dead if this fails... */
+       in_pcbinfo_attach(&divcbinfo);
 
+#if IPFIREWALL
        if (!IPFW_LOADED) {
                load_ipfw();
        }
+#endif
 }
 
 /*
@@ -187,7 +212,7 @@ div_init(void)
 void
 div_input(struct mbuf *m, __unused int off)
 {
-       ipstat.ips_noproto++;
+       OSAddAtomic(1, &ipstat.ips_noproto);
        m_freem(m);
 }
 
@@ -196,7 +221,6 @@ div_input(struct mbuf *m, __unused int off)
  *
  * Setup generic address and protocol structures for div_input routine,
  * then pass them along with mbuf chain.
- * ###LOCK  called in ip_mutex from ip_output/ip_input
  */
 void
 divert_packet(struct mbuf *m, int incoming, int port, int rule)
@@ -209,7 +233,7 @@ divert_packet(struct mbuf *m, int incoming, int port, int rule)
        /* Sanity check */
        KASSERT(port != 0, ("%s: port=0", __FUNCTION__));
 
-       divsrc.sin_port = rule;         /* record matching rule */
+       divsrc.sin_port = rule;         /* record matching rule */
 
        /* Assure header */
        if (m->m_len < sizeof(struct ip) &&
@@ -232,12 +256,14 @@ divert_packet(struct mbuf *m, int incoming, int port, int rule)
                /* Find IP address for receive interface */
                ifnet_lock_shared(m->m_pkthdr.rcvif);
                TAILQ_FOREACH(ifa, &m->m_pkthdr.rcvif->if_addrhead, ifa_link) {
-                       if (ifa->ifa_addr == NULL)
-                               continue;
-                       if (ifa->ifa_addr->sa_family != AF_INET)
+                       IFA_LOCK(ifa);
+                       if (ifa->ifa_addr->sa_family != AF_INET) {
+                               IFA_UNLOCK(ifa);
                                continue;
+                       }
                        divsrc.sin_addr =
-                           ((struct sockaddr_in *) ifa->ifa_addr)->sin_addr;
+                           ((struct sockaddr_in *)(void *) ifa->ifa_addr)->sin_addr;
+                       IFA_UNLOCK(ifa);
                        break;
                }
                ifnet_lock_done(m->m_pkthdr.rcvif);
@@ -248,10 +274,10 @@ divert_packet(struct mbuf *m, int incoming, int port, int rule)
        bzero(&divsrc.sin_zero, sizeof(divsrc.sin_zero));
        if (m->m_pkthdr.rcvif) {
                /*
-                * Hide the actual interface name in there in the 
+                * Hide the actual interface name in there in the
                 * sin_zero array. XXX This needs to be moved to a
                 * different sockaddr type for divert, e.g.
-                * sockaddr_div with multiple fields like 
+                * sockaddr_div with multiple fields like
                 * sockaddr_dl. Presently we have only 7 bytes
                 * but that will do for now as most interfaces
                 * are 4 or less + 2 or less bytes for unit.
@@ -264,34 +290,35 @@ divert_packet(struct mbuf *m, int incoming, int port, int rule)
                 * and re-uses the sockaddr_in as suggested in the man pages,
                 * this iface name will come along for the ride.
                 * (see div_output for the other half of this.)
-                */ 
+                */
                snprintf(divsrc.sin_zero, sizeof(divsrc.sin_zero),
-                       "%s%d", m->m_pkthdr.rcvif->if_name,
-                       m->m_pkthdr.rcvif->if_unit);
+                   "%s", if_name(m->m_pkthdr.rcvif));
        }
 
        /* Put packet on socket queue, if any */
        sa = NULL;
        nport = htons((u_int16_t)port);
-       lck_rw_lock_shared(divcbinfo.mtx);      
+       lck_rw_lock_shared(divcbinfo.ipi_lock);
        LIST_FOREACH(inp, &divcb, inp_list) {
-               if (inp->inp_lport == nport)
+               if (inp->inp_lport == nport) {
                        sa = inp->inp_socket;
+               }
        }
        if (sa) {
                int error = 0;
-               
+
                socket_lock(sa, 1);
                if (sbappendaddr(&sa->so_rcv, (struct sockaddr *)&divsrc,
-                               m, (struct mbuf *)0, &error) != 0)
+                   m, (struct mbuf *)0, &error) != 0) {
                        sorwakeup(sa);
+               }
                socket_unlock(sa, 1);
        } else {
                m_freem(m);
-               ipstat.ips_noproto++;
-               ipstat.ips_delivered--;
-        }
-       lck_rw_done(divcbinfo.mtx);     
+               OSAddAtomic(1, &ipstat.ips_noproto);
+               OSAddAtomic(-1, &ipstat.ips_delivered);
+       }
+       lck_rw_done(divcbinfo.ipi_lock);
 }
 
