2 * Copyright (c) 1990, 1991, 1993
3 * The Regents of the University of California. All rights reserved.
5 * This code is derived from the Stanford/CMU enet packet filter,
6 * (net/enet.c) distributed as part of 4.3BSD, and code contributed
7 * to Berkeley by Steven McCanne and Van Jacobson both of Lawrence
10 * Redistribution and use in source and binary forms, with or without
11 * modification, are permitted provided that the following conditions
13 * 1. Redistributions of source code must retain the above copyright
14 * notice, this list of conditions and the following disclaimer.
15 * 2. Redistributions in binary form must reproduce the above copyright
16 * notice, this list of conditions and the following disclaimer in the
17 * documentation and/or other materials provided with the distribution.
18 * 3. All advertising materials mentioning features or use of this software
19 * must display the following acknowledgement:
20 * This product includes software developed by the University of
21 * California, Berkeley and its contributors.
22 * 4. Neither the name of the University nor the names of its contributors
23 * may be used to endorse or promote products derived from this software
24 * without specific prior written permission.
26 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
27 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
28 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
29 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
30 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
31 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
32 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
33 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
34 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
35 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
38 * @(#)bpf.c 8.2 (Berkeley) 3/28/94
40 * $FreeBSD: src/sys/net/bpf.c,v 1.59.2.12 2002/04/14 21:41:48 luigi Exp $
41 * $DragonFly: src/sys/net/bpf.c,v 1.50 2008/09/23 11:28:49 sephe Exp $
46 #include <sys/param.h>
47 #include <sys/systm.h>
49 #include <sys/device.h>
50 #include <sys/malloc.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>
62 #include <sys/socket.h>
63 #include <sys/vnode.h>
65 #include <sys/thread2.h>
66 #include <sys/mplock2.h>
70 #include <net/bpfdesc.h>
71 #include <net/netmsg2.h>
73 #include <netinet/in.h>
74 #include <netinet/if_ether.h>
75 #include <sys/kernel.h>
76 #include <sys/sysctl.h>
78 #include <sys/devfs.h>
80 struct netmsg_bpf_output
{
81 struct netmsg nm_netmsg
;
84 struct sockaddr
*nm_dst
;
87 MALLOC_DEFINE(M_BPF
, "BPF", "BPF data");
88 DEVFS_DECLARE_CLONE_BITMAP(bpf
);
91 #define BPF_PREALLOCATED_UNITS 4
93 #define BPF_PREALLOCATED_UNITS NBPF
99 * The default read buffer size is patchable.
101 static int bpf_bufsize
= BPF_DEFAULTBUFSIZE
;
102 SYSCTL_INT(_debug
, OID_AUTO
, bpf_bufsize
, CTLFLAG_RW
,
103 &bpf_bufsize
, 0, "");
104 int bpf_maxbufsize
= BPF_MAXBUFSIZE
;
105 SYSCTL_INT(_debug
, OID_AUTO
, bpf_maxbufsize
, CTLFLAG_RW
,
106 &bpf_maxbufsize
, 0, "");
109 * bpf_iflist is the list of interfaces; each corresponds to an ifnet
111 static struct bpf_if
*bpf_iflist
;
113 static int bpf_allocbufs(struct bpf_d
*);
114 static void bpf_attachd(struct bpf_d
*d
, struct bpf_if
*bp
);
115 static void bpf_detachd(struct bpf_d
*d
);
116 static void bpf_resetd(struct bpf_d
*);
117 static void bpf_freed(struct bpf_d
*);
118 static void bpf_mcopy(const void *, void *, size_t);
119 static int bpf_movein(struct uio
*, int, struct mbuf
**,
120 struct sockaddr
*, int *, struct bpf_insn
*);
121 static int bpf_setif(struct bpf_d
*, struct ifreq
*);
122 static void bpf_timed_out(void *);
123 static void bpf_wakeup(struct bpf_d
*);
124 static void catchpacket(struct bpf_d
*, u_char
*, u_int
, u_int
,
125 void (*)(const void *, void *, size_t),
126 const struct timeval
*);
127 static int bpf_setf(struct bpf_d
*, struct bpf_program
*, u_long cmd
);
128 static int bpf_getdltlist(struct bpf_d
*, struct bpf_dltlist
*);
129 static int bpf_setdlt(struct bpf_d
*, u_int
);
130 static void bpf_drvinit(void *unused
);
132 static d_open_t bpfopen
;
133 static d_clone_t bpfclone
;
134 static d_close_t bpfclose
;
135 static d_read_t bpfread
;
136 static d_write_t bpfwrite
;
137 static d_ioctl_t bpfioctl
;
138 static d_poll_t bpfpoll
;
140 #define CDEV_MAJOR 23
141 static struct dev_ops bpf_ops
= {
142 { "bpf", CDEV_MAJOR
, 0 },
153 bpf_movein(struct uio
*uio
, int linktype
, struct mbuf
**mp
,
154 struct sockaddr
*sockp
, int *datlen
, struct bpf_insn
*wfilter
)
166 * Build a sockaddr based on the data link layer type.
