Home | History | Annotate | Line # | Download | only in net
bpf.c revision 1.105.6.1
      1  1.105.6.1      yamt /*	$NetBSD: bpf.c,v 1.105.6.1 2005/03/19 08:36:31 yamt Exp $	*/
      2       1.13       cgd 
      3       1.12   mycroft /*
      4       1.12   mycroft  * Copyright (c) 1990, 1991, 1993
      5       1.12   mycroft  *	The Regents of the University of California.  All rights reserved.
      6        1.1       cgd  *
      7        1.1       cgd  * This code is derived from the Stanford/CMU enet packet filter,
      8        1.1       cgd  * (net/enet.c) distributed as part of 4.3BSD, and code contributed
      9       1.36  christos  * to Berkeley by Steven McCanne and Van Jacobson both of Lawrence
     10        1.2       cgd  * Berkeley Laboratory.
     11        1.1       cgd  *
     12        1.1       cgd  * Redistribution and use in source and binary forms, with or without
     13        1.1       cgd  * modification, are permitted provided that the following conditions
     14        1.1       cgd  * are met:
     15        1.1       cgd  * 1. Redistributions of source code must retain the above copyright
     16        1.1       cgd  *    notice, this list of conditions and the following disclaimer.
     17        1.1       cgd  * 2. Redistributions in binary form must reproduce the above copyright
     18        1.1       cgd  *    notice, this list of conditions and the following disclaimer in the
     19        1.1       cgd  *    documentation and/or other materials provided with the distribution.
     20       1.83       agc  * 3. Neither the name of the University nor the names of its contributors
     21        1.1       cgd  *    may be used to endorse or promote products derived from this software
     22        1.1       cgd  *    without specific prior written permission.
     23        1.1       cgd  *
     24        1.1       cgd  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
     25        1.1       cgd  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
     26        1.1       cgd  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
     27        1.1       cgd  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
     28        1.1       cgd  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
     29        1.1       cgd  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
     30        1.1       cgd  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
     31        1.1       cgd  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
     32        1.1       cgd  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
     33        1.1       cgd  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
     34        1.1       cgd  * SUCH DAMAGE.
     35        1.1       cgd  *
     36       1.39      fvdl  *	@(#)bpf.c	8.4 (Berkeley) 1/9/95
     37       1.36  christos  * static char rcsid[] =
     38       1.36  christos  * "Header: bpf.c,v 1.67 96/09/26 22:00:52 leres Exp ";
     39        1.1       cgd  */
     40       1.63     lukem 
     41       1.63     lukem #include <sys/cdefs.h>
     42  1.105.6.1      yamt __KERNEL_RCSID(0, "$NetBSD: bpf.c,v 1.105.6.1 2005/03/19 08:36:31 yamt Exp $");
     43        1.1       cgd 
     44        1.1       cgd #include <sys/param.h>
     45        1.1       cgd #include <sys/systm.h>
     46        1.1       cgd #include <sys/mbuf.h>
     47        1.1       cgd #include <sys/buf.h>
     48       1.12   mycroft #include <sys/time.h>
     49        1.1       cgd #include <sys/proc.h>
     50        1.1       cgd #include <sys/user.h>
     51        1.1       cgd #include <sys/ioctl.h>
     52       1.25  christos #include <sys/conf.h>
     53       1.51     enami #include <sys/vnode.h>
     54      1.105  christos #include <sys/queue.h>
     55       1.11   deraadt 
     56        1.1       cgd #include <sys/file.h>
     57      1.105  christos #include <sys/filedesc.h>
     58        1.1       cgd #include <sys/tty.h>
     59        1.1       cgd #include <sys/uio.h>
     60       1.11   deraadt 
     61        1.1       cgd #include <sys/protosw.h>
     62        1.1       cgd #include <sys/socket.h>
     63       1.30   mycroft #include <sys/errno.h>
     64       1.30   mycroft #include <sys/kernel.h>
     65       1.30   mycroft #include <sys/poll.h>
     66       1.89  jonathan #include <sys/sysctl.h>
     67       1.30   mycroft 
     68       1.11   deraadt #include <net/if.h>
     69      1.104  christos #include <net/slip.h>
     70        1.8   mycroft 
     71        1.1       cgd #include <net/bpf.h>
     72        1.1       cgd #include <net/bpfdesc.h>
     73        1.1       cgd 
     74       1.35    scottr #include <net/if_arc.h>
     75       1.34        is #include <net/if_ether.h>
     76       1.34        is 
     77       1.12   mycroft #include <netinet/in.h>
     78       1.34        is #include <netinet/if_inarp.h>
     79       1.65  wrstuden 
     80       1.65  wrstuden #if defined(_KERNEL_OPT)
     81       1.84  wrstuden #include "opt_bpf.h"
     82      1.104  christos #include "sl.h"
     83      1.104  christos #include "strip.h"
     84       1.65  wrstuden #endif
     85       1.11   deraadt 
     86       1.55  jonathan #ifndef BPF_BUFSIZE
     87       1.87  jonathan /*
     88       1.87  jonathan  * 4096 is too small for FDDI frames. 8192 is too small for gigabit Ethernet
     89       1.87  jonathan  * jumbos (circa 9k), ATM, or Intel gig/10gig ethernet jumbos (16k).
     90       1.87  jonathan  */
     91       1.87  jonathan # define BPF_BUFSIZE 32768
     92       1.55  jonathan #endif
     93        1.1       cgd 
     94       1.12   mycroft #define PRINET  26			/* interruptible */
     95        1.2       cgd 
     96        1.1       cgd /*
     97       1.89  jonathan  * The default read buffer size, and limit for BIOCSBLEN, is sysctl'able.
     98       1.89  jonathan  * XXX the default values should be computed dynamically based
     99       1.89  jonathan  * on available memory size and available mbuf clusters.
    100        1.1       cgd  */
    101        1.2       cgd int bpf_bufsize = BPF_BUFSIZE;
    102       1.89  jonathan int bpf_maxbufsize = BPF_DFLTBUFSIZE;	/* XXX set dynamically, see above */
    103        1.1       cgd 
    104        1.1       cgd /*
    105        1.1       cgd  *  bpf_iflist is the list of interfaces; each corresponds to an ifnet
    106        1.1       cgd  *  bpf_dtab holds the descriptors, indexed by minor device #
    107        1.1       cgd  */
    108       1.12   mycroft struct bpf_if	*bpf_iflist;
    109      1.105  christos LIST_HEAD(, bpf_d) bpf_list;
    110        1.1       cgd 
    111      1.103  christos static int	bpf_allocbufs(struct bpf_d *);
    112      1.101    dyoung static void	bpf_deliver(struct bpf_if *,
    113      1.101    dyoung 		            void *(*cpfn)(void *, const void *, size_t),
    114      1.101    dyoung 			    void *, u_int, u_int, struct ifnet *);
    115      1.103  christos static void	bpf_freed(struct bpf_d *);
    116      1.103  christos static void	bpf_ifname(struct ifnet *, struct ifreq *);
    117      1.103  christos static void	*bpf_mcpy(void *, const void *, size_t);
    118      1.103  christos static int	bpf_movein(struct uio *, int, int,
    119      1.103  christos 			        struct mbuf **, struct sockaddr *);
    120      1.103  christos static void	bpf_attachd(struct bpf_d *, struct bpf_if *);
    121      1.103  christos static void	bpf_detachd(struct bpf_d *);
    122      1.103  christos static int	bpf_setif(struct bpf_d *, struct ifreq *);
    123      1.103  christos static void	bpf_timed_out(void *);
    124       1.15       cgd static __inline void
    125      1.103  christos 		bpf_wakeup(struct bpf_d *);
    126      1.103  christos static void	catchpacket(struct bpf_d *, u_char *, u_int, u_int,
    127      1.103  christos 				 void *(*)(void *, const void *, size_t));
    128      1.103  christos static void	reset_d(struct bpf_d *);
    129      1.103  christos static int	bpf_getdltlist(struct bpf_d *, struct bpf_dltlist *);
    130      1.103  christos static int	bpf_setdlt(struct bpf_d *, u_int);
    131       1.12   mycroft 
    132      1.105  christos static int	bpf_read(struct file *, off_t *, struct uio *, struct ucred *,
    133      1.105  christos     int);
    134      1.105  christos static int	bpf_write(struct file *, off_t *, struct uio *, struct ucred *,
    135      1.105  christos     int);
    136      1.105  christos static int	bpf_ioctl(struct file *, u_long, void *, struct proc *);
    137      1.105  christos static int	bpf_poll(struct file *, int, struct proc *);
    138      1.105  christos static int	bpf_close(struct file *, struct proc *);
    139      1.105  christos static int	bpf_kqfilter(struct file *, struct knote *);
    140      1.105  christos 
    141      1.105  christos static const struct fileops bpf_fileops = {
    142      1.105  christos 	bpf_read,
    143      1.105  christos 	bpf_write,
    144      1.105  christos 	bpf_ioctl,
    145      1.105  christos 	fnullop_fcntl,
    146      1.105  christos 	bpf_poll,
    147      1.105  christos 	fbadop_stat,
    148      1.105  christos 	bpf_close,
    149      1.105  christos 	bpf_kqfilter,
    150      1.105  christos };
    151      1.105  christos 
    152       1.67   gehenna dev_type_open(bpfopen);
    153       1.67   gehenna 
    154       1.67   gehenna const struct cdevsw bpf_cdevsw = {
    155      1.105  christos 	bpfopen, noclose, noread, nowrite, noioctl,
    156      1.105  christos 	nostop, notty, nopoll, nommap, nokqfilter,
    157       1.67   gehenna };
    158       1.67   gehenna 
    159       1.12   mycroft static int
    160       1.36  christos bpf_movein(uio, linktype, mtu, mp, sockp)
    161       1.53  augustss 	struct uio *uio;
    162       1.20   mycroft 	int linktype;
    163       1.36  christos 	int mtu;
    164       1.53  augustss 	struct mbuf **mp;
    165       1.53  augustss 	struct sockaddr *sockp;
    166       1.12   mycroft {
    167       1.12   mycroft 	struct mbuf *m;
    168       1.12   mycroft 	int error;
    169       1.12   mycroft 	int len;
    170       1.12   mycroft 	int hlen;
    171       1.36  christos 	int align;
    172       1.12   mycroft 
    173       1.12   mycroft 	/*
    174       1.12   mycroft 	 * Build a sockaddr based on the data link layer type.
    175       1.12   mycroft 	 * We do this at this level because the ethernet header
    176       1.12   mycroft 	 * is copied directly into the data field of the sockaddr.
    177       1.12   mycroft 	 * In the case of SLIP, there is no header and the packet
    178       1.12   mycroft 	 * is forwarded as is.
    179       1.12   mycroft 	 * Also, we are careful to leave room at the front of the mbuf
    180       1.12   mycroft 	 * for the link level header.