 /*
@@ -300,124 +327,146 @@ divert_packet(struct mbuf *m, int incoming, int port, int rule)
  * If no address specified, or address is 0.0.0.0, send to ip_output();
  * otherwise, send to ip_input() and mark as having been received on
  * the interface with that address.
- * ###LOCK  called in inet_proto mutex when from div_send. 
+ * ###LOCK  called in inet_proto mutex when from div_send.
  */
 static int
-div_output(so, m, addr, control)
-       struct socket *so;
-       register struct mbuf *m;
-       struct sockaddr *addr;
-       struct mbuf *control;
+div_output(struct socket *so, struct mbuf *m, struct sockaddr_in *sin,
+    struct mbuf *control)
 {
-       register struct inpcb *const inp = sotoinpcb(so);
-       register struct ip *const ip = mtod(m, struct ip *);
-       struct sockaddr_in *sin = (struct sockaddr_in *)addr;
+       struct inpcb *const inp = sotoinpcb(so);
+       struct ip *const ip = mtod(m, struct ip *);
        int error = 0;
+       int sotc = SO_TC_UNSPEC;
 
-       if (control)
-               m_freem(control);               /* XXX */
+       if (control != NULL) {
+               int ignored;
+
+               (void) so_tc_from_control(contro, &sotc, &ignored);
+
+               m_freem(control);               /* XXX */
+               control = NULL;
+       }
+       if (sotc == SO_TC_UNSPEC) {
+               sotc = so->so_traffic_class;
+       }
 
        /* Loopback avoidance and state recovery */
        if (sin) {
                struct m_tag *mtag;
                struct divert_tag *dt;
-               int     len = 0;
-               char    *c = sin->sin_zero;
+               int     len = 0;
+               char    *c = sin->sin_zero;
 
-               mtag = m_tag_alloc(KERNEL_MODULE_TAG_ID, KERNEL_TAG_TYPE_DIVERT,
-                               sizeof(struct divert_tag), M_NOWAIT);
+               mtag = m_tag_create(KERNEL_MODULE_TAG_ID, KERNEL_TAG_TYPE_DIVERT,
+                   sizeof(struct divert_tag), M_NOWAIT, m);
                if (mtag == NULL) {
                        error = ENOBUFS;
                        goto cantsend;
                }
-               dt = (struct divert_tag *)(mtag+1);
+               dt = (struct divert_tag *)(mtag + 1);
                dt->info = 0;
                dt->cookie = sin->sin_port;
                m_tag_prepend(m, mtag);
 
                /*
                 * Find receive interface with the given name or IP address.
-                * The name is user supplied data so don't trust it's size or 
+                * The name is user supplied data so don't trust it's size or
                 * that it is zero terminated. The name has priority.
-                * We are presently assuming that the sockaddr_in 
+                * We are presently assuming that the sockaddr_in
                 * has not been replaced by a sockaddr_div, so we limit it
                 * to 16 bytes in total. the name is stuffed (if it exists)
                 * in the sin_zero[] field.
                 */
-               while (*c++ && (len++ < sizeof(sin->sin_zero)));
-               if ((len > 0) && (len < sizeof(sin->sin_zero)))
+               while (*c++ && (len++ < sizeof(sin->sin_zero))) {
+                       ;
+               }
+               if ((len > 0) && (len < sizeof(sin->sin_zero))) {
                        m->m_pkthdr.rcvif = ifunit(sin->sin_zero);
+               }
        }
 
