mbuf.h revision 152580
1/*-
2 * Copyright (c) 1982, 1986, 1988, 1993
3 *	The Regents of the University of California.  All rights reserved.
4 *
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions
7 * are met:
8 * 1. Redistributions of source code must retain the above copyright
9 *    notice, this list of conditions and the following disclaimer.
10 * 2. Redistributions in binary form must reproduce the above copyright
11 *    notice, this list of conditions and the following disclaimer in the
12 *    documentation and/or other materials provided with the distribution.
13 * 3. Neither the name of the University nor the names of its contributors
14 *    may be used to endorse or promote products derived from this software
15 *    without specific prior written permission.
16 *
17 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
18 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
19 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
20 * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
21 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
22 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
23 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
24 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
25 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
26 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
27 * SUCH DAMAGE.
28 *
29 *	@(#)mbuf.h	8.5 (Berkeley) 2/19/95
30 * $FreeBSD: head/sys/sys/mbuf.h 152580 2005-11-18 14:40:43Z andre $
31 */
32
33#ifndef _SYS_MBUF_H_
34#define	_SYS_MBUF_H_
35
36/* XXX: These includes suck. Sorry! */
37#include <sys/queue.h>
38#ifdef _KERNEL
39#include <sys/systm.h>
40#include <vm/uma.h>
41#ifdef WITNESS
42#include <sys/lock.h>
43#endif
44#endif
45
46/*
47 * Mbufs are of a single size, MSIZE (sys/param.h), which
48 * includes overhead.  An mbuf may add a single "mbuf cluster" of size
49 * MCLBYTES (also in sys/param.h), which has no additional overhead
50 * and is used instead of the internal data area; this is done when
51 * at least MINCLSIZE of data must be stored.  Additionally, it is possible
52 * to allocate a separate buffer externally and attach it to the mbuf in
53 * a way similar to that of mbuf clusters.
54 */
55#define	MLEN		(MSIZE - sizeof(struct m_hdr))	/* normal data len */
56#define	MHLEN		(MLEN - sizeof(struct pkthdr))	/* data len w/pkthdr */
57#define	MINCLSIZE	(MHLEN + 1)	/* smallest amount to put in cluster */
58#define	M_MAXCOMPRESS	(MHLEN / 2)	/* max amount to copy for compression */
59
60#ifdef _KERNEL
61/*-
62 * Macros for type conversion:
63 * mtod(m, t)	-- Convert mbuf pointer to data pointer of correct type.
64 * dtom(x)	-- Convert data pointer within mbuf to mbuf pointer (XXX).
65 */
66#define	mtod(m, t)	((t)((m)->m_data))
67#define	dtom(x)		((struct mbuf *)((intptr_t)(x) & ~(MSIZE-1)))
68
69/*
70 * Argument structure passed to UMA routines during mbuf and packet
71 * allocations.
72 */
73struct mb_args {
74	int	flags;	/* Flags for mbuf being allocated */
75	short	type;	/* Type of mbuf being allocated */
76};
77#endif /* _KERNEL */
78
79/*
80 * Header present at the beginning of every mbuf.
81 */
82struct 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 */
94struct 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 */
105struct 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 */
119struct 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 */
133struct 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_JUMBO9	3	/* jumbo cluster 9216 bytes */
189#define	EXT_JUMBO16	4	/* jumbo cluster 16184 bytes */
190#define	EXT_PACKET	5	/* mbuf+cluster from packet zone */
191#define	EXT_NET_DRV	100	/* custom ext_buf provided by net driver(s) */
192#define	EXT_MOD_TYPE	200	/* custom module's ext_buf type */
193#define	EXT_DISPOSABLE	300	/* can throw this buffer away w/page flipping */
194#define	EXT_EXTREF	400	/* has externally maintained ref_cnt ptr */
195
196/*
197 * Flags copied when copying m_pkthdr.
198 */
199#define	M_COPYFLAGS	(M_PKTHDR|M_EOR|M_RDONLY|M_PROTO1|M_PROTO1|M_PROTO2|\
200			    M_PROTO3|M_PROTO4|M_PROTO5|M_SKIP_FIREWALL|\
201			    M_BCAST|M_MCAST|M_FRAG|M_FIRSTFRAG|M_LASTFRAG|\
202			    M_VLANTAG)
203
204/*
205 * Flags indicating hw checksum support and sw checksum requirements.
