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 $
45 #include <sys/param.h>
46 #include <sys/systm.h>
48 #include <sys/device.h>
49 #include <sys/malloc.h>
53 #include <sys/signalvar.h>
54 #include <sys/filio.h>
55 #include <sys/sockio.h>
56 #include <sys/ttycom.h>
57 #include <sys/filedesc.h>
59 #include <sys/event.h>
61 #include <sys/socket.h>
62 #include <sys/vnode.h>
64 #include <sys/thread2.h>
68 #include <net/bpfdesc.h>
69 #include <net/netmsg2.h>
71 #include <netinet/in.h>
72 #include <netinet/if_ether.h>
73 #include <sys/kernel.h>
74 #include <sys/sysctl.h>
76 #include <sys/devfs.h>
78 struct netmsg_bpf_output {
79 struct netmsg_base base;
82 struct sockaddr *nm_dst;
85 MALLOC_DEFINE(M_BPF, "BPF", "BPF data");
86 DEVFS_DECLARE_CLONE_BITMAP(bpf);
89 #define BPF_PREALLOCATED_UNITS 4
91 #define BPF_PREALLOCATED_UNITS NBPF
97 * The default read buffer size is patchable.
99 static int bpf_bufsize = BPF_DEFAULTBUFSIZE;
100 SYSCTL_INT(_debug, OID_AUTO, bpf_bufsize, CTLFLAG_RW,
101 &bpf_bufsize, 0, "Current size of bpf buffer");
102 int bpf_maxbufsize = BPF_MAXBUFSIZE;
103 SYSCTL_INT(_debug, OID_AUTO, bpf_maxbufsize, CTLFLAG_RW,
104 &bpf_maxbufsize, 0, "Maximum size of bpf buffer");
107 * bpf_iflist is the list of interfaces; each corresponds to an ifnet
109 static struct bpf_if *bpf_iflist;
111 static struct lwkt_token bpf_token = LWKT_TOKEN_INITIALIZER(bpf_token);
113 static int bpf_allocbufs(struct bpf_d *);
114 static void bpf_attachd(struct bpf_d *d, struct bpf_if *bp);
115 static void bpf_detachd(struct bpf_d *d);
116 static void bpf_resetd(struct bpf_d *);
117 static void bpf_freed(struct bpf_d *);
118 static void bpf_mcopy(const void *, void *, size_t);
119 static int bpf_movein(struct uio *, int, struct mbuf **,
120 struct sockaddr *, int *, struct bpf_insn *);
121 static int bpf_setif(struct bpf_d *, struct ifreq *);
122 static void bpf_timed_out(void *);
123 static void bpf_wakeup(struct bpf_d *);
124 static void catchpacket(struct bpf_d *, u_char *, u_int, u_int,
125 void (*)(const void *, void *, size_t),
126 const struct timeval *);
127 static int bpf_setf(struct bpf_d *, struct bpf_program *, u_long cmd);
128 static int bpf_getdltlist(struct bpf_d *, struct bpf_dltlist *);
129 static int bpf_setdlt(struct bpf_d *, u_int);
130 static void bpf_drvinit(void *unused);
131 static void bpf_filter_detach(struct knote *kn);
132 static int bpf_filter_read(struct knote *kn, long hint);
134 static d_open_t bpfopen;
135 static d_clone_t bpfclone;
136 static d_close_t bpfclose;
137 static d_read_t bpfread;
138 static d_write_t bpfwrite;
139 static d_ioctl_t bpfioctl;
140 static d_kqfilter_t bpfkqfilter;
142 #define CDEV_MAJOR 23
143 static struct dev_ops bpf_ops = {
144 { "bpf", 0, D_MPSAFE },
150 .d_kqfilter = bpfkqfilter
155 bpf_movein(struct uio *uio, int linktype, struct mbuf **mp,
156 struct sockaddr *sockp, int *datlen, struct bpf_insn *wfilter)
168 * Build a sockaddr based on the data link layer type.
169 * We do this at this level because the ethernet header
170 * is copied directly into the data field of the sockaddr.
171 * In the case of SLIP, there is no header and the packet
172 * is forwarded as is.
173 * Also, we are careful to leave room at the front of the mbuf
174 * for the link level header.
178 sockp->sa_family = AF_INET;
183 sockp->sa_family = AF_UNSPEC;
184 /* XXX Would MAXLINKHDR be better? */
185 hlen = sizeof(struct ether_header);
190 sockp->sa_family = AF_UNSPEC;
194 case DLT_ATM_RFC1483:
196 * en atm driver requires 4-byte atm pseudo header.