        /* Reinject packet into the system as incoming or outgoing */
        if (!sin || sin->sin_addr.s_addr == 0) {
+               struct ip_out_args ipoa;
+               struct route ro;
+               struct ip_moptions *imo;
+
+               bzero(&ipoa, sizeof(ipoa));
+               ipoa.ipoa_boundif = IFSCOPE_NONE;
+               ipoa.ipoa_flags = IPOAF_SELECT_SRCIF;
+               ipoa.ipoa_sotc = SO_TC_UNSPEC;
+               ipoa.ipoa_netsvctype = _NET_SERVICE_TYPE_UNSPEC;
+
                /*
                 * Don't allow both user specified and setsockopt options,
                 * and don't allow packet length sizes that will crash
                 */
-               if (((ip->ip_hl != (sizeof (*ip) >> 2)) && inp->inp_options) ||
-                    ((u_short)ntohs(ip->ip_len) > m->m_pkthdr.len)) {
+               if (((ip->ip_hl != (sizeof(*ip) >> 2)) && inp->inp_options) ||
+                   ((u_short)ntohs(ip->ip_len) > m->m_pkthdr.len)) {
                        error = EINVAL;
                        goto cantsend;
                }
 
                /* Convert fields to host order for ip_output() */
+#if BYTE_ORDER != BIG_ENDIAN
                NTOHS(ip->ip_len);
                NTOHS(ip->ip_off);
+#endif
 
-               /* Send packet to output processing */
-               ipstat.ips_rawout++;                    /* XXX */
+               OSAddAtomic(1, &ipstat.ips_rawout);
+               /* Copy the cached route and take an extra reference */
+               inp_route_copyout(inp, &ro);
+
+               if (sotc != SO_TC_UNSPEC) {
+                       ipoa.ipoa_flags |= IPOAF_QOSMARKING_ALLOWED;
+                       ipoa.ipoa_sotc = sotc;
+               }
+               set_packet_service_class(m, so, sotc, 0);
+
+               imo = inp->inp_moptions;
+               if (imo != NULL) {
+                       IMO_ADDREF(imo);
+               }
                socket_unlock(so, 0);
-               error = ip_output(m,
-                           inp->inp_options, &inp->inp_route,
-                       (so->so_options & SO_DONTROUTE) |
-                       IP_ALLOWBROADCAST | IP_RAWOUTPUT,
-                       inp->inp_moptions);
+#if CONFIG_MACF_NET
+               mac_mbuf_label_associate_inpcb(inp, m);
+#endif
+               /* Send packet to output processing */
+               error = ip_output(m, inp->inp_options, &ro,
+                   (so->so_options & SO_DONTROUTE) |
+                   IP_ALLOWBROADCAST | IP_RAWOUTPUT | IP_OUTARGS,
+                   imo, &ipoa);
+
                socket_lock(so, 0);
+               if (imo != NULL) {
+                       IMO_REMREF(imo);
+               }
+               /* Synchronize cached PCB route */
+               inp_route_copyin(inp, &ro);
        } else {
-               struct  ifaddr *ifa;
+               struct  ifaddr *ifa;
 
                /* If no luck with the name above. check by IP address.  */
                if (m->m_pkthdr.rcvif == NULL) {
+                       struct sockaddr_in _sin;
                        /*
-                        * Make sure there are no distractions
-                        * for ifa_ifwithaddr. Clear the port and the ifname.
-                        * Maybe zap all 8 bytes at once using a 64bit write?
+                        * Make sure there are no distractions for
+                        * ifa_ifwithaddr; use sanitized version.
                         */
-                       bzero(sin->sin_zero, sizeof(sin->sin_zero));
-                       /* *((u_int64_t *)sin->sin_zero) = 0; */ /* XXX ?? */
-                       sin->sin_port = 0;
-                       if (!(ifa = ifa_ifwithaddr((struct sockaddr *) sin))) {
+                       bzero(&_sin, sizeof(_sin));
+                       _sin.sin_family = AF_INET;
+                       _sin.sin_len = sizeof(struct sockaddr_in);
+                       _sin.sin_addr.s_addr = sin->sin_addr.s_addr;
+                       if (!(ifa = ifa_ifwithaddr(SA(&_sin)))) {
                                error = EADDRNOTAVAIL;
                                goto cantsend;
                        }
                        m->m_pkthdr.rcvif = ifa->ifa_ifp;
-                       ifafree(ifa);
-               }
-               
-               if ((~IF_HWASSIST_CSUM_FLAGS(m->m_pkthdr.rcvif->if_hwassist) & 
-                               m->m_pkthdr.csum_flags) == 0) {
-                       if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
-                               m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
-                       }
-                       m->m_pkthdr.csum_flags |=
-                               CSUM_DATA_VALID | CSUM_PSEUDO_HDR |
-                               CSUM_IP_CHECKED | CSUM_IP_VALID;
-                       m->m_pkthdr.csum_data = 0xffff;
+                       IFA_REMREF(ifa);
                }
-               else if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
-                       int     hlen;
-                       
-#ifdef _IP_VHL
-                       hlen = IP_VHL_HL(ip->ip_vhl) << 2;
-#else
-                       hlen = ip->ip_hl << 2;
+#if CONFIG_MACF_NET
+               mac_mbuf_label_associate_socket(so, m);
 #endif
-                       in_delayed_cksum(m);
-                       m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
-                       ip->ip_sum = in_cksum(m, hlen);
-               }
-
                /* Send packet to input processing */
                proto_inject(PF_INET, m);
        }
@@ -437,28 +486,35 @@ div_attach(struct socket *so, int proto, struct proc *p)
 
