Merge branch 'master' of ssh://crater.dragonflybsd.org/repository/git/dragonfly
[dragonfly.git] / sys / net / bpf.c
blob9ac8291defbb1d6128b46743c2dd036af0298661
1 /*
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
8 * Berkeley Laboratory.
10 * Redistribution and use in source and binary forms, with or without
11 * modification, are permitted provided that the following conditions
12 * are met:
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
36 * SUCH DAMAGE.
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 $
44 #include "use_bpf.h"
46 #include <sys/param.h>
47 #include <sys/systm.h>
48 #include <sys/conf.h>
49 #include <sys/device.h>
50 #include <sys/malloc.h>
51 #include <sys/mbuf.h>
52 #include <sys/time.h>
53 #include <sys/proc.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>
60 #include <sys/poll.h>
62 #include <sys/socket.h>
63 #include <sys/vnode.h>
65 #include <sys/thread2.h>
67 #include <net/if.h>
68 #include <net/bpf.h>
69 #include <net/bpfdesc.h>
70 #include <net/netmsg2.h>
72 #include <netinet/in.h>
73 #include <netinet/if_ether.h>
74 #include <sys/kernel.h>
75 #include <sys/sysctl.h>
77 struct netmsg_bpf_output {
78 struct netmsg nm_netmsg;
79 struct mbuf *nm_mbuf;
80 struct ifnet *nm_ifp;
81 struct sockaddr *nm_dst;
84 MALLOC_DEFINE(M_BPF, "BPF", "BPF data");
86 #if NBPF > 0
89 * The default read buffer size is patchable.
91 static int bpf_bufsize = BPF_DEFAULTBUFSIZE;
92 SYSCTL_INT(_debug, OID_AUTO, bpf_bufsize, CTLFLAG_RW,
93 &bpf_bufsize, 0, "");
94 int bpf_maxbufsize = BPF_MAXBUFSIZE;
95 SYSCTL_INT(_debug, OID_AUTO, bpf_maxbufsize, CTLFLAG_RW,
96 &bpf_maxbufsize, 0, "");
99 * bpf_iflist is the list of interfaces; each corresponds to an ifnet
101 static struct bpf_if *bpf_iflist;
103 static int bpf_allocbufs(struct bpf_d *);
104 static void bpf_attachd(struct bpf_d *d, struct bpf_if *bp);
105 static void bpf_detachd(struct bpf_d *d);
106 static void bpf_resetd(struct bpf_d *);
107 static void bpf_freed(struct bpf_d *);
108 static void bpf_mcopy(const void *, void *, size_t);
109 static int bpf_movein(struct uio *, int, struct mbuf **,
110 struct sockaddr *, int *, struct bpf_insn *);
111 static int bpf_setif(struct bpf_d *, struct ifreq *);
112 static void bpf_timed_out(void *);
113 static void bpf_wakeup(struct bpf_d *);
114 static void catchpacket(struct bpf_d *, u_char *, u_int, u_int,
115 void (*)(const void *, void *, size_t),
116 const struct timeval *);
117 static int bpf_setf(struct bpf_d *, struct bpf_program *, u_long cmd);
118 static int bpf_getdltlist(struct bpf_d *, struct bpf_dltlist *);
119 static int bpf_setdlt(struct bpf_d *, u_int);
120 static void bpf_drvinit(void *unused);
122 static d_open_t bpfopen;
123 static d_close_t bpfclose;
124 static d_read_t bpfread;
125 static d_write_t bpfwrite;
126 static d_ioctl_t bpfioctl;
127 static d_poll_t bpfpoll;
129 #define CDEV_MAJOR 23
130 static struct dev_ops bpf_ops = {
131 { "bpf", CDEV_MAJOR, 0 },
132 .d_open = bpfopen,
133 .d_close = bpfclose,
134 .d_read = bpfread,
135 .d_write = bpfwrite,
136 .d_ioctl = bpfioctl,
137 .d_poll = bpfpoll,
141 static int
142 bpf_movein(struct uio *uio, int linktype, struct mbuf **mp,
143 struct sockaddr *sockp, int *datlen, struct bpf_insn *wfilter)
145 struct mbuf *m;
146 int error;
147 int len;
148 int hlen;
149 int slen;
151 *datlen = 0;
152 *mp = NULL;
155 * Build a sockaddr based on the data link layer type.
156 * We do this at this level because the ethernet header
157 * is copied directly into the data field of the sockaddr.
158 * In the case of SLIP, there is no header and the packet
159 * is forwarded as is.
160 * Also, we are careful to leave room at the front of the mbuf
161 * for the link level header.
163 switch (linktype) {
164 case DLT_SLIP:
165 sockp->sa_family = AF_INET;
166 hlen = 0;
167 break;
169 case DLT_EN10MB:
170 sockp->sa_family = AF_UNSPEC;
171 /* XXX Would MAXLINKHDR be better? */
172 hlen = sizeof(struct ether_header);
173 break;
175 case DLT_RAW:
176 case DLT_NULL:
177 sockp->sa_family = AF_UNSPEC;
178 hlen = 0;
179 break;
181 case DLT_ATM_RFC1483:
183 * en atm driver requires 4-byte atm pseudo header.
184 * though it isn't standard, vpi:vci needs to be
185 * specified anyway.