    181       1.12   mycroft 	 */
    182       1.12   mycroft 	switch (linktype) {
    183       1.12   mycroft 
    184       1.12   mycroft 	case DLT_SLIP:
    185       1.12   mycroft 		sockp->sa_family = AF_INET;
    186       1.12   mycroft 		hlen = 0;
    187       1.36  christos 		align = 0;
    188       1.12   mycroft 		break;
    189       1.12   mycroft 
    190       1.12   mycroft 	case DLT_PPP:
    191       1.12   mycroft 		sockp->sa_family = AF_UNSPEC;
    192       1.12   mycroft 		hlen = 0;
    193       1.36  christos 		align = 0;
    194       1.12   mycroft 		break;
    195       1.12   mycroft 
    196       1.12   mycroft 	case DLT_EN10MB:
    197       1.12   mycroft 		sockp->sa_family = AF_UNSPEC;
    198       1.12   mycroft 		/* XXX Would MAXLINKHDR be better? */
    199       1.36  christos  		/* 6(dst)+6(src)+2(type) */
    200       1.12   mycroft 		hlen = sizeof(struct ether_header);
    201       1.36  christos 		align = 2;
    202       1.17     glass 		break;
    203       1.17     glass 
    204       1.17     glass 	case DLT_ARCNET:
    205       1.17     glass 		sockp->sa_family = AF_UNSPEC;
    206       1.17     glass 		hlen = ARC_HDRLEN;
    207       1.36  christos 		align = 5;
    208       1.12   mycroft 		break;
    209       1.12   mycroft 
    210       1.12   mycroft 	case DLT_FDDI:
    211       1.56      matt 		sockp->sa_family = AF_LINK;
    212       1.56      matt 		/* XXX 4(FORMAC)+6(dst)+6(src) */
    213       1.56      matt 		hlen = 16;
    214       1.36  christos 		align = 0;
    215       1.62     bjh21 		break;
    216       1.62     bjh21 
    217       1.62     bjh21 	case DLT_ECONET:
    218       1.62     bjh21 		sockp->sa_family = AF_UNSPEC;
    219       1.62     bjh21 		hlen = 6;
    220       1.62     bjh21 		align = 2;
    221       1.12   mycroft 		break;
    222       1.12   mycroft 
    223       1.12   mycroft 	case DLT_NULL:
    224       1.12   mycroft 		sockp->sa_family = AF_UNSPEC;
    225       1.12   mycroft 		hlen = 0;
    226       1.36  christos 		align = 0;
    227       1.12   mycroft 		break;
    228       1.12   mycroft 
    229       1.12   mycroft 	default:
    230       1.12   mycroft 		return (EIO);
    231       1.12   mycroft 	}
    232       1.12   mycroft 
    233       1.12   mycroft 	len = uio->uio_resid;
    234       1.36  christos 	/*
    235       1.36  christos 	 * If there aren't enough bytes for a link level header or the
    236       1.36  christos 	 * packet length exceeds the interface mtu, return an error.
    237       1.36  christos 	 */
    238       1.36  christos 	if (len < hlen || len - hlen > mtu)
    239       1.36  christos 		return (EMSGSIZE);
    240       1.36  christos 
    241       1.36  christos 	/*
    242       1.36  christos 	 * XXX Avoid complicated buffer chaining ---
    243       1.36  christos 	 * bail if it won't fit in a single mbuf.
    244       1.36  christos 	 * (Take into account possible alignment bytes)
    245       1.36  christos 	 */
    246       1.36  christos 	if ((unsigned)len > MCLBYTES - align)
    247       1.12   mycroft 		return (EIO);
    248       1.12   mycroft 
    249       1.77      matt 	m = m_gethdr(M_WAIT, MT_DATA);
    250       1.20   mycroft 	m->m_pkthdr.rcvif = 0;
    251       1.20   mycroft 	m->m_pkthdr.len = len - hlen;
    252       1.36  christos 	if (len > MHLEN - align) {
    253       1.77      matt 		m_clget(m, M_WAIT);
    254       1.12   mycroft 		if ((m->m_flags & M_EXT) == 0) {
    255       1.12   mycroft 			error = ENOBUFS;
    256       1.12   mycroft 			goto bad;
    257       1.12   mycroft 		}
    258       1.12   mycroft 	}
    259       1.36  christos 
    260       1.36  christos 	/* Insure the data is properly aligned */
    261       1.36  christos 	if (align > 0) {
    262       1.36  christos 		m->m_data += align;
    263       1.36  christos 		m->m_len -= align;
    264       1.36  christos 	}
    265       1.36  christos 
    266      1.103  christos 	error = uiomove(mtod(m, void *), len, uio);
    267       1.36  christos 	if (error)
    268       1.36  christos 		goto bad;
    269       1.12   mycroft 	if (hlen != 0) {
    270      1.103  christos 		memcpy(sockp->sa_data, mtod(m, void *), hlen);
    271       1.12   mycroft 		m->m_data += hlen; /* XXX */
    272       1.36  christos 		len -= hlen;
    273       1.12   mycroft 	}
    274       1.36  christos 	m->m_len = len;
    275       1.36  christos 	*mp = m;
    276       1.38   mycroft 	return (0);
    277       1.38   mycroft 
    278       1.38   mycroft bad:
    279       1.12   mycroft 	m_freem(m);
    280       1.12   mycroft 	return (error);
    281       1.12   mycroft }
    282        1.1       cgd 
    283        1.1       cgd /*
    284        1.2       cgd  * Attach file to the bpf interface, i.e. make d listen on bp.
    285       1.61   thorpej  * Must be called at splnet.
    286        1.1       cgd  */
    287        1.1       cgd static void
    288        1.1       cgd bpf_attachd(d, bp)
    289        1.1       cgd 	struct bpf_d *d;
    290        1.1       cgd 	struct bpf_if *bp;
    291        1.1       cgd {
    292        1.2       cgd 	/*
    293        1.2       cgd 	 * Point d at bp, and add d to the interface's list of listeners.
    294        1.2       cgd 	 * Finally, point the driver's bpf cookie at the interface so
    295        1.2       cgd 	 * it will divert packets to bpf.
    296        1.2       cgd 	 */
    297        1.1       cgd 	d->bd_bif = bp;
    298        1.1       cgd 	d->bd_next = bp->bif_dlist;
    299        1.1       cgd 	bp->bif_dlist = d;
    300        1.1       cgd 
    301        1.1       cgd 	*bp->bif_driverp = bp;
    302        1.1       cgd }
    303        1.1       cgd 
    304        1.2       cgd /*
    305        1.2       cgd  * Detach a file from its interface.
    306        1.2       cgd  */
    307        1.1       cgd static void
    308        1.1       cgd bpf_detachd(d)
    309        1.1       cgd 	struct bpf_d *d;
    310        1.1       cgd {
    311        1.1       cgd 	struct bpf_d **p;
    312        1.1       cgd 	struct bpf_if *bp;
    313        1.1       cgd 
    314        1.1       cgd 	bp = d->bd_bif;
    315        1.1       cgd 	/*
    316        1.1       cgd 	 * Check if this descriptor had requested promiscuous mode.
    317        1.1       cgd 	 * If so, turn it off.
    318        1.1       cgd 	 */
    319        1.1       cgd 	if (d->bd_promisc) {
    320       1.18   mycroft 		int error;
    321       1.18   mycroft 
    322        1.1       cgd 		d->bd_promisc = 0;
    323       1.36  christos 		/*
    324       1.36  christos 		 * Take device out of promiscuous mode.  Since we were
    325       1.36  christos 		 * able to enter promiscuous mode, we should be able
    326       1.36  christos 		 * to turn it off.  But we can get an error if
    327       1.36  christos 		 * the interface was configured down, so only panic
    328       1.36  christos 		 * if we don't get an unexpected error.
    329       1.36  christos 		 */
    330       1.36  christos   		error = ifpromisc(bp->bif_ifp, 0);
    331       1.18   mycroft 		if (error && error != EINVAL)
    332        1.2       cgd 			panic("bpf: ifpromisc failed");
    333        1.1       cgd 	}
    334        1.2       cgd 	/* Remove d from the interface's descriptor list. */
    335        1.1       cgd 	p = &bp->bif_dlist;
    336        1.1       cgd 	while (*p != d) {
    337        1.1       cgd 		p = &(*p)->bd_next;
    338        1.1       cgd 		if (*p == 0)
    339        1.1       cgd 			panic("bpf_detachd: descriptor not in list");
    340        1.1       cgd 	}
    341        1.1       cgd 	*p = (*p)->bd_next;
    342        1.1       cgd 	if (bp->bif_dlist == 0)
    343        1.1       cgd 		/*
    344        1.1       cgd 		 * Let the driver know that there are no more listeners.
    345        1.1       cgd 		 */
    346        1.1       cgd 		*d->bd_bif->bif_driverp = 0;
    347        1.1       cgd 	d->bd_bif = 0;
    348        1.1       cgd }
    349        1.1       cgd 
    350        1.1       cgd 
    351        1.1       cgd /*
    352       1.12   mycroft  * Mark a descriptor free by making it point to itself.
    353        1.1       cgd  * This is probably cheaper than marking with a constant since
    354        1.1       cgd  * the address should be in a register anyway.
    355        1.1       cgd  */
    356        1.1       cgd 
    357        1.1       cgd /*
    358       1.46    bouyer  * bpfilterattach() is called at boot time.
    359       1.46    bouyer  */
    360       1.46    bouyer /* ARGSUSED */
    361       1.46    bouyer void
    362       1.46    bouyer bpfilterattach(n)
    363       1.46    bouyer 	int n;
    364       1.46    bouyer {
    365      1.105  christos 	LIST_INIT(&bpf_list);
    366       1.46    bouyer }
    367       1.46    bouyer 
    368       1.46    bouyer /*
    369      1.105  christos  * Open ethernet device. Clones.
    370        1.1       cgd  */
    371        1.1       cgd /* ARGSUSED */
    372        1.1       cgd int
    373       1.82      fvdl bpfopen(dev, flag, mode, p)
    374        1.1       cgd 	dev_t dev;
    375        1.1       cgd 	int flag;
    376       1.24  christos 	int mode;
    377       1.82      fvdl 	struct proc *p;
    378        1.1       cgd {
    379       1.53  augustss 	struct bpf_d *d;
    380      1.105  christos 	struct file *fp;
    381      1.105  christos 	int error, fd;
    382       1.12   mycroft 
    383      1.105  christos 	/* falloc() will use the descriptor for us. */
    384      1.105  christos 	if ((error = falloc(p, &fp, &fd)) != 0)
    385      1.105  christos 		return error;
    386        1.2       cgd 
    387      1.105  christos 	d = malloc(sizeof(*d), M_DEVBUF, M_WAITOK);
    388      1.105  christos 	(void)memset(d, 0, sizeof(*d));
    389        1.2       cgd 	d->bd_bufsize = bpf_bufsize;
    390       1.91   darrenr 	d->bd_seesent = 1;
    391       1.91   darrenr 	callout_init(&d->bd_callout);
    392        1.1       cgd 
    393      1.105  christos 	LIST_INSERT_HEAD(&bpf_list, d, bd_list);
    394      1.105  christos 
    395  1.105.6.1      yamt 	return fdclone(p, fp, fd, flag, &bpf_fileops, d);
    396        1.1       cgd }
    397        1.1       cgd 
    398        1.1       cgd /*
    399        1.1       cgd  * Close the descriptor by detaching it from its interface,
    400        1.1       cgd  * deallocating its buffers, and marking it free.
    401        1.1       cgd  */
    402        1.1       cgd /* ARGSUSED */
    403      1.105  christos static int
    404      1.105  christos bpf_close(struct file *fp, struct proc *p)
    405        1.1       cgd {
    406      1.105  christos 	struct bpf_d *d = fp->f_data;
    407       1.53  augustss 	int s;
    408        1.1       cgd 
    409       1.61   thorpej 	s = splnet();
    410       1.91   darrenr 	if (d->bd_state == BPF_WAITING)
    411       1.91   darrenr 		callout_stop(&d->bd_callout);
    412       1.91   darrenr 	d->bd_state = BPF_IDLE;
    413        1.1       cgd 	if (d->bd_bif)
    414        1.1       cgd 		bpf_detachd(d);
    415        1.1       cgd 	splx(s);
    416        1.2       cgd 	bpf_freed(d);
    417      1.105  christos 	LIST_REMOVE(d, bd_list);
    418      1.105  christos 	free(d, M_DEVBUF);
    419      1.105  christos 	fp->f_data = NULL;
    420        1.2       cgd 
    421        1.2       cgd 	return (0);
    422        1.2       cgd }
    423        1.2       cgd 
    424        1.2       cgd /*
    425        1.1       cgd  * Rotate the packet buffers in descriptor d.  Move the store buffer
    426       1.12   mycroft  * into the hold slot, and the free buffer into the store slot.
    427        1.1       cgd  * Zero the length of the new store buffer.
