mbuf.h revision 123929
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. All advertising materials mentioning features or use of this software
14 *    must display the following acknowledgement:
15 *	This product includes software developed by the University of
16 *	California, Berkeley and its contributors.
17 * 4. Neither the name of the University nor the names of its contributors
18 *    may be used to endorse or promote products derived from this software
19 *    without specific prior written permission.
20 *
21 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
22 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
23 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
24 * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
25 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
26 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
27 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
28 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
29 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
30 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
31 * SUCH DAMAGE.
32 *
33 *	@(#)mbuf.h	8.5 (Berkeley) 2/19/95
34 * $FreeBSD: head/sys/sys/mbuf.h 123929 2003-12-28 08:57:09Z silby $
35 */
36
37#ifndef _SYS_MBUF_H_
38#define	_SYS_MBUF_H_
39
40#include <sys/queue.h>
41
42/*
43 * Mbufs are of a single size, MSIZE (sys/param.h), which
44 * includes overhead.  An mbuf may add a single "mbuf cluster" of size
45 * MCLBYTES (also in sys/param.h), which has no additional overhead
46 * and is used instead of the internal data area; this is done when
47 * at least MINCLSIZE of data must be stored.  Additionally, it is possible
48 * to allocate a separate buffer externally and attach it to the mbuf in
49 * a way similar to that of mbuf clusters.
50 */
51#define	MLEN		(MSIZE - sizeof(struct m_hdr))	/* normal data len */
52#define	MHLEN		(MLEN - sizeof(struct pkthdr))	/* data len w/pkthdr */
53#define	MINCLSIZE	(MHLEN + 1)	/* smallest amount to put in cluster */
54#define	M_MAXCOMPRESS	(MHLEN / 2)	/* max amount to copy for compression */
55
56#ifdef _KERNEL
57/*-
58 * Macros for type conversion:
59 * mtod(m, t)	-- Convert mbuf pointer to data pointer of correct type.
60 * dtom(x)	-- Convert data pointer within mbuf to mbuf pointer (XXX).
61 */
62#define	mtod(m, t)	((t)((m)->m_data))
63#define	dtom(x)		((struct mbuf *)((intptr_t)(x) & ~(MSIZE-1)))
64#endif /* _KERNEL */
65
66/*
67 * Header present at the beginning of every mbuf.
68 */
69struct m_hdr {
70	struct	mbuf *mh_next;		/* next buffer in chain */
71	struct	mbuf *mh_nextpkt;	/* next chain in queue/record */
72	caddr_t	mh_data;		/* location of data */
73	int	mh_len;			/* amount of data in this mbuf */
74	int	mh_flags;		/* flags; see below */
75	short	mh_type;		/* type of data in this mbuf */
76};
77
78/*
79 * Packet tag structure (see below for details).
80 */
81struct m_tag {
82	SLIST_ENTRY(m_tag)	m_tag_link;	/* List of packet tags */
83	u_int16_t		m_tag_id;	/* Tag ID */
84	u_int16_t		m_tag_len;	/* Length of data */
85	u_int32_t		m_tag_cookie;	/* ABI/Module ID */
86};
87
88/*
89 * Record/packet header in first mbuf of chain; valid only if M_PKTHDR is set.
90 */
91struct pkthdr {
92	struct	ifnet *rcvif;		/* rcv interface */
93	int	len;			/* total packet length */
94	/* variables for ip and tcp reassembly */
95	void	*header;		/* pointer to packet header */
96	/* variables for hardware checksum */
97	int	csum_flags;		/* flags regarding checksum */
98	int	csum_data;		/* data field used by csum routines */
99	SLIST_HEAD(packet_tags, m_tag) tags; /* list of packet tags */
100};
101
102/*
103 * Description of external storage mapped into mbuf; valid only if M_EXT is set.
104 */
105struct m_ext {
106	caddr_t	ext_buf;		/* start of buffer */
107	void	(*ext_free)		/* free routine if not the usual */
108		    (void *, void *);
109	void	*ext_args;		/* optional argument pointer */
110	u_int	ext_size;		/* size of buffer, for ext_free */
111	u_int	*ref_cnt;		/* pointer to ref count info */
112	int	ext_type;		/* type of external storage */
113};
114
115/*
116 * The core of the mbuf object along with some shortcut defines for
117 * practical purposes.