206 */
207#define	CSUM_IP			0x0001		/* will csum IP */
208#define	CSUM_TCP		0x0002		/* will csum TCP */
209#define	CSUM_UDP		0x0004		/* will csum UDP */
210#define	CSUM_IP_FRAGS		0x0008		/* will csum IP fragments */
211#define	CSUM_FRAGMENT		0x0010		/* will do IP fragmentation */
212
213#define	CSUM_IP_CHECKED		0x0100		/* did csum IP */
214#define	CSUM_IP_VALID		0x0200		/*   ... the csum is valid */
215#define	CSUM_DATA_VALID		0x0400		/* csum_data field is valid */
216#define	CSUM_PSEUDO_HDR		0x0800		/* csum_data has pseudo hdr */
217
218#define	CSUM_DELAY_DATA		(CSUM_TCP | CSUM_UDP)
219#define	CSUM_DELAY_IP		(CSUM_IP)	/* XXX add ipv6 here too? */
220
221/*
222 * mbuf types.
223 */
224#define	MT_NOTMBUF	0	/* USED INTERNALLY ONLY! Object is not mbuf */
225#define	MT_DATA		1	/* dynamic (data) allocation */
226#define	MT_HEADER	MT_DATA	/* packet header, use M_PKTHDR instead */
227#if 0
228#define	MT_SOCKET	3	/* socket structure */
229#define	MT_PCB		4	/* protocol control block */
230#define	MT_RTABLE	5	/* routing tables */
231#define	MT_HTABLE	6	/* IMP host tables */
232#define	MT_ATABLE	7	/* address resolution tables */
233#endif
234#define	MT_SONAME	8	/* socket name */
235#if 0
236#define	MT_SOOPTS	10	/* socket options */
237#endif
238#define	MT_FTABLE	11	/* fragment reassembly header */
239#if 0
240#define	MT_RIGHTS	12	/* access rights */
241#define	MT_IFADDR	13	/* interface address */
242#endif
243#define	MT_CONTROL	14	/* extra-data protocol message */
244#define	MT_OOBDATA	15	/* expedited data  */
245#define	MT_NTYPES	16	/* number of mbuf types for mbtypes[] */
246
247#define	MT_NOINIT	255	/* Not a type but a flag to allocate
248				   a non-initialized mbuf */
249
250/*
251 * General mbuf allocator statistics structure.
252 */
253struct mbstat {
254	u_long	m_mbufs;	/* XXX */
255	u_long	m_mclusts;	/* XXX */
256
257	u_long	m_drain;	/* times drained protocols for space */
258	u_long	m_mcfail;	/* XXX: times m_copym failed */
259	u_long	m_mpfail;	/* XXX: times m_pullup failed */
260	u_long	m_msize;	/* length of an mbuf */
261	u_long	m_mclbytes;	/* length of an mbuf cluster */
262	u_long	m_minclsize;	/* min length of data to allocate a cluster */
263	u_long	m_mlen;		/* length of data in an mbuf */
264	u_long	m_mhlen;	/* length of data in a header mbuf */
265
266	/* Number of mbtypes (gives # elems in mbtypes[] array: */
267	short	m_numtypes;
268
269	/* XXX: Sendfile stats should eventually move to their own struct */
270	u_long	sf_iocnt;	/* times sendfile had to do disk I/O */
271	u_long	sf_allocfail;	/* times sfbuf allocation failed */
272	u_long	sf_allocwait;	/* times sfbuf allocation had to wait */
273};
274
275/*
276 * Flags specifying how an allocation should be made.
277 *
278 * The flag to use is as follows:
279 * - M_DONTWAIT or M_NOWAIT from an interrupt handler to not block allocation.
280 * - M_WAIT or M_WAITOK or M_TRYWAIT from wherever it is safe to block.
281 *
282 * M_DONTWAIT/M_NOWAIT means that we will not block the thread explicitly
283 * and if we cannot allocate immediately we may return NULL,
284 * whereas M_WAIT/M_WAITOK/M_TRYWAIT means that if we cannot allocate
285 * resources we will block until they are available, and thus never
286 * return NULL.
287 *
288 * XXX Eventually just phase this out to use M_WAITOK/M_NOWAIT.
