xref: /freebsd/sys/sys/mbuf.h (revision d056fa046c6a91b90cd98165face0e42a33a5173)
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_JUMBOP	3	/* jumbo cluster 4096 bytes */
189 #define	EXT_JUMBO9	4	/* jumbo cluster 9216 bytes */
190 #define	EXT_JUMBO16	5	/* jumbo cluster 16184 bytes */
191 #define	EXT_PACKET	6	/* mbuf+cluster from packet zone */
192 #define	EXT_NET_DRV	100	/* custom ext_buf provided by net driver(s) */
193 #define	EXT_MOD_TYPE	200	/* custom module's ext_buf type */
194 #define	EXT_DISPOSABLE	300	/* can throw this buffer away w/page flipping */
195 #define	EXT_EXTREF	400	/* has externally maintained ref_cnt ptr */
196 
197 /*
198  * Flags copied when copying m_pkthdr.
199  */
200 #define	M_COPYFLAGS	(M_PKTHDR|M_EOR|M_RDONLY|M_PROTO1|M_PROTO1|M_PROTO2|\
201 			    M_PROTO3|M_PROTO4|M_PROTO5|M_SKIP_FIREWALL|\
202 			    M_BCAST|M_MCAST|M_FRAG|M_FIRSTFRAG|M_LASTFRAG|\
203 			    M_VLANTAG)
204 
205 /*
206  * Flags to purge when crossing layers.
207  */
208 #define	M_PROTOFLAGS	(M_PROTO1|M_PROTO2|M_PROTO3|M_PROTO4|M_PROTO5)
209 
210 /*
211  * Flags indicating hw checksum support and sw checksum requirements.
212  */
213 #define	CSUM_IP			0x0001		/* will csum IP */
214 #define	CSUM_TCP		0x0002		/* will csum TCP */
215 #define	CSUM_UDP		0x0004		/* will csum UDP */
216 #define	CSUM_IP_FRAGS		0x0008		/* will csum IP fragments */
217 #define	CSUM_FRAGMENT		0x0010		/* will do IP fragmentation */
218 
219 #define	CSUM_IP_CHECKED		0x0100		/* did csum IP */
220 #define	CSUM_IP_VALID		0x0200		/*   ... the csum is valid */
221 #define	CSUM_DATA_VALID		0x0400		/* csum_data field is valid */
222 #define	CSUM_PSEUDO_HDR		0x0800		/* csum_data has pseudo hdr */
223 
224 #define	CSUM_DELAY_DATA		(CSUM_TCP | CSUM_UDP)
225 #define	CSUM_DELAY_IP		(CSUM_IP)	/* XXX add ipv6 here too? */
226 
227 /*
228  * mbuf types.
229  */
230 #define	MT_NOTMBUF	0	/* USED INTERNALLY ONLY! Object is not mbuf */
231 #define	MT_DATA		1	/* dynamic (data) allocation */
232 #define	MT_HEADER	MT_DATA	/* packet header, use M_PKTHDR instead */
233 #if 0
234 #define	MT_SOCKET	3	/* socket structure */
235 #define	MT_PCB		4	/* protocol control block */
236 #define	MT_RTABLE	5	/* routing tables */
237 #define	MT_HTABLE	6	/* IMP host tables */
238 #define	MT_ATABLE	7	/* address resolution tables */
239 #endif
240 #define	MT_SONAME	8	/* socket name */
241 #if 0
242 #define	MT_SOOPTS	10	/* socket options */
243 #endif
244 #define	MT_FTABLE	11	/* fragment reassembly header */
245 #if 0
246 #define	MT_RIGHTS	12	/* access rights */
247 #define	MT_IFADDR	13	/* interface address */
248 #endif
249 #define	MT_CONTROL	14	/* extra-data protocol message */
250 #define	MT_OOBDATA	15	/* expedited data  */
251 #define	MT_NTYPES	16	/* number of mbuf types for mbtypes[] */
252 
253 #define	MT_NOINIT	255	/* Not a type but a flag to allocate
254 				   a non-initialized mbuf */
255 
256 /*
257  * General mbuf allocator statistics structure.
