xref: /freebsd/sys/sys/mbuf.h (revision a4eb85b6acb49cb60c72c2cab0d0d3f00eaa6d46)
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 #define	MT_SONAME	8	/* socket name */
234 #define	MT_CONTROL	14	/* extra-data protocol message */
235 #define	MT_OOBDATA	15	/* expedited data  */
236 #define	MT_NTYPES	16	/* number of mbuf types for mbtypes[] */
237 
238 #define	MT_NOINIT	255	/* Not a type but a flag to allocate
239 				   a non-initialized mbuf */
240 
241 /*
242  * General mbuf allocator statistics structure.
243  *
244  * Many of these statistics are no longer used; we instead track many
245  * allocator statistics through UMA's built in statistics mechanism.
246  */
247 struct 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 /*
290  * String names of mbuf-related UMA(9) and malloc(9) types.  Exposed to
291  * !_KERNEL so that monitoring tools can look up the zones with
292  * libmemstat(3).
293  */
294 #define	MBUF_MEM_NAME		"mbuf"
295 #define	MBUF_CLUSTER_MEM_NAME	"mbuf_cluster"
296 #define	MBUF_PACKET_MEM_NAME	"mbuf_packet"
297 #define	MBUF_JUMBOP_MEM_NAME	"mbuf_jumbo_pagesize"
298 #define	MBUF_JUMBO9_MEM_NAME	"mbuf_jumbo_9k"
299 #define	MBUF_JUMBO16_MEM_NAME	"mbuf_jumbo_16k"
300 #define	MBUF_TAG_MEM_NAME	"mbuf_tag"
301 #define	MBUF_EXTREFCNT_MEM_NAME	"mbuf_ext_refcnt"
302 
303 #ifdef _KERNEL
304 
305 #ifdef WITNESS
306 #define MBUF_CHECKSLEEP(how) do {					\
307 	if (how == M_WAITOK)						\
308 		WITNESS_WARN(WARN_GIANTOK | WARN_SLEEPOK, NULL,		\
309 		    "Sleeping in \"%s\"", __func__);			\
310 } while (0)
311 #else
312 #define MBUF_CHECKSLEEP(how)
313 #endif
314 
315 /*
316  * Network buffer allocation API
317  *
318  * The rest of it is defined in kern/kern_mbuf.c
319  */
320 
321 extern uma_zone_t	zone_mbuf;
322 extern uma_zone_t	zone_clust;
323 extern uma_zone_t	zone_pack;
324 extern uma_zone_t	zone_jumbop;
325 extern uma_zone_t	zone_jumbo9;
326 extern uma_zone_t	zone_jumbo16;
327 extern uma_zone_t	zone_ext_refcnt;
328 extern uma_zone_t	zone_mtag_vlan;
329 
330 static __inline struct mbuf	*m_get(int how, short type);
331 static __inline struct mbuf	*m_gethdr(int how, short type);
332 static __inline struct mbuf	*m_getcl(int how, short type, int flags);
333 static __inline struct mbuf	*m_getjcl(int how, short type, int flags, int size);
334 static __inline struct mbuf	*m_getclr(int how, short type);	/* XXX */
335 static __inline struct mbuf	*m_free(struct mbuf *m);
336 static __inline void		 m_clget(struct mbuf *m, int how);
337 static __inline void		*m_cljget(struct mbuf *m, int how, int size);
338 static __inline void		 m_chtype(struct mbuf *m, short new_type);
339 void				 mb_free_ext(struct mbuf *);
340 
341 static __inline
342 struct mbuf *
343 m_get(int how, short type)
344 {
345 	struct mb_args args;
346 
347 	args.flags = 0;
348 	args.type = type;
349 	return (struct mbuf *)(uma_zalloc_arg(zone_mbuf, &args, how));
350 }
351 
352 /* XXX This should be depracated, very little use */
353 static __inline
354 struct mbuf *
355 m_getclr(int how, short type)
356 {
357 	struct mbuf *m;
358 	struct mb_args args;
359 
360 	args.flags = 0;
361 	args.type = type;
362 	m = uma_zalloc_arg(zone_mbuf, &args, how);
363 	if (m != NULL)
364 		bzero(m->m_data, MLEN);
365 	return m;
366 }
367 
368 static __inline
369 struct mbuf *
370 m_gethdr(int how, short type)
371 {
372 	struct mb_args args;
373 
374 	args.flags = M_PKTHDR;
375 	args.type = type;
376 	return (struct mbuf *)(uma_zalloc_arg(zone_mbuf, &args, how));
377 }
378 
379 static __inline
380 struct mbuf *
381 m_getcl(int how, short type, int flags)
382 {
383 	struct mb_args args;
384 
385 	args.flags = flags;
386 	args.type = type;
387 	return (struct mbuf *)(uma_zalloc_arg(zone_pack, &args, how));
388 }
389 
390 /*
391  * m_getjcl() returns an mbuf with a cluster of the specified size attached.