289 */
290#define	MBTOM(how)	(how)
291#define	M_DONTWAIT	M_NOWAIT
292#define	M_TRYWAIT	M_WAITOK
293#define	M_WAIT		M_WAITOK
294
295/*
296 * String names of mbuf-related UMA(9) and malloc(9) types.  Exposed to
297 * !_KERNEL so that monitoring tools can look up the zones with
298 * libmemstat(3).
299 */
300#define	MBUF_MEM_NAME		"mbuf"
301#define	MBUF_CLUSTER_MEM_NAME	"mbuf_cluster"
302#define	MBUF_PACKET_MEM_NAME	"mbuf_packet"
303#define	MBUF_JUMBO9_MEM_NAME	"mbuf_jumbo_9k"
304#define	MBUF_JUMBO16_MEM_NAME	"mbuf_jumbo_16k"
305#define	MBUF_TAG_MEM_NAME	"mbuf_tag"
306#define	MBUF_EXTREFCNT_MEM_NAME	"mbuf_ext_refcnt"
307
308#ifdef _KERNEL
309
310#ifdef WITNESS
311#define MBUF_CHECKSLEEP(how) do {					\
312	if (how == M_WAITOK)						\
313		WITNESS_WARN(WARN_GIANTOK | WARN_SLEEPOK, NULL,		\
314		    "Sleeping in \"%s\"", __func__);			\
315} while (0)
316#else
317#define MBUF_CHECKSLEEP(how)
318#endif
319
320/*
321 * Network buffer allocation API
322 *
323 * The rest of it is defined in kern/kern_mbuf.c
324 */
325
326extern uma_zone_t	zone_mbuf;
327extern uma_zone_t	zone_clust;
328extern uma_zone_t	zone_pack;
329extern uma_zone_t	zone_jumbo9;
330extern uma_zone_t	zone_jumbo16;
331extern uma_zone_t	zone_ext_refcnt;
332
333static __inline struct mbuf	*m_get(int how, short type);
334static __inline struct mbuf	*m_gethdr(int how, short type);
335static __inline struct mbuf	*m_getcl(int how, short type, int flags);
336static __inline struct mbuf	*m_getclr(int how, short type);	/* XXX */
337static __inline struct mbuf	*m_free(struct mbuf *m);
338static __inline void		 m_clget(struct mbuf *m, int how);
339static __inline void		 m_chtype(struct mbuf *m, short new_type);
340void				 mb_free_ext(struct mbuf *);
341
342static __inline
343struct mbuf *
344m_get(int how, short type)
345{
346	struct mb_args args;
347
348	args.flags = 0;
349	args.type = type;
350	return (uma_zalloc_arg(zone_mbuf, &args, how));
351}
352
353/* XXX This should be depracated, very little use */
354static __inline
355struct mbuf *
356m_getclr(int how, short type)
357{
358	struct mbuf *m;
359	struct mb_args args;
360
361	args.flags = 0;
362	args.type = type;
363	m = uma_zalloc_arg(zone_mbuf, &args, how);
364	if (m != NULL)
365		bzero(m->m_data, MLEN);
366	return m;
367}
368
369static __inline
370struct mbuf *
371m_gethdr(int how, short type)
372{
373	struct mb_args args;
374
375	args.flags = M_PKTHDR;
376	args.type = type;
377	return (uma_zalloc_arg(zone_mbuf, &args, how));
378}
379
380static __inline
381struct mbuf *
382m_getcl(int how, short type, int flags)
383{
384	struct mb_args args;
385
386	args.flags = flags;
387	args.type = type;
388	return (uma_zalloc_arg(zone_pack, &args, how));
389}
390
391static __inline
392struct mbuf *
393m_free(struct mbuf *m)
394{
395	struct mbuf *n = m->m_next;
396
397	if (m->m_flags & M_EXT)
398		mb_free_ext(m);
399	else
400		uma_zfree(zone_mbuf, m);
401	return n;
402}
403
404static __inline
405void
406m_clget(struct mbuf *m, int how)
407{
408	if (m->m_flags & M_EXT)
409		printf("%s: %p mbuf already has cluster\n", __func__, m);
410	m->m_ext.ext_buf = NULL;
411	uma_zalloc_arg(zone_clust, m, how);
412}
413
414static __inline
415void
416m_chtype(struct mbuf *m, short new_type)
417{
418	m->m_type = new_type;
419}
420
421/*
422 * mbuf, cluster, and external object allocation macros
423 * (for compatibility purposes).