118 */
119struct mbuf {
120	struct	m_hdr m_hdr;
121	union {
122		struct {
123			struct	pkthdr MH_pkthdr;	/* M_PKTHDR set */
124			union {
125				struct	m_ext MH_ext;	/* M_EXT set */
126				char	MH_databuf[MHLEN];
127			} MH_dat;
128		} MH;
129		char	M_databuf[MLEN];		/* !M_PKTHDR, !M_EXT */
130	} M_dat;
131};
132#define	m_next		m_hdr.mh_next
133#define	m_len		m_hdr.mh_len
134#define	m_data		m_hdr.mh_data
135#define	m_type		m_hdr.mh_type
136#define	m_flags		m_hdr.mh_flags
137#define	m_nextpkt	m_hdr.mh_nextpkt
138#define	m_act		m_nextpkt
139#define	m_pkthdr	M_dat.MH.MH_pkthdr
140#define	m_ext		M_dat.MH.MH_dat.MH_ext
141#define	m_pktdat	M_dat.MH.MH_dat.MH_databuf
142#define	m_dat		M_dat.M_databuf
143
144/*
145 * mbuf flags.
146 */
147#define	M_EXT		0x0001	/* has associated external storage */
148#define	M_PKTHDR	0x0002	/* start of record */
149#define	M_EOR		0x0004	/* end of record */
150#define	M_RDONLY	0x0008	/* associated data is marked read-only */
151#define	M_PROTO1	0x0010	/* protocol-specific */
152#define	M_PROTO2	0x0020	/* protocol-specific */
153#define	M_PROTO3	0x0040	/* protocol-specific */
154#define	M_PROTO4	0x0080	/* protocol-specific */
155#define	M_PROTO5	0x0100	/* protocol-specific */
156#define M_PROTO6	0x4000	/* protocol-specific (avoid M_BCAST conflict) */
157#define	M_FREELIST	0x8000	/* mbuf is on the free list */
158
159/*
160 * mbuf pkthdr flags (also stored in m_flags).
161 */
162#define	M_BCAST		0x0200	/* send/received as link-level broadcast */
163#define	M_MCAST		0x0400	/* send/received as link-level multicast */
164#define	M_FRAG		0x0800	/* packet is a fragment of a larger packet */
165#define	M_FIRSTFRAG	0x1000	/* packet is first fragment */
166#define	M_LASTFRAG	0x2000	/* packet is last fragment */
167
168/*
169 * External buffer types: identify ext_buf type.
170 */
171#define	EXT_CLUSTER	1	/* mbuf cluster */
172#define	EXT_SFBUF	2	/* sendfile(2)'s sf_bufs */
173#define	EXT_NET_DRV	100	/* custom ext_buf provided by net driver(s) */
174#define	EXT_MOD_TYPE	200	/* custom module's ext_buf type */
175#define	EXT_DISPOSABLE	300	/* can throw this buffer away w/page flipping */
176#define	EXT_EXTREF	400	/* has externally maintained ref_cnt ptr*/
177
178/*
179 * Flags copied when copying m_pkthdr.
180 */
181#define	M_COPYFLAGS	(M_PKTHDR|M_EOR|M_RDONLY|M_PROTO1|M_PROTO1|M_PROTO2|\
182			    M_PROTO3|M_PROTO4|M_PROTO5|M_PROTO6|\
183			    M_BCAST|M_MCAST|M_FRAG|M_FIRSTFRAG|M_LASTFRAG)
184
185/*
186 * Flags indicating hw checksum support and sw checksum requirements.