392  * For size it takes MCLBYTES, MJUMPAGESIZE, MJUM9BYTES, MJUM16BYTES.
393  */
394 static __inline	/* XXX: This is rather large, should be real function maybe. */
395 struct mbuf *
396 m_getjcl(int how, short type, int flags, int size)
397 {
398 	struct mb_args args;
399 	struct mbuf *m, *n;
400 	uma_zone_t zone;
401 
402 	args.flags = flags;
403 	args.type = type;
404 
405 	m = uma_zalloc_arg(zone_mbuf, &args, how);
406 	if (m == NULL)
407 		return NULL;
408 
409 	switch (size) {
410 	case MCLBYTES:
411 		zone = zone_clust;
412 		break;
413 #if MJUMPAGESIZE != MCLBYTES
414 	case MJUMPAGESIZE:
415 		zone = zone_jumbop;
416 		break;
417 #endif
418 	case MJUM9BYTES:
419 		zone = zone_jumbo9;
420 		break;
421 	case MJUM16BYTES:
422 		zone = zone_jumbo16;
423 		break;
424 	default:
425 		panic("%s: m_getjcl: invalid cluster type", __func__);
426 	}
427 	n = uma_zalloc_arg(zone, m, how);
428 	if (n == NULL) {
429 		uma_zfree(zone_mbuf, m);
430 		return NULL;
431 	}
432 	return m;
433 }
434 
435 static __inline
436 struct mbuf *
437 m_free(struct mbuf *m)
438 {
439 	struct mbuf *n = m->m_next;
440 
441 	if (m->m_flags & M_EXT)
442 		mb_free_ext(m);
443 	else
444 		uma_zfree(zone_mbuf, m);
445 	return n;
446 }
447 
448 static __inline
449 void
450 m_clget(struct mbuf *m, int how)
451 {
452 	if (m->m_flags & M_EXT)
453 		printf("%s: %p mbuf already has cluster\n", __func__, m);
454 	m->m_ext.ext_buf = (char *)NULL;
455 	uma_zalloc_arg(zone_clust, m, how);
456 }
457 
458 /*
459  * m_cljget() is different from m_clget() as it can allocate clusters
460  * without attaching them to an mbuf.  In that case the return value
461  * is the pointer to the cluster of the requested size.  If an mbuf was
462  * specified, it gets the cluster attached to it and the return value
463  * can be safely ignored.
464  * For size it takes MCLBYTES, MJUMPAGESIZE, MJUM9BYTES, MJUM16BYTES.
465  */
466 static __inline
467 void *
468 m_cljget(struct mbuf *m, int how, int size)
469 {
470 	uma_zone_t zone;
471 
472 	if (m && m->m_flags & M_EXT)
473 		printf("%s: %p mbuf already has cluster\n", __func__, m);
474 	if (m != NULL)
475 		m->m_ext.ext_buf = NULL;
476 
477 	switch (size) {
478 	case MCLBYTES:
479 		zone = zone_clust;
480 		break;
481 #if MJUMPAGESIZE != MCLBYTES
482 	case MJUMPAGESIZE:
483 		zone = zone_jumbop;
484 		break;
485 #endif
486 	case MJUM9BYTES:
487 		zone = zone_jumbo9;
488 		break;
489 	case MJUM16BYTES:
490 		zone = zone_jumbo16;
491 		break;
492 	default:
493 		panic("%s: m_getjcl: invalid cluster type", __func__);
494 	}
495 
496 	return (uma_zalloc_arg(zone, m, how));
497 }
498 
499 static __inline
500 void
501 m_chtype(struct mbuf *m, short new_type)
502 {
503 	m->m_type = new_type;
504 }
505 
506 /*
507  * mbuf, cluster, and external object allocation macros
508  * (for compatibility purposes).