 
        inp  = sotoinpcb(so);
-       if (inp)
+       if (inp) {
                panic("div_attach");
-       if (p && (error = proc_suser(p)) != 0)
+       }
+       if ((error = proc_suser(p)) != 0) {
                return error;
+       }
 
        error = soreserve(so, div_sendspace, div_recvspace);
-       if (error)
+       if (error) {
                return error;
+       }
        error = in_pcballoc(so, &divcbinfo, p);
-       if (error)
+       if (error) {
                return error;
+       }
        inp = (struct inpcb *)so->so_pcb;
        inp->inp_ip_p = proto;
        inp->inp_vflag |= INP_IPV4;
        inp->inp_flags |= INP_HDRINCL;
        /* The socket is always "connected" because
-          we always know "where" to send the packet */
+        *  we always know "where" to send the packet */
        so->so_state |= SS_ISCONNECTED;
 
 #ifdef MORE_DICVLOCK_DEBUG
-       printf("div_attach: so=%x sopcb=%x lock=%x ref=%x\n",
-                       so, so->so_pcb, ((struct inpcb *)so->so_pcb)->inpcb_mtx, so->so_usecount);
+       printf("div_attach: so=0x%llx sopcb=0x%llx lock=0x%llx ref=%x\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);
 #endif
        return 0;
 }
@@ -469,12 +525,16 @@ div_detach(struct socket *so)
        struct inpcb *inp;
 
 #ifdef MORE_DICVLOCK_DEBUG
-       printf("div_detach: so=%x sopcb=%x lock=%x ref=%x\n",
-                       so, so->so_pcb, ((struct inpcb *)so->so_pcb)->inpcb_mtx, so->so_usecount);
+       printf("div_detach: so=0x%llx sopcb=0x%llx lock=0x%llx ref=%x\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);
 #endif
        inp = sotoinpcb(so);
-       if (inp == 0)
-               panic("div_detach: so=%x null inp\n", so);
+       if (inp == 0) {
+               panic("div_detach: so=%p null inp\n", so);
+       }
        in_pcbdetach(inp);
        inp->inp_state = INPCB_STATE_DEAD;
        return 0;
@@ -490,8 +550,9 @@ div_abort(struct socket *so)
 static int
 div_disconnect(struct socket *so)
 {
-       if ((so->so_state & SS_ISCONNECTED) == 0)
+       if ((so->so_state & SS_ISCONNECTED) == 0) {
                return ENOTCONN;
+       }
        return div_abort(so);
 }
 
@@ -503,16 +564,16 @@ div_bind(struct socket *so, struct sockaddr *nam, struct proc *p)
 
        inp = sotoinpcb(so);
        /* in_pcbbind assumes that the socket is a sockaddr_in
-       * and in_pcbbind requires a valid address. Since divert
-       * sockets don't we need to make sure the address is
-       * filled in properly.
-       * XXX -- divert should not be abusing in_pcbind
-       * and should probably have its own family.
-       */
+        * and in_pcbbind requires a valid address. Since divert
+        * sockets don't we need to make sure the address is
+        * filled in properly.
+        * XXX -- divert should not be abusing in_pcbind
+        * and should probably have its own family.
+        */
        if (nam->sa_family != AF_INET) {
                error = EAFNOSUPPORT;
        } else {
-               ((struct sockaddr_in *)nam)->sin_addr.s_addr = INADDR_ANY;
+               ((struct sockaddr_in *)(void *)nam)->sin_addr.s_addr = INADDR_ANY;
                error = in_pcbbind(inp, nam, p);
        }
        return error;
@@ -527,23 +588,25 @@ div_shutdown(struct socket *so)
 