167 * We do this at this level because the ethernet header
168 * is copied directly into the data field of the sockaddr.
169 * In the case of SLIP, there is no header and the packet
170 * is forwarded as is.
171 * Also, we are careful to leave room at the front of the mbuf
172 * for the link level header.
176 sockp
->sa_family
= AF_INET
;
181 sockp
->sa_family
= AF_UNSPEC
;
182 /* XXX Would MAXLINKHDR be better? */
183 hlen
= sizeof(struct ether_header
);
188 sockp
->sa_family
= AF_UNSPEC
;
192 case DLT_ATM_RFC1483
:
194 * en atm driver requires 4-byte atm pseudo header.
195 * though it isn't standard, vpi:vci needs to be
198 sockp
->sa_family
= AF_UNSPEC
;
199 hlen
= 12; /* XXX 4(ATM_PH) + 3(LLC) + 5(SNAP) */
203 sockp
->sa_family
= AF_UNSPEC
;
204 hlen
= 4; /* This should match PPP_HDRLEN */
211 len
= uio
->uio_resid
;
212 *datlen
= len
- hlen
;
213 if ((unsigned)len
> MCLBYTES
)
216 m
= m_getl(len
, MB_WAIT
, MT_DATA
, M_PKTHDR
, NULL
);
219 m
->m_pkthdr
.len
= m
->m_len
= len
;
220 m
->m_pkthdr
.rcvif
= NULL
;
223 if (m
->m_len
< hlen
) {
228 error
= uiomove(mtod(m
, u_char
*), len
, uio
);
232 slen
= bpf_filter(wfilter
, mtod(m
, u_char
*), len
, len
);
239 * Make room for link header, and copy it to sockaddr.
242 bcopy(m
->m_data
, sockp
->sa_data
, hlen
);
243 m
->m_pkthdr
.len
-= hlen
;
245 m
->m_data
+= hlen
; /* XXX */
254 * Attach file to the bpf interface, i.e. make d listen on bp.
255 * Must be called at splimp.
258 bpf_attachd(struct bpf_d
*d
, struct bpf_if
*bp
)
261 * Point d at bp, and add d to the interface's list of listeners.
262 * Finally, point the driver's bpf cookie at the interface so
263 * it will divert packets to bpf.
266 SLIST_INSERT_HEAD(&bp
->bif_dlist
, d
, bd_next
);
267 *bp
->bif_driverp
= bp
;
271 * Detach a file from its interface.
274 bpf_detachd(struct bpf_d
*d
)
283 /* Remove d from the interface's descriptor list. */
284 SLIST_REMOVE(&bp
->bif_dlist
, d
, bpf_d
, bd_next
);
286 if (SLIST_EMPTY(&bp
->bif_dlist
)) {
288 * Let the driver know that there are no more listeners.
290 *bp
->bif_driverp
= NULL
;
294 * Check if this descriptor had requested promiscuous mode.
295 * If so, turn it off.
299 error
= ifpromisc(ifp
, 0);
300 if (error
!= 0 && error
!= ENXIO
) {
302 * ENXIO can happen if a pccard is unplugged,
303 * Something is really wrong if we were able to put
304 * the driver into promiscuous mode, but can't
307 if_printf(ifp
, "bpf_detach: ifpromisc failed(%d)\n",
314 * Open ethernet device. Returns ENXIO for illegal minor device number,
315 * EBUSY if file is open by another process.
319 bpfopen(struct dev_open_args
*ap
)
321 cdev_t dev
= ap
->a_head
.a_dev
;
324 if (ap
->a_cred
->cr_prison
)
329 * Each minor can be opened by only one process. If the requested
330 * minor is in use, return EBUSY.
335 MALLOC(d
, struct bpf_d
*, sizeof *d
, M_BPF
, M_WAITOK
| M_ZERO
);
337 d
->bd_bufsize
= bpf_bufsize
;
340 callout_init(&d
->bd_callout
);
345 bpfclone(struct dev_clone_args
*ap
)
349 unit
= devfs_clone_bitmap_get(&DEVFS_CLONE_BITMAP(bpf
), 0);
350 ap
->a_dev
= make_only_dev(&bpf_ops
, unit
, 0, 0, 0600, "bpf%d", unit
);
356 * Close the descriptor by detaching it from its interface,
357 * deallocating its buffers, and marking it free.