258  */
259 struct mbstat {
260 	u_long	m_mbufs;	/* XXX */
261 	u_long	m_mclusts;	/* XXX */
262 
263 	u_long	m_drain;	/* times drained protocols for space */
264 	u_long	m_mcfail;	/* XXX: times m_copym failed */
265 	u_long	m_mpfail;	/* XXX: times m_pullup failed */
266 	u_long	m_msize;	/* length of an mbuf */
267 	u_long	m_mclbytes;	/* length of an mbuf cluster */
268 	u_long	m_minclsize;	/* min length of data to allocate a cluster */
269 	u_long	m_mlen;		/* length of data in an mbuf */
270 	u_long	m_mhlen;	/* length of data in a header mbuf */
271 
272 	/* Number of mbtypes (gives # elems in mbtypes[] array: */
273 	short	m_numtypes;
274 
275 	/* XXX: Sendfile stats should eventually move to their own struct */
276 	u_long	sf_iocnt;	/* times sendfile had to do disk I/O */
277 	u_long	sf_allocfail;	/* times sfbuf allocation failed */
278 	u_long	sf_allocwait;	/* times sfbuf allocation had to wait */
279 };
280 
281 /*
282  * Flags specifying how an allocation should be made.
283  *
284  * The flag to use is as follows:
285  * - M_DONTWAIT or M_NOWAIT from an interrupt handler to not block allocation.
286  * - M_WAIT or M_WAITOK or M_TRYWAIT from wherever it is safe to block.
287  *
288  * M_DONTWAIT/M_NOWAIT means that we will not block the thread explicitly
289  * and if we cannot allocate immediately we may return NULL,
290  * whereas M_WAIT/M_WAITOK/M_TRYWAIT means that if we cannot allocate
291  * resources we will block until they are available, and thus never
292  * return NULL.
293  *
294  * XXX Eventually just phase this out to use M_WAITOK/M_NOWAIT.
295  */
296 #define	MBTOM(how)	(how)
297 #define	M_DONTWAIT	M_NOWAIT
298 #define	M_TRYWAIT	M_WAITOK
299 #define	M_WAIT		M_WAITOK
300 
301 /*
302  * String names of mbuf-related UMA(9) and malloc(9) types.  Exposed to
303  * !_KERNEL so that monitoring tools can look up the zones with
304  * libmemstat(3).
305  */
306 #define	MBUF_MEM_NAME		"mbuf"
307 #define	MBUF_CLUSTER_MEM_NAME	"mbuf_cluster"
308 #define	MBUF_PACKET_MEM_NAME	"mbuf_packet"
309 #define	MBUF_JUMBOP_MEM_NAME	"mbuf_jumbo_pagesize"
310 #define	MBUF_JUMBO9_MEM_NAME	"mbuf_jumbo_9k"
311 #define	MBUF_JUMBO16_MEM_NAME	"mbuf_jumbo_16k"
312 #define	MBUF_TAG_MEM_NAME	"mbuf_tag"
313 #define	MBUF_EXTREFCNT_MEM_NAME	"mbuf_ext_refcnt"
314 
315 #ifdef _KERNEL
316 
317 #ifdef WITNESS
318 #define MBUF_CHECKSLEEP(how) do {					\
319 	if (how == M_WAITOK)						\
320 		WITNESS_WARN(WARN_GIANTOK | WARN_SLEEPOK, NULL,		\
321 		    "Sleeping in \"%s\"", __func__);			\
322 } while (0)
323 #else
324 #define MBUF_CHECKSLEEP(how)
325 #endif
326 
327 /*
328  * Network buffer allocation API
329  *
330  * The rest of it is defined in kern/kern_mbuf.c
331  */
332 
333 extern uma_zone_t	zone_mbuf;
334 extern uma_zone_t	zone_clust;
335 extern uma_zone_t	zone_pack;
336 extern uma_zone_t	zone_jumbop;
337 extern uma_zone_t	zone_jumbo9;
338 extern uma_zone_t	zone_jumbo16;
339 extern uma_zone_t	zone_ext_refcnt;
340 extern uma_zone_t	zone_mtag_vlan;
341 
342 static __inline struct mbuf	*m_get(int how, short type);
343 static __inline struct mbuf	*m_gethdr(int how, short type);
344 static __inline struct mbuf	*m_getcl(int how, short type, int flags);
345 static __inline struct mbuf	*m_getjcl(int how, short type, int flags, int size);
