2 * Copyright (c) 1990, 1991, 1993
3 * The Regents of the University of California. All rights reserved.
5 * This code is derived from the Stanford/CMU enet packet filter,
6 * (net/enet.c) distributed as part of 4.3BSD, and code contributed
7 * to Berkeley by Steven McCanne and Van Jacobson both of Lawrence
10 * Redistribution and use in source and binary forms, with or without
11 * modification, are permitted provided that the following conditions
13 * 1. Redistributions of source code must retain the above copyright
14 * notice, this list of conditions and the following disclaimer.
15 * 2. Redistributions in binary form must reproduce the above copyright
16 * notice, this list of conditions and the following disclaimer in the
17 * documentation and/or other materials provided with the distribution.
18 * 3. All advertising materials mentioning features or use of this software
19 * must display the following acknowledgement:
20 * This product includes software developed by the University of
21 * California, Berkeley and its contributors.
22 * 4. Neither the name of the University nor the names of its contributors
23 * may be used to endorse or promote products derived from this software
24 * without specific prior written permission.
26 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
27 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
28 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
29 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
30 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
31 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
32 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
33 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
34 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
35 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
38 * @(#)bpf.c 8.2 (Berkeley) 3/28/94
40 * $FreeBSD: src/sys/net/bpf.c,v 1.59.2.12 2002/04/14 21:41:48 luigi Exp $
41 * $DragonFly: src/sys/net/bpf.c,v 1.50 2008/09/23 11:28:49 sephe Exp $
46 #include <sys/param.h>
47 #include <sys/systm.h>
49 #include <sys/device.h>
50 #include <sys/malloc.h>
54 #include <sys/signalvar.h>
55 #include <sys/filio.h>
56 #include <sys/sockio.h>
57 #include <sys/ttycom.h>
58 #include <sys/filedesc.h>
61 #include <sys/event.h>
63 #include <sys/socket.h>
64 #include <sys/vnode.h>
66 #include <sys/thread2.h>
67 #include <sys/mplock2.h>
71 #include <net/bpfdesc.h>
72 #include <net/netmsg2.h>
74 #include <netinet/in.h>
75 #include <netinet/if_ether.h>
76 #include <sys/kernel.h>
77 #include <sys/sysctl.h>
79 #include <sys/devfs.h>
81 struct netmsg_bpf_output {
82 struct netmsg nm_netmsg;
85 struct sockaddr *nm_dst;
88 MALLOC_DEFINE(M_BPF, "BPF", "BPF data");
89 DEVFS_DECLARE_CLONE_BITMAP(bpf);
92 #define BPF_PREALLOCATED_UNITS 4
94 #define BPF_PREALLOCATED_UNITS NBPF
100 * The default read buffer size is patchable.
102 static int bpf_bufsize = BPF_DEFAULTBUFSIZE;
103 SYSCTL_INT(_debug, OID_AUTO, bpf_bufsize, CTLFLAG_RW,
104 &bpf_bufsize, 0, "");
105 int bpf_maxbufsize = BPF_MAXBUFSIZE;
106 SYSCTL_INT(_debug, OID_AUTO, bpf_maxbufsize, CTLFLAG_RW,
107 &bpf_maxbufsize, 0, "");
110 * bpf_iflist is the list of interfaces; each corresponds to an ifnet
112 static struct bpf_if *bpf_iflist;
114 static int bpf_allocbufs(struct bpf_d *);
115 static void bpf_attachd(struct bpf_d *d, struct bpf_if *bp);
116 static void bpf_detachd(struct bpf_d *d);
117 static void bpf_resetd(struct bpf_d *);
118 static void bpf_freed(struct bpf_d *);
119 static void bpf_mcopy(const void *, void *, size_t);
120 static int bpf_movein(struct uio *, int, struct mbuf **,
121 struct sockaddr *, int *, struct bpf_insn *);
122 static int bpf_setif(struct bpf_d *, struct ifreq *);
123 static void bpf_timed_out(void *);
124 static void bpf_wakeup(struct bpf_d *);
125 static void catchpacket(struct bpf_d *, u_char *, u_int, u_int,
126 void (*)(const void *, void *, size_t),
127 const struct timeval *);
128 static int bpf_setf(struct bpf_d *, struct bpf_program *, u_long cmd);
129 static int bpf_getdltlist(struct bpf_d *, struct bpf_dltlist *);
130 static int bpf_setdlt(struct bpf_d *, u_int);
131 static void bpf_drvinit(void *unused);
132 static void bpf_filter_detach(struct knote *kn);
133 static int bpf_filter_read(struct knote *kn, long hint);
135 static d_open_t bpfopen;
136 static d_clone_t bpfclone;
137 static d_close_t bpfclose;
138 static d_read_t bpfread;
139 static d_write_t bpfwrite;
140 static d_ioctl_t bpfioctl;
141 static d_poll_t bpfpoll;
142 static d_kqfilter_t bpfkqfilter;
144 #define CDEV_MAJOR 23
145 static struct dev_ops bpf_ops = {
146 { "bpf", CDEV_MAJOR, D_KQFILTER },
153 .d_kqfilter = bpfkqfilter
158 bpf_movein(struct uio *uio, int linktype, struct mbuf **mp,
159 struct sockaddr *sockp, int *datlen, struct bpf_insn *wfilter)
171 * Build a sockaddr based on the data link layer type.