424 */
425/* NB: M_COPY_PKTHDR is deprecated.  Use M_MOVE_PKTHDR or m_dup_pktdr. */
426#define	M_MOVE_PKTHDR(to, from)	m_move_pkthdr((to), (from))
427#define	MGET(m, how, type)	((m) = m_get((how), (type)))
428#define	MGETHDR(m, how, type)	((m) = m_gethdr((how), (type)))
429#define	MCLGET(m, how)		m_clget((m), (how))
430#define	MEXTADD(m, buf, size, free, args, flags, type) 			\
431    m_extadd((m), (caddr_t)(buf), (size), (free), (args), (flags), (type))
432
433/*
434 * Evaluate TRUE if it's safe to write to the mbuf m's data region (this
435 * can be both the local data payload, or an external buffer area,
436 * depending on whether M_EXT is set).
437 */
438#define	M_WRITABLE(m)	(!((m)->m_flags & M_RDONLY) &&			\
439			 (!(((m)->m_flags & M_EXT)) ||			\
440			 (*((m)->m_ext.ref_cnt) == 1)) )		\
441
442/* Check if the supplied mbuf has a packet header, or else panic. */
443#define	M_ASSERTPKTHDR(m)						\
444	KASSERT(m != NULL && m->m_flags & M_PKTHDR,			\
445	    ("%s: no mbuf packet header!", __func__))
446
447/* Ensure that the supplied mbuf is a valid, non-free mbuf. */
448/* XXX: Broken at the moment.  Need some UMA magic to make it work again. */
449#define	M_ASSERTVALID(m)						\
450	KASSERT((((struct mbuf *)m)->m_flags & 0) == 0,			\
451	    ("%s: attempted use of a free mbuf!", __func__))
452
453/*
454 * Set the m_data pointer of a newly-allocated mbuf (m_get/MGET) to place
455 * an object of the specified size at the end of the mbuf, longword aligned.
456 */
457#define	M_ALIGN(m, len) do {						\
458	KASSERT(!((m)->m_flags & (M_PKTHDR|M_EXT)),			\
459		("%s: M_ALIGN not normal mbuf", __func__));		\
460	KASSERT((m)->m_data == (m)->m_dat,				\
461		("%s: M_ALIGN not a virgin mbuf", __func__));		\
462	(m)->m_data += (MLEN - (len)) & ~(sizeof(long) - 1);		\
463} while (0)
464
465/*
466 * As above, for mbufs allocated with m_gethdr/MGETHDR
467 * or initialized by M_COPY_PKTHDR.
468 */
469#define	MH_ALIGN(m, len) do {						\
470	KASSERT((m)->m_flags & M_PKTHDR && !((m)->m_flags & M_EXT),	\
471		("%s: MH_ALIGN not PKTHDR mbuf", __func__));		\
472	KASSERT((m)->m_data == (m)->m_pktdat,				\
473		("%s: MH_ALIGN not a virgin mbuf", __func__));		\
474	(m)->m_data += (MHLEN - (len)) & ~(sizeof(long) - 1);		\
475} while (0)
476
477/*
478 * Compute the amount of space available
479 * before the current start of data in an mbuf.
480 *
481 * The M_WRITABLE() is a temporary, conservative safety measure: the burden
482 * of checking writability of the mbuf data area rests solely with the caller.
483 */
484#define	M_LEADINGSPACE(m)						\
485	((m)->m_flags & M_EXT ?						\
486	    (M_WRITABLE(m) ? (m)->m_data - (m)->m_ext.ext_buf : 0):	\
487	    (m)->m_flags & M_PKTHDR ? (m)->m_data - (m)->m_pktdat :	\
488	    (m)->m_data - (m)->m_dat)
489
490/*
491 * Compute the amount of space available
492 * after the end of data in an mbuf.
493 *
494 * The M_WRITABLE() is a temporary, conservative safety measure: the burden
495 * of checking writability of the mbuf data area rests solely with the caller.
