]> git.saurik.com Git - apple/xnu.git/blobdiff - bsd/net/bpf.c
xnu-3789.70.16.tar.gz
[apple/xnu.git] / bsd / net / bpf.c
index 74ac085698f7ebed0785bfe721e4cc549cbf7e8f..66e0c3b549cc5b03d0a3fcded261f6a69ef0c719 100644 (file)
@@ -1,5 +1,5 @@
 /*
 /*
- * Copyright (c) 2000-2008 Apple Inc. All rights reserved.
+ * Copyright (c) 2000-2017 Apple Inc. All rights reserved.
  *
  * @APPLE_OSREFERENCE_LICENSE_HEADER_START@
  * 
  *
  * @APPLE_OSREFERENCE_LICENSE_HEADER_START@
  * 
 #include <sys/ttycom.h>
 #include <sys/filedesc.h>
 #include <sys/uio_internal.h>
 #include <sys/ttycom.h>
 #include <sys/filedesc.h>
 #include <sys/uio_internal.h>
-#include <sys/fcntl.h>
 #include <sys/file_internal.h>
 #include <sys/event.h>
 
 #include <sys/file_internal.h>
 #include <sys/event.h>
 
-#if defined(sparc) && BSD < 199103
-#include <sys/stream.h>
-#endif
 #include <sys/poll.h>
 
 #include <sys/socket.h>
 #include <sys/poll.h>
 
 #include <sys/socket.h>
+#include <sys/socketvar.h>
 #include <sys/vnode.h>
 
 #include <net/if.h>
 #include <sys/vnode.h>
 
 #include <net/if.h>
 #include <net/bpfdesc.h>
 
 #include <netinet/in.h>
 #include <net/bpfdesc.h>
 
 #include <netinet/in.h>
+#include <netinet/in_pcb.h>
+#include <netinet/in_var.h>
+#include <netinet/ip_var.h>
+#include <netinet/tcp.h>
+#include <netinet/tcp_var.h>
+#include <netinet/udp.h>
+#include <netinet/udp_var.h>
 #include <netinet/if_ether.h>
 #include <sys/kernel.h>
 #include <sys/sysctl.h>
 #include <netinet/if_ether.h>
 #include <sys/kernel.h>
 #include <sys/sysctl.h>
 
 #include <miscfs/devfs/devfs.h>
 #include <net/dlil.h>
 
 #include <miscfs/devfs/devfs.h>
 #include <net/dlil.h>
+#include <net/pktap.h>
 
 #include <kern/locks.h>
 
 #include <kern/locks.h>
+#include <kern/thread_call.h>
 
 #if CONFIG_MACF_NET
 #include <security/mac_framework.h>
 
 #if CONFIG_MACF_NET
 #include <security/mac_framework.h>
 
 extern int tvtohz(struct timeval *);
 
 
 extern int tvtohz(struct timeval *);
 
-/*
- * Older BSDs don't have kernel malloc.
- */
-#if BSD < 199103
-extern bcopy();
-static caddr_t bpf_alloc();
-#include <net/bpf_compat.h>
-#define BPF_BUFSIZE (MCLBYTES-8)
-#define UIOMOVE(cp, len, code, uio) uiomove(cp, len, code, uio)
-#else
 #define BPF_BUFSIZE 4096
 #define UIOMOVE(cp, len, code, uio) uiomove(cp, len, uio)
 #define BPF_BUFSIZE 4096
 #define UIOMOVE(cp, len, code, uio) uiomove(cp, len, uio)
-#endif
 
 
 #define PRINET  26                     /* interruptible */
 
 
 #define PRINET  26                     /* interruptible */
@@ -148,14 +143,26 @@ static caddr_t bpf_alloc();
  * The default read buffer size is patchable.
  */
 static unsigned int bpf_bufsize = BPF_BUFSIZE;
  * The default read buffer size is patchable.
  */
 static unsigned int bpf_bufsize = BPF_BUFSIZE;
-SYSCTL_INT(_debug, OID_AUTO, bpf_bufsize, CTLFLAG_RW
+SYSCTL_INT(_debug, OID_AUTO, bpf_bufsize, CTLFLAG_RW | CTLFLAG_LOCKED,
        &bpf_bufsize, 0, "");
        &bpf_bufsize, 0, "");
-static unsigned int bpf_maxbufsize = BPF_MAXBUFSIZE;
-SYSCTL_INT(_debug, OID_AUTO, bpf_maxbufsize, CTLFLAG_RW
+__private_extern__ unsigned int bpf_maxbufsize = BPF_MAXBUFSIZE;
+SYSCTL_INT(_debug, OID_AUTO, bpf_maxbufsize, CTLFLAG_RW | CTLFLAG_LOCKED,
        &bpf_maxbufsize, 0, "");
 static unsigned int bpf_maxdevices = 256;
        &bpf_maxbufsize, 0, "");
 static unsigned int bpf_maxdevices = 256;
-SYSCTL_UINT(_debug, OID_AUTO, bpf_maxdevices, CTLFLAG_RW
+SYSCTL_UINT(_debug, OID_AUTO, bpf_maxdevices, CTLFLAG_RW | CTLFLAG_LOCKED,
        &bpf_maxdevices, 0, "");
        &bpf_maxdevices, 0, "");
+/*
+ * bpf_wantpktap controls the defaul visibility of DLT_PKTAP
+ * For OS X is off by default so process need to use the ioctl BPF_WANT_PKTAP
+ * explicitly to be able to use DLT_PKTAP.
+ */
+static unsigned int bpf_wantpktap = 0;
+SYSCTL_UINT(_debug, OID_AUTO, bpf_wantpktap, CTLFLAG_RW | CTLFLAG_LOCKED,
+       &bpf_wantpktap, 0, "");
+
+static int bpf_debug = 0;
+SYSCTL_INT(_debug, OID_AUTO, bpf_debug, CTLFLAG_RW | CTLFLAG_LOCKED,
+       &bpf_debug, 0, "");
 
 /*
  *  bpf_iflist is the list of interfaces; each corresponds to an ifnet
 
 /*
  *  bpf_iflist is the list of interfaces; each corresponds to an ifnet
@@ -177,36 +184,36 @@ static struct bpf_d       **bpf_dtab = NULL;
 static unsigned int bpf_dtab_size = 0;
 static unsigned int    nbpfilter = 0;
 
 static unsigned int bpf_dtab_size = 0;
 static unsigned int    nbpfilter = 0;
 
-static lck_mtx_t               *bpf_mlock;
+decl_lck_mtx_data(static, bpf_mlock_data);
+static lck_mtx_t               *bpf_mlock = &bpf_mlock_data;
 static lck_grp_t               *bpf_mlock_grp;
 static lck_grp_attr_t  *bpf_mlock_grp_attr;
 static lck_attr_t              *bpf_mlock_attr;
 
 static lck_grp_t               *bpf_mlock_grp;
 static lck_grp_attr_t  *bpf_mlock_grp_attr;
 static lck_attr_t              *bpf_mlock_attr;
 
-/*
- * Mark a descriptor free by making it point to itself.
- * This is probably cheaper than marking with a constant since
- * the address should be in a register anyway.
- */
+static mbuf_tag_id_t bpf_mtag_id;
 #endif /* __APPLE__ */
 
 static int     bpf_allocbufs(struct bpf_d *);
 static errno_t bpf_attachd(struct bpf_d *d, struct bpf_if *bp);
 #endif /* __APPLE__ */
 
 static int     bpf_allocbufs(struct bpf_d *);
 static errno_t bpf_attachd(struct bpf_d *d, struct bpf_if *bp);
-static void    bpf_detachd(struct bpf_d *d);
+static int     bpf_detachd(struct bpf_d *d, int);
 static void    bpf_freed(struct bpf_d *);
 static void    bpf_mcopy(const void *, void *, size_t);
 static int     bpf_movein(struct uio *, int,
                    struct mbuf **, struct sockaddr *, int *);
 static int     bpf_setif(struct bpf_d *, ifnet_t ifp, u_int32_t dlt);
 static void    bpf_freed(struct bpf_d *);
 static void    bpf_mcopy(const void *, void *, size_t);
 static int     bpf_movein(struct uio *, int,
                    struct mbuf **, struct sockaddr *, int *);
 static int     bpf_setif(struct bpf_d *, ifnet_t ifp, u_int32_t dlt);
-static void bpf_wakeup(struct bpf_d *);
-static void    catchpacket(struct bpf_d *, u_char *, u_int,
-                   u_int, void (*)(const void *, void *, size_t));
+static void    bpf_timed_out(void *, void *);
+static void    bpf_wakeup(struct bpf_d *);
+static void    catchpacket(struct bpf_d *, u_char *, struct mbuf *, u_int,
+                   u_int, int, void (*)(const void *, void *, size_t));
 static void    reset_d(struct bpf_d *);
 static void    reset_d(struct bpf_d *);
-static int bpf_setf(struct bpf_d *, u_int bf_len, user_addr_t bf_insns);
-static int     bpf_getdltlist(struct bpf_d *, struct bpf_dltlist *,
-    struct proc *);
+static int     bpf_setf(struct bpf_d *, u_int, user_addr_t, u_long);
+static int     bpf_getdltlist(struct bpf_d *, caddr_t, struct proc *);
 static int     bpf_setdlt(struct bpf_d *, u_int);
 static int     bpf_setdlt(struct bpf_d *, u_int);
+static int     bpf_set_traffic_class(struct bpf_d *, int);
+static void    bpf_set_packet_service_class(struct mbuf *, int);
 
 
-/*static  void *bpf_devfs_token[MAXBPFILTER];*/
+static void    bpf_acquire_d(struct bpf_d *);
+static void    bpf_release_d(struct bpf_d *);
 
 static  int bpf_devsw_installed;
 
 
 static  int bpf_devsw_installed;
 
@@ -217,31 +224,31 @@ static int bpf_tap_callback(struct ifnet *ifp, struct mbuf *m);
  * Darwin differs from BSD here, the following are static
  * on BSD and not static on Darwin.
  */
  * Darwin differs from BSD here, the following are static
  * on BSD and not static on Darwin.
  */
-       d_open_t        bpfopen;
-       d_close_t       bpfclose;
-       d_read_t        bpfread;
-       d_write_t       bpfwrite;
-        ioctl_fcn_t    bpfioctl;
-       select_fcn_t    bpfpoll;
+       d_open_t            bpfopen;
+       d_close_t           bpfclose;
+       d_read_t            bpfread;
+       d_write_t           bpfwrite;
+       ioctl_fcn_t         bpfioctl;
+       select_fcn_t        bpfselect;
 
 
 /* Darwin's cdevsw struct differs slightly from BSDs */
 #define CDEV_MAJOR 23
 static struct cdevsw bpf_cdevsw = {
 
 
 /* Darwin's cdevsw struct differs slightly from BSDs */
 #define CDEV_MAJOR 23
 static struct cdevsw bpf_cdevsw = {
-       /* open */      bpfopen,
-       /* close */     bpfclose,
-       /* read */      bpfread,
-       /* write */     bpfwrite,
-       /* ioctl */     bpfioctl,
-       /* stop */              eno_stop,
-       /* reset */             eno_reset,
-       /* tty */               NULL,
-       /* select */    bpfpoll,
-       /* mmap */              eno_mmap,
-       /* strategy*/   eno_strat,
-       /* getc */              eno_getc,
-       /* putc */              eno_putc,
-       /* type */              0
+       /* open */          bpfopen,
+       /* close */         bpfclose,
+       /* read */          bpfread,
+       /* write */         bpfwrite,
+       /* ioctl */         bpfioctl,
+       /* stop */          eno_stop,
+       /* reset */         eno_reset,
+       /* tty */           NULL,
+       /* select */        bpfselect,
+       /* mmap */          eno_mmap,
+       /* strategy*/       eno_strat,
+       /* getc */          eno_getc,
+       /* putc */          eno_putc,
+       /* type */          0
 };
 
 #define SOCKADDR_HDR_LEN          offsetof(struct sockaddr, sa_data)
 };
 
 #define SOCKADDR_HDR_LEN          offsetof(struct sockaddr, sa_data)
@@ -316,6 +323,11 @@ bpf_movein(struct uio *uio, int linktype, struct mbuf **mp, struct sockaddr *soc
                hlen = 0;
                break;
 
                hlen = 0;
                break;
 
+       case DLT_IEEE802_11_RADIO:
+               sa_family = AF_IEEE80211;
+               hlen = 0;
+               break;
+
        default:
                return (EIO);
        }
        default:
                return (EIO);
        }
@@ -354,13 +366,8 @@ bpf_movein(struct uio *uio, int linktype, struct mbuf **mp, struct sockaddr *soc
        if (m == 0)
                return (ENOBUFS);
        if ((unsigned)len > MHLEN) {
        if (m == 0)
                return (ENOBUFS);
        if ((unsigned)len > MHLEN) {
-#if BSD >= 199103
                MCLGET(m, M_WAIT);
                if ((m->m_flags & M_EXT) == 0) {
                MCLGET(m, M_WAIT);
                if ((m->m_flags & M_EXT) == 0) {
-#else
-               MCLGET(m);
-               if (m->m_len != MCLBYTES) {
-#endif
                        error = ENOBUFS;
                        goto bad;
                }
                        error = ENOBUFS;
                        goto bad;
                }
@@ -368,24 +375,38 @@ bpf_movein(struct uio *uio, int linktype, struct mbuf **mp, struct sockaddr *soc
        m->m_pkthdr.len = m->m_len = len;
        m->m_pkthdr.rcvif = NULL;
        *mp = m;
        m->m_pkthdr.len = m->m_len = len;
        m->m_pkthdr.rcvif = NULL;
        *mp = m;
+       
        /*
         * Make room for link header.
         */
        if (hlen != 0) {
                m->m_pkthdr.len -= hlen;
                m->m_len -= hlen;
        /*
         * Make room for link header.
         */
        if (hlen != 0) {
                m->m_pkthdr.len -= hlen;
                m->m_len -= hlen;
-#if BSD >= 199103
                m->m_data += hlen; /* XXX */
                m->m_data += hlen; /* XXX */
-#else
-               m->m_off += hlen;
-#endif
                error = UIOMOVE((caddr_t)sockp->sa_data, hlen, UIO_WRITE, uio);
                if (error)
                        goto bad;
        }
        error = UIOMOVE(mtod(m, caddr_t), len - hlen, UIO_WRITE, uio);
                error = UIOMOVE((caddr_t)sockp->sa_data, hlen, UIO_WRITE, uio);
                if (error)
                        goto bad;
        }
        error = UIOMOVE(mtod(m, caddr_t), len - hlen, UIO_WRITE, uio);
-       if (!error)
-               return (0);
+       if (error)
+               goto bad;
+       
+       /* Check for multicast destination */
+       switch (linktype) {
+               case DLT_EN10MB: {
+                       struct ether_header *eh = mtod(m, struct ether_header *);
+                       
+                       if (ETHER_IS_MULTICAST(eh->ether_dhost)) {
+                               if (_ether_cmp(etherbroadcastaddr, eh->ether_dhost) == 0)
+                                       m->m_flags |= M_BCAST;
+                               else
+                                       m->m_flags |= M_MCAST;
+                       }
+                       break;
+               }
+       }
+       
+       return 0;
  bad:
        m_freem(m);
        return (error);
  bad:
        m_freem(m);
        return (error);
@@ -394,8 +415,9 @@ bpf_movein(struct uio *uio, int linktype, struct mbuf **mp, struct sockaddr *soc
 #ifdef __APPLE__
 
