mbuf.h revision 132488
1/*-
2 * Copyright (c) 1982, 1986, 1988, 1993
3 *	The Regents of the University of California.  All rights reserved.
4 *
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions
7 * are met:
8 * 1. Redistributions of source code must retain the above copyright
9 *    notice, this list of conditions and the following disclaimer.
10 * 2. Redistributions in binary form must reproduce the above copyright
11 *    notice, this list of conditions and the following disclaimer in the
12 *    documentation and/or other materials provided with the distribution.
13 * 3. Neither the name of the University nor the names of its contributors
14 *    may be used to endorse or promote products derived from this software
15 *    without specific prior written permission.
16 *
17 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
18 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
19 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
20 * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
21 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
22 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
23 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
24 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
25 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
26 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
27 * SUCH DAMAGE.
28 *
29 *	@(#)mbuf.h	8.5 (Berkeley) 2/19/95
30 * $FreeBSD: head/sys/sys/mbuf.h 132488 2004-07-21 07:12:24Z alfred $
31 */
32
33#ifndef _SYS_MBUF_H_
34#define	_SYS_MBUF_H_
35
36/* XXX: These includes suck. Sorry! */
37#include <sys/queue.h>
38#ifdef _KERNEL
39#include <sys/systm.h>
40#include <vm/uma.h>
41#ifdef WITNESS
42#include <sys/lock.h>
43#endif
44#endif
45
46/*
47 * Mbufs are of a single size, MSIZE (sys/param.h), which
48 * includes overhead.  An mbuf may add a single "mbuf cluster" of size
49 * MCLBYTES (also in sys/param.h), which has no additional overhead
50 * and is used instead of the internal data area; this is done when
51 * at least MINCLSIZE of data must be stored.  Additionally, it is possible
52 * to allocate a separate buffer externally and attach it to the mbuf in
53 * a way similar to that of mbuf clusters.
54 */
55#define	MLEN		(MSIZE - sizeof(struct m_hdr))	/* normal data len */
56#define	MHLEN		(MLEN - sizeof(struct pkthdr))	/* data len w/pkthdr */
57#define	MINCLSIZE	(MHLEN + 1)	/* smallest amount to put in cluster */
58#define	M_MAXCOMPRESS	(MHLEN / 2)	/* max amount to copy for compression */
59
60#ifdef _KERNEL
61/*-
62 * Macros for type conversion:
63 * mtod(m, t)	-- Convert mbuf pointer to data pointer of correct type.
64 * dtom(x)	-- Convert data pointer within mbuf to mbuf pointer (XXX).
65 */
66#define	mtod(m, t)	((t)((m)->m_data))
67#define	dtom(x)		((struct mbuf *)((intptr_t)(x) & ~(MSIZE-1)))
68
69/*
70 * Argument structure passed to UMA routines during mbuf and packet
71 * allocations.
72 */
73struct mb_args {
74	int	flags;	/* Flags for mbuf being allocated */
75	int	how;	/* How to allocate: M_WAITOK or M_DONTWAIT */
76	short	type;	/* Type of mbuf being allocated */
77};
78#endif /* _KERNEL */
79
80/*
81 * Header present at the beginning of every mbuf.
82 */
83struct m_hdr {
84	struct	mbuf *mh_next;		/* next buffer in chain */
85	struct	mbuf *mh_nextpkt;	/* next chain in queue/record */
86	caddr_t	mh_data;		/* location of data */
87	int	mh_len;			/* amount of data in this mbuf */
88	int	mh_flags;		/* flags; see below */
89	short	mh_type;		/* type of data in this mbuf */
90};
91
92/*
93 * Packet tag structure (see below for details).
94 */
95struct m_tag {
96	SLIST_ENTRY(m_tag)	m_tag_link;	/* List of packet tags */
97	u_int16_t		m_tag_id;	/* Tag ID */
98	u_int16_t		m_tag_len;	/* Length of data */
99	u_int32_t		m_tag_cookie;	/* ABI/Module ID */
100	void			(*m_tag_free)(struct m_tag *);
101};
102
103/*
104 * Record/packet header in first mbuf of chain; valid only if M_PKTHDR is set.