197 * though it isn't standard, vpi:vci needs to be
200 sockp->sa_family = AF_UNSPEC;
201 hlen = 12; /* XXX 4(ATM_PH) + 3(LLC) + 5(SNAP) */
205 sockp->sa_family = AF_UNSPEC;
206 hlen = 4; /* This should match PPP_HDRLEN */
213 len = uio->uio_resid;
214 *datlen = len - hlen;
215 if ((unsigned)len > MCLBYTES)
218 m = m_getl(len, MB_WAIT, MT_DATA, M_PKTHDR, NULL);
221 m->m_pkthdr.len = m->m_len = len;
222 m->m_pkthdr.rcvif = NULL;
225 if (m->m_len < hlen) {
230 error = uiomove(mtod(m, u_char *), len, uio);
234 slen = bpf_filter(wfilter, mtod(m, u_char *), len, len);
241 * Make room for link header, and copy it to sockaddr.
244 bcopy(m->m_data, sockp->sa_data, hlen);
245 m->m_pkthdr.len -= hlen;
247 m->m_data += hlen; /* XXX */
256 * Attach file to the bpf interface, i.e. make d listen on bp.
257 * Must be called at splimp.
260 bpf_attachd(struct bpf_d *d, struct bpf_if *bp)
263 * Point d at bp, and add d to the interface's list of listeners.
264 * Finally, point the driver's bpf cookie at the interface so
265 * it will divert packets to bpf.
267 lwkt_gettoken(&bpf_token);
269 SLIST_INSERT_HEAD(&bp->bif_dlist, d, bd_next);
270 *bp->bif_driverp = bp;
272 EVENTHANDLER_INVOKE(bpf_track, bp->bif_ifp, bp->bif_dlt, 1);
273 lwkt_reltoken(&bpf_token);
277 * Detach a file from its interface.
280 bpf_detachd(struct bpf_d *d)
286 lwkt_gettoken(&bpf_token);
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",
321 lwkt_reltoken(&bpf_token);
325 * Open ethernet device. Returns ENXIO for illegal minor device number,
326 * EBUSY if file is open by another process.
330 bpfopen(struct dev_open_args *ap)
332 cdev_t dev = ap->a_head.a_dev;
335 lwkt_gettoken(&bpf_token);
336 if (ap->a_cred->cr_prison) {
337 lwkt_reltoken(&bpf_token);
343 * Each minor can be opened by only one process. If the requested
344 * minor is in use, return EBUSY.
347 lwkt_reltoken(&bpf_token);
351 d = kmalloc(sizeof *d, M_BPF, M_WAITOK | M_ZERO);
353 d->bd_bufsize = bpf_bufsize;
356 callout_init(&d->bd_callout);
357 lwkt_reltoken(&bpf_token);
363 bpfclone(struct dev_clone_args *ap)
367 unit = devfs_clone_bitmap_get(&DEVFS_CLONE_BITMAP(bpf), 0);
368 ap->a_dev = make_only_dev(&bpf_ops, unit, 0, 0, 0600, "bpf%d", unit);
374 * Close the descriptor by detaching it from its interface,
375 * deallocating its buffers, and marking it free.
379 bpfclose(struct dev_close_args *ap)
381 cdev_t dev = ap->a_head.a_dev;
382 struct bpf_d *d = dev->si_drv1;
384 lwkt_gettoken(&bpf_token);
385 funsetown(&d->bd_sigio);
387 if (d->bd_state == BPF_WAITING)
388 callout_stop(&d->bd_callout);
389 d->bd_state = BPF_IDLE;
390 if (d->bd_bif != NULL)
395 if (dev->si_uminor >= BPF_PREALLOCATED_UNITS) {
396 devfs_clone_bitmap_put(&DEVFS_CLONE_BITMAP(bpf), dev->si_uminor);
400 lwkt_reltoken(&bpf_token);
406 * Rotate the packet buffers in descriptor d. Move the store buffer
407 * into the hold slot, and the free buffer into the store slot.
408 * Zero the length of the new store buffer.
410 #define ROTATE_BUFFERS(d) \
411 (d)->bd_hbuf = (d)->bd_sbuf; \
412 (d)->bd_hlen = (d)->bd_slen; \
413 (d)->bd_sbuf = (d)->bd_fbuf; \
417 * bpfread - read next chunk of packets from buffers
420 bpfread(struct dev_read_args *ap)
422 cdev_t dev = ap->a_head.a_dev;
423 struct bpf_d *d = dev->si_drv1;
427 lwkt_gettoken(&bpf_token);
429 * Restrict application to use a buffer the same size as
432 if (ap->a_uio->uio_resid != d->bd_bufsize) {
433 lwkt_reltoken(&bpf_token);
438 if (d->bd_state == BPF_WAITING)
439 callout_stop(&d->bd_callout);
440 timed_out = (d->bd_state == BPF_TIMED_OUT);
441 d->bd_state = BPF_IDLE;
443 * If the hold buffer is empty, then do a timed sleep, which
444 * ends when the timeout expires or when enough packets
445 * have arrived to fill the store buffer.