 /*
 #ifdef __APPLE__
 
 /*
- * The dynamic addition of a new device node must block all processes that are opening 
- * the last device so that no process will get an unexpected ENOENT 
+ * The dynamic addition of a new device node must block all processes that
+ * are opening the last device so that no process will get an unexpected
+ * ENOENT 
  */
 static void
 bpf_make_dev_t(int maj)
  */
 static void
 bpf_make_dev_t(int maj)
@@ -468,57 +490,86 @@ bpf_attachd(struct bpf_d *d, struct bpf_if *bp)
        d->bd_bif = bp;
        d->bd_next = bp->bif_dlist;
        bp->bif_dlist = d;
        d->bd_bif = bp;
        d->bd_next = bp->bif_dlist;
        bp->bif_dlist = d;
-       
-       if (first) {
-               bpf_tap_mode tap_mode;
 
 
-               switch ((d->bd_oflags & (FREAD | FWRITE))) {
-                       case FREAD:
-                               tap_mode = BPF_TAP_INPUT;
-                               break;
-                       case FWRITE:
-                               tap_mode = BPF_TAP_OUTPUT;
-                               break;
-                       default:
-                               tap_mode = BPF_TAP_INPUT_OUTPUT;
-                               break;
-               }
+       /*
+        * Take a reference on the device even if an error is returned
+        * because we keep the device in the interface's list of listeners
+        */
+       bpf_acquire_d(d);
 
 
+       if (first) {
                /* Find the default bpf entry for this ifp */
                if (bp->bif_ifp->if_bpf == NULL) {
                /* Find the default bpf entry for this ifp */
                if (bp->bif_ifp->if_bpf == NULL) {
-                       struct bpf_if   *primary;
+                       struct bpf_if   *tmp, *primary = NULL;
                        
                        
-                       for (primary = bpf_iflist; primary && primary->bif_ifp != bp->bif_ifp;
-                                primary = primary->bif_next)
-                               ;
-               
+                       for (tmp = bpf_iflist; tmp; tmp = tmp->bif_next) {
+                               if (tmp->bif_ifp != bp->bif_ifp)
+                                       continue;
+                               primary = tmp;
+                               /*
+                                * Make DLT_PKTAP only if process knows how
+                                * to deal with it, otherwise find another one
+                                */
+                               if (tmp->bif_dlt == DLT_PKTAP &&
+                                       !(d->bd_flags & BPF_WANT_PKTAP))
+                                       continue;
+                               break;
+                       }
                        bp->bif_ifp->if_bpf = primary;
                }
                
                /* Only call dlil_set_bpf_tap for primary dlt */
                if (bp->bif_ifp->if_bpf == bp)
                        bp->bif_ifp->if_bpf = primary;
                }
                
                /* Only call dlil_set_bpf_tap for primary dlt */
                if (bp->bif_ifp->if_bpf == bp)
-                       dlil_set_bpf_tap(bp->bif_ifp, tap_mode, bpf_tap_callback);
+                       dlil_set_bpf_tap(bp->bif_ifp, BPF_TAP_INPUT_OUTPUT, bpf_tap_callback);
                
                if (bp->bif_tap)
                
                if (bp->bif_tap)
-                       error = bp->bif_tap(bp->bif_ifp, bp->bif_dlt, tap_mode);
+                       error = bp->bif_tap(bp->bif_ifp, bp->bif_dlt, BPF_TAP_INPUT_OUTPUT);
        }
 
        }
 
+       /*
+        * Reset the detach flags in case we previously detached an interface
+        */
+       d->bd_flags &= ~(BPF_DETACHING | BPF_DETACHED);
+
+       if (bp->bif_ifp->if_bpf != NULL &&
+               bp->bif_ifp->if_bpf->bif_dlt == DLT_PKTAP)
+               d->bd_flags |= BPF_FINALIZE_PKTAP;
+       else
+               d->bd_flags &= ~BPF_FINALIZE_PKTAP;
+
        return error;
 }
 
 /*
  * Detach a file from its interface.
        return error;
 }
 
 /*
  * Detach a file from its interface.
+ *
+ * Return 1 if was closed by some thread, 0 otherwise
  */
  */
-static void
-bpf_detachd(struct bpf_d *d)
+static int
+bpf_detachd(struct bpf_d *d, int closing)
 {
        struct bpf_d **p;
        struct bpf_if *bp;
        struct ifnet  *ifp;
 
 {
        struct bpf_d **p;
        struct bpf_if *bp;
        struct ifnet  *ifp;
 
+       /*
+        * Some other thread already detached
+        */
+       if ((d->bd_flags & (BPF_DETACHED | BPF_DETACHING)) != 0)
+               goto done;
+       /*
+        * This thread is doing the detach
+        */
+       d->bd_flags |= BPF_DETACHING;
+
        ifp = d->bd_bif->bif_ifp;
        bp = d->bd_bif;
        ifp = d->bd_bif->bif_ifp;
        bp = d->bd_bif;
-       
+
+       if (bpf_debug != 0)
+               printf("%s: %llx %s%s\n",
+                   __func__, (uint64_t)VM_KERNEL_ADDRPERM(d),
+                   if_name(ifp), closing ? " closing" : "");
+
        /* Remove d from the interface's descriptor list. */
        p = &bp->bif_dlist;
        while (*p != d) {
        /* Remove d from the interface's descriptor list. */
        p = &bp->bif_dlist;
        while (*p != d) {
@@ -558,12 +609,127 @@ bpf_detachd(struct bpf_d *d)
                         * take it out.
                         * Most likely the network interface is gone.
                         */
                         * take it out.
                         * Most likely the network interface is gone.
                         */
-                       printf("bpf: ifnet_set_promiscuous failed");
+                       printf("%s: ifnet_set_promiscuous failed\n", __func__);
                }
                lck_mtx_lock(bpf_mlock);
        }
                }
                lck_mtx_lock(bpf_mlock);
        }
+
+       /*
+        * Wake up other thread that are waiting for this thread to finish
+        * detaching
+        */
+       d->bd_flags &= ~BPF_DETACHING;
+       d->bd_flags |= BPF_DETACHED;
+       /*
+        * Note that We've kept the reference because we may have dropped
+        * the lock when turning off promiscuous mode
+        */
+       bpf_release_d(d);
+
+done:
+       /*
+        * When closing makes sure no other thread refer to the bpf_d
+        */
+       if (bpf_debug != 0)
+               printf("%s: %llx done\n",
+                   __func__, (uint64_t)VM_KERNEL_ADDRPERM(d));
+       /*
+        * Let the caller know the bpf_d is closed
+        */
+       if ((d->bd_flags & BPF_CLOSING))
+               return (1);
+       else
+               return (0);
+}
+
+
+/*
+ * Start asynchronous timer, if necessary.
+ * Must be called with bpf_mlock held.
+ */
+static void
+bpf_start_timer(struct bpf_d *d)
+{
+       uint64_t deadline;
+       struct timeval tv;
+
+       if (d->bd_rtout > 0 && d->bd_state == BPF_IDLE) {
+               tv.tv_sec = d->bd_rtout / hz;
+               tv.tv_usec = (d->bd_rtout % hz) * tick;
+
+               clock_interval_to_deadline(
+                   (uint64_t)tv.tv_sec * USEC_PER_SEC + tv.tv_usec,
+                   NSEC_PER_USEC, &deadline);
+               /*
+                * The state is BPF_IDLE, so the timer hasn't 
+                * been started yet, and hasn't gone off yet;
+                * there is no thread call scheduled, so this
+                * won't change the schedule.
+                *
+                * XXX - what if, by the time it gets entered,
+                * the deadline has already passed?
+                */
+               thread_call_enter_delayed(d->bd_thread_call, deadline);
+               d->bd_state = BPF_WAITING;
+       }
+}
+
+/*
+ * Cancel asynchronous timer.
+ * Must be called with bpf_mlock held.
+ */
+static boolean_t
+bpf_stop_timer(struct bpf_d *d)
+{
+       /*
+        * If the timer has already gone off, this does nothing.
+        * Our caller is expected to set d->bd_state to BPF_IDLE,
+        * with the bpf_mlock, after we are called. bpf_timed_out()
+        * also grabs bpf_mlock, so, if the timer has gone off and 
+        * bpf_timed_out() hasn't finished, it's waiting for the
+        * lock; when this thread releases the lock, it will 
+        * find the state is BPF_IDLE, and just release the 
+        * lock and return.
+        */
+       return (thread_call_cancel(d->bd_thread_call));
+}
+
+void
+bpf_acquire_d(struct bpf_d *d)
+{
+       void *lr_saved =  __builtin_return_address(0);
+
+       lck_mtx_assert(bpf_mlock, LCK_MTX_ASSERT_OWNED);
+
+       d->bd_refcnt += 1;
+
+       d->bd_ref_lr[d->bd_next_ref_lr] = lr_saved;
+       d->bd_next_ref_lr = (d->bd_next_ref_lr + 1) % BPF_REF_HIST;
 }
 
 }
 
+void
+bpf_release_d(struct bpf_d *d)
+{
+       void *lr_saved =  __builtin_return_address(0);
+
+       lck_mtx_assert(bpf_mlock, LCK_MTX_ASSERT_OWNED);
+
+       if (d->bd_refcnt <= 0)
+               panic("%s: %p refcnt <= 0", __func__, d);
+
+       d->bd_refcnt -= 1;
+
+       d->bd_unref_lr[d->bd_next_unref_lr] = lr_saved;
+       d->bd_next_unref_lr = (d->bd_next_unref_lr + 1) % BPF_REF_HIST;
+
+       if (d->bd_refcnt == 0) {
+               /* Assert the device is detached */
+               if ((d->bd_flags & BPF_DETACHED) == 0)
+                       panic("%s: %p BPF_DETACHED not set", __func__, d);
+
+               _FREE(d, M_DEVBUF);
+       }
+}
 
 /*
  * Open ethernet device.  Returns ENXIO for illegal minor device number,
 
 /*
  * Open ethernet device.  Returns ENXIO for illegal minor device number,
@@ -607,7 +773,8 @@ bpfopen(dev_t dev, int flags, __unused int fmt,
                lck_mtx_unlock(bpf_mlock);
                return (EBUSY);
        }
                lck_mtx_unlock(bpf_mlock);
                return (EBUSY);
        }
-       d = (struct bpf_d *)_MALLOC(sizeof(struct bpf_d), M_DEVBUF, M_WAIT);
+       d = (struct bpf_d *)_MALLOC(sizeof(struct bpf_d), M_DEVBUF,
+           M_WAIT | M_ZERO);
        if (d == NULL) {
                /* this really is a catastrophic failure */
                printf("bpfopen: malloc bpf_d failed\n");
        if (d == NULL) {
                /* this really is a catastrophic failure */
                printf("bpfopen: malloc bpf_d failed\n");
@@ -615,18 +782,29 @@ bpfopen(dev_t dev, int flags, __unused int fmt,
                lck_mtx_unlock(bpf_mlock);
                return ENOMEM;
        }
                lck_mtx_unlock(bpf_mlock);
                return ENOMEM;
        }
-       bzero(d, sizeof(struct bpf_d));
-       
-       /*
-        * It is not necessary to take the BPF lock here because no other 
-        * thread can access the device until it is marked opened...
-        */
-       
+
        /* Mark "in use" and do most initialization. */
        /* Mark "in use" and do most initialization. */
+       bpf_acquire_d(d);
        d->bd_bufsize = bpf_bufsize;
        d->bd_sig = SIGIO;
        d->bd_seesent = 1;
        d->bd_oflags = flags;
        d->bd_bufsize = bpf_bufsize;
        d->bd_sig = SIGIO;
        d->bd_seesent = 1;
        d->bd_oflags = flags;
+       d->bd_state = BPF_IDLE;
+       d->bd_traffic_class = SO_TC_BE;
+       d->bd_flags |= BPF_DETACHED;
+       if (bpf_wantpktap)
+               d->bd_flags |= BPF_WANT_PKTAP;
+       else
+               d->bd_flags &= ~BPF_WANT_PKTAP;
+       d->bd_thread_call = thread_call_allocate(bpf_timed_out, d);
+       if (d->bd_thread_call == NULL) {
+               printf("bpfopen: malloc thread call failed\n");
+               bpf_dtab[minor(dev)] = NULL;
+               bpf_release_d(d);
+               lck_mtx_unlock(bpf_mlock);
+
+               return (ENOMEM);
+       }
 #if CONFIG_MACF_NET
        mac_bpfdesc_label_init(d);
        mac_bpfdesc_label_associate(kauth_cred_get(), d);
 #if CONFIG_MACF_NET
        mac_bpfdesc_label_init(d);
        mac_bpfdesc_label_associate(kauth_cred_get(), d);
@@ -655,23 +833,93 @@ bpfclose(dev_t dev, __unused int flags, __unused int fmt,
        if (d == 0 || d == (void *)1) {
                lck_mtx_unlock(bpf_mlock);
                return (ENXIO);
        if (d == 0 || d == (void *)1) {
                lck_mtx_unlock(bpf_mlock);
                return (ENXIO);
-       }       
+       }
+
+       /*
+        * Other threads may call bpd_detachd() if we drop the bpf_mlock
+        */
+       d->bd_flags |= BPF_CLOSING;
+
+       if (bpf_debug != 0)
+               printf("%s: %llx\n",
+                   __func__, (uint64_t)VM_KERNEL_ADDRPERM(d));
+
        bpf_dtab[minor(dev)] = (void *)1;               /* Mark closing */
 
        bpf_dtab[minor(dev)] = (void *)1;               /* Mark closing */
 
+       /*
+        * Deal with any in-progress timeouts.
+        */
+       switch (d->bd_state) {
+               case BPF_IDLE:
+                       /*
+                        * Not waiting for a timeout, and no timeout happened.
+                        */
+                       break;
+
+               case BPF_WAITING:
+                       /*
+                        * Waiting for a timeout.
+                        * Cancel any timer that has yet to go off,
+                        * and mark the state as "closing".
+                        * Then drop the lock to allow any timers that
+                        * *have* gone off to run to completion, and wait
+                        * for them to finish.
+                        */
+                       if (!bpf_stop_timer(d)) {
+                               /*
+                                * There was no pending call, so the call must 
+                                * have been in progress. Wait for the call to
+                                * complete; we have to drop the lock while 
+                                * waiting. to let the in-progrss call complete
+                                */
+                               d->bd_state = BPF_DRAINING;
+                               while (d->bd_state == BPF_DRAINING)
+                                       msleep((caddr_t)d, bpf_mlock, PRINET,
+                                                       "bpfdraining", NULL);
+                       }
+                       d->bd_state = BPF_IDLE;
+                       break;
+
+               case BPF_TIMED_OUT:
+                       /*
+                        * Timer went off, and the timeout routine finished.
+                        */
+                       d->bd_state = BPF_IDLE;
+                       break;
+
+               case BPF_DRAINING:
+                       /*
+                        * Another thread is blocked on a close waiting for
+                        * a timeout to finish.
+                        * This "shouldn't happen", as the first thread to enter
+                        * bpfclose() will set bpf_dtab[minor(dev)] to 1, and 
+                        * all subsequent threads should see that and fail with 
+                        * ENXIO.
+                        */
+                       panic("Two threads blocked in a BPF close");
+                       break;
+       }
+
        if (d->bd_bif)
        if (d->bd_bif)
-               bpf_detachd(d);
+               bpf_detachd(d, 1);
        selthreadclear(&d->bd_sel);
 #if CONFIG_MACF_NET
        mac_bpfdesc_label_destroy(d);
 #endif
        selthreadclear(&d->bd_sel);
 #if CONFIG_MACF_NET
        mac_bpfdesc_label_destroy(d);
 #endif
+       thread_call_free(d->bd_thread_call);
+
+       while (d->bd_hbuf_read)
+               msleep((caddr_t)d, bpf_mlock, PRINET, "bpf_reading", NULL);
+
        bpf_freed(d);
 