172 * We do this at this level because the ethernet header
173 * is copied directly into the data field of the sockaddr.
174 * In the case of SLIP, there is no header and the packet
175 * is forwarded as is.
176 * Also, we are careful to leave room at the front of the mbuf
177 * for the link level header.
181 sockp->sa_family = AF_INET;
186 sockp->sa_family = AF_UNSPEC;
187 /* XXX Would MAXLINKHDR be better? */
188 hlen = sizeof(struct ether_header);
193 sockp->sa_family = AF_UNSPEC;
197 case DLT_ATM_RFC1483:
199 * en atm driver requires 4-byte atm pseudo header.
200 * though it isn't standard, vpi:vci needs to be
203 sockp->sa_family = AF_UNSPEC;
204 hlen = 12; /* XXX 4(ATM_PH) + 3(LLC) + 5(SNAP) */
208 sockp->sa_family = AF_UNSPEC;
209 hlen = 4; /* This should match PPP_HDRLEN */
216 len = uio->uio_resid;
217 *datlen = len - hlen;
218 if ((unsigned)len > MCLBYTES)
221 m = m_getl(len, MB_WAIT, MT_DATA, M_PKTHDR, NULL);
224 m->m_pkthdr.len = m->m_len = len;
225 m->m_pkthdr.rcvif = NULL;
228 if (m->m_len < hlen) {
233 error = uiomove(mtod(m, u_char *), len, uio);
237 slen = bpf_filter(wfilter, mtod(m, u_char *), len, len);
244 * Make room for link header, and copy it to sockaddr.
247 bcopy(m->m_data, sockp->sa_data, hlen);
248 m->m_pkthdr.len -= hlen;
250 m->m_data += hlen; /* XXX */
259 * Attach file to the bpf interface, i.e. make d listen on bp.
260 * Must be called at splimp.
263 bpf_attachd(struct bpf_d *d, struct bpf_if *bp)
266 * Point d at bp, and add d to the interface's list of listeners.
267 * Finally, point the driver's bpf cookie at the interface so
268 * it will divert packets to bpf.
271 SLIST_INSERT_HEAD(&bp->bif_dlist, d, bd_next);
272 *bp->bif_driverp = bp;
274 EVENTHANDLER_INVOKE(bpf_track, bp->bif_ifp, bp->bif_dlt, 1);
278 * Detach a file from its interface.
281 bpf_detachd(struct bpf_d *d)
290 /* Remove d from the interface's descriptor list. */
291 SLIST_REMOVE(&bp->bif_dlist, d, bpf_d, bd_next);
293 if (SLIST_EMPTY(&bp->bif_dlist)) {
295 * Let the driver know that there are no more listeners.
297 *bp->bif_driverp = NULL;
301 EVENTHANDLER_INVOKE(bpf_track, ifp, bp->bif_dlt, 0);
304 * Check if this descriptor had requested promiscuous mode.
305 * If so, turn it off.
309 error = ifpromisc(ifp, 0);
310 if (error != 0 && error != ENXIO) {
312 * ENXIO can happen if a pccard is unplugged,
313 * Something is really wrong if we were able to put
314 * the driver into promiscuous mode, but can't
317 if_printf(ifp, "bpf_detach: ifpromisc failed(%d)\n",
324 * Open ethernet device. Returns ENXIO for illegal minor device number,
325 * EBUSY if file is open by another process.
329 bpfopen(struct dev_open_args *ap)
331 cdev_t dev = ap->a_head.a_dev;
334 if (ap->a_cred->cr_prison)
339 * Each minor can be opened by only one process. If the requested
340 * minor is in use, return EBUSY.
345 MALLOC(d, struct bpf_d *, sizeof *d, M_BPF, M_WAITOK | M_ZERO);
347 d->bd_bufsize = bpf_bufsize;
350 callout_init(&d->bd_callout);
355 bpfclone(struct dev_clone_args *ap)
359 unit = devfs_clone_bitmap_get(&DEVFS_CLONE_BITMAP(bpf), 0);
360 ap->a_dev = make_only_dev(&bpf_ops, unit, 0, 0, 0600, "bpf%d", unit);
366 * Close the descriptor by detaching it from its interface,
367 * deallocating its buffers, and marking it free.