361 bpfclose(struct dev_close_args
*ap
)
363 cdev_t dev
= ap
->a_head
.a_dev
;
364 struct bpf_d
*d
= dev
->si_drv1
;
366 funsetown(d
->bd_sigio
);
368 if (d
->bd_state
== BPF_WAITING
)
369 callout_stop(&d
->bd_callout
);
370 d
->bd_state
= BPF_IDLE
;
371 if (d
->bd_bif
!= NULL
)
376 if (dev
->si_uminor
>= BPF_PREALLOCATED_UNITS
) {
377 devfs_clone_bitmap_put(&DEVFS_CLONE_BITMAP(bpf
), dev
->si_uminor
);
385 * Rotate the packet buffers in descriptor d. Move the store buffer
386 * into the hold slot, and the free buffer into the store slot.
387 * Zero the length of the new store buffer.
389 #define ROTATE_BUFFERS(d) \
390 (d)->bd_hbuf = (d)->bd_sbuf; \
391 (d)->bd_hlen = (d)->bd_slen; \
392 (d)->bd_sbuf = (d)->bd_fbuf; \
396 * bpfread - read next chunk of packets from buffers
399 bpfread(struct dev_read_args
*ap
)
401 cdev_t dev
= ap
->a_head
.a_dev
;
402 struct bpf_d
*d
= dev
->si_drv1
;
407 * Restrict application to use a buffer the same size as
410 if (ap
->a_uio
->uio_resid
!= d
->bd_bufsize
)
414 if (d
->bd_state
== BPF_WAITING
)
415 callout_stop(&d
->bd_callout
);
416 timed_out
= (d
->bd_state
== BPF_TIMED_OUT
);
417 d
->bd_state
= BPF_IDLE
;
419 * If the hold buffer is empty, then do a timed sleep, which
420 * ends when the timeout expires or when enough packets
421 * have arrived to fill the store buffer.
423 while (d
->bd_hbuf
== NULL
) {
424 if ((d
->bd_immediate
|| timed_out
) && d
->bd_slen
!= 0) {
426 * A packet(s) either arrived since the previous
427 * read or arrived while we were asleep.
428 * Rotate the buffers and return what's here.
435 * No data is available, check to see if the bpf device
436 * is still pointed at a real interface. If not, return
437 * ENXIO so that the userland process knows to rebind
438 * it before using it again.
440 if (d
->bd_bif
== NULL
) {
445 if (ap
->a_ioflag
& IO_NDELAY
) {
449 error
= tsleep(d
, PCATCH
, "bpf", d
->bd_rtout
);
450 if (error
== EINTR
|| error
== ERESTART
) {
454 if (error
== EWOULDBLOCK
) {
456 * On a timeout, return what's in the buffer,
457 * which may be nothing. If there is something
458 * in the store buffer, we can rotate the buffers.
462 * We filled up the buffer in between
463 * getting the timeout and arriving
464 * here, so we don't need to rotate.
468 if (d
->bd_slen
== 0) {
477 * At this point, we know we have something in the hold slot.
482 * Move data from hold buffer into user space.
483 * We know the entire buffer is transferred since
484 * we checked above that the read buffer is bpf_bufsize bytes.
486 error
= uiomove(d
->bd_hbuf
, d
->bd_hlen
, ap
->a_uio
);
489 d
->bd_fbuf
= d
->bd_hbuf
;
499 * If there are processes sleeping on this descriptor, wake them up.
502 bpf_wakeup(struct bpf_d
*d
)
504 if (d
->bd_state
== BPF_WAITING
) {
505 callout_stop(&d
->bd_callout
);
506 d
->bd_state
= BPF_IDLE
;
509 if (d
->bd_async
&& d
->bd_sig
&& d
->bd_sigio
)
510 pgsigio(d
->bd_sigio
, d
->bd_sig
, 0);
513 selwakeup(&d
->bd_sel
);
516 d
->bd_sel
.si_pid
= 0;
520 bpf_timed_out(void *arg
)
522 struct bpf_d
*d
= (struct bpf_d
*)arg
;
525 if (d
->bd_state
== BPF_WAITING
) {
526 d
->bd_state
= BPF_TIMED_OUT
;
534 bpf_output_dispatch(struct netmsg
*nmsg
)
536 struct netmsg_bpf_output
*bmsg
= (struct netmsg_bpf_output
*)nmsg
;
537 struct ifnet
*ifp
= bmsg
->nm_ifp
;
541 * The driver frees the mbuf.