187 */
188#define	CSUM_IP			0x0001		/* will csum IP */
189#define	CSUM_TCP		0x0002		/* will csum TCP */
190#define	CSUM_UDP		0x0004		/* will csum UDP */
191#define	CSUM_IP_FRAGS		0x0008		/* will csum IP fragments */
192#define	CSUM_FRAGMENT		0x0010		/* will do IP fragmentation */
193
194#define	CSUM_IP_CHECKED		0x0100		/* did csum IP */
195#define	CSUM_IP_VALID		0x0200		/*   ... the csum is valid */
196#define	CSUM_DATA_VALID		0x0400		/* csum_data field is valid */
197#define	CSUM_PSEUDO_HDR		0x0800		/* csum_data has pseudo hdr */
198
199#define	CSUM_DELAY_DATA		(CSUM_TCP | CSUM_UDP)
200#define	CSUM_DELAY_IP		(CSUM_IP)	/* XXX add ipv6 here too? */
201
202/*
203 * mbuf types.
204 */
205#define	MT_NOTMBUF	0	/* USED INTERNALLY ONLY! Object is not mbuf */
206#define	MT_DATA		1	/* dynamic (data) allocation */
207#define	MT_HEADER	2	/* packet header */
208#if 0
209#define	MT_SOCKET	3	/* socket structure */
210#define	MT_PCB		4	/* protocol control block */
211#define	MT_RTABLE	5	/* routing tables */
212#define	MT_HTABLE	6	/* IMP host tables */
213#define	MT_ATABLE	7	/* address resolution tables */
214#endif
215#define	MT_SONAME	8	/* socket name */
216#if 0
217#define	MT_SOOPTS	10	/* socket options */
218#endif
219#define	MT_FTABLE	11	/* fragment reassembly header */
220#if 0
221#define	MT_RIGHTS	12	/* access rights */
222#define	MT_IFADDR	13	/* interface address */
223#endif
224#define	MT_TAG		13	/* volatile metadata associated to pkts */
225#define	MT_CONTROL	14	/* extra-data protocol message */
226#define	MT_OOBDATA	15	/* expedited data  */
227#define	MT_NTYPES	16	/* number of mbuf types for mbtypes[] */
228
229/*
230 * Mbuf and cluster allocation statistics PCPU structure.
231 */
232struct mbpstat {
233	u_long	mb_mbfree;
234	u_long	mb_mbbucks;
235	u_long	mb_clfree;
236	u_long	mb_clbucks;
237	long	mb_mbtypes[MT_NTYPES];
238	short	mb_active;
239};
240
241/*
242 * General mbuf allocator statistics structure.
243 * XXX: Modifications of these are not protected by any mutex locks nor by
244 * any atomic() manipulations.  As a result, we may occasionally lose
245 * a count or two.  Luckily, not all of these fields are modified at all
246 * and remain static, and those that are manipulated are only manipulated
247 * in failure situations, which do not occur (hopefully) very often.
248 */
249struct mbstat {
250	u_long	m_drops;	/* times failed to allocate */
251	u_long	m_wait;		/* times succesfully returned from wait */
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	u_int	m_mbperbuck;	/* number of mbufs per "bucket" */
261	u_int	m_clperbuck;	/* number of clusters per "bucket" */
262	/* Number of mbtypes (gives # elems in mbpstat's mb_mbtypes[] array: */
263	short	m_numtypes;
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 * M_DONTWAIT means "don't block if nothing is available" whereas
273 * M_TRYWAIT means "block for mbuf_wait ticks at most if nothing is
274 * available."
275 */
276#define	M_DONTWAIT	0x4		/* don't conflict with M_NOWAIT */
277#define	M_TRYWAIT	0x8		/* or M_WAITOK */
278#define	M_WAIT		M_TRYWAIT	/* XXX: deprecated */
279#define	MBTOM(how)	((how) & M_TRYWAIT ? M_WAITOK : M_NOWAIT)
280
281#ifdef _KERNEL
282/*-
283 * mbuf external reference count management macros.
284 *
285 * MEXT_IS_REF(m): true if (m) is not the only mbuf referencing
286 *     the external buffer ext_buf.
287 *
288 * MEXT_REM_REF(m): remove reference to m_ext object.
289 *
290 * MEXT_ADD_REF(m): add reference to m_ext object already
291 *     referred to by (m).
