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 * 4. Neither the name of the University nor the names of its contributors
19 * may be used to endorse or promote products derived from this software
20 * without specific prior written permission.
22 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
23 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
24 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
25 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
26 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
27 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
28 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
29 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
30 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
31 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
34 * @(#)bpf.c 8.2 (Berkeley) 3/28/94
36 * $FreeBSD: src/sys/net/bpf.c,v 1.59.2.12 2002/04/14 21:41:48 luigi Exp $
41 #include <sys/param.h>
42 #include <sys/systm.h>
44 #include <sys/device.h>
45 #include <sys/malloc.h>
49 #include <sys/signalvar.h>
50 #include <sys/filio.h>
51 #include <sys/sockio.h>
52 #include <sys/ttycom.h>
53 #include <sys/filedesc.h>
55 #include <sys/event.h>
57 #include <sys/socket.h>
58 #include <sys/vnode.h>
60 #include <sys/thread2.h>
64 #include <net/bpfdesc.h>
65 #include <net/netmsg2.h>
66 #include <net/netisr2.h>
68 #include <netinet/in.h>
69 #include <netinet/if_ether.h>
70 #include <sys/kernel.h>
71 #include <sys/sysctl.h>
73 #include <sys/devfs.h>
75 struct netmsg_bpf_output
{
76 struct netmsg_base base
;
79 struct sockaddr
*nm_dst
;
82 MALLOC_DEFINE(M_BPF
, "BPF", "BPF data");
83 DEVFS_DECLARE_CLONE_BITMAP(bpf
);
86 #define BPF_PREALLOCATED_UNITS 4
88 #define BPF_PREALLOCATED_UNITS NBPF
94 * The default read buffer size is patchable.
96 static int bpf_bufsize
= BPF_DEFAULTBUFSIZE
;
97 SYSCTL_INT(_debug
, OID_AUTO
, bpf_bufsize
, CTLFLAG_RW
,
98 &bpf_bufsize
, 0, "Current size of bpf buffer");
99 int bpf_maxbufsize
= BPF_MAXBUFSIZE
;
100 SYSCTL_INT(_debug
, OID_AUTO
, bpf_maxbufsize
, CTLFLAG_RW
,
101 &bpf_maxbufsize
, 0, "Maximum size of bpf buffer");
104 * bpf_iflist is the list of interfaces; each corresponds to an ifnet
106 static struct bpf_if
*bpf_iflist
;
108 static struct lwkt_token bpf_token
= LWKT_TOKEN_INITIALIZER(bpf_token
);
110 static int bpf_allocbufs(struct bpf_d
*);
111 static void bpf_attachd(struct bpf_d
*d
, struct bpf_if
*bp
);
112 static void bpf_detachd(struct bpf_d
*d
);
113 static void bpf_resetd(struct bpf_d
*);
114 static void bpf_freed(struct bpf_d
*);
115 static void bpf_mcopy(const void *, void *, size_t);
116 static int bpf_movein(struct uio
*, int, struct mbuf
**,
117 struct sockaddr
*, int *, struct bpf_insn
*);
118 static int bpf_setif(struct bpf_d
*, struct ifreq
*);
119 static void bpf_timed_out(void *);
120 static void bpf_wakeup(struct bpf_d
*);
121 static void catchpacket(struct bpf_d
*, u_char
*, u_int
, u_int
,
122 void (*)(const void *, void *, size_t),
123 const struct timeval
*);
124 static int bpf_setf(struct bpf_d
*, struct bpf_program
*, u_long cmd
);
125 static int bpf_getdltlist(struct bpf_d
*, struct bpf_dltlist
*);
126 static int bpf_setdlt(struct bpf_d
*, u_int
);
127 static void bpf_drvinit(void *unused
);
128 static void bpf_filter_detach(struct knote
*kn
);
129 static int bpf_filter_read(struct knote
*kn
, long hint
);
131 static d_open_t bpfopen
;
132 static d_clone_t bpfclone
;
133 static d_close_t bpfclose
;
134 static d_read_t bpfread
;
135 static d_write_t bpfwrite
;
136 static d_ioctl_t bpfioctl
;
137 static d_kqfilter_t bpfkqfilter
;
139 #define CDEV_MAJOR 23
140 static struct dev_ops bpf_ops
= {
141 { "bpf", 0, D_MPSAFE
},
147 .d_kqfilter
= bpfkqfilter
152 bpf_movein(struct uio
*uio
, int linktype
, struct mbuf
**mp
,
153 struct sockaddr
*sockp
, int *datlen
, struct bpf_insn
*wfilter
)
165 * Build a sockaddr based on the data link layer type.
166 * We do this at this level because the ethernet header
167 * is copied directly into the data field of the sockaddr.
168 * In the case of SLIP, there is no header and the packet
169 * is forwarded as is.
170 * Also, we are careful to leave room at the front of the mbuf
171 * for the link level header.
175 sockp
->sa_family
= AF_INET
;
180 sockp
->sa_family
= AF_UNSPEC
;
181 /* XXX Would MAXLINKHDR be better? */
182 hlen
= sizeof(struct ether_header
);
187 sockp
->sa_family
= AF_UNSPEC
;
191 case DLT_ATM_RFC1483
:
193 * en atm driver requires 4-byte atm pseudo header.