543 error
= ifp
->if_output(ifp
, bmsg
->nm_mbuf
, bmsg
->nm_dst
, NULL
);
544 lwkt_replymsg(&nmsg
->nm_lmsg
, error
);
548 bpfwrite(struct dev_write_args
*ap
)
550 cdev_t dev
= ap
->a_head
.a_dev
;
551 struct bpf_d
*d
= dev
->si_drv1
;
557 struct netmsg_bpf_output bmsg
;
559 if (d
->bd_bif
== NULL
)
562 ifp
= d
->bd_bif
->bif_ifp
;
564 if (ap
->a_uio
->uio_resid
== 0)
567 error
= bpf_movein(ap
->a_uio
, (int)d
->bd_bif
->bif_dlt
, &m
,
568 &dst
, &datlen
, d
->bd_wfilter
);
572 if (datlen
> ifp
->if_mtu
) {
578 dst
.sa_family
= pseudo_AF_HDRCMPLT
;
580 netmsg_init(&bmsg
.nm_netmsg
, NULL
, &curthread
->td_msgport
,
581 MSGF_MPSAFE
, bpf_output_dispatch
);
586 return lwkt_domsg(cpu_portfn(0), &bmsg
.nm_netmsg
.nm_lmsg
, 0);
590 * Reset a descriptor by flushing its packet buffer and clearing the
591 * receive and drop counts. Should be called at splimp.
594 bpf_resetd(struct bpf_d
*d
)
597 /* Free the hold buffer. */
598 d
->bd_fbuf
= d
->bd_hbuf
;
608 * FIONREAD Check for read packet available.
609 * SIOCGIFADDR Get interface address - convenient hook to driver.
610 * BIOCGBLEN Get buffer len [for read()].
611 * BIOCSETF Set ethernet read filter.
612 * BIOCSETWF Set ethernet write filter.
613 * BIOCFLUSH Flush read packet buffer.
614 * BIOCPROMISC Put interface into promiscuous mode.
615 * BIOCGDLT Get link layer type.
616 * BIOCGETIF Get interface name.
617 * BIOCSETIF Set interface.
618 * BIOCSRTIMEOUT Set read timeout.
619 * BIOCGRTIMEOUT Get read timeout.
620 * BIOCGSTATS Get packet stats.
621 * BIOCIMMEDIATE Set immediate mode.
622 * BIOCVERSION Get filter language version.
623 * BIOCGHDRCMPLT Get "header already complete" flag
624 * BIOCSHDRCMPLT Set "header already complete" flag
625 * BIOCGSEESENT Get "see packets sent" flag
626 * BIOCSSEESENT Set "see packets sent" flag
627 * BIOCLOCK Set "locked" flag
631 bpfioctl(struct dev_ioctl_args
*ap
)
633 cdev_t dev
= ap
->a_head
.a_dev
;
634 struct bpf_d
*d
= dev
->si_drv1
;
638 if (d
->bd_state
== BPF_WAITING
)
639 callout_stop(&d
->bd_callout
);
640 d
->bd_state
= BPF_IDLE
;
643 if (d
->bd_locked
== 1) {
671 * Check for read packet available.
683 *(int *)ap
->a_data
= n
;
691 if (d
->bd_bif
== NULL
) {
694 ifp
= d
->bd_bif
->bif_ifp
;
695 ifnet_serialize_all(ifp
);
696 error
= ifp
->if_ioctl(ifp
, ap
->a_cmd
,
697 ap
->a_data
, ap
->a_cred
);
698 ifnet_deserialize_all(ifp
);
704 * Get buffer len [for read()].
707 *(u_int
*)ap
->a_data
= d
->bd_bufsize
;
714 if (d
->bd_bif
!= NULL
) {
717 u_int size
= *(u_int
*)ap
->a_data
;
719 if (size
> bpf_maxbufsize
)
720 *(u_int
*)ap
->a_data
= size
= bpf_maxbufsize
;
721 else if (size
< BPF_MINBUFSIZE
)
722 *(u_int
*)ap
->a_data
= size
= BPF_MINBUFSIZE
;
723 d
->bd_bufsize
= size
;
728 * Set link layer read filter.
732 error
= bpf_setf(d
, (struct bpf_program
*)ap
->a_data
,
737 * Flush read packet buffer.
746 * Put interface into promiscuous mode.
749 if (d
->bd_bif
== NULL
) {
751 * No interface attached yet.
757 if (d
->bd_promisc
== 0) {
758 error
= ifpromisc(d
->bd_bif
->bif_ifp
, 1);
766 * Get device parameters.
769 if (d
->bd_bif
== NULL
)
772 *(u_int
*)ap
->a_data
= d
->bd_bif
->bif_dlt
;
776 * Get a list of supported data link types.
779 if (d
->bd_bif
== NULL
) {
782 error
= bpf_getdltlist(d
,
783 (struct bpf_dltlist
*)ap
->a_data
);
788 * Set data link type.
791 if (d
->bd_bif
== NULL
)
794 error
= bpf_setdlt(d
, *(u_int
*)ap
->a_data
);
798 * Get interface name.
