1 .\" Copyright (c) 2007 Seccuris Inc.
2 .\" All rights reserved.
4 .\" This sofware was developed by Robert N. M. Watson under contract to
7 .\" Redistribution and use in source and binary forms, with or without
8 .\" modification, are permitted provided that the following conditions
10 .\" 1. Redistributions of source code must retain the above copyright
11 .\" notice, this list of conditions and the following disclaimer.
12 .\" 2. Redistributions in binary form must reproduce the above copyright
13 .\" notice, this list of conditions and the following disclaimer in the
14 .\" documentation and/or other materials provided with the distribution.
16 .\" THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
17 .\" ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
18 .\" IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
19 .\" ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
20 .\" FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
21 .\" DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
22 .\" OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
23 .\" HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
24 .\" LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
25 .\" OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
28 .\" Copyright (c) 1990 The Regents of the University of California.
29 .\" All rights reserved.
31 .\" Redistribution and use in source and binary forms, with or without
32 .\" modification, are permitted provided that: (1) source code distributions
33 .\" retain the above copyright notice and this paragraph in its entirety, (2)
34 .\" distributions including binary code include the above copyright notice and
35 .\" this paragraph in its entirety in the documentation or other materials
36 .\" provided with the distribution, and (3) all advertising materials mentioning
37 .\" features or use of this software display the following acknowledgement:
38 .\" ``This product includes software developed by the University of California,
39 .\" Lawrence Berkeley Laboratory and its contributors.'' Neither the name of
40 .\" the University nor the names of its contributors may be used to endorse
41 .\" or promote products derived from this software without specific prior
42 .\" written permission.
43 .\" THIS SOFTWARE IS PROVIDED ``AS IS'' AND WITHOUT ANY EXPRESS OR IMPLIED
44 .\" WARRANTIES, INCLUDING, WITHOUT LIMITATION, THE IMPLIED WARRANTIES OF
45 .\" MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE.
47 .\" This document is derived in part from the enet man page (enet.4)
48 .\" distributed with 4.3BSD Unix.
57 .Nd Berkeley Packet Filter
61 The Berkeley Packet Filter
62 provides a raw interface to data link layers in a protocol
64 All packets on the network, even those destined for other hosts,
65 are accessible through this mechanism.
67 The packet filter appears as a character special device,
71 After opening the device, the file descriptor must be bound to a
72 specific network interface with the
75 A given interface can be shared by multiple listeners, and the filter
76 underlying each descriptor will see an identical packet stream.
78 A separate device file is required for each minor device.
79 If a file is in use, the open will fail and
84 Associated with each open instance of a
86 file is a user-settable packet filter.
87 Whenever a packet is received by an interface,
88 all file descriptors listening on that interface apply their filter.
89 Each descriptor that accepts the packet receives its own copy.
91 The packet filter will support any link level protocol that has fixed length
93 Currently, only Ethernet,
97 drivers have been modified to interact with
100 Since packet data is in network byte order, applications should use the
102 macros to extract multi-byte values.
104 A packet can be sent out on the network by writing to a
107 The writes are unbuffered, meaning only one packet can be processed per write.
108 Currently, only writes to Ethernets and
113 devices deliver packet data to the application via memory buffers provided by
115 The buffer mode is set using the
117 ioctl, and read using the
120 .Ss Buffered read mode
123 devices operate in the
124 .Dv BPF_BUFMODE_BUFFER
125 mode, in which packet data is copied explicitly from kernel to user memory
129 The user process will declare a fixed buffer size that will be used both for
130 sizing internal buffers and for all
132 operations on the file.
133 This size is queried using the
135 ioctl, and is set using the
138 Note that an individual packet larger than the buffer size is necessarily
140 .Ss Zero-copy buffer mode
142 devices may also operate in the
143 .Dv BPF_BUFMODE_ZEROCOPY
144 mode, in which packet data is written directly into two user memory buffers
145 by the kernel, avoiding both system call and copying overhead.
146 Buffers are of fixed (and equal) size, page-aligned, and an even multiple of
148 The maximum zero-copy buffer size is returned by the
151 Note that an individual packet larger than the buffer size is necessarily
154 The user process registers two memory buffers using the
156 ioctl, which accepts a
158 pointer as an argument:
168 is a pointer to the userspace address of the first buffer that will be
171 is a pointer to the second buffer.
173 will then cycle between the two buffers as they fill and are acknowledged.
