xref: /freebsd/sys/sys/mbuf.h (revision 3642298923e528d795e3a30ec165d2b469e28b40)
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  * $FreeBSD$
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  */
73 struct mb_args {
74 	int	flags;	/* Flags for mbuf being allocated */
75 	short	type;	/* Type of mbuf being allocated */
76 };
77 #endif /* _KERNEL */
78 
79 /*
80  * Header present at the beginning of every mbuf.
81  */
82 struct m_hdr {
83 	struct	mbuf *mh_next;		/* next buffer in chain */
84 	struct	mbuf *mh_nextpkt;	/* next chain in queue/record */
85 	caddr_t	mh_data;		/* location of data */
86 	int	mh_len;			/* amount of data in this mbuf */
87 	int	mh_flags;		/* flags; see below */
88 	short	mh_type;		/* type of data in this mbuf */
89 };
90 
91 /*
92  * Packet tag structure (see below for details).
93  */
94 struct m_tag {
95 	SLIST_ENTRY(m_tag)	m_tag_link;	/* List of packet tags */
96 	u_int16_t		m_tag_id;	/* Tag ID */
97 	u_int16_t		m_tag_len;	/* Length of data */
98 	u_int32_t		m_tag_cookie;	/* ABI/Module ID */
99 	void			(*m_tag_free)(struct m_tag *);
100 };
101 
102 /*
103  * Record/packet header in first mbuf of chain; valid only if M_PKTHDR is set.
104  */
105 struct pkthdr {
106 	struct	ifnet *rcvif;		/* rcv interface */
107 	int	len;			/* total packet length */
108 	/* variables for ip and tcp reassembly */
109 	void	*header;		/* pointer to packet header */
110 	/* variables for hardware checksum */
111 	int	csum_flags;		/* flags regarding checksum */
112 	int	csum_data;		/* data field used by csum routines */
113 	SLIST_HEAD(packet_tags, m_tag) tags; /* list of packet tags */
114 };
115 
116 /*
117  * Description of external storage mapped into mbuf; valid only if M_EXT is set.
118  */
119 struct m_ext {
120 	caddr_t	ext_buf;		/* start of buffer */
121 	void	(*ext_free)		/* free routine if not the usual */
122 		    (void *, void *);
123 	void	*ext_args;		/* optional argument pointer */
124 	u_int	ext_size;		/* size of buffer, for ext_free */
125 	volatile u_int *ref_cnt;	/* pointer to ref count info */
126 	int	ext_type;		/* type of external storage */
127 };
128 
129 /*
130  * The core of the mbuf object along with some shortcut defines for
131  * practical purposes.
132  */
133 struct mbuf {
134 	struct	m_hdr m_hdr;
135 	union {
136 		struct {
137 			struct	pkthdr MH_pkthdr;	/* M_PKTHDR set */
138 			union {
139 				struct	m_ext MH_ext;	/* M_EXT set */
140 				char	MH_databuf[MHLEN];
141 			} MH_dat;
142 		} MH;
143 		char	M_databuf[MLEN];		/* !M_PKTHDR, !M_EXT */
144 	} M_dat;
145 };
146 #define	m_next		m_hdr.mh_next
147 #define	m_len		m_hdr.mh_len
148 #define	m_data		m_hdr.mh_data
149 #define	m_type		m_hdr.mh_type
150 #define	m_flags		m_hdr.mh_flags
151 #define	m_nextpkt	m_hdr.mh_nextpkt
152 #define	m_act		m_nextpkt
153 #define	m_pkthdr	M_dat.MH.MH_pkthdr
154 #define	m_ext		M_dat.MH.MH_dat.MH_ext
155 #define	m_pktdat	M_dat.MH.MH_dat.MH_databuf
156 #define	m_dat		M_dat.M_databuf
157 
158 /*
159  * mbuf flags.