    428        1.1       cgd  */
    429        1.1       cgd #define ROTATE_BUFFERS(d) \
    430      1.100   darrenr 	(d)->bd_hbuf = (d)->bd_sbuf; \
    431      1.100   darrenr 	(d)->bd_hlen = (d)->bd_slen; \
    432      1.100   darrenr 	(d)->bd_sbuf = (d)->bd_fbuf; \
    433      1.100   darrenr 	(d)->bd_slen = 0; \
    434      1.100   darrenr 	(d)->bd_fbuf = 0;
    435        1.1       cgd /*
    436        1.1       cgd  *  bpfread - read next chunk of packets from buffers
    437        1.1       cgd  */
    438      1.105  christos static int
    439      1.105  christos bpf_read(struct file *fp, off_t *offp, struct uio *uio,
    440      1.105  christos     struct ucred *cred, int flags)
    441        1.1       cgd {
    442      1.105  christos 	struct bpf_d *d = fp->f_data;
    443       1.91   darrenr 	int timed_out;
    444        1.1       cgd 	int error;
    445        1.1       cgd 	int s;
    446        1.1       cgd 
    447        1.1       cgd 	/*
    448       1.12   mycroft 	 * Restrict application to use a buffer the same size as
    449        1.1       cgd 	 * as kernel buffers.
    450        1.1       cgd 	 */
    451        1.1       cgd 	if (uio->uio_resid != d->bd_bufsize)
    452        1.1       cgd 		return (EINVAL);
    453        1.1       cgd 
    454       1.61   thorpej 	s = splnet();
    455       1.91   darrenr 	if (d->bd_state == BPF_WAITING)
    456       1.91   darrenr 		callout_stop(&d->bd_callout);
    457       1.91   darrenr 	timed_out = (d->bd_state == BPF_TIMED_OUT);
    458       1.91   darrenr 	d->bd_state = BPF_IDLE;
    459        1.1       cgd 	/*
    460        1.2       cgd 	 * If the hold buffer is empty, then do a timed sleep, which
    461        1.2       cgd 	 * ends when the timeout expires or when enough packets
    462        1.2       cgd 	 * have arrived to fill the store buffer.
    463        1.1       cgd 	 */
    464        1.1       cgd 	while (d->bd_hbuf == 0) {
    465      1.105  christos 		if (fp->f_flag & FNONBLOCK) {
    466       1.93   darrenr 			if (d->bd_slen == 0) {
    467      1.100   darrenr 				splx(s);
    468       1.93   darrenr 				return (EWOULDBLOCK);
    469       1.93   darrenr 			}
    470       1.93   darrenr 			ROTATE_BUFFERS(d);
    471       1.93   darrenr 			break;
    472       1.93   darrenr 		}
    473       1.93   darrenr 
    474       1.91   darrenr 		if ((d->bd_immediate || timed_out) && d->bd_slen != 0) {
    475        1.1       cgd 			/*
    476        1.1       cgd 			 * A packet(s) either arrived since the previous
    477        1.1       cgd 			 * read or arrived while we were asleep.
    478        1.1       cgd 			 * Rotate the buffers and return what's here.
    479        1.1       cgd 			 */
    480        1.1       cgd 			ROTATE_BUFFERS(d);
    481        1.1       cgd 			break;
    482        1.1       cgd 		}
    483      1.103  christos 		error = tsleep(d, PRINET|PCATCH, "bpf",
    484       1.93   darrenr 				d->bd_rtout);
    485       1.12   mycroft 		if (error == EINTR || error == ERESTART) {
    486       1.12   mycroft 			splx(s);
    487       1.12   mycroft 			return (error);
    488       1.12   mycroft 		}
    489       1.12   mycroft 		if (error == EWOULDBLOCK) {
    490       1.12   mycroft 			/*
    491       1.12   mycroft 			 * On a timeout, return what's in the buffer,
    492       1.12   mycroft 			 * which may be nothing.  If there is something
    493       1.12   mycroft 			 * in the store buffer, we can rotate the buffers.
    494       1.12   mycroft 			 */
    495       1.12   mycroft 			if (d->bd_hbuf)
    496        1.1       cgd 				/*
    497       1.12   mycroft 				 * We filled up the buffer in between
    498       1.12   mycroft 				 * getting the timeout and arriving
    499       1.12   mycroft 				 * here, so we don't need to rotate.
    500        1.1       cgd 				 */
    501        1.1       cgd 				break;
    502       1.12   mycroft 
    503      1.100   darrenr 			if (d->bd_slen == 0) {
    504      1.100   darrenr 				splx(s);
    505      1.100   darrenr 				return (0);
    506        1.1       cgd 			}
    507      1.100   darrenr 			ROTATE_BUFFERS(d);
    508      1.100   darrenr 			break;
    509        1.1       cgd 		}
    510       1.36  christos 		if (error != 0)
    511      1.100   darrenr 			goto done;
    512        1.1       cgd 	}
    513      1.100   darrenr 	/*
    514      1.100   darrenr 	 * At this point, we know we have something in the hold slot.
    515      1.100   darrenr 	 */
    516      1.100   darrenr 	splx(s);
    517      1.100   darrenr 
    518      1.100   darrenr 	/*
    519      1.100   darrenr 	 * Move data from hold buffer into user space.
    520      1.100   darrenr 	 * We know the entire buffer is transferred since
    521      1.100   darrenr 	 * we checked above that the read buffer is bpf_bufsize bytes.
    522      1.100   darrenr 	 */
    523      1.100   darrenr 	error = uiomove(d->bd_hbuf, d->bd_hlen, uio);
    524       1.12   mycroft 
    525      1.100   darrenr 	s = splnet();
    526      1.100   darrenr 	d->bd_fbuf = d->bd_hbuf;
    527      1.100   darrenr 	d->bd_hbuf = 0;
    528      1.100   darrenr 	d->bd_hlen = 0;
    529      1.100   darrenr done:
    530      1.100   darrenr 	splx(s);
    531      1.100   darrenr 	return (error);
    532        1.1       cgd }
    533        1.1       cgd 
    534        1.1       cgd 
    535        1.1       cgd /*
    536       1.12   mycroft  * If there are processes sleeping on this descriptor, wake them up.
    537        1.1       cgd  */
    538       1.15       cgd static __inline void
    539        1.1       cgd bpf_wakeup(d)
    540       1.53  augustss 	struct bpf_d *d;
    541        1.1       cgd {
    542      1.103  christos 	wakeup(d);
    543       1.85  jdolecek 	if (d->bd_async)
    544       1.86  christos 		fownsignal(d->bd_pgid, SIGIO, 0, 0, NULL);
    545       1.23   thorpej 
    546       1.75  jdolecek 	selnotify(&d->bd_sel, 0);
    547        1.2       cgd 	/* XXX */
    548       1.76  christos 	d->bd_sel.sel_pid = 0;
    549        1.1       cgd }
    550        1.1       cgd 
    551       1.91   darrenr 
    552       1.91   darrenr static void
    553       1.91   darrenr bpf_timed_out(arg)
    554       1.91   darrenr 	void *arg;
    555       1.91   darrenr {
    556      1.103  christos 	struct bpf_d *d = arg;
    557       1.91   darrenr 	int s;
    558       1.91   darrenr 
    559       1.91   darrenr 	s = splnet();
    560       1.91   darrenr 	if (d->bd_state == BPF_WAITING) {
    561       1.91   darrenr 		d->bd_state = BPF_TIMED_OUT;
    562       1.91   darrenr 		if (d->bd_slen != 0)
    563       1.91   darrenr 			bpf_wakeup(d);
    564       1.91   darrenr 	}
    565       1.91   darrenr 	splx(s);
    566       1.91   darrenr }
    567       1.91   darrenr 
    568       1.91   darrenr 
    569      1.105  christos static int
    570      1.105  christos bpf_write(struct file *fp, off_t *offp, struct uio *uio,
    571      1.105  christos     struct ucred *cred, int flags)
    572        1.1       cgd {
    573      1.105  christos 	struct bpf_d *d = fp->f_data;
    574       1.12   mycroft 	struct ifnet *ifp;
    575       1.12   mycroft 	struct mbuf *m;
    576       1.12   mycroft 	int error, s;
    577       1.56      matt 	static struct sockaddr_storage dst;
    578        1.1       cgd 
    579       1.12   mycroft 	if (d->bd_bif == 0)
    580        1.1       cgd 		return (ENXIO);
    581        1.1       cgd 
    582       1.12   mycroft 	ifp = d->bd_bif->bif_ifp;
    583       1.11   deraadt 
    584       1.12   mycroft 	if (uio->uio_resid == 0)
    585       1.12   mycroft 		return (0);
    586       1.11   deraadt 
    587       1.56      matt 	error = bpf_movein(uio, (int)d->bd_bif->bif_dlt, ifp->if_mtu, &m,
    588       1.56      matt 		(struct sockaddr *) &dst);
    589       1.12   mycroft 	if (error)
    590       1.12   mycroft 		return (error);
    591       1.11   deraadt 
    592       1.20   mycroft 	if (m->m_pkthdr.len > ifp->if_mtu)
    593        1.1       cgd 		return (EMSGSIZE);
    594        1.1       cgd 
    595       1.40   thorpej 	if (d->bd_hdrcmplt)
    596       1.56      matt 		dst.ss_family = pseudo_AF_HDRCMPLT;
    597       1.40   thorpej 
    598       1.21   mycroft 	s = splsoftnet();
    599       1.56      matt 	error = (*ifp->if_output)(ifp, m, (struct sockaddr *) &dst, NULL);
    600        1.1       cgd 	splx(s);
    601        1.1       cgd 	/*
    602       1.12   mycroft 	 * The driver frees the mbuf.
    603        1.1       cgd 	 */
    604        1.1       cgd 	return (error);
    605        1.1       cgd }
    606        1.1       cgd 
    607        1.1       cgd /*
    608        1.2       cgd  * Reset a descriptor by flushing its packet buffer and clearing the
    609       1.61   thorpej  * receive and drop counts.  Should be called at splnet.
    610        1.1       cgd  */
    611        1.1       cgd static void
    612        1.1       cgd reset_d(d)
    613        1.1       cgd 	struct bpf_d *d;
    614        1.1       cgd {
    615        1.1       cgd 	if (d->bd_hbuf) {
    616        1.1       cgd 		/* Free the hold buffer. */
    617        1.1       cgd 		d->bd_fbuf = d->bd_hbuf;
    618        1.1       cgd 		d->bd_hbuf = 0;
    619        1.1       cgd 	}
    620        1.1       cgd 	d->bd_slen = 0;
    621        1.2       cgd 	d->bd_hlen = 0;
    622        1.1       cgd 	d->bd_rcount = 0;
    623        1.1       cgd 	d->bd_dcount = 0;
    624       1.94   darrenr 	d->bd_ccount = 0;
    625        1.1       cgd }
    626        1.1       cgd 
    627       1.36  christos #ifdef BPF_KERN_FILTER
    628       1.36  christos extern struct bpf_insn *bpf_tcp_filter;
    629       1.36  christos extern struct bpf_insn *bpf_udp_filter;
    630       1.36  christos #endif
    631       1.36  christos 
    632        1.1       cgd /*
    633        1.1       cgd  *  FIONREAD		Check for read packet available.
    634        1.1       cgd  *  BIOCGBLEN		Get buffer len [for read()].
    635        1.1       cgd  *  BIOCSETF		Set ethernet read filter.
    636        1.1       cgd  *  BIOCFLUSH		Flush read packet buffer.
    637        1.1       cgd  *  BIOCPROMISC		Put interface into promiscuous mode.
    638        1.1       cgd  *  BIOCGDLT		Get link layer type.
    639        1.1       cgd  *  BIOCGETIF		Get interface name.
    640        1.1       cgd  *  BIOCSETIF		Set interface.
    641        1.1       cgd  *  BIOCSRTIMEOUT	Set read timeout.
    642        1.1       cgd  *  BIOCGRTIMEOUT	Get read timeout.
    643        1.1       cgd  *  BIOCGSTATS		Get packet stats.
    644        1.1       cgd  *  BIOCIMMEDIATE	Set immediate mode.
    645        1.2       cgd  *  BIOCVERSION		Get filter language version.
    646       1.40   thorpej  *  BIOGHDRCMPLT	Get "header already complete" flag.
    647       1.40   thorpej  *  BIOSHDRCMPLT	Set "header already complete" flag.