194 * though it isn't standard, vpi:vci needs to be
197 sockp
->sa_family
= AF_UNSPEC
;
198 hlen
= 12; /* XXX 4(ATM_PH) + 3(LLC) + 5(SNAP) */
202 sockp
->sa_family
= AF_UNSPEC
;
203 hlen
= 4; /* This should match PPP_HDRLEN */
210 len
= uio
->uio_resid
;
211 *datlen
= len
- hlen
;
212 if ((unsigned)len
> MCLBYTES
)
215 m
= m_getl(len
, MB_WAIT
, MT_DATA
, M_PKTHDR
, NULL
);
218 m
->m_pkthdr
.len
= m
->m_len
= len
;
219 m
->m_pkthdr
.rcvif
= NULL
;
222 if (m
->m_len
< hlen
) {
227 error
= uiomove(mtod(m
, u_char
*), len
, uio
);
231 slen
= bpf_filter(wfilter
, mtod(m
, u_char
*), len
, len
);
238 * Make room for link header, and copy it to sockaddr.
241 bcopy(m
->m_data
, sockp
->sa_data
, hlen
);
242 m
->m_pkthdr
.len
-= hlen
;
244 m
->m_data
+= hlen
; /* XXX */
253 * Attach file to the bpf interface, i.e. make d listen on bp.
254 * Must be called at splimp.
257 bpf_attachd(struct bpf_d
*d
, struct bpf_if
*bp
)
260 * Point d at bp, and add d to the interface's list of listeners.
261 * Finally, point the driver's bpf cookie at the interface so
262 * it will divert packets to bpf.
264 lwkt_gettoken(&bpf_token
);
266 SLIST_INSERT_HEAD(&bp
->bif_dlist
, d
, bd_next
);
267 *bp
->bif_driverp
= bp
;
269 EVENTHANDLER_INVOKE(bpf_track
, bp
->bif_ifp
, bp
->bif_dlt
, 1);
270 lwkt_reltoken(&bpf_token
);
274 * Detach a file from its interface.
277 bpf_detachd(struct bpf_d
*d
)
283 lwkt_gettoken(&bpf_token
);
287 /* Remove d from the interface's descriptor list. */
288 SLIST_REMOVE(&bp
->bif_dlist
, d
, bpf_d
, bd_next
);
290 if (SLIST_EMPTY(&bp
->bif_dlist
)) {
292 * Let the driver know that there are no more listeners.
294 *bp
->bif_driverp
= NULL
;
298 EVENTHANDLER_INVOKE(bpf_track
, ifp
, bp
->bif_dlt
, 0);
301 * Check if this descriptor had requested promiscuous mode.
302 * If so, turn it off.
306 error
= ifpromisc(ifp
, 0);
307 if (error
!= 0 && error
!= ENXIO
) {
309 * ENXIO can happen if a pccard is unplugged,
310 * Something is really wrong if we were able to put
311 * the driver into promiscuous mode, but can't
314 if_printf(ifp
, "bpf_detach: ifpromisc failed(%d)\n",
318 lwkt_reltoken(&bpf_token
);
322 * Open ethernet device. Returns ENXIO for illegal minor device number,
323 * EBUSY if file is open by another process.
327 bpfopen(struct dev_open_args
*ap
)
329 cdev_t dev
= ap
->a_head
.a_dev
;
332 lwkt_gettoken(&bpf_token
);
333 if (ap
->a_cred
->cr_prison
) {
334 lwkt_reltoken(&bpf_token
);
340 * Each minor can be opened by only one process. If the requested
341 * minor is in use, return EBUSY.
344 lwkt_reltoken(&bpf_token
);
348 d
= kmalloc(sizeof *d
, M_BPF
, M_WAITOK
| M_ZERO
);
350 d
->bd_bufsize
= bpf_bufsize
;
353 callout_init(&d
->bd_callout
);
354 lwkt_reltoken(&bpf_token
);
360 bpfclone(struct dev_clone_args
*ap
)
364 unit
= devfs_clone_bitmap_get(&DEVFS_CLONE_BITMAP(bpf
), 0);
365 ap
->a_dev
= make_only_dev(&bpf_ops
, unit
, 0, 0, 0600, "bpf%d", unit
);
371 * Close the descriptor by detaching it from its interface,
372 * deallocating its buffers, and marking it free.
376 bpfclose(struct dev_close_args
*ap
)
378 cdev_t dev
= ap
->a_head
.a_dev
;
379 struct bpf_d
*d
= dev
->si_drv1
;
381 lwkt_gettoken(&bpf_token
);
382 funsetown(&d
->bd_sigio
);
383 if (d
->bd_state
== BPF_WAITING
)
384 callout_stop(&d
->bd_callout
);
385 d
->bd_state
= BPF_IDLE
;
386 if (d
->bd_bif
!= NULL
)
390 if (dev
->si_uminor
>= BPF_PREALLOCATED_UNITS
) {
391 devfs_clone_bitmap_put(&DEVFS_CLONE_BITMAP(bpf
), dev
->si_uminor
);
395 lwkt_reltoken(&bpf_token
);
401 * Rotate the packet buffers in descriptor d. Move the store buffer
402 * into the hold slot, and the free buffer into the store slot.