175 Each buffer begins with a fixed-length header to hold synchronization and
176 data length information for the buffer:
178 struct bpf_zbuf_header {
179 volatile u_int bzh_kernel_gen; /* Kernel generation number. */
180 volatile u_int bzh_kernel_len; /* Length of data in the buffer. */
181 volatile u_int bzh_user_gen; /* User generation number. */
182 /* ...padding for future use... */
186 The header structure of each buffer, including all padding, should be zeroed
187 before it is configured using
189 Remaining space in the buffer will be used by the kernel to store packet
190 data, laid out in the same format as with buffered read mode.
192 The kernel and the user process follow a simple acknowledgement protocol via
193 the buffer header to synchronize access to the buffer: when the header
198 hold the same value, the kernel owns the buffer, and when they differ,
199 userspace owns the buffer.
201 While the kernel owns the buffer, the contents are unstable and may change
202 asynchronously; while the user process owns the buffer, its contents are
203 stable and will not be changed until the buffer has been acknowledged.
205 Initializing the buffer headers to all 0's before registering the buffer has
206 the effect of assigning initial ownership of both buffers to the kernel.
207 The kernel signals that a buffer has been assigned to userspace by modifying
209 and userspace acknowledges the buffer and returns it to the kernel by setting
215 In order to avoid caching and memory re-ordering effects, the user process
216 must use atomic operations and memory barriers when checking for and
217 acknowledging buffers:
219 #include <machine/atomic.h>
222 * Return ownership of a buffer to the kernel for reuse.
225 buffer_acknowledge(struct bpf_zbuf_header *bzh)
228 atomic_store_rel_int(&bzh->bzh_user_gen, bzh->bzh_kernel_gen);
232 * Check whether a buffer has been assigned to userspace by the kernel.
233 * Return true if userspace owns the buffer, and false otherwise.
236 buffer_check(struct bpf_zbuf_header *bzh)
239 return (bzh->bzh_user_gen !=
240 atomic_load_acq_int(&bzh->bzh_kernel_gen));
244 The user process may force the assignment of the next buffer, if any data
245 is pending, to userspace using the
248 This allows the user process to retrieve data in a partially filled buffer
249 before the buffer is full, such as following a timeout; the process must
250 recheck for buffer ownership using the header generation numbers, as the
251 buffer will not be assigned to userspace if no data was present.
253 As in the buffered read mode,
258 may be used to sleep awaiting the availbility of a completed buffer.
259 They will return a readable file descriptor when ownership of the next buffer
260 is assigned to user space.
262 In the current implementation, the kernel may assign zero, one, or both
263 buffers to the user process; however, an earlier implementation maintained
264 the invariant that at most one buffer could be assigned to the user process
266 In order to both ensure progress and high performance, user processes should
267 acknowledge a completely processed buffer as quickly as possible, returning
268 it for reuse, and not block waiting on a second buffer while holding another
273 command codes below are defined in
278 #include <sys/types.h>
279 #include <sys/time.h>
280 #include <sys/ioctl.h>
297 the following commands may be applied to any open
300 The (third) argument to
302 should be a pointer to the type indicated.
303 .Bl -tag -width BIOCGETBUFMODE
306 Returns the required buffer length for reads on
311 Sets the buffer length for reads on
314 The buffer must be set before the file is attached to an interface
317 If the requested buffer size cannot be accommodated, the closest
318 allowable size will be set and returned in the argument.
319 A read call will result in
321 if it is passed a buffer that is not this size.
324 Returns the type of the data link layer underlying the attached interface.
326 is returned if no interface has been specified.
327 The device types, prefixed with
332 Forces the interface into promiscuous mode.
333 All packets, not just those destined for the local host, are processed.
334 Since more than one file can be listening on a given interface,
335 a listener that opened its interface non-promiscuously may receive
336 packets promiscuously.
337 This problem can be remedied with an appropriate filter.
339 Flushes the buffer of incoming packets,
340 and resets the statistics that are returned by BIOCGSTATS.
342 .Pq Li "struct ifreq"
343 Returns the name of the hardware interface that the file is listening on.
344 The name is returned in the ifr_name field of
348 All other fields are undefined.
350 .Pq Li "struct ifreq"
351 Sets the hardware interface associate with the file.
353 command must be performed before any packets can be read.