    648        1.1       cgd  */
    649        1.1       cgd /* ARGSUSED */
    650      1.105  christos static int
    651      1.105  christos bpf_ioctl(struct file *fp, u_long cmd, void *addr, struct proc *p)
    652        1.1       cgd {
    653      1.105  christos 	struct bpf_d *d = fp->f_data;
    654        1.1       cgd 	int s, error = 0;
    655       1.36  christos #ifdef BPF_KERN_FILTER
    656       1.53  augustss 	struct bpf_insn **p;
    657       1.36  christos #endif
    658        1.1       cgd 
    659       1.91   darrenr 	s = splnet();
    660       1.91   darrenr 	if (d->bd_state == BPF_WAITING)
    661       1.91   darrenr 		callout_stop(&d->bd_callout);
    662       1.91   darrenr 	d->bd_state = BPF_IDLE;
    663       1.91   darrenr 	splx(s);
    664       1.91   darrenr 
    665        1.1       cgd 	switch (cmd) {
    666        1.1       cgd 
    667        1.1       cgd 	default:
    668        1.1       cgd 		error = EINVAL;
    669        1.1       cgd 		break;
    670        1.1       cgd 
    671        1.1       cgd 	/*
    672        1.1       cgd 	 * Check for read packet available.
    673        1.1       cgd 	 */
    674        1.1       cgd 	case FIONREAD:
    675        1.1       cgd 		{
    676        1.1       cgd 			int n;
    677       1.12   mycroft 
    678       1.61   thorpej 			s = splnet();
    679        1.1       cgd 			n = d->bd_slen;
    680       1.12   mycroft 			if (d->bd_hbuf)
    681        1.1       cgd 				n += d->bd_hlen;
    682        1.1       cgd 			splx(s);
    683        1.1       cgd 
    684        1.1       cgd 			*(int *)addr = n;
    685        1.1       cgd 			break;
    686        1.1       cgd 		}
    687        1.1       cgd 
    688        1.1       cgd 	/*
    689        1.2       cgd 	 * Get buffer len [for read()].
    690        1.1       cgd 	 */
    691        1.2       cgd 	case BIOCGBLEN:
    692        1.2       cgd 		*(u_int *)addr = d->bd_bufsize;
    693        1.1       cgd 		break;
    694        1.2       cgd 
    695        1.1       cgd 	/*
    696        1.2       cgd 	 * Set buffer length.
    697        1.1       cgd 	 */
    698        1.2       cgd 	case BIOCSBLEN:
    699        1.2       cgd 		if (d->bd_bif != 0)
    700        1.2       cgd 			error = EINVAL;
    701        1.2       cgd 		else {
    702       1.53  augustss 			u_int size = *(u_int *)addr;
    703        1.2       cgd 
    704       1.87  jonathan 			if (size > bpf_maxbufsize)
    705       1.87  jonathan 				*(u_int *)addr = size = bpf_maxbufsize;
    706        1.2       cgd 			else if (size < BPF_MINBUFSIZE)
    707        1.2       cgd 				*(u_int *)addr = size = BPF_MINBUFSIZE;
    708        1.2       cgd 			d->bd_bufsize = size;
    709        1.2       cgd 		}
    710        1.1       cgd 		break;
    711        1.1       cgd 
    712        1.1       cgd 	/*
    713        1.2       cgd 	 * Set link layer read filter.
    714        1.1       cgd 	 */
    715       1.12   mycroft 	case BIOCSETF:
    716      1.103  christos 		error = bpf_setf(d, addr);
    717        1.1       cgd 		break;
    718        1.1       cgd 
    719       1.36  christos #ifdef BPF_KERN_FILTER
    720       1.36  christos 	/*
    721       1.36  christos 	 * Set TCP or UDP reject filter.
    722       1.36  christos 	 */
    723       1.36  christos 	case BIOCSTCPF:
    724       1.36  christos 	case BIOCSUDPF:
    725       1.36  christos 		if (!suser()) {
    726       1.36  christos 			error = EPERM;
    727       1.36  christos 			break;
    728       1.36  christos 		}
    729       1.36  christos 
    730       1.36  christos 		/* Validate and store filter */
    731      1.103  christos 		error = bpf_setf(d, addr);
    732       1.36  christos 
    733       1.36  christos 		/* Free possible old filter */
    734       1.36  christos 		if (cmd == BIOCSTCPF)
    735       1.36  christos 			p = &bpf_tcp_filter;
    736       1.36  christos 		else
    737       1.36  christos 			p = &bpf_udp_filter;
    738       1.36  christos 		if (*p != NULL)
    739      1.103  christos 			free(*p, M_DEVBUF);
    740       1.36  christos 
    741       1.36  christos 		/* Steal new filter (noop if error) */
    742       1.61   thorpej 		s = splnet();
    743       1.36  christos 		*p = d->bd_filter;
    744       1.36  christos 		d->bd_filter = NULL;
    745       1.36  christos 		splx(s);
    746       1.36  christos 		break;
    747       1.36  christos #endif
    748       1.36  christos 
    749        1.1       cgd 	/*
    750        1.1       cgd 	 * Flush read packet buffer.
    751        1.1       cgd 	 */
    752        1.1       cgd 	case BIOCFLUSH:
    753       1.61   thorpej 		s = splnet();
    754        1.1       cgd 		reset_d(d);
    755        1.1       cgd 		splx(s);
    756        1.1       cgd 		break;
    757        1.1       cgd 
    758        1.1       cgd 	/*
    759        1.1       cgd 	 * Put interface into promiscuous mode.
    760        1.1       cgd 	 */
    761        1.1       cgd 	case BIOCPROMISC:
    762        1.1       cgd 		if (d->bd_bif == 0) {
    763        1.1       cgd 			/*
    764        1.1       cgd 			 * No interface attached yet.
    765        1.1       cgd 			 */
    766        1.1       cgd 			error = EINVAL;
    767        1.1       cgd 			break;
    768        1.1       cgd 		}
    769       1.61   thorpej 		s = splnet();
    770        1.1       cgd 		if (d->bd_promisc == 0) {
    771        1.1       cgd 			error = ifpromisc(d->bd_bif->bif_ifp, 1);
    772        1.2       cgd 			if (error == 0)
    773        1.2       cgd 				d->bd_promisc = 1;
    774        1.1       cgd 		}
    775        1.1       cgd 		splx(s);
    776        1.1       cgd 		break;
    777        1.1       cgd 
    778        1.1       cgd 	/*
    779        1.1       cgd 	 * Get device parameters.
    780        1.1       cgd 	 */
    781        1.1       cgd 	case BIOCGDLT:
    782        1.1       cgd 		if (d->bd_bif == 0)
    783        1.1       cgd 			error = EINVAL;
    784        1.1       cgd 		else
    785        1.1       cgd 			*(u_int *)addr = d->bd_bif->bif_dlt;
    786        1.1       cgd 		break;
    787        1.1       cgd 
    788        1.1       cgd 	/*
    789       1.66      onoe 	 * Get a list of supported device parameters.
    790       1.66      onoe 	 */
    791       1.66      onoe 	case BIOCGDLTLIST:
    792       1.66      onoe 		if (d->bd_bif == 0)
    793       1.66      onoe 			error = EINVAL;
    794       1.66      onoe 		else
    795      1.103  christos 			error = bpf_getdltlist(d, addr);
    796       1.66      onoe 		break;
    797       1.66      onoe 
    798       1.66      onoe 	/*
    799       1.66      onoe 	 * Set device parameters.
    800       1.66      onoe 	 */
    801       1.66      onoe 	case BIOCSDLT:
    802       1.66      onoe 		if (d->bd_bif == 0)
    803       1.66      onoe 			error = EINVAL;
    804       1.66      onoe 		else
    805       1.66      onoe 			error = bpf_setdlt(d, *(u_int *)addr);
    806       1.66      onoe 		break;
    807       1.66      onoe 
    808       1.66      onoe 	/*
    809        1.1       cgd 	 * Set interface name.
    810        1.1       cgd 	 */
    811        1.1       cgd 	case BIOCGETIF:
    812        1.1       cgd 		if (d->bd_bif == 0)
    813        1.1       cgd 			error = EINVAL;
    814        1.1       cgd 		else
    815      1.103  christos 			bpf_ifname(d->bd_bif->bif_ifp, addr);
    816        1.1       cgd 		break;
    817        1.1       cgd 
    818        1.1       cgd 	/*
    819        1.1       cgd 	 * Set interface.
    820        1.1       cgd 	 */
    821        1.1       cgd 	case BIOCSETIF:
    822      1.103  christos 		error = bpf_setif(d, addr);
    823        1.1       cgd 		break;
    824        1.1       cgd 
    825        1.1       cgd 	/*
    826        1.1       cgd 	 * Set read timeout.
    827        1.1       cgd 	 */
    828       1.12   mycroft 	case BIOCSRTIMEOUT:
    829        1.1       cgd 		{
    830      1.103  christos 			struct timeval *tv = addr;
    831        1.1       cgd 
    832       1.19       cgd 			/* Compute number of ticks. */
    833       1.19       cgd 			d->bd_rtout = tv->tv_sec * hz + tv->tv_usec / tick;
    834       1.33   thorpej 			if ((d->bd_rtout == 0) && (tv->tv_usec != 0))
    835       1.33   thorpej 				d->bd_rtout = 1;
    836        1.1       cgd 			break;
    837        1.1       cgd 		}
    838        1.1       cgd 
    839        1.1       cgd 	/*
    840        1.1       cgd 	 * Get read timeout.
    841        1.1       cgd 	 */
    842       1.12   mycroft 	case BIOCGRTIMEOUT:
    843        1.1       cgd 		{
    844      1.103  christos 			struct timeval *tv = addr;
    845        1.1       cgd 
    846       1.19       cgd 			tv->tv_sec = d->bd_rtout / hz;
    847       1.19       cgd 			tv->tv_usec = (d->bd_rtout % hz) * tick;
    848        1.1       cgd 			break;
    849        1.1       cgd 		}
    850        1.1       cgd 
    851        1.1       cgd 	/*
    852        1.1       cgd 	 * Get packet stats.
    853        1.1       cgd 	 */
    854        1.1       cgd 	case BIOCGSTATS:
    855        1.1       cgd 		{
    856      1.103  christos 			struct bpf_stat *bs = addr;
    857        1.1       cgd 
    858        1.1       cgd 			bs->bs_recv = d->bd_rcount;
    859        1.1       cgd 			bs->bs_drop = d->bd_dcount;
    860       1.94   darrenr 			bs->bs_capt = d->bd_ccount;
    861       1.94   darrenr 			break;
    862       1.94   darrenr 		}
    863       1.94   darrenr 
    864       1.94   darrenr 	case BIOCGSTATSOLD:
    865       1.94   darrenr 		{
    866      1.103  christos 			struct bpf_stat_old *bs = addr;
    867       1.94   darrenr 
    868       1.94   darrenr 			bs->bs_recv = d->bd_rcount;
    869       1.94   darrenr 			bs->bs_drop = d->bd_dcount;
    870        1.1       cgd 			break;
    871        1.1       cgd 		}
    872        1.1       cgd 
    873        1.1       cgd 	/*
    874        1.1       cgd 	 * Set immediate mode.
    875        1.1       cgd 	 */
    876        1.1       cgd 	case BIOCIMMEDIATE:
    877        1.1       cgd 		d->bd_immediate = *(u_int *)addr;
    878        1.1       cgd 		break;
    879        1.2       cgd 
    880        1.2       cgd 	case BIOCVERSION:
    881        1.2       cgd 		{
    882      1.103  christos 			struct bpf_version *bv = addr;
    883        1.2       cgd 
    884        1.2       cgd 			bv->bv_major = BPF_MAJOR_VERSION;
    885        1.2       cgd 			bv->bv_minor = BPF_MINOR_VERSION;
    886        1.2       cgd 			break;
    887       1.12   mycroft 		}
    888       1.23   thorpej 
    889       1.40   thorpej 	case BIOCGHDRCMPLT:	/* get "header already complete" flag */
    890       1.40   thorpej 		*(u_int *)addr = d->bd_hdrcmplt;
    891       1.40   thorpej 		break;
    892       1.40   thorpej 
    893       1.40   thorpej 	case BIOCSHDRCMPLT:	/* set "header already complete" flag */
    894       1.40   thorpej 		d->bd_hdrcmplt = *(u_int *)addr ? 1 : 0;
    895       1.40   thorpej 		break;
    896       1.23   thorpej 
    897       1.91   darrenr 	/*
    898       1.91   darrenr 	 * Get "see sent packets" flag
    899       1.91   darrenr 	 */
    900       1.91   darrenr 	case BIOCGSEESENT:
    901       1.91   darrenr 		*(u_int *)addr = d->bd_seesent;
    902       1.91   darrenr 		break;
    903       1.91   darrenr 
    904       1.91   darrenr 	/*
    905       1.91   darrenr 	 * Set "see sent" packets flag
    906       1.91   darrenr 	 */
    907       1.91   darrenr 	case BIOCSSEESENT:
    908       1.91   darrenr 		d->bd_seesent = *(u_int *)addr;
    909       1.91   darrenr 		break;
    910       1.91   darrenr 
    911       1.23   thorpej 	case FIONBIO:		/* Non-blocking I/O */
    912       1.93   darrenr 		/*
    913       1.93   darrenr 		 * No need to do anything special as we use IO_NDELAY in
    914       1.93   darrenr 		 * bpfread() as an indication of whether or not to block
    915       1.93   darrenr 		 * the read.