105 */
106struct pkthdr {
107	struct	ifnet *rcvif;		/* rcv interface */
108	int	len;			/* total packet length */
109	/* variables for ip and tcp reassembly */
110	void	*header;		/* pointer to packet header */
111	/* variables for hardware checksum */
112	int	csum_flags;		/* flags regarding checksum */
113	int	csum_data;		/* data field used by csum routines */
114	SLIST_HEAD(packet_tags, m_tag) tags; /* list of packet tags */
115};
116
117/*
118 * Description of external storage mapped into mbuf; valid only if M_EXT is set.
119 */
120struct m_ext {
121	caddr_t	ext_buf;		/* start of buffer */
122	void	(*ext_free)		/* free routine if not the usual */
123		    (void *, void *);
124	void	*ext_args;		/* optional argument pointer */
125	u_int	ext_size;		/* size of buffer, for ext_free */
126	u_int	*ref_cnt;		/* pointer to ref count info */
127	int	ext_type;		/* type of external storage */
128};
129
130/*
131 * The core of the mbuf object along with some shortcut defines for
132 * practical purposes.
133 */
134struct mbuf {
135	struct	m_hdr m_hdr;
136	union {
137		struct {
138			struct	pkthdr MH_pkthdr;	/* M_PKTHDR set */
139			union {
140				struct	m_ext MH_ext;	/* M_EXT set */
141				char	MH_databuf[MHLEN];
142			} MH_dat;
143		} MH;
144		char	M_databuf[MLEN];		/* !M_PKTHDR, !M_EXT */
145	} M_dat;
146};
147#define	m_next		m_hdr.mh_next
148#define	m_len		m_hdr.mh_len
149#define	m_data		m_hdr.mh_data
150#define	m_type		m_hdr.mh_type
151#define	m_flags		m_hdr.mh_flags
152#define	m_nextpkt	m_hdr.mh_nextpkt
153#define	m_act		m_nextpkt
154#define	m_pkthdr	M_dat.MH.MH_pkthdr
155#define	m_ext		M_dat.MH.MH_dat.MH_ext
156#define	m_pktdat	M_dat.MH.MH_dat.MH_databuf
157#define	m_dat		M_dat.M_databuf
158
159/*
160 * mbuf flags.
161 */
162#define	M_EXT		0x0001	/* has associated external storage */
163#define	M_PKTHDR	0x0002	/* start of record */
164#define	M_EOR		0x0004	/* end of record */
165#define	M_RDONLY	0x0008	/* associated data is marked read-only */
166#define	M_PROTO1	0x0010	/* protocol-specific */
167#define	M_PROTO2	0x0020	/* protocol-specific */
168#define	M_PROTO3	0x0040	/* protocol-specific */
169#define	M_PROTO4	0x0080	/* protocol-specific */
170#define	M_PROTO5	0x0100	/* protocol-specific */
171#define	M_SKIP_FIREWALL	0x4000	/* skip firewall processing */
172#define	M_FREELIST	0x8000	/* mbuf is on the free list */
173
174/*
175 * mbuf pkthdr flags (also stored in m_flags).
176 */
177#define	M_BCAST		0x0200	/* send/received as link-level broadcast */
178#define	M_MCAST		0x0400	/* send/received as link-level multicast */
179#define	M_FRAG		0x0800	/* packet is a fragment of a larger packet */
180#define	M_FIRSTFRAG	0x1000	/* packet is first fragment */
181#define	M_LASTFRAG	0x2000	/* packet is last fragment */
182
183/*
184 * External buffer types: identify ext_buf type.
185 */
186#define	EXT_CLUSTER	1	/* mbuf cluster */
187#define	EXT_SFBUF	2	/* sendfile(2)'s sf_bufs */
188#define	EXT_PACKET	3	/* came out of Packet zone */
189#define	EXT_NET_DRV	100	/* custom ext_buf provided by net driver(s) */
190#define	EXT_MOD_TYPE	200	/* custom module's ext_buf type */
191#define	EXT_DISPOSABLE	300	/* can throw this buffer away w/page flipping */
192#define	EXT_EXTREF	400	/* has externally maintained ref_cnt ptr */
193
194/*
195 * Flags copied when copying m_pkthdr.