 static int
 div_send(struct socket *so, __unused int flags, struct mbuf *m, struct sockaddr *nam,
-        struct mbuf *control, __unused struct proc *p)
+    struct mbuf *control, __unused struct proc *p)
 {
        /* Packet must have a header (but that's about it) */
-       if (m->m_len < sizeof (struct ip) &&
-           (m = m_pullup(m, sizeof (struct ip))) == 0) {
-               ipstat.ips_toosmall++;
+       if (m->m_len < sizeof(struct ip) &&
+           (m = m_pullup(m, sizeof(struct ip))) == 0) {
+               OSAddAtomic(1, &ipstat.ips_toosmall);
                m_freem(m);
                return EINVAL;
        }
 
        /* Send packet */
-       return div_output(so, m, nam, control);
+       return div_output(so, m, SIN(nam), control);
 }
 
+#if 0
 static int
 div_pcblist SYSCTL_HANDLER_ARGS
 {
+#pragma unused(oidp, arg1, arg2)
        int error, i, n;
        struct inpcb *inp, **inp_list;
        inp_gen_t gencnt;
@@ -553,17 +616,17 @@ div_pcblist SYSCTL_HANDLER_ARGS
         * The process of preparing the TCB list is too time-consuming and
         * resource-intensive to repeat twice on every request.
         */
-       lck_rw_lock_exclusive(divcbinfo.mtx);
+       lck_rw_lock_exclusive(divcbinfo.ipi_lock);
        if (req->oldptr == USER_ADDR_NULL) {
                n = divcbinfo.ipi_count;
                req->oldidx = 2 * (sizeof xig)
-                       + (n + n/8) * sizeof(struct xinpcb);
-               lck_rw_done(divcbinfo.mtx);
+                   + (n + n / 8) * sizeof(struct xinpcb);
+               lck_rw_done(divcbinfo.ipi_lock);
                return 0;
        }
 
        if (req->newptr != USER_ADDR_NULL) {
-               lck_rw_done(divcbinfo.mtx);
+               lck_rw_done(divcbinfo.ipi_lock);
                return EPERM;
        }
 
@@ -580,24 +643,24 @@ div_pcblist SYSCTL_HANDLER_ARGS
        xig.xig_sogen = so_gencnt;
        error = SYSCTL_OUT(req, &xig, sizeof xig);
        if (error) {
-               lck_rw_done(divcbinfo.mtx);
+               lck_rw_done(divcbinfo.ipi_lock);
                return error;
        }
 
        inp_list = _MALLOC(n * sizeof *inp_list, M_TEMP, M_WAITOK);
        if (inp_list == 0) {
-               lck_rw_done(divcbinfo.mtx);
+               lck_rw_done(divcbinfo.ipi_lock);
                return ENOMEM;
        }
-       
-       for (inp = LIST_FIRST(divcbinfo.listhead), i = 0; inp && i < n;
-            inp = LIST_NEXT(inp, inp_list)) {
+
+       for (inp = LIST_FIRST(divcbinfo.ipi_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
-                       inp_list[i++] = inp;
+               { inp_list[i++] = inp;}
        }
        n = i;
 
@@ -611,8 +674,9 @@ div_pcblist SYSCTL_HANDLER_ARGS
                        xi.xi_len = sizeof xi;
                        /* XXX should avoid extra copy */
                        inpcb_to_compat(inp, &xi.xi_inp);
-                       if (inp->inp_socket)
+                       if (inp->inp_socket) {
                                sotoxsocket(inp->inp_socket, &xi.xi_socket);
+                       }
                        error = SYSCTL_OUT(req, &xi, sizeof xi);
                }
        }
@@ -632,113 +696,134 @@ div_pcblist SYSCTL_HANDLER_ARGS
                error = SYSCTL_OUT(req, &xig, sizeof xig);
        }
        FREE(inp_list, M_TEMP);
-       lck_rw_done(divcbinfo.mtx);
+       lck_rw_done(divcbinfo.ipi_lock);
        return error;
 }
+#endif
 