    916       1.93   darrenr 		 */
    917       1.23   thorpej 		break;
    918       1.23   thorpej 
    919       1.23   thorpej 	case FIOASYNC:		/* Send signal on receive packets */
    920       1.23   thorpej 		d->bd_async = *(int *)addr;
    921       1.23   thorpej 		break;
    922       1.23   thorpej 
    923       1.23   thorpej 	case TIOCSPGRP:		/* Process or group to send signals to */
    924       1.85  jdolecek 	case FIOSETOWN:
    925       1.85  jdolecek 		error = fsetown(p, &d->bd_pgid, cmd, addr);
    926       1.23   thorpej 		break;
    927       1.23   thorpej 
    928       1.23   thorpej 	case TIOCGPGRP:
    929       1.85  jdolecek 	case FIOGETOWN:
    930       1.85  jdolecek 		error = fgetown(p, d->bd_pgid, cmd, addr);
    931       1.23   thorpej 		break;
    932        1.1       cgd 	}
    933        1.1       cgd 	return (error);
    934        1.1       cgd }
    935        1.1       cgd 
    936       1.12   mycroft /*
    937        1.2       cgd  * Set d's packet filter program to fp.  If this file already has a filter,
    938        1.1       cgd  * free it and replace it.  Returns EINVAL for bogus requests.
    939        1.1       cgd  */
    940        1.1       cgd int
    941      1.103  christos bpf_setf(struct bpf_d *d, struct bpf_program *fp)
    942        1.1       cgd {
    943        1.1       cgd 	struct bpf_insn *fcode, *old;
    944        1.1       cgd 	u_int flen, size;
    945        1.1       cgd 	int s;
    946        1.1       cgd 
    947        1.1       cgd 	old = d->bd_filter;
    948        1.1       cgd 	if (fp->bf_insns == 0) {
    949        1.1       cgd 		if (fp->bf_len != 0)
    950        1.1       cgd 			return (EINVAL);
    951       1.61   thorpej 		s = splnet();
    952        1.1       cgd 		d->bd_filter = 0;
    953        1.1       cgd 		reset_d(d);
    954        1.1       cgd 		splx(s);
    955        1.1       cgd 		if (old != 0)
    956      1.103  christos 			free(old, M_DEVBUF);
    957        1.1       cgd 		return (0);
    958        1.1       cgd 	}
    959        1.1       cgd 	flen = fp->bf_len;
    960        1.1       cgd 	if (flen > BPF_MAXINSNS)
    961        1.1       cgd 		return (EINVAL);
    962        1.1       cgd 
    963        1.1       cgd 	size = flen * sizeof(*fp->bf_insns);
    964      1.103  christos 	fcode = malloc(size, M_DEVBUF, M_WAITOK);
    965      1.103  christos 	if (copyin(fp->bf_insns, fcode, size) == 0 &&
    966        1.2       cgd 	    bpf_validate(fcode, (int)flen)) {
    967       1.61   thorpej 		s = splnet();
    968        1.1       cgd 		d->bd_filter = fcode;
    969        1.1       cgd 		reset_d(d);
    970        1.1       cgd 		splx(s);
    971        1.1       cgd 		if (old != 0)
    972      1.103  christos 			free(old, M_DEVBUF);
    973        1.1       cgd 
    974        1.1       cgd 		return (0);
    975        1.1       cgd 	}
    976      1.103  christos 	free(fcode, M_DEVBUF);
    977        1.1       cgd 	return (EINVAL);
    978        1.1       cgd }
    979        1.1       cgd 
    980        1.1       cgd /*
    981        1.2       cgd  * Detach a file from its current interface (if attached at all) and attach
    982        1.2       cgd  * to the interface indicated by the name stored in ifr.
    983        1.2       cgd  * Return an errno or 0.
    984        1.1       cgd  */
    985        1.1       cgd static int
    986      1.103  christos bpf_setif(struct bpf_d *d, struct ifreq *ifr)
    987        1.1       cgd {
    988        1.1       cgd 	struct bpf_if *bp;
    989        1.1       cgd 	char *cp;
    990       1.26   thorpej 	int unit_seen, i, s, error;
    991        1.1       cgd 
    992        1.1       cgd 	/*
    993       1.26   thorpej 	 * Make sure the provided name has a unit number, and default
    994       1.26   thorpej 	 * it to '0' if not specified.
    995       1.26   thorpej 	 * XXX This is ugly ... do this differently?
    996        1.1       cgd 	 */
    997       1.26   thorpej 	unit_seen = 0;
    998        1.1       cgd 	cp = ifr->ifr_name;
    999       1.26   thorpej 	cp[sizeof(ifr->ifr_name) - 1] = '\0';	/* sanity */
   1000       1.26   thorpej 	while (*cp++)
   1001       1.26   thorpej 		if (*cp >= '0' && *cp <= '9')
   1002       1.26   thorpej 			unit_seen = 1;
   1003       1.26   thorpej 	if (!unit_seen) {
   1004       1.26   thorpej 		/* Make sure to leave room for the '\0'. */
   1005       1.26   thorpej 		for (i = 0; i < (IFNAMSIZ - 1); ++i) {
   1006       1.26   thorpej 			if ((ifr->ifr_name[i] >= 'a' &&
   1007       1.26   thorpej 			     ifr->ifr_name[i] <= 'z') ||
   1008       1.26   thorpej 			    (ifr->ifr_name[i] >= 'A' &&
   1009       1.26   thorpej 			     ifr->ifr_name[i] <= 'Z'))
   1010       1.26   thorpej 				continue;
   1011       1.26   thorpej 			ifr->ifr_name[i] = '0';
   1012        1.1       cgd 		}
   1013        1.1       cgd 	}
   1014       1.26   thorpej 
   1015        1.1       cgd 	/*
   1016        1.1       cgd 	 * Look through attached interfaces for the named one.
   1017        1.1       cgd 	 */
   1018        1.1       cgd 	for (bp = bpf_iflist; bp != 0; bp = bp->bif_next) {
   1019        1.1       cgd 		struct ifnet *ifp = bp->bif_ifp;
   1020        1.1       cgd 
   1021       1.26   thorpej 		if (ifp == 0 ||
   1022       1.26   thorpej 		    strcmp(ifp->if_xname, ifr->ifr_name) != 0)
   1023        1.1       cgd 			continue;
   1024       1.66      onoe 		/* skip additional entry */
   1025       1.66      onoe 		if (bp->bif_driverp != (struct bpf_if **)&ifp->if_bpf)
   1026       1.66      onoe 			continue;
   1027        1.1       cgd 		/*
   1028        1.2       cgd 		 * We found the requested interface.
   1029        1.2       cgd 		 * Allocate the packet buffers if we need to.
   1030        1.2       cgd 		 * If we're already attached to requested interface,
   1031        1.2       cgd 		 * just flush the buffer.
   1032        1.1       cgd 		 */
   1033        1.2       cgd 		if (d->bd_sbuf == 0) {
   1034        1.2       cgd 			error = bpf_allocbufs(d);
   1035        1.2       cgd 			if (error != 0)
   1036        1.2       cgd 				return (error);
   1037        1.2       cgd 		}
   1038       1.61   thorpej 		s = splnet();
   1039        1.1       cgd 		if (bp != d->bd_bif) {
   1040        1.1       cgd 			if (d->bd_bif)
   1041       1.12   mycroft 				/*
   1042        1.1       cgd 				 * Detach if attached to something else.
   1043        1.1       cgd 				 */
   1044        1.1       cgd 				bpf_detachd(d);
   1045        1.1       cgd 
   1046        1.1       cgd 			bpf_attachd(d, bp);
   1047        1.1       cgd 		}
   1048        1.1       cgd 		reset_d(d);
   1049        1.1       cgd 		splx(s);
   1050        1.1       cgd 		return (0);
   1051        1.1       cgd 	}
   1052        1.1       cgd 	/* Not found. */
   1053        1.1       cgd 	return (ENXIO);
   1054        1.1       cgd }
   1055        1.1       cgd 
   1056        1.1       cgd /*
   1057       1.26   thorpej  * Copy the interface name to the ifreq.
   1058        1.1       cgd  */
   1059        1.1       cgd static void
   1060      1.103  christos bpf_ifname(struct ifnet *ifp, struct ifreq *ifr)
   1061        1.1       cgd {
   1062       1.41     perry 	memcpy(ifr->ifr_name, ifp->if_xname, IFNAMSIZ);
   1063        1.1       cgd }
   1064        1.1       cgd 
   1065        1.1       cgd /*
   1066       1.36  christos  * Support for poll() system call
   1067        1.1       cgd  *
   1068       1.64   darrenr  * Return true iff the specific operation will not block indefinitely - with
   1069       1.64   darrenr  * the assumption that it is safe to positively acknowledge a request for the
   1070       1.64   darrenr  * ability to write to the BPF device.
   1071        1.2       cgd  * Otherwise, return false but make a note that a selwakeup() must be done.
   1072        1.1       cgd  */
   1073      1.105  christos static int
   1074      1.105  christos bpf_poll(struct file *fp, int events, struct proc *p)
   1075        1.1       cgd {
   1076      1.105  christos 	struct bpf_d *d = fp->f_data;
   1077       1.61   thorpej 	int s = splnet();
   1078       1.64   darrenr 	int revents;
   1079       1.12   mycroft 
   1080       1.64   darrenr 	revents = events & (POLLOUT | POLLWRNORM);
   1081       1.44   thorpej 	if (events & (POLLIN | POLLRDNORM)) {
   1082       1.64   darrenr 		/*
   1083       1.64   darrenr 		 * An imitation of the FIONREAD ioctl code.
