xref: /netbsd-src/share/man/man9/mbuf.9 (revision b1c86f5f087524e68db12794ee9c3e3da1ab17a0)
1.\"	$NetBSD: mbuf.9,v 1.49 2009/03/09 17:28:12 joerg Exp $
2.\"
3.\" Copyright (c) 1997 The NetBSD Foundation, Inc.
4.\" All rights reserved.
5.\"
6.\" This documentation is derived from text contributed to The NetBSD Foundation
7.\" by S.P.Zeidler (aka stargazer).
8.\"
9.\" Redistribution and use in source and binary forms, with or without
10.\" modification, are permitted provided that the following conditions
11.\" are met:
12.\" 1. Redistributions of source code must retain the above copyright
13.\"    notice, this list of conditions and the following disclaimer.
14.\" 2. Redistributions in binary form must reproduce the above copyright
15.\"    notice, this list of conditions and the following disclaimer in the
16.\"    documentation and/or other materials provided with the distribution.
17.\"
18.\" THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
19.\" ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
20.\" TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
21.\" PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
22.\" BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
23.\" CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
24.\" SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
25.\" INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
26.\" CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
27.\" ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
28.\" POSSIBILITY OF SUCH DAMAGE.
29.\"
30.Dd March 24, 2008
31.Dt MBUF 9
32.Os
33.Sh NAME
34.Nm mbuf ,
35.Nm m_get ,
36.Nm m_getclr ,
37.Nm m_gethdr ,
38.Nm m_devget ,
39.Nm m_copym ,
40.Nm m_copypacket ,
41.Nm m_copydata ,
42.Nm m_copyback ,
43.Nm m_copyback_cow ,
44.Nm m_cat ,
45.Nm m_dup ,
46.Nm m_makewritable ,
47.Nm m_prepend ,
48.Nm m_pulldown ,
49.Nm m_pullup ,
50.Nm m_split ,
51.Nm m_adj ,
52.Nm m_apply ,
53.Nm m_free ,
54.Nm m_freem ,
55.Nm mtod ,
56.Nm MGET ,
57.Nm MGETHDR ,
58.Nm MEXTMALLOC ,
59.Nm MEXTADD ,
60.Nm MCLGET ,
61.Nm M_COPY_PKTHDR ,
62.Nm M_MOVE_PKTHDR ,
63.Nm M_ALIGN ,
64.Nm MH_ALIGN ,
65.Nm M_LEADINGSPACE ,
66.Nm M_TRAILINGSPACE ,
67.Nm M_PREPEND ,
68.Nm MCHTYPE ,
69.Nm MFREE
70.Nd "functions and macros for managing memory used by networking code"
71.Sh SYNOPSIS
72.In sys/mbuf.h
73.Ft struct mbuf *
74.Fn m_get "int nowait" "int type"
75.Ft struct mbuf *
76.Fn m_getclr "int nowait" "int type"
77.Ft struct mbuf *
78.Fn m_gethdr "int nowait" "int type"
79.Ft struct mbuf *
80.Fn m_devget "char *buf" "int totlen" "int off0" "struct ifnet *ifp" "void (*copy)(const void *, void *, size_t)"
81.Ft struct mbuf *
82.Fn m_copym "struct mbuf *m" "int off0" "int len" "int wait"
83.Ft struct mbuf *
84.Fn m_copypacket "struct mbuf *m" "int how"
85.Ft void
86.Fn m_copydata "struct mbuf *m" "int off" "int len" "void *cp"
87.Ft void
88.Fn m_copyback "struct mbuf *m0" "int off" "int len" "void *cp"
89.Ft struct mbuf *
90.Fn m_copyback_cow "struct mbuf *m0" "int off" "int len" "void *cp" "int how"
91.Ft int
92.Fn m_makewritable "struct mbuf **mp" "int off" "int len" "int how"
93.Ft void
94.Fn m_cat "struct mbuf *m" "struct mbuf *n"
95.Ft struct mbuf *
96.Fn m_dup "struct mbuf *m" "int off0" "int len" "int wait"
97.Ft struct mbuf *
98.Fn m_prepend "struct mbuf *m" "int len" "int how"
99.Ft struct mbuf *