371 bpfclose(struct dev_close_args *ap)
373 cdev_t dev = ap->a_head.a_dev;
374 struct bpf_d *d = dev->si_drv1;
376 funsetown(d->bd_sigio);
378 if (d->bd_state == BPF_WAITING)
379 callout_stop(&d->bd_callout);
380 d->bd_state = BPF_IDLE;
381 if (d->bd_bif != NULL)
386 if (dev->si_uminor >= BPF_PREALLOCATED_UNITS) {
387 devfs_clone_bitmap_put(&DEVFS_CLONE_BITMAP(bpf), dev->si_uminor);
395 * Rotate the packet buffers in descriptor d. Move the store buffer
396 * into the hold slot, and the free buffer into the store slot.
397 * Zero the length of the new store buffer.
399 #define ROTATE_BUFFERS(d) \
400 (d)->bd_hbuf = (d)->bd_sbuf; \
401 (d)->bd_hlen = (d)->bd_slen; \
402 (d)->bd_sbuf = (d)->bd_fbuf; \
406 * bpfread - read next chunk of packets from buffers
409 bpfread(struct dev_read_args *ap)
411 cdev_t dev = ap->a_head.a_dev;
412 struct bpf_d *d = dev->si_drv1;
417 * Restrict application to use a buffer the same size as
420 if (ap->a_uio->uio_resid != d->bd_bufsize)
424 if (d->bd_state == BPF_WAITING)
425 callout_stop(&d->bd_callout);
426 timed_out = (d->bd_state == BPF_TIMED_OUT);
427 d->bd_state = BPF_IDLE;
429 * If the hold buffer is empty, then do a timed sleep, which
430 * ends when the timeout expires or when enough packets
431 * have arrived to fill the store buffer.
433 while (d->bd_hbuf == NULL) {
434 if ((d->bd_immediate || (ap->a_ioflag & IO_NDELAY) || timed_out)
435 && d->bd_slen != 0) {
437 * A packet(s) either arrived since the previous,
438 * We're in immediate mode, or are reading
439 * in non-blocking mode, and a packet(s)
440 * either arrived since the previous
441 * read or arrived while we were asleep.
442 * Rotate the buffers and return what's here.
449 * No data is available, check to see if the bpf device
450 * is still pointed at a real interface. If not, return
451 * ENXIO so that the userland process knows to rebind
452 * it before using it again.
454 if (d->bd_bif == NULL) {
459 if (ap->a_ioflag & IO_NDELAY) {
463 error = tsleep(d, PCATCH, "bpf", d->bd_rtout);
464 if (error == EINTR || error == ERESTART) {
468 if (error == EWOULDBLOCK) {
470 * On a timeout, return what's in the buffer,
471 * which may be nothing. If there is something
472 * in the store buffer, we can rotate the buffers.
476 * We filled up the buffer in between
477 * getting the timeout and arriving
478 * here, so we don't need to rotate.
482 if (d->bd_slen == 0) {
491 * At this point, we know we have something in the hold slot.
496 * Move data from hold buffer into user space.
497 * We know the entire buffer is transferred since
498 * we checked above that the read buffer is bpf_bufsize bytes.
500 error = uiomove(d->bd_hbuf, d->bd_hlen, ap->a_uio);
503 d->bd_fbuf = d->bd_hbuf;
513 * If there are processes sleeping on this descriptor, wake them up.
516 bpf_wakeup(struct bpf_d *d)
518 if (d->bd_state == BPF_WAITING) {
519 callout_stop(&d->bd_callout);
520 d->bd_state = BPF_IDLE;
523 if (d->bd_async && d->bd_sig && d->bd_sigio)
524 pgsigio(d->bd_sigio, d->bd_sig, 0);
527 selwakeup(&d->bd_sel);
530 d->bd_sel.si_pid = 0;
534 bpf_timed_out(void *arg)
536 struct bpf_d *d = (struct bpf_d *)arg;
539 if (d->bd_state == BPF_WAITING) {
540 d->bd_state = BPF_TIMED_OUT;
548 bpf_output_dispatch(struct netmsg *nmsg)
550 struct netmsg_bpf_output *bmsg = (struct netmsg_bpf_output *)nmsg;
551 struct ifnet *ifp = bmsg->nm_ifp;
555 * The driver frees the mbuf.