187 sockp->sa_family = AF_UNSPEC;
188 hlen = 12; /* XXX 4(ATM_PH) + 3(LLC) + 5(SNAP) */
189 break;
191 case DLT_PPP:
192 sockp->sa_family = AF_UNSPEC;
193 hlen = 4; /* This should match PPP_HDRLEN */
194 break;
196 default:
197 return(EIO);
200 len = uio->uio_resid;
201 *datlen = len - hlen;
202 if ((unsigned)len > MCLBYTES)
203 return(EIO);
205 m = m_getl(len, MB_WAIT, MT_DATA, M_PKTHDR, NULL);
206 if (m == NULL)
207 return(ENOBUFS);
208 m->m_pkthdr.len = m->m_len = len;
209 m->m_pkthdr.rcvif = NULL;
210 *mp = m;
212 if (m->m_len < hlen) {
213 error = EPERM;
214 goto bad;
217 error = uiomove(mtod(m, u_char *), len, uio);
218 if (error)
219 goto bad;
221 slen = bpf_filter(wfilter, mtod(m, u_char *), len, len);
222 if (slen == 0) {
223 error = EPERM;
224 goto bad;
228 * Make room for link header, and copy it to sockaddr.
230 if (hlen != 0) {
231 bcopy(m->m_data, sockp->sa_data, hlen);
232 m->m_pkthdr.len -= hlen;
233 m->m_len -= hlen;
234 m->m_data += hlen; /* XXX */
236 return (0);
237 bad:
238 m_freem(m);
239 return(error);
243 * Attach file to the bpf interface, i.e. make d listen on bp.
244 * Must be called at splimp.
246 static void
247 bpf_attachd(struct bpf_d *d, struct bpf_if *bp)
250 * Point d at bp, and add d to the interface's list of listeners.
251 * Finally, point the driver's bpf cookie at the interface so
252 * it will divert packets to bpf.
254 d->bd_bif = bp;
255 SLIST_INSERT_HEAD(&bp->bif_dlist, d, bd_next);
256 *bp->bif_driverp = bp;
260 * Detach a file from its interface.
262 static void
263 bpf_detachd(struct bpf_d *d)
265 int error;
266 struct bpf_if *bp;
267 struct ifnet *ifp;
269 bp = d->bd_bif;
270 ifp = bp->bif_ifp;
272 /* Remove d from the interface's descriptor list. */
273 SLIST_REMOVE(&bp->bif_dlist, d, bpf_d, bd_next);
275 if (SLIST_EMPTY(&bp->bif_dlist)) {
277 * Let the driver know that there are no more listeners.
279 *bp->bif_driverp = NULL;
281 d->bd_bif = NULL;
283 * Check if this descriptor had requested promiscuous mode.
284 * If so, turn it off.
286 if (d->bd_promisc) {
287 d->bd_promisc = 0;
288 error = ifpromisc(ifp, 0);
289 if (error != 0 && error != ENXIO) {
291 * ENXIO can happen if a pccard is unplugged,
292 * Something is really wrong if we were able to put
293 * the driver into promiscuous mode, but can't
294 * take it out.
296 if_printf(ifp, "bpf_detach: ifpromisc failed(%d)\n",
297 error);
303 * Open ethernet device. Returns ENXIO for illegal minor device number,
304 * EBUSY if file is open by another process.
306 /* ARGSUSED */
307 static int
308 bpfopen(struct dev_open_args *ap)
310 cdev_t dev = ap->a_head.a_dev;
311 struct bpf_d *d;
313 if (ap->a_cred->cr_prison)
314 return(EPERM);
316 d = dev->si_drv1;
318 * Each minor can be opened by only one process. If the requested
319 * minor is in use, return EBUSY.
321 if (d != NULL)
322 return(EBUSY);
323 make_dev(&bpf_ops, minor(dev), 0, 0, 0600, "bpf%d", lminor(dev));
324 MALLOC(d, struct bpf_d *, sizeof *d, M_BPF, M_WAITOK | M_ZERO);
325 dev->si_drv1 = d;
326 d->bd_bufsize = bpf_bufsize;
327 d->bd_sig = SIGIO;
328 d->bd_seesent = 1;
329 callout_init(&d->bd_callout);
330 return(0);
334 * Close the descriptor by detaching it from its interface,
335 * deallocating its buffers, and marking it free.
337 /* ARGSUSED */
338 static int
339 bpfclose(struct dev_close_args *ap)
341 cdev_t dev = ap->a_head.a_dev;
342 struct bpf_d *d = dev->si_drv1;
344 funsetown(d->bd_sigio);
345 crit_enter();
346 if (d->bd_state == BPF_WAITING)
347 callout_stop(&d->bd_callout);
348 d->bd_state = BPF_IDLE;
349 if (d->bd_bif != NULL)
350 bpf_detachd(d);
351 crit_exit();
352 bpf_freed(d);
353 dev->si_drv1 = NULL;
354 kfree(d, M_BPF);
356 return(0);
360 * Rotate the packet buffers in descriptor d. Move the store buffer
361 * into the hold slot, and the free buffer into the store slot.
362 * Zero the length of the new store buffer.
364 #define ROTATE_BUFFERS(d) \
365 (d)->bd_hbuf = (d)->bd_sbuf; \
366 (d)->bd_hlen = (d)->bd_slen; \
367 (d)->bd_sbuf = (d)->bd_fbuf; \
368 (d)->bd_slen = 0; \
369 (d)->bd_fbuf = NULL;
371 * bpfread - read next chunk of packets from buffers
373 static int
374 bpfread(struct dev_read_args *ap)
376 cdev_t dev = ap->a_head.a_dev;
377 struct bpf_d *d = dev->si_drv1;
378 int timed_out;
379 int error;
382 * Restrict application to use a buffer the same size as
383 * as kernel buffers.
