mbuf.h revision 129906
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 129906 2004-05-31 21:46:06Z bmilekic $
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#endif
42
43/*
44 * Mbufs are of a single size, MSIZE (sys/param.h), which
45 * includes overhead.  An mbuf may add a single "mbuf cluster" of size
46 * MCLBYTES (also in sys/param.h), which has no additional overhead
47 * and is used instead of the internal data area; this is done when
48 * at least MINCLSIZE of data must be stored.  Additionally, it is possible
49 * to allocate a separate buffer externally and attach it to the mbuf in
50 * a way similar to that of mbuf clusters.
51 */
52#define	MLEN		(MSIZE - sizeof(struct m_hdr))	/* normal data len */
53#define	MHLEN		(MLEN - sizeof(struct pkthdr))	/* data len w/pkthdr */
54#define	MINCLSIZE	(MHLEN + 1)	/* smallest amount to put in cluster */
55#define	M_MAXCOMPRESS	(MHLEN / 2)	/* max amount to copy for compression */
56
57#ifdef _KERNEL
58/*-
59 * Macros for type conversion:
60 * mtod(m, t)	-- Convert mbuf pointer to data pointer of correct type.
61 * dtom(x)	-- Convert data pointer within mbuf to mbuf pointer (XXX).
62 */
63#define	mtod(m, t)	((t)((m)->m_data))
64#define	dtom(x)		((struct mbuf *)((intptr_t)(x) & ~(MSIZE-1)))
65
66/*
67 * Argument structure passed to UMA routines during mbuf and packet
68 * allocations.
69 */
70struct mb_args {
71	int	flags;	/* Flags for mbuf being allocated */
72	int	how;	/* How to allocate: M_WAITOK or M_DONTWAIT */
73	short	type;	/* Type of mbuf being allocated */
74};
75#endif /* _KERNEL */
76
77/*
78 * Header present at the beginning of every mbuf.
79 */
80struct m_hdr {
81	struct	mbuf *mh_next;		/* next buffer in chain */
82	struct	mbuf *mh_nextpkt;	/* next chain in queue/record */
83	caddr_t	mh_data;		/* location of data */
84	int	mh_len;			/* amount of data in this mbuf */
85	int	mh_flags;		/* flags; see below */
86	short	mh_type;		/* type of data in this mbuf */
87};
88
89/*
90 * Packet tag structure (see below for details).
91 */
92struct m_tag {
93	SLIST_ENTRY(m_tag)	m_tag_link;	/* List of packet tags */
94	u_int16_t		m_tag_id;	/* Tag ID */
95	u_int16_t		m_tag_len;	/* Length of data */
96	u_int32_t		m_tag_cookie;	/* ABI/Module ID */
97	void			(*m_tag_free)(struct m_tag *);
98};
99
100/*
101 * Record/packet header in first mbuf of chain; valid only if M_PKTHDR is set.
102 */
103struct pkthdr {
104	struct	ifnet *rcvif;		/* rcv interface */
105	int	len;			/* total packet length */
106	/* variables for ip and tcp reassembly */
107	void	*header;		/* pointer to packet header */
108	/* variables for hardware checksum */
109	int	csum_flags;		/* flags regarding checksum */
110	int	csum_data;		/* data field used by csum routines */
111	SLIST_HEAD(packet_tags, m_tag) tags; /* list of packet tags */
112};
113
114/*
115 * Description of external storage mapped into mbuf; valid only if M_EXT is set.
116 */
117struct m_ext {
118	caddr_t	ext_buf;		/* start of buffer */
119	void	(*ext_free)		/* free routine if not the usual */
120		    (void *, void *);
121	void	*ext_args;		/* optional argument pointer */
122	u_int	ext_size;		/* size of buffer, for ext_free */
123	u_int	*ref_cnt;		/* pointer to ref count info */
124	int	ext_type;		/* type of external storage */
125};
126
127/*
128 * The core of the mbuf object along with some shortcut defines for
129 * practical purposes.
