xref: /dflybsd-src/sys/net/bpf.c (revision ca86d83e7d8d6bfef814ef3683c37d99ad62f11c)
1 /*
2  * Copyright (c) 1990, 1991, 1993
3  *	The Regents of the University of California.  All rights reserved.
4  *
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.
9  *
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.
25  *
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.
37  *
38  *      @(#)bpf.c	8.2 (Berkeley) 3/28/94
39  *
40  * $FreeBSD: src/sys/net/bpf.c,v 1.59.2.12 2002/04/14 21:41:48 luigi Exp $
41  */
42 
43 #include "use_bpf.h"
44 
45 #include <sys/param.h>
46 #include <sys/systm.h>
47 #include <sys/conf.h>
48 #include <sys/device.h>
49 #include <sys/malloc.h>
50 #include <sys/mbuf.h>
51 #include <sys/time.h>
52 #include <sys/proc.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>
58 
59 #include <sys/event.h>
60 
61 #include <sys/socket.h>
62 #include <sys/vnode.h>
63 
64 #include <sys/thread2.h>
65 
66 #include <net/if.h>
67 #include <net/bpf.h>
68 #include <net/bpfdesc.h>
69 #include <net/netmsg2.h>
70 #include <net/netisr2.h>
71 
72 #include <netinet/in.h>
73 #include <netinet/if_ether.h>
74 #include <sys/kernel.h>
75 #include <sys/sysctl.h>
76 
77 #include <sys/devfs.h>
78 
79 struct netmsg_bpf_output {
80 	struct netmsg_base base;
81 	struct mbuf	*nm_mbuf;
82 	struct ifnet	*nm_ifp;
83 	struct sockaddr	*nm_dst;
84 };
85 
86 MALLOC_DEFINE(M_BPF, "BPF", "BPF data");
87 DEVFS_DECLARE_CLONE_BITMAP(bpf);
88 
89 #if NBPF <= 1
90 #define BPF_PREALLOCATED_UNITS	4
91 #else
92 #define BPF_PREALLOCATED_UNITS	NBPF
93 #endif
94 
95 #if NBPF > 0
96 
97 /*
98  * The default read buffer size is patchable.
99  */
100 static int bpf_bufsize = BPF_DEFAULTBUFSIZE;
101 SYSCTL_INT(_debug, OID_AUTO, bpf_bufsize, CTLFLAG_RW,
102    &bpf_bufsize, 0, "Current size of bpf buffer");
103 int bpf_maxbufsize = BPF_MAXBUFSIZE;
104 SYSCTL_INT(_debug, OID_AUTO, bpf_maxbufsize, CTLFLAG_RW,
105    &bpf_maxbufsize, 0, "Maximum size of bpf buffer");
106 
107 /*
108  *  bpf_iflist is the list of interfaces; each corresponds to an ifnet
109  */
110 static struct bpf_if	*bpf_iflist;
111 
112 static struct lwkt_token bpf_token = LWKT_TOKEN_INITIALIZER(bpf_token);
113 
114 static int	bpf_allocbufs(struct bpf_d *);
115 static void	bpf_attachd(struct bpf_d *d, struct bpf_if *bp);
116 static void	bpf_detachd(struct bpf_d *d);
117 static void	bpf_resetd(struct bpf_d *);
118 static void	bpf_freed(struct bpf_d *);
119 static void	bpf_mcopy(const void *, void *, size_t);
120 static int	bpf_movein(struct uio *, int, struct mbuf **,
121 			   struct sockaddr *, int *, struct bpf_insn *);
122 static int	bpf_setif(struct bpf_d *, struct ifreq *);
123 static void	bpf_timed_out(void *);
124 static void	bpf_wakeup(struct bpf_d *);
125 static void	catchpacket(struct bpf_d *, u_char *, u_int, u_int,
126 			    void (*)(const void *, void *, size_t),
127 			    const struct timeval *);
128 static int	bpf_setf(struct bpf_d *, struct bpf_program *, u_long cmd);
129 static int	bpf_getdltlist(struct bpf_d *, struct bpf_dltlist *);
130 static int	bpf_setdlt(struct bpf_d *, u_int);
131 static void	bpf_drvinit(void *unused);
132 static void	bpf_filter_detach(struct knote *kn);
133 static int	bpf_filter_read(struct knote *kn, long hint);
134 
135 static d_open_t		bpfopen;
136 static d_clone_t	bpfclone;
137 static d_close_t	bpfclose;
138 static d_read_t		bpfread;
139 static d_write_t	bpfwrite;
140 static d_ioctl_t	bpfioctl;
141 static d_kqfilter_t	bpfkqfilter;
142 
143 #define CDEV_MAJOR 23
144 static struct dev_ops bpf_ops = {
145 	{ "bpf", 0, D_MPSAFE },
146 	.d_open =	bpfopen,
147 	.d_close =	bpfclose,
148 	.d_read =	bpfread,
149 	.d_write =	bpfwrite,
150 	.d_ioctl =	bpfioctl,
151 	.d_kqfilter =	bpfkqfilter
152 };
153 
154 
155 static int
156 bpf_movein(struct uio *uio, int linktype, struct mbuf **mp,
157 	   struct sockaddr *sockp, int *datlen, struct bpf_insn *wfilter)
158 {
159 	struct mbuf *m;
160 	int error;
161 	int len;
162 	int hlen;
163 	int slen;
164 
165 	*datlen = 0;
166 	*mp = NULL;
167 
168 	/*
169 	 * Build a sockaddr based on the data link layer type.
170 	 * We do this at this level because the ethernet header
171 	 * is copied directly into the data field of the sockaddr.
172 	 * In the case of SLIP, there is no header and the packet
173 	 * is forwarded as is.
174 	 * Also, we are careful to leave room at the front of the mbuf
175 	 * for the link level header.
176 	 */
177 	switch (linktype) {
178 	case DLT_SLIP:
179 		sockp->sa_family = AF_INET;
180 		hlen = 0;
181 		break;
182 
183 	case DLT_EN10MB:
184 		sockp->sa_family = AF_UNSPEC;
185 		/* XXX Would MAXLINKHDR be better? */
186 		hlen = sizeof(struct ether_header);
187 		break;
188 
189 	case DLT_RAW:
190 	case DLT_NULL:
191 		sockp->sa_family = AF_UNSPEC;
192 		hlen = 0;
193 		break;
194 
195 	case DLT_ATM_RFC1483:
196 		/*
197 		 * en atm driver requires 4-byte atm pseudo header.
198 		 * though it isn't standard, vpi:vci needs to be
199 		 * specified anyway.
200 		 */
201 		sockp->sa_family = AF_UNSPEC;
202 		hlen = 12;	/* XXX 4(ATM_PH) + 3(LLC) + 5(SNAP) */
203 		break;
204 
205 	case DLT_PPP:
206 		sockp->sa_family = AF_UNSPEC;
207 		hlen = 4;	/* This should match PPP_HDRLEN */
208 		break;
209 
210 	default:
211 		return(EIO);
212 	}
213 
214 	len = uio->uio_resid;
215 	*datlen = len - hlen;
216 	if ((unsigned)len > MCLBYTES)
217 		return(EIO);
218 
219 	m = m_getl(len, MB_WAIT, MT_DATA, M_PKTHDR, NULL);
220 	if (m == NULL)
221 		return(ENOBUFS);
222 	m->m_pkthdr.len = m->m_len = len;
223 	m->m_pkthdr.rcvif = NULL;
224 	*mp = m;
225 
226 	if (m->m_len < hlen) {
227 		error = EPERM;
228 		goto bad;
229 	}
230 
231 	error = uiomove(mtod(m, u_char *), len, uio);
232 	if (error)
233 		goto bad;
234 
235 	slen = bpf_filter(wfilter, mtod(m, u_char *), len, len);
236 	if (slen == 0) {
237 		error = EPERM;
238 		goto bad;
239 	}
240 
241 	/*
242 	 * Make room for link header, and copy it to sockaddr.