354 The device is indicated by name using the
359 Additionally, performs the actions of
363 .Pq Li "struct timeval"
364 Set or get the read timeout parameter.
366 specifies the length of time to wait before timing
367 out on a read request.
368 This parameter is initialized to zero by
370 indicating no timeout.
372 .Pq Li "struct bpf_stat"
373 Returns the following structure of packet statistics:
376 u_int bs_recv; /* number of packets received */
377 u_int bs_drop; /* number of packets dropped */
382 .Bl -hang -offset indent
384 the number of packets received by the descriptor since opened or reset
385 (including any buffered since the last read call);
388 the number of packets which were accepted by the filter but dropped by the
389 kernel because of buffer overflows
390 (i.e., the application's reads are not keeping up with the packet traffic).
396 based on the truth value of the argument.
397 When immediate mode is enabled, reads return immediately upon packet
399 Otherwise, a read will block until either the kernel buffer
400 becomes full or a timeout occurs.
401 This is useful for programs like
403 which must respond to messages in real time.
404 The default for a new file is off.
407 .Pq Li "struct bpf_program"
408 Sets the read filter program used by the kernel to discard uninteresting
410 An array of instructions and its length is passed in using
411 the following structure:
415 struct bpf_insn *bf_insns;
419 The filter program is pointed to by the
421 field while its length in units of
422 .Sq Li struct bpf_insn
428 for an explanation of the filter language.
429 The only difference between
435 performs the actions of
441 .Pq Li "struct bpf_program"
442 Sets the write filter program used by the kernel to control what type of
443 packets can be written to the interface.
451 .Pq Li "struct bpf_version"
452 Returns the major and minor version numbers of the filter language currently
453 recognized by the kernel.
454 Before installing a filter, applications must check
455 that the current version is compatible with the running kernel.
456 Version numbers are compatible if the major numbers match and the application minor
457 is less than or equal to the kernel minor.
458 The kernel version number is returned in the following structure:
466 The current version numbers are given by
467 .Dv BPF_MAJOR_VERSION
469 .Dv BPF_MINOR_VERSION
472 An incompatible filter
473 may result in undefined behavior (most likely, an error returned by
475 or haphazard packet matching).
479 Set or get the status of the
482 Set to zero if the link level source address should be filled in automatically
483 by the interface output routine.
484 Set to one if the link level source
485 address will be written, as provided, to the wire.
486 This flag is initialized to zero by default.
490 These commands are obsolete but left for compatibility.
496 Set or get the flag determining whether locally generated packets on the
497 interface should be returned by BPF.
498 Set to zero to see only incoming packets on the interface.
499 Set to one to see packets originating locally and remotely on the interface.
500 This flag is initialized to one by default.
501 .It Dv BIOCSDIRECTION
502 .It Dv BIOCGDIRECTION
504 Set or get the setting determining whether incoming, outgoing, or all packets
505 on the interface should be returned by BPF.
508 to see only incoming packets on the interface.
511 to see packets originating locally and remotely on the interface.
514 to see only outgoing packets on the interface.
515 This setting is initialized to
520 Set packet feedback mode.
521 This allows injected packets to be fed back as input to the interface when
522 output via the interface is successful.
525 direction is set, injected outgoing packet is not returned by BPF to avoid
526 duplication. This flag is initialized to zero by default.
528 Set the locked flag on the
531 This prevents the execution of
532 ioctl commands which could change the underlying operating parameters of
534 .It Dv BIOCGETBUFMODE
535 .It Dv BIOCSETBUFMODE
537 Get or set the current
539 buffering mode; possible values are
540 .Dv BPF_BUFMODE_BUFFER ,
541 buffered read mode, and
542 .Dv BPF_BUFMODE_ZBUF ,
543 zero-copy buffer mode.
545 .Pq Li struct bpf_zbuf
546 Set the current zero-copy buffer locations; buffer locations may be
547 set only once zero-copy buffer mode has been selected, and prior to attaching
549 Buffers must be of identical size, page-aligned, and an integer multiple of
557 If buffers have already been set for this device, the ioctl will fail.
560 Get the largest individual zero-copy buffer size allowed.
561 As two buffers are used in zero-copy buffer mode, the limit (in practice) is
562 twice the returned size.
563 As zero-copy buffers consume kernel address space, conservative selection of
564 buffer size is suggested, especially when there are multiple
566 descriptors in use on 32-bit systems.
568 Force ownership of the next buffer to be assigned to userspace, if any data
569 present in the buffer.