   1084       1.64   darrenr 		 */
   1085       1.93   darrenr 		if ((d->bd_hlen != 0) ||
   1086       1.93   darrenr 		    (d->bd_immediate && d->bd_slen != 0)) {
   1087       1.30   mycroft 			revents |= events & (POLLIN | POLLRDNORM);
   1088       1.93   darrenr 		} else if (d->bd_state == BPF_TIMED_OUT) {
   1089       1.95   darrenr 			if (d->bd_slen != 0)
   1090       1.95   darrenr 				revents |= events & (POLLIN | POLLRDNORM);
   1091       1.95   darrenr 			else
   1092       1.95   darrenr 				revents |= events & POLLIN;
   1093       1.92   darrenr 		} else {
   1094       1.82      fvdl 			selrecord(p, &d->bd_sel);
   1095       1.91   darrenr 			/* Start the read timeout if necessary */
   1096       1.91   darrenr 			if (d->bd_rtout > 0 && d->bd_state == BPF_IDLE) {
   1097       1.91   darrenr 				callout_reset(&d->bd_callout, d->bd_rtout,
   1098       1.91   darrenr 					      bpf_timed_out, d);
   1099       1.91   darrenr 				d->bd_state = BPF_WAITING;
   1100       1.91   darrenr 			}
   1101       1.91   darrenr 		}
   1102       1.44   thorpej 	}
   1103       1.12   mycroft 
   1104       1.12   mycroft 	splx(s);
   1105       1.30   mycroft 	return (revents);
   1106       1.75  jdolecek }
   1107       1.75  jdolecek 
   1108       1.75  jdolecek static void
   1109       1.75  jdolecek filt_bpfrdetach(struct knote *kn)
   1110       1.75  jdolecek {
   1111       1.75  jdolecek 	struct bpf_d *d = kn->kn_hook;
   1112       1.75  jdolecek 	int s;
   1113       1.75  jdolecek 
   1114       1.75  jdolecek 	s = splnet();
   1115       1.76  christos 	SLIST_REMOVE(&d->bd_sel.sel_klist, kn, knote, kn_selnext);
   1116       1.75  jdolecek 	splx(s);
   1117       1.75  jdolecek }
   1118       1.75  jdolecek 
   1119       1.75  jdolecek static int
   1120       1.75  jdolecek filt_bpfread(struct knote *kn, long hint)
   1121       1.75  jdolecek {
   1122       1.75  jdolecek 	struct bpf_d *d = kn->kn_hook;
   1123       1.75  jdolecek 
   1124       1.75  jdolecek 	kn->kn_data = d->bd_hlen;
   1125       1.75  jdolecek 	if (d->bd_immediate)
   1126       1.75  jdolecek 		kn->kn_data += d->bd_slen;
   1127       1.75  jdolecek 	return (kn->kn_data > 0);
   1128       1.75  jdolecek }
   1129       1.75  jdolecek 
   1130       1.75  jdolecek static const struct filterops bpfread_filtops =
   1131       1.75  jdolecek 	{ 1, NULL, filt_bpfrdetach, filt_bpfread };
   1132       1.75  jdolecek 
   1133      1.105  christos static int
   1134      1.105  christos bpf_kqfilter(struct file *fp, struct knote *kn)
   1135       1.75  jdolecek {
   1136      1.105  christos 	struct bpf_d *d = fp->f_data;
   1137       1.75  jdolecek 	struct klist *klist;
   1138       1.75  jdolecek 	int s;
   1139       1.75  jdolecek 
   1140       1.75  jdolecek 	switch (kn->kn_filter) {
   1141       1.75  jdolecek 	case EVFILT_READ:
   1142       1.76  christos 		klist = &d->bd_sel.sel_klist;
   1143       1.75  jdolecek 		kn->kn_fop = &bpfread_filtops;
   1144       1.75  jdolecek 		break;
   1145       1.75  jdolecek 
   1146       1.75  jdolecek 	default:
   1147       1.75  jdolecek 		return (1);
   1148       1.75  jdolecek 	}
   1149       1.75  jdolecek 
   1150       1.75  jdolecek 	kn->kn_hook = d;
   1151       1.75  jdolecek 
   1152       1.75  jdolecek 	s = splnet();
   1153       1.75  jdolecek 	SLIST_INSERT_HEAD(klist, kn, kn_selnext);
   1154       1.75  jdolecek 	splx(s);
   1155       1.75  jdolecek 
   1156       1.75  jdolecek 	return (0);
   1157        1.1       cgd }
   1158        1.1       cgd 
   1159        1.1       cgd /*
   1160        1.2       cgd  * Incoming linkage from device drivers.  Process the packet pkt, of length
   1161        1.2       cgd  * pktlen, which is stored in a contiguous buffer.  The packet is parsed
   1162        1.2       cgd  * by each process' filter, and if accepted, stashed into the corresponding
   1163        1.2       cgd  * buffer.
   1164        1.1       cgd  */
   1165        1.1       cgd void
   1166      1.103  christos bpf_tap(void *arg, u_char *pkt, u_int pktlen)
   1167        1.1       cgd {
   1168        1.1       cgd 	struct bpf_if *bp;
   1169       1.53  augustss 	struct bpf_d *d;
   1170       1.53  augustss 	u_int slen;
   1171        1.1       cgd 	/*
   1172        1.1       cgd 	 * Note that the ipl does not have to be raised at this point.
   1173        1.1       cgd 	 * The only problem that could arise here is that if two different
   1174        1.1       cgd 	 * interfaces shared any data.  This is not the case.
   1175        1.1       cgd 	 */
   1176      1.103  christos 	bp = arg;
   1177        1.1       cgd 	for (d = bp->bif_dlist; d != 0; d = d->bd_next) {
   1178        1.1       cgd 		++d->bd_rcount;
   1179        1.1       cgd 		slen = bpf_filter(d->bd_filter, pkt, pktlen, pktlen);
   1180        1.1       cgd 		if (slen != 0)
   1181       1.41     perry 			catchpacket(d, pkt, pktlen, slen, memcpy);
   1182        1.1       cgd 	}
   1183        1.1       cgd }
   1184        1.1       cgd 
   1185        1.1       cgd /*
   1186        1.1       cgd  * Copy data from an mbuf chain into a buffer.  This code is derived
   1187        1.1       cgd  * from m_copydata in sys/uipc_mbuf.c.
   1188        1.1       cgd  */
   1189       1.43     perry static void *
   1190      1.103  christos bpf_mcpy(void *dst_arg, const void *src_arg, size_t len)
   1191       1.12   mycroft {
   1192       1.53  augustss 	const struct mbuf *m;
   1193       1.53  augustss 	u_int count;
   1194        1.1       cgd 	u_char *dst;
   1195        1.1       cgd 
   1196       1.12   mycroft 	m = src_arg;
   1197       1.12   mycroft 	dst = dst_arg;
   1198        1.1       cgd 	while (len > 0) {
   1199        1.1       cgd 		if (m == 0)
   1200       1.41     perry 			panic("bpf_mcpy");
   1201       1.12   mycroft 		count = min(m->m_len, len);
   1202      1.103  christos 		memcpy(dst, mtod(m, void *), count);
   1203        1.1       cgd 		m = m->m_next;
   1204        1.1       cgd 		dst += count;
   1205        1.1       cgd 		len -= count;
   1206        1.1       cgd 	}
   1207       1.68    itojun 	return (dst_arg);
   1208        1.1       cgd }
   1209        1.1       cgd 
   1210        1.1       cgd /*
   1211       1.96    dyoung  * Dispatch a packet to all the listeners on interface bp.
   1212       1.96    dyoung  *
   1213       1.96    dyoung  * marg    pointer to the packet, either a data buffer or an mbuf chain
   1214       1.96    dyoung  * buflen  buffer length, if marg is a data buffer
   1215       1.96    dyoung  * cpfn    a function that can copy marg into the listener's buffer
   1216       1.96    dyoung  * pktlen  length of the packet
   1217       1.96    dyoung  * rcvif   either NULL or the interface the packet came in on.
   1218       1.96    dyoung  */
   1219       1.96    dyoung static __inline void
   1220       1.96    dyoung bpf_deliver(struct bpf_if *bp, void *(*cpfn)(void *, const void *, size_t),
   1221       1.96    dyoung     void *marg, u_int pktlen, u_int buflen, struct ifnet *rcvif)
   1222       1.96    dyoung {
   1223       1.96    dyoung 	u_int slen;
   1224       1.96    dyoung 	struct bpf_d *d;
   1225       1.96    dyoung 
   1226       1.96    dyoung 	for (d = bp->bif_dlist; d != 0; d = d->bd_next) {
   1227       1.96    dyoung 		if (!d->bd_seesent && (rcvif == NULL))
   1228       1.96    dyoung 			continue;
   1229       1.96    dyoung 		++d->bd_rcount;
   1230       1.96    dyoung 		slen = bpf_filter(d->bd_filter, marg, pktlen, buflen);
   1231       1.96    dyoung 		if (slen != 0)
   1232       1.96    dyoung 			catchpacket(d, marg, pktlen, slen, cpfn);
   1233       1.96    dyoung 	}
   1234       1.96    dyoung }
   1235       1.96    dyoung 
   1236       1.96    dyoung /*
   1237       1.96    dyoung  * Incoming linkage from device drivers, when the head of the packet is in
   1238       1.96    dyoung  * a buffer, and the tail is in an mbuf chain.
   1239       1.96    dyoung  */
   1240       1.96    dyoung void
   1241      1.103  christos bpf_mtap2(void *arg, void *data, u_int dlen, struct mbuf *m)
   1242       1.96    dyoung {
   1243      1.103  christos 	struct bpf_if *bp = arg;
   1244       1.96    dyoung 	u_int pktlen;
   1245       1.96    dyoung 	struct mbuf mb;
   1246       1.96    dyoung 
   1247      1.101    dyoung 	pktlen = m_length(m) + dlen;
   1248       1.96    dyoung 
   1249       1.96    dyoung 	/*
   1250       1.96    dyoung 	 * Craft on-stack mbuf suitable for passing to bpf_filter.
   1251       1.96    dyoung 	 * Note that we cut corners here; we only setup what's
   1252       1.96    dyoung 	 * absolutely needed--this mbuf should never go anywhere else.
   1253       1.96    dyoung 	 */
   1254      1.101    dyoung 	(void)memset(&mb, 0, sizeof(mb));
   1255       1.96    dyoung 	mb.m_next = m;
   1256       1.96    dyoung 	mb.m_data = data;
   1257       1.96    dyoung 	mb.m_len = dlen;
   1258       1.96    dyoung 
   1259       1.96    dyoung 	bpf_deliver(bp, bpf_mcpy, &mb, pktlen, 0, m->m_pkthdr.rcvif);
   1260       1.96    dyoung }
   1261       1.96    dyoung 
   1262       1.96    dyoung /*
   1263        1.2       cgd  * Incoming linkage from device drivers, when packet is in an mbuf chain.
   1264        1.1       cgd  */
   1265        1.1       cgd void
   1266      1.103  christos bpf_mtap(void *arg, struct mbuf *m)
   1267        1.1       cgd {
   1268      1.103  christos 	void *(*cpfn)(void *, const void *, size_t);
   1269      1.103  christos 	struct bpf_if *bp = arg;
   1270       1.96    dyoung 	u_int pktlen, buflen;
   1271       1.91   darrenr 	void *marg;
   1272        1.1       cgd 
   1273      1.101    dyoung 	pktlen = m_length(m);
   1274        1.1       cgd 
   1275       1.91   darrenr 	if (pktlen == m->m_len) {
   1276       1.91   darrenr 		cpfn = memcpy;
   1277       1.91   darrenr 		marg = mtod(m, void *);
   1278       1.91   darrenr 		buflen = pktlen;
   1279       1.91   darrenr 	} else {
   1280       1.91   darrenr 		cpfn = bpf_mcpy;
   1281       1.91   darrenr 		marg = m;
   1282       1.91   darrenr 		buflen = 0;
   1283       1.91   darrenr 	}
   1284       1.91   darrenr 
   1285       1.96    dyoung 	bpf_deliver(bp, cpfn, marg, pktlen, buflen, m->m_pkthdr.rcvif);
   1286        1.1       cgd }
   1287        1.1       cgd 
   1288        1.1       cgd /*
   1289      1.104  christos  * We need to prepend the address family as
   1290      1.104  christos  * a four byte field.  Cons up a dummy header
   1291      1.104  christos  * to pacify bpf.  This is safe because bpf
   1292      1.104  christos  * will only read from the mbuf (i.e., it won't
   1293      1.104  christos  * try to free it or keep a pointer a to it).
   1294      1.104  christos  */
   1295      1.104  christos void
   1296      1.104  christos bpf_mtap_af(void *arg, u_int32_t af, struct mbuf *m)
   1297      1.104  christos {
   1298      1.104  christos 	struct mbuf m0;
   1299  1.105.6.1      yamt 
   1300      1.104  christos 	m0.m_flags = 0;
   1301      1.104  christos 	m0.m_next = m;
   1302      1.104  christos 	m0.m_len = 4;
   1303      1.104  christos 	m0.m_data = (char *)&af;
   1304      1.104  christos 
   1305      1.104  christos 	bpf_mtap(arg, &m0);
   1306      1.104  christos }
   1307      1.104  christos 
   1308      1.104  christos void
   1309      1.104  christos bpf_mtap_et(void *arg, u_int16_t et, struct mbuf *m)
   1310      1.104  christos {
   1311      1.104  christos 	struct mbuf m0;
   1312      1.104  christos 
   1313      1.104  christos 	m0.m_flags = 0;
   1314      1.104  christos 	m0.m_next = m;
   1315      1.104  christos 	m0.m_len = 14;
   1316      1.104  christos 	m0.m_data = m0.m_dat;
   1317      1.104  christos 
   1318      1.104  christos 	((u_int32_t *)m0.m_data)[0] = 0;
   1319      1.104  christos 	((u_int32_t *)m0.m_data)[1] = 0;
   1320      1.104  christos 	((u_int32_t *)m0.m_data)[2] = 0;
   1321      1.104  christos 	((u_int16_t *)m0.m_data)[6] = et;
   1322      1.104  christos 
   1323      1.104  christos 	bpf_mtap(arg, &m0);
   1324      1.104  christos }
   1325      1.104  christos 
   1326      1.104  christos #if NSL > 0 || NSTRIP > 0
   1327      1.104  christos /*
   1328      1.104  christos  * Put the SLIP pseudo-"link header" in place.