509  */
510 /* NB: M_COPY_PKTHDR is deprecated.  Use M_MOVE_PKTHDR or m_dup_pktdr. */
511 #define	M_MOVE_PKTHDR(to, from)	m_move_pkthdr((to), (from))
512 #define	MGET(m, how, type)	((m) = m_get((how), (type)))
513 #define	MGETHDR(m, how, type)	((m) = m_gethdr((how), (type)))
514 #define	MCLGET(m, how)		m_clget((m), (how))
515 #define	MEXTADD(m, buf, size, free, args, flags, type) 			\
516     m_extadd((m), (caddr_t)(buf), (size), (free), (args), (flags), (type))
517 
518 /*
519  * Evaluate TRUE if it's safe to write to the mbuf m's data region (this
520  * can be both the local data payload, or an external buffer area,
521  * depending on whether M_EXT is set).
522  */
523 #define	M_WRITABLE(m)	(!((m)->m_flags & M_RDONLY) &&			\
524 			 (!(((m)->m_flags & M_EXT)) ||			\
525 			 (*((m)->m_ext.ref_cnt) == 1)) )		\
526 
527 /* Check if the supplied mbuf has a packet header, or else panic. */
528 #define	M_ASSERTPKTHDR(m)						\
529 	KASSERT(m != NULL && m->m_flags & M_PKTHDR,			\
530 	    ("%s: no mbuf packet header!", __func__))
531 
532 /* Ensure that the supplied mbuf is a valid, non-free mbuf. */
533 /* XXX: Broken at the moment.  Need some UMA magic to make it work again. */
534 #define	M_ASSERTVALID(m)						\
535 	KASSERT((((struct mbuf *)m)->m_flags & 0) == 0,			\
536 	    ("%s: attempted use of a free mbuf!", __func__))
537 
538 /*
539  * Set the m_data pointer of a newly-allocated mbuf (m_get/MGET) to place
540  * an object of the specified size at the end of the mbuf, longword aligned.
541  */
542 #define	M_ALIGN(m, len) do {						\
543 	KASSERT(!((m)->m_flags & (M_PKTHDR|M_EXT)),			\
544 		("%s: M_ALIGN not normal mbuf", __func__));		\
545 	KASSERT((m)->m_data == (m)->m_dat,				\
546 		("%s: M_ALIGN not a virgin mbuf", __func__));		\
547 	(m)->m_data += (MLEN - (len)) & ~(sizeof(long) - 1);		\
548 } while (0)
549 
550 /*
551  * As above, for mbufs allocated with m_gethdr/MGETHDR
552  * or initialized by M_COPY_PKTHDR.
553  */
554 #define	MH_ALIGN(m, len) do {						\
555 	KASSERT((m)->m_flags & M_PKTHDR && !((m)->m_flags & M_EXT),	\
556 		("%s: MH_ALIGN not PKTHDR mbuf", __func__));		\
557 	KASSERT((m)->m_data == (m)->m_pktdat,				\
558 		("%s: MH_ALIGN not a virgin mbuf", __func__));		\
559 	(m)->m_data += (MHLEN - (len)) & ~(sizeof(long) - 1);		\
560 } while (0)
561 
562 /*
563  * Compute the amount of space available
564  * before the current start of data in an mbuf.
565  *
566  * The M_WRITABLE() is a temporary, conservative safety measure: the burden
567  * of checking writability of the mbuf data area rests solely with the caller.
568  */
569 #define	M_LEADINGSPACE(m)						\
570 	((m)->m_flags & M_EXT ?						\
571 	    (M_WRITABLE(m) ? (m)->m_data - (m)->m_ext.ext_buf : 0):	\
572 	    (m)->m_flags & M_PKTHDR ? (m)->m_data - (m)->m_pktdat :	\
573 	    (m)->m_data - (m)->m_dat)
574 
575 /*
576  * Compute the amount of space available
577  * after the end of data in an mbuf.
578  *
579  * The M_WRITABLE() is a temporary, conservative safety measure: the burden
580  * of checking writability of the mbuf data area rests solely with the caller.
581  */
582 #define	M_TRAILINGSPACE(m)						\
583 	((m)->m_flags & M_EXT ?						\
584 	    (M_WRITABLE(m) ? (m)->m_ext.ext_buf + (m)->m_ext.ext_size	\
585 		- ((m)->m_data + (m)->m_len) : 0) :			\
586 	    &(m)->m_dat[MLEN] - ((m)->m_data + (m)->m_len))
587 
588 /*
589  * Arrange to prepend space of size plen to mbuf m.
590  * If a new mbuf must be allocated, how specifies whether to wait.