385 if (ap->a_uio->uio_resid != d->bd_bufsize)
386 return(EINVAL);
388 crit_enter();
389 if (d->bd_state == BPF_WAITING)
390 callout_stop(&d->bd_callout);
391 timed_out = (d->bd_state == BPF_TIMED_OUT);
392 d->bd_state = BPF_IDLE;
394 * If the hold buffer is empty, then do a timed sleep, which
395 * ends when the timeout expires or when enough packets
396 * have arrived to fill the store buffer.
398 while (d->bd_hbuf == NULL) {
399 if ((d->bd_immediate || timed_out) && d->bd_slen != 0) {
401 * A packet(s) either arrived since the previous
402 * read or arrived while we were asleep.
403 * Rotate the buffers and return what's here.
405 ROTATE_BUFFERS(d);
406 break;
410 * No data is available, check to see if the bpf device
411 * is still pointed at a real interface. If not, return
412 * ENXIO so that the userland process knows to rebind
413 * it before using it again.
415 if (d->bd_bif == NULL) {
416 crit_exit();
417 return(ENXIO);
420 if (ap->a_ioflag & IO_NDELAY) {
421 crit_exit();
422 return(EWOULDBLOCK);
424 error = tsleep(d, PCATCH, "bpf", d->bd_rtout);
425 if (error == EINTR || error == ERESTART) {
426 crit_exit();
427 return(error);
429 if (error == EWOULDBLOCK) {
431 * On a timeout, return what's in the buffer,
432 * which may be nothing. If there is something
433 * in the store buffer, we can rotate the buffers.
435 if (d->bd_hbuf)
437 * We filled up the buffer in between
438 * getting the timeout and arriving
439 * here, so we don't need to rotate.
441 break;
443 if (d->bd_slen == 0) {
444 crit_exit();
445 return(0);
447 ROTATE_BUFFERS(d);
448 break;
452 * At this point, we know we have something in the hold slot.
454 crit_exit();
457 * Move data from hold buffer into user space.
458 * We know the entire buffer is transferred since
459 * we checked above that the read buffer is bpf_bufsize bytes.
461 error = uiomove(d->bd_hbuf, d->bd_hlen, ap->a_uio);
463 crit_enter();
464 d->bd_fbuf = d->bd_hbuf;
465 d->bd_hbuf = NULL;
466 d->bd_hlen = 0;
467 crit_exit();
469 return(error);
474 * If there are processes sleeping on this descriptor, wake them up.
476 static void
477 bpf_wakeup(struct bpf_d *d)
479 if (d->bd_state == BPF_WAITING) {
480 callout_stop(&d->bd_callout);
481 d->bd_state = BPF_IDLE;
483 wakeup(d);
484 if (d->bd_async && d->bd_sig && d->bd_sigio)
485 pgsigio(d->bd_sigio, d->bd_sig, 0);
487 get_mplock();
488 selwakeup(&d->bd_sel);
489 rel_mplock();
490 /* XXX */
491 d->bd_sel.si_pid = 0;
494 static void
495 bpf_timed_out(void *arg)
497 struct bpf_d *d = (struct bpf_d *)arg;
499 crit_enter();
500 if (d->bd_state == BPF_WAITING) {
501 d->bd_state = BPF_TIMED_OUT;
502 if (d->bd_slen != 0)
503 bpf_wakeup(d);
505 crit_exit();
508 static void
509 bpf_output_dispatch(struct netmsg *nmsg)
511 struct netmsg_bpf_output *bmsg = (struct netmsg_bpf_output *)nmsg;
512 struct ifnet *ifp = bmsg->nm_ifp;
513 int error;
516 * The driver frees the mbuf.
518 error = ifp->if_output(ifp, bmsg->nm_mbuf, bmsg->nm_dst, NULL);
519 lwkt_replymsg(&nmsg->nm_lmsg, error);
522 static int
523 bpfwrite(struct dev_write_args *ap)
525 cdev_t dev = ap->a_head.a_dev;
526 struct bpf_d *d = dev->si_drv1;
527 struct ifnet *ifp;
528 struct mbuf *m;
529 int error;
530 struct sockaddr dst;
531 int datlen;
532 struct netmsg_bpf_output bmsg;
534 if (d->bd_bif == NULL)
535 return(ENXIO);
537 ifp = d->bd_bif->bif_ifp;
539 if (ap->a_uio->uio_resid == 0)
540 return(0);
542 error = bpf_movein(ap->a_uio, (int)d->bd_bif->bif_dlt, &m,
543 &dst, &datlen, d->bd_wfilter);
544 if (error)
545 return(error);
547 if (datlen > ifp->if_mtu) {
548 m_freem(m);
549 return(EMSGSIZE);
552 if (d->bd_hdrcmplt)
553 dst.sa_family = pseudo_AF_HDRCMPLT;
555 netmsg_init(&bmsg.nm_netmsg, &curthread->td_msgport, MSGF_MPSAFE,
556 bpf_output_dispatch);
557 bmsg.nm_mbuf = m;
558 bmsg.nm_ifp = ifp;
559 bmsg.nm_dst = &dst;
561 return lwkt_domsg(cpu_portfn(0), &bmsg.nm_netmsg.nm_lmsg, 0);
565 * Reset a descriptor by flushing its packet buffer and clearing the
566 * receive and drop counts. Should be called at splimp.
568 static void
569 bpf_resetd(struct bpf_d *d)
571 if (d->bd_hbuf) {
572 /* Free the hold buffer. */
573 d->bd_fbuf = d->bd_hbuf;
574 d->bd_hbuf = NULL;
576 d->bd_slen = 0;
577 d->bd_hlen = 0;
578 d->bd_rcount = 0;
579 d->bd_dcount = 0;
583 * FIONREAD Check for read packet available.