557 error = ifp->if_output(ifp, bmsg->nm_mbuf, bmsg->nm_dst, NULL);
558 lwkt_replymsg(&nmsg->nm_lmsg, error);
562 bpfwrite(struct dev_write_args *ap)
564 cdev_t dev = ap->a_head.a_dev;
565 struct bpf_d *d = dev->si_drv1;
571 struct netmsg_bpf_output bmsg;
573 if (d->bd_bif == NULL)
576 ifp = d->bd_bif->bif_ifp;
578 if (ap->a_uio->uio_resid == 0)
581 error = bpf_movein(ap->a_uio, (int)d->bd_bif->bif_dlt, &m,
582 &dst, &datlen, d->bd_wfilter);
586 if (datlen > ifp->if_mtu) {
592 dst.sa_family = pseudo_AF_HDRCMPLT;
594 netmsg_init(&bmsg.nm_netmsg, NULL, &curthread->td_msgport,
595 MSGF_MPSAFE, bpf_output_dispatch);
600 return lwkt_domsg(cpu_portfn(0), &bmsg.nm_netmsg.nm_lmsg, 0);
604 * Reset a descriptor by flushing its packet buffer and clearing the
605 * receive and drop counts. Should be called at splimp.
608 bpf_resetd(struct bpf_d *d)
611 /* Free the hold buffer. */
612 d->bd_fbuf = d->bd_hbuf;
622 * FIONREAD Check for read packet available.
623 * SIOCGIFADDR Get interface address - convenient hook to driver.
624 * BIOCGBLEN Get buffer len [for read()].
625 * BIOCSETF Set ethernet read filter.
626 * BIOCSETWF Set ethernet write filter.
627 * BIOCFLUSH Flush read packet buffer.
628 * BIOCPROMISC Put interface into promiscuous mode.
629 * BIOCGDLT Get link layer type.
630 * BIOCGETIF Get interface name.
631 * BIOCSETIF Set interface.
632 * BIOCSRTIMEOUT Set read timeout.
633 * BIOCGRTIMEOUT Get read timeout.
634 * BIOCGSTATS Get packet stats.
635 * BIOCIMMEDIATE Set immediate mode.
636 * BIOCVERSION Get filter language version.
637 * BIOCGHDRCMPLT Get "header already complete" flag
638 * BIOCSHDRCMPLT Set "header already complete" flag
639 * BIOCGSEESENT Get "see packets sent" flag
640 * BIOCSSEESENT Set "see packets sent" flag
641 * BIOCLOCK Set "locked" flag
645 bpfioctl(struct dev_ioctl_args *ap)
647 cdev_t dev = ap->a_head.a_dev;
648 struct bpf_d *d = dev->si_drv1;
652 if (d->bd_state == BPF_WAITING)
653 callout_stop(&d->bd_callout);
654 d->bd_state = BPF_IDLE;
657 if (d->bd_locked == 1) {
685 * Check for read packet available.
697 *(int *)ap->a_data = n;
705 if (d->bd_bif == NULL) {
708 ifp = d->bd_bif->bif_ifp;
709 ifnet_serialize_all(ifp);
710 error = ifp->if_ioctl(ifp, ap->a_cmd,
711 ap->a_data, ap->a_cred);
712 ifnet_deserialize_all(ifp);
718 * Get buffer len [for read()].
721 *(u_int *)ap->a_data = d->bd_bufsize;
728 if (d->bd_bif != NULL) {
731 u_int size = *(u_int *)ap->a_data;
733 if (size > bpf_maxbufsize)
734 *(u_int *)ap->a_data = size = bpf_maxbufsize;
735 else if (size < BPF_MINBUFSIZE)
736 *(u_int *)ap->a_data = size = BPF_MINBUFSIZE;
737 d->bd_bufsize = size;
742 * Set link layer read filter.
746 error = bpf_setf(d, (struct bpf_program *)ap->a_data,
751 * Flush read packet buffer.
760 * Put interface into promiscuous mode.
763 if (d->bd_bif == NULL) {
765 * No interface attached yet.
771 if (d->bd_promisc == 0) {
772 error = ifpromisc(d->bd_bif->bif_ifp, 1);
780 * Get device parameters.
783 if (d->bd_bif == NULL)
786 *(u_int *)ap->a_data = d->bd_bif->bif_dlt;
790 * Get a list of supported data link types.
793 if (d->bd_bif == NULL) {
796 error = bpf_getdltlist(d,
797 (struct bpf_dltlist *)ap->a_data);
802 * Set data link type.
805 if (d->bd_bif == NULL)
808 error = bpf_setdlt(d, *(u_int *)ap->a_data);
812 * Get interface name.