591  * If the allocation fails, the original mbuf chain is freed and m is
592  * set to NULL.
593  */
594 #define	M_PREPEND(m, plen, how) do {					\
595 	struct mbuf **_mmp = &(m);					\
596 	struct mbuf *_mm = *_mmp;					\
597 	int _mplen = (plen);						\
598 	int __mhow = (how);						\
599 									\
600 	MBUF_CHECKSLEEP(how);						\
601 	if (M_LEADINGSPACE(_mm) >= _mplen) {				\
602 		_mm->m_data -= _mplen;					\
603 		_mm->m_len += _mplen;					\
604 	} else								\
605 		_mm = m_prepend(_mm, _mplen, __mhow);			\
606 	if (_mm != NULL && _mm->m_flags & M_PKTHDR)			\
607 		_mm->m_pkthdr.len += _mplen;				\
608 	*_mmp = _mm;							\
609 } while (0)
610 
611 /*
612  * Change mbuf to new type.
613  * This is a relatively expensive operation and should be avoided.
614  */
615 #define	MCHTYPE(m, t)	m_chtype((m), (t))
616 
617 /* Length to m_copy to copy all. */
618 #define	M_COPYALL	1000000000
619 
620 /* Compatibility with 4.3. */
621 #define	m_copy(m, o, l)	m_copym((m), (o), (l), M_DONTWAIT)
622 
623 extern	int max_datalen;		/* MHLEN - max_hdr */
624 extern	int max_hdr;			/* Largest link + protocol header */
625 extern	int max_linkhdr;		/* Largest link-level header */
626 extern	int max_protohdr;		/* Largest protocol header */
627 extern	struct mbstat mbstat;		/* General mbuf stats/infos */
628 extern	int nmbclusters;		/* Maximum number of clusters */
629 
630 struct uio;
631 
632 void		 m_adj(struct mbuf *, int);
633 void		 m_align(struct mbuf *, int);
634 int		 m_apply(struct mbuf *, int, int,
635 		    int (*)(void *, void *, u_int), void *);
636 int		 m_append(struct mbuf *, int, c_caddr_t);
637 void		 m_cat(struct mbuf *, struct mbuf *);
638 void		 m_extadd(struct mbuf *, caddr_t, u_int,
639 		    void (*)(void *, void *), void *, int, int);
640 void		 m_copyback(struct mbuf *, int, int, c_caddr_t);
641 void		 m_copydata(const struct mbuf *, int, int, caddr_t);
642 struct	mbuf	*m_copym(struct mbuf *, int, int, int);
643 struct	mbuf	*m_copymdata(struct mbuf *, struct mbuf *,
644 		    int, int, int, int);
645 struct	mbuf	*m_copypacket(struct mbuf *, int);
646 void		 m_copy_pkthdr(struct mbuf *, struct mbuf *);
647 struct	mbuf	*m_copyup(struct mbuf *n, int len, int dstoff);
648 struct	mbuf	*m_defrag(struct mbuf *, int);
649 void		 m_demote(struct mbuf *, int);
650 struct	mbuf	*m_devget(char *, int, int, struct ifnet *,
651 		    void (*)(char *, caddr_t, u_int));
652 struct	mbuf	*m_dup(struct mbuf *, int);
653 int		 m_dup_pkthdr(struct mbuf *, struct mbuf *, int);
654 u_int		 m_fixhdr(struct mbuf *);
655 struct	mbuf	*m_fragment(struct mbuf *, int, int);
656 void		 m_freem(struct mbuf *);
657 struct	mbuf	*m_getm(struct mbuf *, int, int, short);
658 struct	mbuf	*m_getptr(struct mbuf *, int, int *);
659 u_int		 m_length(struct mbuf *, struct mbuf **);
660 void		 m_move_pkthdr(struct mbuf *, struct mbuf *);
661 struct	mbuf	*m_prepend(struct mbuf *, int, int);
662 void		 m_print(const struct mbuf *, int);
663 struct	mbuf	*m_pulldown(struct mbuf *, int, int, int *);
664 struct	mbuf	*m_pullup(struct mbuf *, int);
665 int		m_sanity(struct mbuf *, int);
666 struct	mbuf	*m_split(struct mbuf *, int, int);
667 struct	mbuf	*m_uiotombuf(struct uio *, int, int, int);
668 struct	mbuf	*m_unshare(struct mbuf *, int how);
669 
670 /*-
671  * Network packets may have annotations attached by affixing a list
672  * of "packet tags" to the pkthdr structure.  Packet tags are
673  * dynamically allocated semi-opaque data structures that have
674  * a fixed header (struct m_tag) that specifies the size of the
675  * memory block and a <cookie,type> pair that identifies it.