160  */
161 #define	M_EXT		0x0001	/* has associated external storage */
162 #define	M_PKTHDR	0x0002	/* start of record */
163 #define	M_EOR		0x0004	/* end of record */
164 #define	M_RDONLY	0x0008	/* associated data is marked read-only */
165 #define	M_PROTO1	0x0010	/* protocol-specific */
166 #define	M_PROTO2	0x0020	/* protocol-specific */
167 #define	M_PROTO3	0x0040	/* protocol-specific */
168 #define	M_PROTO4	0x0080	/* protocol-specific */
169 #define	M_PROTO5	0x0100	/* protocol-specific */
170 #define	M_SKIP_FIREWALL	0x4000	/* skip firewall processing */
171 #define	M_FREELIST	0x8000	/* mbuf is on the free list */
172 
173 /*
174  * mbuf pkthdr flags (also stored in m_flags).
175  */
176 #define	M_BCAST		0x0200	/* send/received as link-level broadcast */
177 #define	M_MCAST		0x0400	/* send/received as link-level multicast */
178 #define	M_FRAG		0x0800	/* packet is a fragment of a larger packet */
179 #define	M_FIRSTFRAG	0x1000	/* packet is first fragment */
180 #define	M_LASTFRAG	0x2000	/* packet is last fragment */
181 #define	M_VLANTAG	0x10000	/* packet has VLAN tag attached */
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 			    M_VLANTAG)
201 
202 /*
203  * Flags indicating hw checksum support and sw checksum requirements.
204  */
205 #define	CSUM_IP			0x0001		/* will csum IP */
206 #define	CSUM_TCP		0x0002		/* will csum TCP */
207 #define	CSUM_UDP		0x0004		/* will csum UDP */
208 #define	CSUM_IP_FRAGS		0x0008		/* will csum IP fragments */
209 #define	CSUM_FRAGMENT		0x0010		/* will do IP fragmentation */
210 
211 #define	CSUM_IP_CHECKED		0x0100		/* did csum IP */
212 #define	CSUM_IP_VALID		0x0200		/*   ... the csum is valid */
213 #define	CSUM_DATA_VALID		0x0400		/* csum_data field is valid */
214 #define	CSUM_PSEUDO_HDR		0x0800		/* csum_data has pseudo hdr */
215 
216 #define	CSUM_DELAY_DATA		(CSUM_TCP | CSUM_UDP)
217 #define	CSUM_DELAY_IP		(CSUM_IP)	/* XXX add ipv6 here too? */
218 
219 /*
220  * mbuf types.
221  */
222 #define	MT_NOTMBUF	0	/* USED INTERNALLY ONLY! Object is not mbuf */
223 #define	MT_DATA		1	/* dynamic (data) allocation */
224 #define	MT_HEADER	2	/* packet header */
225 #if 0
226 #define	MT_SOCKET	3	/* socket structure */
227 #define	MT_PCB		4	/* protocol control block */
228 #define	MT_RTABLE	5	/* routing tables */
229 #define	MT_HTABLE	6	/* IMP host tables */
230 #define	MT_ATABLE	7	/* address resolution tables */
231 #endif
232 #define	MT_SONAME	8	/* socket name */
233 #if 0
234 #define	MT_SOOPTS	10	/* socket options */
235 #endif
236 #define	MT_FTABLE	11	/* fragment reassembly header */
237 #if 0
238 #define	MT_RIGHTS	12	/* access rights */
239 #define	MT_IFADDR	13	/* interface address */
240 #endif
241 #define	MT_CONTROL	14	/* extra-data protocol message */
242 #define	MT_OOBDATA	15	/* expedited data  */
243 #define	MT_NTYPES	16	/* number of mbuf types for mbtypes[] */
244 
245 /*
246  * General mbuf allocator statistics structure.
247  */
248 struct mbstat {
249 	u_long	m_mbufs;	/* XXX */
250 	u_long	m_mclusts;	/* XXX */
251 
252 	u_long	m_drain;	/* times drained protocols for space */
253 	u_long	m_mcfail;	/* XXX: times m_copym failed */
254 	u_long	m_mpfail;	/* XXX: times m_pullup failed */
255 	u_long	m_msize;	/* length of an mbuf */
256 	u_long	m_mclbytes;	/* length of an mbuf cluster */
257 	u_long	m_minclsize;	/* min length of data to allocate a cluster */
258 	u_long	m_mlen;		/* length of data in an mbuf */
259 	u_long	m_mhlen;	/* length of data in a header mbuf */
260 
261 	/* Number of mbtypes (gives # elems in mbtypes[] array: */
262 	short	m_numtypes;
263 
264 	/* XXX: Sendfile stats should eventually move to their own struct */
265 	u_long	sf_iocnt;	/* times sendfile had to do disk I/O */
266 	u_long	sf_allocfail;	/* times sfbuf allocation failed */
267 	u_long	sf_allocwait;	/* times sfbuf allocation had to wait */
268 };
269 
270 /*
271  * Flags specifying how an allocation should be made.