346 static __inline struct mbuf	*m_getclr(int how, short type);	/* XXX */
347 static __inline struct mbuf	*m_free(struct mbuf *m);
348 static __inline void		 m_clget(struct mbuf *m, int how);
349 static __inline void		*m_cljget(struct mbuf *m, int how, int size);
350 static __inline void		 m_chtype(struct mbuf *m, short new_type);
351 void				 mb_free_ext(struct mbuf *);
352 
353 static __inline
354 struct mbuf *
355 m_get(int how, short type)
356 {
357 	struct mb_args args;
358 
359 	args.flags = 0;
360 	args.type = type;
361 	return (uma_zalloc_arg(zone_mbuf, &args, how));
362 }
363 
364 /* XXX This should be depracated, very little use */
365 static __inline
366 struct mbuf *
367 m_getclr(int how, short type)
368 {
369 	struct mbuf *m;
370 	struct mb_args args;
371 
372 	args.flags = 0;
373 	args.type = type;
374 	m = uma_zalloc_arg(zone_mbuf, &args, how);
375 	if (m != NULL)
376 		bzero(m->m_data, MLEN);
377 	return m;
378 }
379 
380 static __inline
381 struct mbuf *
382 m_gethdr(int how, short type)
383 {
384 	struct mb_args args;
385 
386 	args.flags = M_PKTHDR;
387 	args.type = type;
388 	return (uma_zalloc_arg(zone_mbuf, &args, how));
389 }
390 
391 static __inline
392 struct mbuf *
393 m_getcl(int how, short type, int flags)
394 {
395 	struct mb_args args;
396 
397 	args.flags = flags;
398 	args.type = type;
399 	return (uma_zalloc_arg(zone_pack, &args, how));
400 }
401 
402 /*
403  * m_getjcl() returns an mbuf with a cluster of the specified size attached.
404  * For size it takes MCLBYTES, MJUMPAGESIZE, MJUM9BYTES, MJUM16BYTES.
405  */
406 static __inline	/* XXX: This is rather large, should be real function maybe. */
407 struct mbuf *
408 m_getjcl(int how, short type, int flags, int size)
409 {
410 	struct mb_args args;
411 	struct mbuf *m, *n;
412 	uma_zone_t zone;
413 
414 	args.flags = flags;
415 	args.type = type;
416 
417 	m = uma_zalloc_arg(zone_mbuf, &args, how);
418 	if (m == NULL)
419 		return NULL;
420 
421 	switch (size) {
422 	case MCLBYTES:
423 		zone = zone_clust;
424 		break;
425 #if MJUMPAGESIZE != MCLBYTES
426 	case MJUMPAGESIZE:
427 		zone = zone_jumbop;
428 		break;
429 #endif
430 	case MJUM9BYTES:
431 		zone = zone_jumbo9;
432 		break;
433 	case MJUM16BYTES:
434 		zone = zone_jumbo16;
435 		break;
436 	default:
437 		panic("%s: m_getjcl: invalid cluster type", __func__);
438 	}
439 	n = uma_zalloc_arg(zone, m, how);
440 	if (n == NULL) {
441 		uma_zfree(zone_mbuf, m);
442 		return NULL;
443 	}
444 	return m;
445 }
446 
447 static __inline
448 struct mbuf *
449 m_free(struct mbuf *m)
450 {
451 	struct mbuf *n = m->m_next;
452 
453 	if (m->m_flags & M_EXT)
454 		mb_free_ext(m);
455 	else
456 		uma_zfree(zone_mbuf, m);
457 	return n;
458 }
459 
460 static __inline
461 void
462 m_clget(struct mbuf *m, int how)
463 {
464 	if (m->m_flags & M_EXT)
465 		printf("%s: %p mbuf already has cluster\n", __func__, m);
466 	m->m_ext.ext_buf = NULL;
467 	uma_zalloc_arg(zone_clust, m, how);
468 }
469 
470 /*
471  * m_cljget() is different from m_clget() as it can allocate clusters
472  * without attaching them to an mbuf.  In that case the return value
473  * is the pointer to the cluster of the requested size.  If an mbuf was
474  * specified, it gets the cluster attached to it and the return value
475  * can be safely ignored.