676  * The cookie is a 32-bit unique unsigned value used to identify
677  * a module or ABI.  By convention this value is chose as the
678  * date+time that the module is created, expressed as the number of
679  * seconds since the epoch (e.g., using date -u +'%s').  The type value
680  * is an ABI/module-specific value that identifies a particular annotation
681  * and is private to the module.  For compatibility with systems
682  * like OpenBSD that define packet tags w/o an ABI/module cookie,
683  * the value PACKET_ABI_COMPAT is used to implement m_tag_get and
684  * m_tag_find compatibility shim functions and several tag types are
685  * defined below.  Users that do not require compatibility should use
686  * a private cookie value so that packet tag-related definitions
687  * can be maintained privately.
688  *
689  * Note that the packet tag returned by m_tag_alloc has the default
690  * memory alignment implemented by malloc.  To reference private data
691  * one can use a construct like:
692  *
693  *	struct m_tag *mtag = m_tag_alloc(...);
694  *	struct foo *p = (struct foo *)(mtag+1);
695  *
696  * if the alignment of struct m_tag is sufficient for referencing members
697  * of struct foo.  Otherwise it is necessary to embed struct m_tag within
698  * the private data structure to insure proper alignment; e.g.,
699  *
700  *	struct foo {
701  *		struct m_tag	tag;
702  *		...
703  *	};
704  *	struct foo *p = (struct foo *) m_tag_alloc(...);
705  *	struct m_tag *mtag = &p->tag;
706  */
707 
708 /*
709  * Persistent tags stay with an mbuf until the mbuf is reclaimed.
710  * Otherwise tags are expected to ``vanish'' when they pass through
711  * a network interface.  For most interfaces this happens normally
712  * as the tags are reclaimed when the mbuf is free'd.  However in
713  * some special cases reclaiming must be done manually.  An example
714  * is packets that pass through the loopback interface.  Also, one
715  * must be careful to do this when ``turning around'' packets (e.g.,
716  * icmp_reflect).
717  *
718  * To mark a tag persistent bit-or this flag in when defining the
719  * tag id.  The tag will then be treated as described above.
720  */
721 #define	MTAG_PERSISTENT				0x800
722 
723 #define	PACKET_TAG_NONE				0  /* Nadda */
724 
725 /* Packet tags for use with PACKET_ABI_COMPAT. */
726 #define	PACKET_TAG_IPSEC_IN_DONE		1  /* IPsec applied, in */
727 #define	PACKET_TAG_IPSEC_OUT_DONE		2  /* IPsec applied, out */
728 #define	PACKET_TAG_IPSEC_IN_CRYPTO_DONE		3  /* NIC IPsec crypto done */
729 #define	PACKET_TAG_IPSEC_OUT_CRYPTO_NEEDED	4  /* NIC IPsec crypto req'ed */
730 #define	PACKET_TAG_IPSEC_IN_COULD_DO_CRYPTO	5  /* NIC notifies IPsec */
731 #define	PACKET_TAG_IPSEC_PENDING_TDB		6  /* Reminder to do IPsec */
732 #define	PACKET_TAG_BRIDGE			7  /* Bridge processing done */
733 #define	PACKET_TAG_GIF				8  /* GIF processing done */
734 #define	PACKET_TAG_GRE				9  /* GRE processing done */
735 #define	PACKET_TAG_IN_PACKET_CHECKSUM		10 /* NIC checksumming done */
736 #define	PACKET_TAG_ENCAP			11 /* Encap.  processing */
737 #define	PACKET_TAG_IPSEC_SOCKET			12 /* IPSEC socket ref */
738 #define	PACKET_TAG_IPSEC_HISTORY		13 /* IPSEC history */
739 #define	PACKET_TAG_IPV6_INPUT			14 /* IPV6 input processing */
740 #define	PACKET_TAG_DUMMYNET			15 /* dummynet info */
741 #define	PACKET_TAG_DIVERT			17 /* divert info */
742 #define	PACKET_TAG_IPFORWARD			18 /* ipforward info */