130 */
131struct mbuf {
132	struct	m_hdr m_hdr;
133	union {
134		struct {
135			struct	pkthdr MH_pkthdr;	/* M_PKTHDR set */
136			union {
137				struct	m_ext MH_ext;	/* M_EXT set */
138				char	MH_databuf[MHLEN];
139			} MH_dat;
140		} MH;
141		char	M_databuf[MLEN];		/* !M_PKTHDR, !M_EXT */
142	} M_dat;
143};
144#define	m_next		m_hdr.mh_next
145#define	m_len		m_hdr.mh_len
146#define	m_data		m_hdr.mh_data
147#define	m_type		m_hdr.mh_type
148#define	m_flags		m_hdr.mh_flags
149#define	m_nextpkt	m_hdr.mh_nextpkt
150#define	m_act		m_nextpkt
151#define	m_pkthdr	M_dat.MH.MH_pkthdr
152#define	m_ext		M_dat.MH.MH_dat.MH_ext
153#define	m_pktdat	M_dat.MH.MH_dat.MH_databuf
154#define	m_dat		M_dat.M_databuf
155
156/*
157 * mbuf flags.
158 */
159#define	M_EXT		0x0001	/* has associated external storage */
160#define	M_PKTHDR	0x0002	/* start of record */
161#define	M_EOR		0x0004	/* end of record */
162#define	M_RDONLY	0x0008	/* associated data is marked read-only */
163#define	M_PROTO1	0x0010	/* protocol-specific */
164#define	M_PROTO2	0x0020	/* protocol-specific */
165#define	M_PROTO3	0x0040	/* protocol-specific */
166#define	M_PROTO4	0x0080	/* protocol-specific */
167#define	M_PROTO5	0x0100	/* protocol-specific */
168#define	M_PROTO6	0x4000	/* protocol-specific (avoid M_BCAST conflict) */
169#define	M_FREELIST	0x8000	/* mbuf is on the free list */
170
171/*
172 * mbuf pkthdr flags (also stored in m_flags).
173 */
174#define	M_BCAST		0x0200	/* send/received as link-level broadcast */
175#define	M_MCAST		0x0400	/* send/received as link-level multicast */
176#define	M_FRAG		0x0800	/* packet is a fragment of a larger packet */
177#define	M_FIRSTFRAG	0x1000	/* packet is first fragment */
178#define	M_LASTFRAG	0x2000	/* packet is last fragment */
179
180/*
181 * External buffer types: identify ext_buf type.
182 */
183#define	EXT_CLUSTER	1	/* mbuf cluster */
184#define	EXT_SFBUF	2	/* sendfile(2)'s sf_bufs */
185#define	EXT_PACKET	3	/* came out of Packet zone */
186#define	EXT_NET_DRV	100	/* custom ext_buf provided by net driver(s) */
187#define	EXT_MOD_TYPE	200	/* custom module's ext_buf type */
188#define	EXT_DISPOSABLE	300	/* can throw this buffer away w/page flipping */
189#define	EXT_EXTREF	400	/* has externally maintained ref_cnt ptr */
190
191/*
192 * Flags copied when copying m_pkthdr.
193 */
194#define	M_COPYFLAGS	(M_PKTHDR|M_EOR|M_RDONLY|M_PROTO1|M_PROTO1|M_PROTO2|\
195			    M_PROTO3|M_PROTO4|M_PROTO5|M_PROTO6|\
196			    M_BCAST|M_MCAST|M_FRAG|M_FIRSTFRAG|M_LASTFRAG)
197
198/*
199 * Flags indicating hw checksum support and sw checksum requirements.