801 if (d
->bd_bif
== NULL
) {
804 struct ifnet
*const ifp
= d
->bd_bif
->bif_ifp
;
805 struct ifreq
*const ifr
= (struct ifreq
*)ap
->a_data
;
807 strlcpy(ifr
->ifr_name
, ifp
->if_xname
,
808 sizeof ifr
->ifr_name
);
816 error
= bpf_setif(d
, (struct ifreq
*)ap
->a_data
);
824 struct timeval
*tv
= (struct timeval
*)ap
->a_data
;
827 * Subtract 1 tick from tvtohz() since this isn't
830 if ((error
= itimerfix(tv
)) == 0)
831 d
->bd_rtout
= tvtohz_low(tv
);
840 struct timeval
*tv
= (struct timeval
*)ap
->a_data
;
842 tv
->tv_sec
= d
->bd_rtout
/ hz
;
843 tv
->tv_usec
= (d
->bd_rtout
% hz
) * ustick
;
852 struct bpf_stat
*bs
= (struct bpf_stat
*)ap
->a_data
;
854 bs
->bs_recv
= d
->bd_rcount
;
855 bs
->bs_drop
= d
->bd_dcount
;
860 * Set immediate mode.
863 d
->bd_immediate
= *(u_int
*)ap
->a_data
;
868 struct bpf_version
*bv
= (struct bpf_version
*)ap
->a_data
;
870 bv
->bv_major
= BPF_MAJOR_VERSION
;
871 bv
->bv_minor
= BPF_MINOR_VERSION
;
876 * Get "header already complete" flag
879 *(u_int
*)ap
->a_data
= d
->bd_hdrcmplt
;
883 * Set "header already complete" flag
886 d
->bd_hdrcmplt
= *(u_int
*)ap
->a_data
? 1 : 0;
890 * Get "see sent packets" flag
893 *(u_int
*)ap
->a_data
= d
->bd_seesent
;
897 * Set "see sent packets" flag
900 d
->bd_seesent
= *(u_int
*)ap
->a_data
;
903 case FIOASYNC
: /* Send signal on receive packets */
904 d
->bd_async
= *(int *)ap
->a_data
;
908 error
= fsetown(*(int *)ap
->a_data
, &d
->bd_sigio
);
912 *(int *)ap
->a_data
= fgetown(d
->bd_sigio
);
915 /* This is deprecated, FIOSETOWN should be used instead. */
917 error
= fsetown(-(*(int *)ap
->a_data
), &d
->bd_sigio
);
920 /* This is deprecated, FIOGETOWN should be used instead. */
922 *(int *)ap
->a_data
= -fgetown(d
->bd_sigio
);
925 case BIOCSRSIG
: /* Set receive signal */
929 sig
= *(u_int
*)ap
->a_data
;
938 *(u_int
*)ap
->a_data
= d
->bd_sig
;
948 * Set d's packet filter program to fp. If this file already has a filter,
949 * free it and replace it. Returns EINVAL for bogus requests.
952 bpf_setf(struct bpf_d
*d
, struct bpf_program
*fp
, u_long cmd
)
954 struct bpf_insn
*fcode
, *old
;
955 u_int wfilter
, flen
, size
;
957 if (cmd
== BIOCSETWF
) {
964 if (fp
->bf_insns
== NULL
) {
969 d
->bd_wfilter
= NULL
;
971 d
->bd_rfilter
= NULL
;
979 if (flen
> BPF_MAXINSNS
)
982 size
= flen
* sizeof *fp
->bf_insns
;
983 fcode
= (struct bpf_insn
*)kmalloc(size
, M_BPF
, M_WAITOK
);
984 if (copyin(fp
->bf_insns
, fcode
, size
) == 0 &&
985 bpf_validate(fcode
, (int)flen
)) {
988 d
->bd_wfilter
= fcode
;
990 d
->bd_rfilter
= fcode
;
1003 * Detach a file from its current interface (if attached at all) and attach
1004 * to the interface indicated by the name stored in ifr.
1005 * Return an errno or 0.
1008 bpf_setif(struct bpf_d
*d
, struct ifreq
*ifr
)
1012 struct ifnet
*theywant
;
1014 theywant
= ifunit(ifr
->ifr_name
);
1015 if (theywant
== NULL
)
1019 * Look through attached interfaces for the named one.
1021 for (bp
= bpf_iflist
; bp
!= NULL
; bp
= bp
->bif_next
) {
1022 struct ifnet
*ifp
= bp
->bif_ifp
;
1024 if (ifp
== NULL
|| ifp
!= theywant
)
1026 /* skip additional entry */
1027 if (bp
->bif_driverp
!= &ifp
->if_bpf
)
1030 * We found the requested interface.
1031 * If it's not up, return an error.
1032 * Allocate the packet buffers if we need to.
1033 * If we're already attached to requested interface,
1034 * just flush the buffer.
