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);
386 if (d->bd_state == BPF_WAITING)
387 callout_stop(&d->bd_callout);
388 d->bd_state = BPF_IDLE;
389 if (d->bd_bif != NULL)
393 if (dev->si_uminor >= BPF_PREALLOCATED_UNITS) {
394 devfs_clone_bitmap_put(&DEVFS_CLONE_BITMAP(bpf), dev->si_uminor);
398 lwkt_reltoken(&bpf_token);
404 * Rotate the packet buffers in descriptor d. Move the store buffer
405 * into the hold slot, and the free buffer into the store slot.
406 * Zero the length of the new store buffer.
408 #define ROTATE_BUFFERS(d) \
409 (d)->bd_hbuf = (d)->bd_sbuf; \
410 (d)->bd_hlen = (d)->bd_slen; \
411 (d)->bd_sbuf = (d)->bd_fbuf; \
415 * bpfread - read next chunk of packets from buffers
418 bpfread(struct dev_read_args *ap)
420 cdev_t dev = ap->a_head.a_dev;
421 struct bpf_d *d = dev->si_drv1;
425 lwkt_gettoken(&bpf_token);
427 * Restrict application to use a buffer the same size as
430 if (ap->a_uio->uio_resid != d->bd_bufsize) {
431 lwkt_reltoken(&bpf_token);
435 if (d->bd_state == BPF_WAITING)
436 callout_stop(&d->bd_callout);
437 timed_out = (d->bd_state == BPF_TIMED_OUT);
438 d->bd_state = BPF_IDLE;
440 * If the hold buffer is empty, then do a timed sleep, which
441 * ends when the timeout expires or when enough packets
442 * have arrived to fill the store buffer.
444 while (d->bd_hbuf == NULL) {
445 if ((d->bd_immediate || (ap->a_ioflag & IO_NDELAY) || timed_out)
446 && d->bd_slen != 0) {
448 * A packet(s) either arrived since the previous,
449 * We're in immediate mode, or are reading
450 * in non-blocking mode, and a packet(s)
451 * either arrived since the previous
452 * read or arrived while we were asleep.
453 * Rotate the buffers and return what's here.
460 * No data is available, check to see if the bpf device
461 * is still pointed at a real interface. If not, return
462 * ENXIO so that the userland process knows to rebind
463 * it before using it again.
465 if (d->bd_bif == NULL) {
466 lwkt_reltoken(&bpf_token);
470 if (ap->a_ioflag & IO_NDELAY) {
471 lwkt_reltoken(&bpf_token);
474 error = tsleep(d, PCATCH, "bpf", d->bd_rtout);
475 if (error == EINTR || error == ERESTART) {
476 lwkt_reltoken(&bpf_token);
479 if (error == EWOULDBLOCK) {
481 * On a timeout, return what's in the buffer,
482 * which may be nothing. If there is something
483 * in the store buffer, we can rotate the buffers.
487 * We filled up the buffer in between
488 * getting the timeout and arriving
489 * here, so we don't need to rotate.
493 if (d->bd_slen == 0) {
494 lwkt_reltoken(&bpf_token);
502 * At this point, we know we have something in the hold slot.
506 * Move data from hold buffer into user space.
507 * We know the entire buffer is transferred since
508 * we checked above that the read buffer is bpf_bufsize bytes.
510 error = uiomove(d->bd_hbuf, d->bd_hlen, ap->a_uio);
512 d->bd_fbuf = d->bd_hbuf;
515 lwkt_reltoken(&bpf_token);
522 * If there are processes sleeping on this descriptor, wake them up.
525 bpf_wakeup(struct bpf_d *d)
527 if (d->bd_state == BPF_WAITING) {
528 callout_stop(&d->bd_callout);
529 d->bd_state = BPF_IDLE;
532 if (d->bd_async && d->bd_sig && d->bd_sigio)
533 pgsigio(d->bd_sigio, d->bd_sig, 0);
535 KNOTE(&d->bd_kq.ki_note, 0);
539 bpf_timed_out(void *arg)
541 struct bpf_d *d = (struct bpf_d *)arg;
543 if (d->bd_state == BPF_WAITING) {
544 d->bd_state = BPF_TIMED_OUT;
551 bpf_output_dispatch(netmsg_t msg)
553 struct netmsg_bpf_output *bmsg = (struct netmsg_bpf_output *)msg;
554 struct ifnet *ifp = bmsg->nm_ifp;
558 * The driver frees the mbuf.