200 */
201#define	CSUM_IP			0x0001		/* will csum IP */
202#define	CSUM_TCP		0x0002		/* will csum TCP */
203#define	CSUM_UDP		0x0004		/* will csum UDP */
204#define	CSUM_IP_FRAGS		0x0008		/* will csum IP fragments */
205#define	CSUM_FRAGMENT		0x0010		/* will do IP fragmentation */
206
207#define	CSUM_IP_CHECKED		0x0100		/* did csum IP */
208#define	CSUM_IP_VALID		0x0200		/*   ... the csum is valid */
209#define	CSUM_DATA_VALID		0x0400		/* csum_data field is valid */
210#define	CSUM_PSEUDO_HDR		0x0800		/* csum_data has pseudo hdr */
211
212#define	CSUM_DELAY_DATA		(CSUM_TCP | CSUM_UDP)
213#define	CSUM_DELAY_IP		(CSUM_IP)	/* XXX add ipv6 here too? */
214
215/*
216 * mbuf types.
217 */
218#define	MT_NOTMBUF	0	/* USED INTERNALLY ONLY! Object is not mbuf */
219#define	MT_DATA		1	/* dynamic (data) allocation */
220#define	MT_HEADER	2	/* packet header */
221#if 0
222#define	MT_SOCKET	3	/* socket structure */
223#define	MT_PCB		4	/* protocol control block */
224#define	MT_RTABLE	5	/* routing tables */
225#define	MT_HTABLE	6	/* IMP host tables */
226#define	MT_ATABLE	7	/* address resolution tables */
227#endif
228#define	MT_SONAME	8	/* socket name */
229#if 0
230#define	MT_SOOPTS	10	/* socket options */
231#endif
232#define	MT_FTABLE	11	/* fragment reassembly header */
233#if 0
234#define	MT_RIGHTS	12	/* access rights */
235#define	MT_IFADDR	13	/* interface address */
236#endif
237#define	MT_CONTROL	14	/* extra-data protocol message */
238#define	MT_OOBDATA	15	/* expedited data  */
239#define	MT_NTYPES	16	/* number of mbuf types for mbtypes[] */
240
241/*
242 * General mbuf allocator statistics structure.
243 */
244struct mbstat {
245	u_long	m_mbufs;	/* XXX */
246	u_long	m_mclusts;	/* XXX */
247
248	u_long	m_drain;	/* times drained protocols for space */
249	u_long	m_mcfail;	/* XXX: times m_copym failed */
250	u_long	m_mpfail;	/* XXX: times m_pullup failed */
251	u_long	m_msize;	/* length of an mbuf */
252	u_long	m_mclbytes;	/* length of an mbuf cluster */
253	u_long	m_minclsize;	/* min length of data to allocate a cluster */
254	u_long	m_mlen;		/* length of data in an mbuf */
255	u_long	m_mhlen;	/* length of data in a header mbuf */
256
257	/* Number of mbtypes (gives # elems in mbtypes[] array: */
258	short	m_numtypes;
259
260	/* XXX: Sendfile stats should eventually move to their own struct */
261	u_long	sf_iocnt;	/* times sendfile had to do disk I/O */
262	u_long	sf_allocfail;	/* times sfbuf allocation failed */
263	u_long	sf_allocwait;	/* times sfbuf allocation had to wait */
264};
265
266/*
267 * Flags specifying how an allocation should be made.
268 *
269 * The flag to use is as follows:
270 * - M_DONTWAIT or M_NOWAIT from an interrupt handler to not block allocation.
271 * - M_WAIT or M_WAITOK or M_TRYWAIT from wherever it is safe to block.
272 *
273 * M_DONTWAIT/M_NOWAIT means that we will not block the thread explicitly
274 * and if we cannot allocate immediately we may return NULL,
275 * whereas M_WAIT/M_WAITOK/M_TRYWAIT means that if we cannot allocate
276 * resources we will block until they are available, and thus never
277 * return NULL.
278 *
279 * XXX Eventually just phase this out to use M_WAITOK/M_NOWAIT.
280 */
281#define	MBTOM(how)	(how)
282#define	M_DONTWAIT	M_NOWAIT
283#define	M_TRYWAIT	M_WAITOK
284#define	M_WAIT		M_WAITOK
285
286#ifdef _KERNEL
287/*-
288 * mbuf external reference count management macros.