   1329      1.104  christos  * Note this M_PREPEND() should never fail,
   1330      1.104  christos  * swince we know we always have enough space
   1331      1.104  christos  * in the input buffer.
   1332      1.104  christos  */
   1333      1.104  christos void
   1334      1.104  christos bpf_mtap_sl_in(void *arg, u_char *chdr, struct mbuf **m)
   1335      1.104  christos {
   1336      1.104  christos 	int s;
   1337      1.104  christos 	u_char *hp;
   1338      1.104  christos 
   1339      1.104  christos 	M_PREPEND(*m, SLIP_HDRLEN, M_DONTWAIT);
   1340      1.104  christos 	if (*m == NULL)
   1341      1.104  christos 		return;
   1342      1.104  christos 
   1343      1.104  christos 	hp = mtod(*m, u_char *);
   1344      1.104  christos 	hp[SLX_DIR] = SLIPDIR_IN;
   1345      1.104  christos 	(void)memcpy(&hp[SLX_CHDR], chdr, CHDR_LEN);
   1346      1.104  christos 
   1347      1.104  christos 	s = splnet();
   1348      1.104  christos 	bpf_mtap(arg, *m);
   1349      1.104  christos 	splx(s);
   1350      1.104  christos 
   1351      1.104  christos 	m_adj(*m, SLIP_HDRLEN);
   1352      1.104  christos }
   1353      1.104  christos 
   1354      1.104  christos /*
   1355      1.104  christos  * Put the SLIP pseudo-"link header" in
   1356      1.104  christos  * place.  The compressed header is now
   1357      1.104  christos  * at the beginning of the mbuf.
   1358      1.104  christos  */
   1359      1.104  christos void
   1360      1.104  christos bpf_mtap_sl_out(void *arg, u_char *chdr, struct mbuf *m)
   1361      1.104  christos {
   1362      1.104  christos 	struct mbuf m0;
   1363      1.104  christos 	u_char *hp;
   1364      1.104  christos 	int s;
   1365      1.104  christos 
   1366      1.104  christos 	m0.m_flags = 0;
   1367      1.104  christos 	m0.m_next = m;
   1368      1.104  christos 	m0.m_data = m0.m_dat;
   1369      1.104  christos 	m0.m_len = SLIP_HDRLEN;
   1370      1.104  christos 
   1371      1.104  christos 	hp = mtod(&m0, u_char *);
   1372      1.104  christos 
   1373      1.104  christos 	hp[SLX_DIR] = SLIPDIR_OUT;
   1374      1.104  christos 	(void)memcpy(&hp[SLX_CHDR], chdr, CHDR_LEN);
   1375      1.104  christos 
   1376      1.104  christos 	s = splnet();
   1377      1.104  christos 	bpf_mtap(arg, &m0);
   1378      1.104  christos 	splx(s);
   1379      1.104  christos 	m_freem(m);
   1380      1.104  christos }
   1381      1.104  christos #endif
   1382      1.104  christos 
   1383      1.104  christos /*
   1384        1.1       cgd  * Move the packet data from interface memory (pkt) into the
   1385        1.1       cgd  * store buffer.  Return 1 if it's time to wakeup a listener (buffer full),
   1386        1.2       cgd  * otherwise 0.  "copy" is the routine called to do the actual data
   1387       1.41     perry  * transfer.  memcpy is passed in to copy contiguous chunks, while
   1388       1.41     perry  * bpf_mcpy is passed in to copy mbuf chains.  In the latter case,
   1389        1.2       cgd  * pkt is really an mbuf.
   1390        1.1       cgd  */
   1391        1.1       cgd static void
   1392      1.103  christos catchpacket(struct bpf_d *d, u_char *pkt, u_int pktlen, u_int snaplen,
   1393      1.103  christos     void *(*cpfn)(void *, const void *, size_t))
   1394       1.53  augustss {
   1395       1.53  augustss 	struct bpf_hdr *hp;
   1396       1.53  augustss 	int totlen, curlen;
   1397       1.53  augustss 	int hdrlen = d->bd_bif->bif_hdrlen;
   1398       1.94   darrenr 
   1399       1.94   darrenr 	++d->bd_ccount;
   1400        1.1       cgd 	/*
   1401        1.1       cgd 	 * Figure out how many bytes to move.  If the packet is
   1402        1.1       cgd 	 * greater or equal to the snapshot length, transfer that
   1403        1.1       cgd 	 * much.  Otherwise, transfer the whole packet (unless
   1404        1.1       cgd 	 * we hit the buffer size limit).
   1405        1.1       cgd 	 */
   1406       1.12   mycroft 	totlen = hdrlen + min(snaplen, pktlen);
   1407        1.1       cgd 	if (totlen > d->bd_bufsize)
   1408        1.1       cgd 		totlen = d->bd_bufsize;
   1409        1.1       cgd 
   1410        1.1       cgd 	/*
   1411        1.1       cgd 	 * Round up the end of the previous packet to the next longword.
   1412        1.1       cgd 	 */
   1413        1.1       cgd 	curlen = BPF_WORDALIGN(d->bd_slen);
   1414        1.1       cgd 	if (curlen + totlen > d->bd_bufsize) {
   1415        1.1       cgd 		/*
   1416        1.1       cgd 		 * This packet will overflow the storage buffer.
   1417        1.1       cgd 		 * Rotate the buffers if we can, then wakeup any
   1418        1.1       cgd 		 * pending reads.
   1419        1.1       cgd 		 */
   1420      1.100   darrenr 		if (d->bd_fbuf == 0) {
   1421       1.12   mycroft 			/*
   1422       1.12   mycroft 			 * We haven't completed the previous read yet,
   1423        1.1       cgd 			 * so drop the packet.
   1424        1.1       cgd 			 */
   1425        1.1       cgd 			++d->bd_dcount;
   1426        1.1       cgd 			return;
   1427        1.1       cgd 		}
   1428        1.1       cgd 		ROTATE_BUFFERS(d);
   1429        1.1       cgd 		bpf_wakeup(d);
   1430        1.1       cgd 		curlen = 0;
   1431       1.97   darrenr 	}
   1432        1.1       cgd 
   1433        1.1       cgd 	/*
   1434        1.1       cgd 	 * Append the bpf header.
   1435        1.1       cgd 	 */
   1436        1.1       cgd 	hp = (struct bpf_hdr *)(d->bd_sbuf + curlen);
   1437        1.2       cgd 	microtime(&hp->bh_tstamp);
   1438        1.1       cgd 	hp->bh_datalen = pktlen;
   1439        1.1       cgd 	hp->bh_hdrlen = hdrlen;
   1440        1.1       cgd 	/*
   1441        1.1       cgd 	 * Copy the packet data into the store buffer and update its length.
   1442        1.1       cgd 	 */
   1443       1.41     perry 	(*cpfn)((u_char *)hp + hdrlen, pkt, (hp->bh_caplen = totlen - hdrlen));
   1444        1.1       cgd 	d->bd_slen = curlen + totlen;
   1445       1.97   darrenr 
   1446       1.97   darrenr 	/*
   1447       1.97   darrenr 	 * Call bpf_wakeup after bd_slen has been updated so that kevent(2)
   1448       1.97   darrenr 	 * will cause filt_bpfread() to be called with it adjusted.
   1449       1.97   darrenr 	 */
   1450       1.97   darrenr 	if (d->bd_immediate || d->bd_state == BPF_TIMED_OUT)
   1451       1.97   darrenr 		/*
   1452       1.97   darrenr 		 * Immediate mode is set, or the read timeout has
   1453       1.97   darrenr 		 * already expired during a select call.  A packet
   1454       1.97   darrenr 		 * arrived, so the reader should be woken up.
   1455       1.97   darrenr 		 */
   1456       1.97   darrenr 		bpf_wakeup(d);
   1457        1.1       cgd }
   1458        1.1       cgd 
   1459       1.12   mycroft /*
   1460        1.1       cgd  * Initialize all nonzero fields of a descriptor.
   1461        1.1       cgd  */
   1462        1.1       cgd static int
   1463      1.103  christos bpf_allocbufs(struct bpf_d *d)
   1464        1.1       cgd {
   1465       1.50     enami 
   1466      1.103  christos 	d->bd_fbuf = malloc(d->bd_bufsize, M_DEVBUF, M_NOWAIT);
   1467      1.100   darrenr 	if (!d->bd_fbuf)
   1468       1.79    itojun 		return (ENOBUFS);
   1469      1.103  christos 	d->bd_sbuf = malloc(d->bd_bufsize, M_DEVBUF, M_NOWAIT);
   1470      1.100   darrenr 	if (!d->bd_sbuf) {
   1471      1.100   darrenr 		free(d->bd_fbuf, M_DEVBUF);
   1472       1.79    itojun 		return (ENOBUFS);
   1473       1.79    itojun 	}
   1474        1.1       cgd 	d->bd_slen = 0;
   1475        1.1       cgd 	d->bd_hlen = 0;
   1476        1.1       cgd 	return (0);
   1477        1.1       cgd }
   1478        1.1       cgd 
   1479        1.1       cgd /*
   1480        1.2       cgd  * Free buffers currently in use by a descriptor.
   1481        1.2       cgd  * Called on close.
   1482        1.2       cgd  */
   1483        1.2       cgd static void
   1484      1.103  christos bpf_freed(struct bpf_d *d)
   1485        1.2       cgd {
   1486        1.2       cgd 	/*
   1487        1.2       cgd 	 * We don't need to lock out interrupts since this descriptor has
   1488        1.2       cgd 	 * been detached from its interface and it yet hasn't been marked
   1489        1.2       cgd 	 * free.
   1490        1.2       cgd 	 */
   1491      1.100   darrenr 	if (d->bd_sbuf != 0) {
   1492      1.100   darrenr 		free(d->bd_sbuf, M_DEVBUF);
   1493      1.100   darrenr 		if (d->bd_hbuf != 0)
   1494      1.100   darrenr 			free(d->bd_hbuf, M_DEVBUF);
   1495      1.100   darrenr 		if (d->bd_fbuf != 0)
   1496      1.100   darrenr 			free(d->bd_fbuf, M_DEVBUF);
   1497      1.100   darrenr 	}
   1498        1.2       cgd 	if (d->bd_filter)
   1499      1.103  christos 		free(d->bd_filter, M_DEVBUF);
   1500        1.2       cgd }
   1501        1.2       cgd 
   1502        1.2       cgd /*
   1503       1.60   thorpej  * Attach an interface to bpf.  dlt is the link layer type; hdrlen is the
   1504       1.60   thorpej  * fixed size of the link header (variable length headers not yet supported).
   1505        1.1       cgd  */
   1506        1.1       cgd void
   1507      1.103  christos bpfattach(struct ifnet *ifp, u_int dlt, u_int hdrlen)
   1508        1.1       cgd {
   1509       1.66      onoe 
   1510       1.66      onoe 	bpfattach2(ifp, dlt, hdrlen, &ifp->if_bpf);
   1511       1.66      onoe }
   1512       1.66      onoe 
   1513       1.66      onoe /*
   1514       1.66      onoe  * Attach additional dlt for a interface to bpf.  dlt is the link layer type;
   1515       1.66      onoe  * hdrlen is the fixed size of the link header for the specified dlt
   1516       1.66      onoe  * (variable length headers not yet supported).