447 while (d->bd_hbuf == NULL) {
448 if ((d->bd_immediate || (ap->a_ioflag & IO_NDELAY) || timed_out)
449 && d->bd_slen != 0) {
451 * A packet(s) either arrived since the previous,
452 * We're in immediate mode, or are reading
453 * in non-blocking mode, and a packet(s)
454 * either arrived since the previous
455 * read or arrived while we were asleep.
456 * Rotate the buffers and return what's here.
463 * No data is available, check to see if the bpf device
464 * is still pointed at a real interface. If not, return
465 * ENXIO so that the userland process knows to rebind
466 * it before using it again.
468 if (d->bd_bif == NULL) {
470 lwkt_reltoken(&bpf_token);
474 if (ap->a_ioflag & IO_NDELAY) {
476 lwkt_reltoken(&bpf_token);
479 error = tsleep(d, PCATCH, "bpf", d->bd_rtout);
480 if (error == EINTR || error == ERESTART) {
482 lwkt_reltoken(&bpf_token);
485 if (error == EWOULDBLOCK) {
487 * On a timeout, return what's in the buffer,
488 * which may be nothing. If there is something
489 * in the store buffer, we can rotate the buffers.
493 * We filled up the buffer in between
494 * getting the timeout and arriving
495 * here, so we don't need to rotate.
499 if (d->bd_slen == 0) {
501 lwkt_reltoken(&bpf_token);
509 * At this point, we know we have something in the hold slot.
514 * Move data from hold buffer into user space.
515 * We know the entire buffer is transferred since
516 * we checked above that the read buffer is bpf_bufsize bytes.
518 error = uiomove(d->bd_hbuf, d->bd_hlen, ap->a_uio);
521 d->bd_fbuf = d->bd_hbuf;
525 lwkt_reltoken(&bpf_token);
532 * If there are processes sleeping on this descriptor, wake them up.
535 bpf_wakeup(struct bpf_d *d)
537 if (d->bd_state == BPF_WAITING) {
538 callout_stop(&d->bd_callout);
539 d->bd_state = BPF_IDLE;
542 if (d->bd_async && d->bd_sig && d->bd_sigio)
543 pgsigio(d->bd_sigio, d->bd_sig, 0);
545 KNOTE(&d->bd_kq.ki_note, 0);
549 bpf_timed_out(void *arg)
551 struct bpf_d *d = (struct bpf_d *)arg;
554 if (d->bd_state == BPF_WAITING) {
555 d->bd_state = BPF_TIMED_OUT;
563 bpf_output_dispatch(netmsg_t msg)
565 struct netmsg_bpf_output *bmsg = (struct netmsg_bpf_output *)msg;
566 struct ifnet *ifp = bmsg->nm_ifp;
570 * The driver frees the mbuf.
572 error = ifp->if_output(ifp, bmsg->nm_mbuf, bmsg->nm_dst, NULL);
573 lwkt_replymsg(&msg->lmsg, error);
577 bpfwrite(struct dev_write_args *ap)
579 cdev_t dev = ap->a_head.a_dev;
580 struct bpf_d *d = dev->si_drv1;
586 struct netmsg_bpf_output bmsg;
588 lwkt_gettoken(&bpf_token);
589 if (d->bd_bif == NULL) {
590 lwkt_reltoken(&bpf_token);
594 ifp = d->bd_bif->bif_ifp;
596 if (ap->a_uio->uio_resid == 0) {
597 lwkt_reltoken(&bpf_token);
601 error = bpf_movein(ap->a_uio, (int)d->bd_bif->bif_dlt, &m,
602 &dst, &datlen, d->bd_wfilter);
604 lwkt_reltoken(&bpf_token);
608 if (datlen > ifp->if_mtu) {
610 lwkt_reltoken(&bpf_token);
615 dst.sa_family = pseudo_AF_HDRCMPLT;
617 netmsg_init(&bmsg.base, NULL, &curthread->td_msgport,
618 0, bpf_output_dispatch);
623 ret = lwkt_domsg(netisr_portfn(0), &bmsg.base.lmsg, 0);
624 lwkt_reltoken(&bpf_token);
630 * Reset a descriptor by flushing its packet buffer and clearing the
631 * receive and drop counts. Should be called at splimp.