1036 if (!(ifp
->if_flags
& IFF_UP
))
1039 if (d
->bd_sbuf
== NULL
) {
1040 error
= bpf_allocbufs(d
);
1045 if (bp
!= d
->bd_bif
) {
1046 if (d
->bd_bif
!= NULL
) {
1048 * Detach if attached to something else.
1065 * Support for select() and poll() system calls
1067 * Return true iff the specific operation will not block indefinitely.
1068 * Otherwise, return false but make a note that a selwakeup() must be done.
1071 bpfpoll(struct dev_poll_args
*ap
)
1073 cdev_t dev
= ap
->a_head
.a_dev
;
1078 if (d
->bd_bif
== NULL
)
1081 revents
= ap
->a_events
& (POLLOUT
| POLLWRNORM
);
1083 if (ap
->a_events
& (POLLIN
| POLLRDNORM
)) {
1085 * An imitation of the FIONREAD ioctl code.
1086 * XXX not quite. An exact imitation:
1087 * if (d->b_slen != 0 ||
1088 * (d->bd_hbuf != NULL && d->bd_hlen != 0)
1090 if (d
->bd_hlen
!= 0 ||
1091 ((d
->bd_immediate
|| d
->bd_state
== BPF_TIMED_OUT
) &&
1093 revents
|= ap
->a_events
& (POLLIN
| POLLRDNORM
);
1095 selrecord(curthread
, &d
->bd_sel
);
1096 /* Start the read timeout if necessary. */
1097 if (d
->bd_rtout
> 0 && d
->bd_state
== BPF_IDLE
) {
1098 callout_reset(&d
->bd_callout
, d
->bd_rtout
,
1100 d
->bd_state
= BPF_WAITING
;
1105 ap
->a_events
= revents
;
1110 * Process the packet pkt of length pktlen. The packet is parsed
1111 * by each listener's filter, and if accepted, stashed into the
1112 * corresponding buffer.
1115 bpf_tap(struct bpf_if
*bp
, u_char
*pkt
, u_int pktlen
)
1131 * Note that the ipl does not have to be raised at this point.
1132 * The only problem that could arise here is that if two different
1133 * interfaces shared any data. This is not the case.
1135 SLIST_FOREACH(d
, &bp
->bif_dlist
, bd_next
) {
1137 slen
= bpf_filter(d
->bd_rfilter
, pkt
, pktlen
, pktlen
);
1143 catchpacket(d
, pkt
, pktlen
, slen
, ovbcopy
, &tv
);
1151 * Copy data from an mbuf chain into a buffer. This code is derived
1152 * from m_copydata in sys/uipc_mbuf.c.
1155 bpf_mcopy(const void *src_arg
, void *dst_arg
, size_t len
)
1157 const struct mbuf
*m
;
1166 count
= min(m
->m_len
, len
);
1167 bcopy(mtod(m
, void *), dst
, count
);
1175 * Process the packet in the mbuf chain m. The packet is parsed by each
1176 * listener's filter, and if accepted, stashed into the corresponding
1180 bpf_mtap(struct bpf_if
*bp
, struct mbuf
*m
)
1195 /* Don't compute pktlen, if no descriptor is attached. */
1196 if (SLIST_EMPTY(&bp
->bif_dlist
)) {
1201 pktlen
= m_lengthm(m
, NULL
);
1203 SLIST_FOREACH(d
, &bp
->bif_dlist
, bd_next
) {
1204 if (!d
->bd_seesent
&& (m
->m_pkthdr
.rcvif
== NULL
))
1207 slen
= bpf_filter(d
->bd_rfilter
, (u_char
*)m
, pktlen
, 0);
1213 catchpacket(d
, (u_char
*)m
, pktlen
, slen
, bpf_mcopy
,
1222 bpf_mtap_family(struct bpf_if
*bp
, struct mbuf
*m
, sa_family_t family
)
1226 KKASSERT(family
!= AF_UNSPEC
);
1228 family4
= (u_int
)family
;
1229 bpf_ptap(bp
, m
, &family4
, sizeof(family4
));
1233 * Process the packet in the mbuf chain m with the header in m prepended.
1234 * The packet is parsed by each listener's filter, and if accepted,
1235 * stashed into the corresponding buffer.
1238 bpf_ptap(struct bpf_if
*bp
, struct mbuf
*m
, const void *data
, u_int dlen
)
1243 * Craft on-stack mbuf suitable for passing to bpf_mtap.
1244 * Note that we cut corners here; we only setup what's
1245 * absolutely needed--this mbuf should never go anywhere else.