584 * SIOCGIFADDR Get interface address - convenient hook to driver.
585 * BIOCGBLEN Get buffer len [for read()].
586 * BIOCSETF Set ethernet read filter.
587 * BIOCSETWF Set ethernet write filter.
588 * BIOCFLUSH Flush read packet buffer.
589 * BIOCPROMISC Put interface into promiscuous mode.
590 * BIOCGDLT Get link layer type.
591 * BIOCGETIF Get interface name.
592 * BIOCSETIF Set interface.
593 * BIOCSRTIMEOUT Set read timeout.
594 * BIOCGRTIMEOUT Get read timeout.
595 * BIOCGSTATS Get packet stats.
596 * BIOCIMMEDIATE Set immediate mode.
597 * BIOCVERSION Get filter language version.
598 * BIOCGHDRCMPLT Get "header already complete" flag
599 * BIOCSHDRCMPLT Set "header already complete" flag
600 * BIOCGSEESENT Get "see packets sent" flag
601 * BIOCSSEESENT Set "see packets sent" flag
602 * BIOCLOCK Set "locked" flag
604 /* ARGSUSED */
605 static int
606 bpfioctl(struct dev_ioctl_args *ap)
608 cdev_t dev = ap->a_head.a_dev;
609 struct bpf_d *d = dev->si_drv1;
610 int error = 0;
612 crit_enter();
613 if (d->bd_state == BPF_WAITING)
614 callout_stop(&d->bd_callout);
615 d->bd_state = BPF_IDLE;
616 crit_exit();
618 if (d->bd_locked == 1) {
619 switch (ap->a_cmd) {
620 case BIOCGBLEN:
621 case BIOCFLUSH:
622 case BIOCGDLT:
623 case BIOCGDLTLIST:
624 case BIOCGETIF:
625 case BIOCGRTIMEOUT:
626 case BIOCGSTATS:
627 case BIOCVERSION:
628 case BIOCGRSIG:
629 case BIOCGHDRCMPLT:
630 case FIONREAD:
631 case BIOCLOCK:
632 case BIOCSRTIMEOUT:
633 case BIOCIMMEDIATE:
634 case TIOCGPGRP:
635 break;
636 default:
637 return (EPERM);
640 switch (ap->a_cmd) {
641 default:
642 error = EINVAL;
643 break;
646 * Check for read packet available.
648 case FIONREAD:
650 int n;
652 crit_enter();
653 n = d->bd_slen;
654 if (d->bd_hbuf)
655 n += d->bd_hlen;
656 crit_exit();
658 *(int *)ap->a_data = n;
659 break;
662 case SIOCGIFADDR:
664 struct ifnet *ifp;
666 if (d->bd_bif == NULL) {
667 error = EINVAL;
668 } else {
669 ifp = d->bd_bif->bif_ifp;
670 lwkt_serialize_enter(ifp->if_serializer);
671 error = ifp->if_ioctl(ifp, ap->a_cmd,
672 ap->a_data, ap->a_cred);
673 lwkt_serialize_exit(ifp->if_serializer);
675 break;
679 * Get buffer len [for read()].
681 case BIOCGBLEN:
682 *(u_int *)ap->a_data = d->bd_bufsize;
683 break;
686 * Set buffer length.
688 case BIOCSBLEN:
689 if (d->bd_bif != NULL) {
690 error = EINVAL;
691 } else {
692 u_int size = *(u_int *)ap->a_data;
694 if (size > bpf_maxbufsize)
695 *(u_int *)ap->a_data = size = bpf_maxbufsize;
696 else if (size < BPF_MINBUFSIZE)
697 *(u_int *)ap->a_data = size = BPF_MINBUFSIZE;
698 d->bd_bufsize = size;
700 break;
703 * Set link layer read filter.
705 case BIOCSETF:
706 case BIOCSETWF:
707 error = bpf_setf(d, (struct bpf_program *)ap->a_data,
708 ap->a_cmd);
709 break;
712 * Flush read packet buffer.
714 case BIOCFLUSH:
715 crit_enter();
716 bpf_resetd(d);
717 crit_exit();
718 break;
721 * Put interface into promiscuous mode.
723 case BIOCPROMISC:
724 if (d->bd_bif == NULL) {
726 * No interface attached yet.
728 error = EINVAL;
729 break;
731 crit_enter();
732 if (d->bd_promisc == 0) {
733 error = ifpromisc(d->bd_bif->bif_ifp, 1);
734 if (error == 0)
735 d->bd_promisc = 1;
737 crit_exit();
738 break;
741 * Get device parameters.
743 case BIOCGDLT:
744 if (d->bd_bif == NULL)
745 error = EINVAL;
746 else
747 *(u_int *)ap->a_data = d->bd_bif->bif_dlt;
748 break;
751 * Get a list of supported data link types.
753 case BIOCGDLTLIST:
754 if (d->bd_bif == NULL) {
755 error = EINVAL;
756 } else {
757 error = bpf_getdltlist(d,
758 (struct bpf_dltlist *)ap->a_data);
760 break;
763 * Set data link type.
765 case BIOCSDLT:
766 if (d->bd_bif == NULL)
767 error = EINVAL;
768 else
769 error = bpf_setdlt(d, *(u_int *)ap->a_data);
770 break;
773 * Get interface name.
775 case BIOCGETIF:
776 if (d->bd_bif == NULL) {
777 error = EINVAL;
778 } else {
779 struct ifnet *const ifp = d->bd_bif->bif_ifp;
780 struct ifreq *const ifr = (struct ifreq *)ap->a_data;
782 strlcpy(ifr->ifr_name, ifp->if_xname,
783 sizeof ifr->ifr_name);
785 break;
788 * Set interface.