570 If no data is present, the buffer will remain owned by the kernel.
571 This allows consumers of zero-copy buffering to implement timeouts and
572 retrieve partially filled buffers.
573 In order to handle the case where no data is present in the buffer and
574 therefore ownership is not assigned, the user process must check
580 The following structure is prepended to each packet returned by
582 or via a zero-copy buffer:
585 struct timeval bh_tstamp; /* time stamp */
586 u_long bh_caplen; /* length of captured portion */
587 u_long bh_datalen; /* original length of packet */
588 u_short bh_hdrlen; /* length of bpf header (this struct
589 plus alignment padding */
593 The fields, whose values are stored in host order, and are:
595 .Bl -tag -compact -width bh_datalen
597 The time at which the packet was processed by the packet filter.
599 The length of the captured portion of the packet.
600 This is the minimum of
601 the truncation amount specified by the filter and the length of the packet.
603 The length of the packet off the wire.
604 This value is independent of the truncation amount specified by the filter.
608 header, which may not be equal to
609 .\" XXX - not really a function call
610 .Fn sizeof "struct bpf_hdr" .
615 field exists to account for
616 padding between the header and the link level protocol.
617 The purpose here is to guarantee proper alignment of the packet
618 data structures, which is required on alignment sensitive
619 architectures and improves performance on many other architectures.
620 The packet filter insures that the
622 and the network layer
623 header will be word aligned.
625 must be taken when accessing the link layer protocol fields on alignment
627 (This is not a problem on an Ethernet, since
628 the type field is a short falling on an even offset,
629 and the addresses are probably accessed in a bytewise fashion).
631 Additionally, individual packets are padded so that each starts
633 This requires that an application
634 has some knowledge of how to get from packet to packet.
641 It rounds up its argument to the nearest word aligned value (where a word is
647 points to the start of a packet, this expression
648 will advance it to the next packet:
649 .Dl p = (char *)p + BPF_WORDALIGN(p->bh_hdrlen + p->bh_caplen)
651 For the alignment mechanisms to work properly, the
654 must itself be word aligned.
658 will always return an aligned buffer.
660 A filter program is an array of instructions, with all branches forwardly
661 directed, terminated by a
664 Each instruction performs some action on the pseudo-machine state,
665 which consists of an accumulator, index register, scratch memory store,
666 and implicit program counter.
668 The following structure defines the instruction format:
680 field is used in different ways by different instructions,
685 fields are used as offsets
686 by the branch instructions.
687 The opcodes are encoded in a semi-hierarchical fashion.
688 There are eight classes of instructions:
698 Various other mode and
699 operator bits are or'd into the class to give the actual instructions.
700 The classes and modes are defined in
703 Below are the semantics for each defined
706 We use the convention that A is the accumulator, X is the index register,
707 P[] packet data, and M[] scratch memory store.
708 P[i:n] gives the data at byte offset
711 interpreted as a word (n=4),
712 unsigned halfword (n=2), or unsigned byte (n=1).
713 M[i] gives the i'th word in the scratch memory store, which is only
714 addressed in word units.
715 The memory store is indexed from 0 to
722 are the corresponding fields in the
723 instruction definition.
725 refers to the length of the packet.
727 .Bl -tag -width BPF_STXx
729 These instructions copy a value into the accumulator.
730 The type of the source operand is specified by an
732 and can be a constant
734 packet data at a fixed offset
736 packet data at a variable offset
740 or a word in the scratch memory store
746 the data size must be specified as a word
752 The semantics of all the recognized
757 BPF_LD+BPF_W+BPF_ABS A <- P[k:4]
758 BPF_LD+BPF_H+BPF_ABS A <- P[k:2]
759 BPF_LD+BPF_B+BPF_ABS A <- P[k:1]
760 BPF_LD+BPF_W+BPF_IND A <- P[X+k:4]
761 BPF_LD+BPF_H+BPF_IND A <- P[X+k:2]
762 BPF_LD+BPF_B+BPF_IND A <- P[X+k:1]
763 BPF_LD+BPF_W+BPF_LEN A <- len
764 BPF_LD+BPF_IMM A <- k
765 BPF_LD+BPF_MEM A <- M[k]
768 These instructions load a value into the index register.
770 the addressing modes are more restrictive than those of the accumulator loads,
773 a hack for efficiently loading the IP header length.