634 bpf_resetd(struct bpf_d *d)
637 /* Free the hold buffer. */
638 d->bd_fbuf = d->bd_hbuf;
648 * FIONREAD Check for read packet available.
649 * SIOCGIFADDR Get interface address - convenient hook to driver.
650 * BIOCGBLEN Get buffer len [for read()].
651 * BIOCSETF Set ethernet read filter.
652 * BIOCSETWF Set ethernet write filter.
653 * BIOCFLUSH Flush read packet buffer.
654 * BIOCPROMISC Put interface into promiscuous mode.
655 * BIOCGDLT Get link layer type.
656 * BIOCGETIF Get interface name.
657 * BIOCSETIF Set interface.
658 * BIOCSRTIMEOUT Set read timeout.
659 * BIOCGRTIMEOUT Get read timeout.
660 * BIOCGSTATS Get packet stats.
661 * BIOCIMMEDIATE Set immediate mode.
662 * BIOCVERSION Get filter language version.
663 * BIOCGHDRCMPLT Get "header already complete" flag
664 * BIOCSHDRCMPLT Set "header already complete" flag
665 * BIOCGSEESENT Get "see packets sent" flag
666 * BIOCSSEESENT Set "see packets sent" flag
667 * BIOCLOCK Set "locked" flag
671 bpfioctl(struct dev_ioctl_args *ap)
673 cdev_t dev = ap->a_head.a_dev;
674 struct bpf_d *d = dev->si_drv1;
677 lwkt_gettoken(&bpf_token);
679 if (d->bd_state == BPF_WAITING)
680 callout_stop(&d->bd_callout);
681 d->bd_state = BPF_IDLE;
684 if (d->bd_locked == 1) {
703 lwkt_reltoken(&bpf_token);
713 * Check for read packet available.
725 *(int *)ap->a_data = n;
733 if (d->bd_bif == NULL) {
736 ifp = d->bd_bif->bif_ifp;
737 ifnet_serialize_all(ifp);
738 error = ifp->if_ioctl(ifp, ap->a_cmd,
739 ap->a_data, ap->a_cred);
740 ifnet_deserialize_all(ifp);
746 * Get buffer len [for read()].
749 *(u_int *)ap->a_data = d->bd_bufsize;
756 if (d->bd_bif != NULL) {
759 u_int size = *(u_int *)ap->a_data;
761 if (size > bpf_maxbufsize)
762 *(u_int *)ap->a_data = size = bpf_maxbufsize;
763 else if (size < BPF_MINBUFSIZE)
764 *(u_int *)ap->a_data = size = BPF_MINBUFSIZE;
765 d->bd_bufsize = size;
770 * Set link layer read filter.
774 error = bpf_setf(d, (struct bpf_program *)ap->a_data,
779 * Flush read packet buffer.
788 * Put interface into promiscuous mode.
791 if (d->bd_bif == NULL) {
793 * No interface attached yet.
799 if (d->bd_promisc == 0) {
800 error = ifpromisc(d->bd_bif->bif_ifp, 1);
808 * Get device parameters.
811 if (d->bd_bif == NULL)
814 *(u_int *)ap->a_data = d->bd_bif->bif_dlt;
818 * Get a list of supported data link types.
821 if (d->bd_bif == NULL) {
824 error = bpf_getdltlist(d,
825 (struct bpf_dltlist *)ap->a_data);
830 * Set data link type.
833 if (d->bd_bif == NULL)
836 error = bpf_setdlt(d, *(u_int *)ap->a_data);
840 * Get interface name.
843 if (d->bd_bif == NULL) {
846 struct ifnet *const ifp = d->bd_bif->bif_ifp;
847 struct ifreq *const ifr = (struct ifreq *)ap->a_data;
849 strlcpy(ifr->ifr_name, ifp->if_xname,
850 sizeof ifr->ifr_name);
858 error = bpf_setif(d, (struct ifreq *)ap->a_data);
866 struct timeval *tv = (struct timeval *)ap->a_data;
869 * Subtract 1 tick from tvtohz() since this isn't
872 if ((error = itimerfix(tv)) == 0)
873 d->bd_rtout = tvtohz_low(tv);
882 struct timeval *tv = (struct timeval *)ap->a_data;
884 tv->tv_sec = d->bd_rtout / hz;
885 tv->tv_usec = (d->bd_rtout % hz) * ustick;
894 struct bpf_stat *bs = (struct bpf_stat *)ap->a_data;
896 bs->bs_recv = d->bd_rcount;
897 bs->bs_drop = d->bd_dcount;
902 * Set immediate mode.