560 error = ifp->if_output(ifp, bmsg->nm_mbuf, bmsg->nm_dst, NULL);
561 lwkt_replymsg(&msg->lmsg, error);
565 bpfwrite(struct dev_write_args *ap)
567 cdev_t dev = ap->a_head.a_dev;
568 struct bpf_d *d = dev->si_drv1;
574 struct netmsg_bpf_output bmsg;
576 lwkt_gettoken(&bpf_token);
577 if (d->bd_bif == NULL) {
578 lwkt_reltoken(&bpf_token);
582 ifp = d->bd_bif->bif_ifp;
584 if (ap->a_uio->uio_resid == 0) {
585 lwkt_reltoken(&bpf_token);
589 error = bpf_movein(ap->a_uio, (int)d->bd_bif->bif_dlt, &m,
590 &dst, &datlen, d->bd_wfilter);
592 lwkt_reltoken(&bpf_token);
596 if (datlen > ifp->if_mtu) {
598 lwkt_reltoken(&bpf_token);
603 dst.sa_family = pseudo_AF_HDRCMPLT;
605 netmsg_init(&bmsg.base, NULL, &curthread->td_msgport,
606 0, bpf_output_dispatch);
611 ret = lwkt_domsg(netisr_portfn(0), &bmsg.base.lmsg, 0);
612 lwkt_reltoken(&bpf_token);
618 * Reset a descriptor by flushing its packet buffer and clearing the
619 * receive and drop counts. Should be called at splimp.
622 bpf_resetd(struct bpf_d *d)
625 /* Free the hold buffer. */
626 d->bd_fbuf = d->bd_hbuf;
636 * FIONREAD Check for read packet available.
637 * SIOCGIFADDR Get interface address - convenient hook to driver.
638 * BIOCGBLEN Get buffer len [for read()].
639 * BIOCSETF Set ethernet read filter.
640 * BIOCSETWF Set ethernet write filter.
641 * BIOCFLUSH Flush read packet buffer.
642 * BIOCPROMISC Put interface into promiscuous mode.
643 * BIOCGDLT Get link layer type.
644 * BIOCGETIF Get interface name.
645 * BIOCSETIF Set interface.
646 * BIOCSRTIMEOUT Set read timeout.
647 * BIOCGRTIMEOUT Get read timeout.
648 * BIOCGSTATS Get packet stats.
649 * BIOCIMMEDIATE Set immediate mode.
650 * BIOCVERSION Get filter language version.
651 * BIOCGHDRCMPLT Get "header already complete" flag
652 * BIOCSHDRCMPLT Set "header already complete" flag
653 * BIOCGSEESENT Get "see packets sent" flag
654 * BIOCSSEESENT Set "see packets sent" flag
655 * BIOCLOCK Set "locked" flag
659 bpfioctl(struct dev_ioctl_args *ap)
661 cdev_t dev = ap->a_head.a_dev;
662 struct bpf_d *d = dev->si_drv1;
665 lwkt_gettoken(&bpf_token);
666 if (d->bd_state == BPF_WAITING)
667 callout_stop(&d->bd_callout);
668 d->bd_state = BPF_IDLE;
670 if (d->bd_locked == 1) {
689 lwkt_reltoken(&bpf_token);
699 * Check for read packet available.
709 *(int *)ap->a_data = n;
717 if (d->bd_bif == NULL) {
720 ifp = d->bd_bif->bif_ifp;
721 ifnet_serialize_all(ifp);
722 error = ifp->if_ioctl(ifp, ap->a_cmd,
723 ap->a_data, ap->a_cred);
724 ifnet_deserialize_all(ifp);
730 * Get buffer len [for read()].