243 	 */
244 	if (hlen != 0) {
245 		bcopy(m->m_data, sockp->sa_data, hlen);
246 		m->m_pkthdr.len -= hlen;
247 		m->m_len -= hlen;
248 		m->m_data += hlen; /* XXX */
249 	}
250 	return (0);
251 bad:
252 	m_freem(m);
253 	return(error);
254 }
255 
256 /*
257  * Attach file to the bpf interface, i.e. make d listen on bp.
258  * Must be called at splimp.
259  */
260 static void
261 bpf_attachd(struct bpf_d *d, struct bpf_if *bp)
262 {
263 	/*
264 	 * Point d at bp, and add d to the interface's list of listeners.
265 	 * Finally, point the driver's bpf cookie at the interface so
266 	 * it will divert packets to bpf.
267 	 */
268 	lwkt_gettoken(&bpf_token);
269 	d->bd_bif = bp;
270 	SLIST_INSERT_HEAD(&bp->bif_dlist, d, bd_next);
271 	*bp->bif_driverp = bp;
272 
273 	EVENTHANDLER_INVOKE(bpf_track, bp->bif_ifp, bp->bif_dlt, 1);
274 	lwkt_reltoken(&bpf_token);
275 }
276 
277 /*
278  * Detach a file from its interface.
279  */
280 static void
281 bpf_detachd(struct bpf_d *d)
282 {
283 	int error;
284 	struct bpf_if *bp;
285 	struct ifnet *ifp;
286 
287 	lwkt_gettoken(&bpf_token);
288 	bp = d->bd_bif;
289 	ifp = bp->bif_ifp;
290 
291 	/* Remove d from the interface's descriptor list. */
292 	SLIST_REMOVE(&bp->bif_dlist, d, bpf_d, bd_next);
293 
294 	if (SLIST_EMPTY(&bp->bif_dlist)) {
295 		/*
296 		 * Let the driver know that there are no more listeners.
297 		 */
298 		*bp->bif_driverp = NULL;
299 	}
300 	d->bd_bif = NULL;
301 
302 	EVENTHANDLER_INVOKE(bpf_track, ifp, bp->bif_dlt, 0);
303 
304 	/*
305 	 * Check if this descriptor had requested promiscuous mode.
306 	 * If so, turn it off.
307 	 */
308 	if (d->bd_promisc) {
309 		d->bd_promisc = 0;
310 		error = ifpromisc(ifp, 0);
311 		if (error != 0 && error != ENXIO) {
312 			/*
313 			 * ENXIO can happen if a pccard is unplugged,
314 			 * Something is really wrong if we were able to put
315 			 * the driver into promiscuous mode, but can't
316 			 * take it out.
317 			 */
318 			if_printf(ifp, "bpf_detach: ifpromisc failed(%d)\n",
319 				  error);
320 		}
321 	}
322 	lwkt_reltoken(&bpf_token);
323 }
324 
325 /*
326  * Open ethernet device.  Returns ENXIO for illegal minor device number,
327  * EBUSY if file is open by another process.
328  */
329 /* ARGSUSED */
330 static int
331 bpfopen(struct dev_open_args *ap)
332 {
333 	cdev_t dev = ap->a_head.a_dev;
334 	struct bpf_d *d;
335 
336 	lwkt_gettoken(&bpf_token);
337 	if (ap->a_cred->cr_prison) {
338 		lwkt_reltoken(&bpf_token);
339 		return(EPERM);
340 	}
341 
342 	d = dev->si_drv1;
343 	/*
344 	 * Each minor can be opened by only one process.  If the requested
345 	 * minor is in use, return EBUSY.
346 	 */
347 	if (d != NULL) {
348 		lwkt_reltoken(&bpf_token);
349 		return(EBUSY);
350 	}
351 
352 	d = kmalloc(sizeof *d, M_BPF, M_WAITOK | M_ZERO);
353 	dev->si_drv1 = d;
354 	d->bd_bufsize = bpf_bufsize;
355 	d->bd_sig = SIGIO;
356 	d->bd_seesent = 1;
357 	callout_init(&d->bd_callout);
358 	lwkt_reltoken(&bpf_token);
359 
360 	return(0);
361 }
362 
363 static int
364 bpfclone(struct dev_clone_args *ap)
365 {
366 	int unit;
367 
368 	unit = devfs_clone_bitmap_get(&DEVFS_CLONE_BITMAP(bpf), 0);
369 	ap->a_dev = make_only_dev(&bpf_ops, unit, 0, 0, 0600, "bpf%d", unit);
370 
371 	return 0;
372 }
373 
374 /*
375  * Close the descriptor by detaching it from its interface,
376  * deallocating its buffers, and marking it free.
377  */
378 /* ARGSUSED */
379 static int
380 bpfclose(struct dev_close_args *ap)
381 {
382 	cdev_t dev = ap->a_head.a_dev;
383 	struct bpf_d *d = dev->si_drv1;
384 
385 	lwkt_gettoken(&bpf_token);
386 	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)
391 		bpf_detachd(d);
392 	bpf_freed(d);
393 	dev->si_drv1 = NULL;
394 	if (dev->si_uminor >= BPF_PREALLOCATED_UNITS) {
395 		devfs_clone_bitmap_put(&DEVFS_CLONE_BITMAP(bpf), dev->si_uminor);
396 		destroy_dev(dev);
397 	}
398 	kfree(d, M_BPF);
399 	lwkt_reltoken(&bpf_token);
400 
401 	return(0);
402 }
403 
404 /*
405  * Rotate the packet buffers in descriptor d.  Move the store buffer
406  * into the hold slot, and the free buffer into the store slot.
407  * Zero the length of the new store buffer.
408  */
409 #define ROTATE_BUFFERS(d) \
410 	(d)->bd_hbuf = (d)->bd_sbuf; \
411 	(d)->bd_hlen = (d)->bd_slen; \
412 	(d)->bd_sbuf = (d)->bd_fbuf; \
413 	(d)->bd_slen = 0; \
414 	(d)->bd_fbuf = NULL;
415 /*
416  *  bpfread - read next chunk of packets from buffers
417  */
418 static int
419 bpfread(struct dev_read_args *ap)
420 {
421 	cdev_t dev = ap->a_head.a_dev;
422 	struct bpf_d *d = dev->si_drv1;
423 	int timed_out;
424 	int error;
425 
426 	lwkt_gettoken(&bpf_token);
427 	/*
428 	 * Restrict application to use a buffer the same size as
429 	 * as kernel buffers.