905 d->bd_immediate = *(u_int *)ap->a_data;
910 struct bpf_version *bv = (struct bpf_version *)ap->a_data;
912 bv->bv_major = BPF_MAJOR_VERSION;
913 bv->bv_minor = BPF_MINOR_VERSION;
918 * Get "header already complete" flag
921 *(u_int *)ap->a_data = d->bd_hdrcmplt;
925 * Set "header already complete" flag
928 d->bd_hdrcmplt = *(u_int *)ap->a_data ? 1 : 0;
932 * Get "see sent packets" flag
935 *(u_int *)ap->a_data = d->bd_seesent;
939 * Set "see sent packets" flag
942 d->bd_seesent = *(u_int *)ap->a_data;
945 case FIOASYNC: /* Send signal on receive packets */
946 d->bd_async = *(int *)ap->a_data;
950 error = fsetown(*(int *)ap->a_data, &d->bd_sigio);
954 *(int *)ap->a_data = fgetown(&d->bd_sigio);
957 /* This is deprecated, FIOSETOWN should be used instead. */
959 error = fsetown(-(*(int *)ap->a_data), &d->bd_sigio);
962 /* This is deprecated, FIOGETOWN should be used instead. */
964 *(int *)ap->a_data = -fgetown(&d->bd_sigio);
967 case BIOCSRSIG: /* Set receive signal */
971 sig = *(u_int *)ap->a_data;
980 *(u_int *)ap->a_data = d->bd_sig;
986 lwkt_reltoken(&bpf_token);
992 * Set d's packet filter program to fp. If this file already has a filter,
993 * free it and replace it. Returns EINVAL for bogus requests.
996 bpf_setf(struct bpf_d *d, struct bpf_program *fp, u_long cmd)
998 struct bpf_insn *fcode, *old;
999 u_int wfilter, flen, size;
1001 if (cmd == BIOCSETWF) {
1002 old = d->bd_wfilter;
1006 old = d->bd_rfilter;
1008 if (fp->bf_insns == NULL) {
1009 if (fp->bf_len != 0)
1013 d->bd_wfilter = NULL;
1015 d->bd_rfilter = NULL;
1023 if (flen > BPF_MAXINSNS)
1026 size = flen * sizeof *fp->bf_insns;
1027 fcode = (struct bpf_insn *)kmalloc(size, M_BPF, M_WAITOK);
1028 if (copyin(fp->bf_insns, fcode, size) == 0 &&
1029 bpf_validate(fcode, (int)flen)) {
1032 d->bd_wfilter = fcode;
1034 d->bd_rfilter = fcode;
1042 kfree(fcode, M_BPF);
1047 * Detach a file from its current interface (if attached at all) and attach
1048 * to the interface indicated by the name stored in ifr.
1049 * Return an errno or 0.
1052 bpf_setif(struct bpf_d *d, struct ifreq *ifr)
1056 struct ifnet *theywant;
1058 theywant = ifunit(ifr->ifr_name);
1059 if (theywant == NULL)
1063 * Look through attached interfaces for the named one.
1065 for (bp = bpf_iflist; bp != NULL; bp = bp->bif_next) {
1066 struct ifnet *ifp = bp->bif_ifp;
1068 if (ifp == NULL || ifp != theywant)
1070 /* skip additional entry */
1071 if (bp->bif_driverp != &ifp->if_bpf)
1074 * We found the requested interface.
1075 * Allocate the packet buffers if we need to.
1076 * If we're already attached to requested interface,
1077 * just flush the buffer.
1079 if (d->bd_sbuf == NULL) {
1080 error = bpf_allocbufs(d);
1085 if (bp != d->bd_bif) {
1086 if (d->bd_bif != NULL) {
1088 * Detach if attached to something else.