476  * For size it takes MCLBYTES, MJUMPAGESIZE, MJUM9BYTES, MJUM16BYTES.
477  */
478 static __inline
479 void *
480 m_cljget(struct mbuf *m, int how, int size)
481 {
482 	uma_zone_t zone;
483 
484 	if (m && m->m_flags & M_EXT)
485 		printf("%s: %p mbuf already has cluster\n", __func__, m);
486 	if (m != NULL)
487 		m->m_ext.ext_buf = NULL;
488 
489 	switch (size) {
490 	case MCLBYTES:
491 		zone = zone_clust;
492 		break;
493 #if MJUMPAGESIZE != MCLBYTES
494 	case MJUMPAGESIZE:
495 		zone = zone_jumbop;
496 		break;
497 #endif
498 	case MJUM9BYTES:
499 		zone = zone_jumbo9;
500 		break;
501 	case MJUM16BYTES:
502 		zone = zone_jumbo16;
503 		break;
504 	default:
505 		panic("%s: m_getjcl: invalid cluster type", __func__);
506 	}
507 
508 	return (uma_zalloc_arg(zone, m, how));
509 }
510 
511 static __inline
512 void
513 m_chtype(struct mbuf *m, short new_type)
514 {
515 	m->m_type = new_type;
516 }
517 
518 /*
519  * mbuf, cluster, and external object allocation macros
520  * (for compatibility purposes).
521  */
522 /* NB: M_COPY_PKTHDR is deprecated.  Use M_MOVE_PKTHDR or m_dup_pktdr. */
523 #define	M_MOVE_PKTHDR(to, from)	m_move_pkthdr((to), (from))
524 #define	MGET(m, how, type)	((m) = m_get((how), (type)))
525 #define	MGETHDR(m, how, type)	((m) = m_gethdr((how), (type)))
526 #define	MCLGET(m, how)		m_clget((m), (how))
527 #define	MEXTADD(m, buf, size, free, args, flags, type) 			\
528     m_extadd((m), (caddr_t)(buf), (size), (free), (args), (flags), (type))
529 
530 /*
531  * Evaluate TRUE if it's safe to write to the mbuf m's data region (this
532  * can be both the local data payload, or an external buffer area,
533  * depending on whether M_EXT is set).
534  */
535 #define	M_WRITABLE(m)	(!((m)->m_flags & M_RDONLY) &&			\
536 			 (!(((m)->m_flags & M_EXT)) ||			\
537 			 (*((m)->m_ext.ref_cnt) == 1)) )		\
538 
539 /* Check if the supplied mbuf has a packet header, or else panic. */
540 #define	M_ASSERTPKTHDR(m)						\
541 	KASSERT(m != NULL && m->m_flags & M_PKTHDR,			\
542 	    ("%s: no mbuf packet header!", __func__))
543 
544 /* Ensure that the supplied mbuf is a valid, non-free mbuf. */
545 /* XXX: Broken at the moment.  Need some UMA magic to make it work again. */
546 #define	M_ASSERTVALID(m)						\
547 	KASSERT((((struct mbuf *)m)->m_flags & 0) == 0,			\
548 	    ("%s: attempted use of a free mbuf!", __func__))
549 
550 /*
551  * Set the m_data pointer of a newly-allocated mbuf (m_get/MGET) to place
552  * an object of the specified size at the end of the mbuf, longword aligned.