272  *
273  * The flag to use is as follows:
274  * - M_DONTWAIT or M_NOWAIT from an interrupt handler to not block allocation.
275  * - M_WAIT or M_WAITOK or M_TRYWAIT from wherever it is safe to block.
276  *
277  * M_DONTWAIT/M_NOWAIT means that we will not block the thread explicitly
278  * and if we cannot allocate immediately we may return NULL,
279  * whereas M_WAIT/M_WAITOK/M_TRYWAIT means that if we cannot allocate
280  * resources we will block until they are available, and thus never
281  * return NULL.
282  *
283  * XXX Eventually just phase this out to use M_WAITOK/M_NOWAIT.
284  */
285 #define	MBTOM(how)	(how)
286 #define	M_DONTWAIT	M_NOWAIT
287 #define	M_TRYWAIT	M_WAITOK
288 #define	M_WAIT		M_WAITOK
289 
290 /*
291  * String names of mbuf-related UMA(9) and malloc(9) types.  Exposed to
292  * !_KERNEL so that monitoring tools can look up the zones with
293  * libmemstat(3).
294  */
295 #define	MBUF_MEM_NAME		"mbuf"
296 #define	MBUF_CLUSTER_MEM_NAME	"mbuf_cluster"
297 #define	MBUF_PACKET_MEM_NAME	"mbuf_packet"
298 #define	MBUF_TAG_MEM_NAME	"mbuf_tag"
299 
300 #ifdef _KERNEL
301 /*-
302  * mbuf external reference count management macros.
303  *
304  * MEXT_IS_REF(m): true if (m) is not the only mbuf referencing
305  *     the external buffer ext_buf.
306  *
307  * MEXT_REM_REF(m): remove reference to m_ext object.
308  *
309  * MEXT_ADD_REF(m): add reference to m_ext object already
310  *     referred to by (m).  XXX Note that it is VERY important that you
311  *     always set the second mbuf's m_ext.ref_cnt to point to the first
312  *     one's (i.e., n->m_ext.ref_cnt = m->m_ext.ref_cnt) AFTER you run
313  *     MEXT_ADD_REF(m).  This is because m might have a lazy initialized
314  *     ref_cnt (NULL) before this is run and it will only be looked up
315  *     from here.  We should make MEXT_ADD_REF() always take two mbufs
316  *     as arguments so that it can take care of this itself.