292 */
293#define	MEXT_IS_REF(m)	(*((m)->m_ext.ref_cnt) > 1)
294
295#define	MEXT_REM_REF(m) do {						\
296	KASSERT(*((m)->m_ext.ref_cnt) > 0, ("m_ext refcnt < 0"));	\
297	atomic_subtract_int((m)->m_ext.ref_cnt, 1);			\
298} while(0)
299
300#define	MEXT_ADD_REF(m)	atomic_add_int((m)->m_ext.ref_cnt, 1)
301
302/*
303 * mbuf, cluster, and external object allocation macros
304 * (for compatibility purposes).
305 */
306/* NB: M_COPY_PKTHDR is deprecated.  Use M_MOVE_PKTHDR or m_dup_pktdr. */
307#define	M_MOVE_PKTHDR(to, from)	m_move_pkthdr((to), (from))
308#define	m_getclr(how, type)	m_get_clrd((how), (type))
309#define	MGET(m, how, type)	((m) = m_get((how), (type)))
310#define	MGETHDR(m, how, type)	((m) = m_gethdr((how), (type)))
311#define	MCLGET(m, how)		m_clget((m), (how))
312#define	MEXTADD(m, buf, size, free, args, flags, type) 			\
313    m_extadd((m), (caddr_t)(buf), (size), (free), (args), (flags), (type))
314
315/*
316 * MEXTFREE(m): disassociate (and possibly free) an external object from (m).
317 *
318 * If the atomic_cmpset_int() returns 0, then we effectively do nothing
319 * in terms of "cleaning up" (freeing the ext buf and ref. counter) as
320 * this means that either there are still references, or another thread
321 * is taking care of the clean-up.
322 */
323#define	MEXTFREE(m) do {						\
324	struct mbuf *_mb = (m);						\
325									\
326	MEXT_REM_REF(_mb);						\
327	if (atomic_cmpset_int(_mb->m_ext.ref_cnt, 0, 1))		\
328		_mext_free(_mb);					\
329	_mb->m_flags &= ~M_EXT;						\
330} while (0)
331
332/*
333 * Evaluate TRUE if it's safe to write to the mbuf m's data region (this
334 * can be both the local data payload, or an external buffer area,
335 * depending on whether M_EXT is set).
336 */
337#define	M_WRITABLE(m)	(!((m)->m_flags & M_RDONLY) && (!((m)->m_flags  \
338			    & M_EXT) || !MEXT_IS_REF(m)))
339
340/*
341 * Check if the supplied mbuf has a packet header, or else panic.
342 */
343#define M_ASSERTPKTHDR(m)				\
344	KASSERT(m != NULL && m->m_flags & M_PKTHDR,	\
345		("%s: no mbuf packet header!", __func__))
346
347/*
348 * Ensure that the supplied mbuf is a valid, non-free mbuf.
349 */
350#define M_ASSERTVALID(m)					\
351	KASSERT((((struct mbuf *)m)->m_flags & M_FREELIST) == 0,			\
352		("%s: attempted use of a free mbuf!", __func__))
353
354/*
355 * Set the m_data pointer of a newly-allocated mbuf (m_get/MGET) to place
356 * an object of the specified size at the end of the mbuf, longword aligned.
357 */
358#define	M_ALIGN(m, len) do {						\
359	(m)->m_data += (MLEN - (len)) & ~(sizeof(long) - 1);		\
360} while (0)
361
362/*
363 * As above, for mbufs allocated with m_gethdr/MGETHDR
364 * or initialized by M_COPY_PKTHDR.
365 */
366#define	MH_ALIGN(m, len) do {						\
367	(m)->m_data += (MHLEN - (len)) & ~(sizeof(long) - 1);		\
368} while (0)
369
370/*
371 * Compute the amount of space available
372 * before the current start of data in an mbuf.
373 *
374 * The M_WRITABLE() is a temporary, conservative safety measure: the burden
375 * of checking writability of the mbuf data area rests solely with the caller.