790 case BIOCSETIF:
791 error = bpf_setif(d, (struct ifreq *)ap->a_data);
792 break;
795 * Set read timeout.
797 case BIOCSRTIMEOUT:
799 struct timeval *tv = (struct timeval *)ap->a_data;
802 * Subtract 1 tick from tvtohz() since this isn't
803 * a one-shot timer.
805 if ((error = itimerfix(tv)) == 0)
806 d->bd_rtout = tvtohz_low(tv);
807 break;
811 * Get read timeout.
813 case BIOCGRTIMEOUT:
815 struct timeval *tv = (struct timeval *)ap->a_data;
817 tv->tv_sec = d->bd_rtout / hz;
818 tv->tv_usec = (d->bd_rtout % hz) * tick;
819 break;
823 * Get packet stats.
825 case BIOCGSTATS:
827 struct bpf_stat *bs = (struct bpf_stat *)ap->a_data;
829 bs->bs_recv = d->bd_rcount;
830 bs->bs_drop = d->bd_dcount;
831 break;
835 * Set immediate mode.
837 case BIOCIMMEDIATE:
838 d->bd_immediate = *(u_int *)ap->a_data;
839 break;
841 case BIOCVERSION:
843 struct bpf_version *bv = (struct bpf_version *)ap->a_data;
845 bv->bv_major = BPF_MAJOR_VERSION;
846 bv->bv_minor = BPF_MINOR_VERSION;
847 break;
851 * Get "header already complete" flag
853 case BIOCGHDRCMPLT:
854 *(u_int *)ap->a_data = d->bd_hdrcmplt;
855 break;
858 * Set "header already complete" flag
860 case BIOCSHDRCMPLT:
861 d->bd_hdrcmplt = *(u_int *)ap->a_data ? 1 : 0;
862 break;
865 * Get "see sent packets" flag
867 case BIOCGSEESENT:
868 *(u_int *)ap->a_data = d->bd_seesent;
869 break;
872 * Set "see sent packets" flag
874 case BIOCSSEESENT:
875 d->bd_seesent = *(u_int *)ap->a_data;
876 break;
878 case FIOASYNC: /* Send signal on receive packets */
879 d->bd_async = *(int *)ap->a_data;
880 break;
882 case FIOSETOWN:
883 error = fsetown(*(int *)ap->a_data, &d->bd_sigio);
884 break;
886 case FIOGETOWN:
887 *(int *)ap->a_data = fgetown(d->bd_sigio);
888 break;
890 /* This is deprecated, FIOSETOWN should be used instead. */
891 case TIOCSPGRP:
892 error = fsetown(-(*(int *)ap->a_data), &d->bd_sigio);
893 break;
895 /* This is deprecated, FIOGETOWN should be used instead. */
896 case TIOCGPGRP:
897 *(int *)ap->a_data = -fgetown(d->bd_sigio);
898 break;
900 case BIOCSRSIG: /* Set receive signal */
902 u_int sig;
904 sig = *(u_int *)ap->a_data;
906 if (sig >= NSIG)
907 error = EINVAL;
908 else
909 d->bd_sig = sig;
910 break;
912 case BIOCGRSIG:
913 *(u_int *)ap->a_data = d->bd_sig;
914 break;
915 case BIOCLOCK:
916 d->bd_locked = 1;
917 break;
919 return(error);
923 * Set d's packet filter program to fp. If this file already has a filter,
924 * free it and replace it. Returns EINVAL for bogus requests.
926 static int
927 bpf_setf(struct bpf_d *d, struct bpf_program *fp, u_long cmd)
929 struct bpf_insn *fcode, *old;
930 u_int wfilter, flen, size;
932 if (cmd == BIOCSETWF) {
933 old = d->bd_wfilter;
934 wfilter = 1;
935 } else {
936 wfilter = 0;
937 old = d->bd_rfilter;
939 if (fp->bf_insns == NULL) {
940 if (fp->bf_len != 0)
941 return(EINVAL);
942 crit_enter();
943 if (wfilter)
944 d->bd_wfilter = NULL;
945 else
946 d->bd_rfilter = NULL;
947 bpf_resetd(d);
948 crit_exit();
949 if (old != NULL)
950 kfree(old, M_BPF);
951 return(0);
953 flen = fp->bf_len;
954 if (flen > BPF_MAXINSNS)
955 return(EINVAL);
957 size = flen * sizeof *fp->bf_insns;
958 fcode = (struct bpf_insn *)kmalloc(size, M_BPF, M_WAITOK);
959 if (copyin(fp->bf_insns, fcode, size) == 0 &&
960 bpf_validate(fcode, (int)flen)) {
961 crit_enter();
962 if (wfilter)
963 d->bd_wfilter = fcode;
964 else
965 d->bd_rfilter = fcode;
966 bpf_resetd(d);
967 crit_exit();
968 if (old != NULL)
969 kfree(old, M_BPF);
971 return(0);
973 kfree(fcode, M_BPF);
974 return(EINVAL);
978 * Detach a file from its current interface (if attached at all) and attach
979 * to the interface indicated by the name stored in ifr.
980 * Return an errno or 0.
982 static int
983 bpf_setif(struct bpf_d *d, struct ifreq *ifr)
985 struct bpf_if *bp;
986 int error;
987 struct ifnet *theywant;
989 theywant = ifunit(ifr->ifr_name);
990 if (theywant == NULL)
991 return(ENXIO);
994 * Look through attached interfaces for the named one.