317  */
318 #define	MEXT_IS_REF(m)	(((m)->m_ext.ref_cnt != NULL)			\
319     && (*((m)->m_ext.ref_cnt) > 1))
320 
321 #define	MEXT_REM_REF(m) do {						\
322 	KASSERT((m)->m_ext.ref_cnt != NULL, ("m_ext refcnt lazy NULL")); \
323 	KASSERT(*((m)->m_ext.ref_cnt) > 0, ("m_ext refcnt < 0"));	\
324 	atomic_subtract_int((m)->m_ext.ref_cnt, 1);			\
325 } while(0)
326 
327 #define	MEXT_ADD_REF(m)	do {						\
328 	if ((m)->m_ext.ref_cnt == NULL) {				\
329 		KASSERT((m)->m_ext.ext_type == EXT_CLUSTER ||		\
330 		    (m)->m_ext.ext_type == EXT_PACKET,			\
331 		    ("Unexpected mbuf type has lazy refcnt"));		\
332 		(m)->m_ext.ref_cnt = (u_int *)uma_find_refcnt(		\
333 		    zone_clust, (m)->m_ext.ext_buf);			\
334 		*((m)->m_ext.ref_cnt) = 2;				\
335 	} else								\
336 		atomic_add_int((m)->m_ext.ref_cnt, 1);			\
337 } while (0)
338 
339 #ifdef WITNESS
340 #define MBUF_CHECKSLEEP(how) do {					\
341 	if (how == M_WAITOK)						\
342 		WITNESS_WARN(WARN_GIANTOK | WARN_SLEEPOK, NULL,		\
343 		    "Sleeping in \"%s\"", __func__);			\
344 } while(0)
345 #else
346 #define MBUF_CHECKSLEEP(how)
347 #endif
348 
349 /*
350  * Network buffer allocation API
351  *
352  * The rest of it is defined in kern/kern_mbuf.c
353  */
354 
355 extern uma_zone_t	zone_mbuf;
356 extern uma_zone_t	zone_clust;
357 extern uma_zone_t	zone_pack;
358 
359 static __inline struct mbuf	*m_get(int how, short type);
360 static __inline struct mbuf	*m_gethdr(int how, short type);
361 static __inline struct mbuf	*m_getcl(int how, short type, int flags);
362 static __inline struct mbuf	*m_getclr(int how, short type);	/* XXX */
363 static __inline struct mbuf	*m_free(struct mbuf *m);
364 static __inline void		 m_clget(struct mbuf *m, int how);
365 static __inline void		 m_chtype(struct mbuf *m, short new_type);
366 void				 mb_free_ext(struct mbuf *);
367 
368 static __inline
369 struct mbuf *
370 m_get(int how, short type)
371 {
372 	struct mb_args args;
373 
374 	args.flags = 0;
375 	args.type = type;
376 	return (uma_zalloc_arg(zone_mbuf, &args, how));
377 }
378 
379 /* XXX This should be depracated, very little use */
380 static __inline
381 struct mbuf *
382 m_getclr(int how, short type)
383 {
384 	struct mbuf *m;
385 	struct mb_args args;
386 
387 	args.flags = 0;
388 	args.type = type;
389 	m = uma_zalloc_arg(zone_mbuf, &args, how);
390 	if (m != NULL)
391 		bzero(m->m_data, MLEN);
392 	return m;
393 }
394 
395 static __inline
396 struct mbuf *
397 m_gethdr(int how, short type)
398 {
399 	struct mb_args args;
400 
401 	args.flags = M_PKTHDR;
402 	args.type = type;
403 	return (uma_zalloc_arg(zone_mbuf, &args, how));
404 }
405 
406 static __inline
407 struct mbuf *
408 m_getcl(int how, short type, int flags)
409 {
410 	struct mb_args args;
411 
412 	args.flags = flags;
413 	args.type = type;
414 	return (uma_zalloc_arg(zone_pack, &args, how));
415 }
416 
417 static __inline
418 struct mbuf *
419 m_free(struct mbuf *m)
420 {
421 	struct mbuf *n = m->m_next;
422 
423 #ifdef INVARIANTS
424 	m->m_flags |= M_FREELIST;
425 #endif
426 	if (m->m_flags & M_EXT)
427 		mb_free_ext(m);
428 	else
429 		uma_zfree(zone_mbuf, m);
430 	return n;
431 }
432 
433 static __inline
434 void
435 m_clget(struct mbuf *m, int how)
436 {
437 
438 	m->m_ext.ext_buf = NULL;
439 	uma_zalloc_arg(zone_clust, m, how);
440 }
441 
442 static __inline
443 void
444 m_chtype(struct mbuf *m, short new_type)
445 {
446 	m->m_type = new_type;
447 }
448 
449 /*
450  * mbuf, cluster, and external object allocation macros
451  * (for compatibility purposes).