815 if (d->bd_bif == NULL) {
818 struct ifnet *const ifp = d->bd_bif->bif_ifp;
819 struct ifreq *const ifr = (struct ifreq *)ap->a_data;
821 strlcpy(ifr->ifr_name, ifp->if_xname,
822 sizeof ifr->ifr_name);
830 error = bpf_setif(d, (struct ifreq *)ap->a_data);
838 struct timeval *tv = (struct timeval *)ap->a_data;
841 * Subtract 1 tick from tvtohz() since this isn't
844 if ((error = itimerfix(tv)) == 0)
845 d->bd_rtout = tvtohz_low(tv);
854 struct timeval *tv = (struct timeval *)ap->a_data;
856 tv->tv_sec = d->bd_rtout / hz;
857 tv->tv_usec = (d->bd_rtout % hz) * ustick;
866 struct bpf_stat *bs = (struct bpf_stat *)ap->a_data;
868 bs->bs_recv = d->bd_rcount;
869 bs->bs_drop = d->bd_dcount;
874 * Set immediate mode.
877 d->bd_immediate = *(u_int *)ap->a_data;
882 struct bpf_version *bv = (struct bpf_version *)ap->a_data;
884 bv->bv_major = BPF_MAJOR_VERSION;
885 bv->bv_minor = BPF_MINOR_VERSION;
890 * Get "header already complete" flag
893 *(u_int *)ap->a_data = d->bd_hdrcmplt;
897 * Set "header already complete" flag
900 d->bd_hdrcmplt = *(u_int *)ap->a_data ? 1 : 0;
904 * Get "see sent packets" flag
907 *(u_int *)ap->a_data = d->bd_seesent;
911 * Set "see sent packets" flag
914 d->bd_seesent = *(u_int *)ap->a_data;
917 case FIOASYNC: /* Send signal on receive packets */
918 d->bd_async = *(int *)ap->a_data;
922 error = fsetown(*(int *)ap->a_data, &d->bd_sigio);
926 *(int *)ap->a_data = fgetown(d->bd_sigio);
929 /* This is deprecated, FIOSETOWN should be used instead. */
931 error = fsetown(-(*(int *)ap->a_data), &d->bd_sigio);
934 /* This is deprecated, FIOGETOWN should be used instead. */
936 *(int *)ap->a_data = -fgetown(d->bd_sigio);
939 case BIOCSRSIG: /* Set receive signal */
943 sig = *(u_int *)ap->a_data;
952 *(u_int *)ap->a_data = d->bd_sig;
962 * Set d's packet filter program to fp. If this file already has a filter,
963 * free it and replace it. Returns EINVAL for bogus requests.
966 bpf_setf(struct bpf_d *d, struct bpf_program *fp, u_long cmd)
968 struct bpf_insn *fcode, *old;
969 u_int wfilter, flen, size;
971 if (cmd == BIOCSETWF) {
978 if (fp->bf_insns == NULL) {
983 d->bd_wfilter = NULL;
985 d->bd_rfilter = NULL;
993 if (flen > BPF_MAXINSNS)
996 size = flen * sizeof *fp->bf_insns;
997 fcode = (struct bpf_insn *)kmalloc(size, M_BPF, M_WAITOK);
998 if (copyin(fp->bf_insns, fcode, size) == 0 &&
999 bpf_validate(fcode, (int)flen)) {
1002 d->bd_wfilter = fcode;
1004 d->bd_rfilter = fcode;
1012 kfree(fcode, M_BPF);
1017 * Detach a file from its current interface (if attached at all) and attach
1018 * to the interface indicated by the name stored in ifr.
1019 * Return an errno or 0.
1022 bpf_setif(struct bpf_d *d, struct ifreq *ifr)
1026 struct ifnet *theywant;
1028 theywant = ifunit(ifr->ifr_name);
1029 if (theywant == NULL)
1033 * Look through attached interfaces for the named one.
1035 for (bp = bpf_iflist; bp != NULL; bp = bp->bif_next) {
1036 struct ifnet *ifp = bp->bif_ifp;
1038 if (ifp == NULL || ifp != theywant)
1040 /* skip additional entry */
1041 if (bp->bif_driverp != &ifp->if_bpf)
1044 * We found the requested interface.
1045 * Allocate the packet buffers if we need to.
1046 * If we're already attached to requested interface,
1047 * just flush the buffer.
1049 if (d->bd_sbuf == NULL) {
1050 error = bpf_allocbufs(d);
1055 if (bp != d->bd_bif) {
1056 if (d->bd_bif != NULL) {
1058 * Detach if attached to something else.
1075 * Support for select() and poll() system calls
1077 * Return true iff the specific operation will not block indefinitely.
1078 * Otherwise, return false but make a note that a selwakeup() must be done.