196 */
197#define	M_COPYFLAGS	(M_PKTHDR|M_EOR|M_RDONLY|M_PROTO1|M_PROTO1|M_PROTO2|\
198			    M_PROTO3|M_PROTO4|M_PROTO5|M_SKIP_FIREWALL|\
199			    M_BCAST|M_MCAST|M_FRAG|M_FIRSTFRAG|M_LASTFRAG)
200
201/*
202 * Flags indicating hw checksum support and sw checksum requirements.
203 */
204#define	CSUM_IP			0x0001		/* will csum IP */
205#define	CSUM_TCP		0x0002		/* will csum TCP */
206#define	CSUM_UDP		0x0004		/* will csum UDP */
207#define	CSUM_IP_FRAGS		0x0008		/* will csum IP fragments */
208#define	CSUM_FRAGMENT		0x0010		/* will do IP fragmentation */
209
210#define	CSUM_IP_CHECKED		0x0100		/* did csum IP */
211#define	CSUM_IP_VALID		0x0200		/*   ... the csum is valid */
212#define	CSUM_DATA_VALID		0x0400		/* csum_data field is valid */
213#define	CSUM_PSEUDO_HDR		0x0800		/* csum_data has pseudo hdr */
214
215#define	CSUM_DELAY_DATA		(CSUM_TCP | CSUM_UDP)
216#define	CSUM_DELAY_IP		(CSUM_IP)	/* XXX add ipv6 here too? */
217
218/*
219 * mbuf types.
220 */
221#define	MT_NOTMBUF	0	/* USED INTERNALLY ONLY! Object is not mbuf */
222#define	MT_DATA		1	/* dynamic (data) allocation */
223#define	MT_HEADER	2	/* packet header */
224#if 0
225#define	MT_SOCKET	3	/* socket structure */
226#define	MT_PCB		4	/* protocol control block */
227#define	MT_RTABLE	5	/* routing tables */
228#define	MT_HTABLE	6	/* IMP host tables */
229#define	MT_ATABLE	7	/* address resolution tables */
230#endif
231#define	MT_SONAME	8	/* socket name */
232#if 0
233#define	MT_SOOPTS	10	/* socket options */
234#endif
235#define	MT_FTABLE	11	/* fragment reassembly header */
236#if 0
237#define	MT_RIGHTS	12	/* access rights */
238#define	MT_IFADDR	13	/* interface address */
239#endif
240#define	MT_CONTROL	14	/* extra-data protocol message */
241#define	MT_OOBDATA	15	/* expedited data  */
242#define	MT_NTYPES	16	/* number of mbuf types for mbtypes[] */
243
244/*
245 * General mbuf allocator statistics structure.
246 */
247struct mbstat {
248	u_long	m_mbufs;	/* XXX */
249	u_long	m_mclusts;	/* XXX */
250
251	u_long	m_drain;	/* times drained protocols for space */
252	u_long	m_mcfail;	/* XXX: times m_copym failed */
253	u_long	m_mpfail;	/* XXX: times m_pullup failed */
254	u_long	m_msize;	/* length of an mbuf */
255	u_long	m_mclbytes;	/* length of an mbuf cluster */
256	u_long	m_minclsize;	/* min length of data to allocate a cluster */
257	u_long	m_mlen;		/* length of data in an mbuf */
258	u_long	m_mhlen;	/* length of data in a header mbuf */
259
260	/* Number of mbtypes (gives # elems in mbtypes[] array: */
261	short	m_numtypes;
262
263	/* XXX: Sendfile stats should eventually move to their own struct */
264	u_long	sf_iocnt;	/* times sendfile had to do disk I/O */
265	u_long	sf_allocfail;	/* times sfbuf allocation failed */
266	u_long	sf_allocwait;	/* times sfbuf allocation had to wait */
267};
268
269/*
270 * Flags specifying how an allocation should be made.
271 *
272 * The flag to use is as follows:
273 * - M_DONTWAIT or M_NOWAIT from an interrupt handler to not block allocation.
274 * - M_WAIT or M_WAITOK or M_TRYWAIT from wherever it is safe to block.
275 *
276 * M_DONTWAIT/M_NOWAIT means that we will not block the thread explicitly
277 * and if we cannot allocate immediately we may return NULL,
278 * whereas M_WAIT/M_WAITOK/M_TRYWAIT means that if we cannot allocate
279 * resources we will block until they are available, and thus never
280 * return NULL.