452  */
453 /* NB: M_COPY_PKTHDR is deprecated.  Use M_MOVE_PKTHDR or m_dup_pktdr. */
454 #define	M_MOVE_PKTHDR(to, from)	m_move_pkthdr((to), (from))
455 #define	MGET(m, how, type)	((m) = m_get((how), (type)))
456 #define	MGETHDR(m, how, type)	((m) = m_gethdr((how), (type)))
457 #define	MCLGET(m, how)		m_clget((m), (how))
458 #define	MEXTADD(m, buf, size, free, args, flags, type) 			\
459     m_extadd((m), (caddr_t)(buf), (size), (free), (args), (flags), (type))
460 
461 /*
462  * Evaluate TRUE if it's safe to write to the mbuf m's data region (this
463  * can be both the local data payload, or an external buffer area,
464  * depending on whether M_EXT is set).
465  */
466 #define	M_WRITABLE(m)	(!((m)->m_flags & M_RDONLY) && (!((m)->m_flags  \
467 			    & M_EXT) || !MEXT_IS_REF(m)))
468 
469 /* Check if the supplied mbuf has a packet header, or else panic. */
470 #define	M_ASSERTPKTHDR(m)						\
471 	KASSERT(m != NULL && m->m_flags & M_PKTHDR,			\
472 	    ("%s: no mbuf packet header!", __func__))
473 
474 /* Ensure that the supplied mbuf is a valid, non-free mbuf. */
475 #define	M_ASSERTVALID(m)						\
476 	KASSERT((((struct mbuf *)m)->m_flags & M_FREELIST) == 0,	\
477 	    ("%s: attempted use of a free mbuf!", __func__))
478 
479 /*
480  * Set the m_data pointer of a newly-allocated mbuf (m_get/MGET) to place
481  * an object of the specified size at the end of the mbuf, longword aligned.
482  */
483 #define	M_ALIGN(m, len) do {						\
484 	(m)->m_data += (MLEN - (len)) & ~(sizeof(long) - 1);		\
485 } while (0)
486 
487 /*
488  * As above, for mbufs allocated with m_gethdr/MGETHDR
489  * or initialized by M_COPY_PKTHDR.
490  */
491 #define	MH_ALIGN(m, len) do {						\
492 	(m)->m_data += (MHLEN - (len)) & ~(sizeof(long) - 1);		\
493 } while (0)
494 
495 /*
496  * Compute the amount of space available
497  * before the current start of data in an mbuf.
498  *
499  * The M_WRITABLE() is a temporary, conservative safety measure: the burden
500  * of checking writability of the mbuf data area rests solely with the caller.
501  */
502 #define	M_LEADINGSPACE(m)						\
503 	((m)->m_flags & M_EXT ?						\
504 	    (M_WRITABLE(m) ? (m)->m_data - (m)->m_ext.ext_buf : 0):	\
505 	    (m)->m_flags & M_PKTHDR ? (m)->m_data - (m)->m_pktdat :	\
506 	    (m)->m_data - (m)->m_dat)
507 
508 /*
509  * Compute the amount of space available
510  * after the end of data in an mbuf.
511  *
512  * The M_WRITABLE() is a temporary, conservative safety measure: the burden
513  * of checking writability of the mbuf data area rests solely with the caller.
514  */
515 #define	M_TRAILINGSPACE(m)						\
516 	((m)->m_flags & M_EXT ?						\
517 	    (M_WRITABLE(m) ? (m)->m_ext.ext_buf + (m)->m_ext.ext_size	\
518 		- ((m)->m_data + (m)->m_len) : 0) :			\
519 	    &(m)->m_dat[MLEN] - ((m)->m_data + (m)->m_len))
520 
521 /*
522  * Arrange to prepend space of size plen to mbuf m.
523  * If a new mbuf must be allocated, how specifies whether to wait.
524  * If the allocation fails, the original mbuf chain is freed and m is
525  * set to NULL.
526  */
527 #define	M_PREPEND(m, plen, how) do {					\
528 	struct mbuf **_mmp = &(m);					\
529 	struct mbuf *_mm = *_mmp;					\
530 	int _mplen = (plen);						\
531 	int __mhow = (how);						\
532 									\
533 	MBUF_CHECKSLEEP(how);						\
534 	if (M_LEADINGSPACE(_mm) >= _mplen) {				\
535 		_mm->m_data -= _mplen;					\
536 		_mm->m_len += _mplen;					\
537 	} else								\
538 		_mm = m_prepend(_mm, _mplen, __mhow);			\
539 	if (_mm != NULL && _mm->m_flags & M_PKTHDR)			\
540 		_mm->m_pkthdr.len += _mplen;				\
541 	*_mmp = _mm;							\
542 } while (0)
543 
544 /*
545  * Change mbuf to new type.