376 */
377#define	M_LEADINGSPACE(m)						\
378	((m)->m_flags & M_EXT ?						\
379	    (M_WRITABLE(m) ? (m)->m_data - (m)->m_ext.ext_buf : 0):	\
380	    (m)->m_flags & M_PKTHDR ? (m)->m_data - (m)->m_pktdat :	\
381	    (m)->m_data - (m)->m_dat)
382
383/*
384 * Compute the amount of space available
385 * after the end of data in an mbuf.
386 *
387 * The M_WRITABLE() is a temporary, conservative safety measure: the burden
388 * of checking writability of the mbuf data area rests solely with the caller.
389 */
390#define	M_TRAILINGSPACE(m)						\
391	((m)->m_flags & M_EXT ?						\
392	    (M_WRITABLE(m) ? (m)->m_ext.ext_buf + (m)->m_ext.ext_size	\
393		- ((m)->m_data + (m)->m_len) : 0) :			\
394	    &(m)->m_dat[MLEN] - ((m)->m_data + (m)->m_len))
395
396/*
397 * Arrange to prepend space of size plen to mbuf m.
398 * If a new mbuf must be allocated, how specifies whether to wait.
399 * If the allocation fails, the original mbuf chain is freed and m is
400 * set to NULL.
401 */
402#define	M_PREPEND(m, plen, how) do {					\
403	struct mbuf **_mmp = &(m);					\
404	struct mbuf *_mm = *_mmp;					\
405	int _mplen = (plen);						\
406	int __mhow = (how);						\
407									\
408	if (M_LEADINGSPACE(_mm) >= _mplen) {				\
409		_mm->m_data -= _mplen;					\
410		_mm->m_len += _mplen;					\
411	} else								\
412		_mm = m_prepend(_mm, _mplen, __mhow);			\
413	if (_mm != NULL && _mm->m_flags & M_PKTHDR)			\
414		_mm->m_pkthdr.len += _mplen;				\
415	*_mmp = _mm;							\
416} while (0)
417
418/*
419 * Change mbuf to new type.
420 * This is a relatively expensive operation and should be avoided.
421 */
422#define	MCHTYPE(m, t)	m_chtype((m), (t))
423
424/* Length to m_copy to copy all. */
425#define	M_COPYALL	1000000000
426
427/* Compatibility with 4.3. */
428#define	m_copy(m, o, l)	m_copym((m), (o), (l), M_DONTWAIT)
429
430extern	int max_datalen;		/* MHLEN - max_hdr */
431extern	int max_hdr;			/* Largest link + protocol header */
432extern	int max_linkhdr;		/* Largest link-level header */
433extern	int max_protohdr;		/* Largest protocol header */
434extern	struct mbstat mbstat;		/* General mbuf stats/infos */
435extern	int nmbclusters;		/* Maximum number of clusters */
436extern	int nmbcnt;			/* Scale kmem_map for counter space */
437extern	int nmbufs;			/* Maximum number of mbufs */
438extern	int nsfbufs;			/* Number of sendfile(2) bufs alloced */
439extern	int nsfbufspeak;		/* Peak of nsfbufsused */
440extern	int nsfbufsused;		/* Number of sendfile(2) bufs in use */
441
442void		 _mext_free(struct mbuf *);
443void		 m_adj(struct mbuf *, int);
444int		 m_apply(struct mbuf *, int, int,
445		 int (*)(void *, void *, unsigned int), void *);
446void		 m_cat(struct mbuf *, struct mbuf *);
447void		 m_chtype(struct mbuf *, short);
448void		 m_clget(struct mbuf *, int);
449void		 m_extadd(struct mbuf *, caddr_t, u_int,
450		    void (*)(void *, void *), void *, int, int);
451void		 m_copyback(struct mbuf *, int, int, caddr_t);
452void		 m_copydata(const struct mbuf *, int, int, caddr_t);
453struct	mbuf	*m_copym(struct mbuf *, int, int, int);
454struct	mbuf	*m_copypacket(struct mbuf *, int);
455void		 m_copy_pkthdr(struct mbuf *, struct mbuf *);
456struct	mbuf	*m_defrag(struct mbuf *, int);
457struct	mbuf	*m_devget(char *, int, int, struct ifnet *,
458		    void (*)(char *, caddr_t, u_int));
459struct	mbuf	*m_dup(struct mbuf *, int);
