Home | History | Annotate | Line # | Download | only in net
bpf.c revision 1.104
      1  1.104  christos /*	$NetBSD: bpf.c,v 1.104 2004/08/19 20:58:23 christos 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  1.104  christos __KERNEL_RCSID(0, "$NetBSD: bpf.c,v 1.104 2004/08/19 20:58:23 christos Exp $");
     43    1.1       cgd 
     44    1.1       cgd #include "bpfilter.h"
     45    1.1       cgd 
     46    1.1       cgd #include <sys/param.h>
     47    1.1       cgd #include <sys/systm.h>
     48    1.1       cgd #include <sys/mbuf.h>
     49    1.1       cgd #include <sys/buf.h>
     50   1.12   mycroft #include <sys/time.h>
     51    1.1       cgd #include <sys/proc.h>
     52    1.1       cgd #include <sys/user.h>
     53    1.1       cgd #include <sys/ioctl.h>
     54   1.25  christos #include <sys/conf.h>
     55   1.51     enami #include <sys/vnode.h>
     56   1.11   deraadt 
     57    1.1       cgd #include <sys/file.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    1.2       cgd struct bpf_d	bpf_dtab[NBPFILTER];
    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   1.67   gehenna dev_type_open(bpfopen);
    133   1.67   gehenna dev_type_close(bpfclose);
    134   1.67   gehenna dev_type_read(bpfread);
    135   1.67   gehenna dev_type_write(bpfwrite);
    136   1.67   gehenna dev_type_ioctl(bpfioctl);
    137   1.67   gehenna dev_type_poll(bpfpoll);
    138   1.75  jdolecek dev_type_kqfilter(bpfkqfilter);
    139   1.67   gehenna 
    140   1.67   gehenna const struct cdevsw bpf_cdevsw = {
    141   1.67   gehenna 	bpfopen, bpfclose, bpfread, bpfwrite, bpfioctl,
    142  1.100   darrenr 	nostop, notty, bpfpoll, nommap, bpfkqfilter,
    143   1.67   gehenna };
    144   1.67   gehenna 
    145   1.12   mycroft static int
    146   1.36  christos bpf_movein(uio, linktype, mtu, mp, sockp)
    147   1.53  augustss 	struct uio *uio;
    148   1.20   mycroft 	int linktype;
    149   1.36  christos 	int mtu;
    150   1.53  augustss 	struct mbuf **mp;
    151   1.53  augustss 	struct sockaddr *sockp;
    152   1.12   mycroft {
    153   1.12   mycroft 	struct mbuf *m;
    154   1.12   mycroft 	int error;
    155   1.12   mycroft 	int len;
    156   1.12   mycroft 	int hlen;
    157   1.36  christos 	int align;
    158   1.12   mycroft 
    159   1.12   mycroft 	/*
    160   1.12   mycroft 	 * Build a sockaddr based on the data link layer type.
    161   1.12   mycroft 	 * We do this at this level because the ethernet header
    162   1.12   mycroft 	 * is copied directly into the data field of the sockaddr.
    163   1.12   mycroft 	 * In the case of SLIP, there is no header and the packet
    164   1.12   mycroft 	 * is forwarded as is.
    165   1.12   mycroft 	 * Also, we are careful to leave room at the front of the mbuf
    166   1.12   mycroft 	 * for the link level header.
    167   1.12   mycroft 	 */
    168   1.12   mycroft 	switch (linktype) {
    169   1.12   mycroft 
    170   1.12   mycroft 	case DLT_SLIP:
    171   1.12   mycroft 		sockp->sa_family = AF_INET;
    172   1.12   mycroft 		hlen = 0;
    173   1.36  christos 		align = 0;
    174   1.12   mycroft 		break;
    175   1.12   mycroft 
    176   1.12   mycroft 	case DLT_PPP:
    177   1.12   mycroft 		sockp->sa_family = AF_UNSPEC;
    178   1.12   mycroft 		hlen = 0;
    179   1.36  christos 		align = 0;
    180   1.12   mycroft 		break;
    181   1.12   mycroft 
    182   1.12   mycroft 	case DLT_EN10MB:
    183   1.12   mycroft 		sockp->sa_family = AF_UNSPEC;
    184   1.12   mycroft 		/* XXX Would MAXLINKHDR be better? */
    185   1.36  christos  		/* 6(dst)+6(src)+2(type) */
    186   1.12   mycroft 		hlen = sizeof(struct ether_header);
    187   1.36  christos 		align = 2;
    188   1.17     glass 		break;
    189   1.17     glass 
    190   1.17     glass 	case DLT_ARCNET:
    191   1.17     glass 		sockp->sa_family = AF_UNSPEC;
    192   1.17     glass 		hlen = ARC_HDRLEN;
    193   1.36  christos 		align = 5;
    194   1.12   mycroft 		break;
    195   1.12   mycroft 
    196   1.12   mycroft 	case DLT_FDDI:
    197   1.56      matt 		sockp->sa_family = AF_LINK;
    198   1.56      matt 		/* XXX 4(FORMAC)+6(dst)+6(src) */
    199   1.56      matt 		hlen = 16;
    200   1.36  christos 		align = 0;
    201   1.62     bjh21 		break;
    202   1.62     bjh21 
    203   1.62     bjh21 	case DLT_ECONET:
    204   1.62     bjh21 		sockp->sa_family = AF_UNSPEC;
    205   1.62     bjh21 		hlen = 6;
    206   1.62     bjh21 		align = 2;
    207   1.12   mycroft 		break;
    208   1.12   mycroft 
    209   1.12   mycroft 	case DLT_NULL:
    210   1.12   mycroft 		sockp->sa_family = AF_UNSPEC;
    211   1.12   mycroft 		hlen = 0;
    212   1.36  christos 		align = 0;
    213   1.12   mycroft 		break;
    214   1.12   mycroft 
    215   1.12   mycroft 	default:
    216   1.12   mycroft 		return (EIO);
    217   1.12   mycroft 	}
    218   1.12   mycroft 
    219   1.12   mycroft 	len = uio->uio_resid;
    220   1.36  christos 	/*
    221   1.36  christos 	 * If there aren't enough bytes for a link level header or the
    222   1.36  christos 	 * packet length exceeds the interface mtu, return an error.
    223   1.36  christos 	 */
    224   1.36  christos 	if (len < hlen || len - hlen > mtu)
    225   1.36  christos 		return (EMSGSIZE);
    226   1.36  christos 
    227   1.36  christos 	/*
    228   1.36  christos 	 * XXX Avoid complicated buffer chaining ---
    229   1.36  christos 	 * bail if it won't fit in a single mbuf.
    230   1.36  christos 	 * (Take into account possible alignment bytes)
    231   1.36  christos 	 */
    232   1.36  christos 	if ((unsigned)len > MCLBYTES - align)
    233   1.12   mycroft 		return (EIO);
    234   1.12   mycroft 
    235   1.77      matt 	m = m_gethdr(M_WAIT, MT_DATA);
    236   1.20   mycroft 	m->m_pkthdr.rcvif = 0;
    237   1.20   mycroft 	m->m_pkthdr.len = len - hlen;
    238   1.36  christos 	if (len > MHLEN - align) {
    239   1.77      matt 		m_clget(m, M_WAIT);
    240   1.12   mycroft 		if ((m->m_flags & M_EXT) == 0) {
    241   1.12   mycroft 			error = ENOBUFS;
    242   1.12   mycroft 			goto bad;
    243   1.12   mycroft 		}
    244   1.12   mycroft 	}
    245   1.36  christos 
    246   1.36  christos 	/* Insure the data is properly aligned */
    247   1.36  christos 	if (align > 0) {
    248   1.36  christos 		m->m_data += align;
    249   1.36  christos 		m->m_len -= align;
    250   1.36  christos 	}
    251   1.36  christos 
    252  1.103  christos 	error = uiomove(mtod(m, void *), len, uio);
    253   1.36  christos 	if (error)
    254   1.36  christos 		goto bad;
    255   1.12   mycroft 	if (hlen != 0) {
    256  1.103  christos 		memcpy(sockp->sa_data, mtod(m, void *), hlen);
    257   1.12   mycroft 		m->m_data += hlen; /* XXX */
    258   1.36  christos 		len -= hlen;
    259   1.12   mycroft 	}
    260   1.36  christos 	m->m_len = len;
    261   1.36  christos 	*mp = m;
    262   1.38   mycroft 	return (0);
    263   1.38   mycroft 
    264   1.38   mycroft bad:
    265   1.12   mycroft 	m_freem(m);
    266   1.12   mycroft 	return (error);
    267   1.12   mycroft }
    268    1.1       cgd 
    269    1.1       cgd /*
    270    1.2       cgd  * Attach file to the bpf interface, i.e. make d listen on bp.
    271   1.61   thorpej  * Must be called at splnet.
    272    1.1       cgd  */
    273    1.1       cgd static void
    274    1.1       cgd bpf_attachd(d, bp)
    275    1.1       cgd 	struct bpf_d *d;
    276    1.1       cgd 	struct bpf_if *bp;
    277    1.1       cgd {
    278    1.2       cgd 	/*
    279    1.2       cgd 	 * Point d at bp, and add d to the interface's list of listeners.
    280    1.2       cgd 	 * Finally, point the driver's bpf cookie at the interface so
    281    1.2       cgd 	 * it will divert packets to bpf.
    282    1.2       cgd 	 */
    283    1.1       cgd 	d->bd_bif = bp;
    284    1.1       cgd 	d->bd_next = bp->bif_dlist;
    285    1.1       cgd 	bp->bif_dlist = d;
    286    1.1       cgd 
    287    1.1       cgd 	*bp->bif_driverp = bp;
    288    1.1       cgd }
    289    1.1       cgd 
    290    1.2       cgd /*
    291    1.2       cgd  * Detach a file from its interface.
    292    1.2       cgd  */
    293    1.1       cgd static void
    294    1.1       cgd bpf_detachd(d)
    295    1.1       cgd 	struct bpf_d *d;
    296    1.1       cgd {
    297    1.1       cgd 	struct bpf_d **p;
    298    1.1       cgd 	struct bpf_if *bp;
    299    1.1       cgd 
    300    1.1       cgd 	bp = d->bd_bif;
    301    1.1       cgd 	/*
    302    1.1       cgd 	 * Check if this descriptor had requested promiscuous mode.
    303    1.1       cgd 	 * If so, turn it off.
    304    1.1       cgd 	 */
    305    1.1       cgd 	if (d->bd_promisc) {
    306   1.18   mycroft 		int error;
    307   1.18   mycroft 
    308    1.1       cgd 		d->bd_promisc = 0;
    309   1.36  christos 		/*
    310   1.36  christos 		 * Take device out of promiscuous mode.  Since we were
    311   1.36  christos 		 * able to enter promiscuous mode, we should be able
    312   1.36  christos 		 * to turn it off.  But we can get an error if
    313   1.36  christos 		 * the interface was configured down, so only panic
    314   1.36  christos 		 * if we don't get an unexpected error.