430 	 */
431 	if (ap->a_uio->uio_resid != d->bd_bufsize) {
432 		lwkt_reltoken(&bpf_token);
433 		return(EINVAL);
434 	}
435 
436 	if (d->bd_state == BPF_WAITING)
437 		callout_stop(&d->bd_callout);
438 	timed_out = (d->bd_state == BPF_TIMED_OUT);
439 	d->bd_state = BPF_IDLE;
440 	/*
441 	 * If the hold buffer is empty, then do a timed sleep, which
442 	 * ends when the timeout expires or when enough packets
443 	 * have arrived to fill the store buffer.
444 	 */
445 	while (d->bd_hbuf == NULL) {
446 		if ((d->bd_immediate || (ap->a_ioflag & IO_NDELAY) || timed_out)
447 		    && d->bd_slen != 0) {
448 			/*
449 			 * A packet(s) either arrived since the previous,
450 			 * We're in immediate mode, or are reading
451 			 * in non-blocking mode, and a packet(s)
452 			 * either arrived since the previous
453 			 * read or arrived while we were asleep.
454 			 * Rotate the buffers and return what's here.
455 			 */
456 			ROTATE_BUFFERS(d);
457 			break;
458 		}
459 
460 		/*
461 		 * No data is available, check to see if the bpf device
462 		 * is still pointed at a real interface.  If not, return
463 		 * ENXIO so that the userland process knows to rebind
464 		 * it before using it again.
465 		 */
466 		if (d->bd_bif == NULL) {
467 			lwkt_reltoken(&bpf_token);
468 			return(ENXIO);
469 		}
470 
471 		if (ap->a_ioflag & IO_NDELAY) {
472 			lwkt_reltoken(&bpf_token);
473 			return(EWOULDBLOCK);
474 		}
475 		error = tsleep(d, PCATCH, "bpf", d->bd_rtout);
476 		if (error == EINTR || error == ERESTART) {
477 			lwkt_reltoken(&bpf_token);
478 			return(error);
479 		}
480 		if (error == EWOULDBLOCK) {
481 			/*
482 			 * On a timeout, return what's in the buffer,
483 			 * which may be nothing.  If there is something
484 			 * in the store buffer, we can rotate the buffers.
485 			 */
486 			if (d->bd_hbuf)
487 				/*
488 				 * We filled up the buffer in between
489 				 * getting the timeout and arriving
490 				 * here, so we don't need to rotate.
491 				 */
492 				break;
493 
494 			if (d->bd_slen == 0) {
495 				lwkt_reltoken(&bpf_token);
496 				return(0);
497 			}
498 			ROTATE_BUFFERS(d);
499 			break;
500 		}
501 	}
502 	/*
503 	 * At this point, we know we have something in the hold slot.
504 	 */
505 
506 	/*
507 	 * Move data from hold buffer into user space.
508 	 * We know the entire buffer is transferred since
509 	 * we checked above that the read buffer is bpf_bufsize bytes.
510 	 */
511 	error = uiomove(d->bd_hbuf, d->bd_hlen, ap->a_uio);
512 
513 	d->bd_fbuf = d->bd_hbuf;
514 	d->bd_hbuf = NULL;
515 	d->bd_hlen = 0;
516 	lwkt_reltoken(&bpf_token);
517 
518 	return(error);
519 }
520 
521 
522 /*
523  * If there are processes sleeping on this descriptor, wake them up.
524  */
525 static void
526 bpf_wakeup(struct bpf_d *d)
527 {
528 	if (d->bd_state == BPF_WAITING) {
529 		callout_stop(&d->bd_callout);
530 		d->bd_state = BPF_IDLE;
531 	}
532 	wakeup(d);
533 	if (d->bd_async && d->bd_sig && d->bd_sigio)
534 		pgsigio(d->bd_sigio, d->bd_sig, 0);
535 
536 	KNOTE(&d->bd_kq.ki_note, 0);
537 }
538 
539 static void
540 bpf_timed_out(void *arg)
541 {
542 	struct bpf_d *d = (struct bpf_d *)arg;
543 
544 	if (d->bd_state == BPF_WAITING) {
545 		d->bd_state = BPF_TIMED_OUT;
546 		if (d->bd_slen != 0)
547 			bpf_wakeup(d);
548 	}
549 }
550 
551 static void
552 bpf_output_dispatch(netmsg_t msg)
553 {
554 	struct netmsg_bpf_output *bmsg = (struct netmsg_bpf_output *)msg;
555 	struct ifnet *ifp = bmsg->nm_ifp;
556 	int error;
557 
558 	/*
559 	 * The driver frees the mbuf.
560 	 */
561 	error = ifp->if_output(ifp, bmsg->nm_mbuf, bmsg->nm_dst, NULL);
562 	lwkt_replymsg(&msg->lmsg, error);
563 }
564 
565 static int
566 bpfwrite(struct dev_write_args *ap)
567 {
568 	cdev_t dev = ap->a_head.a_dev;
569 	struct bpf_d *d = dev->si_drv1;
570 	struct ifnet *ifp;
571 	struct mbuf *m;
572 	int error, ret;
573 	struct sockaddr dst;
574 	int datlen;
575 	struct netmsg_bpf_output bmsg;
576 
577 	lwkt_gettoken(&bpf_token);
578 	if (d->bd_bif == NULL) {
579 		lwkt_reltoken(&bpf_token);
580 		return(ENXIO);
581 	}
582 
583 	ifp = d->bd_bif->bif_ifp;
584 
585 	if (ap->a_uio->uio_resid == 0) {
586 		lwkt_reltoken(&bpf_token);
587 		return(0);
588 	}
589 
590 	error = bpf_movein(ap->a_uio, (int)d->bd_bif->bif_dlt, &m,
591 			   &dst, &datlen, d->bd_wfilter);
592 	if (error) {
593 		lwkt_reltoken(&bpf_token);
594 		return(error);
595 	}
596 
597 	if (datlen > ifp->if_mtu) {
598 		m_freem(m);
599 		lwkt_reltoken(&bpf_token);
600 		return(EMSGSIZE);
601 	}
602 
603 	if (d->bd_hdrcmplt)
604 		dst.sa_family = pseudo_AF_HDRCMPLT;
605 
606 	netmsg_init(&bmsg.base, NULL, &curthread->td_msgport,
607 		    0, bpf_output_dispatch);
608 	bmsg.nm_mbuf = m;
609 	bmsg.nm_ifp = ifp;
610 	bmsg.nm_dst = &dst;
611 
612 	ret = lwkt_domsg(netisr_cpuport(0), &bmsg.base.lmsg, 0);
613 	lwkt_reltoken(&bpf_token);
614 
615 	return ret;
616 }
617 
618 /*
619  * Reset a descriptor by flushing its packet buffer and clearing the
620  * receive and drop counts.  Should be called at splimp.
621  */
622 static void
623 bpf_resetd(struct bpf_d *d)
624 {
625 	if (d->bd_hbuf) {
626 		/* Free the hold buffer. */
627 		d->bd_fbuf = d->bd_hbuf;
628 		d->bd_hbuf = NULL;
629 	}
630 	d->bd_slen = 0;
631 	d->bd_hlen = 0;
632 	d->bd_rcount = 0;
633 	d->bd_dcount = 0;
634 }
635 
636 /*
637  *  FIONREAD		Check for read packet available.
638  *  SIOCGIFADDR		Get interface address - convenient hook to driver.
639  *  BIOCGBLEN		Get buffer len [for read()].