460int		 m_dup_pkthdr(struct mbuf *, struct mbuf *, int);
461u_int		 m_fixhdr(struct mbuf *);
462struct	mbuf	*m_fragment(struct mbuf *, int, int);
463struct	mbuf	*m_free(struct mbuf *);
464void		 m_freem(struct mbuf *);
465struct	mbuf	*m_get(int, short);
466struct	mbuf	*m_get_clrd(int, short);
467struct	mbuf	*m_getcl(int, short, int);
468struct	mbuf	*m_gethdr(int, short);
469struct	mbuf	*m_gethdr_clrd(int, short);
470struct	mbuf	*m_getm(struct mbuf *, int, int, short);
471struct	mbuf	*m_getptr(struct mbuf *, int, int *);
472u_int		 m_length(struct mbuf *, struct mbuf **);
473void		 m_move_pkthdr(struct mbuf *, struct mbuf *);
474struct	mbuf	*m_prepend(struct mbuf *, int, int);
475void		 m_print(const struct mbuf *);
476struct	mbuf	*m_pulldown(struct mbuf *, int, int, int *);
477struct	mbuf	*m_pullup(struct mbuf *, int);
478struct	mbuf	*m_split(struct mbuf *, int, int);
479
480/*
481 * Packets may have annotations attached by affixing a list
482 * of "packet tags" to the pkthdr structure.  Packet tags are
483 * dynamically allocated semi-opaque data structures that have
484 * a fixed header (struct m_tag) that specifies the size of the
485 * memory block and a <cookie,type> pair that identifies it.
486 * The cookie is a 32-bit unique unsigned value used to identify
487 * a module or ABI.  By convention this value is chose as the
488 * date+time that the module is created, expressed as the number of
489 * seconds since the epoch (e.g., using date -u +'%s').  The type value
490 * is an ABI/module-specific value that identifies a particular annotation
491 * and is private to the module.  For compatibility with systems
492 * like OpenBSD that define packet tags w/o an ABI/module cookie,
493 * the value PACKET_ABI_COMPAT is used to implement m_tag_get and
494 * m_tag_find compatibility shim functions and several tag types are
495 * defined below.  Users that do not require compatibility should use
496 * a private cookie value so that packet tag-related definitions
497 * can be maintained privately.
498 *
499 * Note that the packet tag returned by m_tag_allocate has the default
500 * memory alignment implemented by malloc.  To reference private data
501 * one can use a construct like:
502 *
503 *	struct m_tag *mtag = m_tag_allocate(...);
504 *	struct foo *p = (struct foo *)(mtag+1);
505 *
506 * if the alignment of struct m_tag is sufficient for referencing members
507 * of struct foo.  Otherwise it is necessary to embed struct m_tag within
508 * the private data structure to insure proper alignment; e.g.,
509 *
510 *	struct foo {
511 *		struct m_tag	tag;
512 *		...
513 *	};
514 *	struct foo *p = (struct foo *) m_tag_allocate(...);
515 *	struct m_tag *mtag = &p->tag;
516 */
517
518/*
519 * Persistent tags stay with an mbuf until the mbuf is reclaimed.
520 * Otherwise tags are expected to ``vanish'' when they pass through
521 * a network interface.  For most interfaces this happens normally
522 * as the tags are reclaimed when the mbuf is free'd.  However in
523 * some special cases reclaiming must be done manually.  An example
524 * is packets that pass through the loopback interface.  Also, one
525 * must be careful to do this when ``turning around'' packets (e.g.,
526 * icmp_reflect).
527 *
528 * To mark a tag persistent bit-or this flag in when defining the
529 * tag id.  The tag will then be treated as described above.