    315   1.36  christos 		 */
    316   1.36  christos   		error = ifpromisc(bp->bif_ifp, 0);
    317   1.18   mycroft 		if (error && error != EINVAL)
    318    1.2       cgd 			panic("bpf: ifpromisc failed");
    319    1.1       cgd 	}
    320    1.2       cgd 	/* Remove d from the interface's descriptor list. */
    321    1.1       cgd 	p = &bp->bif_dlist;
    322    1.1       cgd 	while (*p != d) {
    323    1.1       cgd 		p = &(*p)->bd_next;
    324    1.1       cgd 		if (*p == 0)
    325    1.1       cgd 			panic("bpf_detachd: descriptor not in list");
    326    1.1       cgd 	}
    327    1.1       cgd 	*p = (*p)->bd_next;
    328    1.1       cgd 	if (bp->bif_dlist == 0)
    329    1.1       cgd 		/*
    330    1.1       cgd 		 * Let the driver know that there are no more listeners.
    331    1.1       cgd 		 */
    332    1.1       cgd 		*d->bd_bif->bif_driverp = 0;
    333    1.1       cgd 	d->bd_bif = 0;
    334    1.1       cgd }
    335    1.1       cgd 
    336    1.1       cgd 
    337    1.1       cgd /*
    338   1.12   mycroft  * Mark a descriptor free by making it point to itself.
    339    1.1       cgd  * This is probably cheaper than marking with a constant since
    340    1.1       cgd  * the address should be in a register anyway.
    341    1.1       cgd  */
    342    1.1       cgd #define D_ISFREE(d) ((d) == (d)->bd_next)
    343    1.1       cgd #define D_MARKFREE(d) ((d)->bd_next = (d))
    344    1.1       cgd #define D_MARKUSED(d) ((d)->bd_next = 0)
    345    1.1       cgd 
    346    1.1       cgd /*
    347   1.46    bouyer  * bpfilterattach() is called at boot time.
    348   1.46    bouyer  */
    349   1.46    bouyer /* ARGSUSED */
    350   1.46    bouyer void
    351   1.46    bouyer bpfilterattach(n)
    352   1.46    bouyer 	int n;
    353   1.46    bouyer {
    354   1.46    bouyer 	int i;
    355   1.46    bouyer 	/*
    356   1.46    bouyer 	 * Mark all the descriptors free.
    357   1.46    bouyer 	 */
    358   1.46    bouyer 	for (i = 0; i < NBPFILTER; ++i)
    359   1.46    bouyer 		D_MARKFREE(&bpf_dtab[i]);
    360   1.46    bouyer 
    361   1.46    bouyer }
    362   1.46    bouyer 
    363   1.46    bouyer /*
    364    1.2       cgd  * Open ethernet device.  Returns ENXIO for illegal minor device number,
    365    1.2       cgd  * EBUSY if file is open by another process.
    366    1.1       cgd  */
    367    1.1       cgd /* ARGSUSED */
    368    1.1       cgd int
    369   1.82      fvdl bpfopen(dev, flag, mode, p)
    370    1.1       cgd 	dev_t dev;
    371    1.1       cgd 	int flag;
    372   1.24  christos 	int mode;
    373   1.82      fvdl 	struct proc *p;
    374    1.1       cgd {
    375   1.53  augustss 	struct bpf_d *d;
    376   1.12   mycroft 
    377    1.1       cgd 	if (minor(dev) >= NBPFILTER)
    378    1.1       cgd 		return (ENXIO);
    379    1.1       cgd 	/*
    380    1.1       cgd 	 * Each minor can be opened by only one process.  If the requested
    381    1.1       cgd 	 * minor is in use, return EBUSY.
    382    1.1       cgd 	 */
    383    1.1       cgd 	d = &bpf_dtab[minor(dev)];
    384    1.2       cgd 	if (!D_ISFREE(d))
    385    1.1       cgd 		return (EBUSY);
    386    1.2       cgd 
    387    1.2       cgd 	/* Mark "free" and do most initialization. */
    388   1.41     perry 	memset((char *)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    1.1       cgd 	return (0);
    394    1.1       cgd }
    395    1.1       cgd 
    396    1.1       cgd /*
    397    1.1       cgd  * Close the descriptor by detaching it from its interface,
    398    1.1       cgd  * deallocating its buffers, and marking it free.
    399    1.1       cgd  */
    400    1.1       cgd /* ARGSUSED */
    401    1.2       cgd int
    402   1.82      fvdl bpfclose(dev, flag, mode, p)
    403    1.1       cgd 	dev_t dev;
    404    1.1       cgd 	int flag;
    405   1.24  christos 	int mode;
    406   1.82      fvdl 	struct proc *p;
    407    1.1       cgd {
    408   1.53  augustss 	struct bpf_d *d = &bpf_dtab[minor(dev)];
    409   1.53  augustss 	int s;
    410    1.1       cgd 
    411   1.61   thorpej 	s = splnet();
    412   1.91   darrenr 	if (d->bd_state == BPF_WAITING)
    413   1.91   darrenr 		callout_stop(&d->bd_callout);
    414   1.91   darrenr 	d->bd_state = BPF_IDLE;
    415    1.1       cgd 	if (d->bd_bif)
    416    1.1       cgd 		bpf_detachd(d);
    417    1.1       cgd 	splx(s);
    418    1.2       cgd 	bpf_freed(d);
    419    1.2       cgd 
    420    1.2       cgd 	return (0);
    421    1.2       cgd }
    422    1.2       cgd 
    423    1.2       cgd /*
    424    1.1       cgd  * Rotate the packet buffers in descriptor d.  Move the store buffer
    425   1.12   mycroft  * into the hold slot, and the free buffer into the store slot.
    426    1.1       cgd  * Zero the length of the new store buffer.
    427    1.1       cgd  */
    428    1.1       cgd #define ROTATE_BUFFERS(d) \
    429  1.100   darrenr 	(d)->bd_hbuf = (d)->bd_sbuf; \
    430  1.100   darrenr 	(d)->bd_hlen = (d)->bd_slen; \
    431  1.100   darrenr 	(d)->bd_sbuf = (d)->bd_fbuf; \
    432  1.100   darrenr 	(d)->bd_slen = 0; \
    433  1.100   darrenr 	(d)->bd_fbuf = 0;
    434    1.1       cgd /*
    435    1.1       cgd  *  bpfread - read next chunk of packets from buffers
    436    1.1       cgd  */
    437    1.1       cgd int
    438   1.24  christos bpfread(dev, uio, ioflag)
    439    1.1       cgd 	dev_t dev;
    440   1.53  augustss 	struct uio *uio;
    441   1.24  christos 	int ioflag;
    442    1.1       cgd {
    443   1.53  augustss 	struct bpf_d *d = &bpf_dtab[minor(dev)];
    444   1.91   darrenr 	int timed_out;
    445    1.1       cgd 	int error;
    446    1.1       cgd 	int s;
    447    1.1       cgd 
    448    1.1       cgd 	/*
    449   1.12   mycroft 	 * Restrict application to use a buffer the same size as
    450    1.1       cgd 	 * as kernel buffers.
    451    1.1       cgd 	 */
    452    1.1       cgd 	if (uio->uio_resid != d->bd_bufsize)
    453    1.1       cgd 		return (EINVAL);
    454    1.1       cgd 
    455   1.61   thorpej 	s = splnet();
    456   1.91   darrenr 	if (d->bd_state == BPF_WAITING)
    457   1.91   darrenr 		callout_stop(&d->bd_callout);
    458   1.91   darrenr 	timed_out = (d->bd_state == BPF_TIMED_OUT);
    459   1.91   darrenr 	d->bd_state = BPF_IDLE;
    460    1.1       cgd 	/*
    461    1.2       cgd 	 * If the hold buffer is empty, then do a timed sleep, which
    462    1.2       cgd 	 * ends when the timeout expires or when enough packets
    463    1.2       cgd 	 * have arrived to fill the store buffer.
    464    1.1       cgd 	 */
    465    1.1       cgd 	while (d->bd_hbuf == 0) {
    466   1.93   darrenr 		if (ioflag & IO_NDELAY) {
    467   1.93   darrenr 			if (d->bd_slen == 0) {
    468  1.100   darrenr 				splx(s);
    469   1.93   darrenr 				return (EWOULDBLOCK);
    470   1.93   darrenr 			}
    471   1.93   darrenr 			ROTATE_BUFFERS(d);
    472   1.93   darrenr 			break;
    473   1.93   darrenr 		}
    474   1.93   darrenr 
    475   1.91   darrenr 		if ((d->bd_immediate || timed_out) && d->bd_slen != 0) {
    476    1.1       cgd 			/*
    477    1.1       cgd 			 * A packet(s) either arrived since the previous
    478    1.1       cgd 			 * read or arrived while we were asleep.
    479    1.1       cgd 			 * Rotate the buffers and return what's here.
    480    1.1       cgd 			 */
    481    1.1       cgd 			ROTATE_BUFFERS(d);
    482    1.1       cgd 			break;
    483    1.1       cgd 		}
    484  1.103  christos 		error = tsleep(d, PRINET|PCATCH, "bpf",
    485   1.93   darrenr 				d->bd_rtout);
    486   1.12   mycroft 		if (error == EINTR || error == ERESTART) {
    487   1.12   mycroft 			splx(s);
    488   1.12   mycroft 			return (error);
    489   1.12   mycroft 		}
    490   1.12   mycroft 		if (error == EWOULDBLOCK) {
    491   1.12   mycroft 			/*
    492   1.12   mycroft 			 * On a timeout, return what's in the buffer,
    493   1.12   mycroft 			 * which may be nothing.  If there is something
    494   1.12   mycroft 			 * in the store buffer, we can rotate the buffers.
    495   1.12   mycroft 			 */
    496   1.12   mycroft 			if (d->bd_hbuf)
    497    1.1       cgd 				/*
    498   1.12   mycroft 				 * We filled up the buffer in between
    499   1.12   mycroft 				 * getting the timeout and arriving
    500   1.12   mycroft 				 * here, so we don't need to rotate.
    501    1.1       cgd 				 */
    502    1.1       cgd 				break;
    503   1.12   mycroft 
    504  1.100   darrenr 			if (d->bd_slen == 0) {
    505  1.100   darrenr 				splx(s);
    506  1.100   darrenr 				return (0);
    507    1.1       cgd 			}
    508  1.100   darrenr 			ROTATE_BUFFERS(d);
    509  1.100   darrenr 			break;
    510    1.1       cgd 		}
    511   1.36  christos 		if (error != 0)
    512  1.100   darrenr 			goto done;
    513    1.1       cgd 	}
    514  1.100   darrenr 	/*
    515  1.100   darrenr 	 * At this point, we know we have something in the hold slot.
    516  1.100   darrenr 	 */
    517  1.100   darrenr 	splx(s);
    518  1.100   darrenr 
    519  1.100   darrenr 	/*
    520  1.100   darrenr 	 * Move data from hold buffer into user space.
    521  1.100   darrenr 	 * We know the entire buffer is transferred since
    522  1.100   darrenr 	 * we checked above that the read buffer is bpf_bufsize bytes.
    523  1.100   darrenr 	 */
    524  1.100   darrenr 	error = uiomove(d->bd_hbuf, d->bd_hlen, uio);
    525   1.12   mycroft 
    526  1.100   darrenr 	s = splnet();
    527  1.100   darrenr 	d->bd_fbuf = d->bd_hbuf;
    528  1.100   darrenr 	d->bd_hbuf = 0;
    529  1.100   darrenr 	d->bd_hlen = 0;
    530  1.100   darrenr done:
    531  1.100   darrenr 	splx(s);
    532  1.100   darrenr 	return (error);
    533    1.1       cgd }
    534    1.1       cgd 
    535    1.1       cgd 
    536    1.1       cgd /*
    537   1.12   mycroft  * If there are processes sleeping on this descriptor, wake them up.