100.Fn m_pulldown "struct mbuf *m" "int off" "int len" "int *offp"
101.Ft struct mbuf *
102.Fn m_pullup "struct mbuf *n" "int len"
103.Ft struct mbuf *
104.Fn m_split "struct mbuf *m0" "int len0" "int wait"
105.Ft void
106.Fn m_adj "struct mbuf *mp" "int req_len"
107.Ft int
108.Fn m_apply "struct mbuf *m" "int off" "int len" "int *f(void *, void *, unsigned int)" "void *arg"
109.Ft struct mbuf *
110.Fn m_free "struct mbuf *m"
111.Ft void
112.Fn m_freem "struct mbuf *m"
113.Ft datatype
114.Fn mtod "struct mbuf *m" "datatype"
115.Ft void
116.Fn MGET "struct mbuf *m" "int how" "int type"
117.Ft void
118.Fn MGETHDR "struct mbuf *m" "int how" "int type"
119.Ft void
120.Fn MEXTMALLOC "struct mbuf *m" "int len" "int how"
121.Ft void
122.Fn MEXTADD "struct mbuf *m" "void *buf" "int size" "int type" "void (*free)(struct mbuf *, void *, size_t, void *)" "void *arg"
123.Ft void
124.Fn MCLGET "struct mbuf *m" "int how"
125.Ft void
126.Fn M_COPY_PKTHDR "struct mbuf *to" "struct mbuf *from"
127.Ft void
128.Fn M_MOVE_PKTHDR "struct mbuf *to" "struct mbuf *from"
129.Ft void
130.Fn M_ALIGN "struct mbuf *m" "int len"
131.Ft void
132.Fn MH_ALIGN "struct mbuf *m" "int len"
133.Ft int
134.Fn M_LEADINGSPACE "struct mbuf *m"
135.Ft int
136.Fn M_TRAILINGSPACE "struct mbuf *m"
137.Ft void
138.Fn M_PREPEND "struct mbuf *m" "int plen" "int how"
139.Ft void
140.Fn MCHTYPE "struct mbuf *m" "int type"
141.Ft void
142.Fn MFREE "struct mbuf *m" "struct mbuf *n"
143.Sh DESCRIPTION
144The
145.Nm
146functions and macros provide an easy and consistent way to handle
147a networking stack's memory management needs.
148An
149.Nm
150consists of a header and a data area.
151It is of a fixed size,
152.Dv MSIZE
153.Pq defined in Aq Pa machine/param.h ,
154which includes overhead.
155The header contains a pointer to the next
156.Nm
157in the
158.Sq "mbuf chain" ,
159a pointer to the next
160.Sq "mbuf chain" ,
161a pointer to the data area, the amount of data in this mbuf, its type
162and a
163.Dv flags
164field.
165.Pp
166The
167.Dv type
168variable can signify:
169.Bl -tag -compact -offset indent -width "XXXXXXXXXXX"
170.It Dv MT_FREE
171the mbuf should be on the ``free'' list
172.It Dv MT_DATA
173data was dynamically allocated
174.It Dv MT_HEADER
175data is a packet header
176.It Dv MT_SONAME
177data is a socket name
178.It Dv MT_SOOPTS
179data is socket options
180.It Dv MT_FTABLE
181data is the fragment reassembly header
182.It Dv MT_CONTROL
183mbuf contains ancillary \&(protocol control\&) data
184.It Dv MT_OOBDATA
185mbuf contains out-of-band data.
186.El
187.Pp
188The
189.Dv flags
190variable contains information describing the
191.Nm ,
192notably:
193.Bl -tag -compact -offset indent -width "XXXXXXXXXXX"
194.It Dv M_EXT
195has external storage
196.It Dv M_PKTHDR
197is start of record
198.It Dv M_EOR
199is end of record
200.It Dv M_CLUSTER
201external storage is a cluster.
202.El
203.Pp
204If an
205.Nm
206designates the start of a record
207.Pq Dv M_PKTHDR ,
208its
209.Dv flags
210field may contain additional information describing the content of
211the record:
212.Bl -tag -compact -offset indent -width "XXXXXXXXXXX"
213.It Dv M_BCAST
214sent/received as link-level broadcast
215.It Dv M_MCAST
216sent/received as link-level multicast
217.It Dv M_LINK0 ,
218.It Dv M_LINK1 ,
219.It Dv M_LINK2
220three link-level specific flags.