530 */
531#define	MTAG_PERSISTENT				0x800
532
533#define	PACKET_TAG_NONE				0  /* Nadda */
534
535/* Packet tag for use with PACKET_ABI_COMPAT. */
536#define	PACKET_TAG_IPSEC_IN_DONE		1  /* IPsec applied, in */
537#define	PACKET_TAG_IPSEC_OUT_DONE		2  /* IPsec applied, out */
538#define	PACKET_TAG_IPSEC_IN_CRYPTO_DONE		3  /* NIC IPsec crypto done */
539#define	PACKET_TAG_IPSEC_OUT_CRYPTO_NEEDED	4  /* NIC IPsec crypto req'ed */
540#define	PACKET_TAG_IPSEC_IN_COULD_DO_CRYPTO	5  /* NIC notifies IPsec */
541#define	PACKET_TAG_IPSEC_PENDING_TDB		6  /* Reminder to do IPsec */
542#define	PACKET_TAG_BRIDGE			7  /* Bridge processing done */
543#define	PACKET_TAG_GIF				8  /* GIF processing done */
544#define	PACKET_TAG_GRE				9  /* GRE processing done */
545#define	PACKET_TAG_IN_PACKET_CHECKSUM		10 /* NIC checksumming done */
546#define	PACKET_TAG_ENCAP			11 /* Encap.  processing */
547#define	PACKET_TAG_IPSEC_SOCKET			12 /* IPSEC socket ref */
548#define	PACKET_TAG_IPSEC_HISTORY		13 /* IPSEC history */
549#define	PACKET_TAG_IPV6_INPUT			14 /* IPV6 input processing */
550
551/*
552 * As a temporary and low impact solution to replace the even uglier
553 * approach used so far in some parts of the network stack (which relies
554 * on global variables), packet tag-like annotations are stored in MT_TAG
555 * mbufs (or lookalikes) prepended to the actual mbuf chain.
556 *
557 *	m_type	= MT_TAG
558 *	m_flags = m_tag_id
559 *	m_next	= next buffer in chain.
560 *
561 * BE VERY CAREFUL not to pass these blocks to the mbuf handling routines.
562 */
563#define	_m_tag_id	m_hdr.mh_flags
564
565/* Packet tags used in the FreeBSD network stack. */
566#define	PACKET_TAG_DUMMYNET			15 /* dummynet info */
567#define	PACKET_TAG_IPFW				16 /* ipfw classification */
568#define	PACKET_TAG_DIVERT			17 /* divert info */
569#define	PACKET_TAG_IPFORWARD			18 /* ipforward info */
570#define	PACKET_TAG_MACLABEL	(19 | MTAG_PERSISTENT) /* MAC label */
571#define	PACKET_TAG_IPFASTFWD_OURS		20 /* IP fastforward dropback */
572
573/* Packet tag routines. */
574struct	m_tag 	*m_tag_alloc(u_int32_t, int, int, int);
575void		 m_tag_free(struct m_tag *);
576void		 m_tag_prepend(struct mbuf *, struct m_tag *);
577void		 m_tag_unlink(struct mbuf *, struct m_tag *);
578void		 m_tag_delete(struct mbuf *, struct m_tag *);
579void		 m_tag_delete_chain(struct mbuf *, struct m_tag *);
580struct	m_tag	*m_tag_locate(struct mbuf *, u_int32_t, int, struct m_tag *);
581struct	m_tag	*m_tag_copy(struct m_tag *, int);
582int		 m_tag_copy_chain(struct mbuf *, struct mbuf *, int);
583void		 m_tag_init(struct mbuf *);
584struct	m_tag	*m_tag_first(struct mbuf *);
585struct	m_tag	*m_tag_next(struct mbuf *, struct m_tag *);
586void		 m_tag_delete_nonpersistent(struct mbuf *);
587
588/* These are for OpenBSD compatibility. */
589#define	MTAG_ABI_COMPAT		0		/* compatibility ABI */
590
591static __inline struct m_tag *
592m_tag_get(int type, int length, int wait)
593{
594	return m_tag_alloc(MTAG_ABI_COMPAT, type, length, wait);
595}
596
597static __inline struct m_tag *
598m_tag_find(struct mbuf *m, int type, struct m_tag *start)
599{
600	return m_tag_locate(m, MTAG_ABI_COMPAT, type, start);
601}
602#endif /* _KERNEL */
603
604#endif /* !_SYS_MBUF_H_ */
605