546  * This is a relatively expensive operation and should be avoided.
547  */
548 #define	MCHTYPE(m, t)	m_chtype((m), (t))
549 
550 /* Length to m_copy to copy all. */
551 #define	M_COPYALL	1000000000
552 
553 /* Compatibility with 4.3. */
554 #define	m_copy(m, o, l)	m_copym((m), (o), (l), M_DONTWAIT)
555 
556 extern	int max_datalen;		/* MHLEN - max_hdr */
557 extern	int max_hdr;			/* Largest link + protocol header */
558 extern	int max_linkhdr;		/* Largest link-level header */
559 extern	int max_protohdr;		/* Largest protocol header */
560 extern	struct mbstat mbstat;		/* General mbuf stats/infos */
561 extern	int nmbclusters;		/* Maximum number of clusters */
562 
563 struct uio;
564 
565 void		 m_adj(struct mbuf *, int);
566 void		 m_align(struct mbuf *, int);
567 int		 m_apply(struct mbuf *, int, int,
568 		    int (*)(void *, void *, u_int), void *);
569 int		 m_append(struct mbuf *, int, c_caddr_t);
570 void		 m_cat(struct mbuf *, struct mbuf *);
571 void		 m_extadd(struct mbuf *, caddr_t, u_int,
572 		    void (*)(void *, void *), void *, int, int);
573 void		 m_copyback(struct mbuf *, int, int, c_caddr_t);
574 void		 m_copydata(const struct mbuf *, int, int, caddr_t);
575 struct	mbuf	*m_copym(struct mbuf *, int, int, int);
576 struct	mbuf	*m_copymdata(struct mbuf *, struct mbuf *,
577 		    int, int, int, int);
578 struct	mbuf	*m_copypacket(struct mbuf *, int);
579 void		 m_copy_pkthdr(struct mbuf *, struct mbuf *);
580 struct	mbuf	*m_copyup(struct mbuf *n, int len, int dstoff);
581 struct	mbuf	*m_defrag(struct mbuf *, int);
582 void		 m_demote(struct mbuf *, int);
583 struct	mbuf	*m_devget(char *, int, int, struct ifnet *,
584 		    void (*)(char *, caddr_t, u_int));
585 struct	mbuf	*m_dup(struct mbuf *, int);
586 int		 m_dup_pkthdr(struct mbuf *, struct mbuf *, int);
587 u_int		 m_fixhdr(struct mbuf *);
588 struct	mbuf	*m_fragment(struct mbuf *, int, int);
589 void		 m_freem(struct mbuf *);
590 struct	mbuf	*m_getm(struct mbuf *, int, int, short);
591 struct	mbuf	*m_getptr(struct mbuf *, int, int *);
592 u_int		 m_length(struct mbuf *, struct mbuf **);
593 void		 m_move_pkthdr(struct mbuf *, struct mbuf *);
594 struct	mbuf	*m_prepend(struct mbuf *, int, int);
595 void		 m_print(const struct mbuf *, int);
596 struct	mbuf	*m_pulldown(struct mbuf *, int, int, int *);
597 struct	mbuf	*m_pullup(struct mbuf *, int);
598 int		m_sanity(struct mbuf *, int);
599 struct	mbuf	*m_split(struct mbuf *, int, int);
600 struct	mbuf	*m_uiotombuf(struct uio *, int, int, int);
601 
602 /*-
603  * Network packets may have annotations attached by affixing a list
604  * of "packet tags" to the pkthdr structure.  Packet tags are
605  * dynamically allocated semi-opaque data structures that have
606  * a fixed header (struct m_tag) that specifies the size of the
607  * memory block and a <cookie,type> pair that identifies it.