221.El
222.Pp
223An
224.Nm
225may add a single
226.Sq "mbuf cluster"
227of
228.Dv MCLBYTES
229bytes
230.Pq also defined in Aq Pa machine/param.h ,
231which has no additional overhead
232and is used instead of the internal data area; this is done when at least
233.Dv MINCLSIZE
234bytes of data must be stored.
235.Pp
236When the
237.Dv M_EXT
238flag is raised for an mbuf,
239the external storage area could be shared among multiple mbufs.
240Be careful when you attempt to overwrite the data content of the mbuf.
241.Bl -tag -width compact
242.It Fn m_get "int nowait" "int type"
243Allocates an mbuf and initializes it to contain internal data.
244The
245.Fa nowait
246parameter is a choice of
247.Dv M_WAIT / M_DONTWAIT
248from caller.
249.Dv M_WAIT
250means the call cannot fail, but may take forever.
251The
252.Fa type
253parameter is an mbuf type.
254.It Fn m_getclr "int nowait" "int type"
255Allocates an mbuf and initializes it to contain internal data, then
256zeros the data area.
257The
258.Fa nowait
259parameter is a choice of
260.Dv M_WAIT / M_DONTWAIT
261from caller.
262The
263.Fa type
264parameter is an mbuf type.
265.It Fn m_gethdr "int nowait" "int type"
266Allocates an mbuf and initializes it to contain a packet header and internal
267data.
268The
269.Fa nowait
270parameter is a choice of
271.Dv M_WAIT / M_DONTWAIT
272from caller.
273The
274.Fa type
275parameter is an mbuf type.
276.It Fn m_devget "char *buf" "int totlen" "int off0" "struct ifnet *ifp" "void (*copy)(const void *, void *, size_t)"
277Copies
278.Fa len
279bytes from device local memory into mbufs using copy routine
280.Fa copy .
281If parameter
282.Fa off
283is non-zero, the packet is supposed to be trailer-encapsulated and
284.Fa off
285bytes plus the type and length fields will be skipped before copying.
286Returns the top of the mbuf chain it created.
287.It Fn m_copym "struct mbuf *m" "int off0" "int len" "int wait"
288Creates a copy of an mbuf chain starting
289.Fa off0
290bytes from the beginning, continuing for
291.Fa len
292bytes.
293If the
294.Fa len
295requested is
296.Dv M_COPYALL ,
297the complete mbuf chain will be copied.
298The
299.Fa wait
300parameter is a choice of
301.Dv M_WAIT / M_DONTWAIT
302from caller.
303.It Fn m_copypacket "struct mbuf *m" "int how"
304Copies an entire packet, including header (which must be present).
305This function is an optimization of the common case
306.Li m_copym(m, 0, Dv M_COPYALL, Fa how ) .
307.It Fn m_copydata "struct mbuf *m" "int off" "int len" "void *cp"
308Copies
309.Fa len
310bytes data from mbuf chain
311.Fa m
312into the buffer
313.Fa cp ,
314starting
315.Fa off
316bytes from the beginning.
317.It Fn m_copyback "struct mbuf *m0" "int off" "int len" "void *cp"
318Copies
319.Fa len
320bytes data from buffer
321.Fa cp
322back into the mbuf chain
323.Fa m0 ,
324starting
325.Fa off
326bytes from the beginning of the chain, extending the mbuf chain if necessary.
327.Fn m_copyback
328can only fail when extending the chain.
329The caller should check for this kind of failure
330by checking the resulting length of the chain in that case.
331It is an error to use
332.Fn m_copyback
333on read-only mbufs.
334.It Fn m_copyback_cow "struct mbuf *m0" "int off" "int len" "void *cp" \
335"int how"
336Copies
337.Fa len
338bytes data from buffer
339.Fa cp
340back into the mbuf chain
341.Fa m0
342as
343.Fn m_copyback
344does.
345Unlike
346.Fn m_copyback ,
347it is safe to use
348.Fn m_copyback_cow
349on read-only mbufs.