733 *(u_int *)ap->a_data = d->bd_bufsize;
740 if (d->bd_bif != NULL) {
743 u_int size = *(u_int *)ap->a_data;
745 if (size > bpf_maxbufsize)
746 *(u_int *)ap->a_data = size = bpf_maxbufsize;
747 else if (size < BPF_MINBUFSIZE)
748 *(u_int *)ap->a_data = size = BPF_MINBUFSIZE;
749 d->bd_bufsize = size;
754 * Set link layer read filter.
758 error = bpf_setf(d, (struct bpf_program *)ap->a_data,
763 * Flush read packet buffer.
770 * Put interface into promiscuous mode.
773 if (d->bd_bif == NULL) {
775 * No interface attached yet.
780 if (d->bd_promisc == 0) {
781 error = ifpromisc(d->bd_bif->bif_ifp, 1);
788 * Get device parameters.
791 if (d->bd_bif == NULL)
794 *(u_int *)ap->a_data = d->bd_bif->bif_dlt;
798 * Get a list of supported data link types.
801 if (d->bd_bif == NULL) {
804 error = bpf_getdltlist(d,
805 (struct bpf_dltlist *)ap->a_data);
810 * Set data link type.
813 if (d->bd_bif == NULL)
816 error = bpf_setdlt(d, *(u_int *)ap->a_data);
820 * Get interface name.
823 if (d->bd_bif == NULL) {
826 struct ifnet *const ifp = d->bd_bif->bif_ifp;
827 struct ifreq *const ifr = (struct ifreq *)ap->a_data;
829 strlcpy(ifr->ifr_name, ifp->if_xname,
830 sizeof ifr->ifr_name);
838 error = bpf_setif(d, (struct ifreq *)ap->a_data);
846 struct timeval *tv = (struct timeval *)ap->a_data;
849 * Subtract 1 tick from tvtohz() since this isn't
852 if ((error = itimerfix(tv)) == 0)
853 d->bd_rtout = tvtohz_low(tv);
862 struct timeval *tv = (struct timeval *)ap->a_data;
864 tv->tv_sec = d->bd_rtout / hz;
865 tv->tv_usec = (d->bd_rtout % hz) * ustick;
874 struct bpf_stat *bs = (struct bpf_stat *)ap->a_data;
876 bs->bs_recv = d->bd_rcount;
877 bs->bs_drop = d->bd_dcount;
882 * Set immediate mode.
885 d->bd_immediate = *(u_int *)ap->a_data;
890 struct bpf_version *bv = (struct bpf_version *)ap->a_data;
892 bv->bv_major = BPF_MAJOR_VERSION;
893 bv->bv_minor = BPF_MINOR_VERSION;
898 * Get "header already complete" flag
901 *(u_int *)ap->a_data = d->bd_hdrcmplt;
905 * Set "header already complete" flag
908 d->bd_hdrcmplt = *(u_int *)ap->a_data ? 1 : 0;
912 * Get "see sent packets" flag
915 *(u_int *)ap->a_data = d->bd_seesent;
919 * Set "see sent packets" flag
922 d->bd_seesent = *(u_int *)ap->a_data;
925 case FIOASYNC: /* Send signal on receive packets */
926 d->bd_async = *(int *)ap->a_data;
930 error = fsetown(*(int *)ap->a_data, &d->bd_sigio);
934 *(int *)ap->a_data = fgetown(&d->bd_sigio);
937 /* This is deprecated, FIOSETOWN should be used instead. */
939 error = fsetown(-(*(int *)ap->a_data), &d->bd_sigio);
942 /* This is deprecated, FIOGETOWN should be used instead. */
944 *(int *)ap->a_data = -fgetown(&d->bd_sigio);
947 case BIOCSRSIG: /* Set receive signal */
951 sig = *(u_int *)ap->a_data;
960 *(u_int *)ap->a_data = d->bd_sig;
966 lwkt_reltoken(&bpf_token);
972 * Set d's packet filter program to fp. If this file already has a filter,
973 * free it and replace it. Returns EINVAL for bogus requests.