403 * Zero the length of the new store buffer.
405 #define ROTATE_BUFFERS(d) \
406 (d)->bd_hbuf = (d)->bd_sbuf; \
407 (d)->bd_hlen = (d)->bd_slen; \
408 (d)->bd_sbuf = (d)->bd_fbuf; \
412 * bpfread - read next chunk of packets from buffers
415 bpfread(struct dev_read_args
*ap
)
417 cdev_t dev
= ap
->a_head
.a_dev
;
418 struct bpf_d
*d
= dev
->si_drv1
;
422 lwkt_gettoken(&bpf_token
);
424 * Restrict application to use a buffer the same size as
427 if (ap
->a_uio
->uio_resid
!= d
->bd_bufsize
) {
428 lwkt_reltoken(&bpf_token
);
432 if (d
->bd_state
== BPF_WAITING
)
433 callout_stop(&d
->bd_callout
);
434 timed_out
= (d
->bd_state
== BPF_TIMED_OUT
);
435 d
->bd_state
= BPF_IDLE
;
437 * If the hold buffer is empty, then do a timed sleep, which
438 * ends when the timeout expires or when enough packets
439 * have arrived to fill the store buffer.
441 while (d
->bd_hbuf
== NULL
) {
442 if ((d
->bd_immediate
|| (ap
->a_ioflag
& IO_NDELAY
) || timed_out
)
443 && d
->bd_slen
!= 0) {
445 * A packet(s) either arrived since the previous,
446 * We're in immediate mode, or are reading
447 * in non-blocking mode, and a packet(s)
448 * either arrived since the previous
449 * read or arrived while we were asleep.
450 * Rotate the buffers and return what's here.
457 * No data is available, check to see if the bpf device
458 * is still pointed at a real interface. If not, return
459 * ENXIO so that the userland process knows to rebind
460 * it before using it again.
462 if (d
->bd_bif
== NULL
) {
463 lwkt_reltoken(&bpf_token
);
467 if (ap
->a_ioflag
& IO_NDELAY
) {
468 lwkt_reltoken(&bpf_token
);
471 error
= tsleep(d
, PCATCH
, "bpf", d
->bd_rtout
);
472 if (error
== EINTR
|| error
== ERESTART
) {
473 lwkt_reltoken(&bpf_token
);
476 if (error
== EWOULDBLOCK
) {
478 * On a timeout, return what's in the buffer,
479 * which may be nothing. If there is something
480 * in the store buffer, we can rotate the buffers.
484 * We filled up the buffer in between
485 * getting the timeout and arriving
486 * here, so we don't need to rotate.
490 if (d
->bd_slen
== 0) {
491 lwkt_reltoken(&bpf_token
);
499 * At this point, we know we have something in the hold slot.
503 * Move data from hold buffer into user space.
504 * We know the entire buffer is transferred since
505 * we checked above that the read buffer is bpf_bufsize bytes.
507 error
= uiomove(d
->bd_hbuf
, d
->bd_hlen
, ap
->a_uio
);
509 d
->bd_fbuf
= d
->bd_hbuf
;
512 lwkt_reltoken(&bpf_token
);
519 * If there are processes sleeping on this descriptor, wake them up.
522 bpf_wakeup(struct bpf_d
*d
)
524 if (d
->bd_state
== BPF_WAITING
) {
525 callout_stop(&d
->bd_callout
);
526 d
->bd_state
= BPF_IDLE
;
529 if (d
->bd_async
&& d
->bd_sig
&& d
->bd_sigio
)
530 pgsigio(d
->bd_sigio
, d
->bd_sig
, 0);
532 KNOTE(&d
->bd_kq
.ki_note
, 0);
536 bpf_timed_out(void *arg
)
538 struct bpf_d
*d
= (struct bpf_d
*)arg
;
540 if (d
->bd_state
== BPF_WAITING
) {
541 d
->bd_state
= BPF_TIMED_OUT
;
548 bpf_output_dispatch(netmsg_t msg
)
550 struct netmsg_bpf_output
*bmsg
= (struct netmsg_bpf_output
*)msg
;
551 struct ifnet
*ifp
= bmsg
->nm_ifp
;
555 * The driver frees the mbuf.