1248 mb
.m_data
= __DECONST(void *, data
); /* LINTED */
1250 mb
.m_pkthdr
.rcvif
= m
->m_pkthdr
.rcvif
;
1256 * Move the packet data from interface memory (pkt) into the
1257 * store buffer. Return 1 if it's time to wakeup a listener (buffer full),
1258 * otherwise 0. "copy" is the routine called to do the actual data
1259 * transfer. bcopy is passed in to copy contiguous chunks, while
1260 * bpf_mcopy is passed in to copy mbuf chains. In the latter case,
1261 * pkt is really an mbuf.
1264 catchpacket(struct bpf_d
*d
, u_char
*pkt
, u_int pktlen
, u_int snaplen
,
1265 void (*cpfn
)(const void *, void *, size_t),
1266 const struct timeval
*tv
)
1270 int hdrlen
= d
->bd_bif
->bif_hdrlen
;
1272 * Figure out how many bytes to move. If the packet is
1273 * greater or equal to the snapshot length, transfer that
1274 * much. Otherwise, transfer the whole packet (unless
1275 * we hit the buffer size limit).
1277 totlen
= hdrlen
+ min(snaplen
, pktlen
);
1278 if (totlen
> d
->bd_bufsize
)
1279 totlen
= d
->bd_bufsize
;
1282 * Round up the end of the previous packet to the next longword.
1284 curlen
= BPF_WORDALIGN(d
->bd_slen
);
1285 if (curlen
+ totlen
> d
->bd_bufsize
) {
1287 * This packet will overflow the storage buffer.
1288 * Rotate the buffers if we can, then wakeup any
1291 if (d
->bd_fbuf
== NULL
) {
1293 * We haven't completed the previous read yet,
1294 * so drop the packet.
1302 } else if (d
->bd_immediate
|| d
->bd_state
== BPF_TIMED_OUT
) {
1304 * Immediate mode is set, or the read timeout has
1305 * already expired during a select call. A packet
1306 * arrived, so the reader should be woken up.
1312 * Append the bpf header.
1314 hp
= (struct bpf_hdr
*)(d
->bd_sbuf
+ curlen
);
1315 hp
->bh_tstamp
= *tv
;
1316 hp
->bh_datalen
= pktlen
;
1317 hp
->bh_hdrlen
= hdrlen
;
1319 * Copy the packet data into the store buffer and update its length.
1321 (*cpfn
)(pkt
, (u_char
*)hp
+ hdrlen
, (hp
->bh_caplen
= totlen
- hdrlen
));
1322 d
->bd_slen
= curlen
+ totlen
;
1326 * Initialize all nonzero fields of a descriptor.
1329 bpf_allocbufs(struct bpf_d
*d
)
1331 d
->bd_fbuf
= kmalloc(d
->bd_bufsize
, M_BPF
, M_WAITOK
);
1332 d
->bd_sbuf
= kmalloc(d
->bd_bufsize
, M_BPF
, M_WAITOK
);
1339 * Free buffers and packet filter program currently in use by a descriptor.
1343 bpf_freed(struct bpf_d
*d
)
1346 * We don't need to lock out interrupts since this descriptor has
1347 * been detached from its interface and it yet hasn't been marked
1350 if (d
->bd_sbuf
!= NULL
) {
1351 kfree(d
->bd_sbuf
, M_BPF
);
1352 if (d
->bd_hbuf
!= NULL
)
1353 kfree(d
->bd_hbuf
, M_BPF
);
1354 if (d
->bd_fbuf
!= NULL
)
1355 kfree(d
->bd_fbuf
, M_BPF
);
1358 kfree(d
->bd_rfilter
, M_BPF
);
1360 kfree(d
->bd_wfilter
, M_BPF
);
1364 * Attach an interface to bpf. ifp is a pointer to the structure
1365 * defining the interface to be attached, dlt is the link layer type,
1366 * and hdrlen is the fixed size of the link header (variable length
1367 * headers are not yet supported).
1370 bpfattach(struct ifnet
*ifp
, u_int dlt
, u_int hdrlen
)
1372 bpfattach_dlt(ifp
, dlt
, hdrlen
, &ifp
->if_bpf
);
1376 bpfattach_dlt(struct ifnet
*ifp
, u_int dlt
, u_int hdrlen
, struct bpf_if
**driverp
)
1380 bp
= kmalloc(sizeof *bp
, M_BPF
, M_WAITOK
| M_ZERO
);
1382 SLIST_INIT(&bp
->bif_dlist
);
1385 bp
->bif_driverp
= driverp
;
1386 *bp
->bif_driverp
= NULL
;
1388 bp
->bif_next
= bpf_iflist
;
1392 * Compute the length of the bpf header. This is not necessarily
1393 * equal to SIZEOF_BPF_HDR because we want to insert spacing such
1394 * that the network layer header begins on a longword boundary (for
1395 * performance reasons and to alleviate alignment restrictions).