281 *
282 * XXX Eventually just phase this out to use M_WAITOK/M_NOWAIT.
283 */
284#define	MBTOM(how)	(how)
285#define	M_DONTWAIT	M_NOWAIT
286#define	M_TRYWAIT	M_WAITOK
287#define	M_WAIT		M_WAITOK
288
289#ifdef _KERNEL
290/*-
291 * mbuf external reference count management macros.
292 *
293 * MEXT_IS_REF(m): true if (m) is not the only mbuf referencing
294 *     the external buffer ext_buf.
295 *
296 * MEXT_REM_REF(m): remove reference to m_ext object.
297 *
298 * MEXT_ADD_REF(m): add reference to m_ext object already
299 *     referred to by (m).
300 */
301#define	MEXT_IS_REF(m)	(*((m)->m_ext.ref_cnt) > 1)
302
303#define	MEXT_REM_REF(m) do {						\
304	KASSERT(*((m)->m_ext.ref_cnt) > 0, ("m_ext refcnt < 0"));	\
305	atomic_subtract_int((m)->m_ext.ref_cnt, 1);			\
306} while(0)
307
308#define	MEXT_ADD_REF(m)	atomic_add_int((m)->m_ext.ref_cnt, 1)
309
310#define MBUF_CHECKSLEEP(how) do {					\
311	if (how == M_WAITOK)						\
312		WITNESS_WARN(WARN_GIANTOK | WARN_SLEEPOK, NULL,		\
313		    "Sleeping in \"%s\"", __func__);			\
314} while(0)
315
316/*
317 * Network buffer allocation API
318 *
319 * The rest of it is defined in kern/subr_mbuf.c
320 */
321
322extern uma_zone_t	zone_mbuf;
323extern uma_zone_t	zone_clust;
324extern uma_zone_t	zone_pack;
325
326static __inline struct mbuf	*m_get(int how, short type);
327static __inline struct mbuf	*m_gethdr(int how, short type);
328static __inline struct mbuf	*m_getcl(int how, short type, int flags);
329static __inline struct mbuf	*m_getclr(int how, short type);	/* XXX */
330static __inline struct mbuf	*m_free(struct mbuf *m);
331static __inline void		 m_clget(struct mbuf *m, int how);
332static __inline void		 m_chtype(struct mbuf *m, short new_type);
333void				 mb_free_ext(struct mbuf *);
334
335static __inline
336struct mbuf *
337m_get(int how, short type)
338{
339	struct mb_args args;
340
341	MBUF_CHECKSLEEP(how);
342	args.flags = 0;
343	args.how = how;
344	args.type = type;
345	return (uma_zalloc_arg(zone_mbuf, &args, how));
346}
347
348/* XXX This should be depracated, very little use */
349static __inline
350struct mbuf *
351m_getclr(int how, short type)
352{
353	struct mbuf *m;
354	struct mb_args args;
355
356	MBUF_CHECKSLEEP(how);
357	args.flags = 0;
358	args.how = how;
359	args.type = type;
360	m = uma_zalloc_arg(zone_mbuf, &args, how);
361	if (m != NULL)
362		bzero(m->m_data, MLEN);
363	return m;
364}
365
366static __inline
367struct mbuf *
368m_gethdr(int how, short type)
369{
370	struct mb_args args;
371
372	MBUF_CHECKSLEEP(how);
373	args.flags = M_PKTHDR;
374	args.how = how;
375	args.type = type;
376	return (uma_zalloc_arg(zone_mbuf, &args, how));
377}
378
379static __inline
380struct mbuf *
381m_getcl(int how, short type, int flags)
382{
383	struct mb_args args;
384
385	MBUF_CHECKSLEEP(how);
386	args.flags = flags;
387	args.how = how;
388	args.type = type;
389	return (uma_zalloc_arg(zone_pack, &args, how));
390}
391
392static __inline
393struct mbuf *
394m_free(struct mbuf *m)
395{
396	struct mbuf *n = m->m_next;
397
398#ifdef INVARIANTS
399	m->m_flags |= M_FREELIST;
400#endif
401	if (m->m_flags & M_EXT)
402		mb_free_ext(m);
403	else
404		uma_zfree(zone_mbuf, m);
405	return n;
406}
407
408static __inline
409void
410m_clget(struct mbuf *m, int how)
411{
412
413	MBUF_CHECKSLEEP(how);
414	m->m_ext.ext_buf = NULL;
415	uma_zalloc_arg(zone_clust, m, how);
416}
417
418static __inline
419void
420m_chtype(struct mbuf *m, short new_type)
421{
422	m->m_type = new_type;
423}
424
425/*
426 * mbuf, cluster, and external object allocation macros
427 * (for compatibility purposes).