 __private_extern__ int
-div_lock(struct socket *so, int refcount, int lr)
- {
-       int lr_saved;
-#ifdef __ppc__
-       if (lr == 0) {
-                       __asm__ volatile("mflr %0" : "=r" (lr_saved));
-       }
-       else lr_saved = lr;
-#endif
-       
+div_lock(struct socket *so, int refcount, void *lr)
+{
+       void *lr_saved;
+
+       if (lr == NULL) {
+               lr_saved = __builtin_return_address(0);
+       } else {
+               lr_saved = lr;
+       }
+
 #ifdef MORE_DICVLOCK_DEBUG
-       printf("div_lock: so=%x sopcb=%x lock=%x ref=%x lr=%x\n",
-                       so, 
-                       so->so_pcb, 
-                       so->so_pcb ? ((struct inpcb *)so->so_pcb)->inpcb_mtx : 0, 
-                       so->so_usecount, 
-                       lr_saved);
+       printf("div_lock: 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), so->so_pcb ?
+           (uint64_t)VM_KERNEL_ADDRPERM(&(sotoinpcb(so)->inpcb_mtx)) : NULL,
+           so->so_usecount, (uint64_t)VM_KERNEL_ADDRPERM(lr_saved));
 #endif
        if (so->so_pcb) {
-               lck_mtx_lock(((struct inpcb *)so->so_pcb)->inpcb_mtx);
-       } else  {
-               panic("div_lock: so=%x NO PCB! lr=%x\n", so, lr_saved);
-               lck_mtx_lock(so->so_proto->pr_domain->dom_mtx);
-       }
-       
-       if (so->so_usecount < 0)
-               panic("div_lock: so=%x so_pcb=%x lr=%x ref=%x\n",
-               so, so->so_pcb, lr_saved, so->so_usecount);
-       
-       if (refcount)
+               lck_mtx_lock(&((struct inpcb *)so->so_pcb)->inpcb_mtx);
+       } else {
+               panic("div_lock: so=%p NO PCB! lr=%p lrh= lrh= %s\n",
+                   so, lr_saved, solockhistory_nr(so));
+               /* NOTREACHED */
+       }
+
+       if (so->so_usecount < 0) {
+               panic("div_lock: so=%p so_pcb=%p lr=%p ref=%x lrh= %s\n",
+                   so, so->so_pcb, lr_saved, so->so_usecount,
+                   solockhistory_nr(so));
+               /* NOTREACHED */
+       }
+
+       if (refcount) {
                so->so_usecount++;
-       so->reserved3 = (void *)lr_saved;
+       }
+       so->lock_lr[so->next_lock_lr] = lr_saved;
+       so->next_lock_lr = (so->next_lock_lr + 1) % SO_LCKDBG_MAX;
 
-       return (0);
+       return 0;
 }
 
 __private_extern__ int
-div_unlock(struct socket *so, int refcount, int lr)
+div_unlock(struct socket *so, int refcount, void *lr)
 {
-       int lr_saved;
+       void *lr_saved;
        lck_mtx_t * mutex_held;
-       struct inpcb *inp = sotoinpcb(so);      
-#ifdef __ppc__
-       if (lr == 0) {
-               __asm__ volatile("mflr %0" : "=r" (lr_saved));
+       struct inpcb *inp = sotoinpcb(so);
+
+       if (lr == NULL) {
+               lr_saved = __builtin_return_address(0);
+       } else {
+               lr_saved = lr;
        }
-       else lr_saved = lr;
-#endif
-       
+
 #ifdef MORE_DICVLOCK_DEBUG
-       printf("div_unlock: so=%x sopcb=%x lock=%x ref=%x lr=%x\n",
-                       so, 
-                       so->so_pcb, 
-                       so->so_pcb ? ((struct inpcb *)so->so_pcb)->inpcb_mtx : 0, 
-                       so->so_usecount, 
-                       lr_saved);
+       printf("div_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), so->so_pcb ?
+           (uint64_t)VM_KERNEL_ADDRPERM(&(sotoinpcb(so)->inpcb_mtx)) : NULL,
+           so->so_usecount, lr_saved);
 #endif
-       if (refcount)
+       if (refcount) {
                so->so_usecount--;
-       
-       if (so->so_usecount < 0)
-               panic("div_unlock: so=%x usecount=%x\n", so, so->so_usecount);
+       }
+
+       if (so->so_usecount < 0) {
+               panic("div_unlock: so=%p usecount=%x lrh= %s\n",
+                   so, so->so_usecount, solockhistory_nr(so));
+               /* NOTREACHED */
+       }
        if (so->so_pcb == NULL) {
-               panic("div_unlock: so=%x NO PCB usecount=%x lr=%x\n", so, so->so_usecount, lr_saved);
-               mutex_held = so->so_proto->pr_domain->dom_mtx;
-       } else {
-               mutex_held = ((struct inpcb *)so->so_pcb)->inpcb_mtx;
+               panic("div_unlock: so=%p NO PCB usecount=%x lr=%p lrh= %s\n",
+                   so, so->so_usecount, lr_saved, solockhistory_nr(so));
+               /* NOTREACHED */
        }
+       mutex_held = &((struct inpcb *)so->so_pcb)->inpcb_mtx;
 