289 *
290 * MEXT_IS_REF(m): true if (m) is not the only mbuf referencing
291 *     the external buffer ext_buf.
292 *
293 * MEXT_REM_REF(m): remove reference to m_ext object.
294 *
295 * MEXT_ADD_REF(m): add reference to m_ext object already
296 *     referred to by (m).
297 */
298#define	MEXT_IS_REF(m)	(*((m)->m_ext.ref_cnt) > 1)
299
300#define	MEXT_REM_REF(m) do {						\
301	KASSERT(*((m)->m_ext.ref_cnt) > 0, ("m_ext refcnt < 0"));	\
302	atomic_subtract_int((m)->m_ext.ref_cnt, 1);			\
303} while(0)
304
305#define	MEXT_ADD_REF(m)	atomic_add_int((m)->m_ext.ref_cnt, 1)
306
307/*
308 * Network buffer allocation API
309 *
310 * The rest of it is defined in kern/subr_mbuf.c
311 */
312
313extern uma_zone_t	zone_mbuf;
314extern uma_zone_t	zone_clust;
315extern uma_zone_t	zone_pack;
316
317static __inline struct mbuf	*m_get(int how, short type);
318static __inline struct mbuf	*m_gethdr(int how, short type);
319static __inline struct mbuf	*m_getcl(int how, short type, int flags);
320static __inline struct mbuf	*m_getclr(int how, short type);	/* XXX */
321static __inline struct mbuf	*m_free(struct mbuf *m);
322static __inline void		 m_clget(struct mbuf *m, int how);
323static __inline void		 m_chtype(struct mbuf *m, short new_type);
324void				 mb_free_ext(struct mbuf *);
325
326static __inline
327struct mbuf *
328m_get(int how, short type)
329{
330	struct mb_args args;
331
332	args.flags = 0;
333	args.how = how;
334	args.type = type;
335	return (uma_zalloc_arg(zone_mbuf, &args, how));
336}
337
338/* XXX This should be depracated, very little use */
339static __inline
340struct mbuf *
341m_getclr(int how, short type)
342{
343	struct mbuf *m;
344	struct mb_args args;
345
346	args.flags = 0;
347	args.how = how;
348	args.type = type;
349	m = uma_zalloc_arg(zone_mbuf, &args, how);
350	if (m != NULL)
351		bzero(m->m_data, MLEN);
352	return m;
353}
354
355static __inline
356struct mbuf *
357m_gethdr(int how, short type)
358{
359	struct mb_args args;
360
361	args.flags = M_PKTHDR;
362	args.how = how;
363	args.type = type;
364	return (uma_zalloc_arg(zone_mbuf, &args, how));
365}
366
367static __inline
368struct mbuf *
369m_getcl(int how, short type, int flags)
370{
371	struct mb_args args;
372
373	args.flags = flags;
374	args.how = how;
375	args.type = type;
376	return (uma_zalloc_arg(zone_pack, &args, how));
377}
378
379static __inline
380struct mbuf *
381m_free(struct mbuf *m)
382{
383	struct mbuf *n = m->m_next;
384
385#ifdef INVARIANTS
386	m->m_flags |= M_FREELIST;
387#endif
388	if (m->m_flags & M_EXT)
389		mb_free_ext(m);
390	else
391		uma_zfree(zone_mbuf, m);
392	return n;
393}
394
395static __inline
396void
397m_clget(struct mbuf *m, int how)
398{
399	m->m_ext.ext_buf = NULL;
400	uma_zalloc_arg(zone_clust, m, how);
401}
402
403static __inline
404void
405m_chtype(struct mbuf *m, short new_type)
406{
407	m->m_type = new_type;
408}
409
410/*
411 * mbuf, cluster, and external object allocation macros
412 * (for compatibility purposes).