1081 bpfpoll(struct dev_poll_args *ap)
1083 cdev_t dev = ap->a_head.a_dev;
1088 if (d->bd_bif == NULL)
1091 revents = ap->a_events & (POLLOUT | POLLWRNORM);
1093 if (ap->a_events & (POLLIN | POLLRDNORM)) {
1095 * An imitation of the FIONREAD ioctl code.
1096 * XXX not quite. An exact imitation:
1097 * if (d->b_slen != 0 ||
1098 * (d->bd_hbuf != NULL && d->bd_hlen != 0)
1100 if (d->bd_hlen != 0 ||
1101 ((d->bd_immediate || d->bd_state == BPF_TIMED_OUT) &&
1103 revents |= ap->a_events & (POLLIN | POLLRDNORM);
1105 selrecord(curthread, &d->bd_sel);
1106 /* Start the read timeout if necessary. */
1107 if (d->bd_rtout > 0 && d->bd_state == BPF_IDLE) {
1108 callout_reset(&d->bd_callout, d->bd_rtout,
1110 d->bd_state = BPF_WAITING;
1115 ap->a_events = revents;
1119 static struct filterops bpf_read_filtops =
1120 { 1, NULL, bpf_filter_detach, bpf_filter_read };
1123 bpfkqfilter(struct dev_kqfilter_args *ap)
1125 cdev_t dev = ap->a_head.a_dev;
1126 struct knote *kn = ap->a_kn;
1127 struct klist *klist;
1131 if (d->bd_bif == NULL) {
1137 switch (kn->kn_filter) {
1139 kn->kn_fop = &bpf_read_filtops;
1140 kn->kn_hook = (caddr_t)d;
1148 klist = &d->bd_sel.si_note;
1149 SLIST_INSERT_HEAD(klist, kn, kn_selnext);
1156 bpf_filter_detach(struct knote *kn)
1158 struct klist *klist;
1162 d = (struct bpf_d *)kn->kn_hook;
1163 klist = &d->bd_sel.si_note;
1164 SLIST_REMOVE(klist, kn, knote, kn_selnext);
1169 bpf_filter_read(struct knote *kn, long hint)
1175 d = (struct bpf_d *)kn->kn_hook;
1176 if (d->bd_hlen != 0 ||
1177 ((d->bd_immediate || d->bd_state == BPF_TIMED_OUT) &&
1181 /* Start the read timeout if necessary. */
1182 if (d->bd_rtout > 0 && d->bd_state == BPF_IDLE) {
1183 callout_reset(&d->bd_callout, d->bd_rtout,
1185 d->bd_state = BPF_WAITING;
1195 * Process the packet pkt of length pktlen. The packet is parsed
1196 * by each listener's filter, and if accepted, stashed into the
1197 * corresponding buffer.
1200 bpf_tap(struct bpf_if *bp, u_char *pkt, u_int pktlen)
1216 * Note that the ipl does not have to be raised at this point.
1217 * The only problem that could arise here is that if two different
1218 * interfaces shared any data. This is not the case.
1220 SLIST_FOREACH(d, &bp->bif_dlist, bd_next) {
1222 slen = bpf_filter(d->bd_rfilter, pkt, pktlen, pktlen);
1228 catchpacket(d, pkt, pktlen, slen, ovbcopy, &tv);
1236 * Copy data from an mbuf chain into a buffer. This code is derived
1237 * from m_copydata in sys/uipc_mbuf.c.
1240 bpf_mcopy(const void *src_arg, void *dst_arg, size_t len)
1242 const struct mbuf *m;
1251 count = min(m->m_len, len);
1252 bcopy(mtod(m, void *), dst, count);
1260 * Process the packet in the mbuf chain m. The packet is parsed by each
1261 * listener's filter, and if accepted, stashed into the corresponding
1265 bpf_mtap(struct bpf_if *bp, struct mbuf *m)
1280 /* Don't compute pktlen, if no descriptor is attached. */
1281 if (SLIST_EMPTY(&bp->bif_dlist)) {
1286 pktlen = m_lengthm(m, NULL);
1288 SLIST_FOREACH(d, &bp->bif_dlist, bd_next) {
1289 if (!d->bd_seesent && (m->m_pkthdr.rcvif == NULL))
1292 slen = bpf_filter(d->bd_rfilter, (u_char *)m, pktlen, 0);
1298 catchpacket(d, (u_char *)m, pktlen, slen, bpf_mcopy,
1307 bpf_mtap_family(struct bpf_if *bp, struct mbuf *m, sa_family_t family)
1311 KKASSERT(family != AF_UNSPEC);
1313 family4 = (u_int)family;
1314 bpf_ptap(bp, m, &family4, sizeof(family4));
1318 * Process the packet in the mbuf chain m with the header in m prepended.