640  *  BIOCSETF		Set ethernet read filter.
641  *  BIOCSETWF		Set ethernet write filter.
642  *  BIOCFLUSH		Flush read packet buffer.
643  *  BIOCPROMISC		Put interface into promiscuous mode.
644  *  BIOCGDLT		Get link layer type.
645  *  BIOCGETIF		Get interface name.
646  *  BIOCSETIF		Set interface.
647  *  BIOCSRTIMEOUT	Set read timeout.
648  *  BIOCGRTIMEOUT	Get read timeout.
649  *  BIOCGSTATS		Get packet stats.
650  *  BIOCIMMEDIATE	Set immediate mode.
651  *  BIOCVERSION		Get filter language version.
652  *  BIOCGHDRCMPLT	Get "header already complete" flag
653  *  BIOCSHDRCMPLT	Set "header already complete" flag
654  *  BIOCGSEESENT	Get "see packets sent" flag
655  *  BIOCSSEESENT	Set "see packets sent" flag
656  *  BIOCLOCK		Set "locked" flag
657  */
658 /* ARGSUSED */
659 static int
660 bpfioctl(struct dev_ioctl_args *ap)
661 {
662 	cdev_t dev = ap->a_head.a_dev;
663 	struct bpf_d *d = dev->si_drv1;
664 	int error = 0;
665 
666 	lwkt_gettoken(&bpf_token);
667 	if (d->bd_state == BPF_WAITING)
668 		callout_stop(&d->bd_callout);
669 	d->bd_state = BPF_IDLE;
670 
671 	if (d->bd_locked == 1) {
672 		switch (ap->a_cmd) {
673 		case BIOCGBLEN:
674 		case BIOCFLUSH:
675 		case BIOCGDLT:
676 		case BIOCGDLTLIST:
677 		case BIOCGETIF:
678 		case BIOCGRTIMEOUT:
679 		case BIOCGSTATS:
680 		case BIOCVERSION:
681 		case BIOCGRSIG:
682 		case BIOCGHDRCMPLT:
683 		case FIONREAD:
684 		case BIOCLOCK:
685 		case BIOCSRTIMEOUT:
686 		case BIOCIMMEDIATE:
687 		case TIOCGPGRP:
688 			break;
689 		default:
690 			lwkt_reltoken(&bpf_token);
691 			return (EPERM);
692 		}
693 	}
694 	switch (ap->a_cmd) {
695 	default:
696 		error = EINVAL;
697 		break;
698 
699 	/*
700 	 * Check for read packet available.
701 	 */
702 	case FIONREAD:
703 		{
704 			int n;
705 
706 			n = d->bd_slen;
707 			if (d->bd_hbuf)
708 				n += d->bd_hlen;
709 
710 			*(int *)ap->a_data = n;
711 			break;
712 		}
713 
714 	case SIOCGIFADDR:
715 		{
716 			struct ifnet *ifp;
717 
718 			if (d->bd_bif == NULL) {
719 				error = EINVAL;
720 			} else {
721 				ifp = d->bd_bif->bif_ifp;
722 				ifnet_serialize_all(ifp);
723 				error = ifp->if_ioctl(ifp, ap->a_cmd,
724 						      ap->a_data, ap->a_cred);
725 				ifnet_deserialize_all(ifp);
726 			}
727 			break;
728 		}
729 
730 	/*
731 	 * Get buffer len [for read()].
732 	 */
733 	case BIOCGBLEN:
734 		*(u_int *)ap->a_data = d->bd_bufsize;
735 		break;
736 
737 	/*
738 	 * Set buffer length.
739 	 */
740 	case BIOCSBLEN:
741 		if (d->bd_bif != NULL) {
742 			error = EINVAL;
743 		} else {
744 			u_int size = *(u_int *)ap->a_data;
745 
746 			if (size > bpf_maxbufsize)
747 				*(u_int *)ap->a_data = size = bpf_maxbufsize;
748 			else if (size < BPF_MINBUFSIZE)
749 				*(u_int *)ap->a_data = size = BPF_MINBUFSIZE;
750 			d->bd_bufsize = size;
751 		}
752 		break;
753 
754 	/*
755 	 * Set link layer read filter.
756 	 */
757 	case BIOCSETF:
758 	case BIOCSETWF:
759 		error = bpf_setf(d, (struct bpf_program *)ap->a_data,
760 			ap->a_cmd);
761 		break;
762 
763 	/*
764 	 * Flush read packet buffer.
765 	 */
766 	case BIOCFLUSH:
767 		bpf_resetd(d);
768 		break;
769 
770 	/*
771 	 * Put interface into promiscuous mode.
772 	 */
773 	case BIOCPROMISC:
774 		if (d->bd_bif == NULL) {
775 			/*
776 			 * No interface attached yet.
777 			 */
778 			error = EINVAL;
779 			break;
780 		}
781 		if (d->bd_promisc == 0) {
782 			error = ifpromisc(d->bd_bif->bif_ifp, 1);
783 			if (error == 0)
784 				d->bd_promisc = 1;
785 		}
786 		break;
787 
788 	/*
789 	 * Get device parameters.
790 	 */
791 	case BIOCGDLT:
792 		if (d->bd_bif == NULL)
793 			error = EINVAL;
794 		else
795 			*(u_int *)ap->a_data = d->bd_bif->bif_dlt;
796 		break;
797 
798 	/*
799 	 * Get a list of supported data link types.
800 	 */
801 	case BIOCGDLTLIST:
802 		if (d->bd_bif == NULL) {
803 			error = EINVAL;
804 		} else {
805 			error = bpf_getdltlist(d,
806 				(struct bpf_dltlist *)ap->a_data);
807 		}
808 		break;
809 
810 	/*
811 	 * Set data link type.
812 	 */
813 	case BIOCSDLT:
814 		if (d->bd_bif == NULL)
815 			error = EINVAL;
816 		else
817 			error = bpf_setdlt(d, *(u_int *)ap->a_data);
818 		break;
819 
820 	/*
821 	 * Get interface name.
822 	 */
823 	case BIOCGETIF:
824 		if (d->bd_bif == NULL) {
825 			error = EINVAL;
826 		} else {
827 			struct ifnet *const ifp = d->bd_bif->bif_ifp;
828 			struct ifreq *const ifr = (struct ifreq *)ap->a_data;
829 
830 			strlcpy(ifr->ifr_name, ifp->if_xname,
831 				sizeof ifr->ifr_name);
832 		}
833 		break;
834 
835 	/*
836 	 * Set interface.
837 	 */
838 	case BIOCSETIF:
839 		error = bpf_setif(d, (struct ifreq *)ap->a_data);
840 		break;
841 
842 	/*
843 	 * Set read timeout.
844 	 */
845 	case BIOCSRTIMEOUT:
846 		{
847 			struct timeval *tv = (struct timeval *)ap->a_data;
848 
849 			/*
850 			 * Subtract 1 tick from tvtohz() since this isn't
851 			 * a one-shot timer.
852 			 */
853 			if ((error = itimerfix(tv)) == 0)
854 				d->bd_rtout = tvtohz_low(tv);
855 			break;
856 		}
857 
858 	/*
859 	 * Get read timeout.