350If needed,
351.Fn m_copyback_cow
352automatically allocates new mbufs and adjusts the chain.
353On success, it returns a pointer to the resulting mbuf chain,
354and frees the original mbuf
355.Fa m0 .
356Otherwise, it returns
357.Dv NULL .
358The
359.Fa how
360parameter is a choice of
361.Dv M_WAIT / M_DONTWAIT
362from the caller.
363Unlike
364.Fn m_copyback ,
365extending the mbuf chain isn't supported.
366It is an error to attempt to extend the mbuf chain using
367.Fn m_copyback_cow .
368.It Fn m_makewritable "struct mbuf **mp" "int off" "int len" "int how"
369Rearranges an mbuf chain so that
370.Fa len
371bytes from offset
372.Fa off
373are writable.
374When it meets read-only mbufs, it allocates new mbufs, adjusts the chain as
375.Fn m_copyback_cow
376does, and copies the original content into them.
377.Fn m_makewritable
378does
379.Em not
380guarantee that all
381.Fa len
382bytes at
383.Fa off
384are consecutive.
385The
386.Fa how
387parameter is a choice of
388.Dv M_WAIT / M_DONTWAIT
389from the caller.
390.Fn m_makewritable
391preserves the contents of the mbuf chain even in the case of failure.
392It updates a pointer to the mbuf chain pointed to by
393.Fa mp .
394It returns 0 on success.
395Otherwise, it returns an error code, typically
396.Er ENOBUFS .
397.It Fn m_cat "struct mbuf *m" "struct mbuf *n"
398Concatenates mbuf chain
399.Fa n
400to
401.Fa m .
402Both chains must be of the same type; packet headers will
403.Em not
404be updated if present.
405.It Fn m_dup "struct mbuf *m" "int off0" "int len" "int wait"
406Similarly to
407.Fn m_copym ,
408the function creates a copy of an mbuf chain starting
409.Fa off0
410bytes from the beginning, continuing for
411.Fa len
412bytes.
413While
414.Fn m_copym
415tries to share external storage for mbufs with
416.Dv M_EXT
417flag,
418.Fn m_dup
419will deep-copy the whole data content into new mbuf chain
420and avoids shared external storage.
421.It Fn m_prepend "struct mbuf *m" "int len" "int how"
422Lesser-used path for
423.Fn M_PREPEND :
424allocates new mbuf
425.Fa m
426of size
427.Fa len
428to prepend to the chain, copying junk along.
429The
430.Fa how
431parameter is a choice of
432.Dv M_WAIT / M_DONTWAIT
433from caller.
434.It Fn m_pulldown "struct mbuf *m" "int off" "int len" "int *offp"
435Rearranges an mbuf chain so that
436.Fa len
437bytes from offset
438.Fa off
439are contiguous and in the data area of an mbuf.
440The return value points to an mbuf in the middle of the mbuf chain
441.Fa m .
442If we call the return value
443.Fa n ,
444the contiguous data region is available at
445.Li "mtod(n, void *) + *offp" ,
446or
447.Li "mtod(n, void *)"
448if
449.Fa offp
450is
451.Dv NULL .
452The top of the mbuf chain
453.Fa m ,
454and mbufs up to
455.Fa off ,
456will not be modified.
457On successful return, it is guaranteed that the mbuf pointed to by
458.Fa n
459does not have a shared external storage,
460therefore it is safe to update the contiguous region.
461Returns
462.Dv NULL
463and frees the mbuf chain on failure.
464.Fa len
465must be smaller or equal than
466.Dv MCLBYTES .
467.It Fn m_pullup "struct mbuf *m" "int len"
468Rearranges an mbuf chain so that
469.Fa len
470bytes are contiguous
471and in the data area of an mbuf (so that
472.Fn mtod
473will work for a structure of size
474.Fa len ) .
475Returns the resulting
476mbuf chain on success, frees it and returns
477.Dv NULL
478on failure.
479If there is room, it will add up to
480.Dv max_protohdr
481-
482.Fa len
483extra bytes to the
484contiguous region to possibly avoid being called again.
485.Fa len
486must be smaller or equal than
487.Dv MHLEN .