1319 * The packet is parsed by each listener's filter, and if accepted,
1320 * stashed into the corresponding buffer.
1323 bpf_ptap(struct bpf_if *bp, struct mbuf *m, const void *data, u_int dlen)
1328 * Craft on-stack mbuf suitable for passing to bpf_mtap.
1329 * Note that we cut corners here; we only setup what's
1330 * absolutely needed--this mbuf should never go anywhere else.
1333 mb.m_data = __DECONST(void *, data); /* LINTED */
1335 mb.m_pkthdr.rcvif = m->m_pkthdr.rcvif;
1341 * Move the packet data from interface memory (pkt) into the
1342 * store buffer. Return 1 if it's time to wakeup a listener (buffer full),
1343 * otherwise 0. "copy" is the routine called to do the actual data
1344 * transfer. bcopy is passed in to copy contiguous chunks, while
1345 * bpf_mcopy is passed in to copy mbuf chains. In the latter case,
1346 * pkt is really an mbuf.
1349 catchpacket(struct bpf_d *d, u_char *pkt, u_int pktlen, u_int snaplen,
1350 void (*cpfn)(const void *, void *, size_t),
1351 const struct timeval *tv)
1355 int hdrlen = d->bd_bif->bif_hdrlen;
1357 * Figure out how many bytes to move. If the packet is
1358 * greater or equal to the snapshot length, transfer that
1359 * much. Otherwise, transfer the whole packet (unless
1360 * we hit the buffer size limit).
1362 totlen = hdrlen + min(snaplen, pktlen);
1363 if (totlen > d->bd_bufsize)
1364 totlen = d->bd_bufsize;
1367 * Round up the end of the previous packet to the next longword.
1369 curlen = BPF_WORDALIGN(d->bd_slen);
1370 if (curlen + totlen > d->bd_bufsize) {
1372 * This packet will overflow the storage buffer.
1373 * Rotate the buffers if we can, then wakeup any
1376 if (d->bd_fbuf == NULL) {
1378 * We haven't completed the previous read yet,
1379 * so drop the packet.
1387 } else if (d->bd_immediate || d->bd_state == BPF_TIMED_OUT) {
1389 * Immediate mode is set, or the read timeout has
1390 * already expired during a select call. A packet
1391 * arrived, so the reader should be woken up.
1397 * Append the bpf header.
1399 hp = (struct bpf_hdr *)(d->bd_sbuf + curlen);
1400 hp->bh_tstamp = *tv;
1401 hp->bh_datalen = pktlen;
1402 hp->bh_hdrlen = hdrlen;
1404 * Copy the packet data into the store buffer and update its length.
1406 (*cpfn)(pkt, (u_char *)hp + hdrlen, (hp->bh_caplen = totlen - hdrlen));
1407 d->bd_slen = curlen + totlen;
1411 * Initialize all nonzero fields of a descriptor.
1414 bpf_allocbufs(struct bpf_d *d)
1416 d->bd_fbuf = kmalloc(d->bd_bufsize, M_BPF, M_WAITOK);
1417 d->bd_sbuf = kmalloc(d->bd_bufsize, M_BPF, M_WAITOK);
1424 * Free buffers and packet filter program currently in use by a descriptor.
1428 bpf_freed(struct bpf_d *d)
1431 * We don't need to lock out interrupts since this descriptor has
1432 * been detached from its interface and it yet hasn't been marked
1435 if (d->bd_sbuf != NULL) {
1436 kfree(d->bd_sbuf, M_BPF);
1437 if (d->bd_hbuf != NULL)
1438 kfree(d->bd_hbuf, M_BPF);
1439 if (d->bd_fbuf != NULL)
1440 kfree(d->bd_fbuf, M_BPF);
1443 kfree(d->bd_rfilter, M_BPF);
1445 kfree(d->bd_wfilter, M_BPF);
1449 * Attach an interface to bpf. ifp is a pointer to the structure
1450 * defining the interface to be attached, dlt is the link layer type,
1451 * and hdrlen is the fixed size of the link header (variable length
1452 * headers are not yet supported).
1455 bpfattach(struct ifnet *ifp, u_int dlt, u_int hdrlen)
1457 bpfattach_dlt(ifp, dlt, hdrlen, &ifp->if_bpf);
1461 bpfattach_dlt(struct ifnet *ifp, u_int dlt, u_int hdrlen, struct bpf_if **driverp)
1465 bp = kmalloc(sizeof *bp, M_BPF, M_WAITOK | M_ZERO);
1467 SLIST_INIT(&bp->bif_dlist);
1470 bp->bif_driverp = driverp;
1471 *bp->bif_driverp = NULL;
1473 bp->bif_next = bpf_iflist;
1477 * Compute the length of the bpf header. This is not necessarily
1478 * equal to SIZEOF_BPF_HDR because we want to insert spacing such
1479 * that the network layer header begins on a longword boundary (for
1480 * performance reasons and to alleviate alignment restrictions).