        /* Mark free in same context as bpfopen comes to check */
        bpf_dtab[minor(dev)] = NULL;                    /* Mark closed */
        bpf_freed(d);
 
        /* Mark free in same context as bpfopen comes to check */
        bpf_dtab[minor(dev)] = NULL;                    /* Mark closed */
+
+       bpf_release_d(d);
+
        lck_mtx_unlock(bpf_mlock);
        lck_mtx_unlock(bpf_mlock);
-       
-       _FREE(d, M_DEVBUF);
-       
+
        return (0);
 }
 
        return (0);
 }
 
@@ -681,15 +929,12 @@ bpfclose(dev_t dev, __unused int flags, __unused int fmt,
 static int
 bpf_sleep(struct bpf_d *d, int pri, const char *wmesg, int timo)
 {
 static int
 bpf_sleep(struct bpf_d *d, int pri, const char *wmesg, int timo)
 {
-       int st;
+       u_int64_t abstime = 0;
 
 
-       lck_mtx_unlock(bpf_mlock);
+       if(timo)
+               clock_interval_to_deadline(timo, NSEC_PER_SEC / hz, &abstime);
        
        
-       st = tsleep((caddr_t)d, pri, wmesg, timo);
-       
-       lck_mtx_lock(bpf_mlock);
-       
-       return st;
+       return msleep1((caddr_t)d, bpf_mlock, pri, wmesg, abstime);
 }
 
 /*
 }
 
 /*
@@ -698,10 +943,14 @@ bpf_sleep(struct bpf_d *d, int pri, const char *wmesg, int timo)
  * Zero the length of the new store buffer.
  */
 #define ROTATE_BUFFERS(d) \
  * Zero the length of the new store buffer.
  */
 #define ROTATE_BUFFERS(d) \
+       if (d->bd_hbuf_read) \
+               panic("rotating bpf buffers during read"); \
        (d)->bd_hbuf = (d)->bd_sbuf; \
        (d)->bd_hlen = (d)->bd_slen; \
        (d)->bd_hbuf = (d)->bd_sbuf; \
        (d)->bd_hlen = (d)->bd_slen; \
+       (d)->bd_hcnt = (d)->bd_scnt; \
        (d)->bd_sbuf = (d)->bd_fbuf; \
        (d)->bd_slen = 0; \
        (d)->bd_sbuf = (d)->bd_fbuf; \
        (d)->bd_slen = 0; \
+       (d)->bd_scnt = 0; \
        (d)->bd_fbuf = NULL;
 /*
  *  bpfread - read next chunk of packets from buffers
        (d)->bd_fbuf = NULL;
 /*
  *  bpfread - read next chunk of packets from buffers
@@ -710,35 +959,59 @@ int
 bpfread(dev_t dev, struct uio *uio, int ioflag)
 {
        struct bpf_d *d;
 bpfread(dev_t dev, struct uio *uio, int ioflag)
 {
        struct bpf_d *d;
+       caddr_t hbuf; 
+       int timed_out, hbuf_len;
        int error;
        int error;
+       int flags;
 
        lck_mtx_lock(bpf_mlock);
 
        d = bpf_dtab[minor(dev)];
 
        lck_mtx_lock(bpf_mlock);
 
        d = bpf_dtab[minor(dev)];
-       if (d == 0 || d == (void *)1) {
+       if (d == 0 || d == (void *)1 || (d->bd_flags & BPF_CLOSING) != 0) {
                lck_mtx_unlock(bpf_mlock);
                return (ENXIO);
        }
 
                lck_mtx_unlock(bpf_mlock);
                return (ENXIO);
        }
 
+       bpf_acquire_d(d);
 
        /*
         * Restrict application to use a buffer the same size as
         * as kernel buffers.
         */
        if (uio_resid(uio) != d->bd_bufsize) {
 
        /*
         * Restrict application to use a buffer the same size as
         * as kernel buffers.
         */
        if (uio_resid(uio) != d->bd_bufsize) {
+               bpf_release_d(d);
                lck_mtx_unlock(bpf_mlock);
                return (EINVAL);
        }
                lck_mtx_unlock(bpf_mlock);
                return (EINVAL);
        }
+       
+       if (d->bd_state == BPF_WAITING)
+               bpf_stop_timer(d);
+       
+       timed_out = (d->bd_state == BPF_TIMED_OUT);
+       d->bd_state = BPF_IDLE;
+
+       while (d->bd_hbuf_read) 
+               msleep((caddr_t)d, bpf_mlock, PRINET, "bpf_reading", NULL);
 
 
+       if ((d->bd_flags & BPF_CLOSING) != 0) {
+               bpf_release_d(d);
+               lck_mtx_unlock(bpf_mlock);
+               return (ENXIO);
+       }
        /*
         * If the hold buffer is empty, then do a timed sleep, which
         * ends when the timeout expires or when enough packets
         * have arrived to fill the store buffer.
         */
        while (d->bd_hbuf == 0) {
        /*
         * If the hold buffer is empty, then do a timed sleep, which
         * ends when the timeout expires or when enough packets
         * have arrived to fill the store buffer.
         */
        while (d->bd_hbuf == 0) {
-               if (d->bd_immediate && d->bd_slen != 0) {
+               if ((d->bd_immediate || timed_out || (ioflag & IO_NDELAY)) 
+                       && d->bd_slen != 0) {
                        /*
                        /*
-                        * A packet(s) either arrived since the previous
+                        * We're in immediate mode, or are reading
+                        * in non-blocking mode, or a timer was
+                        * started before the read (e.g., by select()
+                        * or poll()) and has expired and a packet(s)
+                        * either arrived since the previous
                         * read or arrived while we were asleep.
                         * Rotate the buffers and return what's here.
                         */
                         * read or arrived while we were asleep.
                         * Rotate the buffers and return what's here.
                         */
@@ -753,10 +1026,12 @@ bpfread(dev_t dev, struct uio *uio, int ioflag)
                 * it before using it again.
                 */
                if (d->bd_bif == NULL) {
                 * it before using it again.
                 */
                if (d->bd_bif == NULL) {
+                       bpf_release_d(d);
                        lck_mtx_unlock(bpf_mlock);
                        return (ENXIO);
                }
                if (ioflag & IO_NDELAY) {
                        lck_mtx_unlock(bpf_mlock);
                        return (ENXIO);
                }
                if (ioflag & IO_NDELAY) {
+                       bpf_release_d(d);
                        lck_mtx_unlock(bpf_mlock);
                        return (EWOULDBLOCK);
                }
                        lck_mtx_unlock(bpf_mlock);
                        return (EWOULDBLOCK);
                }
@@ -765,12 +1040,42 @@ bpfread(dev_t dev, struct uio *uio, int ioflag)
                /*
                 * Make sure device is still opened
                 */
                /*
                 * Make sure device is still opened
                 */
-               d = bpf_dtab[minor(dev)];
-               if (d == 0 || d == (void *)1) {
+               if ((d->bd_flags & BPF_CLOSING) != 0) {
+                       bpf_release_d(d);
+                       lck_mtx_unlock(bpf_mlock);
+                       return (ENXIO);
+               }
+
+               while (d->bd_hbuf_read)
+                       msleep((caddr_t)d, bpf_mlock, PRINET, "bpf_reading", NULL);
+
+               if ((d->bd_flags & BPF_CLOSING) != 0) {
+                       bpf_release_d(d);
                        lck_mtx_unlock(bpf_mlock);
                        return (ENXIO);
                }
                        lck_mtx_unlock(bpf_mlock);
                        return (ENXIO);
                }
+
                if (error == EINTR || error == ERESTART) {
                if (error == EINTR || error == ERESTART) {
+                       if (d->bd_hbuf) {
+                               /*
+                                * Because we msleep, the hold buffer might
+                                * be filled when we wake up.  Avoid rotating
+                                * in this case.
+                                */
+                               break;
+                       }
+                       if (d->bd_slen) {
+                               /*
+                                * Sometimes we may be interrupted often and
+                                * the sleep above will not timeout.
+                                * Regardless, we should rotate the buffers
+                                * if there's any new data pending and
+                                * return it.
+                                */
+                               ROTATE_BUFFERS(d);
+                               break;
+                       }
+                       bpf_release_d(d);
                        lck_mtx_unlock(bpf_mlock);
                        return (error);
                }
                        lck_mtx_unlock(bpf_mlock);
                        return (error);
                }
@@ -789,6 +1094,7 @@ bpfread(dev_t dev, struct uio *uio, int ioflag)
                                break;
 
                        if (d->bd_slen == 0) {
                                break;
 
                        if (d->bd_slen == 0) {
+                               bpf_release_d(d);
                                lck_mtx_unlock(bpf_mlock);
                                return (0);
                        }
                                lck_mtx_unlock(bpf_mlock);
                                return (0);
                        }
@@ -800,18 +1106,120 @@ bpfread(dev_t dev, struct uio *uio, int ioflag)
         * At this point, we know we have something in the hold slot.
         */
 
         * At this point, we know we have something in the hold slot.
         */
 
+       /*
+        * Set the hold buffer read. So we do not 
+        * rotate the buffers until the hold buffer
+        * read is complete. Also to avoid issues resulting
+        * from page faults during disk sleep (<rdar://problem/13436396>).
+        */
+       d->bd_hbuf_read = 1;
+       hbuf = d->bd_hbuf;
+       hbuf_len = d->bd_hlen;
+       flags = d->bd_flags;
+       lck_mtx_unlock(bpf_mlock);
+
+#ifdef __APPLE__
+       /*
+        * Before we move data to userland, we fill out the extended
+        * header fields.
+        */
+       if (flags & BPF_EXTENDED_HDR) {
+               char *p;
+
+               p = hbuf;
+               while (p < hbuf + hbuf_len) {
+                       struct bpf_hdr_ext *ehp;
+                       uint32_t flowid;
+                       struct so_procinfo soprocinfo;
+                       int found = 0;
+
+                       ehp = (struct bpf_hdr_ext *)(void *)p;
+                       if ((flowid = ehp->bh_flowid)) {
+                               if (ehp->bh_proto == IPPROTO_TCP)
+                                       found = inp_findinpcb_procinfo(&tcbinfo,
+                                           flowid, &soprocinfo);
+                               else if (ehp->bh_proto == IPPROTO_UDP)
+                                       found = inp_findinpcb_procinfo(&udbinfo,
+                                           flowid, &soprocinfo);
+                               if (found == 1) {
+                                       ehp->bh_pid = soprocinfo.spi_pid;
+                                       proc_name(ehp->bh_pid, ehp->bh_comm, MAXCOMLEN);
+                               }
+                               ehp->bh_flowid = 0;
+                       }
+                       if (flags & BPF_FINALIZE_PKTAP) {
+                               struct pktap_header *pktaphdr;
+                               
+                               pktaphdr = (struct pktap_header *)(void *)
+                                   (p + BPF_WORDALIGN(ehp->bh_hdrlen));
+
+                               if (pktaphdr->pth_flags & PTH_FLAG_DELAY_PKTAP)
+                                       pktap_finalize_proc_info(pktaphdr);
+
+                               if (pktaphdr->pth_flags & PTH_FLAG_TSTAMP) {
+                                       ehp->bh_tstamp.tv_sec =
+                                               pktaphdr->pth_tstamp.tv_sec;
+                                       ehp->bh_tstamp.tv_usec =
+                                               pktaphdr->pth_tstamp.tv_usec;
+                               }
+                       }
+                       p += BPF_WORDALIGN(ehp->bh_hdrlen + ehp->bh_caplen);
+               }
+       } else if (flags & BPF_FINALIZE_PKTAP) {
+               char *p;
+
+               p = hbuf;
+               while (p < hbuf + hbuf_len) {
+                       struct bpf_hdr *hp;
+                       struct pktap_header *pktaphdr;
+                       
+                       hp = (struct bpf_hdr *)(void *)p;
+                       pktaphdr = (struct pktap_header *)(void *)
+                           (p + BPF_WORDALIGN(hp->bh_hdrlen));
+
+                       if (pktaphdr->pth_flags & PTH_FLAG_DELAY_PKTAP)
+                               pktap_finalize_proc_info(pktaphdr);
+
+                       if (pktaphdr->pth_flags & PTH_FLAG_TSTAMP) {
+                               hp->bh_tstamp.tv_sec =
+                                       pktaphdr->pth_tstamp.tv_sec;
+                               hp->bh_tstamp.tv_usec =
+                                       pktaphdr->pth_tstamp.tv_usec;
+                       }
+
+                       p += BPF_WORDALIGN(hp->bh_hdrlen + hp->bh_caplen);
+               }
+       }
+#endif
+
        /*
         * Move data from hold buffer into user space.
         * We know the entire buffer is transferred since
         * we checked above that the read buffer is bpf_bufsize bytes.
         */
        /*
         * Move data from hold buffer into user space.
         * We know the entire buffer is transferred since
         * we checked above that the read buffer is bpf_bufsize bytes.
         */
-       error = UIOMOVE(d->bd_hbuf, d->bd_hlen, UIO_READ, uio);
-
+       error = UIOMOVE(hbuf, hbuf_len, UIO_READ, uio);
+       
+       lck_mtx_lock(bpf_mlock);
+       /*
+        * Make sure device is still opened
+        */
+       if ((d->bd_flags & BPF_CLOSING) != 0) {
+               bpf_release_d(d);
+               lck_mtx_unlock(bpf_mlock);
+               return (ENXIO);
+       }
+       
+       d->bd_hbuf_read = 0;
        d->bd_fbuf = d->bd_hbuf;
        d->bd_hbuf = NULL;
        d->bd_hlen = 0;
        d->bd_fbuf = d->bd_hbuf;
        d->bd_hbuf = NULL;
        d->bd_hlen = 0;
+       d->bd_hcnt = 0;
+       wakeup((caddr_t)d);
+
+       bpf_release_d(d);
        lck_mtx_unlock(bpf_mlock);
        return (error);
        lck_mtx_unlock(bpf_mlock);
        return (error);
+
 }
 