    538    1.1       cgd  */
    539   1.15       cgd static __inline void
    540    1.1       cgd bpf_wakeup(d)
    541   1.53  augustss 	struct bpf_d *d;
    542    1.1       cgd {
    543  1.103  christos 	wakeup(d);
    544   1.85  jdolecek 	if (d->bd_async)
    545   1.86  christos 		fownsignal(d->bd_pgid, SIGIO, 0, 0, NULL);
    546   1.23   thorpej 
    547   1.75  jdolecek 	selnotify(&d->bd_sel, 0);
    548    1.2       cgd 	/* XXX */
    549   1.76  christos 	d->bd_sel.sel_pid = 0;
    550    1.1       cgd }
    551    1.1       cgd 
    552   1.91   darrenr 
    553   1.91   darrenr static void
    554   1.91   darrenr bpf_timed_out(arg)
    555   1.91   darrenr 	void *arg;
    556   1.91   darrenr {
    557  1.103  christos 	struct bpf_d *d = arg;
    558   1.91   darrenr 	int s;
    559   1.91   darrenr 
    560   1.91   darrenr 	s = splnet();
    561   1.91   darrenr 	if (d->bd_state == BPF_WAITING) {
    562   1.91   darrenr 		d->bd_state = BPF_TIMED_OUT;
    563   1.91   darrenr 		if (d->bd_slen != 0)
    564   1.91   darrenr 			bpf_wakeup(d);
    565   1.91   darrenr 	}
    566   1.91   darrenr 	splx(s);
    567   1.91   darrenr }
    568   1.91   darrenr 
    569   1.91   darrenr 
    570    1.1       cgd int
    571   1.24  christos bpfwrite(dev, uio, ioflag)
    572    1.1       cgd 	dev_t dev;
    573    1.1       cgd 	struct uio *uio;
    574   1.24  christos 	int ioflag;
    575    1.1       cgd {
    576   1.53  augustss 	struct bpf_d *d = &bpf_dtab[minor(dev)];
    577   1.12   mycroft 	struct ifnet *ifp;
    578   1.12   mycroft 	struct mbuf *m;
    579   1.12   mycroft 	int error, s;
    580   1.56      matt 	static struct sockaddr_storage dst;
    581    1.1       cgd 
    582   1.12   mycroft 	if (d->bd_bif == 0)
    583    1.1       cgd 		return (ENXIO);
    584    1.1       cgd 
    585   1.12   mycroft 	ifp = d->bd_bif->bif_ifp;
    586   1.11   deraadt 
    587   1.12   mycroft 	if (uio->uio_resid == 0)
    588   1.12   mycroft 		return (0);
    589   1.11   deraadt 
    590   1.56      matt 	error = bpf_movein(uio, (int)d->bd_bif->bif_dlt, ifp->if_mtu, &m,
    591   1.56      matt 		(struct sockaddr *) &dst);
    592   1.12   mycroft 	if (error)
    593   1.12   mycroft 		return (error);
    594   1.11   deraadt 
    595   1.20   mycroft 	if (m->m_pkthdr.len > ifp->if_mtu)
    596    1.1       cgd 		return (EMSGSIZE);
    597    1.1       cgd 
    598   1.40   thorpej 	if (d->bd_hdrcmplt)
    599   1.56      matt 		dst.ss_family = pseudo_AF_HDRCMPLT;
    600   1.40   thorpej 
    601   1.21   mycroft 	s = splsoftnet();
    602   1.56      matt 	error = (*ifp->if_output)(ifp, m, (struct sockaddr *) &dst, NULL);
    603    1.1       cgd 	splx(s);
    604    1.1       cgd 	/*
    605   1.12   mycroft 	 * The driver frees the mbuf.
    606    1.1       cgd 	 */
    607    1.1       cgd 	return (error);
    608    1.1       cgd }
    609    1.1       cgd 
    610    1.1       cgd /*
    611    1.2       cgd  * Reset a descriptor by flushing its packet buffer and clearing the
    612   1.61   thorpej  * receive and drop counts.  Should be called at splnet.
    613    1.1       cgd  */
    614    1.1       cgd static void
    615    1.1       cgd reset_d(d)
    616    1.1       cgd 	struct bpf_d *d;
    617    1.1       cgd {
    618    1.1       cgd 	if (d->bd_hbuf) {
    619    1.1       cgd 		/* Free the hold buffer. */
    620    1.1       cgd 		d->bd_fbuf = d->bd_hbuf;
    621    1.1       cgd 		d->bd_hbuf = 0;
    622    1.1       cgd 	}
    623    1.1       cgd 	d->bd_slen = 0;
    624    1.2       cgd 	d->bd_hlen = 0;
    625    1.1       cgd 	d->bd_rcount = 0;
    626    1.1       cgd 	d->bd_dcount = 0;
    627   1.94   darrenr 	d->bd_ccount = 0;
    628    1.1       cgd }
    629    1.1       cgd 
    630   1.36  christos #ifdef BPF_KERN_FILTER
    631   1.36  christos extern struct bpf_insn *bpf_tcp_filter;
    632   1.36  christos extern struct bpf_insn *bpf_udp_filter;
    633   1.36  christos #endif
    634   1.36  christos 
    635    1.1       cgd /*
    636    1.1       cgd  *  FIONREAD		Check for read packet available.
    637    1.1       cgd  *  BIOCGBLEN		Get buffer len [for read()].
    638    1.1       cgd  *  BIOCSETF		Set ethernet read filter.
    639    1.1       cgd  *  BIOCFLUSH		Flush read packet buffer.
    640    1.1       cgd  *  BIOCPROMISC		Put interface into promiscuous mode.
    641    1.1       cgd  *  BIOCGDLT		Get link layer type.
    642    1.1       cgd  *  BIOCGETIF		Get interface name.
    643    1.1       cgd  *  BIOCSETIF		Set interface.
    644    1.1       cgd  *  BIOCSRTIMEOUT	Set read timeout.
    645    1.1       cgd  *  BIOCGRTIMEOUT	Get read timeout.
    646    1.1       cgd  *  BIOCGSTATS		Get packet stats.
    647    1.1       cgd  *  BIOCIMMEDIATE	Set immediate mode.
    648    1.2       cgd  *  BIOCVERSION		Get filter language version.
    649   1.40   thorpej  *  BIOGHDRCMPLT	Get "header already complete" flag.
    650   1.40   thorpej  *  BIOSHDRCMPLT	Set "header already complete" flag.
    651    1.1       cgd  */
    652    1.1       cgd /* ARGSUSED */
    653    1.1       cgd int
    654  1.103  christos bpfioctl(dev_t dev, u_long cmd, caddr_t arg, int flag, struct proc *p)
    655    1.1       cgd {
    656   1.53  augustss 	struct bpf_d *d = &bpf_dtab[minor(dev)];
    657    1.1       cgd 	int s, error = 0;
    658   1.36  christos #ifdef BPF_KERN_FILTER
    659   1.53  augustss 	struct bpf_insn **p;
    660   1.36  christos #endif
    661  1.103  christos 	void *addr = arg;
    662    1.1       cgd 
    663   1.91   darrenr 	s = splnet();
    664   1.91   darrenr 	if (d->bd_state == BPF_WAITING)
    665   1.91   darrenr 		callout_stop(&d->bd_callout);
    666   1.91   darrenr 	d->bd_state = BPF_IDLE;
    667   1.91   darrenr 	splx(s);
    668   1.91   darrenr 
    669    1.1       cgd 	switch (cmd) {
    670    1.1       cgd 
    671    1.1       cgd 	default:
    672    1.1       cgd 		error = EINVAL;
    673    1.1       cgd 		break;
    674    1.1       cgd 
    675    1.1       cgd 	/*
    676    1.1       cgd 	 * Check for read packet available.
    677    1.1       cgd 	 */
    678    1.1       cgd 	case FIONREAD:
    679    1.1       cgd 		{
    680    1.1       cgd 			int n;
    681   1.12   mycroft 
    682   1.61   thorpej 			s = splnet();
    683    1.1       cgd 			n = d->bd_slen;
    684   1.12   mycroft 			if (d->bd_hbuf)
    685    1.1       cgd 				n += d->bd_hlen;
    686    1.1       cgd 			splx(s);
    687    1.1       cgd 
    688    1.1       cgd 			*(int *)addr = n;
    689    1.1       cgd 			break;
    690    1.1       cgd 		}
    691    1.1       cgd 
    692    1.1       cgd 	/*
    693    1.2       cgd 	 * Get buffer len [for read()].
    694    1.1       cgd 	 */
    695    1.2       cgd 	case BIOCGBLEN:
    696    1.2       cgd 		*(u_int *)addr = d->bd_bufsize;
    697    1.1       cgd 		break;
    698    1.2       cgd 
    699    1.1       cgd 	/*
    700    1.2       cgd 	 * Set buffer length.
    701    1.1       cgd 	 */
    702    1.2       cgd 	case BIOCSBLEN:
    703    1.2       cgd 		if (d->bd_bif != 0)
    704    1.2       cgd 			error = EINVAL;
    705    1.2       cgd 		else {
    706   1.53  augustss 			u_int size = *(u_int *)addr;
    707    1.2       cgd 
    708   1.87  jonathan 			if (size > bpf_maxbufsize)
    709   1.87  jonathan 				*(u_int *)addr = size = bpf_maxbufsize;
    710    1.2       cgd 			else if (size < BPF_MINBUFSIZE)
    711    1.2       cgd 				*(u_int *)addr = size = BPF_MINBUFSIZE;
    712    1.2       cgd 			d->bd_bufsize = size;
    713    1.2       cgd 		}
    714    1.1       cgd 		break;
    715    1.1       cgd 
    716    1.1       cgd 	/*
    717    1.2       cgd 	 * Set link layer read filter.
    718    1.1       cgd 	 */
    719   1.12   mycroft 	case BIOCSETF:
    720  1.103  christos 		error = bpf_setf(d, addr);
    721    1.1       cgd 		break;
    722    1.1       cgd 
    723   1.36  christos #ifdef BPF_KERN_FILTER
    724   1.36  christos 	/*
    725   1.36  christos 	 * Set TCP or UDP reject filter.
    726   1.36  christos 	 */
    727   1.36  christos 	case BIOCSTCPF:
    728   1.36  christos 	case BIOCSUDPF:
    729   1.36  christos 		if (!suser()) {
    730   1.36  christos 			error = EPERM;
    731   1.36  christos 			break;
    732   1.36  christos 		}
    733   1.36  christos 
    734   1.36  christos 		/* Validate and store filter */
    735  1.103  christos 		error = bpf_setf(d, addr);
    736   1.36  christos 
    737   1.36  christos 		/* Free possible old filter */
    738   1.36  christos 		if (cmd == BIOCSTCPF)
    739   1.36  christos 			p = &bpf_tcp_filter;
    740   1.36  christos 		else
    741   1.36  christos 			p = &bpf_udp_filter;
    742   1.36  christos 		if (*p != NULL)
    743  1.103  christos 			free(*p, M_DEVBUF);
    744   1.36  christos 
    745   1.36  christos 		/* Steal new filter (noop if error) */
    746   1.61   thorpej 		s = splnet();
    747   1.36  christos 		*p = d->bd_filter;
    748   1.36  christos 		d->bd_filter = NULL;
    749   1.36  christos 		splx(s);
    750   1.36  christos 		break;
    751   1.36  christos #endif
    752   1.36  christos 
    753    1.1       cgd 	/*
    754    1.1       cgd 	 * Flush read packet buffer.