428 */
429/* NB: M_COPY_PKTHDR is deprecated.  Use M_MOVE_PKTHDR or m_dup_pktdr. */
430#define	M_MOVE_PKTHDR(to, from)	m_move_pkthdr((to), (from))
431#define	MGET(m, how, type)	((m) = m_get((how), (type)))
432#define	MGETHDR(m, how, type)	((m) = m_gethdr((how), (type)))
433#define	MCLGET(m, how)		m_clget((m), (how))
434#define	MEXTADD(m, buf, size, free, args, flags, type) 			\
435    m_extadd((m), (caddr_t)(buf), (size), (free), (args), (flags), (type))
436
437/*
438 * Evaluate TRUE if it's safe to write to the mbuf m's data region (this
439 * can be both the local data payload, or an external buffer area,
440 * depending on whether M_EXT is set).
441 */
442#define	M_WRITABLE(m)	(!((m)->m_flags & M_RDONLY) && (!((m)->m_flags  \
443			    & M_EXT) || !MEXT_IS_REF(m)))
444
445/* Check if the supplied mbuf has a packet header, or else panic. */
446#define	M_ASSERTPKTHDR(m)						\
447	KASSERT(m != NULL && m->m_flags & M_PKTHDR,			\
448	    ("%s: no mbuf packet header!", __func__))
449
450/* Ensure that the supplied mbuf is a valid, non-free mbuf. */
451#define	M_ASSERTVALID(m)						\
452	KASSERT((((struct mbuf *)m)->m_flags & M_FREELIST) == 0,	\
453	    ("%s: attempted use of a free mbuf!", __func__))
454
455/*
456 * Set the m_data pointer of a newly-allocated mbuf (m_get/MGET) to place
457 * an object of the specified size at the end of the mbuf, longword aligned.
458 */
459#define	M_ALIGN(m, len) do {						\
460	(m)->m_data += (MLEN - (len)) & ~(sizeof(long) - 1);		\
461} while (0)
462
463/*
464 * As above, for mbufs allocated with m_gethdr/MGETHDR
465 * or initialized by M_COPY_PKTHDR.
466 */
467#define	MH_ALIGN(m, len) do {						\
468	(m)->m_data += (MHLEN - (len)) & ~(sizeof(long) - 1);		\
469} while (0)
470
471/*
472 * Compute the amount of space available
473 * before the current start of data in an mbuf.
474 *
475 * The M_WRITABLE() is a temporary, conservative safety measure: the burden
476 * of checking writability of the mbuf data area rests solely with the caller.
477 */
478#define	M_LEADINGSPACE(m)						\
479	((m)->m_flags & M_EXT ?						\
480	    (M_WRITABLE(m) ? (m)->m_data - (m)->m_ext.ext_buf : 0):	\
481	    (m)->m_flags & M_PKTHDR ? (m)->m_data - (m)->m_pktdat :	\
482	    (m)->m_data - (m)->m_dat)
483
484/*
485 * Compute the amount of space available
486 * after the end of data in an mbuf.
487 *
488 * The M_WRITABLE() is a temporary, conservative safety measure: the burden
489 * of checking writability of the mbuf data area rests solely with the caller.
490 */
491#define	M_TRAILINGSPACE(m)						\
492	((m)->m_flags & M_EXT ?						\
493	    (M_WRITABLE(m) ? (m)->m_ext.ext_buf + (m)->m_ext.ext_size	\
494		- ((m)->m_data + (m)->m_len) : 0) :			\
495	    &(m)->m_dat[MLEN] - ((m)->m_data + (m)->m_len))
496
497/*
498 * Arrange to prepend space of size plen to mbuf m.