976 bpf_setf(struct bpf_d *d, struct bpf_program *fp, u_long cmd)
978 struct bpf_insn *fcode, *old;
979 u_int wfilter, flen, size;
981 if (cmd == BIOCSETWF) {
988 if (fp->bf_insns == NULL) {
992 d->bd_wfilter = NULL;
994 d->bd_rfilter = NULL;
1001 if (flen > BPF_MAXINSNS)
1004 size = flen * sizeof *fp->bf_insns;
1005 fcode = (struct bpf_insn *)kmalloc(size, M_BPF, M_WAITOK);
1006 if (copyin(fp->bf_insns, fcode, size) == 0 &&
1007 bpf_validate(fcode, (int)flen)) {
1009 d->bd_wfilter = fcode;
1011 d->bd_rfilter = fcode;
1018 kfree(fcode, M_BPF);
1023 * Detach a file from its current interface (if attached at all) and attach
1024 * to the interface indicated by the name stored in ifr.
1025 * Return an errno or 0.
1028 bpf_setif(struct bpf_d *d, struct ifreq *ifr)
1032 struct ifnet *theywant;
1034 theywant = ifunit(ifr->ifr_name);
1035 if (theywant == NULL)
1039 * Look through attached interfaces for the named one.
1041 for (bp = bpf_iflist; bp != NULL; bp = bp->bif_next) {
1042 struct ifnet *ifp = bp->bif_ifp;
1044 if (ifp == NULL || ifp != theywant)
1046 /* skip additional entry */
1047 if (bp->bif_driverp != &ifp->if_bpf)
1050 * We found the requested interface.
1051 * Allocate the packet buffers if we need to.
1052 * If we're already attached to requested interface,
1053 * just flush the buffer.
1055 if (d->bd_sbuf == NULL) {
1056 error = bpf_allocbufs(d);
1060 if (bp != d->bd_bif) {
1061 if (d->bd_bif != NULL) {
1063 * Detach if attached to something else.
1078 static struct filterops bpf_read_filtops =
1079 { FILTEROP_ISFD, NULL, bpf_filter_detach, bpf_filter_read };
1082 bpfkqfilter(struct dev_kqfilter_args *ap)
1084 cdev_t dev = ap->a_head.a_dev;
1085 struct knote *kn = ap->a_kn;
1086 struct klist *klist;
1089 lwkt_gettoken(&bpf_token);
1091 if (d->bd_bif == NULL) {
1093 lwkt_reltoken(&bpf_token);
1098 switch (kn->kn_filter) {
1100 kn->kn_fop = &bpf_read_filtops;
1101 kn->kn_hook = (caddr_t)d;
1104 ap->a_result = EOPNOTSUPP;
1105 lwkt_reltoken(&bpf_token);
1109 klist = &d->bd_kq.ki_note;
1110 knote_insert(klist, kn);
1111 lwkt_reltoken(&bpf_token);
1117 bpf_filter_detach(struct knote *kn)
1119 struct klist *klist;
1122 d = (struct bpf_d *)kn->kn_hook;
1123 klist = &d->bd_kq.ki_note;
1124 knote_remove(klist, kn);
1128 bpf_filter_read(struct knote *kn, long hint)
1133 d = (struct bpf_d *)kn->kn_hook;
1134 if (d->bd_hlen != 0 ||
1135 ((d->bd_immediate || d->bd_state == BPF_TIMED_OUT) &&
1139 /* Start the read timeout if necessary. */
1140 if (d->bd_rtout > 0 && d->bd_state == BPF_IDLE) {
1141 callout_reset(&d->bd_callout, d->bd_rtout,
1143 d->bd_state = BPF_WAITING;
1152 * Process the packet pkt of length pktlen. The packet is parsed
1153 * by each listener's filter, and if accepted, stashed into the
1154 * corresponding buffer.
1157 bpf_tap(struct bpf_if *bp, u_char *pkt, u_int pktlen)
1164 lwkt_gettoken(&bpf_token);
1167 lwkt_reltoken(&bpf_token);
1172 * Note that the ipl does not have to be raised at this point.
1173 * The only problem that could arise here is that if two different
1174 * interfaces shared any data. This is not the case.