 
 }
 
 
@@ -821,25 +1229,48 @@ bpfread(dev_t dev, struct uio *uio, int ioflag)
 static void
 bpf_wakeup(struct bpf_d *d)
 {
 static void
 bpf_wakeup(struct bpf_d *d)
 {
+       if (d->bd_state == BPF_WAITING) {
+               bpf_stop_timer(d);
+               d->bd_state = BPF_IDLE;
+       }
        wakeup((caddr_t)d);
        if (d->bd_async && d->bd_sig && d->bd_sigio)
                pgsigio(d->bd_sigio, d->bd_sig);
 
        wakeup((caddr_t)d);
        if (d->bd_async && d->bd_sig && d->bd_sigio)
                pgsigio(d->bd_sigio, d->bd_sig);
 
-#if BSD >= 199103
        selwakeup(&d->bd_sel);
        selwakeup(&d->bd_sel);
-       KNOTE(&d->bd_sel.si_note, 1);
-#ifndef __APPLE__
-       /* XXX */
-       d->bd_sel.si_pid = 0;
-#endif
-#else
-       if (d->bd_selproc) {
-               selwakeup(d->bd_selproc, (int)d->bd_selcoll);
-               d->bd_selcoll = 0;
-               d->bd_selproc = 0;
+       if ((d->bd_flags & BPF_KNOTE))
+               KNOTE(&d->bd_sel.si_note, 1);
+}
+
+
+static void
+bpf_timed_out(void *arg, __unused void *dummy)
+{
+       struct bpf_d *d = (struct bpf_d *)arg;
+
+       lck_mtx_lock(bpf_mlock);
+       if (d->bd_state == BPF_WAITING) {
+               /*
+                * There's a select or kqueue waiting for this; if there's 
+                * now stuff to read, wake it up.
+                */
+               d->bd_state = BPF_TIMED_OUT;
+               if (d->bd_slen != 0)
+                       bpf_wakeup(d);
+       } else if (d->bd_state == BPF_DRAINING) {
+               /*
+                * A close is waiting for this to finish.
+                * Mark it as finished, and wake the close up.
+                */
+               d->bd_state = BPF_IDLE;
+               bpf_wakeup(d);
        }
        }
-#endif
+       lck_mtx_unlock(bpf_mlock);
 }
 }
+       
+
+
+
 
 /* keep in sync with bpf_movein above: */
 #define MAX_DATALINK_HDR_LEN   (sizeof(struct firewire_header))
 
 /* keep in sync with bpf_movein above: */
 #define MAX_DATALINK_HDR_LEN   (sizeof(struct firewire_header))
@@ -853,67 +1284,109 @@ bpfwrite(dev_t dev, struct uio *uio, __unused int ioflag)
        int error;
        char              dst_buf[SOCKADDR_HDR_LEN + MAX_DATALINK_HDR_LEN];
        int datlen = 0;
        int error;
        char              dst_buf[SOCKADDR_HDR_LEN + MAX_DATALINK_HDR_LEN];
        int datlen = 0;
+       int bif_dlt;
+       int bd_hdrcmplt;
 
        lck_mtx_lock(bpf_mlock);
 
        d = bpf_dtab[minor(dev)];
 
        lck_mtx_lock(bpf_mlock);
 
        d = bpf_dtab[minor(dev)];
-       if (d == 0 || d == (void *)1) {
+       if (d == 0 || d == (void *)1 || (d->bd_flags & BPF_CLOSING) != 0) {
                lck_mtx_unlock(bpf_mlock);
                return (ENXIO);
        }
                lck_mtx_unlock(bpf_mlock);
                return (ENXIO);
        }
+
+       bpf_acquire_d(d);
+
        if (d->bd_bif == 0) {
        if (d->bd_bif == 0) {
+               bpf_release_d(d);
                lck_mtx_unlock(bpf_mlock);
                return (ENXIO);
        }
 
        ifp = d->bd_bif->bif_ifp;
 
                lck_mtx_unlock(bpf_mlock);
                return (ENXIO);
        }
 
        ifp = d->bd_bif->bif_ifp;
 
+       if ((ifp->if_flags & IFF_UP) == 0) {
+               bpf_release_d(d);
+               lck_mtx_unlock(bpf_mlock);
+               return (ENETDOWN);
+       }
        if (uio_resid(uio) == 0) {
        if (uio_resid(uio) == 0) {
+               bpf_release_d(d);
                lck_mtx_unlock(bpf_mlock);
                return (0);
        }
        ((struct sockaddr *)dst_buf)->sa_len = sizeof(dst_buf);
                lck_mtx_unlock(bpf_mlock);
                return (0);
        }
        ((struct sockaddr *)dst_buf)->sa_len = sizeof(dst_buf);
-       error = bpf_movein(uio, (int)d->bd_bif->bif_dlt, &m, 
-                          d->bd_hdrcmplt ? NULL : (struct sockaddr *)dst_buf,
-                          &datlen);
+
+       /*
+        * fix for PR-6849527
+        * geting variables onto stack before dropping lock for bpf_movein()
+        */
+       bif_dlt = (int)d->bd_bif->bif_dlt;
+       bd_hdrcmplt  = d->bd_hdrcmplt;
+
+       /* bpf_movein allocating mbufs; drop lock */
+       lck_mtx_unlock(bpf_mlock);
+
+       error = bpf_movein(uio, bif_dlt, &m, 
+       bd_hdrcmplt ? NULL : (struct sockaddr *)dst_buf,
+       &datlen);
+
+       /* take the lock again */
+       lck_mtx_lock(bpf_mlock);
        if (error) {
        if (error) {
+               bpf_release_d(d);
                lck_mtx_unlock(bpf_mlock);
                return (error);
        }
 
                lck_mtx_unlock(bpf_mlock);
                return (error);
        }
 
-       if ((unsigned)datlen > ifp->if_mtu) {
+       /* verify the device is still open */
+       if ((d->bd_flags & BPF_CLOSING) != 0) {
+               bpf_release_d(d);
                lck_mtx_unlock(bpf_mlock);
                m_freem(m);
                lck_mtx_unlock(bpf_mlock);
                m_freem(m);
-               return (EMSGSIZE);
+               return (ENXIO);
        }
        }
-       
-       if ((error = ifp_use(ifp, kIfNetUseCount_MustNotBeZero)) != 0) {
+
+       if (d->bd_bif == NULL) {
+               bpf_release_d(d);
+               lck_mtx_unlock(bpf_mlock);
+               m_free(m);
+               return (ENXIO);
+       }
+
+       if ((unsigned)datlen > ifp->if_mtu) {
+               bpf_release_d(d);
                lck_mtx_unlock(bpf_mlock);
                m_freem(m);
                lck_mtx_unlock(bpf_mlock);
                m_freem(m);
-               return (error);
+               return (EMSGSIZE);
        }
 
        }
 
+
 #if CONFIG_MACF_NET
        mac_mbuf_label_associate_bpfdesc(d, m);
 #endif
 #if CONFIG_MACF_NET
        mac_mbuf_label_associate_bpfdesc(d, m);
 #endif
+
+       bpf_set_packet_service_class(m, d->bd_traffic_class);
+
        lck_mtx_unlock(bpf_mlock);
 
        lck_mtx_unlock(bpf_mlock);
 
+       /*
+        * The driver frees the mbuf.
+        */
        if (d->bd_hdrcmplt) {
                if (d->bd_bif->bif_send)
                        error = d->bd_bif->bif_send(ifp, d->bd_bif->bif_dlt, m);
                else
        if (d->bd_hdrcmplt) {
                if (d->bd_bif->bif_send)
                        error = d->bd_bif->bif_send(ifp, d->bd_bif->bif_dlt, m);
                else
-                       error = dlil_output(ifp, 0, m, NULL, NULL, 1);
-       }
-       else {
-               error = dlil_output(ifp, PF_INET, m, NULL, (struct sockaddr *)dst_buf, 0);
+                       error = dlil_output(ifp, 0, m, NULL, NULL, 1, NULL);
+       } else {
+               error = dlil_output(ifp, PF_INET, m, NULL,
+                   (struct sockaddr *)dst_buf, 0, NULL);
        }
        }
-       
-       if (ifp_unuse(ifp) != 0)
-               ifp_use_reached_zero(ifp);
-       
-       /*
-        * The driver frees the mbuf.
-        */
+
+       lck_mtx_lock(bpf_mlock);
+       bpf_release_d(d);
+       lck_mtx_unlock(bpf_mlock);
+
        return (error);
 }
 
        return (error);
 }
 
@@ -924,6 +1397,9 @@ bpfwrite(dev_t dev, struct uio *uio, __unused int ioflag)
 static void
 reset_d(struct bpf_d *d)
 {
 static void
 reset_d(struct bpf_d *d)
 {
+       if (d->bd_hbuf_read)
+               panic("resetting buffers during read");
+
        if (d->bd_hbuf) {
                /* Free the hold buffer. */
                d->bd_fbuf = d->bd_hbuf;
        if (d->bd_hbuf) {
                /* Free the hold buffer. */
                d->bd_fbuf = d->bd_hbuf;
@@ -931,6 +1407,8 @@ reset_d(struct bpf_d *d)
        }
        d->bd_slen = 0;
        d->bd_hlen = 0;
        }
        d->bd_slen = 0;
        d->bd_hlen = 0;
+       d->bd_scnt = 0;
+       d->bd_hcnt = 0;
        d->bd_rcount = 0;
        d->bd_dcount = 0;
 }
        d->bd_rcount = 0;
        d->bd_dcount = 0;
 }
@@ -954,6 +1432,11 @@ reset_d(struct bpf_d *d)
  *  BIOCSHDRCMPLT      Set "header already complete" flag
  *  BIOCGSEESENT       Get "see packets sent" flag
  *  BIOCSSEESENT       Set "see packets sent" flag
  *  BIOCSHDRCMPLT      Set "header already complete" flag
  *  BIOCGSEESENT       Get "see packets sent" flag
  *  BIOCSSEESENT       Set "see packets sent" flag
+ *  BIOCSETTC          Set traffic class.
+ *  BIOCGETTC          Get traffic class.
+ *  BIOCSEXTHDR                Set "extended header" flag
+ *  BIOCSHEADDROP      Drop head of the buffer if user is not reading
+ *  BIOCGHEADDROP      Get "head-drop" flag
  */
 /* ARGSUSED */
 int
  */
 /* ARGSUSED */
 int
@@ -962,15 +1445,23 @@ bpfioctl(dev_t dev, u_long cmd, caddr_t addr, __unused int flags,
 {
        struct bpf_d *d;
        int error = 0;
 {
        struct bpf_d *d;
        int error = 0;
+       u_int int_arg;
+       struct ifreq ifr;
 
        lck_mtx_lock(bpf_mlock);
 
        d = bpf_dtab[minor(dev)];
 
        lck_mtx_lock(bpf_mlock);
 
        d = bpf_dtab[minor(dev)];
-       if (d == 0 || d == (void *)1) {
+       if (d == 0 || d == (void *)1 || (d->bd_flags & BPF_CLOSING) != 0) {
                lck_mtx_unlock(bpf_mlock);
                return (ENXIO);
        }
 
                lck_mtx_unlock(bpf_mlock);
                return (ENXIO);
        }
 
+       bpf_acquire_d(d);
+
+       if (d->bd_state == BPF_WAITING)
+               bpf_stop_timer(d);
+       d->bd_state = BPF_IDLE;
+
        switch (cmd) {
 
        default:
        switch (cmd) {
 
        default:
@@ -980,19 +1471,19 @@ bpfioctl(dev_t dev, u_long cmd, caddr_t addr, __unused int flags,
        /*
         * Check for read packet available.
         */
        /*
         * Check for read packet available.
         */
-       case FIONREAD:
+       case FIONREAD:                  /* int */
                {
                        int n;
 
                        n = d->bd_slen;
                {
                        int n;
 
                        n = d->bd_slen;
-                       if (d->bd_hbuf)
+                       if (d->bd_hbuf && d->bd_hbuf_read == 0)
                                n += d->bd_hlen;
 
                                n += d->bd_hlen;
 
-                       *(int *)addr = n;
+                       bcopy(&n, addr, sizeof (n));
                        break;
                }
 
                        break;
                }
 
-       case SIOCGIFADDR:
+       case SIOCGIFADDR:               /* struct ifreq */
                {
                        struct ifnet *ifp;
 
                {
                        struct ifnet *ifp;
 
@@ -1008,44 +1499,56 @@ bpfioctl(dev_t dev, u_long cmd, caddr_t addr, __unused int flags,
        /*
         * Get buffer len [for read()].
         */
        /*
         * Get buffer len [for read()].
         */
-       case BIOCGBLEN:
-               *(u_int *)addr = d->bd_bufsize;
+       case BIOCGBLEN:                 /* u_int */
+               bcopy(&d->bd_bufsize, addr, sizeof (u_int));
                break;
 
        /*
         * Set buffer length.
         */
                break;
 
        /*
         * Set buffer length.
         */
-       case BIOCSBLEN:
-#if BSD < 199103
-               error = EINVAL;
-#else
+       case BIOCSBLEN:                 /* u_int */
                if (d->bd_bif != 0)
                        error = EINVAL;
                else {
                if (d->bd_bif != 0)
                        error = EINVAL;
                else {
-                       u_int size = *(u_int *)addr;
+                       u_int size;
+
+                       bcopy(addr, &size, sizeof (size));
 
 
-                       if (size > bpf_maxbufsize)
-                               *(u_int *)addr = size = bpf_maxbufsize;
+                       /*
+                        * Allow larger buffer in head drop mode with the
+                        * assumption the capture is in standby mode to
+                        * keep a cache of recent traffic
+                        */
+                       if (d->bd_headdrop != 0 && size > 2 * bpf_maxbufsize)
+                               size = 2 * bpf_maxbufsize;
+                       else if (size > bpf_maxbufsize)
+                               size = bpf_maxbufsize;
                        else if (size < BPF_MINBUFSIZE)
                        else if (size < BPF_MINBUFSIZE)
-                               *(u_int *)addr = size = BPF_MINBUFSIZE;
+                               size = BPF_MINBUFSIZE;
+                       bcopy(&size, addr, sizeof (size));
                        d->bd_bufsize = size;
                }
                        d->bd_bufsize = size;
                }
-#endif
                break;
 
        /*
         * Set link layer read filter.
         */
                break;
 