553  */
554 #define	M_ALIGN(m, len) do {						\
555 	KASSERT(!((m)->m_flags & (M_PKTHDR|M_EXT)),			\
556 		("%s: M_ALIGN not normal mbuf", __func__));		\
557 	KASSERT((m)->m_data == (m)->m_dat,				\
558 		("%s: M_ALIGN not a virgin mbuf", __func__));		\
559 	(m)->m_data += (MLEN - (len)) & ~(sizeof(long) - 1);		\
560 } while (0)
561 
562 /*
563  * As above, for mbufs allocated with m_gethdr/MGETHDR
564  * or initialized by M_COPY_PKTHDR.
565  */
566 #define	MH_ALIGN(m, len) do {						\
567 	KASSERT((m)->m_flags & M_PKTHDR && !((m)->m_flags & M_EXT),	\
568 		("%s: MH_ALIGN not PKTHDR mbuf", __func__));		\
569 	KASSERT((m)->m_data == (m)->m_pktdat,				\
570 		("%s: MH_ALIGN not a virgin mbuf", __func__));		\
571 	(m)->m_data += (MHLEN - (len)) & ~(sizeof(long) - 1);		\
572 } while (0)
573 
574 /*
575  * Compute the amount of space available
576  * before the current start of data in an mbuf.
577  *
578  * The M_WRITABLE() is a temporary, conservative safety measure: the burden
579  * of checking writability of the mbuf data area rests solely with the caller.
580  */
581 #define	M_LEADINGSPACE(m)						\
582 	((m)->m_flags & M_EXT ?						\
583 	    (M_WRITABLE(m) ? (m)->m_data - (m)->m_ext.ext_buf : 0):	\
584 	    (m)->m_flags & M_PKTHDR ? (m)->m_data - (m)->m_pktdat :	\
585 	    (m)->m_data - (m)->m_dat)
586 
587 /*
588  * Compute the amount of space available
589  * after the end of data in an mbuf.
590  *
591  * The M_WRITABLE() is a temporary, conservative safety measure: the burden
592  * of checking writability of the mbuf data area rests solely with the caller.
593  */
594 #define	M_TRAILINGSPACE(m)						\
595 	((m)->m_flags & M_EXT ?						\
596 	    (M_WRITABLE(m) ? (m)->m_ext.ext_buf + (m)->m_ext.ext_size	\
597 		- ((m)->m_data + (m)->m_len) : 0) :			\
598 	    &(m)->m_dat[MLEN] - ((m)->m_data + (m)->m_len))
599 
600 /*
601  * Arrange to prepend space of size plen to mbuf m.
602  * If a new mbuf must be allocated, how specifies whether to wait.
603  * If the allocation fails, the original mbuf chain is freed and m is
604  * set to NULL.
605  */
606 #define	M_PREPEND(m, plen, how) do {					\
607 	struct mbuf **_mmp = &(m);					\
608 	struct mbuf *_mm = *_mmp;					\
609 	int _mplen = (plen);						\
610 	int __mhow = (how);						\
611 									\
612 	MBUF_CHECKSLEEP(how);						\
613 	if (M_LEADINGSPACE(_mm) >= _mplen) {				\
614 		_mm->m_data -= _mplen;					\
615 		_mm->m_len += _mplen;					\
616 	} else								\
617 		_mm = m_prepend(_mm, _mplen, __mhow);			\
618 	if (_mm != NULL && _mm->m_flags & M_PKTHDR)			\
619 		_mm->m_pkthdr.len += _mplen;				\
620 	*_mmp = _mm;							\
621 } while (0)
622 
623 /*
624  * Change mbuf to new type.
625  * This is a relatively expensive operation and should be avoided.