557 error
= ifp
->if_output(ifp
, bmsg
->nm_mbuf
, bmsg
->nm_dst
, NULL
);
558 lwkt_replymsg(&msg
->lmsg
, error
);
562 bpfwrite(struct dev_write_args
*ap
)
564 cdev_t dev
= ap
->a_head
.a_dev
;
565 struct bpf_d
*d
= dev
->si_drv1
;
571 struct netmsg_bpf_output bmsg
;
573 lwkt_gettoken(&bpf_token
);
574 if (d
->bd_bif
== NULL
) {
575 lwkt_reltoken(&bpf_token
);
579 ifp
= d
->bd_bif
->bif_ifp
;
581 if (ap
->a_uio
->uio_resid
== 0) {
582 lwkt_reltoken(&bpf_token
);
586 error
= bpf_movein(ap
->a_uio
, (int)d
->bd_bif
->bif_dlt
, &m
,
587 &dst
, &datlen
, d
->bd_wfilter
);
589 lwkt_reltoken(&bpf_token
);
593 if (datlen
> ifp
->if_mtu
) {
595 lwkt_reltoken(&bpf_token
);
600 dst
.sa_family
= pseudo_AF_HDRCMPLT
;
602 netmsg_init(&bmsg
.base
, NULL
, &curthread
->td_msgport
,
603 0, bpf_output_dispatch
);
608 ret
= lwkt_domsg(netisr_cpuport(0), &bmsg
.base
.lmsg
, 0);
609 lwkt_reltoken(&bpf_token
);
615 * Reset a descriptor by flushing its packet buffer and clearing the
616 * receive and drop counts. Should be called at splimp.
619 bpf_resetd(struct bpf_d
*d
)
622 /* Free the hold buffer. */
623 d
->bd_fbuf
= d
->bd_hbuf
;
633 * FIONREAD Check for read packet available.
634 * SIOCGIFADDR Get interface address - convenient hook to driver.
635 * BIOCGBLEN Get buffer len [for read()].
636 * BIOCSETF Set ethernet read filter.
637 * BIOCSETWF Set ethernet write filter.
638 * BIOCFLUSH Flush read packet buffer.
639 * BIOCPROMISC Put interface into promiscuous mode.
640 * BIOCGDLT Get link layer type.
641 * BIOCGETIF Get interface name.
642 * BIOCSETIF Set interface.
643 * BIOCSRTIMEOUT Set read timeout.
644 * BIOCGRTIMEOUT Get read timeout.
645 * BIOCGSTATS Get packet stats.
646 * BIOCIMMEDIATE Set immediate mode.
647 * BIOCVERSION Get filter language version.
648 * BIOCGHDRCMPLT Get "header already complete" flag
649 * BIOCSHDRCMPLT Set "header already complete" flag
650 * BIOCGSEESENT Get "see packets sent" flag
651 * BIOCSSEESENT Set "see packets sent" flag
652 * BIOCLOCK Set "locked" flag
656 bpfioctl(struct dev_ioctl_args
*ap
)
658 cdev_t dev
= ap
->a_head
.a_dev
;
659 struct bpf_d
*d
= dev
->si_drv1
;
662 lwkt_gettoken(&bpf_token
);
663 if (d
->bd_state
== BPF_WAITING
)
664 callout_stop(&d
->bd_callout
);
665 d
->bd_state
= BPF_IDLE
;
667 if (d
->bd_locked
== 1) {
686 lwkt_reltoken(&bpf_token
);
696 * Check for read packet available.
706 *(int *)ap
->a_data
= n
;
714 if (d
->bd_bif
== NULL
) {
717 ifp
= d
->bd_bif
->bif_ifp
;
718 ifnet_serialize_all(ifp
);
719 error
= ifp
->if_ioctl(ifp
, ap
->a_cmd
,
720 ap
->a_data
, ap
->a_cred
);
721 ifnet_deserialize_all(ifp
);
727 * Get buffer len [for read()].
730 *(u_int
*)ap
->a_data
= d
->bd_bufsize
;
737 if (d
->bd_bif
!= NULL
) {
740 u_int size
= *(u_int
*)ap
->a_data
;
742 if (size
> bpf_maxbufsize
)
743 *(u_int
*)ap
->a_data
= size
= bpf_maxbufsize
;
744 else if (size
< BPF_MINBUFSIZE
)
745 *(u_int
*)ap
->a_data
= size
= BPF_MINBUFSIZE
;
746 d
->bd_bufsize
= size
;
751 * Set link layer read filter.
755 error
= bpf_setf(d
, (struct bpf_program
*)ap
->a_data
,
760 * Flush read packet buffer.
767 * Put interface into promiscuous mode.
770 if (d
->bd_bif
== NULL
) {
772 * No interface attached yet.
777 if (d
->bd_promisc
== 0) {
778 error
= ifpromisc(d
->bd_bif
->bif_ifp
, 1);
785 * Get device parameters.
788 if (d
->bd_bif
== NULL
)
791 *(u_int
*)ap
->a_data
= d
->bd_bif
->bif_dlt
;
795 * Get a list of supported data link types.
798 if (d
->bd_bif
== NULL
) {
801 error
= bpf_getdltlist(d
,
802 (struct bpf_dltlist
*)ap
->a_data
);
807 * Set data link type.
810 if (d
->bd_bif
== NULL
)
813 error
= bpf_setdlt(d
, *(u_int
*)ap
->a_data
);
817 * Get interface name.