   1517       1.66      onoe  */
   1518       1.66      onoe void
   1519      1.103  christos bpfattach2(struct ifnet *ifp, u_int dlt, u_int hdrlen, void *driverp)
   1520       1.66      onoe {
   1521        1.1       cgd 	struct bpf_if *bp;
   1522      1.103  christos 	bp = malloc(sizeof(*bp), M_DEVBUF, M_DONTWAIT);
   1523       1.12   mycroft 	if (bp == 0)
   1524       1.12   mycroft 		panic("bpfattach");
   1525       1.12   mycroft 
   1526        1.1       cgd 	bp->bif_dlist = 0;
   1527      1.103  christos 	bp->bif_driverp = driverp;
   1528        1.1       cgd 	bp->bif_ifp = ifp;
   1529        1.1       cgd 	bp->bif_dlt = dlt;
   1530        1.1       cgd 
   1531        1.1       cgd 	bp->bif_next = bpf_iflist;
   1532        1.1       cgd 	bpf_iflist = bp;
   1533        1.1       cgd 
   1534        1.1       cgd 	*bp->bif_driverp = 0;
   1535        1.1       cgd 
   1536        1.1       cgd 	/*
   1537        1.1       cgd 	 * Compute the length of the bpf header.  This is not necessarily
   1538       1.12   mycroft 	 * equal to SIZEOF_BPF_HDR because we want to insert spacing such
   1539       1.12   mycroft 	 * that the network layer header begins on a longword boundary (for
   1540        1.1       cgd 	 * performance reasons and to alleviate alignment restrictions).
   1541        1.1       cgd 	 */
   1542        1.1       cgd 	bp->bif_hdrlen = BPF_WORDALIGN(hdrlen + SIZEOF_BPF_HDR) - hdrlen;
   1543        1.1       cgd 
   1544       1.11   deraadt #if 0
   1545       1.32  christos 	printf("bpf: %s attached\n", ifp->if_xname);
   1546       1.11   deraadt #endif
   1547       1.48   thorpej }
   1548       1.48   thorpej 
   1549       1.48   thorpej /*
   1550       1.48   thorpej  * Remove an interface from bpf.
   1551       1.48   thorpej  */
   1552       1.48   thorpej void
   1553      1.103  christos bpfdetach(struct ifnet *ifp)
   1554       1.48   thorpej {
   1555       1.48   thorpej 	struct bpf_if *bp, **pbp;
   1556       1.51     enami 	struct bpf_d *d;
   1557      1.105  christos 	int s;
   1558       1.51     enami 
   1559       1.52     soren 	/* Nuke the vnodes for any open instances */
   1560      1.105  christos 	for (d = LIST_FIRST(&bpf_list); d != NULL; d = LIST_NEXT(d, bd_list)) {
   1561      1.105  christos 		if (d->bd_bif != NULL && d->bd_bif->bif_ifp == ifp) {
   1562       1.51     enami 			/*
   1563       1.51     enami 			 * Detach the descriptor from an interface now.
   1564       1.51     enami 			 * It will be free'ed later by close routine.
   1565       1.51     enami 			 */
   1566       1.61   thorpej 			s = splnet();
   1567       1.51     enami 			d->bd_promisc = 0;	/* we can't touch device. */
   1568       1.51     enami 			bpf_detachd(d);
   1569       1.51     enami 			splx(s);
   1570       1.51     enami 		}
   1571       1.51     enami 	}
   1572       1.48   thorpej 
   1573       1.66      onoe   again:
   1574       1.48   thorpej 	for (bp = bpf_iflist, pbp = &bpf_iflist;
   1575       1.48   thorpej 	     bp != NULL; pbp = &bp->bif_next, bp = bp->bif_next) {
   1576       1.48   thorpej 		if (bp->bif_ifp == ifp) {
   1577       1.48   thorpej 			*pbp = bp->bif_next;
   1578       1.48   thorpej 			free(bp, M_DEVBUF);
   1579       1.66      onoe 			goto again;
   1580       1.48   thorpej 		}
   1581       1.48   thorpej 	}
   1582       1.47   thorpej }
   1583       1.47   thorpej 
   1584       1.47   thorpej /*
   1585       1.66      onoe  * Change the data link type of a interface.
   1586       1.47   thorpej  */
   1587       1.47   thorpej void
   1588      1.103  christos bpf_change_type(struct ifnet *ifp, u_int dlt, u_int hdrlen)
   1589       1.47   thorpej {
   1590       1.47   thorpej 	struct bpf_if *bp;
   1591       1.47   thorpej 
   1592       1.47   thorpej 	for (bp = bpf_iflist; bp != NULL; bp = bp->bif_next) {
   1593       1.59   thorpej 		if (bp->bif_driverp == (struct bpf_if **)&ifp->if_bpf)
   1594       1.47   thorpej 			break;
   1595       1.47   thorpej 	}
   1596       1.47   thorpej 	if (bp == NULL)
   1597       1.47   thorpej 		panic("bpf_change_type");
   1598       1.47   thorpej 
   1599       1.47   thorpej 	bp->bif_dlt = dlt;
   1600       1.47   thorpej 
   1601       1.47   thorpej 	/*
   1602       1.47   thorpej 	 * Compute the length of the bpf header.  This is not necessarily
   1603       1.47   thorpej 	 * equal to SIZEOF_BPF_HDR because we want to insert spacing such
   1604       1.47   thorpej 	 * that the network layer header begins on a longword boundary (for
   1605       1.47   thorpej 	 * performance reasons and to alleviate alignment restrictions).
   1606       1.47   thorpej 	 */
   1607       1.47   thorpej 	bp->bif_hdrlen = BPF_WORDALIGN(hdrlen + SIZEOF_BPF_HDR) - hdrlen;
   1608       1.66      onoe }
   1609       1.66      onoe 
   1610       1.66      onoe /*
   1611       1.66      onoe  * Get a list of available data link type of the interface.
   1612       1.66      onoe  */
   1613       1.66      onoe static int
   1614      1.103  christos bpf_getdltlist(struct bpf_d *d, struct bpf_dltlist *bfl)
   1615       1.66      onoe {
   1616       1.66      onoe 	int n, error;
   1617       1.66      onoe 	struct ifnet *ifp;
   1618       1.66      onoe 	struct bpf_if *bp;
   1619       1.66      onoe 
   1620       1.66      onoe 	ifp = d->bd_bif->bif_ifp;
   1621       1.66      onoe 	n = 0;
   1622       1.66      onoe 	error = 0;
   1623       1.66      onoe 	for (bp = bpf_iflist; bp != NULL; bp = bp->bif_next) {
   1624       1.66      onoe 		if (bp->bif_ifp != ifp)
   1625       1.66      onoe 			continue;
   1626       1.66      onoe 		if (bfl->bfl_list != NULL) {
   1627       1.66      onoe 			if (n >= bfl->bfl_len)
   1628       1.66      onoe 				return ENOMEM;
   1629       1.66      onoe 			error = copyout(&bp->bif_dlt,
   1630       1.66      onoe 			    bfl->bfl_list + n, sizeof(u_int));
   1631       1.66      onoe 		}
   1632       1.66      onoe 		n++;
   1633       1.66      onoe 	}
   1634       1.66      onoe 	bfl->bfl_len = n;
   1635       1.66      onoe 	return error;
   1636       1.66      onoe }
   1637       1.66      onoe 
   1638       1.66      onoe /*
   1639       1.66      onoe  * Set the data link type of a BPF instance.
   1640       1.66      onoe  */
   1641       1.66      onoe static int
   1642      1.103  christos bpf_setdlt(struct bpf_d *d, u_int dlt)
   1643       1.66      onoe {
   1644       1.69   thorpej 	int s, error, opromisc;
   1645       1.66      onoe 	struct ifnet *ifp;
   1646       1.66      onoe 	struct bpf_if *bp;
   1647       1.66      onoe 
   1648       1.66      onoe 	if (d->bd_bif->bif_dlt == dlt)
   1649       1.66      onoe 		return 0;
   1650       1.66      onoe 	ifp = d->bd_bif->bif_ifp;
   1651       1.66      onoe 	for (bp = bpf_iflist; bp != NULL; bp = bp->bif_next) {
   1652       1.66      onoe 		if (bp->bif_ifp == ifp && bp->bif_dlt == dlt)
   1653       1.66      onoe 			break;
   1654       1.66      onoe 	}
   1655       1.66      onoe 	if (bp == NULL)
   1656       1.66      onoe 		return EINVAL;
   1657       1.66      onoe 	s = splnet();
   1658       1.69   thorpej 	opromisc = d->bd_promisc;
   1659       1.66      onoe 	bpf_detachd(d);
   1660       1.66      onoe 	bpf_attachd(d, bp);
   1661       1.66      onoe 	reset_d(d);
   1662       1.69   thorpej 	if (opromisc) {
   1663       1.69   thorpej 		error = ifpromisc(bp->bif_ifp, 1);
   1664       1.69   thorpej 		if (error)
   1665       1.69   thorpej 			printf("%s: bpf_setdlt: ifpromisc failed (%d)\n",
   1666       1.69   thorpej 			    bp->bif_ifp->if_xname, error);
   1667       1.69   thorpej 		else
   1668       1.69   thorpej 			d->bd_promisc = 1;
   1669       1.69   thorpej 	}
   1670       1.66      onoe 	splx(s);
   1671       1.66      onoe 	return 0;
   1672        1.1       cgd }
   1673       1.89  jonathan 
   1674       1.89  jonathan static int
   1675       1.89  jonathan sysctl_net_bpf_maxbufsize(SYSCTLFN_ARGS)
   1676       1.89  jonathan {
   1677       1.89  jonathan 	int newsize, error;
   1678       1.89  jonathan 	struct sysctlnode node;
   1679       1.89  jonathan 
   1680       1.89  jonathan 	node = *rnode;
   1681       1.89  jonathan 	node.sysctl_data = &newsize;
   1682       1.89  jonathan 	newsize = bpf_maxbufsize;
   1683       1.89  jonathan 	error = sysctl_lookup(SYSCTLFN_CALL(&node));
   1684       1.89  jonathan 	if (error || newp == NULL)
   1685       1.89  jonathan 		return (error);
   1686       1.89  jonathan 
   1687       1.89  jonathan 	if (newsize < BPF_MINBUFSIZE || newsize > BPF_MAXBUFSIZE)
   1688       1.89  jonathan 		return (EINVAL);
   1689       1.89  jonathan 
   1690       1.89  jonathan 	bpf_maxbufsize = newsize;
   1691       1.89  jonathan 
   1692       1.89  jonathan 	return (0);
   1693       1.89  jonathan }
   1694       1.89  jonathan 
   1695       1.89  jonathan SYSCTL_SETUP(sysctl_net_bfp_setup, "sysctl net.bpf subtree setup")
   1696       1.89  jonathan {
   1697       1.89  jonathan 	struct sysctlnode *node;
   1698       1.89  jonathan 
   1699       1.90    atatat 	sysctl_createv(clog, 0, NULL, NULL,
   1700       1.90    atatat 		       CTLFLAG_PERMANENT,
   1701       1.89  jonathan 		       CTLTYPE_NODE, "net", NULL,
   1702       1.89  jonathan 		       NULL, 0, NULL, 0,
   1703       1.89  jonathan 		       CTL_NET, CTL_EOL);
   1704       1.89  jonathan 
   1705       1.89  jonathan 	node = NULL;
   1706       1.90    atatat 	sysctl_createv(clog, 0, NULL, &node,
   1707       1.90    atatat 		       CTLFLAG_PERMANENT,
   1708       1.98    atatat 		       CTLTYPE_NODE, "bpf",
   1709       1.98    atatat 		       SYSCTL_DESCR("BPF options"),
   1710       1.89  jonathan 		       NULL, 0, NULL, 0,
   1711       1.89  jonathan 		       CTL_NET, CTL_CREATE, CTL_EOL);
   1712       1.89  jonathan 	if (node != NULL)
   1713       1.90    atatat 		sysctl_createv(clog, 0, NULL, NULL,
   1714       1.90    atatat 			CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
   1715       1.98    atatat 			CTLTYPE_INT, "maxbufsize",
   1716       1.98    atatat 			SYSCTL_DESCR("Maximum size for data capture buffer"),
   1717       1.89  jonathan 			sysctl_net_bpf_maxbufsize, 0, &bpf_maxbufsize, 0,
   1718       1.89  jonathan 			CTL_NET, node->sysctl_num, CTL_CREATE, CTL_EOL);
   1719       1.89  jonathan }
   1720