996 for (bp = bpf_iflist; bp != NULL; bp = bp->bif_next) {
997 struct ifnet *ifp = bp->bif_ifp;
999 if (ifp == NULL || ifp != theywant)
1000 continue;
1001 /* skip additional entry */
1002 if (bp->bif_driverp != &ifp->if_bpf)
1003 continue;
1005 * We found the requested interface.
1006 * If it's not up, return an error.
1007 * Allocate the packet buffers if we need to.
1008 * If we're already attached to requested interface,
1009 * just flush the buffer.
1011 if (!(ifp->if_flags & IFF_UP))
1012 return(ENETDOWN);
1014 if (d->bd_sbuf == NULL) {
1015 error = bpf_allocbufs(d);
1016 if (error != 0)
1017 return(error);
1019 crit_enter();
1020 if (bp != d->bd_bif) {
1021 if (d->bd_bif != NULL) {
1023 * Detach if attached to something else.
1025 bpf_detachd(d);
1028 bpf_attachd(d, bp);
1030 bpf_resetd(d);
1031 crit_exit();
1032 return(0);
1035 /* Not found. */
1036 return(ENXIO);
1040 * Support for select() and poll() system calls
1042 * Return true iff the specific operation will not block indefinitely.
1043 * Otherwise, return false but make a note that a selwakeup() must be done.
1045 static int
1046 bpfpoll(struct dev_poll_args *ap)
1048 cdev_t dev = ap->a_head.a_dev;
1049 struct bpf_d *d;
1050 int revents;
1052 d = dev->si_drv1;
1053 if (d->bd_bif == NULL)
1054 return(ENXIO);
1056 revents = ap->a_events & (POLLOUT | POLLWRNORM);
1057 crit_enter();
1058 if (ap->a_events & (POLLIN | POLLRDNORM)) {
1060 * An imitation of the FIONREAD ioctl code.
1061 * XXX not quite. An exact imitation:
1062 * if (d->b_slen != 0 ||
1063 * (d->bd_hbuf != NULL && d->bd_hlen != 0)
1065 if (d->bd_hlen != 0 ||
1066 ((d->bd_immediate || d->bd_state == BPF_TIMED_OUT) &&
1067 d->bd_slen != 0)) {
1068 revents |= ap->a_events & (POLLIN | POLLRDNORM);
1069 } else {
1070 selrecord(curthread, &d->bd_sel);
1071 /* Start the read timeout if necessary. */
1072 if (d->bd_rtout > 0 && d->bd_state == BPF_IDLE) {
1073 callout_reset(&d->bd_callout, d->bd_rtout,
1074 bpf_timed_out, d);
1075 d->bd_state = BPF_WAITING;
1079 crit_exit();
1080 ap->a_events = revents;
1081 return(0);
1085 * Process the packet pkt of length pktlen. The packet is parsed
1086 * by each listener's filter, and if accepted, stashed into the
1087 * corresponding buffer.
1089 void
1090 bpf_tap(struct bpf_if *bp, u_char *pkt, u_int pktlen)
1092 struct bpf_d *d;
1093 struct timeval tv;
1094 int gottime = 0;
1095 u_int slen;
1097 get_mplock();
1099 /* Re-check */
1100 if (bp == NULL) {
1101 rel_mplock();
1102 return;
1106 * Note that the ipl does not have to be raised at this point.
1107 * The only problem that could arise here is that if two different
1108 * interfaces shared any data. This is not the case.
1110 SLIST_FOREACH(d, &bp->bif_dlist, bd_next) {
1111 ++d->bd_rcount;
1112 slen = bpf_filter(d->bd_rfilter, pkt, pktlen, pktlen);
1113 if (slen != 0) {
1114 if (!gottime) {
1115 microtime(&tv);
1116 gottime = 1;
1118 catchpacket(d, pkt, pktlen, slen, ovbcopy, &tv);
1122 rel_mplock();
1126 * Copy data from an mbuf chain into a buffer. This code is derived
1127 * from m_copydata in sys/uipc_mbuf.c.
1129 static void
1130 bpf_mcopy(const void *src_arg, void *dst_arg, size_t len)
1132 const struct mbuf *m;
1133 u_int count;
1134 u_char *dst;
1136 m = src_arg;
1137 dst = dst_arg;
1138 while (len > 0) {
1139 if (m == NULL)
1140 panic("bpf_mcopy");
1141 count = min(m->m_len, len);
1142 bcopy(mtod(m, void *), dst, count);
1143 m = m->m_next;
1144 dst += count;
1145 len -= count;
1150 * Process the packet in the mbuf chain m. The packet is parsed by each
1151 * listener's filter, and if accepted, stashed into the corresponding
1152 * buffer.
1154 void
1155 bpf_mtap(struct bpf_if *bp, struct mbuf *m)
1157 struct bpf_d *d;
1158 u_int pktlen, slen;
1159 struct timeval tv;
1160 int gottime = 0;
1162 get_mplock();
1164 /* Re-check */
1165 if (bp == NULL) {
1166 rel_mplock();
1167 return;
1170 /* Don't compute pktlen, if no descriptor is attached. */
1171 if (SLIST_EMPTY(&bp->bif_dlist)) {
1172 rel_mplock();
1173 return;
1176 pktlen = m_lengthm(m, NULL);
1178 SLIST_FOREACH(d, &bp->bif_dlist, bd_next) {
1179 if (!d->bd_seesent && (m->m_pkthdr.rcvif == NULL))
1180 continue;
1181 ++d->bd_rcount;
1182 slen = bpf_filter(d->bd_rfilter, (u_char *)m, pktlen, 0);
1183 if (slen != 0) {
1184 if (!gottime) {
1185 microtime(&tv);
1186 gottime = 1;
1188 catchpacket(d, (u_char *)m, pktlen, slen, bpf_mcopy,
1189 &tv);
1193 rel_mplock();
1196 void
1197 bpf_mtap_family(struct bpf_if *bp, struct mbuf *m, sa_family_t family)
1199 u_int family4;
1201 KKASSERT(family != AF_UNSPEC);
1203 family4 = (u_int)family;
1204 bpf_ptap(bp, m, &family4, sizeof(family4));
1208 * Process the packet in the mbuf chain m with the header in m prepended.