1104 static struct filterops bpf_read_filtops =
1105 { FILTEROP_ISFD, NULL, bpf_filter_detach, bpf_filter_read };
1108 bpfkqfilter(struct dev_kqfilter_args *ap)
1110 cdev_t dev = ap->a_head.a_dev;
1111 struct knote *kn = ap->a_kn;
1112 struct klist *klist;
1115 lwkt_gettoken(&bpf_token);
1117 if (d->bd_bif == NULL) {
1119 lwkt_reltoken(&bpf_token);
1124 switch (kn->kn_filter) {
1126 kn->kn_fop = &bpf_read_filtops;
1127 kn->kn_hook = (caddr_t)d;
1130 ap->a_result = EOPNOTSUPP;
1131 lwkt_reltoken(&bpf_token);
1135 klist = &d->bd_kq.ki_note;
1136 knote_insert(klist, kn);
1137 lwkt_reltoken(&bpf_token);
1143 bpf_filter_detach(struct knote *kn)
1145 struct klist *klist;
1148 d = (struct bpf_d *)kn->kn_hook;
1149 klist = &d->bd_kq.ki_note;
1150 knote_remove(klist, kn);
1154 bpf_filter_read(struct knote *kn, long hint)
1160 d = (struct bpf_d *)kn->kn_hook;
1161 if (d->bd_hlen != 0 ||
1162 ((d->bd_immediate || d->bd_state == BPF_TIMED_OUT) &&
1166 /* Start the read timeout if necessary. */
1167 if (d->bd_rtout > 0 && d->bd_state == BPF_IDLE) {
1168 callout_reset(&d->bd_callout, d->bd_rtout,
1170 d->bd_state = BPF_WAITING;
1180 * Process the packet pkt of length pktlen. The packet is parsed
1181 * by each listener's filter, and if accepted, stashed into the
1182 * corresponding buffer.
1185 bpf_tap(struct bpf_if *bp, u_char *pkt, u_int pktlen)
1192 lwkt_gettoken(&bpf_token);
1195 lwkt_reltoken(&bpf_token);
1200 * Note that the ipl does not have to be raised at this point.
1201 * The only problem that could arise here is that if two different
1202 * interfaces shared any data. This is not the case.
1204 SLIST_FOREACH(d, &bp->bif_dlist, bd_next) {
1206 slen = bpf_filter(d->bd_rfilter, pkt, pktlen, pktlen);
1212 catchpacket(d, pkt, pktlen, slen, ovbcopy, &tv);
1215 lwkt_reltoken(&bpf_token);
1219 * Copy data from an mbuf chain into a buffer. This code is derived
1220 * from m_copydata in sys/uipc_mbuf.c.
1223 bpf_mcopy(const void *src_arg, void *dst_arg, size_t len)
1225 const struct mbuf *m;
1234 count = min(m->m_len, len);
1235 bcopy(mtod(m, void *), dst, count);
1243 * Process the packet in the mbuf chain m. The packet is parsed by each
1244 * listener's filter, and if accepted, stashed into the corresponding
1248 bpf_mtap(struct bpf_if *bp, struct mbuf *m)
1255 lwkt_gettoken(&bpf_token);
1258 lwkt_reltoken(&bpf_token);
1262 /* Don't compute pktlen, if no descriptor is attached. */
1263 if (SLIST_EMPTY(&bp->bif_dlist)) {
1264 lwkt_reltoken(&bpf_token);
1268 pktlen = m_lengthm(m, NULL);
1270 SLIST_FOREACH(d, &bp->bif_dlist, bd_next) {
1271 if (!d->bd_seesent && (m->m_pkthdr.rcvif == NULL))
1274 slen = bpf_filter(d->bd_rfilter, (u_char *)m, pktlen, 0);
1280 catchpacket(d, (u_char *)m, pktlen, slen, bpf_mcopy,
1284 lwkt_reltoken(&bpf_token);
1288 * Incoming linkage from device drivers, where we have a mbuf chain
1289 * but need to prepend some arbitrary header from a linear buffer.
1291 * Con up a minimal dummy header to pacify bpf. Allocate (only) a
1292 * struct m_hdr on the stack. This is safe as bpf only reads from the
1293 * fields in this header that we initialize, and will not try to free
1294 * it or keep a pointer to it.
1297 bpf_mtap_hdr(struct bpf_if *arg, caddr_t data, u_int dlen, struct mbuf *m, u_int direction)
1306 return bpf_mtap(arg, (struct mbuf *) &mh);
1310 bpf_mtap_family(struct bpf_if *bp, struct mbuf *m, sa_family_t family)
1314 KKASSERT(family != AF_UNSPEC);
1316 family4 = (u_int)family;
1317 bpf_ptap(bp, m, &family4, sizeof(family4));
1321 * Process the packet in the mbuf chain m with the header in m prepended.
1322 * The packet is parsed by each listener's filter, and if accepted,
1323 * stashed into the corresponding buffer.
1326 bpf_ptap(struct bpf_if *bp, struct mbuf *m, const void *data, u_int dlen)
1331 * Craft on-stack mbuf suitable for passing to bpf_mtap.
1332 * Note that we cut corners here; we only setup what's
1333 * absolutely needed--this mbuf should never go anywhere else.