        /*
         * Set link layer read filter.
         */
-       case BIOCSETF32: {
-               struct bpf_program32 *prg32 = (struct bpf_program32 *)addr;
-               error = bpf_setf(d, prg32->bf_len,
-                   CAST_USER_ADDR_T(prg32->bf_insns));
+       case BIOCSETF32:
+       case BIOCSETFNR32: {            /* struct bpf_program32 */
+               struct bpf_program32 prg32;
+
+               bcopy(addr, &prg32, sizeof (prg32));
+               error = bpf_setf(d, prg32.bf_len,
+                   CAST_USER_ADDR_T(prg32.bf_insns), cmd);
                break;
        }
 
                break;
        }
 
-       case BIOCSETF64: {
-               struct bpf_program64 *prg64 = (struct bpf_program64 *)addr;
-               error = bpf_setf(d, prg64->bf_len, prg64->bf_insns);
+       case BIOCSETF64:
+       case BIOCSETFNR64: {            /* struct bpf_program64 */
+               struct bpf_program64 prg64;
+
+               bcopy(addr, &prg64, sizeof (prg64));
+               error = bpf_setf(d, prg64.bf_len, prg64.bf_insns, cmd);
                break;
        }
 
                break;
        }
 
@@ -1053,6 +1556,13 @@ bpfioctl(dev_t dev, u_long cmd, caddr_t addr, __unused int flags,
         * Flush read packet buffer.
         */
        case BIOCFLUSH:
         * Flush read packet buffer.
         */
        case BIOCFLUSH:
+               while (d->bd_hbuf_read) {
+                       msleep((caddr_t)d, bpf_mlock, PRINET, "bpf_reading", NULL);
+               }
+               if ((d->bd_flags & BPF_CLOSING) != 0) {
+                       error = ENXIO;
+                       break;
+               }
                reset_d(d);
                break;
 
                reset_d(d);
                break;
 
@@ -1079,56 +1589,61 @@ bpfioctl(dev_t dev, u_long cmd, caddr_t addr, __unused int flags,
        /*
         * Get device parameters.
         */
        /*
         * Get device parameters.
         */
-       case BIOCGDLT:
+       case BIOCGDLT:                  /* u_int */
                if (d->bd_bif == 0)
                        error = EINVAL;
                else
                if (d->bd_bif == 0)
                        error = EINVAL;
                else
-                       *(u_int *)addr = d->bd_bif->bif_dlt;
+                       bcopy(&d->bd_bif->bif_dlt, addr, sizeof (u_int));
                break;
 
        /*
         * Get a list of supported data link types.
         */
                break;
 
        /*
         * Get a list of supported data link types.
         */
-       case BIOCGDLTLIST:
+       case BIOCGDLTLIST:              /* struct bpf_dltlist */
                if (d->bd_bif == NULL) {
                        error = EINVAL;
                } else {
                if (d->bd_bif == NULL) {
                        error = EINVAL;
                } else {
-                       error = bpf_getdltlist(d,
-                           (struct bpf_dltlist *)addr, p);
+                       error = bpf_getdltlist(d, addr, p);
                }
                break;
 
        /*
         * Set data link type.
         */
                }
                break;
 
        /*
         * Set data link type.
         */
-       case BIOCSDLT:
-                       if (d->bd_bif == NULL)
-                                       error = EINVAL;
-                       else
-                                       error = bpf_setdlt(d, *(u_int *)addr);
-                       break;
+       case BIOCSDLT:                  /* u_int */
+               if (d->bd_bif == NULL) {
+                       error = EINVAL;
+               } else {
+                       u_int dlt;
+
+                       bcopy(addr, &dlt, sizeof (dlt));
+                       error = bpf_setdlt(d, dlt);
+               }
+               break;
 
        /*
         * Get interface name.
         */
 
        /*
         * Get interface name.
         */
-       case BIOCGETIF:
+       case BIOCGETIF:                 /* struct ifreq */
                if (d->bd_bif == 0)
                        error = EINVAL;
                else {
                        struct ifnet *const ifp = d->bd_bif->bif_ifp;
                if (d->bd_bif == 0)
                        error = EINVAL;
                else {
                        struct ifnet *const ifp = d->bd_bif->bif_ifp;
-                       struct ifreq *const ifr = (struct ifreq *)addr;
 
 
-                       snprintf(ifr->ifr_name, sizeof(ifr->ifr_name),
-                           "%s%d", ifp->if_name, ifp->if_unit);
+                       snprintf(((struct ifreq *)(void *)addr)->ifr_name,
+                           sizeof (ifr.ifr_name), "%s", if_name(ifp));
                }
                break;
 
        /*
         * Set interface.
         */
                }
                break;
 
        /*
         * Set interface.
         */
-       case BIOCSETIF: {
+       case BIOCSETIF: {               /* struct ifreq */
                ifnet_t ifp;
                ifnet_t ifp;
-               ifp = ifunit(((struct ifreq *)addr)->ifr_name);
+
+               bcopy(addr, &ifr, sizeof (ifr));
+               ifr.ifr_name[IFNAMSIZ - 1] = '\0';
+               ifp = ifunit(ifr.ifr_name);
                if (ifp == NULL)
                        error = ENXIO;
                else
                if (ifp == NULL)
                        error = ENXIO;
                else
@@ -1139,96 +1654,145 @@ bpfioctl(dev_t dev, u_long cmd, caddr_t addr, __unused int flags,
        /*
         * Set read timeout.
         */
        /*
         * Set read timeout.
         */
-       case BIOCSRTIMEOUT:
-               {
-                       struct BPF_TIMEVAL *_tv = (struct BPF_TIMEVAL *)addr;
-                       struct timeval tv;
+       case BIOCSRTIMEOUT32: {         /* struct user32_timeval */
+               struct user32_timeval _tv;
+               struct timeval tv;
 
 
-                       tv.tv_sec  = _tv->tv_sec;
-                       tv.tv_usec = _tv->tv_usec;
+               bcopy(addr, &_tv, sizeof (_tv));
+               tv.tv_sec  = _tv.tv_sec;
+               tv.tv_usec = _tv.tv_usec;
 
 
-                       /*
-                        * Subtract 1 tick from tvtohz() since this isn't
-                        * a one-shot timer.
-                        */
-                       if ((error = itimerfix(&tv)) == 0)
-                               d->bd_rtout = tvtohz(&tv) - 1;
-                       break;
-               }
+               /*
+                * Subtract 1 tick from tvtohz() since this isn't
+                * a one-shot timer.
+                */
+               if ((error = itimerfix(&tv)) == 0)
+                       d->bd_rtout = tvtohz(&tv) - 1;
+               break;
+       }
+
+       case BIOCSRTIMEOUT64: {         /* struct user64_timeval */
+               struct user64_timeval _tv;
+               struct timeval tv;
+
+               bcopy(addr, &_tv, sizeof (_tv));
+               tv.tv_sec  = _tv.tv_sec;
+               tv.tv_usec = _tv.tv_usec;
+
+               /*
+                * Subtract 1 tick from tvtohz() since this isn't
+                * a one-shot timer.
+                */
+               if ((error = itimerfix(&tv)) == 0)
+                       d->bd_rtout = tvtohz(&tv) - 1;
+               break;
+       }
 
        /*
         * Get read timeout.
         */
 
        /*
         * Get read timeout.
         */
-       case BIOCGRTIMEOUT:
-               {
-                       struct BPF_TIMEVAL *tv = (struct BPF_TIMEVAL *)addr;
+       case BIOCGRTIMEOUT32: {         /* struct user32_timeval */
+               struct user32_timeval tv;
 
 
-                       tv->tv_sec = d->bd_rtout / hz;
-                       tv->tv_usec = (d->bd_rtout % hz) * tick;
-                       break;
-               }
+               bzero(&tv, sizeof (tv));
+               tv.tv_sec = d->bd_rtout / hz;
+               tv.tv_usec = (d->bd_rtout % hz) * tick;
+               bcopy(&tv, addr, sizeof (tv));
+               break;
+       }
+
+       case BIOCGRTIMEOUT64: {         /* struct user64_timeval */
+               struct user64_timeval tv;
+
+               bzero(&tv, sizeof (tv));
+               tv.tv_sec = d->bd_rtout / hz;
+               tv.tv_usec = (d->bd_rtout % hz) * tick;
+               bcopy(&tv, addr, sizeof (tv));
+               break;
+       }
 
        /*
         * Get packet stats.
         */
 
        /*
         * Get packet stats.
         */
-       case BIOCGSTATS:
-               {
-                       struct bpf_stat *bs = (struct bpf_stat *)addr;
+       case BIOCGSTATS: {              /* struct bpf_stat */
+               struct bpf_stat bs;
 
 
-                       bs->bs_recv = d->bd_rcount;
-                       bs->bs_drop = d->bd_dcount;
-                       break;
-               }
+               bzero(&bs, sizeof (bs));
+               bs.bs_recv = d->bd_rcount;
+               bs.bs_drop = d->bd_dcount;
+               bcopy(&bs, addr, sizeof (bs));
+               break;
+       }
 
        /*
         * Set immediate mode.
         */
 
        /*
         * Set immediate mode.
         */
-       case BIOCIMMEDIATE:
-               d->bd_immediate = *(u_int *)addr;
+       case BIOCIMMEDIATE:             /* u_int */
+               d->bd_immediate = *(u_int *)(void *)addr;
                break;
 
                break;
 
-       case BIOCVERSION:
-               {
-                       struct bpf_version *bv = (struct bpf_version *)addr;
+       case BIOCVERSION: {             /* struct bpf_version */
+               struct bpf_version bv;
 
 
-                       bv->bv_major = BPF_MAJOR_VERSION;
-                       bv->bv_minor = BPF_MINOR_VERSION;
-                       break;
-               }
+               bzero(&bv, sizeof (bv));
+               bv.bv_major = BPF_MAJOR_VERSION;
+               bv.bv_minor = BPF_MINOR_VERSION;
+               bcopy(&bv, addr, sizeof (bv));
+               break;
+       }
 
        /*
         * Get "header already complete" flag
         */
 
        /*
         * Get "header already complete" flag
         */
-       case BIOCGHDRCMPLT:
-               *(u_int *)addr = d->bd_hdrcmplt;
+       case BIOCGHDRCMPLT:             /* u_int */
+               bcopy(&d->bd_hdrcmplt, addr, sizeof (u_int));
                break;
 
        /*
         * Set "header already complete" flag
         */
                break;
 
        /*
         * Set "header already complete" flag
         */
-       case BIOCSHDRCMPLT:
-               d->bd_hdrcmplt = *(u_int *)addr ? 1 : 0;
+       case BIOCSHDRCMPLT:             /* u_int */
+               bcopy(addr, &int_arg, sizeof (int_arg));
+               d->bd_hdrcmplt = int_arg ? 1 : 0;
                break;
 
        /*
         * Get "see sent packets" flag
         */
                break;
 
        /*
         * Get "see sent packets" flag
         */
-       case BIOCGSEESENT:
-               *(u_int *)addr = d->bd_seesent;
+       case BIOCGSEESENT:              /* u_int */
+               bcopy(&d->bd_seesent, addr, sizeof (u_int));
                break;
 
        /*
         * Set "see sent packets" flag
         */
                break;
 
        /*
         * Set "see sent packets" flag
         */
-       case BIOCSSEESENT:
-               d->bd_seesent = *(u_int *)addr;
+       case BIOCSSEESENT:              /* u_int */
+               bcopy(addr, &d->bd_seesent, sizeof (u_int));
+               break;
+
+       /*
+        * Set traffic service class
+        */
+       case BIOCSETTC: {               /* int */
+               int tc;
+
+               bcopy(addr, &tc, sizeof (int));
+               error = bpf_set_traffic_class(d, tc);
+               break;
+       }
+
+       /*
+        * Get traffic service class
+        */
+       case BIOCGETTC:                 /* int */
+               bcopy(&d->bd_traffic_class, addr, sizeof (int));
                break;
 
                break;
 
-       case FIONBIO:           /* Non-blocking I/O */
+       case FIONBIO:           /* Non-blocking I/O; int */
                break;
 
                break;
 
-       case FIOASYNC:          /* Send signal on receive packets */
-               d->bd_async = *(int *)addr;
+       case FIOASYNC:          /* Send signal on receive packets; int */
+               bcopy(addr, &d->bd_async, sizeof (int));
                break;
 #ifndef __APPLE__
        case FIOSETOWN:
                break;
 #ifndef __APPLE__
        case FIOSETOWN:
@@ -1249,23 +1813,78 @@ bpfioctl(dev_t dev, u_long cmd, caddr_t addr, __unused int flags,
                *(int *)addr = -fgetown(d->bd_sigio);
                break;
 #endif
                *(int *)addr = -fgetown(d->bd_sigio);
                break;
 #endif
-       case BIOCSRSIG:         /* Set receive signal */
-               {
-                       u_int sig;
+       case BIOCSRSIG: {       /* Set receive signal; u_int */
+               u_int sig;
 
 
-                       sig = *(u_int *)addr;
+               bcopy(addr, &sig, sizeof (u_int));
 
 
-                       if (sig >= NSIG)
-                               error = EINVAL;
-                       else
-                               d->bd_sig = sig;
+               if (sig >= NSIG)
+                       error = EINVAL;
+               else
+                       d->bd_sig = sig;
+               break;
+       }
+       case BIOCGRSIG:                 /* u_int */
+               bcopy(&d->bd_sig, addr, sizeof (u_int));
+               break;
+#ifdef __APPLE__
+       case BIOCSEXTHDR:               /* u_int */
+               bcopy(addr, &int_arg, sizeof (int_arg));
+               if (int_arg)
+                       d->bd_flags |= BPF_EXTENDED_HDR;
+               else
+                       d->bd_flags &= ~BPF_EXTENDED_HDR;
+               break;
+
+       case BIOCGIFATTACHCOUNT: {              /* struct ifreq */
+               ifnet_t ifp;
+               struct bpf_if *bp;
+
+               bcopy(addr, &ifr, sizeof (ifr));
+               ifr.ifr_name[IFNAMSIZ - 1] = '\0';
+               ifp = ifunit(ifr.ifr_name);
+               if (ifp == NULL) {
+                       error = ENXIO;
                        break;
                }
                        break;
                }
-       case BIOCGRSIG:
-               *(u_int *)addr = d->bd_sig;
+               ifr.ifr_intval = 0;
+               for (bp = bpf_iflist; bp != 0; bp = bp->bif_next) {
+                       struct bpf_d *bpf_d;
+                       
+                       if (bp->bif_ifp == NULL || bp->bif_ifp != ifp)
+                               continue;
+                       for (bpf_d = bp->bif_dlist; bpf_d; bpf_d = bpf_d->bd_next) {
+                               ifr.ifr_intval += 1;
+                       }
+               }
+               bcopy(&ifr, addr, sizeof (ifr));
                break;
        }
                break;
        }
-       
+       case BIOCGWANTPKTAP:                    /* u_int */
+               int_arg = d->bd_flags & BPF_WANT_PKTAP ? 1 : 0;
+               bcopy(&int_arg, addr, sizeof (int_arg));
+               break;
+
+       case BIOCSWANTPKTAP:                    /* u_int */
+                bcopy(addr, &int_arg, sizeof (int_arg));
+                if (int_arg)
+                        d->bd_flags |= BPF_WANT_PKTAP;
+                else
+                        d->bd_flags &= ~BPF_WANT_PKTAP;
+               break;
+#endif
+
+       case BIOCSHEADDROP:
+               bcopy(addr, &int_arg, sizeof (int_arg));
+               d->bd_headdrop = int_arg ? 1 : 0;
+               break;
+
+       case BIOCGHEADDROP:
+               bcopy(&d->bd_headdrop, addr, sizeof (int));
+               break;
+       }
+
+       bpf_release_d(d);
        lck_mtx_unlock(bpf_mlock);
 
        return (error);
        lck_mtx_unlock(bpf_mlock);
 
        return (error);
@@ -1276,11 +1895,18 @@ bpfioctl(dev_t dev, u_long cmd, caddr_t addr, __unused int flags,
  * free it and replace it.  Returns EINVAL for bogus requests.
  */
 static int
  * free it and replace it.  Returns EINVAL for bogus requests.
  */
 static int
-bpf_setf(struct bpf_d *d, u_int bf_len, user_addr_t bf_insns)
+bpf_setf(struct bpf_d *d, u_int bf_len, user_addr_t bf_insns,
+    u_long cmd)
 {
        struct bpf_insn *fcode, *old;
        u_int flen, size;
 