608  * The cookie is a 32-bit unique unsigned value used to identify
609  * a module or ABI.  By convention this value is chose as the
610  * date+time that the module is created, expressed as the number of
611  * seconds since the epoch (e.g., using date -u +'%s').  The type value
612  * is an ABI/module-specific value that identifies a particular annotation
613  * and is private to the module.  For compatibility with systems
614  * like OpenBSD that define packet tags w/o an ABI/module cookie,
615  * the value PACKET_ABI_COMPAT is used to implement m_tag_get and
616  * m_tag_find compatibility shim functions and several tag types are
617  * defined below.  Users that do not require compatibility should use
618  * a private cookie value so that packet tag-related definitions
619  * can be maintained privately.
620  *
621  * Note that the packet tag returned by m_tag_alloc has the default
622  * memory alignment implemented by malloc.  To reference private data
623  * one can use a construct like:
624  *
625  *	struct m_tag *mtag = m_tag_alloc(...);
626  *	struct foo *p = (struct foo *)(mtag+1);
627  *
628  * if the alignment of struct m_tag is sufficient for referencing members
629  * of struct foo.  Otherwise it is necessary to embed struct m_tag within
630  * the private data structure to insure proper alignment; e.g.,
631  *
632  *	struct foo {
633  *		struct m_tag	tag;
634  *		...
635  *	};
636  *	struct foo *p = (struct foo *) m_tag_alloc(...);
637  *	struct m_tag *mtag = &p->tag;
638  */
639 
640 /*
641  * Persistent tags stay with an mbuf until the mbuf is reclaimed.
642  * Otherwise tags are expected to ``vanish'' when they pass through
643  * a network interface.  For most interfaces this happens normally
644  * as the tags are reclaimed when the mbuf is free'd.  However in
645  * some special cases reclaiming must be done manually.  An example
646  * is packets that pass through the loopback interface.  Also, one
647  * must be careful to do this when ``turning around'' packets (e.g.,
648  * icmp_reflect).
649  *
650  * To mark a tag persistent bit-or this flag in when defining the
651  * tag id.  The tag will then be treated as described above.
652  */
653 #define	MTAG_PERSISTENT				0x800
654 
655 #define	PACKET_TAG_NONE				0  /* Nadda */
656 
657 /* Packet tags for use with PACKET_ABI_COMPAT. */
658 #define	PACKET_TAG_IPSEC_IN_DONE		1  /* IPsec applied, in */
659 #define	PACKET_TAG_IPSEC_OUT_DONE		2  /* IPsec applied, out */
660 #define	PACKET_TAG_IPSEC_IN_CRYPTO_DONE		3  /* NIC IPsec crypto done */
661 #define	PACKET_TAG_IPSEC_OUT_CRYPTO_NEEDED	4  /* NIC IPsec crypto req'ed */
662 #define	PACKET_TAG_IPSEC_IN_COULD_DO_CRYPTO	5  /* NIC notifies IPsec */
663 #define	PACKET_TAG_IPSEC_PENDING_TDB		6  /* Reminder to do IPsec */
664 #define	PACKET_TAG_BRIDGE			7  /* Bridge processing done */
665 #define	PACKET_TAG_GIF				8  /* GIF processing done */
666 #define	PACKET_TAG_GRE				9  /* GRE processing done */
667 #define	PACKET_TAG_IN_PACKET_CHECKSUM		10 /* NIC checksumming done */
668 #define	PACKET_TAG_ENCAP			11 /* Encap.  processing */
669 #define	PACKET_TAG_IPSEC_SOCKET			12 /* IPSEC socket ref */
670 #define	PACKET_TAG_IPSEC_HISTORY		13 /* IPSEC history */
671 #define	PACKET_TAG_IPV6_INPUT			14 /* IPV6 input processing */
672 #define	PACKET_TAG_DUMMYNET			15 /* dummynet info */
673 #define	PACKET_TAG_DIVERT			17 /* divert info */