    755    1.1       cgd 	 */
    756    1.1       cgd 	case BIOCFLUSH:
    757   1.61   thorpej 		s = splnet();
    758    1.1       cgd 		reset_d(d);
    759    1.1       cgd 		splx(s);
    760    1.1       cgd 		break;
    761    1.1       cgd 
    762    1.1       cgd 	/*
    763    1.1       cgd 	 * Put interface into promiscuous mode.
    764    1.1       cgd 	 */
    765    1.1       cgd 	case BIOCPROMISC:
    766    1.1       cgd 		if (d->bd_bif == 0) {
    767    1.1       cgd 			/*
    768    1.1       cgd 			 * No interface attached yet.
    769    1.1       cgd 			 */
    770    1.1       cgd 			error = EINVAL;
    771    1.1       cgd 			break;
    772    1.1       cgd 		}
    773   1.61   thorpej 		s = splnet();
    774    1.1       cgd 		if (d->bd_promisc == 0) {
    775    1.1       cgd 			error = ifpromisc(d->bd_bif->bif_ifp, 1);
    776    1.2       cgd 			if (error == 0)
    777    1.2       cgd 				d->bd_promisc = 1;
    778    1.1       cgd 		}
    779    1.1       cgd 		splx(s);
    780    1.1       cgd 		break;
    781    1.1       cgd 
    782    1.1       cgd 	/*
    783    1.1       cgd 	 * Get device parameters.
    784    1.1       cgd 	 */
    785    1.1       cgd 	case BIOCGDLT:
    786    1.1       cgd 		if (d->bd_bif == 0)
    787    1.1       cgd 			error = EINVAL;
    788    1.1       cgd 		else
    789    1.1       cgd 			*(u_int *)addr = d->bd_bif->bif_dlt;
    790    1.1       cgd 		break;
    791    1.1       cgd 
    792    1.1       cgd 	/*
    793   1.66      onoe 	 * Get a list of supported device parameters.
    794   1.66      onoe 	 */
    795   1.66      onoe 	case BIOCGDLTLIST:
    796   1.66      onoe 		if (d->bd_bif == 0)
    797   1.66      onoe 			error = EINVAL;
    798   1.66      onoe 		else
    799  1.103  christos 			error = bpf_getdltlist(d, addr);
    800   1.66      onoe 		break;
    801   1.66      onoe 
    802   1.66      onoe 	/*
    803   1.66      onoe 	 * Set device parameters.
    804   1.66      onoe 	 */
    805   1.66      onoe 	case BIOCSDLT:
    806   1.66      onoe 		if (d->bd_bif == 0)
    807   1.66      onoe 			error = EINVAL;
    808   1.66      onoe 		else
    809   1.66      onoe 			error = bpf_setdlt(d, *(u_int *)addr);
    810   1.66      onoe 		break;
    811   1.66      onoe 
    812   1.66      onoe 	/*
    813    1.1       cgd 	 * Set interface name.
    814    1.1       cgd 	 */
    815    1.1       cgd 	case BIOCGETIF:
    816    1.1       cgd 		if (d->bd_bif == 0)
    817    1.1       cgd 			error = EINVAL;
    818    1.1       cgd 		else
    819  1.103  christos 			bpf_ifname(d->bd_bif->bif_ifp, addr);
    820    1.1       cgd 		break;
    821    1.1       cgd 
    822    1.1       cgd 	/*
    823    1.1       cgd 	 * Set interface.
    824    1.1       cgd 	 */
    825    1.1       cgd 	case BIOCSETIF:
    826  1.103  christos 		error = bpf_setif(d, addr);
    827    1.1       cgd 		break;
    828    1.1       cgd 
    829    1.1       cgd 	/*
    830    1.1       cgd 	 * Set read timeout.
    831    1.1       cgd 	 */
    832   1.12   mycroft 	case BIOCSRTIMEOUT:
    833    1.1       cgd 		{
    834  1.103  christos 			struct timeval *tv = addr;
    835    1.1       cgd 
    836   1.19       cgd 			/* Compute number of ticks. */
    837   1.19       cgd 			d->bd_rtout = tv->tv_sec * hz + tv->tv_usec / tick;
    838   1.33   thorpej 			if ((d->bd_rtout == 0) && (tv->tv_usec != 0))
    839   1.33   thorpej 				d->bd_rtout = 1;
    840    1.1       cgd 			break;
    841    1.1       cgd 		}
    842    1.1       cgd 
    843    1.1       cgd 	/*
    844    1.1       cgd 	 * Get read timeout.
    845    1.1       cgd 	 */
    846   1.12   mycroft 	case BIOCGRTIMEOUT:
    847    1.1       cgd 		{
    848  1.103  christos 			struct timeval *tv = addr;
    849    1.1       cgd 
    850   1.19       cgd 			tv->tv_sec = d->bd_rtout / hz;
    851   1.19       cgd 			tv->tv_usec = (d->bd_rtout % hz) * tick;
    852    1.1       cgd 			break;
    853    1.1       cgd 		}
    854    1.1       cgd 
    855    1.1       cgd 	/*
    856    1.1       cgd 	 * Get packet stats.
    857    1.1       cgd 	 */
    858    1.1       cgd 	case BIOCGSTATS:
    859    1.1       cgd 		{
    860  1.103  christos 			struct bpf_stat *bs = addr;
    861    1.1       cgd 
    862    1.1       cgd 			bs->bs_recv = d->bd_rcount;
    863    1.1       cgd 			bs->bs_drop = d->bd_dcount;
    864   1.94   darrenr 			bs->bs_capt = d->bd_ccount;
    865   1.94   darrenr 			break;
    866   1.94   darrenr 		}
    867   1.94   darrenr 
    868   1.94   darrenr 	case BIOCGSTATSOLD:
    869   1.94   darrenr 		{
    870  1.103  christos 			struct bpf_stat_old *bs = addr;
    871   1.94   darrenr 
    872   1.94   darrenr 			bs->bs_recv = d->bd_rcount;
    873   1.94   darrenr 			bs->bs_drop = d->bd_dcount;
    874    1.1       cgd 			break;
    875    1.1       cgd 		}
    876    1.1       cgd 
    877    1.1       cgd 	/*
    878    1.1       cgd 	 * Set immediate mode.
    879    1.1       cgd 	 */
    880    1.1       cgd 	case BIOCIMMEDIATE:
    881    1.1       cgd 		d->bd_immediate = *(u_int *)addr;
    882    1.1       cgd 		break;
    883    1.2       cgd 
    884    1.2       cgd 	case BIOCVERSION:
    885    1.2       cgd 		{
    886  1.103  christos 			struct bpf_version *bv = addr;
    887    1.2       cgd 
    888    1.2       cgd 			bv->bv_major = BPF_MAJOR_VERSION;
    889    1.2       cgd 			bv->bv_minor = BPF_MINOR_VERSION;
    890    1.2       cgd 			break;
    891   1.12   mycroft 		}
    892   1.23   thorpej 
    893   1.40   thorpej 	case BIOCGHDRCMPLT:	/* get "header already complete" flag */
    894   1.40   thorpej 		*(u_int *)addr = d->bd_hdrcmplt;
    895   1.40   thorpej 		break;
    896   1.40   thorpej 
    897   1.40   thorpej 	case BIOCSHDRCMPLT:	/* set "header already complete" flag */
    898   1.40   thorpej 		d->bd_hdrcmplt = *(u_int *)addr ? 1 : 0;
    899   1.40   thorpej 		break;
    900   1.23   thorpej 
    901   1.91   darrenr 	/*
    902   1.91   darrenr 	 * Get "see sent packets" flag
    903   1.91   darrenr 	 */
    904   1.91   darrenr 	case BIOCGSEESENT:
    905   1.91   darrenr 		*(u_int *)addr = d->bd_seesent;
    906   1.91   darrenr 		break;
    907   1.91   darrenr 
    908   1.91   darrenr 	/*
    909   1.91   darrenr 	 * Set "see sent" packets flag
    910   1.91   darrenr 	 */
    911   1.91   darrenr 	case BIOCSSEESENT:
    912   1.91   darrenr 		d->bd_seesent = *(u_int *)addr;
    913   1.91   darrenr 		break;
    914   1.91   darrenr 
    915   1.23   thorpej 	case FIONBIO:		/* Non-blocking I/O */
    916   1.93   darrenr 		/*
    917   1.93   darrenr 		 * No need to do anything special as we use IO_NDELAY in
    918   1.93   darrenr 		 * bpfread() as an indication of whether or not to block
    919   1.93   darrenr 		 * the read.
    920   1.93   darrenr 		 */
    921   1.23   thorpej 		break;
    922   1.23   thorpej 
    923   1.23   thorpej 	case FIOASYNC:		/* Send signal on receive packets */
    924   1.23   thorpej 		d->bd_async = *(int *)addr;
    925   1.23   thorpej 		break;
    926   1.23   thorpej 
    927   1.23   thorpej 	case TIOCSPGRP:		/* Process or group to send signals to */
    928   1.85  jdolecek 	case FIOSETOWN:
    929   1.85  jdolecek 		error = fsetown(p, &d->bd_pgid, cmd, addr);
    930   1.23   thorpej 		break;
    931   1.23   thorpej 
    932   1.23   thorpej 	case TIOCGPGRP:
    933   1.85  jdolecek 	case FIOGETOWN:
    934   1.85  jdolecek 		error = fgetown(p, d->bd_pgid, cmd, addr);
    935   1.23   thorpej 		break;
    936    1.1       cgd 	}
    937    1.1       cgd 	return (error);
    938    1.1       cgd }
    939    1.1       cgd 
    940   1.12   mycroft /*
    941    1.2       cgd  * Set d's packet filter program to fp.  If this file already has a filter,
    942    1.1       cgd  * free it and replace it.  Returns EINVAL for bogus requests.
    943    1.1       cgd  */
    944    1.1       cgd int
    945  1.103  christos bpf_setf(struct bpf_d *d, struct bpf_program *fp)
    946    1.1       cgd {
    947    1.1       cgd 	struct bpf_insn *fcode, *old;
    948    1.1       cgd 	u_int flen, size;
    949    1.1       cgd 	int s;
    950    1.1       cgd 
    951    1.1       cgd 	old = d->bd_filter;
    952    1.1       cgd 	if (fp->bf_insns == 0) {
    953    1.1       cgd 		if (fp->bf_len != 0)
    954    1.1       cgd 			return (EINVAL);
    955   1.61   thorpej 		s = splnet();
    956    1.1       cgd 		d->bd_filter = 0;
    957    1.1       cgd 		reset_d(d);
    958    1.1       cgd 		splx(s);
    959    1.1       cgd 		if (old != 0)
    960  1.103  christos 			free(old, M_DEVBUF);
    961    1.1       cgd 		return (0);
    962    1.1       cgd 	}
    963    1.1       cgd 	flen = fp->bf_len;
    964    1.1       cgd 	if (flen > BPF_MAXINSNS)
    965    1.1       cgd 		return (EINVAL);
    966    1.1       cgd 
    967    1.1       cgd 	size = flen * sizeof(*fp->bf_insns);
    968  1.103  christos 	fcode = malloc(size, M_DEVBUF, M_WAITOK);
    969  1.103  christos 	if (copyin(fp->bf_insns, fcode, size) == 0 &&
    970    1.2       cgd 	    bpf_validate(fcode, (int)flen)) {
    971   1.61   thorpej 		s = splnet();
    972    1.1       cgd 		d->bd_filter = fcode;
    973    1.1       cgd 		reset_d(d);
    974    1.1       cgd 		splx(s);
    975    1.1       cgd 		if (old != 0)
    976  1.103  christos 			free(old, M_DEVBUF);
    977    1.1       cgd 
    978    1.1       cgd 		return (0);
    979    1.1       cgd 	}
    980  1.103  christos 	free(fcode, M_DEVBUF);
    981    1.1       cgd 	return (EINVAL);
    982    1.1       cgd }
    983    1.1       cgd 
    984    1.1       cgd /*
    985    1.2       cgd  * Detach a file from its current interface (if attached at all) and attach
    986    1.2       cgd  * to the interface indicated by the name stored in ifr.