1209 * The packet is parsed by each listener's filter, and if accepted,
1210 * stashed into the corresponding buffer.
1212 void
1213 bpf_ptap(struct bpf_if *bp, struct mbuf *m, const void *data, u_int dlen)
1215 struct mbuf mb;
1218 * Craft on-stack mbuf suitable for passing to bpf_mtap.
1219 * Note that we cut corners here; we only setup what's
1220 * absolutely needed--this mbuf should never go anywhere else.
1222 mb.m_next = m;
1223 mb.m_data = __DECONST(void *, data); /* LINTED */
1224 mb.m_len = dlen;
1225 mb.m_pkthdr.rcvif = m->m_pkthdr.rcvif;
1227 bpf_mtap(bp, &mb);
1231 * Move the packet data from interface memory (pkt) into the
1232 * store buffer. Return 1 if it's time to wakeup a listener (buffer full),
1233 * otherwise 0. "copy" is the routine called to do the actual data
1234 * transfer. bcopy is passed in to copy contiguous chunks, while
1235 * bpf_mcopy is passed in to copy mbuf chains. In the latter case,
1236 * pkt is really an mbuf.
1238 static void
1239 catchpacket(struct bpf_d *d, u_char *pkt, u_int pktlen, u_int snaplen,
1240 void (*cpfn)(const void *, void *, size_t),
1241 const struct timeval *tv)
1243 struct bpf_hdr *hp;
1244 int totlen, curlen;
1245 int hdrlen = d->bd_bif->bif_hdrlen;
1247 * Figure out how many bytes to move. If the packet is
1248 * greater or equal to the snapshot length, transfer that
1249 * much. Otherwise, transfer the whole packet (unless
1250 * we hit the buffer size limit).
1252 totlen = hdrlen + min(snaplen, pktlen);
1253 if (totlen > d->bd_bufsize)
1254 totlen = d->bd_bufsize;
1257 * Round up the end of the previous packet to the next longword.
1259 curlen = BPF_WORDALIGN(d->bd_slen);
1260 if (curlen + totlen > d->bd_bufsize) {
1262 * This packet will overflow the storage buffer.
1263 * Rotate the buffers if we can, then wakeup any
1264 * pending reads.
1266 if (d->bd_fbuf == NULL) {
1268 * We haven't completed the previous read yet,
1269 * so drop the packet.
1271 ++d->bd_dcount;
1272 return;
1274 ROTATE_BUFFERS(d);
1275 bpf_wakeup(d);
1276 curlen = 0;
1277 } else if (d->bd_immediate || d->bd_state == BPF_TIMED_OUT) {
1279 * Immediate mode is set, or the read timeout has
1280 * already expired during a select call. A packet
1281 * arrived, so the reader should be woken up.
1283 bpf_wakeup(d);
1287 * Append the bpf header.
1289 hp = (struct bpf_hdr *)(d->bd_sbuf + curlen);
1290 hp->bh_tstamp = *tv;
1291 hp->bh_datalen = pktlen;
1292 hp->bh_hdrlen = hdrlen;
1294 * Copy the packet data into the store buffer and update its length.
1296 (*cpfn)(pkt, (u_char *)hp + hdrlen, (hp->bh_caplen = totlen - hdrlen));
1297 d->bd_slen = curlen + totlen;
1301 * Initialize all nonzero fields of a descriptor.
1303 static int
1304 bpf_allocbufs(struct bpf_d *d)
1306 d->bd_fbuf = kmalloc(d->bd_bufsize, M_BPF, M_WAITOK);
1307 d->bd_sbuf = kmalloc(d->bd_bufsize, M_BPF, M_WAITOK);
1308 d->bd_slen = 0;
1309 d->bd_hlen = 0;
1310 return(0);
1314 * Free buffers and packet filter program currently in use by a descriptor.
1315 * Called on close.
1317 static void
1318 bpf_freed(struct bpf_d *d)
1321 * We don't need to lock out interrupts since this descriptor has
1322 * been detached from its interface and it yet hasn't been marked
1323 * free.
1325 if (d->bd_sbuf != NULL) {
1326 kfree(d->bd_sbuf, M_BPF);
1327 if (d->bd_hbuf != NULL)
1328 kfree(d->bd_hbuf, M_BPF);
1329 if (d->bd_fbuf != NULL)
1330 kfree(d->bd_fbuf, M_BPF);
1332 if (d->bd_rfilter)
1333 kfree(d->bd_rfilter, M_BPF);
1334 if (d->bd_wfilter)
1335 kfree(d->bd_wfilter, M_BPF);
1339 * Attach an interface to bpf. ifp is a pointer to the structure
1340 * defining the interface to be attached, dlt is the link layer type,
1341 * and hdrlen is the fixed size of the link header (variable length
1342 * headers are not yet supported).