499 * If a new mbuf must be allocated, how specifies whether to wait.
500 * If the allocation fails, the original mbuf chain is freed and m is
501 * set to NULL.
502 */
503#define	M_PREPEND(m, plen, how) do {					\
504	struct mbuf **_mmp = &(m);					\
505	struct mbuf *_mm = *_mmp;					\
506	int _mplen = (plen);						\
507	int __mhow = (how);						\
508									\
509	MBUF_CHECKSLEEP(how);						\
510	if (M_LEADINGSPACE(_mm) >= _mplen) {				\
511		_mm->m_data -= _mplen;					\
512		_mm->m_len += _mplen;					\
513	} else								\
514		_mm = m_prepend(_mm, _mplen, __mhow);			\
515	if (_mm != NULL && _mm->m_flags & M_PKTHDR)			\
516		_mm->m_pkthdr.len += _mplen;				\
517	*_mmp = _mm;							\
518} while (0)
519
520/*
521 * Change mbuf to new type.
522 * This is a relatively expensive operation and should be avoided.
523 */
524#define	MCHTYPE(m, t)	m_chtype((m), (t))
525
526/* Length to m_copy to copy all. */
527#define	M_COPYALL	1000000000
528
529/* Compatibility with 4.3. */
530#define	m_copy(m, o, l)	m_copym((m), (o), (l), M_DONTWAIT)
531
532extern	int max_datalen;		/* MHLEN - max_hdr */
533extern	int max_hdr;			/* Largest link + protocol header */
534extern	int max_linkhdr;		/* Largest link-level header */
535extern	int max_protohdr;		/* Largest protocol header */
536extern	struct mbstat mbstat;		/* General mbuf stats/infos */
537extern	int nmbclusters;		/* Maximum number of clusters */
538
539struct uio;
540
541void		 m_adj(struct mbuf *, int);
542int		 m_apply(struct mbuf *, int, int,
543		    int (*)(void *, void *, u_int), void *);
544void		 m_cat(struct mbuf *, struct mbuf *);
545void		 m_extadd(struct mbuf *, caddr_t, u_int,
546		    void (*)(void *, void *), void *, int, int);
547void		 m_copyback(struct mbuf *, int, int, c_caddr_t);
548void		 m_copydata(const struct mbuf *, int, int, caddr_t);
549struct	mbuf	*m_copym(struct mbuf *, int, int, int);
550struct	mbuf	*m_copypacket(struct mbuf *, int);
551void		 m_copy_pkthdr(struct mbuf *, struct mbuf *);
552struct	mbuf	*m_defrag(struct mbuf *, int);
553struct	mbuf	*m_devget(char *, int, int, struct ifnet *,
554		    void (*)(char *, caddr_t, u_int));
555struct	mbuf	*m_dup(struct mbuf *, int);
556int		 m_dup_pkthdr(struct mbuf *, struct mbuf *, int);
557u_int		 m_fixhdr(struct mbuf *);
558struct	mbuf	*m_fragment(struct mbuf *, int, int);
559void		 m_freem(struct mbuf *);
560struct	mbuf	*m_getm(struct mbuf *, int, int, short);
561struct	mbuf	*m_getptr(struct mbuf *, int, int *);
562u_int		 m_length(struct mbuf *, struct mbuf **);
563void		 m_move_pkthdr(struct mbuf *, struct mbuf *);
564struct	mbuf	*m_prepend(struct mbuf *, int, int);
565void		 m_print(const struct mbuf *);
566struct	mbuf	*m_pulldown(struct mbuf *, int, int, int *);
567struct	mbuf	*m_pullup(struct mbuf *, int);
568struct	mbuf	*m_split(struct mbuf *, int, int);
569struct	mbuf	*m_uiotombuf(struct uio *, int, int);
570
571/*-
572 * Network packets may have annotations attached by affixing a list
573 * of "packet tags" to the pkthdr structure.  Packet tags are
574 * dynamically allocated semi-opaque data structures that have
575 * a fixed header (struct m_tag) that specifies the size of the
576 * memory block and a <cookie,type> pair that identifies it.