820 if (d
->bd_bif
== NULL
) {
823 struct ifnet
*const ifp
= d
->bd_bif
->bif_ifp
;
824 struct ifreq
*const ifr
= (struct ifreq
*)ap
->a_data
;
826 strlcpy(ifr
->ifr_name
, ifp
->if_xname
,
827 sizeof ifr
->ifr_name
);
835 error
= bpf_setif(d
, (struct ifreq
*)ap
->a_data
);
843 struct timeval
*tv
= (struct timeval
*)ap
->a_data
;
846 * Subtract 1 tick from tvtohz() since this isn't
849 if ((error
= itimerfix(tv
)) == 0)
850 d
->bd_rtout
= tvtohz_low(tv
);
859 struct timeval
*tv
= (struct timeval
*)ap
->a_data
;
861 tv
->tv_sec
= d
->bd_rtout
/ hz
;
862 tv
->tv_usec
= (d
->bd_rtout
% hz
) * ustick
;
871 struct bpf_stat
*bs
= (struct bpf_stat
*)ap
->a_data
;
873 bs
->bs_recv
= d
->bd_rcount
;
874 bs
->bs_drop
= d
->bd_dcount
;
879 * Set immediate mode.
882 d
->bd_immediate
= *(u_int
*)ap
->a_data
;
887 struct bpf_version
*bv
= (struct bpf_version
*)ap
->a_data
;
889 bv
->bv_major
= BPF_MAJOR_VERSION
;
890 bv
->bv_minor
= BPF_MINOR_VERSION
;
895 * Get "header already complete" flag
898 *(u_int
*)ap
->a_data
= d
->bd_hdrcmplt
;
902 * Set "header already complete" flag
905 d
->bd_hdrcmplt
= *(u_int
*)ap
->a_data
? 1 : 0;
909 * Get "see sent packets" flag
912 *(u_int
*)ap
->a_data
= d
->bd_seesent
;
916 * Set "see sent packets" flag
919 d
->bd_seesent
= *(u_int
*)ap
->a_data
;
922 case FIOASYNC
: /* Send signal on receive packets */
923 d
->bd_async
= *(int *)ap
->a_data
;
927 error
= fsetown(*(int *)ap
->a_data
, &d
->bd_sigio
);
931 *(int *)ap
->a_data
= fgetown(&d
->bd_sigio
);
934 /* This is deprecated, FIOSETOWN should be used instead. */
936 error
= fsetown(-(*(int *)ap
->a_data
), &d
->bd_sigio
);
939 /* This is deprecated, FIOGETOWN should be used instead. */
941 *(int *)ap
->a_data
= -fgetown(&d
->bd_sigio
);
944 case BIOCSRSIG
: /* Set receive signal */
948 sig
= *(u_int
*)ap
->a_data
;
957 *(u_int
*)ap
->a_data
= d
->bd_sig
;
963 lwkt_reltoken(&bpf_token
);
969 * Set d's packet filter program to fp. If this file already has a filter,
970 * free it and replace it. Returns EINVAL for bogus requests.
973 bpf_setf(struct bpf_d
*d
, struct bpf_program
*fp
, u_long cmd
)
975 struct bpf_insn
*fcode
, *old
;
976 u_int wfilter
, flen
, size
;
978 if (cmd
== BIOCSETWF
) {
985 if (fp
->bf_insns
== NULL
) {
989 d
->bd_wfilter
= NULL
;
991 d
->bd_rfilter
= NULL
;
998 if (flen
> BPF_MAXINSNS
)
1001 size
= flen
* sizeof *fp
->bf_insns
;
1002 fcode
= (struct bpf_insn
*)kmalloc(size
, M_BPF
, M_WAITOK
);
1003 if (copyin(fp
->bf_insns
, fcode
, size
) == 0 &&
1004 bpf_validate(fcode
, (int)flen
)) {
1006 d
->bd_wfilter
= fcode
;
1008 d
->bd_rfilter
= fcode
;
1015 kfree(fcode
, M_BPF
);
1020 * Detach a file from its current interface (if attached at all) and attach
1021 * to the interface indicated by the name stored in ifr.
1022 * Return an errno or 0.
1025 bpf_setif(struct bpf_d
*d
, struct ifreq
*ifr
)
1029 struct ifnet
*theywant
;
1031 theywant
= ifunit(ifr
->ifr_name
);
1032 if (theywant
== NULL
)
1036 * Look through attached interfaces for the named one.
1038 for (bp
= bpf_iflist
; bp
!= NULL
; bp
= bp
->bif_next
) {
1039 struct ifnet
*ifp
= bp
->bif_ifp
;
1041 if (ifp
== NULL
|| ifp
!= theywant
)
1043 /* skip additional entry */
1044 if (bp
->bif_driverp
!= &ifp
->if_bpf
)
1047 * We found the requested interface.
1048 * Allocate the packet buffers if we need to.
1049 * If we're already attached to requested interface,
1050 * just flush the buffer.
1052 if (d
->bd_sbuf
== NULL
) {
1053 error
= bpf_allocbufs(d
);
1057 if (bp
!= d
->bd_bif
) {
1058 if (d
->bd_bif
!= NULL
) {
1060 * Detach if attached to something else.