776 BPF_LDX+BPF_W+BPF_IMM X <- k
777 BPF_LDX+BPF_W+BPF_MEM X <- M[k]
778 BPF_LDX+BPF_W+BPF_LEN X <- len
779 BPF_LDX+BPF_B+BPF_MSH X <- 4*(P[k:1]&0xf)
782 This instruction stores the accumulator into the scratch memory.
783 We do not need an addressing mode since there is only one possibility
790 This instruction stores the index register in the scratch memory store.
796 The alu instructions perform operations between the accumulator and
797 index register or constant, and store the result back in the accumulator.
798 For binary operations, a source mode is required
804 BPF_ALU+BPF_ADD+BPF_K A <- A + k
805 BPF_ALU+BPF_SUB+BPF_K A <- A - k
806 BPF_ALU+BPF_MUL+BPF_K A <- A * k
807 BPF_ALU+BPF_DIV+BPF_K A <- A / k
808 BPF_ALU+BPF_AND+BPF_K A <- A & k
809 BPF_ALU+BPF_OR+BPF_K A <- A | k
810 BPF_ALU+BPF_LSH+BPF_K A <- A << k
811 BPF_ALU+BPF_RSH+BPF_K A <- A >> k
812 BPF_ALU+BPF_ADD+BPF_X A <- A + X
813 BPF_ALU+BPF_SUB+BPF_X A <- A - X
814 BPF_ALU+BPF_MUL+BPF_X A <- A * X
815 BPF_ALU+BPF_DIV+BPF_X A <- A / X
816 BPF_ALU+BPF_AND+BPF_X A <- A & X
817 BPF_ALU+BPF_OR+BPF_X A <- A | X
818 BPF_ALU+BPF_LSH+BPF_X A <- A << X
819 BPF_ALU+BPF_RSH+BPF_X A <- A >> X
820 BPF_ALU+BPF_NEG A <- -A
823 The jump instructions alter flow of control.
825 compare the accumulator against a constant
827 or the index register
829 If the result is true (or non-zero),
830 the true branch is taken, otherwise the false branch is taken.
831 Jump offsets are encoded in 8 bits so the longest jump is 256 instructions.
832 However, the jump always
834 opcode uses the 32 bit
836 field as the offset, allowing arbitrarily distant destinations.
837 All conditionals use unsigned comparison conventions.
840 BPF_JMP+BPF_JA pc += k
841 BPF_JMP+BPF_JGT+BPF_K pc += (A > k) ? jt : jf
842 BPF_JMP+BPF_JGE+BPF_K pc += (A >= k) ? jt : jf
843 BPF_JMP+BPF_JEQ+BPF_K pc += (A == k) ? jt : jf
844 BPF_JMP+BPF_JSET+BPF_K pc += (A & k) ? jt : jf
845 BPF_JMP+BPF_JGT+BPF_X pc += (A > X) ? jt : jf
846 BPF_JMP+BPF_JGE+BPF_X pc += (A >= X) ? jt : jf
847 BPF_JMP+BPF_JEQ+BPF_X pc += (A == X) ? jt : jf
848 BPF_JMP+BPF_JSET+BPF_X pc += (A & X) ? jt : jf
851 The return instructions terminate the filter program and specify the amount
852 of packet to accept (i.e., they return the truncation amount).
853 A return value of zero indicates that the packet should be ignored.
854 The return value is either a constant
860 BPF_RET+BPF_A accept A bytes
861 BPF_RET+BPF_K accept k bytes
864 The miscellaneous category was created for anything that does not
865 fit into the above classes, and for any new instructions that might need to
867 Currently, these are the register transfer instructions
868 that copy the index register to the accumulator or vice versa.
871 BPF_MISC+BPF_TAX X <- A
872 BPF_MISC+BPF_TXA A <- X
878 interface provides the following macros to facilitate
880 .Fn BPF_STMT opcode operand
882 .Fn BPF_JUMP opcode operand true_offset false_offset .