496 */
497#define	M_TRAILINGSPACE(m)						\
498	((m)->m_flags & M_EXT ?						\
499	    (M_WRITABLE(m) ? (m)->m_ext.ext_buf + (m)->m_ext.ext_size	\
500		- ((m)->m_data + (m)->m_len) : 0) :			\
501	    &(m)->m_dat[MLEN] - ((m)->m_data + (m)->m_len))
502
503/*
504 * Arrange to prepend space of size plen to mbuf m.
505 * If a new mbuf must be allocated, how specifies whether to wait.
506 * If the allocation fails, the original mbuf chain is freed and m is
507 * set to NULL.
508 */
509#define	M_PREPEND(m, plen, how) do {					\
510	struct mbuf **_mmp = &(m);					\
511	struct mbuf *_mm = *_mmp;					\
512	int _mplen = (plen);						\
513	int __mhow = (how);						\
514									\
515	MBUF_CHECKSLEEP(how);						\
516	if (M_LEADINGSPACE(_mm) >= _mplen) {				\
517		_mm->m_data -= _mplen;					\
518		_mm->m_len += _mplen;					\
519	} else								\
520		_mm = m_prepend(_mm, _mplen, __mhow);			\
521	if (_mm != NULL && _mm->m_flags & M_PKTHDR)			\
522		_mm->m_pkthdr.len += _mplen;				\
523	*_mmp = _mm;							\
524} while (0)
525
526/*
527 * Change mbuf to new type.
528 * This is a relatively expensive operation and should be avoided.
529 */
530#define	MCHTYPE(m, t)	m_chtype((m), (t))
531
532/* Length to m_copy to copy all. */
533#define	M_COPYALL	1000000000
534
535/* Compatibility with 4.3. */
536#define	m_copy(m, o, l)	m_copym((m), (o), (l), M_DONTWAIT)
537
538extern	int max_datalen;		/* MHLEN - max_hdr */
539extern	int max_hdr;			/* Largest link + protocol header */
540extern	int max_linkhdr;		/* Largest link-level header */
541extern	int max_protohdr;		/* Largest protocol header */
542extern	struct mbstat mbstat;		/* General mbuf stats/infos */
543extern	int nmbclusters;		/* Maximum number of clusters */
544
545struct uio;
546
547void		 m_adj(struct mbuf *, int);
548void		 m_align(struct mbuf *, int);
549int		 m_apply(struct mbuf *, int, int,
550		    int (*)(void *, void *, u_int), void *);
551int		 m_append(struct mbuf *, int, c_caddr_t);
552void		 m_cat(struct mbuf *, struct mbuf *);
553void		 m_extadd(struct mbuf *, caddr_t, u_int,
554		    void (*)(void *, void *), void *, int, int);
555void		 m_copyback(struct mbuf *, int, int, c_caddr_t);
556void		 m_copydata(const struct mbuf *, int, int, caddr_t);
557struct	mbuf	*m_copym(struct mbuf *, int, int, int);
558struct	mbuf	*m_copymdata(struct mbuf *, struct mbuf *,
559		    int, int, int, int);
560struct	mbuf	*m_copypacket(struct mbuf *, int);
561void		 m_copy_pkthdr(struct mbuf *, struct mbuf *);
562struct	mbuf	*m_copyup(struct mbuf *n, int len, int dstoff);
563struct	mbuf	*m_defrag(struct mbuf *, int);
564void		 m_demote(struct mbuf *, int);
565struct	mbuf	*m_devget(char *, int, int, struct ifnet *,
566		    void (*)(char *, caddr_t, u_int));
567struct	mbuf	*m_dup(struct mbuf *, int);
568int		 m_dup_pkthdr(struct mbuf *, struct mbuf *, int);
569u_int		 m_fixhdr(struct mbuf *);
570struct	mbuf	*m_fragment(struct mbuf *, int, int);
571void		 m_freem(struct mbuf *);
572struct	mbuf	*m_getm(struct mbuf *, int, int, short);
573struct	mbuf	*m_getptr(struct mbuf *, int, int *);
574u_int		 m_length(struct mbuf *, struct mbuf **);
575void		 m_move_pkthdr(struct mbuf *, struct mbuf *);
576struct	mbuf	*m_prepend(struct mbuf *, int, int);
577void		 m_print(const struct mbuf *, int);
578struct	mbuf	*m_pulldown(struct mbuf *, int, int, int *);
579struct	mbuf	*m_pullup(struct mbuf *, int);
580int		m_sanity(struct mbuf *, int);
581struct	mbuf	*m_split(struct mbuf *, int, int);
582struct	mbuf	*m_uiotombuf(struct uio *, int, int, int);
583
584/*-
585 * Network packets may have annotations attached by affixing a list
586 * of "packet tags" to the pkthdr structure.  Packet tags are
587 * dynamically allocated semi-opaque data structures that have
588 * a fixed header (struct m_tag) that specifies the size of the
589 * memory block and a <cookie,type> pair that identifies it.