 {
        struct bpf_insn *fcode, *old;
        u_int flen, size;
 
+       while (d->bd_hbuf_read) 
+               msleep((caddr_t)d, bpf_mlock, PRINET, "bpf_reading", NULL);
+
+       if ((d->bd_flags & BPF_CLOSING) != 0)
+               return (ENXIO);
+       
        old = d->bd_filter;
        if (bf_insns == USER_ADDR_NULL) {
                if (bf_len != 0)
        old = d->bd_filter;
        if (bf_insns == USER_ADDR_NULL) {
                if (bf_len != 0)
@@ -1304,7 +1930,10 @@ bpf_setf(struct bpf_d *d, u_int bf_len, user_addr_t bf_insns)
        if (copyin(bf_insns, (caddr_t)fcode, size) == 0 &&
            bpf_validate(fcode, (int)flen)) {
                d->bd_filter = fcode;
        if (copyin(bf_insns, (caddr_t)fcode, size) == 0 &&
            bpf_validate(fcode, (int)flen)) {
                d->bd_filter = fcode;
-               reset_d(d);
+       
+               if (cmd == BIOCSETF32 || cmd == BIOCSETF64)
+                       reset_d(d);
+       
                if (old != 0)
                        FREE((caddr_t)old, M_DEVBUF);
 
                if (old != 0)
                        FREE((caddr_t)old, M_DEVBUF);
 
@@ -1324,7 +1953,13 @@ bpf_setif(struct bpf_d *d, ifnet_t theywant, u_int32_t dlt)
 {
        struct bpf_if *bp;
        int error;
 {
        struct bpf_if *bp;
        int error;
-       
+
+       while (d->bd_hbuf_read)
+               msleep((caddr_t)d, bpf_mlock, PRINET, "bpf_reading", NULL);
+
+       if ((d->bd_flags & BPF_CLOSING) != 0)
+               return (ENXIO);
+
        /*
         * Look through attached interfaces for the named one.
         */
        /*
         * Look through attached interfaces for the named one.
         */
@@ -1333,31 +1968,30 @@ bpf_setif(struct bpf_d *d, ifnet_t theywant, u_int32_t dlt)
 
                if (ifp == 0 || ifp != theywant || (dlt != 0 && dlt != bp->bif_dlt))
                        continue;
 
                if (ifp == 0 || ifp != theywant || (dlt != 0 && dlt != bp->bif_dlt))
                        continue;
+               /*
+                * If the process knows how to deal with DLT_PKTAP, use it
+                * by default
+                */
+               if (dlt == 0 && bp->bif_dlt == DLT_PKTAP &&
+                       !(d->bd_flags & BPF_WANT_PKTAP))
+                       continue;
                /*
                 * We found the requested interface.
                /*
                 * We found the requested interface.
-                * If it's not up, return an error.
-                * Allocate the packet buffers if we need to.
-                * If we're already attached to requested interface,
-                * just flush the buffer.
+                * Allocate the packet buffers.
+                */
+               error = bpf_allocbufs(d);
+               if (error != 0)
+                       return (error);
+               /*
+                * Detach if attached to something else.
                 */
                 */
-               if ((ifp->if_flags & IFF_UP) == 0)
-                       return (ENETDOWN);
-
-               if (d->bd_sbuf == 0) {
-                       error = bpf_allocbufs(d);
-                       if (error != 0)
-                               return (error);
-               }
                if (bp != d->bd_bif) {
                if (bp != d->bd_bif) {
-                       if (d->bd_bif)
-                               /*
-                                * Detach if attached to something else.
-                                */
-                               bpf_detachd(d);
-
-                       if (bpf_attachd(d, bp) != 0) {
-                               return ENXIO;
+                       if (d->bd_bif != NULL) {
+                               if (bpf_detachd(d, 0) != 0)
+                                       return (ENXIO);
                        }
                        }
+                       if (bpf_attachd(d, bp) != 0)
+                               return (ENXIO);
                }
                reset_d(d);
                return (0);
                }
                reset_d(d);
                return (0);
@@ -1372,37 +2006,49 @@ bpf_setif(struct bpf_d *d, ifnet_t theywant, u_int32_t dlt)
  * Get a list of available data link type of the interface.
  */
 static int
  * Get a list of available data link type of the interface.
  */
 static int
-bpf_getdltlist(struct bpf_d *d, struct bpf_dltlist *bfl, struct proc *p)
+bpf_getdltlist(struct bpf_d *d, caddr_t addr, struct proc *p)
 {
        u_int           n;
        int             error;
        struct ifnet    *ifp;
        struct bpf_if   *bp;
        user_addr_t     dlist;
 {
        u_int           n;
        int             error;
        struct ifnet    *ifp;
        struct bpf_if   *bp;
        user_addr_t     dlist;
+       struct bpf_dltlist bfl;
 
 
+       bcopy(addr, &bfl, sizeof (bfl));
        if (proc_is64bit(p)) {
        if (proc_is64bit(p)) {
-               dlist = (user_addr_t)bfl->bfl_u.bflu_pad;
+               dlist = (user_addr_t)bfl.bfl_u.bflu_pad;
        } else {
        } else {
-               dlist = CAST_USER_ADDR_T(bfl->bfl_u.bflu_list);
+               dlist = CAST_USER_ADDR_T(bfl.bfl_u.bflu_list);
        }
 
        ifp = d->bd_bif->bif_ifp;
        n = 0;
        error = 0;
        }
 
        ifp = d->bd_bif->bif_ifp;
        n = 0;
        error = 0;
+
        for (bp = bpf_iflist; bp; bp = bp->bif_next) {
                if (bp->bif_ifp != ifp)
                        continue;
        for (bp = bpf_iflist; bp; bp = bp->bif_next) {
                if (bp->bif_ifp != ifp)
                        continue;
+               /* 
+                * Return DLT_PKTAP only to processes that know how to handle it
+                */
+               if (bp->bif_dlt == DLT_PKTAP && !(d->bd_flags & BPF_WANT_PKTAP))
+                       continue;
                if (dlist != USER_ADDR_NULL) {
                if (dlist != USER_ADDR_NULL) {
-                       if (n >= bfl->bfl_len) {
+                       if (n >= bfl.bfl_len) {
                                return (ENOMEM);
                        }
                        error = copyout(&bp->bif_dlt, dlist,
                            sizeof (bp->bif_dlt));
                                return (ENOMEM);
                        }
                        error = copyout(&bp->bif_dlt, dlist,
                            sizeof (bp->bif_dlt));
+                       if (error != 0)
+                               break;
                        dlist += sizeof (bp->bif_dlt);
                }
                n++;
        }
                        dlist += sizeof (bp->bif_dlt);
                }
                n++;
        }
-       bfl->bfl_len = n;
+       bfl.bfl_len = n;
+       bcopy(&bfl, addr, sizeof (bfl));
+
        return (error);
 }
 
        return (error);
 }
 
@@ -1411,8 +2057,6 @@ bpf_getdltlist(struct bpf_d *d, struct bpf_dltlist *bfl, struct proc *p)
  */
 static int
 bpf_setdlt(struct bpf_d *d, uint32_t dlt)
  */
 static int
 bpf_setdlt(struct bpf_d *d, uint32_t dlt)
-       
-
 {
        int error, opromisc;
        struct ifnet *ifp;
 {
        int error, opromisc;
        struct ifnet *ifp;
@@ -1420,6 +2064,13 @@ bpf_setdlt(struct bpf_d *d, uint32_t dlt)
        
        if (d->bd_bif->bif_dlt == dlt)
                return (0);
        
        if (d->bd_bif->bif_dlt == dlt)
                return (0);
+       
+       while (d->bd_hbuf_read)
+               msleep((caddr_t)d, bpf_mlock, PRINET, "bpf_reading", NULL);
+
+       if ((d->bd_flags & BPF_CLOSING) != 0)
+               return (ENXIO);
+
        ifp = d->bd_bif->bif_ifp;
        for (bp = bpf_iflist; bp; bp = bp->bif_next) {
                if (bp->bif_ifp == ifp && bp->bif_dlt == dlt)
        ifp = d->bd_bif->bif_ifp;
        for (bp = bpf_iflist; bp; bp = bp->bif_next) {
                if (bp->bif_ifp == ifp && bp->bif_dlt == dlt)
@@ -1427,7 +2078,8 @@ bpf_setdlt(struct bpf_d *d, uint32_t dlt)
        }
        if (bp != NULL) {
                opromisc = d->bd_promisc;
        }
        if (bp != NULL) {
                opromisc = d->bd_promisc;
-               bpf_detachd(d);
+               if (bpf_detachd(d, 0) != 0)
+                       return (ENXIO);
                error = bpf_attachd(d, bp);
                if (error) {
                        printf("bpf_setdlt: bpf_attachd %s%d failed (%d)\n",
                error = bpf_attachd(d, bp);
                if (error) {
                        printf("bpf_setdlt: bpf_attachd %s%d failed (%d)\n",
@@ -1439,16 +2091,41 @@ bpf_setdlt(struct bpf_d *d, uint32_t dlt)
                        lck_mtx_unlock(bpf_mlock);
                        error = ifnet_set_promiscuous(bp->bif_ifp, 1);
                        lck_mtx_lock(bpf_mlock);
                        lck_mtx_unlock(bpf_mlock);
                        error = ifnet_set_promiscuous(bp->bif_ifp, 1);
                        lck_mtx_lock(bpf_mlock);
-                       if (error)
-                               printf("bpf_setdlt: ifpromisc %s%d failed (%d)\n",
-                                          ifnet_name(bp->bif_ifp), ifnet_unit(bp->bif_ifp), error);
-                       else
+                       if (error) {
+                               printf("%s: ifpromisc %s%d failed (%d)\n",
+                                   __func__, ifnet_name(bp->bif_ifp),
+                                   ifnet_unit(bp->bif_ifp), error);
+                       } else {
                                d->bd_promisc = 1;
                                d->bd_promisc = 1;
+                       }
                }
        }
        return (bp == NULL ? EINVAL : 0);
 }
 
                }
        }
        return (bp == NULL ? EINVAL : 0);
 }
 
+static int
+bpf_set_traffic_class(struct bpf_d *d, int tc)
+{
+       int error = 0;
+
+       if (!SO_VALID_TC(tc))
+               error = EINVAL;
+       else
+               d->bd_traffic_class = tc;
+
+       return (error);
+}
+
+static void
+bpf_set_packet_service_class(struct mbuf *m, int tc)
+{
+       if (!(m->m_flags & M_PKTHDR))
+               return;
+
+       VERIFY(SO_VALID_TC(tc));
+       (void) m_set_service_class(m, so_tc2msc(tc));
+}
+
 /*
  * Support for select()
  *
 /*
  * Support for select()
  *
@@ -1456,38 +2133,65 @@ bpf_setdlt(struct bpf_d *d, uint32_t dlt)
  * Otherwise, return false but make a note that a selwakeup() must be done.
  */
 int
  * Otherwise, return false but make a note that a selwakeup() must be done.
  */
 int
-bpfpoll(dev_t dev, int events, void * wql, struct proc *p)
+bpfselect(dev_t dev, int which, void * wql, struct proc *p)
 {
        struct bpf_d *d;
 {
        struct bpf_d *d;
-       int revents = 0;
+       int ret = 0;
 
        lck_mtx_lock(bpf_mlock);
 
        d = bpf_dtab[minor(dev)];
 
        lck_mtx_lock(bpf_mlock);
 
        d = bpf_dtab[minor(dev)];
-       if (d == 0 || d == (void *)1) {
+       if (d == 0 || d == (void *)1 || (d->bd_flags & BPF_CLOSING) != 0) {
                lck_mtx_unlock(bpf_mlock);
                return (ENXIO);
        }
 
                lck_mtx_unlock(bpf_mlock);
                return (ENXIO);
        }
 
-       /*
-        * An imitation of the FIONREAD ioctl code.
-        */
+       bpf_acquire_d(d);
+
        if (d->bd_bif == NULL) {
        if (d->bd_bif == NULL) {
+               bpf_release_d(d);
                lck_mtx_unlock(bpf_mlock);
                return (ENXIO);
        }
 
                lck_mtx_unlock(bpf_mlock);
                return (ENXIO);
        }
 
-       if (events & (POLLIN | POLLRDNORM)) {
-               if (d->bd_hlen != 0 || (d->bd_immediate && d->bd_slen != 0))
-                       revents |= events & (POLLIN | POLLRDNORM);
-               else
-                       selrecord(p, &d->bd_sel, wql);
+       while (d->bd_hbuf_read) 
+               msleep((caddr_t)d, bpf_mlock, PRINET, "bpf_reading", NULL);
+
+       if ((d->bd_flags & BPF_CLOSING) != 0) {
+               bpf_release_d(d);
+               lck_mtx_unlock(bpf_mlock);
+               return (ENXIO);
+       }
+
+       switch (which) {
+               case FREAD:
+                       if (d->bd_hlen != 0 ||
+                                       ((d->bd_immediate || d->bd_state == BPF_TIMED_OUT) &&
+                                        d->bd_slen != 0))
+                               ret = 1; /* read has data to return */
+                       else {
+                               /*
+                                * Read has no data to return.
+                                * Make the select wait, and start a timer if
+                                * necessary.
+                                */
+                               selrecord(p, &d->bd_sel, wql);
+                               bpf_start_timer(d);
+                       }
+                       break;
+
+               case FWRITE:
+                       ret = 1; /* can't determine whether a write would block */
+                       break;
        }
 
        }
 
+       bpf_release_d(d);
        lck_mtx_unlock(bpf_mlock);
        lck_mtx_unlock(bpf_mlock);
-       return (revents);
+
+       return (ret);
 }
 