413 */
414/* NB: M_COPY_PKTHDR is deprecated.  Use M_MOVE_PKTHDR or m_dup_pktdr. */
415#define	M_MOVE_PKTHDR(to, from)	m_move_pkthdr((to), (from))
416#define	MGET(m, how, type)	((m) = m_get((how), (type)))
417#define	MGETHDR(m, how, type)	((m) = m_gethdr((how), (type)))
418#define	MCLGET(m, how)		m_clget((m), (how))
419#define	MEXTADD(m, buf, size, free, args, flags, type) 			\
420    m_extadd((m), (caddr_t)(buf), (size), (free), (args), (flags), (type))
421
422/*
423 * Evaluate TRUE if it's safe to write to the mbuf m's data region (this
424 * can be both the local data payload, or an external buffer area,
425 * depending on whether M_EXT is set).
426 */
427#define	M_WRITABLE(m)	(!((m)->m_flags & M_RDONLY) && (!((m)->m_flags  \
428			    & M_EXT) || !MEXT_IS_REF(m)))
429
430/* Check if the supplied mbuf has a packet header, or else panic. */
431#define	M_ASSERTPKTHDR(m)						\
432	KASSERT(m != NULL && m->m_flags & M_PKTHDR,			\
433	    ("%s: no mbuf packet header!", __func__))
434
435/* Ensure that the supplied mbuf is a valid, non-free mbuf. */
436#define	M_ASSERTVALID(m)						\
437	KASSERT((((struct mbuf *)m)->m_flags & M_FREELIST) == 0,	\
438	    ("%s: attempted use of a free mbuf!", __func__))
439
440/*
441 * Set the m_data pointer of a newly-allocated mbuf (m_get/MGET) to place
442 * an object of the specified size at the end of the mbuf, longword aligned.
443 */
444#define	M_ALIGN(m, len) do {						\
445	(m)->m_data += (MLEN - (len)) & ~(sizeof(long) - 1);		\
446} while (0)
447
448/*
449 * As above, for mbufs allocated with m_gethdr/MGETHDR
450 * or initialized by M_COPY_PKTHDR.
451 */
452#define	MH_ALIGN(m, len) do {						\
453	(m)->m_data += (MHLEN - (len)) & ~(sizeof(long) - 1);		\
454} while (0)
455
456/*
457 * Compute the amount of space available
458 * before the current start of data in an mbuf.
459 *
460 * The M_WRITABLE() is a temporary, conservative safety measure: the burden
461 * of checking writability of the mbuf data area rests solely with the caller.
462 */
463#define	M_LEADINGSPACE(m)						\
464	((m)->m_flags & M_EXT ?						\
465	    (M_WRITABLE(m) ? (m)->m_data - (m)->m_ext.ext_buf : 0):	\
466	    (m)->m_flags & M_PKTHDR ? (m)->m_data - (m)->m_pktdat :	\
467	    (m)->m_data - (m)->m_dat)
468
469/*
470 * Compute the amount of space available
471 * after the end of data in an mbuf.
472 *
473 * The M_WRITABLE() is a temporary, conservative safety measure: the burden
474 * of checking writability of the mbuf data area rests solely with the caller.
475 */
476#define	M_TRAILINGSPACE(m)						\
477	((m)->m_flags & M_EXT ?						\
478	    (M_WRITABLE(m) ? (m)->m_ext.ext_buf + (m)->m_ext.ext_size	\
479		- ((m)->m_data + (m)->m_len) : 0) :			\
480	    &(m)->m_dat[MLEN] - ((m)->m_data + (m)->m_len))
481
482/*
483 * Arrange to prepend space of size plen to mbuf m.
484 * If a new mbuf must be allocated, how specifies whether to wait.
485 * If the allocation fails, the original mbuf chain is freed and m is
486 * set to NULL.