    987    1.2       cgd  * Return an errno or 0.
    988    1.1       cgd  */
    989    1.1       cgd static int
    990  1.103  christos bpf_setif(struct bpf_d *d, struct ifreq *ifr)
    991    1.1       cgd {
    992    1.1       cgd 	struct bpf_if *bp;
    993    1.1       cgd 	char *cp;
    994   1.26   thorpej 	int unit_seen, i, s, error;
    995    1.1       cgd 
    996    1.1       cgd 	/*
    997   1.26   thorpej 	 * Make sure the provided name has a unit number, and default
    998   1.26   thorpej 	 * it to '0' if not specified.
    999   1.26   thorpej 	 * XXX This is ugly ... do this differently?
   1000    1.1       cgd 	 */
   1001   1.26   thorpej 	unit_seen = 0;
   1002    1.1       cgd 	cp = ifr->ifr_name;
   1003   1.26   thorpej 	cp[sizeof(ifr->ifr_name) - 1] = '\0';	/* sanity */
   1004   1.26   thorpej 	while (*cp++)
   1005   1.26   thorpej 		if (*cp >= '0' && *cp <= '9')
   1006   1.26   thorpej 			unit_seen = 1;
   1007   1.26   thorpej 	if (!unit_seen) {
   1008   1.26   thorpej 		/* Make sure to leave room for the '\0'. */
   1009   1.26   thorpej 		for (i = 0; i < (IFNAMSIZ - 1); ++i) {
   1010   1.26   thorpej 			if ((ifr->ifr_name[i] >= 'a' &&
   1011   1.26   thorpej 			     ifr->ifr_name[i] <= 'z') ||
   1012   1.26   thorpej 			    (ifr->ifr_name[i] >= 'A' &&
   1013   1.26   thorpej 			     ifr->ifr_name[i] <= 'Z'))
   1014   1.26   thorpej 				continue;
   1015   1.26   thorpej 			ifr->ifr_name[i] = '0';
   1016    1.1       cgd 		}
   1017    1.1       cgd 	}
   1018   1.26   thorpej 
   1019    1.1       cgd 	/*
   1020    1.1       cgd 	 * Look through attached interfaces for the named one.
   1021    1.1       cgd 	 */
   1022    1.1       cgd 	for (bp = bpf_iflist; bp != 0; bp = bp->bif_next) {
   1023    1.1       cgd 		struct ifnet *ifp = bp->bif_ifp;
   1024    1.1       cgd 
   1025   1.26   thorpej 		if (ifp == 0 ||
   1026   1.26   thorpej 		    strcmp(ifp->if_xname, ifr->ifr_name) != 0)
   1027    1.1       cgd 			continue;
   1028   1.66      onoe 		/* skip additional entry */
   1029   1.66      onoe 		if (bp->bif_driverp != (struct bpf_if **)&ifp->if_bpf)
   1030   1.66      onoe 			continue;
   1031    1.1       cgd 		/*
   1032    1.2       cgd 		 * We found the requested interface.
   1033    1.2       cgd 		 * Allocate the packet buffers if we need to.
   1034    1.2       cgd 		 * If we're already attached to requested interface,
   1035    1.2       cgd 		 * just flush the buffer.
   1036    1.1       cgd 		 */
   1037    1.2       cgd 		if (d->bd_sbuf == 0) {
   1038    1.2       cgd 			error = bpf_allocbufs(d);
   1039    1.2       cgd 			if (error != 0)
   1040    1.2       cgd 				return (error);
   1041    1.2       cgd 		}
   1042   1.61   thorpej 		s = splnet();
   1043    1.1       cgd 		if (bp != d->bd_bif) {
   1044    1.1       cgd 			if (d->bd_bif)
   1045   1.12   mycroft 				/*
   1046    1.1       cgd 				 * Detach if attached to something else.
   1047    1.1       cgd 				 */
   1048    1.1       cgd 				bpf_detachd(d);
   1049    1.1       cgd 
   1050    1.1       cgd 			bpf_attachd(d, bp);
   1051    1.1       cgd 		}
   1052    1.1       cgd 		reset_d(d);
   1053    1.1       cgd 		splx(s);
   1054    1.1       cgd 		return (0);
   1055    1.1       cgd 	}
   1056    1.1       cgd 	/* Not found. */
   1057    1.1       cgd 	return (ENXIO);
   1058    1.1       cgd }
   1059    1.1       cgd 
   1060    1.1       cgd /*
   1061   1.26   thorpej  * Copy the interface name to the ifreq.
   1062    1.1       cgd  */
   1063    1.1       cgd static void
   1064  1.103  christos bpf_ifname(struct ifnet *ifp, struct ifreq *ifr)
   1065    1.1       cgd {
   1066   1.41     perry 	memcpy(ifr->ifr_name, ifp->if_xname, IFNAMSIZ);
   1067    1.1       cgd }
   1068    1.1       cgd 
   1069    1.1       cgd /*
   1070   1.36  christos  * Support for poll() system call
   1071    1.1       cgd  *
   1072   1.64   darrenr  * Return true iff the specific operation will not block indefinitely - with
   1073   1.64   darrenr  * the assumption that it is safe to positively acknowledge a request for the
   1074   1.64   darrenr  * ability to write to the BPF device.
   1075    1.2       cgd  * Otherwise, return false but make a note that a selwakeup() must be done.
   1076    1.1       cgd  */
   1077    1.1       cgd int
   1078  1.103  christos bpfpoll(dev_t dev, int events, struct proc *p)
   1079    1.1       cgd {
   1080   1.53  augustss 	struct bpf_d *d = &bpf_dtab[minor(dev)];
   1081   1.61   thorpej 	int s = splnet();
   1082   1.64   darrenr 	int revents;
   1083   1.12   mycroft 
   1084   1.64   darrenr 	revents = events & (POLLOUT | POLLWRNORM);
   1085   1.44   thorpej 	if (events & (POLLIN | POLLRDNORM)) {
   1086   1.64   darrenr 		/*
   1087   1.64   darrenr 		 * An imitation of the FIONREAD ioctl code.
   1088   1.64   darrenr 		 */
   1089   1.93   darrenr 		if ((d->bd_hlen != 0) ||
   1090   1.93   darrenr 		    (d->bd_immediate && d->bd_slen != 0)) {
   1091   1.30   mycroft 			revents |= events & (POLLIN | POLLRDNORM);
   1092   1.93   darrenr 		} else if (d->bd_state == BPF_TIMED_OUT) {
   1093   1.95   darrenr 			if (d->bd_slen != 0)
   1094   1.95   darrenr 				revents |= events & (POLLIN | POLLRDNORM);
   1095   1.95   darrenr 			else
   1096   1.95   darrenr 				revents |= events & POLLIN;
   1097   1.92   darrenr 		} else {
   1098   1.82      fvdl 			selrecord(p, &d->bd_sel);
   1099   1.91   darrenr 			/* Start the read timeout if necessary */
   1100   1.91   darrenr 			if (d->bd_rtout > 0 && d->bd_state == BPF_IDLE) {
   1101   1.91   darrenr 				callout_reset(&d->bd_callout, d->bd_rtout,
   1102   1.91   darrenr 					      bpf_timed_out, d);
   1103   1.91   darrenr 				d->bd_state = BPF_WAITING;
   1104   1.91   darrenr 			}
   1105   1.91   darrenr 		}
   1106   1.44   thorpej 	}
   1107   1.12   mycroft 
   1108   1.12   mycroft 	splx(s);
   1109   1.30   mycroft 	return (revents);
   1110   1.75  jdolecek }
   1111   1.75  jdolecek 
   1112   1.75  jdolecek static void
   1113   1.75  jdolecek filt_bpfrdetach(struct knote *kn)
   1114   1.75  jdolecek {
   1115   1.75  jdolecek 	struct bpf_d *d = kn->kn_hook;
   1116   1.75  jdolecek 	int s;
   1117   1.75  jdolecek 
   1118   1.75  jdolecek 	s = splnet();
   1119   1.76  christos 	SLIST_REMOVE(&d->bd_sel.sel_klist, kn, knote, kn_selnext);
   1120   1.75  jdolecek 	splx(s);
   1121   1.75  jdolecek }
   1122   1.75  jdolecek 
   1123   1.75  jdolecek static int
   1124   1.75  jdolecek filt_bpfread(struct knote *kn, long hint)
   1125   1.75  jdolecek {
   1126   1.75  jdolecek 	struct bpf_d *d = kn->kn_hook;
   1127   1.75  jdolecek 
   1128   1.75  jdolecek 	kn->kn_data = d->bd_hlen;
   1129   1.75  jdolecek 	if (d->bd_immediate)
   1130   1.75  jdolecek 		kn->kn_data += d->bd_slen;
   1131   1.75  jdolecek 	return (kn->kn_data > 0);
   1132   1.75  jdolecek }
   1133   1.75  jdolecek 
   1134   1.75  jdolecek static const struct filterops bpfread_filtops =
   1135   1.75  jdolecek 	{ 1, NULL, filt_bpfrdetach, filt_bpfread };
   1136   1.75  jdolecek 
   1137   1.75  jdolecek int
   1138  1.103  christos bpfkqfilter(dev_t dev, struct knote *kn)
   1139   1.75  jdolecek {
   1140   1.75  jdolecek 	struct bpf_d *d = &bpf_dtab[minor(dev)];
   1141   1.75  jdolecek 	struct klist *klist;
   1142   1.75  jdolecek 	int s;
   1143   1.75  jdolecek 
   1144   1.75  jdolecek 	switch (kn->kn_filter) {
   1145   1.75  jdolecek 	case EVFILT_READ:
   1146   1.76  christos 		klist = &d->bd_sel.sel_klist;
   1147   1.75  jdolecek 		kn->kn_fop = &bpfread_filtops;
   1148   1.75  jdolecek 		break;
   1149   1.75  jdolecek 
   1150   1.75  jdolecek 	default:
   1151   1.75  jdolecek 		return (1);
   1152   1.75  jdolecek 	}
   1153   1.75  jdolecek 
   1154   1.75  jdolecek 	kn->kn_hook = d;
   1155   1.75  jdolecek 
   1156   1.75  jdolecek 	s = splnet();
   1157   1.75  jdolecek 	SLIST_INSERT_HEAD(klist, kn, kn_selnext);
   1158   1.75  jdolecek 	splx(s);
   1159   1.75  jdolecek 
   1160   1.75  jdolecek 	return (0);
   1161    1.1       cgd }
   1162    1.1       cgd 
   1163    1.1       cgd /*
   1164    1.2       cgd  * Incoming linkage from device drivers.  Process the packet pkt, of length
   1165    1.2       cgd  * pktlen, which is stored in a contiguous buffer.  The packet is parsed
   1166    1.2       cgd  * by each process' filter, and if accepted, stashed into the corresponding
   1167    1.2       cgd  * buffer.