 }
 
+
 /*
  * Support for kevent() system call.  Register EVFILT_READ filters and
  * reject all others.
 /*
  * Support for kevent() system call.  Register EVFILT_READ filters and
  * reject all others.
@@ -1495,50 +2199,115 @@ bpfpoll(dev_t dev, int events, void * wql, struct proc *p)
 int bpfkqfilter(dev_t dev, struct knote *kn);
 static void filt_bpfdetach(struct knote *);
 static int filt_bpfread(struct knote *, long);
 int bpfkqfilter(dev_t dev, struct knote *kn);
 static void filt_bpfdetach(struct knote *);
 static int filt_bpfread(struct knote *, long);
+static int filt_bpftouch(struct knote *kn, struct kevent_internal_s *kev);
+static int filt_bpfprocess(struct knote *kn, struct filt_process_s *data, struct kevent_internal_s *kev);
 
 
-static struct filterops bpfread_filtops = {
+struct filterops bpfread_filtops = {
        .f_isfd = 1, 
        .f_detach = filt_bpfdetach,
        .f_event = filt_bpfread,
        .f_isfd = 1, 
        .f_detach = filt_bpfdetach,
        .f_event = filt_bpfread,
+       .f_touch = filt_bpftouch,
+       .f_process = filt_bpfprocess,
 };
 
 };
 
+static int
+filt_bpfread_common(struct knote *kn, struct bpf_d *d)
+{
+       int ready = 0;
+
+       if (d->bd_immediate) {
+               /*
+                * If there's data in the hold buffer, it's the 
+                * amount of data a read will return.
+                *
+                * If there's no data in the hold buffer, but
+                * there's data in the store buffer, a read will
+                * immediately rotate the store buffer to the 
+                * hold buffer, the amount of data in the store
+                * buffer is the amount of data a read will 
+                * return.
+                *
+                * If there's no data in either buffer, we're not 
+                * ready to read.
+                */
+               kn->kn_data = ((d->bd_hlen == 0  || d->bd_hbuf_read) 
+                   ? d->bd_slen : d->bd_hlen);
+               int64_t lowwat = 1;
+               if (kn->kn_sfflags & NOTE_LOWAT)
+               {
+                       if (kn->kn_sdata > d->bd_bufsize)
+                               lowwat = d->bd_bufsize;
+                       else if (kn->kn_sdata > lowwat)
+                               lowwat = kn->kn_sdata;
+               }
+               ready = (kn->kn_data >= lowwat);
+       } else {
+               /*
+                * If there's data in the hold buffer, it's the 
+                * amount of data a read will return.
+                *
+                * If there's no data in the hold buffer, but 
+                * there's data in the store buffer, if the 
+                * timer has expired a read will immediately
+                * rotate the store buffer to the hold buffer,
+                * so the amount of data in the store buffer is 
+                * the amount of data a read will return.
+                *
+                * If there's no data in either buffer, or there's 
+                * no data in the hold buffer and the timer hasn't 
+                * expired, we're not ready to read.
+                */
+               kn->kn_data = ((d->bd_hlen == 0 || d->bd_hbuf_read) && d->bd_state == BPF_TIMED_OUT ? 
+                               d->bd_slen : d->bd_hlen);
+               ready = (kn->kn_data > 0);
+       }
+       if (!ready)
+               bpf_start_timer(d);
+
+       return (ready);
+}
+
 int
 bpfkqfilter(dev_t dev, struct knote *kn)
 {
        struct bpf_d *d;
 int
 bpfkqfilter(dev_t dev, struct knote *kn)
 {
        struct bpf_d *d;
+       int res;
 
        /*
         * Is this device a bpf?
         */
 
        /*
         * Is this device a bpf?
         */
-       if (major(dev) != CDEV_MAJOR) {
-               return (EINVAL);
-       }
-
-       if (kn->kn_filter != EVFILT_READ) {
-               return (EINVAL);
+       if (major(dev) != CDEV_MAJOR ||
+           kn->kn_filter != EVFILT_READ) {
+               kn->kn_flags = EV_ERROR;
+               kn->kn_data = EINVAL;
+               return 0;
        }
 
        lck_mtx_lock(bpf_mlock);
 
        d = bpf_dtab[minor(dev)];
        }
 
        lck_mtx_lock(bpf_mlock);
 
        d = bpf_dtab[minor(dev)];
-       if (d == 0 || d == (void *)1) {
-               lck_mtx_unlock(bpf_mlock);
-               return (ENXIO);
-       }
 
 
-       /*
-        * An imitation of the FIONREAD ioctl code.
-        */
-       if (d->bd_bif == NULL) {
+       if (d == 0 ||
+           d == (void *)1 ||
+           d->bd_bif == NULL ||
+           (d->bd_flags & BPF_CLOSING) != 0) {
                lck_mtx_unlock(bpf_mlock);
                lck_mtx_unlock(bpf_mlock);
-               return (ENXIO);
+               kn->kn_flags = EV_ERROR;
+               kn->kn_data = ENXIO;
+               return 0;
        }
 
        kn->kn_hook = d;
        }
 
        kn->kn_hook = d;
-       kn->kn_fop = &bpfread_filtops;
+       kn->kn_filtid = EVFILTID_BPFREAD;
        KNOTE_ATTACH(&d->bd_sel.si_note, kn);
        KNOTE_ATTACH(&d->bd_sel.si_note, kn);
+       d->bd_flags |= BPF_KNOTE;
+
+       /* capture the current state */
+       res = filt_bpfread_common(kn, d);
+
        lck_mtx_unlock(bpf_mlock);
        lck_mtx_unlock(bpf_mlock);
-       return 0;
+
+       return (res);
 }
 
 static void
 }
 
 static void
@@ -1547,42 +2316,59 @@ filt_bpfdetach(struct knote *kn)
        struct bpf_d *d = (struct bpf_d *)kn->kn_hook;
 
        lck_mtx_lock(bpf_mlock);
        struct bpf_d *d = (struct bpf_d *)kn->kn_hook;
 
        lck_mtx_lock(bpf_mlock);
-       KNOTE_DETACH(&d->bd_sel.si_note, kn);
+       if (d->bd_flags & BPF_KNOTE) {
+               KNOTE_DETACH(&d->bd_sel.si_note, kn);
+               d->bd_flags &= ~BPF_KNOTE;
+       }
        lck_mtx_unlock(bpf_mlock);
 }
 
 static int
 filt_bpfread(struct knote *kn, long hint)
 {
        lck_mtx_unlock(bpf_mlock);
 }
 
 static int
 filt_bpfread(struct knote *kn, long hint)
 {
+#pragma unused(hint)
        struct bpf_d *d = (struct bpf_d *)kn->kn_hook;
        struct bpf_d *d = (struct bpf_d *)kn->kn_hook;
-       int ready = 0;
 
 
-       if (hint == 0)
-               lck_mtx_lock(bpf_mlock);
+       return filt_bpfread_common(kn, d);
+}
 
 
-       if (d->bd_immediate) {
-               kn->kn_data = (d->bd_hlen == 0 ? d->bd_slen : d->bd_hlen);
-               ready = (kn->kn_data >= ((kn->kn_sfflags & NOTE_LOWAT) ? 
-                                       kn->kn_sdata : 1));
-       } else {
-               kn->kn_data = d->bd_hlen;
-               ready = (kn->kn_data > 0);
-       }
+static int
+filt_bpftouch(struct knote *kn, struct kevent_internal_s *kev)
+{
+       struct bpf_d *d = (struct bpf_d *)kn->kn_hook;
+       int res;
 
 
-       if (hint == 0)
-               lck_mtx_unlock(bpf_mlock);
-       return (ready);
+       lck_mtx_lock(bpf_mlock);
+
+       /* save off the lowat threshold and flag */
+       kn->kn_sdata = kev->data;
+       kn->kn_sfflags = kev->fflags;
+       if ((kn->kn_status & KN_UDATA_SPECIFIC) == 0)
+               kn->kn_udata = kev->udata;
+
+       /* output data will be re-generated here */
+       res = filt_bpfread_common(kn, d);
+
+       lck_mtx_unlock(bpf_mlock);
+
+       return res;
 }
 
 }
 
-static inline void*
-_cast_non_const(const void * ptr) {
-       union {
-               const void*             cval;
-               void*                   val;
-       } ret;
-       
-       ret.cval = ptr;
-       return (ret.val);
+static int
+filt_bpfprocess(struct knote *kn, struct filt_process_s *data, struct kevent_internal_s *kev)
+{
+#pragma unused(data)
+       struct bpf_d *d = (struct bpf_d *)kn->kn_hook;
+       int res;
+
+       lck_mtx_lock(bpf_mlock);
+       res = filt_bpfread_common(kn, d);
+       if (res) {
+               *kev = kn->kn_kevent;
+       }
+       lck_mtx_unlock(bpf_mlock);
+
+       return res;
 }
 
 /*
 }
 
 /*
@@ -1592,7 +2378,7 @@ _cast_non_const(const void * ptr) {
 static void
 bpf_mcopy(const void *src_arg, void *dst_arg, size_t len)
 {
 static void
 bpf_mcopy(const void *src_arg, void *dst_arg, size_t len)
 {
-       struct mbuf *m = _cast_non_const(src_arg);
+       struct mbuf *m = (struct mbuf *)(uintptr_t)(src_arg);
        u_int count;
        u_char *dst;
 
        u_int count;
        u_char *dst;
 
@@ -1615,9 +2401,10 @@ bpf_tap_imp(
        mbuf_t          m,
        void*           hdr,
        size_t          hlen,
        mbuf_t          m,
        void*           hdr,
        size_t          hlen,
-       int                     outbound)
+       int             outbound)
 {
        struct bpf_if *bp;
 {
        struct bpf_if *bp;
+       struct mbuf *savedm = m;
 
        /*
         * It's possible that we get here after the bpf descriptor has been
 
        /*
         * It's possible that we get here after the bpf descriptor has been
@@ -1668,7 +2455,7 @@ bpf_tap_imp(
                        hack_hdr.mh_type = m->m_type;
                        hack_hdr.mh_flags = 0;
                        
                        hack_hdr.mh_type = m->m_type;
                        hack_hdr.mh_flags = 0;
                        
-                       m = (mbuf_t)&hack_hdr;
+                       __IGNORE_WCASTALIGN(m = (mbuf_t)&hack_hdr);
                }
 
                for (m0 = m; m0 != 0; m0 = m0->m_next)
                }
 
                for (m0 = m; m0 != 0; m0 = m0->m_next)
@@ -1684,7 +2471,8 @@ bpf_tap_imp(
                                if (mac_bpfdesc_check_receive(d, bp->bif_ifp) != 0)
                                        continue;
 #endif
                                if (mac_bpfdesc_check_receive(d, bp->bif_ifp) != 0)
                                        continue;
 #endif
-                               catchpacket(d, (u_char *)m, pktlen, slen, bpf_mcopy);
+                               catchpacket(d, (u_char *)m, savedm, pktlen,
+                                   slen, outbound, bpf_mcopy);
                        }
                }
        }
                        }
                }
        }
@@ -1730,12 +2518,22 @@ static int bpf_tap_callback(struct ifnet *ifp, struct mbuf *m)
  * pkt is really an mbuf.
  */
 static void
  * pkt is really an mbuf.
  */
 static void
-catchpacket(struct bpf_d *d, u_char *pkt, u_int pktlen, u_int snaplen, 
+catchpacket(struct bpf_d *d, u_char *pkt, struct mbuf *m, u_int pktlen,
+       u_int snaplen, int outbound,
        void (*cpfn)(const void *, void *, size_t))
 {
        struct bpf_hdr *hp;
        void (*cpfn)(const void *, void *, size_t))
 {
        struct bpf_hdr *hp;
+       struct bpf_hdr_ext *ehp;
        int totlen, curlen;
        int totlen, curlen;
-       int hdrlen = d->bd_bif->bif_hdrlen;
+       int hdrlen, caplen;
+       int do_wakeup = 0;
+       u_char *payload;
+       struct timeval tv;
+       struct m_tag *mt = NULL;
+       struct bpf_mtag *bt = NULL;
+
+       hdrlen = (d->bd_flags & BPF_EXTENDED_HDR) ? d->bd_bif->bif_exthdrlen :
+           d->bd_bif->bif_hdrlen;
        /*
         * Figure out how many bytes to move.  If the packet is
         * greater or equal to the snapshot length, transfer that
        /*
         * Figure out how many bytes to move.  If the packet is
         * greater or equal to the snapshot length, transfer that
@@ -1755,41 +2553,114 @@ catchpacket(struct bpf_d *d, u_char *pkt, u_int pktlen, u_int snaplen,
                 * This packet will overflow the storage buffer.
                 * Rotate the buffers if we can, then wakeup any
                 * pending reads.
                 * This packet will overflow the storage buffer.
                 * Rotate the buffers if we can, then wakeup any
                 * pending reads.
+                *
+                * We cannot rotate buffers if a read is in progress
+                * so drop the packet
                 */
                 */
-               if (d->bd_fbuf == 0) {
-                       /*
-                        * We haven't completed the previous read yet,
-                        * so drop the packet.
-                        */
+               if (d->bd_hbuf_read) {
                        ++d->bd_dcount;
                        return;
                }
                        ++d->bd_dcount;
                        return;
                }
-               ROTATE_BUFFERS(d);
-               bpf_wakeup(d);
+               
+               if (d->bd_fbuf == NULL) {
+                       if (d->bd_headdrop == 0) {
+                               /*
+                                * We haven't completed the previous read yet,
+                                * so drop the packet.
+                                */
+                               ++d->bd_dcount;
+                               return;
+                       }
+                       /*
+                        * Drop the hold buffer as it contains older packets
+                        */
+                       d->bd_dcount += d->bd_hcnt;
+                       d->bd_fbuf = d->bd_hbuf;
+                       ROTATE_BUFFERS(d);
+               } else {
+                       ROTATE_BUFFERS(d);
+               }
+               do_wakeup = 1;
                curlen = 0;
        }
                curlen = 0;
        }
-       else if (d->bd_immediate)
+       else if (d->bd_immediate || d->bd_state == BPF_TIMED_OUT)
                /*
                /*
-                * Immediate mode is set.  A packet arrived so any
-                * reads should be woken up.
+                * Immediate mode is set, or the read timeout has 
+                * already expired during a select call. A packet 
+                * arrived, so the reader should be woken up.
                 */
                 */
-               bpf_wakeup(d);
+               do_wakeup = 1;
 