1336 mb.m_data = __DECONST(void *, data); /* LINTED */
1338 mb.m_pkthdr.rcvif = m->m_pkthdr.rcvif;
1344 * Move the packet data from interface memory (pkt) into the
1345 * store buffer. Return 1 if it's time to wakeup a listener (buffer full),
1346 * otherwise 0. "copy" is the routine called to do the actual data
1347 * transfer. bcopy is passed in to copy contiguous chunks, while
1348 * bpf_mcopy is passed in to copy mbuf chains. In the latter case,
1349 * pkt is really an mbuf.
1352 catchpacket(struct bpf_d *d, u_char *pkt, u_int pktlen, u_int snaplen,
1353 void (*cpfn)(const void *, void *, size_t),
1354 const struct timeval *tv)
1358 int hdrlen = d->bd_bif->bif_hdrlen;
1361 * Figure out how many bytes to move. If the packet is
1362 * greater or equal to the snapshot length, transfer that
1363 * much. Otherwise, transfer the whole packet (unless
1364 * we hit the buffer size limit).
1366 totlen = hdrlen + min(snaplen, pktlen);
1367 if (totlen > d->bd_bufsize)
1368 totlen = d->bd_bufsize;
1371 * Round up the end of the previous packet to the next longword.
1373 curlen = BPF_WORDALIGN(d->bd_slen);
1374 if (curlen + totlen > d->bd_bufsize) {
1376 * This packet will overflow the storage buffer.
1377 * Rotate the buffers if we can, then wakeup any
1380 if (d->bd_fbuf == NULL) {
1382 * We haven't completed the previous read yet,
1383 * so drop the packet.
1391 } else if (d->bd_immediate || d->bd_state == BPF_TIMED_OUT) {
1393 * Immediate mode is set, or the read timeout has
1394 * already expired during a select call. A packet
1395 * arrived, so the reader should be woken up.
1401 * Append the bpf header.
1403 hp = (struct bpf_hdr *)(d->bd_sbuf + curlen);
1404 hp->bh_tstamp = *tv;
1405 hp->bh_datalen = pktlen;
1406 hp->bh_hdrlen = hdrlen;
1408 * Copy the packet data into the store buffer and update its length.
1410 (*cpfn)(pkt, (u_char *)hp + hdrlen, (hp->bh_caplen = totlen - hdrlen));
1411 d->bd_slen = curlen + totlen;
1418 * Initialize all nonzero fields of a descriptor.
1421 bpf_allocbufs(struct bpf_d *d)
1423 d->bd_fbuf = kmalloc(d->bd_bufsize, M_BPF, M_WAITOK);
1424 d->bd_sbuf = kmalloc(d->bd_bufsize, M_BPF, M_WAITOK);
1431 * Free buffers and packet filter program currently in use by a descriptor.
1435 bpf_freed(struct bpf_d *d)
1438 * We don't need to lock out interrupts since this descriptor has
1439 * been detached from its interface and it yet hasn't been marked
1442 if (d->bd_sbuf != NULL) {
1443 kfree(d->bd_sbuf, M_BPF);
1444 if (d->bd_hbuf != NULL)
1445 kfree(d->bd_hbuf, M_BPF);
1446 if (d->bd_fbuf != NULL)
1447 kfree(d->bd_fbuf, M_BPF);
1450 kfree(d->bd_rfilter, M_BPF);
1452 kfree(d->bd_wfilter, M_BPF);
1456 * Attach an interface to bpf. ifp is a pointer to the structure
1457 * defining the interface to be attached, dlt is the link layer type,
1458 * and hdrlen is the fixed size of the link header (variable length
1459 * headers are not yet supported).
1462 bpfattach(struct ifnet *ifp, u_int dlt, u_int hdrlen)
1464 bpfattach_dlt(ifp, dlt, hdrlen, &ifp->if_bpf);
1468 bpfattach_dlt(struct ifnet *ifp, u_int dlt, u_int hdrlen, struct bpf_if **driverp)
1472 bp = kmalloc(sizeof *bp, M_BPF, M_WAITOK | M_ZERO);
1474 lwkt_gettoken(&bpf_token);
1476 SLIST_INIT(&bp->bif_dlist);
1479 bp->bif_driverp = driverp;
1480 *bp->bif_driverp = NULL;
1482 bp->bif_next = bpf_iflist;
1486 * Compute the length of the bpf header. This is not necessarily
1487 * equal to SIZEOF_BPF_HDR because we want to insert spacing such
1488 * that the network layer header begins on a longword boundary (for
1489 * performance reasons and to alleviate alignment restrictions).