860 	 */
861 	case BIOCGRTIMEOUT:
862 		{
863 			struct timeval *tv = (struct timeval *)ap->a_data;
864 
865 			tv->tv_sec = d->bd_rtout / hz;
866 			tv->tv_usec = (d->bd_rtout % hz) * ustick;
867 			break;
868 		}
869 
870 	/*
871 	 * Get packet stats.
872 	 */
873 	case BIOCGSTATS:
874 		{
875 			struct bpf_stat *bs = (struct bpf_stat *)ap->a_data;
876 
877 			bs->bs_recv = d->bd_rcount;
878 			bs->bs_drop = d->bd_dcount;
879 			break;
880 		}
881 
882 	/*
883 	 * Set immediate mode.
884 	 */
885 	case BIOCIMMEDIATE:
886 		d->bd_immediate = *(u_int *)ap->a_data;
887 		break;
888 
889 	case BIOCVERSION:
890 		{
891 			struct bpf_version *bv = (struct bpf_version *)ap->a_data;
892 
893 			bv->bv_major = BPF_MAJOR_VERSION;
894 			bv->bv_minor = BPF_MINOR_VERSION;
895 			break;
896 		}
897 
898 	/*
899 	 * Get "header already complete" flag
900 	 */
901 	case BIOCGHDRCMPLT:
902 		*(u_int *)ap->a_data = d->bd_hdrcmplt;
903 		break;
904 
905 	/*
906 	 * Set "header already complete" flag
907 	 */
908 	case BIOCSHDRCMPLT:
909 		d->bd_hdrcmplt = *(u_int *)ap->a_data ? 1 : 0;
910 		break;
911 
912 	/*
913 	 * Get "see sent packets" flag
914 	 */
915 	case BIOCGSEESENT:
916 		*(u_int *)ap->a_data = d->bd_seesent;
917 		break;
918 
919 	/*
920 	 * Set "see sent packets" flag
921 	 */
922 	case BIOCSSEESENT:
923 		d->bd_seesent = *(u_int *)ap->a_data;
924 		break;
925 
926 	case FIOASYNC:		/* Send signal on receive packets */
927 		d->bd_async = *(int *)ap->a_data;
928 		break;
929 
930 	case FIOSETOWN:
931 		error = fsetown(*(int *)ap->a_data, &d->bd_sigio);
932 		break;
933 
934 	case FIOGETOWN:
935 		*(int *)ap->a_data = fgetown(&d->bd_sigio);
936 		break;
937 
938 	/* This is deprecated, FIOSETOWN should be used instead. */
939 	case TIOCSPGRP:
940 		error = fsetown(-(*(int *)ap->a_data), &d->bd_sigio);
941 		break;
942 
943 	/* This is deprecated, FIOGETOWN should be used instead. */
944 	case TIOCGPGRP:
945 		*(int *)ap->a_data = -fgetown(&d->bd_sigio);
946 		break;
947 
948 	case BIOCSRSIG:		/* Set receive signal */
949 		{
950 			u_int sig;
951 
952 			sig = *(u_int *)ap->a_data;
953 
954 			if (sig >= NSIG)
955 				error = EINVAL;
956 			else
957 				d->bd_sig = sig;
958 			break;
959 		}
960 	case BIOCGRSIG:
961 		*(u_int *)ap->a_data = d->bd_sig;
962 		break;
963 	case BIOCLOCK:
964 		d->bd_locked = 1;
965 		break;
966 	}
967 	lwkt_reltoken(&bpf_token);
968 
969 	return(error);
970 }
971 
972 /*
973  * Set d's packet filter program to fp.  If this file already has a filter,
974  * free it and replace it.  Returns EINVAL for bogus requests.
975  */
976 static int
977 bpf_setf(struct bpf_d *d, struct bpf_program *fp, u_long cmd)
978 {
979 	struct bpf_insn *fcode, *old;
980 	u_int wfilter, flen, size;
981 
982 	if (cmd == BIOCSETWF) {
983 		old = d->bd_wfilter;
984 		wfilter = 1;
985 	} else {
986 		wfilter = 0;
987 		old = d->bd_rfilter;
988 	}
989 	if (fp->bf_insns == NULL) {
990 		if (fp->bf_len != 0)
991 			return(EINVAL);
992 		if (wfilter)
993 			d->bd_wfilter = NULL;
994 		else
995 			d->bd_rfilter = NULL;
996 		bpf_resetd(d);
997 		if (old != NULL)
998 			kfree(old, M_BPF);
999 		return(0);
1000 	}
1001 	flen = fp->bf_len;
1002 	if (flen > BPF_MAXINSNS)
1003 		return(EINVAL);
1004 
1005 	size = flen * sizeof *fp->bf_insns;
1006 	fcode = (struct bpf_insn *)kmalloc(size, M_BPF, M_WAITOK);
1007 	if (copyin(fp->bf_insns, fcode, size) == 0 &&
1008 	    bpf_validate(fcode, (int)flen)) {
1009 		if (wfilter)
1010 			d->bd_wfilter = fcode;
1011 		else
1012 			d->bd_rfilter = fcode;
1013 		bpf_resetd(d);
1014 		if (old != NULL)
1015 			kfree(old, M_BPF);
1016 
1017 		return(0);
1018 	}
1019 	kfree(fcode, M_BPF);
1020 	return(EINVAL);
1021 }
1022 
1023 /*
1024  * Detach a file from its current interface (if attached at all) and attach
1025  * to the interface indicated by the name stored in ifr.
1026  * Return an errno or 0.
1027  */
1028 static int
1029 bpf_setif(struct bpf_d *d, struct ifreq *ifr)
1030 {
1031 	struct bpf_if *bp;
1032 	int error;
1033 	struct ifnet *theywant;
1034 
1035 	theywant = ifunit(ifr->ifr_name);
1036 	if (theywant == NULL)
1037 		return(ENXIO);
1038 
1039 	/*
1040 	 * Look through attached interfaces for the named one.
1041 	 */
1042 	for (bp = bpf_iflist; bp != NULL; bp = bp->bif_next) {
1043 		struct ifnet *ifp = bp->bif_ifp;
1044 
1045 		if (ifp == NULL || ifp != theywant)
1046 			continue;
1047 		/* skip additional entry */
1048 		if (bp->bif_driverp != &ifp->if_bpf)
1049 			continue;
1050 		/*
1051 		 * We found the requested interface.
1052 		 * Allocate the packet buffers if we need to.
1053 		 * If we're already attached to requested interface,
1054 		 * just flush the buffer.
1055 		 */
1056 		if (d->bd_sbuf == NULL) {
1057 			error = bpf_allocbufs(d);
1058 			if (error != 0)
1059 				return(error);
1060 		}
1061 		if (bp != d->bd_bif) {
1062 			if (d->bd_bif != NULL) {
1063 				/*
1064 				 * Detach if attached to something else.