1482 bp->bif_hdrlen = BPF_WORDALIGN(hdrlen + SIZEOF_BPF_HDR) - hdrlen;
1485 if_printf(ifp, "bpf attached\n");
1489 * Detach bpf from an interface. This involves detaching each descriptor
1490 * associated with the interface, and leaving bd_bif NULL. Notify each
1491 * descriptor as it's detached so that any sleepers wake up and get
1495 bpfdetach(struct ifnet *ifp)
1497 struct bpf_if *bp, *bp_prev;
1502 /* Locate BPF interface information */
1504 for (bp = bpf_iflist; bp != NULL; bp = bp->bif_next) {
1505 if (ifp == bp->bif_ifp)
1510 /* Interface wasn't attached */
1511 if (bp->bif_ifp == NULL) {
1513 kprintf("bpfdetach: %s was not attached\n", ifp->if_xname);
1517 while ((d = SLIST_FIRST(&bp->bif_dlist)) != NULL) {
1522 if (bp_prev != NULL)
1523 bp_prev->bif_next = bp->bif_next;
1525 bpf_iflist = bp->bif_next;
1533 * Get a list of available data link type of the interface.
1536 bpf_getdltlist(struct bpf_d *d, struct bpf_dltlist *bfl)
1542 ifp = d->bd_bif->bif_ifp;
1545 for (bp = bpf_iflist; bp != NULL; bp = bp->bif_next) {
1546 if (bp->bif_ifp != ifp)
1548 if (bfl->bfl_list != NULL) {
1549 if (n >= bfl->bfl_len) {
1552 error = copyout(&bp->bif_dlt,
1553 bfl->bfl_list + n, sizeof(u_int));
1562 * Set the data link type of a BPF instance.
1565 bpf_setdlt(struct bpf_d *d, u_int dlt)
1567 int error, opromisc;
1571 if (d->bd_bif->bif_dlt == dlt)
1573 ifp = d->bd_bif->bif_ifp;
1574 for (bp = bpf_iflist; bp != NULL; bp = bp->bif_next) {
1575 if (bp->bif_ifp == ifp && bp->bif_dlt == dlt)
1579 opromisc = d->bd_promisc;
1585 error = ifpromisc(bp->bif_ifp, 1);
1587 if_printf(bp->bif_ifp,
1588 "bpf_setdlt: ifpromisc failed (%d)\n",
1596 return(bp == NULL ? EINVAL : 0);
1600 bpf_drvinit(void *unused)
1604 make_autoclone_dev(&bpf_ops, &DEVFS_CLONE_BITMAP(bpf),
1605 bpfclone, 0, 0, 0600, "bpf");
1606 for (i = 0; i < BPF_PREALLOCATED_UNITS; i++) {
1607 make_dev(&bpf_ops, i, 0, 0, 0600, "bpf%d", i);
1608 devfs_clone_bitmap_set(&DEVFS_CLONE_BITMAP(bpf), i);
1613 bpf_drvuninit(void *unused)
1615 devfs_clone_handler_del("bpf");
1616 dev_ops_remove_all(&bpf_ops);
1617 devfs_clone_bitmap_uninit(&DEVFS_CLONE_BITMAP(bpf));
1620 SYSINIT(bpfdev,SI_SUB_DRIVERS,SI_ORDER_MIDDLE+CDEV_MAJOR,bpf_drvinit,NULL)
1621 SYSUNINIT(bpfdev, SI_SUB_DRIVERS,SI_ORDER_MIDDLE+CDEV_MAJOR,bpf_drvuninit, NULL);
1625 * NOP stubs to allow bpf-using drivers to load and function.
1627 * A 'better' implementation would allow the core bpf functionality
1628 * to be loaded at runtime.
1632 bpf_tap(struct bpf_if *bp, u_char *pkt, u_int pktlen)
1637 bpf_mtap(struct bpf_if *bp, struct mbuf *m)
1642 bpf_ptap(struct bpf_if *bp, struct mbuf *m, const void *data, u_int dlen)
1647 bpfattach(struct ifnet *ifp, u_int dlt, u_int hdrlen)
1652 bpfattach_dlt(struct ifnet *ifp, u_int dlt, u_int hdrlen, struct bpf_if **driverp)
1657 bpfdetach(struct ifnet *ifp)
1662 bpf_filter(const struct bpf_insn *pc, u_char *p, u_int wirelen, u_int buflen)
1664 return -1; /* "no filter" behaviour */