1176 SLIST_FOREACH(d, &bp->bif_dlist, bd_next) {
1178 slen = bpf_filter(d->bd_rfilter, pkt, pktlen, pktlen);
1184 catchpacket(d, pkt, pktlen, slen, ovbcopy, &tv);
1187 lwkt_reltoken(&bpf_token);
1191 * Copy data from an mbuf chain into a buffer. This code is derived
1192 * from m_copydata in sys/uipc_mbuf.c.
1195 bpf_mcopy(const void *src_arg, void *dst_arg, size_t len)
1197 const struct mbuf *m;
1206 count = min(m->m_len, len);
1207 bcopy(mtod(m, void *), dst, count);
1215 * Process the packet in the mbuf chain m. The packet is parsed by each
1216 * listener's filter, and if accepted, stashed into the corresponding
1220 bpf_mtap(struct bpf_if *bp, struct mbuf *m)
1227 lwkt_gettoken(&bpf_token);
1230 lwkt_reltoken(&bpf_token);
1234 /* Don't compute pktlen, if no descriptor is attached. */
1235 if (SLIST_EMPTY(&bp->bif_dlist)) {
1236 lwkt_reltoken(&bpf_token);
1240 pktlen = m_lengthm(m, NULL);
1242 SLIST_FOREACH(d, &bp->bif_dlist, bd_next) {
1243 if (!d->bd_seesent && (m->m_pkthdr.rcvif == NULL))
1246 slen = bpf_filter(d->bd_rfilter, (u_char *)m, pktlen, 0);
1252 catchpacket(d, (u_char *)m, pktlen, slen, bpf_mcopy,
1256 lwkt_reltoken(&bpf_token);
1260 * Incoming linkage from device drivers, where we have a mbuf chain
1261 * but need to prepend some arbitrary header from a linear buffer.
1263 * Con up a minimal dummy header to pacify bpf. Allocate (only) a
1264 * struct m_hdr on the stack. This is safe as bpf only reads from the
1265 * fields in this header that we initialize, and will not try to free
1266 * it or keep a pointer to it.
1269 bpf_mtap_hdr(struct bpf_if *arg, caddr_t data, u_int dlen, struct mbuf *m, u_int direction)
1278 return bpf_mtap(arg, (struct mbuf *) &mh);
1282 bpf_mtap_family(struct bpf_if *bp, struct mbuf *m, sa_family_t family)
1286 KKASSERT(family != AF_UNSPEC);
1288 family4 = (u_int)family;
1289 bpf_ptap(bp, m, &family4, sizeof(family4));
1293 * Process the packet in the mbuf chain m with the header in m prepended.
1294 * The packet is parsed by each listener's filter, and if accepted,
1295 * stashed into the corresponding buffer.
1298 bpf_ptap(struct bpf_if *bp, struct mbuf *m, const void *data, u_int dlen)
1303 * Craft on-stack mbuf suitable for passing to bpf_mtap.
1304 * Note that we cut corners here; we only setup what's
1305 * absolutely needed--this mbuf should never go anywhere else.
1308 mb.m_data = __DECONST(void *, data); /* LINTED */
1310 mb.m_pkthdr.rcvif = m->m_pkthdr.rcvif;
1316 * Move the packet data from interface memory (pkt) into the
1317 * store buffer. Return 1 if it's time to wakeup a listener (buffer full),
1318 * otherwise 0. "copy" is the routine called to do the actual data
1319 * transfer. bcopy is passed in to copy contiguous chunks, while
1320 * bpf_mcopy is passed in to copy mbuf chains. In the latter case,
1321 * pkt is really an mbuf.
1324 catchpacket(struct bpf_d *d, u_char *pkt, u_int pktlen, u_int snaplen,
1325 void (*cpfn)(const void *, void *, size_t),
1326 const struct timeval *tv)
1330 int hdrlen = d->bd_bif->bif_hdrlen;
1333 * Figure out how many bytes to move. If the packet is
1334 * greater or equal to the snapshot length, transfer that
1335 * much. Otherwise, transfer the whole packet (unless
1336 * we hit the buffer size limit).