1065 				 */
1066 				bpf_detachd(d);
1067 			}
1068 
1069 			bpf_attachd(d, bp);
1070 		}
1071 		bpf_resetd(d);
1072 		return(0);
1073 	}
1074 
1075 	/* Not found. */
1076 	return(ENXIO);
1077 }
1078 
1079 static struct filterops bpf_read_filtops =
1080 	{ FILTEROP_ISFD, NULL, bpf_filter_detach, bpf_filter_read };
1081 
1082 static int
1083 bpfkqfilter(struct dev_kqfilter_args *ap)
1084 {
1085 	cdev_t dev = ap->a_head.a_dev;
1086 	struct knote *kn = ap->a_kn;
1087 	struct klist *klist;
1088 	struct bpf_d *d;
1089 
1090 	lwkt_gettoken(&bpf_token);
1091 	d = dev->si_drv1;
1092 	if (d->bd_bif == NULL) {
1093 		ap->a_result = 1;
1094 		lwkt_reltoken(&bpf_token);
1095 		return (0);
1096 	}
1097 
1098 	ap->a_result = 0;
1099 	switch (kn->kn_filter) {
1100 	case EVFILT_READ:
1101 		kn->kn_fop = &bpf_read_filtops;
1102 		kn->kn_hook = (caddr_t)d;
1103 		break;
1104 	default:
1105 		ap->a_result = EOPNOTSUPP;
1106 		lwkt_reltoken(&bpf_token);
1107 		return (0);
1108 	}
1109 
1110 	klist = &d->bd_kq.ki_note;
1111 	knote_insert(klist, kn);
1112 	lwkt_reltoken(&bpf_token);
1113 
1114 	return (0);
1115 }
1116 
1117 static void
1118 bpf_filter_detach(struct knote *kn)
1119 {
1120 	struct klist *klist;
1121 	struct bpf_d *d;
1122 
1123 	d = (struct bpf_d *)kn->kn_hook;
1124 	klist = &d->bd_kq.ki_note;
1125 	knote_remove(klist, kn);
1126 }
1127 
1128 static int
1129 bpf_filter_read(struct knote *kn, long hint)
1130 {
1131 	struct bpf_d *d;
1132 	int ready = 0;
1133 
1134 	d = (struct bpf_d *)kn->kn_hook;
1135 	if (d->bd_hlen != 0 ||
1136 	    ((d->bd_immediate || d->bd_state == BPF_TIMED_OUT) &&
1137 	    d->bd_slen != 0)) {
1138 		ready = 1;
1139 	} else {
1140 		/* Start the read timeout if necessary. */
1141 		if (d->bd_rtout > 0 && d->bd_state == BPF_IDLE) {
1142 			callout_reset(&d->bd_callout, d->bd_rtout,
1143 			    bpf_timed_out, d);
1144 			d->bd_state = BPF_WAITING;
1145 		}
1146 	}
1147 
1148 	return (ready);
1149 }
1150 
1151 
1152 /*
1153  * Process the packet pkt of length pktlen.  The packet is parsed
1154  * by each listener's filter, and if accepted, stashed into the
1155  * corresponding buffer.
1156  */
1157 void
1158 bpf_tap(struct bpf_if *bp, u_char *pkt, u_int pktlen)
1159 {
1160 	struct bpf_d *d;
1161 	struct timeval tv;
1162 	int gottime = 0;
1163 	u_int slen;
1164 
1165 	lwkt_gettoken(&bpf_token);
1166 	/* Re-check */
1167 	if (bp == NULL) {
1168 		lwkt_reltoken(&bpf_token);
1169 		return;
1170 	}
1171 
1172 	/*
1173 	 * Note that the ipl does not have to be raised at this point.
1174 	 * The only problem that could arise here is that if two different
1175 	 * interfaces shared any data.  This is not the case.
1176 	 */
1177 	SLIST_FOREACH(d, &bp->bif_dlist, bd_next) {
1178 		++d->bd_rcount;
1179 		slen = bpf_filter(d->bd_rfilter, pkt, pktlen, pktlen);
1180 		if (slen != 0) {
1181 			if (!gottime) {
1182 				microtime(&tv);
1183 				gottime = 1;
1184 			}
1185 			catchpacket(d, pkt, pktlen, slen, ovbcopy, &tv);
1186 		}
1187 	}
1188 	lwkt_reltoken(&bpf_token);
1189 }
1190 
1191 /*
1192  * Copy data from an mbuf chain into a buffer.  This code is derived
1193  * from m_copydata in sys/uipc_mbuf.c.
1194  */
1195 static void
1196 bpf_mcopy(const void *src_arg, void *dst_arg, size_t len)
1197 {
1198 	const struct mbuf *m;
1199 	u_int count;
1200 	u_char *dst;
1201 
1202 	m = src_arg;
1203 	dst = dst_arg;
1204 	while (len > 0) {
1205 		if (m == NULL)
1206 			panic("bpf_mcopy");
1207 		count = min(m->m_len, len);
1208 		bcopy(mtod(m, void *), dst, count);
1209 		m = m->m_next;
1210 		dst += count;
1211 		len -= count;
1212 	}
1213 }
1214 
1215 /*
1216  * Process the packet in the mbuf chain m.  The packet is parsed by each
1217  * listener's filter, and if accepted, stashed into the corresponding
1218  * buffer.
1219  */
1220 void
1221 bpf_mtap(struct bpf_if *bp, struct mbuf *m)
1222 {
1223 	struct bpf_d *d;
1224 	u_int pktlen, slen;
1225 	struct timeval tv;
1226 	int gottime = 0;
1227 
1228 	lwkt_gettoken(&bpf_token);
1229 	/* Re-check */
1230 	if (bp == NULL) {
1231 		lwkt_reltoken(&bpf_token);
1232 		return;
1233 	}
1234 
1235 	/* Don't compute pktlen, if no descriptor is attached. */
1236 	if (SLIST_EMPTY(&bp->bif_dlist)) {
1237 		lwkt_reltoken(&bpf_token);
1238 		return;
1239 	}
1240 
1241 	pktlen = m_lengthm(m, NULL);
1242 
1243 	SLIST_FOREACH(d, &bp->bif_dlist, bd_next) {
1244 		if (!d->bd_seesent && (m->m_pkthdr.rcvif == NULL))
1245 			continue;
1246 		++d->bd_rcount;
1247 		slen = bpf_filter(d->bd_rfilter, (u_char *)m, pktlen, 0);
1248 		if (slen != 0) {
1249 			if (!gottime) {
1250 				microtime(&tv);
1251 				gottime = 1;
1252 			}
1253 			catchpacket(d, (u_char *)m, pktlen, slen, bpf_mcopy,
1254 				    &tv);
1255 		}
1256 	}
1257 	lwkt_reltoken(&bpf_token);
1258 }
1259 
1260 /*
1261  * Incoming linkage from device drivers, where we have a mbuf chain
1262  * but need to prepend some arbitrary header from a linear buffer.
1263  *
1264  * Con up a minimal dummy header to pacify bpf.  Allocate (only) a
1265  * struct m_hdr on the stack.  This is safe as bpf only reads from the
1266  * fields in this header that we initialize, and will not try to free
1267  * it or keep a pointer to it.
1268  */
1269 void
1270 bpf_mtap_hdr(struct bpf_if *arg, caddr_t data, u_int dlen, struct mbuf *m, u_int direction)
1271 {
1272 	struct m_hdr mh;
1273 
1274 	mh.mh_flags = 0;
1275 	mh.mh_next = m;
1276 	mh.mh_len = dlen;
1277 	mh.mh_data = data;
1278 
1279 	return bpf_mtap(arg, (struct mbuf *) &mh);
1280 }
1281 
1282 void
1283 bpf_mtap_family(struct bpf_if *bp, struct mbuf *m, sa_family_t family)
1284 {
1285 	u_int family4;
1286 
1287 	KKASSERT(family != AF_UNSPEC);
1288 
1289 	family4 = (u_int)family;
1290 	bpf_ptap(bp, m, &family4, sizeof(family4));
1291 }
1292 
1293 /*
1294  * Process the packet in the mbuf chain m with the header in m prepended.