        /*
         * Append the bpf header.
         */
 
        /*
         * Append the bpf header.
         */
-       hp = (struct bpf_hdr *)(d->bd_sbuf + curlen);
-       struct timeval tv;
        microtime(&tv);
        microtime(&tv);
-       hp->bh_tstamp.tv_sec = tv.tv_sec;
-       hp->bh_tstamp.tv_usec = tv.tv_usec;
-       hp->bh_datalen = pktlen;
-       hp->bh_hdrlen = hdrlen;
+       if (d->bd_flags & BPF_EXTENDED_HDR) {
+               ehp = (struct bpf_hdr_ext *)(void *)(d->bd_sbuf + curlen);
+               memset(ehp, 0, sizeof(*ehp));
+               ehp->bh_tstamp.tv_sec = tv.tv_sec;
+               ehp->bh_tstamp.tv_usec = tv.tv_usec;
+               ehp->bh_datalen = pktlen;
+               ehp->bh_hdrlen = hdrlen;
+               ehp->bh_caplen = totlen - hdrlen;
+               mt = m_tag_locate(m, bpf_mtag_id, 0, NULL);
+               if (mt && mt->m_tag_len >= sizeof(*bt)) {
+                       bt = (struct bpf_mtag *)(mt + 1);
+                       ehp->bh_pid = bt->bt_pid;
+                       strlcpy(ehp->bh_comm, bt->bt_comm,
+                           sizeof(ehp->bh_comm));
+                       ehp->bh_svc = so_svc2tc(bt->bt_svc);
+                       if (bt->bt_direction == BPF_MTAG_DIR_OUT)
+                               ehp->bh_flags |= BPF_HDR_EXT_FLAGS_DIR_OUT;
+                       else
+                               ehp->bh_flags |= BPF_HDR_EXT_FLAGS_DIR_IN;
+                       m_tag_delete(m, mt);
+               } else if (outbound) {
+                       /* only do lookups on non-raw INPCB */
+                       if ((m->m_pkthdr.pkt_flags & (PKTF_FLOW_ID|
+                           PKTF_FLOW_LOCALSRC|PKTF_FLOW_RAWSOCK)) ==
+                           (PKTF_FLOW_ID|PKTF_FLOW_LOCALSRC) &&
+                           m->m_pkthdr.pkt_flowsrc == FLOWSRC_INPCB) {
+                               ehp->bh_flowid = m->m_pkthdr.pkt_flowid;
+                               ehp->bh_proto = m->m_pkthdr.pkt_proto;
+                       }
+                       ehp->bh_svc = so_svc2tc(m->m_pkthdr.pkt_svc);
+                       ehp->bh_flags |= BPF_HDR_EXT_FLAGS_DIR_OUT;
+                       if (m->m_pkthdr.pkt_flags & PKTF_TCP_REXMT)
+                               ehp->bh_pktflags |= BPF_PKTFLAGS_TCP_REXMT;
+                       if (m->m_pkthdr.pkt_flags & PKTF_START_SEQ)
+                               ehp->bh_pktflags |= BPF_PKTFLAGS_START_SEQ;
+                       if (m->m_pkthdr.pkt_flags & PKTF_LAST_PKT)
+                               ehp->bh_pktflags |= BPF_PKTFLAGS_LAST_PKT;
+                       if (m->m_pkthdr.pkt_flags & PKTF_VALID_UNSENT_DATA) {
+                               ehp->bh_unsent_bytes =
+                                   m->m_pkthdr.bufstatus_if;
+                               ehp->bh_unsent_snd =
+                                   m->m_pkthdr.bufstatus_sndbuf;
+                       }
+               } else
+                       ehp->bh_flags |= BPF_HDR_EXT_FLAGS_DIR_IN;
+               payload = (u_char *)ehp + hdrlen;
+               caplen = ehp->bh_caplen;
+       } else {
+               hp = (struct bpf_hdr *)(void *)(d->bd_sbuf + curlen);
+               hp->bh_tstamp.tv_sec = tv.tv_sec;
+               hp->bh_tstamp.tv_usec = tv.tv_usec;
+               hp->bh_datalen = pktlen;
+               hp->bh_hdrlen = hdrlen;
+               hp->bh_caplen = totlen - hdrlen;
+               payload = (u_char *)hp + hdrlen;
+               caplen = hp->bh_caplen;
+       }
        /*
         * Copy the packet data into the store buffer and update its length.
         */
        /*
         * Copy the packet data into the store buffer and update its length.
         */
-       (*cpfn)(pkt, (u_char *)hp + hdrlen, (hp->bh_caplen = totlen - hdrlen));
+       (*cpfn)(pkt, payload, caplen);
        d->bd_slen = curlen + totlen;
        d->bd_slen = curlen + totlen;
+       d->bd_scnt += 1;
+
+       if (do_wakeup)
+               bpf_wakeup(d);
 }
 
 /*
 }
 
 /*
@@ -1798,17 +2669,33 @@ catchpacket(struct bpf_d *d, u_char *pkt, u_int pktlen, u_int snaplen,
 static int
 bpf_allocbufs(struct bpf_d *d)
 {
 static int
 bpf_allocbufs(struct bpf_d *d)
 {
+       if (d->bd_sbuf != NULL) {
+               FREE(d->bd_sbuf, M_DEVBUF);
+               d->bd_sbuf = NULL;
+       }
+       if (d->bd_hbuf != NULL) {
+               FREE(d->bd_hbuf, M_DEVBUF);
+               d->bd_hbuf = NULL;
+       }
+       if (d->bd_fbuf != NULL) {
+               FREE(d->bd_fbuf, M_DEVBUF);
+               d->bd_fbuf = NULL;
+       }
+
        d->bd_fbuf = (caddr_t) _MALLOC(d->bd_bufsize, M_DEVBUF, M_WAIT);
        d->bd_fbuf = (caddr_t) _MALLOC(d->bd_bufsize, M_DEVBUF, M_WAIT);
-       if (d->bd_fbuf == 0)
+       if (d->bd_fbuf == NULL)
                return (ENOBUFS);
 
        d->bd_sbuf = (caddr_t) _MALLOC(d->bd_bufsize, M_DEVBUF, M_WAIT);
                return (ENOBUFS);
 
        d->bd_sbuf = (caddr_t) _MALLOC(d->bd_bufsize, M_DEVBUF, M_WAIT);
-       if (d->bd_sbuf == 0) {
+       if (d->bd_sbuf == NULL) {
                FREE(d->bd_fbuf, M_DEVBUF);
                FREE(d->bd_fbuf, M_DEVBUF);
+               d->bd_fbuf = NULL;
                return (ENOBUFS);
        }
        d->bd_slen = 0;
        d->bd_hlen = 0;
                return (ENOBUFS);
        }
        d->bd_slen = 0;
        d->bd_hlen = 0;
+       d->bd_scnt = 0;
+       d->bd_hcnt = 0;
        return (0);
 }
 
        return (0);
 }
 
@@ -1824,9 +2711,12 @@ bpf_freed(struct bpf_d *d)
         * been detached from its interface and it yet hasn't been marked
         * free.
         */
         * been detached from its interface and it yet hasn't been marked
         * free.
         */
+       if (d->bd_hbuf_read)
+               panic("bpf buffer freed during read");
+
        if (d->bd_sbuf != 0) {
                FREE(d->bd_sbuf, M_DEVBUF);
        if (d->bd_sbuf != 0) {
                FREE(d->bd_sbuf, M_DEVBUF);
-               if (d->bd_hbuf != 0)
+               if (d->bd_hbuf != 0) 
                        FREE(d->bd_hbuf, M_DEVBUF);
                if (d->bd_fbuf != 0)
                        FREE(d->bd_fbuf, M_DEVBUF);
                        FREE(d->bd_hbuf, M_DEVBUF);
                if (d->bd_fbuf != 0)
                        FREE(d->bd_fbuf, M_DEVBUF);
@@ -1858,7 +2748,8 @@ bpf_attach(
        struct bpf_if *bp_temp;
        struct bpf_if *bp_first = NULL;
        
        struct bpf_if *bp_temp;
        struct bpf_if *bp_first = NULL;
        
-       bp_new = (struct bpf_if *) _MALLOC(sizeof(*bp_new), M_DEVBUF, M_WAIT);
+       bp_new = (struct bpf_if *) _MALLOC(sizeof(*bp_new), M_DEVBUF,
+           M_WAIT | M_ZERO);
        if (bp_new == 0)
                panic("bpfattach");
 
        if (bp_new == 0)
                panic("bpfattach");
 
@@ -1875,14 +2766,13 @@ bpf_attach(
        }
        
        if (bp_temp != NULL) {
        }
        
        if (bp_temp != NULL) {
-               printf("bpfattach - %s%d with dlt %d is already attached\n",
-                       ifp->if_name, ifp->if_unit, dlt);
+               printf("bpfattach - %s with dlt %d is already attached\n",
+                       if_name(ifp), dlt);
                FREE(bp_new, M_DEVBUF);
                lck_mtx_unlock(bpf_mlock);
                return EEXIST;
        }
        
                FREE(bp_new, M_DEVBUF);
                lck_mtx_unlock(bpf_mlock);
                return EEXIST;
        }
        
-       bzero(bp_new, sizeof(*bp_new));
        bp_new->bif_ifp = ifp;
        bp_new->bif_dlt = dlt;
        bp_new->bif_send = send;
        bp_new->bif_ifp = ifp;
        bp_new->bif_dlt = dlt;
        bp_new->bif_send = send;
@@ -1906,6 +2796,8 @@ bpf_attach(
         * performance reasons and to alleviate alignment restrictions).
         */
        bp_new->bif_hdrlen = BPF_WORDALIGN(hdrlen + SIZEOF_BPF_HDR) - hdrlen;
         * performance reasons and to alleviate alignment restrictions).
         */
        bp_new->bif_hdrlen = BPF_WORDALIGN(hdrlen + SIZEOF_BPF_HDR) - hdrlen;
+       bp_new->bif_exthdrlen = BPF_WORDALIGN(hdrlen +
+           sizeof(struct bpf_hdr_ext)) - hdrlen;
        
        /* Take a reference on the interface */
        ifnet_reference(ifp);
        
        /* Take a reference on the interface */
        ifnet_reference(ifp);
@@ -1914,7 +2806,7 @@ bpf_attach(
 
 #ifndef __APPLE__
        if (bootverbose)
 
 #ifndef __APPLE__
        if (bootverbose)
-               printf("bpf: %s%d attached\n", ifp->if_name, ifp->if_unit);
+               printf("bpf: %s attached\n", if_name(ifp));
 #endif
 
        return 0;
 #endif
 
        return 0;
@@ -1930,42 +2822,49 @@ void
 bpfdetach(struct ifnet *ifp)
 {
        struct bpf_if   *bp, *bp_prev, *bp_next;
 bpfdetach(struct ifnet *ifp)
 {
        struct bpf_if   *bp, *bp_prev, *bp_next;
-       struct bpf_if   *bp_free = NULL;
        struct bpf_d    *d;
 
        struct bpf_d    *d;
 
-       
+       if (bpf_debug != 0)
+               printf("%s: %s\n",
+                   __func__, if_name(ifp));
+
        lck_mtx_lock(bpf_mlock);
 
        lck_mtx_lock(bpf_mlock);
 
-       /* Locate BPF interface information */
+       /*
+        * Build the list of devices attached to that interface
+        * that we need to free while keeping the lock to maintain
+        * the integrity of the interface list
+        */
        bp_prev = NULL;
        for (bp = bpf_iflist; bp != NULL; bp = bp_next) {
                bp_next = bp->bif_next;
        bp_prev = NULL;
        for (bp = bpf_iflist; bp != NULL; bp = bp_next) {
                bp_next = bp->bif_next;
+
                if (ifp != bp->bif_ifp) {
                        bp_prev = bp;
                        continue;
                }
                if (ifp != bp->bif_ifp) {
                        bp_prev = bp;
                        continue;
                }
-               
-               while ((d = bp->bif_dlist) != NULL) {
-                       bpf_detachd(d);
-                       bpf_wakeup(d);
-               }
-       
-               if (bp_prev) {
+               /* Unlink from the interface list */
+               if (bp_prev)
                        bp_prev->bif_next = bp->bif_next;
                        bp_prev->bif_next = bp->bif_next;
-               } else {
+               else
                        bpf_iflist = bp->bif_next;
                        bpf_iflist = bp->bif_next;
+
+               /* Detach the devices attached to the interface */
+               while ((d = bp->bif_dlist) != NULL) {
+                       /*
+                        * Take an extra reference to prevent the device
+                        * from being freed when bpf_detachd() releases
+                        * the reference for the interface list
+                        */
+                       bpf_acquire_d(d);
+                       bpf_detachd(d, 0);
+                       bpf_wakeup(d);
+                       bpf_release_d(d);
                }
                }
-               
-               bp->bif_next = bp_free;
-               bp_free = bp;
-               
                ifnet_release(ifp);
        }
 
        lck_mtx_unlock(bpf_mlock);
                ifnet_release(ifp);
        }
 
        lck_mtx_unlock(bpf_mlock);
-
-       FREE(bp, M_DEVBUF);
-
 }
 
 void
 }
 
 void
@@ -1977,25 +2876,12 @@ bpf_init(__unused void *unused)
 
        if (bpf_devsw_installed == 0) {
                bpf_devsw_installed = 1;
 
        if (bpf_devsw_installed == 0) {
                bpf_devsw_installed = 1;
-
-        bpf_mlock_grp_attr = lck_grp_attr_alloc_init();
-
-        bpf_mlock_grp = lck_grp_alloc_init("bpf", bpf_mlock_grp_attr);
-
-        bpf_mlock_attr = lck_attr_alloc_init();
-
-        bpf_mlock = lck_mtx_alloc_init(bpf_mlock_grp, bpf_mlock_attr);
-
-               if (bpf_mlock == 0) {
-                       printf("bpf_init: failed to allocate bpf_mlock\n");
-                       bpf_devsw_installed = 0;
-                       return;
-               }
-               
+               bpf_mlock_grp_attr = lck_grp_attr_alloc_init();
+               bpf_mlock_grp = lck_grp_alloc_init("bpf", bpf_mlock_grp_attr);
+               bpf_mlock_attr = lck_attr_alloc_init();
+               lck_mtx_init(bpf_mlock, bpf_mlock_grp, bpf_mlock_attr);
                maj = cdevsw_add(CDEV_MAJOR, &bpf_cdevsw);
                if (maj == -1) {
                maj = cdevsw_add(CDEV_MAJOR, &bpf_cdevsw);
                if (maj == -1) {
-                       if (bpf_mlock)
-                               lck_mtx_free(bpf_mlock, bpf_mlock_grp);
                        if (bpf_mlock_attr)
                                lck_attr_free(bpf_mlock_attr);
                        if (bpf_mlock_grp)
                        if (bpf_mlock_attr)
                                lck_attr_free(bpf_mlock_attr);
                        if (bpf_mlock_grp)
@@ -2014,6 +2900,8 @@ bpf_init(__unused void *unused)
 
                for (i = 0 ; i < NBPFILTER; i++)
                        bpf_make_dev_t(maj);
 
                for (i = 0 ; i < NBPFILTER; i++)
                        bpf_make_dev_t(maj);
+
+               VERIFY(mbuf_tag_id_find(BPF_CONTROL_NAME, &bpf_mtag_id) == 0);
        }
 #else
        cdevsw_add(&bpf_cdevsw);
        }
 #else
        cdevsw_add(&bpf_cdevsw);