743 #define	PACKET_TAG_MACLABEL	(19 | MTAG_PERSISTENT) /* MAC label */
744 #define	PACKET_TAG_PF_ROUTED			21 /* PF routed, avoid loops */
745 #define	PACKET_TAG_PF_FRAGCACHE			22 /* PF fragment cached */
746 #define	PACKET_TAG_PF_QID			23 /* PF ALTQ queue id */
747 #define	PACKET_TAG_PF_TAG			24 /* PF tagged */
748 #define	PACKET_TAG_RTSOCKFAM			25 /* rtsock sa family */
749 #define	PACKET_TAG_PF_TRANSLATE_LOCALHOST	26 /* PF translate localhost */
750 #define	PACKET_TAG_IPOPTIONS			27 /* Saved IP options */
751 #define	PACKET_TAG_CARP                         28 /* CARP info */
752 
753 /* Specific cookies and tags. */
754 #define	MTAG_VLAN				1035328035
755 #define	MTAG_VLAN_TAG				0 /* tag of VLAN interface */
756 
757 /* Packet tag routines. */
758 struct	m_tag	*m_tag_alloc(u_int32_t, int, int, int);
759 void		 m_tag_delete(struct mbuf *, struct m_tag *);
760 void		 m_tag_delete_chain(struct mbuf *, struct m_tag *);
761 void		 m_tag_free_default(struct m_tag *);
762 struct	m_tag	*m_tag_locate(struct mbuf *, u_int32_t, int, struct m_tag *);
763 struct	m_tag	*m_tag_copy(struct m_tag *, int);
764 int		 m_tag_copy_chain(struct mbuf *, struct mbuf *, int);
765 void		 m_tag_delete_nonpersistent(struct mbuf *);
766 
767 /*
768  * Initialize the list of tags associated with an mbuf.
769  */
770 static __inline void
771 m_tag_init(struct mbuf *m)
772 {
773 	SLIST_INIT(&m->m_pkthdr.tags);
774 }
775 
776 /*
777  * Set up the contents of a tag.  Note that this does not
778  * fill in the free method; the caller is expected to do that.
779  *
780  * XXX probably should be called m_tag_init, but that was
781  * already taken.
782  */
783 static __inline void
784 m_tag_setup(struct m_tag *t, u_int32_t cookie, int type, int len)
785 {
786 	t->m_tag_id = type;
787 	t->m_tag_len = len;
788 	t->m_tag_cookie = cookie;
789 }
790 
791 /*
792  * Reclaim resources associated with a tag.
793  */
794 static __inline void
795 m_tag_free(struct m_tag *t)
796 {
797 	(*t->m_tag_free)(t);
798 }
799 
800 /*
801  * Return the first tag associated with an mbuf.
802  */
803 static __inline struct m_tag *
804 m_tag_first(struct mbuf *m)
805 {
806 	return (SLIST_FIRST(&m->m_pkthdr.tags));
807 }
808 
809 /*
810  * Return the next tag in the list of tags associated with an mbuf.
811  */
812 static __inline struct m_tag *
813 m_tag_next(struct mbuf *m, struct m_tag *t)
814 {
815 	return (SLIST_NEXT(t, m_tag_link));
816 }
817 
818 /*
819  * Prepend a tag to the list of tags associated with an mbuf.
820  */
821 static __inline void
822 m_tag_prepend(struct mbuf *m, struct m_tag *t)
823 {
824 	SLIST_INSERT_HEAD(&m->m_pkthdr.tags, t, m_tag_link);
825 }
826 
827 /*
828  * Unlink a tag from the list of tags associated with an mbuf.
829  */
830 static __inline void
831 m_tag_unlink(struct mbuf *m, struct m_tag *t)
832 {
833 	SLIST_REMOVE(&m->m_pkthdr.tags, t, m_tag, m_tag_link);
834 }
835 
836 /* These are for OpenBSD compatibility. */
837 #define	MTAG_ABI_COMPAT		0		/* compatibility ABI */
838 
839 static __inline struct m_tag *
840 m_tag_get(int type, int length, int wait)
841 {
842 	return (m_tag_alloc(MTAG_ABI_COMPAT, type, length, wait));
843 }
844 
845 static __inline struct m_tag *
846 m_tag_find(struct mbuf *m, int type, struct m_tag *start)
847 {
848 	return (SLIST_EMPTY(&m->m_pkthdr.tags) ? (struct m_tag *)NULL :
849 	    m_tag_locate(m, MTAG_ABI_COMPAT, type, start));
850 }
851 
852 #endif /* _KERNEL */
853 
854 #endif /* !_SYS_MBUF_H_ */
855