   1168    1.1       cgd  */
   1169    1.1       cgd void
   1170  1.103  christos bpf_tap(void *arg, u_char *pkt, u_int pktlen)
   1171    1.1       cgd {
   1172    1.1       cgd 	struct bpf_if *bp;
   1173   1.53  augustss 	struct bpf_d *d;
   1174   1.53  augustss 	u_int slen;
   1175    1.1       cgd 	/*
   1176    1.1       cgd 	 * Note that the ipl does not have to be raised at this point.
   1177    1.1       cgd 	 * The only problem that could arise here is that if two different
   1178    1.1       cgd 	 * interfaces shared any data.  This is not the case.
   1179    1.1       cgd 	 */
   1180  1.103  christos 	bp = arg;
   1181    1.1       cgd 	for (d = bp->bif_dlist; d != 0; d = d->bd_next) {
   1182    1.1       cgd 		++d->bd_rcount;
   1183    1.1       cgd 		slen = bpf_filter(d->bd_filter, pkt, pktlen, pktlen);
   1184    1.1       cgd 		if (slen != 0)
   1185   1.41     perry 			catchpacket(d, pkt, pktlen, slen, memcpy);
   1186    1.1       cgd 	}
   1187    1.1       cgd }
   1188    1.1       cgd 
   1189    1.1       cgd /*
   1190    1.1       cgd  * Copy data from an mbuf chain into a buffer.  This code is derived
   1191    1.1       cgd  * from m_copydata in sys/uipc_mbuf.c.
   1192    1.1       cgd  */
   1193   1.43     perry static void *
   1194  1.103  christos bpf_mcpy(void *dst_arg, const void *src_arg, size_t len)
   1195   1.12   mycroft {
   1196   1.53  augustss 	const struct mbuf *m;
   1197   1.53  augustss 	u_int count;
   1198    1.1       cgd 	u_char *dst;
   1199    1.1       cgd 
   1200   1.12   mycroft 	m = src_arg;
   1201   1.12   mycroft 	dst = dst_arg;
   1202    1.1       cgd 	while (len > 0) {
   1203    1.1       cgd 		if (m == 0)
   1204   1.41     perry 			panic("bpf_mcpy");
   1205   1.12   mycroft 		count = min(m->m_len, len);
   1206  1.103  christos 		memcpy(dst, mtod(m, void *), count);
   1207    1.1       cgd 		m = m->m_next;
   1208    1.1       cgd 		dst += count;
   1209    1.1       cgd 		len -= count;
   1210    1.1       cgd 	}
   1211   1.68    itojun 	return (dst_arg);
   1212    1.1       cgd }
   1213    1.1       cgd 
   1214    1.1       cgd /*
   1215   1.96    dyoung  * Dispatch a packet to all the listeners on interface bp.
   1216   1.96    dyoung  *
   1217   1.96    dyoung  * marg    pointer to the packet, either a data buffer or an mbuf chain
   1218   1.96    dyoung  * buflen  buffer length, if marg is a data buffer
   1219   1.96    dyoung  * cpfn    a function that can copy marg into the listener's buffer
   1220   1.96    dyoung  * pktlen  length of the packet
   1221   1.96    dyoung  * rcvif   either NULL or the interface the packet came in on.
   1222   1.96    dyoung  */
   1223   1.96    dyoung static __inline void
   1224   1.96    dyoung bpf_deliver(struct bpf_if *bp, void *(*cpfn)(void *, const void *, size_t),
   1225   1.96    dyoung     void *marg, u_int pktlen, u_int buflen, struct ifnet *rcvif)
   1226   1.96    dyoung {
   1227   1.96    dyoung 	u_int slen;
   1228   1.96    dyoung 	struct bpf_d *d;
   1229   1.96    dyoung 
   1230   1.96    dyoung 	for (d = bp->bif_dlist; d != 0; d = d->bd_next) {
   1231   1.96    dyoung 		if (!d->bd_seesent && (rcvif == NULL))
   1232   1.96    dyoung 			continue;
   1233   1.96    dyoung 		++d->bd_rcount;
   1234   1.96    dyoung 		slen = bpf_filter(d->bd_filter, marg, pktlen, buflen);
   1235   1.96    dyoung 		if (slen != 0)
   1236   1.96    dyoung 			catchpacket(d, marg, pktlen, slen, cpfn);
   1237   1.96    dyoung 	}
   1238   1.96    dyoung }
   1239   1.96    dyoung 
   1240   1.96    dyoung /*
   1241   1.96    dyoung  * Incoming linkage from device drivers, when the head of the packet is in
   1242   1.96    dyoung  * a buffer, and the tail is in an mbuf chain.
   1243   1.96    dyoung  */
   1244   1.96    dyoung void
   1245  1.103  christos bpf_mtap2(void *arg, void *data, u_int dlen, struct mbuf *m)
   1246   1.96    dyoung {
   1247  1.103  christos 	struct bpf_if *bp = arg;
   1248   1.96    dyoung 	u_int pktlen;
   1249   1.96    dyoung 	struct mbuf mb;
   1250   1.96    dyoung 
   1251  1.101    dyoung 	pktlen = m_length(m) + dlen;
   1252   1.96    dyoung 
   1253   1.96    dyoung 	/*
   1254   1.96    dyoung 	 * Craft on-stack mbuf suitable for passing to bpf_filter.
   1255   1.96    dyoung 	 * Note that we cut corners here; we only setup what's
   1256   1.96    dyoung 	 * absolutely needed--this mbuf should never go anywhere else.
   1257   1.96    dyoung 	 */
   1258  1.101    dyoung 	(void)memset(&mb, 0, sizeof(mb));
   1259   1.96    dyoung 	mb.m_next = m;
   1260   1.96    dyoung 	mb.m_data = data;
   1261   1.96    dyoung 	mb.m_len = dlen;
   1262   1.96    dyoung 
   1263   1.96    dyoung 	bpf_deliver(bp, bpf_mcpy, &mb, pktlen, 0, m->m_pkthdr.rcvif);
   1264   1.96    dyoung }
   1265   1.96    dyoung 
   1266   1.96    dyoung /*
   1267    1.2       cgd  * Incoming linkage from device drivers, when packet is in an mbuf chain.
   1268    1.1       cgd  */
   1269    1.1       cgd void
   1270  1.103  christos bpf_mtap(void *arg, struct mbuf *m)
   1271    1.1       cgd {
   1272  1.103  christos 	void *(*cpfn)(void *, const void *, size_t);
   1273  1.103  christos 	struct bpf_if *bp = arg;
   1274   1.96    dyoung 	u_int pktlen, buflen;
   1275   1.91   darrenr 	void *marg;
   1276    1.1       cgd 
   1277  1.101    dyoung 	pktlen = m_length(m);
   1278    1.1       cgd 
   1279   1.91   darrenr 	if (pktlen == m->m_len) {
   1280   1.91   darrenr 		cpfn = memcpy;
   1281   1.91   darrenr 		marg = mtod(m, void *);
   1282   1.91   darrenr 		buflen = pktlen;
   1283   1.91   darrenr 	} else {
   1284   1.91   darrenr 		cpfn = bpf_mcpy;
   1285   1.91   darrenr 		marg = m;
   1286   1.91   darrenr 		buflen = 0;
   1287   1.91   darrenr 	}
   1288   1.91   darrenr 
   1289   1.96    dyoung 	bpf_deliver(bp, cpfn, marg, pktlen, buflen, m->m_pkthdr.rcvif);
   1290    1.1       cgd }
   1291    1.1       cgd 
   1292    1.1       cgd /*
   1293  1.104  christos  * We need to prepend the address family as
   1294  1.104  christos  * a four byte field.  Cons up a dummy header
   1295  1.104  christos  * to pacify bpf.  This is safe because bpf
   1296  1.104  christos  * will only read from the mbuf (i.e., it won't
   1297  1.104  christos  * try to free it or keep a pointer a to it).
   1298  1.104  christos  */
   1299  1.104  christos void
   1300  1.104  christos bpf_mtap_af(void *arg, u_int32_t af, struct mbuf *m)
   1301  1.104  christos {
   1302  1.104  christos 	struct mbuf m0;
   1303  1.104  christos 
   1304  1.104  christos 	m0.m_flags = 0;
   1305  1.104  christos 	m0.m_next = m;
   1306  1.104  christos 	m0.m_len = 4;
   1307  1.104  christos 	m0.m_data = (char *)&af;
   1308  1.104  christos 
   1309  1.104  christos 	bpf_mtap(arg, &m0);
   1310  1.104  christos }
   1311  1.104  christos 
   1312  1.104  christos void
   1313  1.104  christos bpf_mtap_et(void *arg, u_int16_t et, struct mbuf *m)
   1314  1.104  christos {
   1315  1.104  christos 	struct mbuf m0;
   1316  1.104  christos 
   1317  1.104  christos 	m0.m_flags = 0;
   1318  1.104  christos 	m0.m_next = m;
   1319  1.104  christos 	m0.m_len = 14;
   1320  1.104  christos 	m0.m_data = m0.m_dat;
   1321  1.104  christos 
   1322  1.104  christos 	((u_int32_t *)m0.m_data)[0] = 0;
   1323  1.104  christos 	((u_int32_t *)m0.m_data)[1] = 0;
   1324  1.104  christos 	((u_int32_t *)m0.m_data)[2] = 0;
   1325  1.104  christos 	((u_int16_t *)m0.m_data)[6] = et;
   1326  1.104  christos 
   1327  1.104  christos 	bpf_mtap(arg, &m0);
   1328  1.104  christos }
   1329  1.104  christos 
   1330  1.104  christos #if NSL > 0 || NSTRIP > 0
   1331  1.104  christos /*
   1332  1.104  christos  * Put the SLIP pseudo-"link header" in place.
   1333  1.104  christos  * Note this M_PREPEND() should never fail,
   1334  1.104  christos  * swince we know we always have enough space
   1335  1.104  christos  * in the input buffer.
   1336  1.104  christos  */
   1337  1.104  christos void
   1338  1.104  christos bpf_mtap_sl_in(void *arg, u_char *chdr, struct mbuf **m)
   1339  1.104  christos {
   1340  1.104  christos 	int s;
   1341  1.104  christos 	u_char *hp;
   1342  1.104  christos 
   1343  1.104  christos 	M_PREPEND(*m, SLIP_HDRLEN, M_DONTWAIT);
   1344  1.104  christos 	if (*m == NULL)
   1345  1.104  christos 		return;
   1346  1.104  christos 
   1347  1.104  christos 	hp = mtod(*m, u_char *);
   1348  1.104  christos 	hp[SLX_DIR] = SLIPDIR_IN;
   1349  1.104  christos 	(void)memcpy(&hp[SLX_CHDR], chdr, CHDR_LEN);
   1350  1.104  christos 
   1351  1.104  christos 	s = splnet();
   1352  1.104  christos 	bpf_mtap(arg, *m);
   1353  1.104  christos 	splx(s);
   1354  1.104  christos 
   1355  1.104  christos 	m_adj(*m, SLIP_HDRLEN);
   1356  1.104  christos }
   1357  1.104  christos 
   1358  1.104  christos /*
   1359  1.104  christos  * Put the SLIP pseudo-"link header" in
   1360  1.104  christos  * place.  The compressed header is now
   1361  1.104  christos  * at the beginning of the mbuf.