1491 bp->bif_hdrlen = BPF_WORDALIGN(hdrlen + SIZEOF_BPF_HDR) - hdrlen;
1493 lwkt_reltoken(&bpf_token);
1496 if_printf(ifp, "bpf attached\n");
1500 * Detach bpf from an interface. This involves detaching each descriptor
1501 * associated with the interface, and leaving bd_bif NULL. Notify each
1502 * descriptor as it's detached so that any sleepers wake up and get
1506 bpfdetach(struct ifnet *ifp)
1508 struct bpf_if *bp, *bp_prev;
1511 lwkt_gettoken(&bpf_token);
1514 /* Locate BPF interface information */
1516 for (bp = bpf_iflist; bp != NULL; bp = bp->bif_next) {
1517 if (ifp == bp->bif_ifp)
1522 /* Interface wasn't attached */
1523 if (bp->bif_ifp == NULL) {
1525 lwkt_reltoken(&bpf_token);
1526 kprintf("bpfdetach: %s was not attached\n", ifp->if_xname);
1530 while ((d = SLIST_FIRST(&bp->bif_dlist)) != NULL) {
1535 if (bp_prev != NULL)
1536 bp_prev->bif_next = bp->bif_next;
1538 bpf_iflist = bp->bif_next;
1543 lwkt_reltoken(&bpf_token);
1547 * Get a list of available data link type of the interface.
1550 bpf_getdltlist(struct bpf_d *d, struct bpf_dltlist *bfl)
1556 ifp = d->bd_bif->bif_ifp;
1559 for (bp = bpf_iflist; bp != NULL; bp = bp->bif_next) {
1560 if (bp->bif_ifp != ifp)
1562 if (bfl->bfl_list != NULL) {
1563 if (n >= bfl->bfl_len) {
1566 error = copyout(&bp->bif_dlt,
1567 bfl->bfl_list + n, sizeof(u_int));
1576 * Set the data link type of a BPF instance.
1579 bpf_setdlt(struct bpf_d *d, u_int dlt)
1581 int error, opromisc;
1585 if (d->bd_bif->bif_dlt == dlt)
1587 ifp = d->bd_bif->bif_ifp;
1588 for (bp = bpf_iflist; bp != NULL; bp = bp->bif_next) {
1589 if (bp->bif_ifp == ifp && bp->bif_dlt == dlt)
1593 opromisc = d->bd_promisc;
1599 error = ifpromisc(bp->bif_ifp, 1);
1601 if_printf(bp->bif_ifp,
1602 "bpf_setdlt: ifpromisc failed (%d)\n",
1610 return(bp == NULL ? EINVAL : 0);
1616 lwkt_gettoken(&bpf_token);
1622 lwkt_reltoken(&bpf_token);
1626 bpf_drvinit(void *unused)
1630 make_autoclone_dev(&bpf_ops, &DEVFS_CLONE_BITMAP(bpf),
1631 bpfclone, 0, 0, 0600, "bpf");
1632 for (i = 0; i < BPF_PREALLOCATED_UNITS; i++) {
1633 make_dev(&bpf_ops, i, 0, 0, 0600, "bpf%d", i);
1634 devfs_clone_bitmap_set(&DEVFS_CLONE_BITMAP(bpf), i);
1639 bpf_drvuninit(void *unused)
1641 devfs_clone_handler_del("bpf");
1642 dev_ops_remove_all(&bpf_ops);
1643 devfs_clone_bitmap_uninit(&DEVFS_CLONE_BITMAP(bpf));
1646 SYSINIT(bpfdev,SI_SUB_DRIVERS,SI_ORDER_MIDDLE+CDEV_MAJOR,bpf_drvinit,NULL)
1647 SYSUNINIT(bpfdev, SI_SUB_DRIVERS,SI_ORDER_MIDDLE+CDEV_MAJOR,bpf_drvuninit, NULL);
1651 * NOP stubs to allow bpf-using drivers to load and function.
1653 * A 'better' implementation would allow the core bpf functionality
1654 * to be loaded at runtime.
1658 bpf_tap(struct bpf_if *bp, u_char *pkt, u_int pktlen)
1663 bpf_mtap(struct bpf_if *bp, struct mbuf *m)
1668 bpf_ptap(struct bpf_if *bp, struct mbuf *m, const void *data, u_int dlen)
1673 bpfattach(struct ifnet *ifp, u_int dlt, u_int hdrlen)
1678 bpfattach_dlt(struct ifnet *ifp, u_int dlt, u_int hdrlen, struct bpf_if **driverp)
1683 bpfdetach(struct ifnet *ifp)
1688 bpf_filter(const struct bpf_insn *pc, u_char *p, u_int wirelen, u_int buflen)
1690 return -1; /* "no filter" behaviour */