1344 void
1345 bpfattach(struct ifnet *ifp, u_int dlt, u_int hdrlen)
1347 bpfattach_dlt(ifp, dlt, hdrlen, &ifp->if_bpf);
1350 void
1351 bpfattach_dlt(struct ifnet *ifp, u_int dlt, u_int hdrlen, struct bpf_if **driverp)
1353 struct bpf_if *bp;
1355 bp = kmalloc(sizeof *bp, M_BPF, M_WAITOK | M_ZERO);
1357 SLIST_INIT(&bp->bif_dlist);
1358 bp->bif_ifp = ifp;
1359 bp->bif_dlt = dlt;
1360 bp->bif_driverp = driverp;
1361 *bp->bif_driverp = NULL;
1363 bp->bif_next = bpf_iflist;
1364 bpf_iflist = bp;
1367 * Compute the length of the bpf header. This is not necessarily
1368 * equal to SIZEOF_BPF_HDR because we want to insert spacing such
1369 * that the network layer header begins on a longword boundary (for
1370 * performance reasons and to alleviate alignment restrictions).
1372 bp->bif_hdrlen = BPF_WORDALIGN(hdrlen + SIZEOF_BPF_HDR) - hdrlen;
1374 if (bootverbose)
1375 if_printf(ifp, "bpf attached\n");
1379 * Detach bpf from an interface. This involves detaching each descriptor
1380 * associated with the interface, and leaving bd_bif NULL. Notify each
1381 * descriptor as it's detached so that any sleepers wake up and get
1382 * ENXIO.
1384 void
1385 bpfdetach(struct ifnet *ifp)
1387 struct bpf_if *bp, *bp_prev;
1388 struct bpf_d *d;
1390 crit_enter();
1392 /* Locate BPF interface information */
1393 bp_prev = NULL;
1394 for (bp = bpf_iflist; bp != NULL; bp = bp->bif_next) {
1395 if (ifp == bp->bif_ifp)
1396 break;
1397 bp_prev = bp;
1400 /* Interface wasn't attached */
1401 if (bp->bif_ifp == NULL) {
1402 crit_exit();
1403 kprintf("bpfdetach: %s was not attached\n", ifp->if_xname);
1404 return;
1407 while ((d = SLIST_FIRST(&bp->bif_dlist)) != NULL) {
1408 bpf_detachd(d);
1409 bpf_wakeup(d);
1412 if (bp_prev != NULL)
1413 bp_prev->bif_next = bp->bif_next;
1414 else
1415 bpf_iflist = bp->bif_next;
1417 kfree(bp, M_BPF);
1419 crit_exit();
1423 * Get a list of available data link type of the interface.
1425 static int
1426 bpf_getdltlist(struct bpf_d *d, struct bpf_dltlist *bfl)
1428 int n, error;
1429 struct ifnet *ifp;
1430 struct bpf_if *bp;
1432 ifp = d->bd_bif->bif_ifp;
1433 n = 0;
1434 error = 0;
1435 for (bp = bpf_iflist; bp != NULL; bp = bp->bif_next) {
1436 if (bp->bif_ifp != ifp)
1437 continue;
1438 if (bfl->bfl_list != NULL) {
1439 if (n >= bfl->bfl_len) {
1440 return (ENOMEM);
1442 error = copyout(&bp->bif_dlt,
1443 bfl->bfl_list + n, sizeof(u_int));
1445 n++;
1447 bfl->bfl_len = n;
1448 return(error);
1452 * Set the data link type of a BPF instance.
1454 static int
1455 bpf_setdlt(struct bpf_d *d, u_int dlt)
1457 int error, opromisc;
1458 struct ifnet *ifp;
1459 struct bpf_if *bp;
1461 if (d->bd_bif->bif_dlt == dlt)
1462 return (0);
1463 ifp = d->bd_bif->bif_ifp;
1464 for (bp = bpf_iflist; bp != NULL; bp = bp->bif_next) {
1465 if (bp->bif_ifp == ifp && bp->bif_dlt == dlt)
1466 break;
1468 if (bp != NULL) {
1469 opromisc = d->bd_promisc;
1470 crit_enter();
1471 bpf_detachd(d);
1472 bpf_attachd(d, bp);
1473 bpf_resetd(d);
1474 if (opromisc) {
1475 error = ifpromisc(bp->bif_ifp, 1);
1476 if (error) {
1477 if_printf(bp->bif_ifp,
1478 "bpf_setdlt: ifpromisc failed (%d)\n",
1479 error);
1480 } else {
1481 d->bd_promisc = 1;
1484 crit_exit();
1486 return(bp == NULL ? EINVAL : 0);
1489 static void
1490 bpf_drvinit(void *unused)
1492 dev_ops_add(&bpf_ops, 0, 0);
1495 SYSINIT(bpfdev,SI_SUB_DRIVERS,SI_ORDER_MIDDLE+CDEV_MAJOR,bpf_drvinit,NULL)
1497 #else /* !BPF */
1499 * NOP stubs to allow bpf-using drivers to load and function.
1501 * A 'better' implementation would allow the core bpf functionality
1502 * to be loaded at runtime.
1505 void
1506 bpf_tap(struct bpf_if *bp, u_char *pkt, u_int pktlen)
1510 void
1511 bpf_mtap(struct bpf_if *bp, struct mbuf *m)
1515 void
1516 bpf_ptap(struct bpf_if *bp, struct mbuf *m, const void *data, u_int dlen)
1520 void
1521 bpfattach(struct ifnet *ifp, u_int dlt, u_int hdrlen)
1525 void
1526 bpfattach_dlt(struct ifnet *ifp, u_int dlt, u_int hdrlen, struct bpf_if **driverp)
1530 void
1531 bpfdetach(struct ifnet *ifp)
1535 u_int
1536 bpf_filter(const struct bpf_insn *pc, u_char *p, u_int wirelen, u_int buflen)
1538 return -1; /* "no filter" behaviour */
1541 #endif /* !BPF */