   1362  1.104  christos  */
   1363  1.104  christos void
   1364  1.104  christos bpf_mtap_sl_out(void *arg, u_char *chdr, struct mbuf *m)
   1365  1.104  christos {
   1366  1.104  christos 	struct mbuf m0;
   1367  1.104  christos 	u_char *hp;
   1368  1.104  christos 	int s;
   1369  1.104  christos 
   1370  1.104  christos 	m0.m_flags = 0;
   1371  1.104  christos 	m0.m_next = m;
   1372  1.104  christos 	m0.m_data = m0.m_dat;
   1373  1.104  christos 	m0.m_len = SLIP_HDRLEN;
   1374  1.104  christos 
   1375  1.104  christos 	hp = mtod(&m0, u_char *);
   1376  1.104  christos 
   1377  1.104  christos 	hp[SLX_DIR] = SLIPDIR_OUT;
   1378  1.104  christos 	(void)memcpy(&hp[SLX_CHDR], chdr, CHDR_LEN);
   1379  1.104  christos 
   1380  1.104  christos 	s = splnet();
   1381  1.104  christos 	bpf_mtap(arg, &m0);
   1382  1.104  christos 	splx(s);
   1383  1.104  christos 	m_freem(m);
   1384  1.104  christos }
   1385  1.104  christos #endif
   1386  1.104  christos 
   1387  1.104  christos /*
   1388    1.1       cgd  * Move the packet data from interface memory (pkt) into the
   1389    1.1       cgd  * store buffer.  Return 1 if it's time to wakeup a listener (buffer full),
   1390    1.2       cgd  * otherwise 0.  "copy" is the routine called to do the actual data
   1391   1.41     perry  * transfer.  memcpy is passed in to copy contiguous chunks, while
   1392   1.41     perry  * bpf_mcpy is passed in to copy mbuf chains.  In the latter case,
   1393    1.2       cgd  * pkt is really an mbuf.
   1394    1.1       cgd  */
   1395    1.1       cgd static void
   1396  1.103  christos catchpacket(struct bpf_d *d, u_char *pkt, u_int pktlen, u_int snaplen,
   1397  1.103  christos     void *(*cpfn)(void *, const void *, size_t))
   1398   1.53  augustss {
   1399   1.53  augustss 	struct bpf_hdr *hp;
   1400   1.53  augustss 	int totlen, curlen;
   1401   1.53  augustss 	int hdrlen = d->bd_bif->bif_hdrlen;
   1402   1.94   darrenr 
   1403   1.94   darrenr 	++d->bd_ccount;
   1404    1.1       cgd 	/*
   1405    1.1       cgd 	 * Figure out how many bytes to move.  If the packet is
   1406    1.1       cgd 	 * greater or equal to the snapshot length, transfer that
   1407    1.1       cgd 	 * much.  Otherwise, transfer the whole packet (unless
   1408    1.1       cgd 	 * we hit the buffer size limit).
   1409    1.1       cgd 	 */
   1410   1.12   mycroft 	totlen = hdrlen + min(snaplen, pktlen);
   1411    1.1       cgd 	if (totlen > d->bd_bufsize)
   1412    1.1       cgd 		totlen = d->bd_bufsize;
   1413    1.1       cgd 
   1414    1.1       cgd 	/*
   1415    1.1       cgd 	 * Round up the end of the previous packet to the next longword.
   1416    1.1       cgd 	 */
   1417    1.1       cgd 	curlen = BPF_WORDALIGN(d->bd_slen);
   1418    1.1       cgd 	if (curlen + totlen > d->bd_bufsize) {
   1419    1.1       cgd 		/*
   1420    1.1       cgd 		 * This packet will overflow the storage buffer.
   1421    1.1       cgd 		 * Rotate the buffers if we can, then wakeup any
   1422    1.1       cgd 		 * pending reads.
   1423    1.1       cgd 		 */
   1424  1.100   darrenr 		if (d->bd_fbuf == 0) {
   1425   1.12   mycroft 			/*
   1426   1.12   mycroft 			 * We haven't completed the previous read yet,
   1427    1.1       cgd 			 * so drop the packet.
   1428    1.1       cgd 			 */
   1429    1.1       cgd 			++d->bd_dcount;
   1430    1.1       cgd 			return;
   1431    1.1       cgd 		}
   1432    1.1       cgd 		ROTATE_BUFFERS(d);
   1433    1.1       cgd 		bpf_wakeup(d);
   1434    1.1       cgd 		curlen = 0;
   1435   1.97   darrenr 	}
   1436    1.1       cgd 
   1437    1.1       cgd 	/*
   1438    1.1       cgd 	 * Append the bpf header.
   1439    1.1       cgd 	 */
   1440    1.1       cgd 	hp = (struct bpf_hdr *)(d->bd_sbuf + curlen);
   1441    1.2       cgd 	microtime(&hp->bh_tstamp);
   1442    1.1       cgd 	hp->bh_datalen = pktlen;
   1443    1.1       cgd 	hp->bh_hdrlen = hdrlen;
   1444    1.1       cgd 	/*
   1445    1.1       cgd 	 * Copy the packet data into the store buffer and update its length.
   1446    1.1       cgd 	 */
   1447   1.41     perry 	(*cpfn)((u_char *)hp + hdrlen, pkt, (hp->bh_caplen = totlen - hdrlen));
   1448    1.1       cgd 	d->bd_slen = curlen + totlen;
   1449   1.97   darrenr 
   1450   1.97   darrenr 	/*
   1451   1.97   darrenr 	 * Call bpf_wakeup after bd_slen has been updated so that kevent(2)
   1452   1.97   darrenr 	 * will cause filt_bpfread() to be called with it adjusted.
   1453   1.97   darrenr 	 */
   1454   1.97   darrenr 	if (d->bd_immediate || d->bd_state == BPF_TIMED_OUT)
   1455   1.97   darrenr 		/*
   1456   1.97   darrenr 		 * Immediate mode is set, or the read timeout has
   1457   1.97   darrenr 		 * already expired during a select call.  A packet
   1458   1.97   darrenr 		 * arrived, so the reader should be woken up.
   1459   1.97   darrenr 		 */
   1460   1.97   darrenr 		bpf_wakeup(d);
   1461    1.1       cgd }
   1462    1.1       cgd 
   1463   1.12   mycroft /*
   1464    1.1       cgd  * Initialize all nonzero fields of a descriptor.
   1465    1.1       cgd  */
   1466    1.1       cgd static int
   1467  1.103  christos bpf_allocbufs(struct bpf_d *d)
   1468    1.1       cgd {
   1469   1.50     enami 
   1470  1.103  christos 	d->bd_fbuf = malloc(d->bd_bufsize, M_DEVBUF, M_NOWAIT);
   1471  1.100   darrenr 	if (!d->bd_fbuf)
   1472   1.79    itojun 		return (ENOBUFS);
   1473  1.103  christos 	d->bd_sbuf = malloc(d->bd_bufsize, M_DEVBUF, M_NOWAIT);
   1474  1.100   darrenr 	if (!d->bd_sbuf) {
   1475  1.100   darrenr 		free(d->bd_fbuf, M_DEVBUF);
   1476   1.79    itojun 		return (ENOBUFS);
   1477   1.79    itojun 	}
   1478    1.1       cgd 	d->bd_slen = 0;
   1479    1.1       cgd 	d->bd_hlen = 0;
   1480    1.1       cgd 	return (0);
   1481    1.1       cgd }
   1482    1.1       cgd 
   1483    1.1       cgd /*
   1484    1.2       cgd  * Free buffers currently in use by a descriptor.
   1485    1.2       cgd  * Called on close.
   1486    1.2       cgd  */
   1487    1.2       cgd static void
   1488  1.103  christos bpf_freed(struct bpf_d *d)
   1489    1.2       cgd {
   1490    1.2       cgd 	/*
   1491    1.2       cgd 	 * We don't need to lock out interrupts since this descriptor has
   1492    1.2       cgd 	 * been detached from its interface and it yet hasn't been marked
   1493    1.2       cgd 	 * free.
   1494    1.2       cgd 	 */
   1495  1.100   darrenr 	if (d->bd_sbuf != 0) {
   1496  1.100   darrenr 		free(d->bd_sbuf, M_DEVBUF);
   1497  1.100   darrenr 		if (d->bd_hbuf != 0)
   1498  1.100   darrenr 			free(d->bd_hbuf, M_DEVBUF);
   1499  1.100   darrenr 		if (d->bd_fbuf != 0)
   1500  1.100   darrenr 			free(d->bd_fbuf, M_DEVBUF);
   1501  1.100   darrenr 	}
   1502    1.2       cgd 	if (d->bd_filter)
   1503  1.103  christos 		free(d->bd_filter, M_DEVBUF);
   1504    1.2       cgd 
   1505    1.2       cgd 	D_MARKFREE(d);
   1506    1.2       cgd }
   1507    1.2       cgd 
   1508    1.2       cgd /*
   1509   1.60   thorpej  * Attach an interface to bpf.  dlt is the link layer type; hdrlen is the
   1510   1.60   thorpej  * fixed size of the link header (variable length headers not yet supported).
   1511    1.1       cgd  */
   1512    1.1       cgd void
   1513  1.103  christos bpfattach(struct ifnet *ifp, u_int dlt, u_int hdrlen)
   1514    1.1       cgd {
   1515   1.66      onoe 
   1516   1.66      onoe 	bpfattach2(ifp, dlt, hdrlen, &ifp->if_bpf);
   1517   1.66      onoe }
   1518   1.66      onoe 
   1519   1.66      onoe /*
   1520   1.66      onoe  * Attach additional dlt for a interface to bpf.  dlt is the link layer type;
   1521   1.66      onoe  * hdrlen is the fixed size of the link header for the specified dlt
   1522   1.66      onoe  * (variable length headers not yet supported).
   1523   1.66      onoe  */
   1524   1.66      onoe void
   1525  1.103  christos bpfattach2(struct ifnet *ifp, u_int dlt, u_int hdrlen, void *driverp)
   1526   1.66      onoe {
   1527    1.1       cgd 	struct bpf_if *bp;
   1528  1.103  christos 	bp = malloc(sizeof(*bp), M_DEVBUF, M_DONTWAIT);
   1529   1.12   mycroft 	if (bp == 0)
   1530   1.12   mycroft 		panic("bpfattach");
   1531   1.12   mycroft 
   1532    1.1       cgd 	bp->bif_dlist = 0;
   1533  1.103  christos 	bp->bif_driverp = driverp;
   1534    1.1       cgd 	bp->bif_ifp = ifp;
   1535    1.1       cgd 	bp->bif_dlt = dlt;
   1536    1.1       cgd 
   1537    1.1       cgd 	bp->bif_next = bpf_iflist;
   1538    1.1       cgd 	bpf_iflist = bp;
   1539    1.1       cgd 
   1540    1.1       cgd 	*bp->bif_driverp = 0;
   1541    1.1       cgd 
   1542    1.1       cgd 	/*
   1543    1.1       cgd 	 * Compute the length of the bpf header.  This is not necessarily
   1544   1.12   mycroft 	 * equal to SIZEOF_BPF_HDR because we want to insert spacing such
   1545   1.12   mycroft 	 * that the network layer header begins on a longword boundary (for
   1546    1.1       cgd 	 * performance reasons and to alleviate alignment restrictions).