1075 static struct filterops bpf_read_filtops
=
1076 { FILTEROP_ISFD
, NULL
, bpf_filter_detach
, bpf_filter_read
};
1079 bpfkqfilter(struct dev_kqfilter_args
*ap
)
1081 cdev_t dev
= ap
->a_head
.a_dev
;
1082 struct knote
*kn
= ap
->a_kn
;
1083 struct klist
*klist
;
1086 lwkt_gettoken(&bpf_token
);
1088 if (d
->bd_bif
== NULL
) {
1090 lwkt_reltoken(&bpf_token
);
1095 switch (kn
->kn_filter
) {
1097 kn
->kn_fop
= &bpf_read_filtops
;
1098 kn
->kn_hook
= (caddr_t
)d
;
1101 ap
->a_result
= EOPNOTSUPP
;
1102 lwkt_reltoken(&bpf_token
);
1106 klist
= &d
->bd_kq
.ki_note
;
1107 knote_insert(klist
, kn
);
1108 lwkt_reltoken(&bpf_token
);
1114 bpf_filter_detach(struct knote
*kn
)
1116 struct klist
*klist
;
1119 d
= (struct bpf_d
*)kn
->kn_hook
;
1120 klist
= &d
->bd_kq
.ki_note
;
1121 knote_remove(klist
, kn
);
1125 bpf_filter_read(struct knote
*kn
, long hint
)
1130 d
= (struct bpf_d
*)kn
->kn_hook
;
1131 if (d
->bd_hlen
!= 0 ||
1132 ((d
->bd_immediate
|| d
->bd_state
== BPF_TIMED_OUT
) &&
1136 /* Start the read timeout if necessary. */
1137 if (d
->bd_rtout
> 0 && d
->bd_state
== BPF_IDLE
) {
1138 callout_reset(&d
->bd_callout
, d
->bd_rtout
,
1140 d
->bd_state
= BPF_WAITING
;
1149 * Process the packet pkt of length pktlen. The packet is parsed
1150 * by each listener's filter, and if accepted, stashed into the
1151 * corresponding buffer.
1154 bpf_tap(struct bpf_if
*bp
, u_char
*pkt
, u_int pktlen
)
1161 lwkt_gettoken(&bpf_token
);
1164 lwkt_reltoken(&bpf_token
);
1169 * Note that the ipl does not have to be raised at this point.
1170 * The only problem that could arise here is that if two different
1171 * interfaces shared any data. This is not the case.
1173 SLIST_FOREACH(d
, &bp
->bif_dlist
, bd_next
) {
1175 slen
= bpf_filter(d
->bd_rfilter
, pkt
, pktlen
, pktlen
);
1181 catchpacket(d
, pkt
, pktlen
, slen
, ovbcopy
, &tv
);
1184 lwkt_reltoken(&bpf_token
);
1188 * Copy data from an mbuf chain into a buffer. This code is derived
1189 * from m_copydata in sys/uipc_mbuf.c.
1192 bpf_mcopy(const void *src_arg
, void *dst_arg
, size_t len
)
1194 const struct mbuf
*m
;
1203 count
= min(m
->m_len
, len
);
1204 bcopy(mtod(m
, void *), dst
, count
);
1212 * Process the packet in the mbuf chain m. The packet is parsed by each
1213 * listener's filter, and if accepted, stashed into the corresponding
1217 bpf_mtap(struct bpf_if
*bp
, struct mbuf
*m
)
1224 lwkt_gettoken(&bpf_token
);
1227 lwkt_reltoken(&bpf_token
);
1231 /* Don't compute pktlen, if no descriptor is attached. */
1232 if (SLIST_EMPTY(&bp
->bif_dlist
)) {
1233 lwkt_reltoken(&bpf_token
);
1237 pktlen
= m_lengthm(m
, NULL
);
1239 SLIST_FOREACH(d
, &bp
->bif_dlist
, bd_next
) {
1240 if (!d
->bd_seesent
&& (m
->m_pkthdr
.rcvif
== NULL
))
1243 slen
= bpf_filter(d
->bd_rfilter
, (u_char
*)m
, pktlen
, 0);
1249 catchpacket(d
, (u_char
*)m
, pktlen
, slen
, bpf_mcopy
,
1253 lwkt_reltoken(&bpf_token
);
1257 * Incoming linkage from device drivers, where we have a mbuf chain
1258 * but need to prepend some arbitrary header from a linear buffer.
1260 * Con up a minimal dummy header to pacify bpf. Allocate (only) a
1261 * struct m_hdr on the stack. This is safe as bpf only reads from the
1262 * fields in this header that we initialize, and will not try to free
1263 * it or keep a pointer to it.
1266 bpf_mtap_hdr(struct bpf_if
*arg
, caddr_t data
, u_int dlen
, struct mbuf
*m
, u_int direction
)
1275 return bpf_mtap(arg
, (struct mbuf
*) &mh
);
1279 bpf_mtap_family(struct bpf_if
*bp
, struct mbuf
*m
, sa_family_t family
)
1283 KKASSERT(family
!= AF_UNSPEC
);
1285 family4
= (u_int
)family
;
1286 bpf_ptap(bp
, m
, &family4
, sizeof(family4
));
1290 * Process the packet in the mbuf chain m with the header in m prepended.
1291 * The packet is parsed by each listener's filter, and if accepted,
1292 * stashed into the corresponding buffer.
1295 bpf_ptap(struct bpf_if
*bp
, struct mbuf
*m
, const void *data
, u_int dlen
)
1300 * Craft on-stack mbuf suitable for passing to bpf_mtap.