1338 totlen = hdrlen + min(snaplen, pktlen);
1339 if (totlen > d->bd_bufsize)
1340 totlen = d->bd_bufsize;
1343 * Round up the end of the previous packet to the next longword.
1345 curlen = BPF_WORDALIGN(d->bd_slen);
1346 if (curlen + totlen > d->bd_bufsize) {
1348 * This packet will overflow the storage buffer.
1349 * Rotate the buffers if we can, then wakeup any
1352 if (d->bd_fbuf == NULL) {
1354 * We haven't completed the previous read yet,
1355 * so drop the packet.
1363 } else if (d->bd_immediate || d->bd_state == BPF_TIMED_OUT) {
1365 * Immediate mode is set, or the read timeout has
1366 * already expired during a select call. A packet
1367 * arrived, so the reader should be woken up.
1373 * Append the bpf header.
1375 hp = (struct bpf_hdr *)(d->bd_sbuf + curlen);
1376 hp->bh_tstamp = *tv;
1377 hp->bh_datalen = pktlen;
1378 hp->bh_hdrlen = hdrlen;
1380 * Copy the packet data into the store buffer and update its length.
1382 (*cpfn)(pkt, (u_char *)hp + hdrlen, (hp->bh_caplen = totlen - hdrlen));
1383 d->bd_slen = curlen + totlen;
1390 * Initialize all nonzero fields of a descriptor.
1393 bpf_allocbufs(struct bpf_d *d)
1395 d->bd_fbuf = kmalloc(d->bd_bufsize, M_BPF, M_WAITOK);
1396 d->bd_sbuf = kmalloc(d->bd_bufsize, M_BPF, M_WAITOK);
1403 * Free buffers and packet filter program currently in use by a descriptor.
1407 bpf_freed(struct bpf_d *d)
1410 * We don't need to lock out interrupts since this descriptor has
1411 * been detached from its interface and it yet hasn't been marked
1414 if (d->bd_sbuf != NULL) {
1415 kfree(d->bd_sbuf, M_BPF);
1416 if (d->bd_hbuf != NULL)
1417 kfree(d->bd_hbuf, M_BPF);
1418 if (d->bd_fbuf != NULL)
1419 kfree(d->bd_fbuf, M_BPF);
1422 kfree(d->bd_rfilter, M_BPF);
1424 kfree(d->bd_wfilter, M_BPF);
1428 * Attach an interface to bpf. ifp is a pointer to the structure
1429 * defining the interface to be attached, dlt is the link layer type,
1430 * and hdrlen is the fixed size of the link header (variable length
1431 * headers are not yet supported).
1434 bpfattach(struct ifnet *ifp, u_int dlt, u_int hdrlen)
1436 bpfattach_dlt(ifp, dlt, hdrlen, &ifp->if_bpf);
1440 bpfattach_dlt(struct ifnet *ifp, u_int dlt, u_int hdrlen, struct bpf_if **driverp)
1444 bp = kmalloc(sizeof *bp, M_BPF, M_WAITOK | M_ZERO);
1446 lwkt_gettoken(&bpf_token);
1448 SLIST_INIT(&bp->bif_dlist);
1451 bp->bif_driverp = driverp;
1452 *bp->bif_driverp = NULL;
1454 bp->bif_next = bpf_iflist;
1458 * Compute the length of the bpf header. This is not necessarily
1459 * equal to SIZEOF_BPF_HDR because we want to insert spacing such
1460 * that the network layer header begins on a longword boundary (for
1461 * performance reasons and to alleviate alignment restrictions).