   1547    1.1       cgd 	 */
   1548    1.1       cgd 	bp->bif_hdrlen = BPF_WORDALIGN(hdrlen + SIZEOF_BPF_HDR) - hdrlen;
   1549    1.1       cgd 
   1550   1.11   deraadt #if 0
   1551   1.32  christos 	printf("bpf: %s attached\n", ifp->if_xname);
   1552   1.11   deraadt #endif
   1553   1.48   thorpej }
   1554   1.48   thorpej 
   1555   1.48   thorpej /*
   1556   1.48   thorpej  * Remove an interface from bpf.
   1557   1.48   thorpej  */
   1558   1.48   thorpej void
   1559  1.103  christos bpfdetach(struct ifnet *ifp)
   1560   1.48   thorpej {
   1561   1.48   thorpej 	struct bpf_if *bp, **pbp;
   1562   1.51     enami 	struct bpf_d *d;
   1563   1.51     enami 	int i, s, cmaj;
   1564   1.51     enami 
   1565   1.51     enami 	/* locate the major number */
   1566   1.67   gehenna 	cmaj = cdevsw_lookup_major(&bpf_cdevsw);
   1567   1.51     enami 
   1568   1.52     soren 	/* Nuke the vnodes for any open instances */
   1569   1.51     enami 	for (i = 0; i < NBPFILTER; ++i) {
   1570   1.51     enami 		d = &bpf_dtab[i];
   1571   1.51     enami 		if (!D_ISFREE(d) && d->bd_bif != NULL &&
   1572   1.51     enami 		    d->bd_bif->bif_ifp == ifp) {
   1573   1.51     enami 			/*
   1574   1.51     enami 			 * Detach the descriptor from an interface now.
   1575   1.51     enami 			 * It will be free'ed later by close routine.
   1576   1.51     enami 			 */
   1577   1.61   thorpej 			s = splnet();
   1578   1.51     enami 			d->bd_promisc = 0;	/* we can't touch device. */
   1579   1.51     enami 			bpf_detachd(d);
   1580   1.51     enami 			splx(s);
   1581   1.51     enami 			vdevgone(cmaj, i, i, VCHR);
   1582   1.51     enami 		}
   1583   1.51     enami 	}
   1584   1.48   thorpej 
   1585   1.66      onoe   again:
   1586   1.48   thorpej 	for (bp = bpf_iflist, pbp = &bpf_iflist;
   1587   1.48   thorpej 	     bp != NULL; pbp = &bp->bif_next, bp = bp->bif_next) {
   1588   1.48   thorpej 		if (bp->bif_ifp == ifp) {
   1589   1.48   thorpej 			*pbp = bp->bif_next;
   1590   1.48   thorpej 			free(bp, M_DEVBUF);
   1591   1.66      onoe 			goto again;
   1592   1.48   thorpej 		}
   1593   1.48   thorpej 	}
   1594   1.47   thorpej }
   1595   1.47   thorpej 
   1596   1.47   thorpej /*
   1597   1.66      onoe  * Change the data link type of a interface.
   1598   1.47   thorpej  */
   1599   1.47   thorpej void
   1600  1.103  christos bpf_change_type(struct ifnet *ifp, u_int dlt, u_int hdrlen)
   1601   1.47   thorpej {
   1602   1.47   thorpej 	struct bpf_if *bp;
   1603   1.47   thorpej 
   1604   1.47   thorpej 	for (bp = bpf_iflist; bp != NULL; bp = bp->bif_next) {
   1605   1.59   thorpej 		if (bp->bif_driverp == (struct bpf_if **)&ifp->if_bpf)
   1606   1.47   thorpej 			break;
   1607   1.47   thorpej 	}
   1608   1.47   thorpej 	if (bp == NULL)
   1609   1.47   thorpej 		panic("bpf_change_type");
   1610   1.47   thorpej 
   1611   1.47   thorpej 	bp->bif_dlt = dlt;
   1612   1.47   thorpej 
   1613   1.47   thorpej 	/*
   1614   1.47   thorpej 	 * Compute the length of the bpf header.  This is not necessarily
   1615   1.47   thorpej 	 * equal to SIZEOF_BPF_HDR because we want to insert spacing such
   1616   1.47   thorpej 	 * that the network layer header begins on a longword boundary (for
   1617   1.47   thorpej 	 * performance reasons and to alleviate alignment restrictions).
   1618   1.47   thorpej 	 */
   1619   1.47   thorpej 	bp->bif_hdrlen = BPF_WORDALIGN(hdrlen + SIZEOF_BPF_HDR) - hdrlen;
   1620   1.66      onoe }
   1621   1.66      onoe 
   1622   1.66      onoe /*
   1623   1.66      onoe  * Get a list of available data link type of the interface.
   1624   1.66      onoe  */
   1625   1.66      onoe static int
   1626  1.103  christos bpf_getdltlist(struct bpf_d *d, struct bpf_dltlist *bfl)
   1627   1.66      onoe {
   1628   1.66      onoe 	int n, error;
   1629   1.66      onoe 	struct ifnet *ifp;
   1630   1.66      onoe 	struct bpf_if *bp;
   1631   1.66      onoe 
   1632   1.66      onoe 	ifp = d->bd_bif->bif_ifp;
   1633   1.66      onoe 	n = 0;
   1634   1.66      onoe 	error = 0;
   1635   1.66      onoe 	for (bp = bpf_iflist; bp != NULL; bp = bp->bif_next) {
   1636   1.66      onoe 		if (bp->bif_ifp != ifp)
   1637   1.66      onoe 			continue;
   1638   1.66      onoe 		if (bfl->bfl_list != NULL) {
   1639   1.66      onoe 			if (n >= bfl->bfl_len)
   1640   1.66      onoe 				return ENOMEM;
   1641   1.66      onoe 			error = copyout(&bp->bif_dlt,
   1642   1.66      onoe 			    bfl->bfl_list + n, sizeof(u_int));
   1643   1.66      onoe 		}
   1644   1.66      onoe 		n++;
   1645   1.66      onoe 	}
   1646   1.66      onoe 	bfl->bfl_len = n;
   1647   1.66      onoe 	return error;
   1648   1.66      onoe }
   1649   1.66      onoe 
   1650   1.66      onoe /*
   1651   1.66      onoe  * Set the data link type of a BPF instance.
   1652   1.66      onoe  */
   1653   1.66      onoe static int
   1654  1.103  christos bpf_setdlt(struct bpf_d *d, u_int dlt)
   1655   1.66      onoe {
   1656   1.69   thorpej 	int s, error, opromisc;
   1657   1.66      onoe 	struct ifnet *ifp;
   1658   1.66      onoe 	struct bpf_if *bp;
   1659   1.66      onoe 
   1660   1.66      onoe 	if (d->bd_bif->bif_dlt == dlt)
   1661   1.66      onoe 		return 0;
   1662   1.66      onoe 	ifp = d->bd_bif->bif_ifp;
   1663   1.66      onoe 	for (bp = bpf_iflist; bp != NULL; bp = bp->bif_next) {
   1664   1.66      onoe 		if (bp->bif_ifp == ifp && bp->bif_dlt == dlt)
   1665   1.66      onoe 			break;
   1666   1.66      onoe 	}
   1667   1.66      onoe 	if (bp == NULL)
   1668   1.66      onoe 		return EINVAL;
   1669   1.66      onoe 	s = splnet();
   1670   1.69   thorpej 	opromisc = d->bd_promisc;
   1671   1.66      onoe 	bpf_detachd(d);
   1672   1.66      onoe 	bpf_attachd(d, bp);
   1673   1.66      onoe 	reset_d(d);
   1674   1.69   thorpej 	if (opromisc) {
   1675   1.69   thorpej 		error = ifpromisc(bp->bif_ifp, 1);
   1676   1.69   thorpej 		if (error)
   1677   1.69   thorpej 			printf("%s: bpf_setdlt: ifpromisc failed (%d)\n",
   1678   1.69   thorpej 			    bp->bif_ifp->if_xname, error);
   1679   1.69   thorpej 		else
   1680   1.69   thorpej 			d->bd_promisc = 1;
   1681   1.69   thorpej 	}
   1682   1.66      onoe 	splx(s);
   1683   1.66      onoe 	return 0;
   1684    1.1       cgd }
   1685   1.89  jonathan 
   1686   1.89  jonathan static int
   1687   1.89  jonathan sysctl_net_bpf_maxbufsize(SYSCTLFN_ARGS)
   1688   1.89  jonathan {
   1689   1.89  jonathan 	int newsize, error;
   1690   1.89  jonathan 	struct sysctlnode node;
   1691   1.89  jonathan 
   1692   1.89  jonathan 	node = *rnode;
   1693   1.89  jonathan 	node.sysctl_data = &newsize;
   1694   1.89  jonathan 	newsize = bpf_maxbufsize;
   1695   1.89  jonathan 	error = sysctl_lookup(SYSCTLFN_CALL(&node));
   1696   1.89  jonathan 	if (error || newp == NULL)
   1697   1.89  jonathan 		return (error);
   1698   1.89  jonathan 
   1699   1.89  jonathan 	if (newsize < BPF_MINBUFSIZE || newsize > BPF_MAXBUFSIZE)
   1700   1.89  jonathan 		return (EINVAL);
   1701   1.89  jonathan 
   1702   1.89  jonathan 	bpf_maxbufsize = newsize;
   1703   1.89  jonathan 
   1704   1.89  jonathan 	return (0);
   1705   1.89  jonathan }
   1706   1.89  jonathan 
   1707   1.89  jonathan SYSCTL_SETUP(sysctl_net_bfp_setup, "sysctl net.bpf subtree setup")
   1708   1.89  jonathan {
   1709   1.89  jonathan 	struct sysctlnode *node;
   1710   1.89  jonathan 
   1711   1.90    atatat 	sysctl_createv(clog, 0, NULL, NULL,
   1712   1.90    atatat 		       CTLFLAG_PERMANENT,
   1713   1.89  jonathan 		       CTLTYPE_NODE, "net", NULL,
   1714   1.89  jonathan 		       NULL, 0, NULL, 0,
   1715   1.89  jonathan 		       CTL_NET, CTL_EOL);
   1716   1.89  jonathan 
   1717   1.89  jonathan 	node = NULL;
   1718   1.90    atatat 	sysctl_createv(clog, 0, NULL, &node,
   1719   1.90    atatat 		       CTLFLAG_PERMANENT,
   1720   1.98    atatat 		       CTLTYPE_NODE, "bpf",
   1721   1.98    atatat 		       SYSCTL_DESCR("BPF options"),
   1722   1.89  jonathan 		       NULL, 0, NULL, 0,
   1723   1.89  jonathan 		       CTL_NET, CTL_CREATE, CTL_EOL);
   1724   1.89  jonathan 	if (node != NULL)
   1725   1.90    atatat 		sysctl_createv(clog, 0, NULL, NULL,
   1726   1.90    atatat 			CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
   1727   1.98    atatat 			CTLTYPE_INT, "maxbufsize",
   1728   1.98    atatat 			SYSCTL_DESCR("Maximum size for data capture buffer"),
   1729   1.89  jonathan 			sysctl_net_bpf_maxbufsize, 0, &bpf_maxbufsize, 0,
   1730   1.89  jonathan 			CTL_NET, node->sysctl_num, CTL_CREATE, CTL_EOL);
   1731   1.89  jonathan }
   1732