884 .Bl -tag -compact -width /dev/bpfXXX
885 .It Pa /dev/bpf Ns Sy n
886 the packet filter device
889 The following filter is taken from the Reverse ARP Daemon.
890 It accepts only Reverse ARP requests.
892 struct bpf_insn insns[] = {
893 BPF_STMT(BPF_LD+BPF_H+BPF_ABS, 12),
894 BPF_JUMP(BPF_JMP+BPF_JEQ+BPF_K, ETHERTYPE_REVARP, 0, 3),
895 BPF_STMT(BPF_LD+BPF_H+BPF_ABS, 20),
896 BPF_JUMP(BPF_JMP+BPF_JEQ+BPF_K, REVARP_REQUEST, 0, 1),
897 BPF_STMT(BPF_RET+BPF_K, sizeof(struct ether_arp) +
898 sizeof(struct ether_header)),
899 BPF_STMT(BPF_RET+BPF_K, 0),
903 This filter accepts only IP packets between host 128.3.112.15 and
906 struct bpf_insn insns[] = {
907 BPF_STMT(BPF_LD+BPF_H+BPF_ABS, 12),
908 BPF_JUMP(BPF_JMP+BPF_JEQ+BPF_K, ETHERTYPE_IP, 0, 8),
909 BPF_STMT(BPF_LD+BPF_W+BPF_ABS, 26),
910 BPF_JUMP(BPF_JMP+BPF_JEQ+BPF_K, 0x8003700f, 0, 2),
911 BPF_STMT(BPF_LD+BPF_W+BPF_ABS, 30),
912 BPF_JUMP(BPF_JMP+BPF_JEQ+BPF_K, 0x80037023, 3, 4),
913 BPF_JUMP(BPF_JMP+BPF_JEQ+BPF_K, 0x80037023, 0, 3),
914 BPF_STMT(BPF_LD+BPF_W+BPF_ABS, 30),
915 BPF_JUMP(BPF_JMP+BPF_JEQ+BPF_K, 0x8003700f, 0, 1),
916 BPF_STMT(BPF_RET+BPF_K, (u_int)-1),
917 BPF_STMT(BPF_RET+BPF_K, 0),
921 Finally, this filter returns only TCP finger packets.
922 We must parse the IP header to reach the TCP header.
926 checks that the IP fragment offset is 0 so we are sure
927 that we have a TCP header.
929 struct bpf_insn insns[] = {
930 BPF_STMT(BPF_LD+BPF_H+BPF_ABS, 12),
931 BPF_JUMP(BPF_JMP+BPF_JEQ+BPF_K, ETHERTYPE_IP, 0, 10),
932 BPF_STMT(BPF_LD+BPF_B+BPF_ABS, 23),
933 BPF_JUMP(BPF_JMP+BPF_JEQ+BPF_K, IPPROTO_TCP, 0, 8),
934 BPF_STMT(BPF_LD+BPF_H+BPF_ABS, 20),
935 BPF_JUMP(BPF_JMP+BPF_JSET+BPF_K, 0x1fff, 6, 0),
936 BPF_STMT(BPF_LDX+BPF_B+BPF_MSH, 14),
937 BPF_STMT(BPF_LD+BPF_H+BPF_IND, 14),
938 BPF_JUMP(BPF_JMP+BPF_JEQ+BPF_K, 79, 2, 0),
939 BPF_STMT(BPF_LD+BPF_H+BPF_IND, 16),
940 BPF_JUMP(BPF_JMP+BPF_JEQ+BPF_K, 79, 0, 1),
941 BPF_STMT(BPF_RET+BPF_K, (u_int)-1),
942 BPF_STMT(BPF_RET+BPF_K, 0),
957 .%T "An efficient, extensible, and portable network monitor"
960 The Enet packet filter was created in 1980 by Mike Accetta and
961 Rick Rashid at Carnegie-Mellon University.
963 Stanford, ported the code to
965 and continued its development from
967 Since then, it has evolved into the Ultrix Packet Filter at
979 of Lawrence Berkeley Laboratory, implemented BPF in
981 Much of the design is due to
984 Support for zero-copy buffers was added by
985 .An Robert N. M. Watson
986 under contract to Seccuris Inc.
988 The read buffer must be of a fixed size (returned by the
992 A file that does not request promiscuous mode may receive promiscuously
993 received packets as a side effect of another file requesting this
994 mode on the same hardware interface.
995 This could be fixed in the kernel with additional processing overhead.
996 However, we favor the model where
997 all files must assume that the interface is promiscuous, and if
998 so desired, must utilize a filter to reject foreign packets.
1000 Data link protocols with variable length headers are not currently supported.
1007 settings have been observed to work incorrectly on some interface
1008 types, including those with hardware loopback rather than software loopback,
1009 and point-to-point interfaces.
1010 They appear to function correctly on a
1011 broad range of Ethernet-style interfaces.