1301 * Note that we cut corners here; we only setup what's
1302 * absolutely needed--this mbuf should never go anywhere else.
1305 mb
.m_data
= __DECONST(void *, data
); /* LINTED */
1307 mb
.m_pkthdr
.rcvif
= m
->m_pkthdr
.rcvif
;
1313 * Move the packet data from interface memory (pkt) into the
1314 * store buffer. Return 1 if it's time to wakeup a listener (buffer full),
1315 * otherwise 0. "copy" is the routine called to do the actual data
1316 * transfer. bcopy is passed in to copy contiguous chunks, while
1317 * bpf_mcopy is passed in to copy mbuf chains. In the latter case,
1318 * pkt is really an mbuf.
1321 catchpacket(struct bpf_d
*d
, u_char
*pkt
, u_int pktlen
, u_int snaplen
,
1322 void (*cpfn
)(const void *, void *, size_t),
1323 const struct timeval
*tv
)
1327 int hdrlen
= d
->bd_bif
->bif_hdrlen
;
1330 * Figure out how many bytes to move. If the packet is
1331 * greater or equal to the snapshot length, transfer that
1332 * much. Otherwise, transfer the whole packet (unless
1333 * we hit the buffer size limit).
1335 totlen
= hdrlen
+ min(snaplen
, pktlen
);
1336 if (totlen
> d
->bd_bufsize
)
1337 totlen
= d
->bd_bufsize
;
1340 * Round up the end of the previous packet to the next longword.
1342 curlen
= BPF_WORDALIGN(d
->bd_slen
);
1343 if (curlen
+ totlen
> d
->bd_bufsize
) {
1345 * This packet will overflow the storage buffer.
1346 * Rotate the buffers if we can, then wakeup any
1349 if (d
->bd_fbuf
== NULL
) {
1351 * We haven't completed the previous read yet,
1352 * so drop the packet.
1360 } else if (d
->bd_immediate
|| d
->bd_state
== BPF_TIMED_OUT
) {
1362 * Immediate mode is set, or the read timeout has
1363 * already expired during a select call. A packet
1364 * arrived, so the reader should be woken up.
1370 * Append the bpf header.
1372 hp
= (struct bpf_hdr
*)(d
->bd_sbuf
+ curlen
);
1373 hp
->bh_tstamp
= *tv
;
1374 hp
->bh_datalen
= pktlen
;
1375 hp
->bh_hdrlen
= hdrlen
;
1377 * Copy the packet data into the store buffer and update its length.
1379 (*cpfn
)(pkt
, (u_char
*)hp
+ hdrlen
, (hp
->bh_caplen
= totlen
- hdrlen
));
1380 d
->bd_slen
= curlen
+ totlen
;
1387 * Initialize all nonzero fields of a descriptor.
1390 bpf_allocbufs(struct bpf_d
*d
)
1392 d
->bd_fbuf
= kmalloc(d
->bd_bufsize
, M_BPF
, M_WAITOK
);
1393 d
->bd_sbuf
= kmalloc(d
->bd_bufsize
, M_BPF
, M_WAITOK
);
1400 * Free buffers and packet filter program currently in use by a descriptor.
1404 bpf_freed(struct bpf_d
*d
)
1407 * We don't need to lock out interrupts since this descriptor has
1408 * been detached from its interface and it yet hasn't been marked
1411 if (d
->bd_sbuf
!= NULL
) {
1412 kfree(d
->bd_sbuf
, M_BPF
);
1413 if (d
->bd_hbuf
!= NULL
)
1414 kfree(d
->bd_hbuf
, M_BPF
);
1415 if (d
->bd_fbuf
!= NULL
)
1416 kfree(d
->bd_fbuf
, M_BPF
);
1419 kfree(d
->bd_rfilter
, M_BPF
);
1421 kfree(d
->bd_wfilter
, M_BPF
);
1425 * Attach an interface to bpf. ifp is a pointer to the structure
1426 * defining the interface to be attached, dlt is the link layer type,
1427 * and hdrlen is the fixed size of the link header (variable length
1428 * headers are not yet supported).
1431 bpfattach(struct ifnet
*ifp
, u_int dlt
, u_int hdrlen
)
1433 bpfattach_dlt(ifp
, dlt
, hdrlen
, &ifp
->if_bpf
);
1437 bpfattach_dlt(struct ifnet
*ifp
, u_int dlt
, u_int hdrlen
, struct bpf_if
**driverp
)
1441 bp
= kmalloc(sizeof *bp
, M_BPF
, M_WAITOK
| M_ZERO
);
1443 lwkt_gettoken(&bpf_token
);
1445 SLIST_INIT(&bp
->bif_dlist
);
1448 bp
->bif_driverp
= driverp
;
1449 *bp
->bif_driverp
= NULL
;
1451 bp
->bif_next
= bpf_iflist
;
1455 * Compute the length of the bpf header. This is not necessarily
1456 * equal to SIZEOF_BPF_HDR because we want to insert spacing such
1457 * that the network layer header begins on a longword boundary (for
1458 * performance reasons and to alleviate alignment restrictions).