1295  * The packet is parsed by each listener's filter, and if accepted,
1296  * stashed into the corresponding buffer.
1297  */
1298 void
1299 bpf_ptap(struct bpf_if *bp, struct mbuf *m, const void *data, u_int dlen)
1300 {
1301 	struct mbuf mb;
1302 
1303 	/*
1304 	 * Craft on-stack mbuf suitable for passing to bpf_mtap.
1305 	 * Note that we cut corners here; we only setup what's
1306 	 * absolutely needed--this mbuf should never go anywhere else.
1307 	 */
1308 	mb.m_next = m;
1309 	mb.m_data = __DECONST(void *, data); /* LINTED */
1310 	mb.m_len = dlen;
1311 	mb.m_pkthdr.rcvif = m->m_pkthdr.rcvif;
1312 
1313 	bpf_mtap(bp, &mb);
1314 }
1315 
1316 /*
1317  * Move the packet data from interface memory (pkt) into the
1318  * store buffer.  Return 1 if it's time to wakeup a listener (buffer full),
1319  * otherwise 0.  "copy" is the routine called to do the actual data
1320  * transfer.  bcopy is passed in to copy contiguous chunks, while
1321  * bpf_mcopy is passed in to copy mbuf chains.  In the latter case,
1322  * pkt is really an mbuf.
1323  */
1324 static void
1325 catchpacket(struct bpf_d *d, u_char *pkt, u_int pktlen, u_int snaplen,
1326 	    void (*cpfn)(const void *, void *, size_t),
1327 	    const struct timeval *tv)
1328 {
1329 	struct bpf_hdr *hp;
1330 	int totlen, curlen;
1331 	int hdrlen = d->bd_bif->bif_hdrlen;
1332 	int wakeup = 0;
1333 	/*
1334 	 * Figure out how many bytes to move.  If the packet is
1335 	 * greater or equal to the snapshot length, transfer that
1336 	 * much.  Otherwise, transfer the whole packet (unless
1337 	 * we hit the buffer size limit).
1338 	 */
1339 	totlen = hdrlen + min(snaplen, pktlen);
1340 	if (totlen > d->bd_bufsize)
1341 		totlen = d->bd_bufsize;
1342 
1343 	/*
1344 	 * Round up the end of the previous packet to the next longword.
1345 	 */
1346 	curlen = BPF_WORDALIGN(d->bd_slen);
1347 	if (curlen + totlen > d->bd_bufsize) {
1348 		/*
1349 		 * This packet will overflow the storage buffer.
1350 		 * Rotate the buffers if we can, then wakeup any
1351 		 * pending reads.
1352 		 */
1353 		if (d->bd_fbuf == NULL) {
1354 			/*
1355 			 * We haven't completed the previous read yet,
1356 			 * so drop the packet.
1357 			 */
1358 			++d->bd_dcount;
1359 			return;
1360 		}
1361 		ROTATE_BUFFERS(d);
1362 		wakeup = 1;
1363 		curlen = 0;
1364 	} else if (d->bd_immediate || d->bd_state == BPF_TIMED_OUT) {
1365 		/*
1366 		 * Immediate mode is set, or the read timeout has
1367 		 * already expired during a select call.  A packet
1368 		 * arrived, so the reader should be woken up.
1369 		 */
1370 		wakeup = 1;
1371 	}
1372 
1373 	/*
1374 	 * Append the bpf header.
1375 	 */
1376 	hp = (struct bpf_hdr *)(d->bd_sbuf + curlen);
1377 	hp->bh_tstamp = *tv;
1378 	hp->bh_datalen = pktlen;
1379 	hp->bh_hdrlen = hdrlen;
1380 	/*
1381 	 * Copy the packet data into the store buffer and update its length.
1382 	 */
1383 	(*cpfn)(pkt, (u_char *)hp + hdrlen, (hp->bh_caplen = totlen - hdrlen));
1384 	d->bd_slen = curlen + totlen;
1385 
1386 	if (wakeup)
1387 		bpf_wakeup(d);
1388 }
1389 
1390 /*
1391  * Initialize all nonzero fields of a descriptor.
1392  */
1393 static int
1394 bpf_allocbufs(struct bpf_d *d)
1395 {
1396 	d->bd_fbuf = kmalloc(d->bd_bufsize, M_BPF, M_WAITOK);
1397 	d->bd_sbuf = kmalloc(d->bd_bufsize, M_BPF, M_WAITOK);
1398 	d->bd_slen = 0;
1399 	d->bd_hlen = 0;
1400 	return(0);
1401 }
1402 
1403 /*
1404  * Free buffers and packet filter program currently in use by a descriptor.
1405  * Called on close.
1406  */
1407 static void
1408 bpf_freed(struct bpf_d *d)
1409 {
1410 	/*
1411 	 * We don't need to lock out interrupts since this descriptor has
1412 	 * been detached from its interface and it yet hasn't been marked
1413 	 * free.
1414 	 */
1415 	if (d->bd_sbuf != NULL) {
1416 		kfree(d->bd_sbuf, M_BPF);
1417 		if (d->bd_hbuf != NULL)
1418 			kfree(d->bd_hbuf, M_BPF);
1419 		if (d->bd_fbuf != NULL)
1420 			kfree(d->bd_fbuf, M_BPF);
1421 	}
1422 	if (d->bd_rfilter)
1423 		kfree(d->bd_rfilter, M_BPF);
1424 	if (d->bd_wfilter)
1425 		kfree(d->bd_wfilter, M_BPF);
1426 }
1427 
1428 /*
1429  * Attach an interface to bpf.  ifp is a pointer to the structure
1430  * defining the interface to be attached, dlt is the link layer type,
1431  * and hdrlen is the fixed size of the link header (variable length
1432  * headers are not yet supported).
1433  */
1434 void
1435 bpfattach(struct ifnet *ifp, u_int dlt, u_int hdrlen)
1436 {
1437 	bpfattach_dlt(ifp, dlt, hdrlen, &ifp->if_bpf);
1438 }
1439 
1440 void
1441 bpfattach_dlt(struct ifnet *ifp, u_int dlt, u_int hdrlen, struct bpf_if **driverp)
1442 {
1443 	struct bpf_if *bp;
1444 
1445 	bp = kmalloc(sizeof *bp, M_BPF, M_WAITOK | M_ZERO);
1446 
1447 	lwkt_gettoken(&bpf_token);
1448 
1449 	SLIST_INIT(&bp->bif_dlist);
1450 	bp->bif_ifp = ifp;
1451 	bp->bif_dlt = dlt;
1452 	bp->bif_driverp = driverp;
1453 	*bp->bif_driverp = NULL;
1454 
1455 	bp->bif_next = bpf_iflist;
1456 	bpf_iflist = bp;
1457 
1458 	/*
1459 	 * Compute the length of the bpf header.  This is not necessarily
1460 	 * equal to SIZEOF_BPF_HDR because we want to insert spacing such
1461 	 * that the network layer header begins on a longword boundary (for
1462 	 * performance reasons and to alleviate alignment restrictions).