590 * The cookie is a 32-bit unique unsigned value used to identify
591 * a module or ABI.  By convention this value is chose as the
592 * date+time that the module is created, expressed as the number of
593 * seconds since the epoch (e.g., using date -u +'%s').  The type value
594 * is an ABI/module-specific value that identifies a particular annotation
595 * and is private to the module.  For compatibility with systems
596 * like OpenBSD that define packet tags w/o an ABI/module cookie,
597 * the value PACKET_ABI_COMPAT is used to implement m_tag_get and
598 * m_tag_find compatibility shim functions and several tag types are
599 * defined below.  Users that do not require compatibility should use
600 * a private cookie value so that packet tag-related definitions
601 * can be maintained privately.
602 *
603 * Note that the packet tag returned by m_tag_alloc has the default
604 * memory alignment implemented by malloc.  To reference private data
605 * one can use a construct like:
606 *
607 *	struct m_tag *mtag = m_tag_alloc(...);
608 *	struct foo *p = (struct foo *)(mtag+1);
609 *
610 * if the alignment of struct m_tag is sufficient for referencing members
611 * of struct foo.  Otherwise it is necessary to embed struct m_tag within
612 * the private data structure to insure proper alignment; e.g.,
613 *
614 *	struct foo {
615 *		struct m_tag	tag;
616 *		...
617 *	};
618 *	struct foo *p = (struct foo *) m_tag_alloc(...);
619 *	struct m_tag *mtag = &p->tag;
620 */
621
622/*
623 * Persistent tags stay with an mbuf until the mbuf is reclaimed.
624 * Otherwise tags are expected to ``vanish'' when they pass through
625 * a network interface.  For most interfaces this happens normally
626 * as the tags are reclaimed when the mbuf is free'd.  However in
627 * some special cases reclaiming must be done manually.  An example
628 * is packets that pass through the loopback interface.  Also, one
629 * must be careful to do this when ``turning around'' packets (e.g.,
630 * icmp_reflect).
631 *
632 * To mark a tag persistent bit-or this flag in when defining the
633 * tag id.  The tag will then be treated as described above.
634 */
635#define	MTAG_PERSISTENT				0x800
636
637#define	PACKET_TAG_NONE				0  /* Nadda */
638
639/* Packet tags for use with PACKET_ABI_COMPAT. */
640#define	PACKET_TAG_IPSEC_IN_DONE		1  /* IPsec applied, in */
641#define	PACKET_TAG_IPSEC_OUT_DONE		2  /* IPsec applied, out */
642#define	PACKET_TAG_IPSEC_IN_CRYPTO_DONE		3  /* NIC IPsec crypto done */
643#define	PACKET_TAG_IPSEC_OUT_CRYPTO_NEEDED	4  /* NIC IPsec crypto req'ed */
644#define	PACKET_TAG_IPSEC_IN_COULD_DO_CRYPTO	5  /* NIC notifies IPsec */
645#define	PACKET_TAG_IPSEC_PENDING_TDB		6  /* Reminder to do IPsec */
646#define	PACKET_TAG_BRIDGE			7  /* Bridge processing done */
647#define	PACKET_TAG_GIF				8  /* GIF processing done */
648#define	PACKET_TAG_GRE				9  /* GRE processing done */
649#define	PACKET_TAG_IN_PACKET_CHECKSUM		10 /* NIC checksumming done */
650#define	PACKET_TAG_ENCAP			11 /* Encap.  processing */
651#define	PACKET_TAG_IPSEC_SOCKET			12 /* IPSEC socket ref */
652#define	PACKET_TAG_IPSEC_HISTORY		13 /* IPSEC history */
653#define	PACKET_TAG_IPV6_INPUT			14 /* IPV6 input processing */
654#define	PACKET_TAG_DUMMYNET			15 /* dummynet info */