674 #define	PACKET_TAG_IPFORWARD			18 /* ipforward info */
675 #define	PACKET_TAG_MACLABEL	(19 | MTAG_PERSISTENT) /* MAC label */
676 #define	PACKET_TAG_PF_ROUTED			21 /* PF routed, avoid loops */
677 #define	PACKET_TAG_PF_FRAGCACHE			22 /* PF fragment cached */
678 #define	PACKET_TAG_PF_QID			23 /* PF ALTQ queue id */
679 #define	PACKET_TAG_PF_TAG			24 /* PF tagged */
680 #define	PACKET_TAG_RTSOCKFAM			25 /* rtsock sa family */
681 #define	PACKET_TAG_PF_TRANSLATE_LOCALHOST	26 /* PF translate localhost */
682 #define	PACKET_TAG_IPOPTIONS			27 /* Saved IP options */
683 #define	PACKET_TAG_CARP                         28 /* CARP info */
684 
685 /* Packet tag routines. */
686 struct	m_tag	*m_tag_alloc(u_int32_t, int, int, int);
687 void		 m_tag_delete(struct mbuf *, struct m_tag *);
688 void		 m_tag_delete_chain(struct mbuf *, struct m_tag *);
689 void		 m_tag_free_default(struct m_tag *);
690 struct	m_tag	*m_tag_locate(struct mbuf *, u_int32_t, int, struct m_tag *);
691 struct	m_tag	*m_tag_copy(struct m_tag *, int);
692 int		 m_tag_copy_chain(struct mbuf *, struct mbuf *, int);
693 void		 m_tag_delete_nonpersistent(struct mbuf *);
694 
695 /*
696  * Initialize the list of tags associated with an mbuf.
697  */
698 static __inline void
699 m_tag_init(struct mbuf *m)
700 {
701 	SLIST_INIT(&m->m_pkthdr.tags);
702 }
703 
704 /*
705  * Set up the contents of a tag.  Note that this does not
706  * fill in the free method; the caller is expected to do that.
707  *
708  * XXX probably should be called m_tag_init, but that was
709  * already taken.
710  */
711 static __inline void
712 m_tag_setup(struct m_tag *t, u_int32_t cookie, int type, int len)
713 {
714 	t->m_tag_id = type;
715 	t->m_tag_len = len;
716 	t->m_tag_cookie = cookie;
717 }
718 
719 /*
720  * Reclaim resources associated with a tag.
721  */
722 static __inline void
723 m_tag_free(struct m_tag *t)
724 {
725 	(*t->m_tag_free)(t);
726 }
727 
728 /*
729  * Return the first tag associated with an mbuf.
730  */
731 static __inline struct m_tag *
732 m_tag_first(struct mbuf *m)
733 {
734 	return (SLIST_FIRST(&m->m_pkthdr.tags));
735 }
736 
737 /*
738  * Return the next tag in the list of tags associated with an mbuf.
739  */
740 static __inline struct m_tag *
741 m_tag_next(struct mbuf *m, struct m_tag *t)
742 {
743 	return (SLIST_NEXT(t, m_tag_link));
744 }
745 
746 /*
747  * Prepend a tag to the list of tags associated with an mbuf.
748  */
749 static __inline void
750 m_tag_prepend(struct mbuf *m, struct m_tag *t)
751 {
752 	SLIST_INSERT_HEAD(&m->m_pkthdr.tags, t, m_tag_link);
753 }
754 
755 /*
756  * Unlink a tag from the list of tags associated with an mbuf.
757  */
758 static __inline void
759 m_tag_unlink(struct mbuf *m, struct m_tag *t)
760 {
761 	SLIST_REMOVE(&m->m_pkthdr.tags, t, m_tag, m_tag_link);
762 }
763 
764 /* These are for OpenBSD compatibility. */
765 #define	MTAG_ABI_COMPAT		0		/* compatibility ABI */
766 
767 static __inline struct m_tag *
768 m_tag_get(int type, int length, int wait)
769 {
770 	return (m_tag_alloc(MTAG_ABI_COMPAT, type, length, wait));
771 }
772 
773 static __inline struct m_tag *
774 m_tag_find(struct mbuf *m, int type, struct m_tag *start)
775 {
776 	return (SLIST_EMPTY(&m->m_pkthdr.tags) ?
777 	    NULL : m_tag_locate(m, MTAG_ABI_COMPAT, type, start));
778 }
779 
780 #endif /* _KERNEL */
781 
782 #endif /* !_SYS_MBUF_H_ */
783