626  */
627 #define	MCHTYPE(m, t)	m_chtype((m), (t))
628 
629 /* Length to m_copy to copy all. */
630 #define	M_COPYALL	1000000000
631 
632 /* Compatibility with 4.3. */
633 #define	m_copy(m, o, l)	m_copym((m), (o), (l), M_DONTWAIT)
634 
635 extern	int max_datalen;		/* MHLEN - max_hdr */
636 extern	int max_hdr;			/* Largest link + protocol header */
637 extern	int max_linkhdr;		/* Largest link-level header */
638 extern	int max_protohdr;		/* Largest protocol header */
639 extern	struct mbstat mbstat;		/* General mbuf stats/infos */
640 extern	int nmbclusters;		/* Maximum number of clusters */
641 
642 struct uio;
643 
644 void		 m_adj(struct mbuf *, int);
645 void		 m_align(struct mbuf *, int);
646 int		 m_apply(struct mbuf *, int, int,
647 		    int (*)(void *, void *, u_int), void *);
648 int		 m_append(struct mbuf *, int, c_caddr_t);
649 void		 m_cat(struct mbuf *, struct mbuf *);
650 void		 m_extadd(struct mbuf *, caddr_t, u_int,
651 		    void (*)(void *, void *), void *, int, int);
652 void		 m_copyback(struct mbuf *, int, int, c_caddr_t);
653 void		 m_copydata(const struct mbuf *, int, int, caddr_t);
654 struct	mbuf	*m_copym(struct mbuf *, int, int, int);
655 struct	mbuf	*m_copymdata(struct mbuf *, struct mbuf *,
656 		    int, int, int, int);
657 struct	mbuf	*m_copypacket(struct mbuf *, int);
658 void		 m_copy_pkthdr(struct mbuf *, struct mbuf *);
659 struct	mbuf	*m_copyup(struct mbuf *n, int len, int dstoff);
660 struct	mbuf	*m_defrag(struct mbuf *, int);
661 void		 m_demote(struct mbuf *, int);
662 struct	mbuf	*m_devget(char *, int, int, struct ifnet *,
663 		    void (*)(char *, caddr_t, u_int));
664 struct	mbuf	*m_dup(struct mbuf *, int);
665 int		 m_dup_pkthdr(struct mbuf *, struct mbuf *, int);
666 u_int		 m_fixhdr(struct mbuf *);
667 struct	mbuf	*m_fragment(struct mbuf *, int, int);
668 void		 m_freem(struct mbuf *);
669 struct	mbuf	*m_getm(struct mbuf *, int, int, short);
670 struct	mbuf	*m_getptr(struct mbuf *, int, int *);
671 u_int		 m_length(struct mbuf *, struct mbuf **);
672 void		 m_move_pkthdr(struct mbuf *, struct mbuf *);
673 struct	mbuf	*m_prepend(struct mbuf *, int, int);
674 void		 m_print(const struct mbuf *, int);
675 struct	mbuf	*m_pulldown(struct mbuf *, int, int, int *);
676 struct	mbuf	*m_pullup(struct mbuf *, int);
677 int		m_sanity(struct mbuf *, int);
678 struct	mbuf	*m_split(struct mbuf *, int, int);
679 struct	mbuf	*m_uiotombuf(struct uio *, int, int, int);
680 struct	mbuf	*m_unshare(struct mbuf *, int how);
681 
682 /*-
683  * Network packets may have annotations attached by affixing a list
684  * of "packet tags" to the pkthdr structure.  Packet tags are
685  * dynamically allocated semi-opaque data structures that have
686  * a fixed header (struct m_tag) that specifies the size of the
687  * memory block and a <cookie,type> pair that identifies it.
688  * The cookie is a 32-bit unique unsigned value used to identify
689  * a module or ABI.  By convention this value is chose as the
690  * date+time that the module is created, expressed as the number of
691  * seconds since the epoch (e.g., using date -u +'%s').  The type value
692  * is an ABI/module-specific value that identifies a particular annotation
693  * and is private to the module.  For compatibility with systems
694  * like OpenBSD that define packet tags w/o an ABI/module cookie,
695  * the value PACKET_ABI_COMPAT is used to implement m_tag_get and
696  * m_tag_find compatibility shim functions and several tag types are
697  * defined below.  Users that do not require compatibility should use
698  * a private cookie value so that packet tag-related definitions
699  * can be maintained privately.