1463 bp->bif_hdrlen = BPF_WORDALIGN(hdrlen + SIZEOF_BPF_HDR) - hdrlen;
1465 lwkt_reltoken(&bpf_token);
1468 if_printf(ifp, "bpf attached\n");
1472 * Detach bpf from an interface. This involves detaching each descriptor
1473 * associated with the interface, and leaving bd_bif NULL. Notify each
1474 * descriptor as it's detached so that any sleepers wake up and get
1478 bpfdetach(struct ifnet *ifp)
1480 struct bpf_if *bp, *bp_prev;
1483 lwkt_gettoken(&bpf_token);
1485 /* Locate BPF interface information */
1487 for (bp = bpf_iflist; bp != NULL; bp = bp->bif_next) {
1488 if (ifp == bp->bif_ifp)
1493 /* Interface wasn't attached */
1494 if (bp->bif_ifp == NULL) {
1495 lwkt_reltoken(&bpf_token);
1496 kprintf("bpfdetach: %s was not attached\n", ifp->if_xname);
1500 while ((d = SLIST_FIRST(&bp->bif_dlist)) != NULL) {
1505 if (bp_prev != NULL)
1506 bp_prev->bif_next = bp->bif_next;
1508 bpf_iflist = bp->bif_next;
1512 lwkt_reltoken(&bpf_token);
1516 * Get a list of available data link type of the interface.
1519 bpf_getdltlist(struct bpf_d *d, struct bpf_dltlist *bfl)
1525 ifp = d->bd_bif->bif_ifp;
1528 for (bp = bpf_iflist; bp != NULL; bp = bp->bif_next) {
1529 if (bp->bif_ifp != ifp)
1531 if (bfl->bfl_list != NULL) {
1532 if (n >= bfl->bfl_len) {
1535 error = copyout(&bp->bif_dlt,
1536 bfl->bfl_list + n, sizeof(u_int));
1545 * Set the data link type of a BPF instance.
1548 bpf_setdlt(struct bpf_d *d, u_int dlt)
1550 int error, opromisc;
1554 if (d->bd_bif->bif_dlt == dlt)
1556 ifp = d->bd_bif->bif_ifp;
1557 for (bp = bpf_iflist; bp != NULL; bp = bp->bif_next) {
1558 if (bp->bif_ifp == ifp && bp->bif_dlt == dlt)
1562 opromisc = d->bd_promisc;
1567 error = ifpromisc(bp->bif_ifp, 1);
1569 if_printf(bp->bif_ifp,
1570 "bpf_setdlt: ifpromisc failed (%d)\n",
1577 return(bp == NULL ? EINVAL : 0);
1583 lwkt_gettoken(&bpf_token);
1589 lwkt_reltoken(&bpf_token);
1593 bpf_drvinit(void *unused)
1597 make_autoclone_dev(&bpf_ops, &DEVFS_CLONE_BITMAP(bpf),
1598 bpfclone, 0, 0, 0600, "bpf");
1599 for (i = 0; i < BPF_PREALLOCATED_UNITS; i++) {
1600 make_dev(&bpf_ops, i, 0, 0, 0600, "bpf%d", i);
1601 devfs_clone_bitmap_set(&DEVFS_CLONE_BITMAP(bpf), i);
1606 bpf_drvuninit(void *unused)
1608 devfs_clone_handler_del("bpf");
1609 dev_ops_remove_all(&bpf_ops);
1610 devfs_clone_bitmap_uninit(&DEVFS_CLONE_BITMAP(bpf));
1613 SYSINIT(bpfdev,SI_SUB_DRIVERS,SI_ORDER_MIDDLE+CDEV_MAJOR,bpf_drvinit,NULL)
1614 SYSUNINIT(bpfdev, SI_SUB_DRIVERS,SI_ORDER_MIDDLE+CDEV_MAJOR,bpf_drvuninit, NULL);
1618 * NOP stubs to allow bpf-using drivers to load and function.
1620 * A 'better' implementation would allow the core bpf functionality
1621 * to be loaded at runtime.
1625 bpf_tap(struct bpf_if *bp, u_char *pkt, u_int pktlen)
1630 bpf_mtap(struct bpf_if *bp, struct mbuf *m)
1635 bpf_ptap(struct bpf_if *bp, struct mbuf *m, const void *data, u_int dlen)
1640 bpfattach(struct ifnet *ifp, u_int dlt, u_int hdrlen)
1645 bpfattach_dlt(struct ifnet *ifp, u_int dlt, u_int hdrlen, struct bpf_if **driverp)
1650 bpfdetach(struct ifnet *ifp)
1655 bpf_filter(const struct bpf_insn *pc, u_char *p, u_int wirelen, u_int buflen)
1657 return -1; /* "no filter" behaviour */