1463 	 */
1464 	bp->bif_hdrlen = BPF_WORDALIGN(hdrlen + SIZEOF_BPF_HDR) - hdrlen;
1465 
1466 	lwkt_reltoken(&bpf_token);
1467 
1468 	if (bootverbose)
1469 		if_printf(ifp, "bpf attached\n");
1470 }
1471 
1472 /*
1473  * Detach bpf from an interface.  This involves detaching each descriptor
1474  * associated with the interface, and leaving bd_bif NULL.  Notify each
1475  * descriptor as it's detached so that any sleepers wake up and get
1476  * ENXIO.
1477  */
1478 void
1479 bpfdetach(struct ifnet *ifp)
1480 {
1481 	struct bpf_if *bp, *bp_prev;
1482 	struct bpf_d *d;
1483 
1484 	lwkt_gettoken(&bpf_token);
1485 
1486 	/* Locate BPF interface information */
1487 	bp_prev = NULL;
1488 	for (bp = bpf_iflist; bp != NULL; bp = bp->bif_next) {
1489 		if (ifp == bp->bif_ifp)
1490 			break;
1491 		bp_prev = bp;
1492 	}
1493 
1494 	/* Interface wasn't attached */
1495 	if (bp->bif_ifp == NULL) {
1496 		lwkt_reltoken(&bpf_token);
1497 		kprintf("bpfdetach: %s was not attached\n", ifp->if_xname);
1498 		return;
1499 	}
1500 
1501 	while ((d = SLIST_FIRST(&bp->bif_dlist)) != NULL) {
1502 		bpf_detachd(d);
1503 		bpf_wakeup(d);
1504 	}
1505 
1506 	if (bp_prev != NULL)
1507 		bp_prev->bif_next = bp->bif_next;
1508 	else
1509 		bpf_iflist = bp->bif_next;
1510 
1511 	kfree(bp, M_BPF);
1512 
1513 	lwkt_reltoken(&bpf_token);
1514 }
1515 
1516 /*
1517  * Get a list of available data link type of the interface.
1518  */
1519 static int
1520 bpf_getdltlist(struct bpf_d *d, struct bpf_dltlist *bfl)
1521 {
1522 	int n, error;
1523 	struct ifnet *ifp;
1524 	struct bpf_if *bp;
1525 
1526 	ifp = d->bd_bif->bif_ifp;
1527 	n = 0;
1528 	error = 0;
1529 	for (bp = bpf_iflist; bp != NULL; bp = bp->bif_next) {
1530 		if (bp->bif_ifp != ifp)
1531 			continue;
1532 		if (bfl->bfl_list != NULL) {
1533 			if (n >= bfl->bfl_len) {
1534 				return (ENOMEM);
1535 			}
1536 			error = copyout(&bp->bif_dlt,
1537 			    bfl->bfl_list + n, sizeof(u_int));
1538 		}
1539 		n++;
1540 	}
1541 	bfl->bfl_len = n;
1542 	return(error);
1543 }
1544 
1545 /*
1546  * Set the data link type of a BPF instance.
1547  */
1548 static int
1549 bpf_setdlt(struct bpf_d *d, u_int dlt)
1550 {
1551 	int error, opromisc;
1552 	struct ifnet *ifp;
1553 	struct bpf_if *bp;
1554 
1555 	if (d->bd_bif->bif_dlt == dlt)
1556 		return (0);
1557 	ifp = d->bd_bif->bif_ifp;
1558 	for (bp = bpf_iflist; bp != NULL; bp = bp->bif_next) {
1559 		if (bp->bif_ifp == ifp && bp->bif_dlt == dlt)
1560 			break;
1561 	}
1562 	if (bp != NULL) {
1563 		opromisc = d->bd_promisc;
1564 		bpf_detachd(d);
1565 		bpf_attachd(d, bp);
1566 		bpf_resetd(d);
1567 		if (opromisc) {
1568 			error = ifpromisc(bp->bif_ifp, 1);
1569 			if (error) {
1570 				if_printf(bp->bif_ifp,
1571 					"bpf_setdlt: ifpromisc failed (%d)\n",
1572 					error);
1573 			} else {
1574 				d->bd_promisc = 1;
1575 			}
1576 		}
1577 	}
1578 	return(bp == NULL ? EINVAL : 0);
1579 }
1580 
1581 void
1582 bpf_gettoken(void)
1583 {
1584 	lwkt_gettoken(&bpf_token);
1585 }
1586 
1587 void
1588 bpf_reltoken(void)
1589 {
1590 	lwkt_reltoken(&bpf_token);
1591 }
1592 
1593 static void
1594 bpf_drvinit(void *unused)
1595 {
1596 	int i;
1597 
1598 	make_autoclone_dev(&bpf_ops, &DEVFS_CLONE_BITMAP(bpf),
1599 		bpfclone, 0, 0, 0600, "bpf");
1600 	for (i = 0; i < BPF_PREALLOCATED_UNITS; i++) {
1601 		make_dev(&bpf_ops, i, 0, 0, 0600, "bpf%d", i);
1602 		devfs_clone_bitmap_set(&DEVFS_CLONE_BITMAP(bpf), i);
1603 	}
1604 }
1605 
1606 static void
1607 bpf_drvuninit(void *unused)
1608 {
1609 	devfs_clone_handler_del("bpf");
1610 	dev_ops_remove_all(&bpf_ops);
1611 	devfs_clone_bitmap_uninit(&DEVFS_CLONE_BITMAP(bpf));
1612 }
1613 
1614 SYSINIT(bpfdev,SI_SUB_DRIVERS,SI_ORDER_MIDDLE+CDEV_MAJOR,bpf_drvinit,NULL)
1615 SYSUNINIT(bpfdev, SI_SUB_DRIVERS,SI_ORDER_MIDDLE+CDEV_MAJOR,bpf_drvuninit, NULL);
1616 
1617 #else /* !BPF */
1618 /*
1619  * NOP stubs to allow bpf-using drivers to load and function.
1620  *
1621  * A 'better' implementation would allow the core bpf functionality
1622  * to be loaded at runtime.
1623  */
1624 
1625 void
1626 bpf_tap(struct bpf_if *bp, u_char *pkt, u_int pktlen)
1627 {
1628 }
1629 
1630 void
1631 bpf_mtap(struct bpf_if *bp, struct mbuf *m)
1632 {
1633 }
1634 
1635 void
1636 bpf_ptap(struct bpf_if *bp, struct mbuf *m, const void *data, u_int dlen)
1637 {
1638 }
1639 
1640 void
1641 bpfattach(struct ifnet *ifp, u_int dlt, u_int hdrlen)
1642 {
1643 }
1644 
1645 void
1646 bpfattach_dlt(struct ifnet *ifp, u_int dlt, u_int hdrlen, struct bpf_if **driverp)
1647 {
1648 }
1649 
1650 void
1651 bpfdetach(struct ifnet *ifp)
1652 {
1653 }
1654 
1655 u_int
1656 bpf_filter(const struct bpf_insn *pc, u_char *p, u_int wirelen, u_int buflen)
1657 {
1658 	return -1;	/* "no filter" behaviour */
1659 }
1660 
1661 void
1662 bpf_gettoken(void)
1663 {
1664 }
1665 
1666 void
1667 bpf_reltoken(void)
1668 {
1669 }
1670 
1671 #endif /* !BPF */
1672