1460 bp
->bif_hdrlen
= BPF_WORDALIGN(hdrlen
+ SIZEOF_BPF_HDR
) - hdrlen
;
1462 lwkt_reltoken(&bpf_token
);
1465 if_printf(ifp
, "bpf attached\n");
1469 * Detach bpf from an interface. This involves detaching each descriptor
1470 * associated with the interface, and leaving bd_bif NULL. Notify each
1471 * descriptor as it's detached so that any sleepers wake up and get
1475 bpfdetach(struct ifnet
*ifp
)
1477 struct bpf_if
*bp
, *bp_prev
;
1480 lwkt_gettoken(&bpf_token
);
1482 /* Locate BPF interface information */
1484 for (bp
= bpf_iflist
; bp
!= NULL
; bp
= bp
->bif_next
) {
1485 if (ifp
== bp
->bif_ifp
)
1490 /* Interface wasn't attached */
1491 if (bp
->bif_ifp
== NULL
) {
1492 lwkt_reltoken(&bpf_token
);
1493 kprintf("bpfdetach: %s was not attached\n", ifp
->if_xname
);
1497 while ((d
= SLIST_FIRST(&bp
->bif_dlist
)) != NULL
) {
1502 if (bp_prev
!= NULL
)
1503 bp_prev
->bif_next
= bp
->bif_next
;
1505 bpf_iflist
= bp
->bif_next
;
1509 lwkt_reltoken(&bpf_token
);
1513 * Get a list of available data link type of the interface.
1516 bpf_getdltlist(struct bpf_d
*d
, struct bpf_dltlist
*bfl
)
1522 ifp
= d
->bd_bif
->bif_ifp
;
1525 for (bp
= bpf_iflist
; bp
!= NULL
; bp
= bp
->bif_next
) {
1526 if (bp
->bif_ifp
!= ifp
)
1528 if (bfl
->bfl_list
!= NULL
) {
1529 if (n
>= bfl
->bfl_len
) {
1532 error
= copyout(&bp
->bif_dlt
,
1533 bfl
->bfl_list
+ n
, sizeof(u_int
));
1542 * Set the data link type of a BPF instance.
1545 bpf_setdlt(struct bpf_d
*d
, u_int dlt
)
1547 int error
, opromisc
;
1551 if (d
->bd_bif
->bif_dlt
== dlt
)
1553 ifp
= d
->bd_bif
->bif_ifp
;
1554 for (bp
= bpf_iflist
; bp
!= NULL
; bp
= bp
->bif_next
) {
1555 if (bp
->bif_ifp
== ifp
&& bp
->bif_dlt
== dlt
)
1559 opromisc
= d
->bd_promisc
;
1564 error
= ifpromisc(bp
->bif_ifp
, 1);
1566 if_printf(bp
->bif_ifp
,
1567 "bpf_setdlt: ifpromisc failed (%d)\n",
1574 return(bp
== NULL
? EINVAL
: 0);
1580 lwkt_gettoken(&bpf_token
);
1586 lwkt_reltoken(&bpf_token
);
1590 bpf_drvinit(void *unused
)
1594 make_autoclone_dev(&bpf_ops
, &DEVFS_CLONE_BITMAP(bpf
),
1595 bpfclone
, 0, 0, 0600, "bpf");
1596 for (i
= 0; i
< BPF_PREALLOCATED_UNITS
; i
++) {
1597 make_dev(&bpf_ops
, i
, 0, 0, 0600, "bpf%d", i
);
1598 devfs_clone_bitmap_set(&DEVFS_CLONE_BITMAP(bpf
), i
);
1603 bpf_drvuninit(void *unused
)
1605 devfs_clone_handler_del("bpf");
1606 dev_ops_remove_all(&bpf_ops
);
1607 devfs_clone_bitmap_uninit(&DEVFS_CLONE_BITMAP(bpf
));
1610 SYSINIT(bpfdev
,SI_SUB_DRIVERS
,SI_ORDER_MIDDLE
+CDEV_MAJOR
,bpf_drvinit
,NULL
)
1611 SYSUNINIT(bpfdev
, SI_SUB_DRIVERS
,SI_ORDER_MIDDLE
+CDEV_MAJOR
,bpf_drvuninit
, NULL
);
1615 * NOP stubs to allow bpf-using drivers to load and function.
1617 * A 'better' implementation would allow the core bpf functionality
1618 * to be loaded at runtime.
1622 bpf_tap(struct bpf_if
*bp
, u_char
*pkt
, u_int pktlen
)
1627 bpf_mtap(struct bpf_if
*bp
, struct mbuf
*m
)
1632 bpf_ptap(struct bpf_if
*bp
, struct mbuf
*m
, const void *data
, u_int dlen
)
1637 bpfattach(struct ifnet
*ifp
, u_int dlt
, u_int hdrlen
)
1642 bpfattach_dlt(struct ifnet
*ifp
, u_int dlt
, u_int hdrlen
, struct bpf_if
**driverp
)
1647 bpfdetach(struct ifnet
*ifp
)
1652 bpf_filter(const struct bpf_insn
*pc
, u_char
*p
, u_int wirelen
, u_int buflen
)
1654 return -1; /* "no filter" behaviour */