577 * The cookie is a 32-bit unique unsigned value used to identify
578 * a module or ABI.  By convention this value is chose as the
579 * date+time that the module is created, expressed as the number of
580 * seconds since the epoch (e.g., using date -u +'%s').  The type value
581 * is an ABI/module-specific value that identifies a particular annotation
582 * and is private to the module.  For compatibility with systems
583 * like OpenBSD that define packet tags w/o an ABI/module cookie,
584 * the value PACKET_ABI_COMPAT is used to implement m_tag_get and
585 * m_tag_find compatibility shim functions and several tag types are
586 * defined below.  Users that do not require compatibility should use
587 * a private cookie value so that packet tag-related definitions
588 * can be maintained privately.
589 *
590 * Note that the packet tag returned by m_tag_alloc has the default
591 * memory alignment implemented by malloc.  To reference private data
592 * one can use a construct like:
593 *
594 *	struct m_tag *mtag = m_tag_alloc(...);
595 *	struct foo *p = (struct foo *)(mtag+1);
596 *
597 * if the alignment of struct m_tag is sufficient for referencing members
598 * of struct foo.  Otherwise it is necessary to embed struct m_tag within
599 * the private data structure to insure proper alignment; e.g.,
600 *
601 *	struct foo {
602 *		struct m_tag	tag;
603 *		...
604 *	};
605 *	struct foo *p = (struct foo *) m_tag_alloc(...);
606 *	struct m_tag *mtag = &p->tag;
607 */
608
609/*
610 * Persistent tags stay with an mbuf until the mbuf is reclaimed.
611 * Otherwise tags are expected to ``vanish'' when they pass through
612 * a network interface.  For most interfaces this happens normally
613 * as the tags are reclaimed when the mbuf is free'd.  However in
614 * some special cases reclaiming must be done manually.  An example
615 * is packets that pass through the loopback interface.  Also, one
616 * must be careful to do this when ``turning around'' packets (e.g.,
617 * icmp_reflect).
618 *
619 * To mark a tag persistent bit-or this flag in when defining the
620 * tag id.  The tag will then be treated as described above.
621 */
622#define	MTAG_PERSISTENT				0x800
623
624#define	PACKET_TAG_NONE				0  /* Nadda */
625
626/* Packet tags for use with PACKET_ABI_COMPAT. */
627#define	PACKET_TAG_IPSEC_IN_DONE		1  /* IPsec applied, in */
628#define	PACKET_TAG_IPSEC_OUT_DONE		2  /* IPsec applied, out */
629#define	PACKET_TAG_IPSEC_IN_CRYPTO_DONE		3  /* NIC IPsec crypto done */
630#define	PACKET_TAG_IPSEC_OUT_CRYPTO_NEEDED	4  /* NIC IPsec crypto req'ed */
631#define	PACKET_TAG_IPSEC_IN_COULD_DO_CRYPTO	5  /* NIC notifies IPsec */
632#define	PACKET_TAG_IPSEC_PENDING_TDB		6  /* Reminder to do IPsec */
633#define	PACKET_TAG_BRIDGE			7  /* Bridge processing done */
634#define	PACKET_TAG_GIF				8  /* GIF processing done */
635#define	PACKET_TAG_GRE				9  /* GRE processing done */
636#define	PACKET_TAG_IN_PACKET_CHECKSUM		10 /* NIC checksumming done */
637#define	PACKET_TAG_ENCAP			11 /* Encap.  processing */
638#define	PACKET_TAG_IPSEC_SOCKET			12 /* IPSEC socket ref */
639#define	PACKET_TAG_IPSEC_HISTORY		13 /* IPSEC history */
640#define	PACKET_TAG_IPV6_INPUT			14 /* IPV6 input processing */
641#define	PACKET_TAG_DUMMYNET			15 /* dummynet info */
642#define	PACKET_TAG_DIVERT			17 /* divert info */
643#define	PACKET_TAG_IPFORWARD			18 /* ipforward info */
644#define	PACKET_TAG_MACLABEL	(19 | MTAG_PERSISTENT) /* MAC label */
645#define	PACKET_TAG_PF_ROUTED			21 /* PF routed, avoid loops */
646#define	PACKET_TAG_PF_FRAGCACHE			22 /* PF fragment cached */
647#define	PACKET_TAG_PF_QID			23 /* PF ALTQ queue id */
648#define	PACKET_TAG_PF_TAG			24 /* PF tagged */
649#define	PACKET_TAG_RTSOCKFAM			25 /* rtsock sa family */
650#define	PACKET_TAG_PF_TRANSLATE_LOCALHOST	26 /* PF translate localhost */
651
652/* Packet tag routines. */
653struct	m_tag	*m_tag_alloc(u_int32_t, int, int, int);
654void		 m_tag_delete(struct mbuf *, struct m_tag *);
655void		 m_tag_delete_chain(struct mbuf *, struct m_tag *);
656struct	m_tag	*m_tag_locate(struct mbuf *, u_int32_t, int, struct m_tag *);
657struct	m_tag	*m_tag_copy(struct m_tag *, int);
658int		 m_tag_copy_chain(struct mbuf *, struct mbuf *, int);
659void		 m_tag_delete_nonpersistent(struct mbuf *);
660
661/*
662 * Initialize the list of tags associated with an mbuf.