655#define	PACKET_TAG_DIVERT			17 /* divert info */
656#define	PACKET_TAG_IPFORWARD			18 /* ipforward info */
657#define	PACKET_TAG_MACLABEL	(19 | MTAG_PERSISTENT) /* MAC label */
658#define	PACKET_TAG_PF_ROUTED			21 /* PF routed, avoid loops */
659#define	PACKET_TAG_PF_FRAGCACHE			22 /* PF fragment cached */
660#define	PACKET_TAG_PF_QID			23 /* PF ALTQ queue id */
661#define	PACKET_TAG_PF_TAG			24 /* PF tagged */
662#define	PACKET_TAG_RTSOCKFAM			25 /* rtsock sa family */
663#define	PACKET_TAG_PF_TRANSLATE_LOCALHOST	26 /* PF translate localhost */
664#define	PACKET_TAG_IPOPTIONS			27 /* Saved IP options */
665#define	PACKET_TAG_CARP                         28 /* CARP info */
666
667/* Packet tag routines. */
668struct	m_tag	*m_tag_alloc(u_int32_t, int, int, int);
669void		 m_tag_delete(struct mbuf *, struct m_tag *);
670void		 m_tag_delete_chain(struct mbuf *, struct m_tag *);
671void		 m_tag_free_default(struct m_tag *);
672struct	m_tag	*m_tag_locate(struct mbuf *, u_int32_t, int, struct m_tag *);
673struct	m_tag	*m_tag_copy(struct m_tag *, int);
674int		 m_tag_copy_chain(struct mbuf *, struct mbuf *, int);
675void		 m_tag_delete_nonpersistent(struct mbuf *);
676
677/*
678 * Initialize the list of tags associated with an mbuf.
679 */
680static __inline void
681m_tag_init(struct mbuf *m)
682{
683	SLIST_INIT(&m->m_pkthdr.tags);
684}
685
686/*
687 * Set up the contents of a tag.  Note that this does not
688 * fill in the free method; the caller is expected to do that.
689 *
690 * XXX probably should be called m_tag_init, but that was
691 * already taken.
692 */
693static __inline void
694m_tag_setup(struct m_tag *t, u_int32_t cookie, int type, int len)
695{
696	t->m_tag_id = type;
697	t->m_tag_len = len;
698	t->m_tag_cookie = cookie;
699}
700
701/*
702 * Reclaim resources associated with a tag.
703 */
704static __inline void
705m_tag_free(struct m_tag *t)
706{
707	(*t->m_tag_free)(t);
708}
709
710/*
711 * Return the first tag associated with an mbuf.
712 */
713static __inline struct m_tag *
714m_tag_first(struct mbuf *m)
715{
716	return (SLIST_FIRST(&m->m_pkthdr.tags));
717}
718
719/*
720 * Return the next tag in the list of tags associated with an mbuf.
721 */
722static __inline struct m_tag *
723m_tag_next(struct mbuf *m, struct m_tag *t)
724{
725	return (SLIST_NEXT(t, m_tag_link));
726}
727
728/*
729 * Prepend a tag to the list of tags associated with an mbuf.
730 */
731static __inline void
732m_tag_prepend(struct mbuf *m, struct m_tag *t)
733{
734	SLIST_INSERT_HEAD(&m->m_pkthdr.tags, t, m_tag_link);
735}
736
737/*
738 * Unlink a tag from the list of tags associated with an mbuf.
739 */
740static __inline void
741m_tag_unlink(struct mbuf *m, struct m_tag *t)
742{
743	SLIST_REMOVE(&m->m_pkthdr.tags, t, m_tag, m_tag_link);
744}
745
746/* These are for OpenBSD compatibility. */
747#define	MTAG_ABI_COMPAT		0		/* compatibility ABI */
748
749static __inline struct m_tag *
750m_tag_get(int type, int length, int wait)
751{
752	return (m_tag_alloc(MTAG_ABI_COMPAT, type, length, wait));
753}
754
755static __inline struct m_tag *
756m_tag_find(struct mbuf *m, int type, struct m_tag *start)
757{
758	return (SLIST_EMPTY(&m->m_pkthdr.tags) ?
759	    NULL : m_tag_locate(m, MTAG_ABI_COMPAT, type, start));
760}
761
762#endif /* _KERNEL */
763
764#endif /* !_SYS_MBUF_H_ */
765