488.It Fn m_split "struct mbuf *m0" "int len0" "int wait"
489Partitions an mbuf chain in two pieces, returning the tail,
490which is all but the first
491.Fa len0
492bytes.
493In case of failure, it returns
494.Dv NULL
495and attempts to
496restore the chain to its original state.
497.It Fn m_adj "struct mbuf *mp" "int req_len"
498Shaves off
499.Fa req_len
500bytes from head or tail of the (valid) data area.
501If
502.Fa req_len
503is greater than zero, front bytes are being shaved off, if it's smaller,
504from the back (and if it is zero, the mbuf will stay bearded).
505This function does not move data in any way, but is used to manipulate the
506data area pointer and data length variable of the mbuf in a non-clobbering
507way.
508.It Fn m_apply "struct mbuf *m" "int off" "int len" "int (*f)(void *, void *, unsigned int)" "void *arg"
509Apply function
510.Fa f
511to the data in an mbuf chain starting
512.Fa off
513bytes from the beginning, continuing for
514.Fa len
515bytes.
516Neither
517.Fa off
518nor
519.Fa len
520may be negative.
521.Fa arg
522will be supplied as first argument for
523.Fa f ,
524the second argument will be the pointer to the data buffer of a
525packet (starting after
526.Fa off
527bytes in the stream), and the third argument is the amount
528of data in bytes in this call.
529If
530.Fa f
531returns something not equal to zero
532.Fn m_apply
533will bail out, returning the return code of
534.Fa f .
535Upon successful completion it will return zero.
536.It Fn m_free "struct mbuf *m"
537Frees mbuf
538.Fa m .
539.It Fn m_freem "struct mbuf *m"
540Frees the mbuf chain beginning with
541.Fa m .
542This function contains the elementary sanity check for a
543.Dv NULL
544pointer.
545.It Fn mtod "struct mbuf *m" "datatype"
546Returns a pointer to the data contained in the specified mbuf
547.Fa m ,
548type-casted to the specified data type
549.Fa datatype .
550Implemented as a macro.
551.It Fn MGET "struct mbuf *m" "int how" "int type"
552Allocates mbuf
553.Fa m
554and initializes it to contain internal data.
555See
556.Fn m_get .
557Implemented as a macro.
558.It Fn MGETHDR "struct mbuf *m" "int how" "int type"
559Allocates mbuf
560.Fa m
561and initializes it to contain a packet header.
562See
563.Fn m_gethdr .
564Implemented as a macro.
565.It Fn MEXTMALLOC "struct mbuf *m" "int len" "int how"
566Allocates external storage of size
567.Fa len
568for mbuf
569.Fa m .
570The
571.Fa how
572parameter is a choice of
573.Dv M_WAIT / M_DONTWAIT
574from caller.
575The flag
576.Dv M_EXT
577is set upon success.
578Implemented as a macro.
579.It Fn MEXTADD "struct mbuf *m" "void *buf" "int size" "int type" "void (*free)(struct mbuf *, void *, size_t, void *)" "void *arg"
580Adds pre-allocated external storage
581.Fa buf
582to a normal mbuf
583.Fa m ;
584the parameters
585.Fa size ,
586.Fa type ,
587.Fa free
588and
589.Fa arg
590describe the external storage.
591.Fa size
592is the size of the storage,
593.Fa type
594describes its
595.Xr malloc 9
596type,
597.Fa free
598is a free routine (if not the usual one), and
599.Fa arg
600is a possible argument to the free routine.
601The flag
602.Dv M_EXT
603is set upon success.
604Implemented as a macro.
605If a free routine is specified, it will be called when the mbuf is freed.
606In the case of former, the first argument for a free routine is the mbuf
607.Fa m
608and the routine is expected to free it in addition to the external storage
609pointed by second argument.
610In the case of latter, the first argument for the routine is NULL.
611.It Fn MCLGET "struct mbuf *m" "int how"
612Allocates and adds an mbuf cluster to a normal mbuf
613.Fa m .
614The
615.Fa how
616parameter is a choice of
617.Dv M_WAIT / M_DONTWAIT
618from caller.
619The flag
620.Dv M_EXT
621is set upon success.
622Implemented as a macro.
623.It Fn M_COPY_PKTHDR "struct mbuf *to" "struct mbuf *from"
624Copies the mbuf pkthdr from mbuf
625.Fa from
626to mbuf
627.Fa to .