663 */
664static __inline void
665m_tag_init(struct mbuf *m)
666{
667	SLIST_INIT(&m->m_pkthdr.tags);
668}
669
670/*
671 * Set up the contents of a tag.  Note that this does not
672 * fill in the free method; the caller is expected to do that.
673 *
674 * XXX probably should be called m_tag_init, but that was
675 * already taken.
676 */
677static __inline void
678m_tag_setup(struct m_tag *t, u_int32_t cookie, int type, int len)
679{
680	t->m_tag_id = type;
681	t->m_tag_len = len;
682	t->m_tag_cookie = cookie;
683}
684
685/*
686 * Reclaim resources associated with a tag.
687 */
688static __inline void
689m_tag_free(struct m_tag *t)
690{
691	(*t->m_tag_free)(t);
692}
693
694/*
695 * Return the first tag associated with an mbuf.
696 */
697static __inline struct m_tag *
698m_tag_first(struct mbuf *m)
699{
700	return (SLIST_FIRST(&m->m_pkthdr.tags));
701}
702
703/*
704 * Return the next tag in the list of tags associated with an mbuf.
705 */
706static __inline struct m_tag *
707m_tag_next(struct mbuf *m, struct m_tag *t)
708{
709	return (SLIST_NEXT(t, m_tag_link));
710}
711
712/*
713 * Prepend a tag to the list of tags associated with an mbuf.
714 */
715static __inline void
716m_tag_prepend(struct mbuf *m, struct m_tag *t)
717{
718	SLIST_INSERT_HEAD(&m->m_pkthdr.tags, t, m_tag_link);
719}
720
721/*
722 * Unlink a tag from the list of tags associated with an mbuf.
723 */
724static __inline void
725m_tag_unlink(struct mbuf *m, struct m_tag *t)
726{
727	SLIST_REMOVE(&m->m_pkthdr.tags, t, m_tag, m_tag_link);
728}
729
730/* These are for OpenBSD compatibility. */
731#define	MTAG_ABI_COMPAT		0		/* compatibility ABI */
732
733static __inline struct m_tag *
734m_tag_get(int type, int length, int wait)
735{
736	return (m_tag_alloc(MTAG_ABI_COMPAT, type, length, wait));
737}
738
739static __inline struct m_tag *
740m_tag_find(struct mbuf *m, int type, struct m_tag *start)
741{
742	return (SLIST_EMPTY(&m->m_pkthdr.tags) ?
743	    NULL : m_tag_locate(m, MTAG_ABI_COMPAT, type, start));
744}
745
746/*
747 * Obtain next_hop information associated with the mbuf, if any.
748 * If a tag is present devalidate it also.
749 */
750static __inline struct sockaddr_in *
751m_claim_next(struct mbuf *m, int type)
752{
753	struct m_tag *mtag = m_tag_find(m, type, NULL);
754	if (mtag) {
755		struct sockaddr_in *sin = *(struct sockaddr_in **)(mtag + 1);
756		mtag->m_tag_id = PACKET_TAG_NONE;
757		return (sin);
758	}
759	return (NULL);
760}
761#endif /* _KERNEL */
762
763#endif /* !_SYS_MBUF_H_ */
764