1397 bp
->bif_hdrlen
= BPF_WORDALIGN(hdrlen
+ SIZEOF_BPF_HDR
) - hdrlen
;
1400 if_printf(ifp
, "bpf attached\n");
1404 * Detach bpf from an interface. This involves detaching each descriptor
1405 * associated with the interface, and leaving bd_bif NULL. Notify each
1406 * descriptor as it's detached so that any sleepers wake up and get
1410 bpfdetach(struct ifnet
*ifp
)
1412 struct bpf_if
*bp
, *bp_prev
;
1417 /* Locate BPF interface information */
1419 for (bp
= bpf_iflist
; bp
!= NULL
; bp
= bp
->bif_next
) {
1420 if (ifp
== bp
->bif_ifp
)
1425 /* Interface wasn't attached */
1426 if (bp
->bif_ifp
== NULL
) {
1428 kprintf("bpfdetach: %s was not attached\n", ifp
->if_xname
);
1432 while ((d
= SLIST_FIRST(&bp
->bif_dlist
)) != NULL
) {
1437 if (bp_prev
!= NULL
)
1438 bp_prev
->bif_next
= bp
->bif_next
;
1440 bpf_iflist
= bp
->bif_next
;
1448 * Get a list of available data link type of the interface.
1451 bpf_getdltlist(struct bpf_d
*d
, struct bpf_dltlist
*bfl
)
1457 ifp
= d
->bd_bif
->bif_ifp
;
1460 for (bp
= bpf_iflist
; bp
!= NULL
; bp
= bp
->bif_next
) {
1461 if (bp
->bif_ifp
!= ifp
)
1463 if (bfl
->bfl_list
!= NULL
) {
1464 if (n
>= bfl
->bfl_len
) {
1467 error
= copyout(&bp
->bif_dlt
,
1468 bfl
->bfl_list
+ n
, sizeof(u_int
));
1477 * Set the data link type of a BPF instance.
1480 bpf_setdlt(struct bpf_d
*d
, u_int dlt
)
1482 int error
, opromisc
;
1486 if (d
->bd_bif
->bif_dlt
== dlt
)
1488 ifp
= d
->bd_bif
->bif_ifp
;
1489 for (bp
= bpf_iflist
; bp
!= NULL
; bp
= bp
->bif_next
) {
1490 if (bp
->bif_ifp
== ifp
&& bp
->bif_dlt
== dlt
)
1494 opromisc
= d
->bd_promisc
;
1500 error
= ifpromisc(bp
->bif_ifp
, 1);
1502 if_printf(bp
->bif_ifp
,
1503 "bpf_setdlt: ifpromisc failed (%d)\n",
1511 return(bp
== NULL
? EINVAL
: 0);
1515 bpf_drvinit(void *unused
)
1519 make_autoclone_dev(&bpf_ops
, &DEVFS_CLONE_BITMAP(bpf
),
1520 bpfclone
, 0, 0, 0600, "bpf");
1521 for (i
= 0; i
< BPF_PREALLOCATED_UNITS
; i
++) {
1522 make_dev(&bpf_ops
, i
, 0, 0, 0600, "bpf%d", i
);
1523 devfs_clone_bitmap_set(&DEVFS_CLONE_BITMAP(bpf
), i
);
1528 bpf_drvuninit(void *unused
)
1530 devfs_clone_handler_del("bpf");
1531 dev_ops_remove_all(&bpf_ops
);
1532 devfs_clone_bitmap_uninit(&DEVFS_CLONE_BITMAP(bpf
));
1535 SYSINIT(bpfdev
,SI_SUB_DRIVERS
,SI_ORDER_MIDDLE
+CDEV_MAJOR
,bpf_drvinit
,NULL
)
1536 SYSUNINIT(bpfdev
, SI_SUB_DRIVERS
,SI_ORDER_MIDDLE
+CDEV_MAJOR
,bpf_drvuninit
, NULL
);
1540 * NOP stubs to allow bpf-using drivers to load and function.
1542 * A 'better' implementation would allow the core bpf functionality
1543 * to be loaded at runtime.
1547 bpf_tap(struct bpf_if
*bp
, u_char
*pkt
, u_int pktlen
)
1552 bpf_mtap(struct bpf_if
*bp
, struct mbuf
*m
)
1557 bpf_ptap(struct bpf_if
*bp
, struct mbuf
*m
, const void *data
, u_int dlen
)
1562 bpfattach(struct ifnet
*ifp
, u_int dlt
, u_int hdrlen
)
1567 bpfattach_dlt(struct ifnet
*ifp
, u_int dlt
, u_int hdrlen
, struct bpf_if
**driverp
)
1572 bpfdetach(struct ifnet
*ifp
)
1577 bpf_filter(const struct bpf_insn
*pc
, u_char
*p
, u_int wirelen
, u_int buflen
)
1579 return -1; /* "no filter" behaviour */