700  *
701  * Note that the packet tag returned by m_tag_alloc has the default
702  * memory alignment implemented by malloc.  To reference private data
703  * one can use a construct like:
704  *
705  *	struct m_tag *mtag = m_tag_alloc(...);
706  *	struct foo *p = (struct foo *)(mtag+1);
707  *
708  * if the alignment of struct m_tag is sufficient for referencing members
709  * of struct foo.  Otherwise it is necessary to embed struct m_tag within
710  * the private data structure to insure proper alignment; e.g.,
711  *
712  *	struct foo {
713  *		struct m_tag	tag;
714  *		...
715  *	};
716  *	struct foo *p = (struct foo *) m_tag_alloc(...);
717  *	struct m_tag *mtag = &p->tag;
718  */
719 
720 /*
721  * Persistent tags stay with an mbuf until the mbuf is reclaimed.
722  * Otherwise tags are expected to ``vanish'' when they pass through
723  * a network interface.  For most interfaces this happens normally
724  * as the tags are reclaimed when the mbuf is free'd.  However in
725  * some special cases reclaiming must be done manually.  An example
726  * is packets that pass through the loopback interface.  Also, one
727  * must be careful to do this when ``turning around'' packets (e.g.,
728  * icmp_reflect).
729  *
730  * To mark a tag persistent bit-or this flag in when defining the
731  * tag id.  The tag will then be treated as described above.
732  */
733 #define	MTAG_PERSISTENT				0x800
734 
735 #define	PACKET_TAG_NONE				0  /* Nadda */
736 
737 /* Packet tags for use with PACKET_ABI_COMPAT. */
738 #define	PACKET_TAG_IPSEC_IN_DONE		1  /* IPsec applied, in */
739 #define	PACKET_TAG_IPSEC_OUT_DONE		2  /* IPsec applied, out */
740 #define	PACKET_TAG_IPSEC_IN_CRYPTO_DONE		3  /* NIC IPsec crypto done */
741 #define	PACKET_TAG_IPSEC_OUT_CRYPTO_NEEDED	4  /* NIC IPsec crypto req'ed */
742 #define	PACKET_TAG_IPSEC_IN_COULD_DO_CRYPTO	5  /* NIC notifies IPsec */
743 #define	PACKET_TAG_IPSEC_PENDING_TDB		6  /* Reminder to do IPsec */
744 #define	PACKET_TAG_BRIDGE			7  /* Bridge processing done */
745 #define	PACKET_TAG_GIF				8  /* GIF processing done */
746 #define	PACKET_TAG_GRE				9  /* GRE processing done */
747 #define	PACKET_TAG_IN_PACKET_CHECKSUM		10 /* NIC checksumming done */
748 #define	PACKET_TAG_ENCAP			11 /* Encap.  processing */
749 #define	PACKET_TAG_IPSEC_SOCKET			12 /* IPSEC socket ref */
750 #define	PACKET_TAG_IPSEC_HISTORY		13 /* IPSEC history */
751 #define	PACKET_TAG_IPV6_INPUT			14 /* IPV6 input processing */
752 #define	PACKET_TAG_DUMMYNET			15 /* dummynet info */
753 #define	PACKET_TAG_DIVERT			17 /* divert info */
754 #define	PACKET_TAG_IPFORWARD			18 /* ipforward info */
755 #define	PACKET_TAG_MACLABEL	(19 | MTAG_PERSISTENT) /* MAC label */
756 #define	PACKET_TAG_PF_ROUTED			21 /* PF routed, avoid loops */
757 #define	PACKET_TAG_PF_FRAGCACHE			22 /* PF fragment cached */
758 #define	PACKET_TAG_PF_QID			23 /* PF ALTQ queue id */
759 #define	PACKET_TAG_PF_TAG			24 /* PF tagged */
760 #define	PACKET_TAG_RTSOCKFAM			25 /* rtsock sa family */
761 #define	PACKET_TAG_PF_TRANSLATE_LOCALHOST	26 /* PF translate localhost */
762 #define	PACKET_TAG_IPOPTIONS			27 /* Saved IP options */
763 #define	PACKET_TAG_CARP                         28 /* CARP info */
764 
765 /* Specific cookies and tags. */
766 #define	MTAG_VLAN				1035328035
767 #define	MTAG_VLAN_TAG				0 /* tag of VLAN interface */
768 
769 /* Packet tag routines. */
770 struct	m_tag	*m_tag_alloc(u_int32_t, int, int, int);
771 void		 m_tag_delete(struct mbuf *, struct m_tag *);
772 void		 m_tag_delete_chain(struct mbuf *, struct m_tag *);
773 void		 m_tag_free_default(struct m_tag *);
774 struct	m_tag	*m_tag_locate(struct mbuf *, u_int32_t, int, struct m_tag *);
775 struct	m_tag	*m_tag_copy(struct m_tag *, int);
776 int		 m_tag_copy_chain(struct mbuf *, struct mbuf *, int);
777 void		 m_tag_delete_nonpersistent(struct mbuf *);
778 
779 /*
780  * Initialize the list of tags associated with an mbuf.