628.Fa from
629must have the type flag
630.Dv M_PKTHDR
631set, and
632.Fa to
633must be empty.
634Implemented as a macro.
635.It Fn M_MOVE_PKTHDR "struct mbuf *to" "struct mbuf *from"
636Moves the mbuf pkthdr from mbuf
637.Fa from
638to mbuf
639.Fa to .
640.Fa from
641must have the type flag
642.Dv M_PKTHDR
643set, and
644.Fa to
645must be empty.
646The flag
647.Dv M_PKTHDR
648in mbuf
649.Fa from
650will be cleared.
651.It Fn M_ALIGN "struct mbuf *m" "int len"
652Sets the data pointer of a newly allocated mbuf
653.Fa m
654to
655.Fa len
656bytes from the end of the mbuf data area, so that
657.Fa len
658bytes of data written to the mbuf
659.Fa m ,
660starting at the data pointer, will be aligned to the end of the data area.
661Implemented as a macro.
662.It Fn MH_ALIGN "struct mbuf *m" "int len"
663Sets the data pointer of a newly allocated packetheader mbuf
664.Fa m
665to
666.Fa len
667bytes from the end of the mbuf data area, so that
668.Fa len
669bytes of data written to the mbuf
670.Fa m ,
671starting at the data pointer, will be aligned to the end of the data area.
672Implemented as a macro.
673.It Fn M_LEADINGSPACE "struct mbuf *m"
674Returns the amount of space available before the current start of valid
675data in mbuf
676.Fa m .
677Returns 0 if the mbuf data part is shared across multiple mbufs
678.Pq i.e. not writable .
679Implemented as a macro.
680.It Fn M_TRAILINGSPACE "struct mbuf *m"
681Returns the amount of space available after the current end of valid
682data in mbuf
683.Fa m .
684Returns 0 if the mbuf data part is shared across multiple mbufs
685.Pq i.e. not writable .
686Implemented as a macro.
687.It Fn M_PREPEND "struct mbuf *m" "int plen" "int how"
688Prepends space of size
689.Fa plen
690to mbuf
691.Fa m .
692If a new mbuf must be allocated,
693.Fa how
694specifies whether to wait.
695If
696.Fa how
697is
698.Dv M_DONTWAIT
699and allocation fails, the original mbuf chain is freed and
700.Fa m
701is set to
702.Dv NULL .
703Implemented as a macro.
704.It Fn MCHTYPE "struct mbuf *m" "int type"
705Change mbuf
706.Fa m
707to new type
708.Fa type .
709Implemented as a macro.
710.It Fn MFREE "struct mbuf *m" "struct mbuf *n"
711Frees a single mbuf
712.Fa m
713and places the successor, if any, in mbuf
714.Fa n .
715Implemented as a macro.
716.El
717.Sh FILES
718The
719.Nm
720management functions are implemented within the file
721.Pa sys/kern/uipc_mbuf.c .
722Function prototypes, and the functions implemented as macros
723are located in
724.Pa sys/sys/mbuf.h .
725Both pathnames are relative to the root of the
726.Nx
727source tree,
728.Pa /usr/src .
729.Sh SEE ALSO
730.Pa /usr/share/doc/smm/18.net ,
731.Xr netstat 1 ,
732.Xr m_tag 9 ,
733.Xr malloc 9
734.Rs
735.%A Jun-ichiro Hagino
736.%T "Mbuf issues in 4.4BSD IPv6/IPsec support (experiences from KAME IPv6/IPsec implementation)"
737.%B "Proceedings of the freenix track: 2000 USENIX annual technical conference"
738.%D June 2000
739.Re
740.Sh AUTHORS
741.An -nosplit
742The original mbuf data structures were designed by Rob Gurwitz
743when he did the initial TCP/IP implementation at BBN.
744.Pp
745Further extensions and enhancements were made by Bill Joy, Sam Leffler,
746and Mike Karels at CSRG.
747.Pp
748Current implementation of external storage by
749.An Matt Thomas
750.Aq matt@3am-software.com
751and
752.An Jason R. Thorpe
753.Aq thorpej@NetBSD.org .
754