        if (so->so_usecount == 0 && (inp->inp_wantcnt == WNT_STOPUSING)) {
-               lck_rw_lock_exclusive(divcbinfo.mtx);
+               lck_rw_lock_exclusive(divcbinfo.ipi_lock);
+               if (inp->inp_state != INPCB_STATE_DEAD) {
+                       in_pcbdetach(inp);
+               }
                in_pcbdispose(inp);
-               lck_rw_done(divcbinfo.mtx);
-               return (0);
+               lck_rw_done(divcbinfo.ipi_lock);
+               return 0;
        }
-       lck_mtx_assert(mutex_held, LCK_MTX_ASSERT_OWNED);
+       LCK_MTX_ASSERT(mutex_held, LCK_MTX_ASSERT_OWNED);
+       so->unlock_lr[so->next_unlock_lr] = lr_saved;
+       so->next_unlock_lr = (so->next_unlock_lr + 1) % SO_LCKDBG_MAX;
        lck_mtx_unlock(mutex_held);
-       so->reserved4 = (void *)lr_saved;
-       return (0);
+       return 0;
 }
 
 __private_extern__ lck_mtx_t *
-div_getlock(struct socket *so, __unused int locktype)
+div_getlock(struct socket *so, __unused int flags)
 {
        struct inpcb *inpcb = (struct inpcb *)so->so_pcb;
-       
-       if (so->so_pcb)  {
-               if (so->so_usecount < 0)
-                       panic("div_getlock: so=%x usecount=%x\n", so, so->so_usecount);
-               return(inpcb->inpcb_mtx);
+
+       if (so->so_pcb) {
+               if (so->so_usecount < 0) {
+                       panic("div_getlock: so=%p usecount=%x lrh= %s\n",
+                           so, so->so_usecount, solockhistory_nr(so));
+               }
+               return &inpcb->inpcb_mtx;
        } else {
-               panic("div_getlock: so=%x NULL so_pcb\n", so);
-               return (so->so_proto->pr_domain->dom_mtx);
+               panic("div_getlock: so=%p NULL NO PCB lrh= %s\n",
+                   so, solockhistory_nr(so));
+               return so->so_proto->pr_domain->dom_mtx;
        }
 }
 
-
 struct pr_usrreqs div_usrreqs = {
-       div_abort, pru_accept_notsupp, div_attach, div_bind,
-       pru_connect_notsupp, pru_connect2_notsupp, in_control, div_detach,
-       div_disconnect, pru_listen_notsupp, in_setpeeraddr, pru_rcvd_notsupp,
-       pru_rcvoob_notsupp, div_send, pru_sense_null, div_shutdown,
-       in_setsockaddr, sosend, soreceive, pru_sopoll_notsupp
+       .pru_abort =            div_abort,
+       .pru_attach =           div_attach,
+       .pru_bind =             div_bind,
+       .pru_control =          in_control,
+       .pru_detach =           div_detach,
+       .pru_disconnect =       div_disconnect,
+       .pru_peeraddr =         in_getpeeraddr,
+       .pru_send =             div_send,
+       .pru_shutdown =         div_shutdown,
+       .pru_sockaddr =         in_getsockaddr,
+       .pru_sosend =           sosend,
+       .pru_soreceive =        soreceive,
 };
-