781  */
782 static __inline void
783 m_tag_init(struct mbuf *m)
784 {
785 	SLIST_INIT(&m->m_pkthdr.tags);
786 }
787 
788 /*
789  * Set up the contents of a tag.  Note that this does not
790  * fill in the free method; the caller is expected to do that.
791  *
792  * XXX probably should be called m_tag_init, but that was
793  * already taken.
794  */
795 static __inline void
796 m_tag_setup(struct m_tag *t, u_int32_t cookie, int type, int len)
797 {
798 	t->m_tag_id = type;
799 	t->m_tag_len = len;
800 	t->m_tag_cookie = cookie;
801 }
802 
803 /*
804  * Reclaim resources associated with a tag.
805  */
806 static __inline void
807 m_tag_free(struct m_tag *t)
808 {
809 	(*t->m_tag_free)(t);
810 }
811 
812 /*
813  * Return the first tag associated with an mbuf.
814  */
815 static __inline struct m_tag *
816 m_tag_first(struct mbuf *m)
817 {
818 	return (SLIST_FIRST(&m->m_pkthdr.tags));
819 }
820 
821 /*
822  * Return the next tag in the list of tags associated with an mbuf.
823  */
824 static __inline struct m_tag *
825 m_tag_next(struct mbuf *m, struct m_tag *t)
826 {
827 	return (SLIST_NEXT(t, m_tag_link));
828 }
829 
830 /*
831  * Prepend a tag to the list of tags associated with an mbuf.
832  */
833 static __inline void
834 m_tag_prepend(struct mbuf *m, struct m_tag *t)
835 {
836 	SLIST_INSERT_HEAD(&m->m_pkthdr.tags, t, m_tag_link);
837 }
838 
839 /*
840  * Unlink a tag from the list of tags associated with an mbuf.
841  */
842 static __inline void
843 m_tag_unlink(struct mbuf *m, struct m_tag *t)
844 {
845 	SLIST_REMOVE(&m->m_pkthdr.tags, t, m_tag, m_tag_link);
846 }
847 
848 /* These are for OpenBSD compatibility. */
849 #define	MTAG_ABI_COMPAT		0		/* compatibility ABI */
850 
851 static __inline struct m_tag *
852 m_tag_get(int type, int length, int wait)
853 {
854 	return (m_tag_alloc(MTAG_ABI_COMPAT, type, length, wait));
855 }
856 
857 static __inline struct m_tag *
858 m_tag_find(struct mbuf *m, int type, struct m_tag *start)
859 {
860 	return (SLIST_EMPTY(&m->m_pkthdr.tags) ?
861 	    NULL : m_tag_locate(m, MTAG_ABI_COMPAT, type, start));
862 }
863 
864 #endif /* _KERNEL */
865 
866 #endif /* !_SYS_MBUF_H_ */
867