487 */
488#define	M_PREPEND(m, plen, how) do {					\
489	struct mbuf **_mmp = &(m);					\
490	struct mbuf *_mm = *_mmp;					\
491	int _mplen = (plen);						\
492	int __mhow = (how);						\
493									\
494	if (M_LEADINGSPACE(_mm) >= _mplen) {				\
495		_mm->m_data -= _mplen;					\
496		_mm->m_len += _mplen;					\
497	} else								\
498		_mm = m_prepend(_mm, _mplen, __mhow);			\
499	if (_mm != NULL && _mm->m_flags & M_PKTHDR)			\
500		_mm->m_pkthdr.len += _mplen;				\
501	*_mmp = _mm;							\
502} while (0)
503
504/*
505 * Change mbuf to new type.
506 * This is a relatively expensive operation and should be avoided.
507 */
508#define	MCHTYPE(m, t)	m_chtype((m), (t))
509
510/* Length to m_copy to copy all. */
511#define	M_COPYALL	1000000000
512
513/* Compatibility with 4.3. */
514#define	m_copy(m, o, l)	m_copym((m), (o), (l), M_DONTWAIT)
515
516extern	int max_datalen;		/* MHLEN - max_hdr */
517extern	int max_hdr;			/* Largest link + protocol header */
518extern	int max_linkhdr;		/* Largest link-level header */
519extern	int max_protohdr;		/* Largest protocol header */
520extern	struct mbstat mbstat;		/* General mbuf stats/infos */
521extern	int nmbclusters;		/* Maximum number of clusters */
522
523struct uio;
524
525void		 m_adj(struct mbuf *, int);
526int		 m_apply(struct mbuf *, int, int,
527		    int (*)(void *, void *, u_int), void *);
528void		 m_cat(struct mbuf *, struct mbuf *);
529void		 m_extadd(struct mbuf *, caddr_t, u_int,
530		    void (*)(void *, void *), void *, int, int);
531void		 m_copyback(struct mbuf *, int, int, c_caddr_t);
532void		 m_copydata(const struct mbuf *, int, int, caddr_t);
533struct	mbuf	*m_copym(struct mbuf *, int, int, int);
534struct	mbuf	*m_copypacket(struct mbuf *, int);
535void		 m_copy_pkthdr(struct mbuf *, struct mbuf *);
536struct	mbuf	*m_defrag(struct mbuf *, int);
537struct	mbuf	*m_devget(char *, int, int, struct ifnet *,
538		    void (*)(char *, caddr_t, u_int));
539struct	mbuf	*m_dup(struct mbuf *, int);
540int		 m_dup_pkthdr(struct mbuf *, struct mbuf *, int);
541u_int		 m_fixhdr(struct mbuf *);
542struct	mbuf	*m_fragment(struct mbuf *, int, int);
543void		 m_freem(struct mbuf *);
544struct	mbuf	*m_getm(struct mbuf *, int, int, short);
545struct	mbuf	*m_getptr(struct mbuf *, int, int *);
546u_int		 m_length(struct mbuf *, struct mbuf **);
547void		 m_move_pkthdr(struct mbuf *, struct mbuf *);
548struct	mbuf	*m_prepend(struct mbuf *, int, int);
549void		 m_print(const struct mbuf *);
550struct	mbuf	*m_pulldown(struct mbuf *, int, int, int *);
551struct	mbuf	*m_pullup(struct mbuf *, int);
552struct	mbuf	*m_split(struct mbuf *, int, int);
553struct	mbuf	*m_uiotombuf(struct uio *, int, int);
554
555/*-
556 * Network packets may have annotations attached by affixing a list
557 * of "packet tags" to the pkthdr structure.  Packet tags are
558 * dynamically allocated semi-opaque data structures that have
559 * a fixed header (struct m_tag) that specifies the size of the
560 * memory block and a <cookie,type> pair that identifies it.
561 * The cookie is a 32-bit unique unsigned value used to identify
562 * a module or ABI.  By convention this value is chose as the
563 * date+time that the module is created, expressed as the number of
564 * seconds since the epoch (e.g., using date -u +'%s').  The type value
565 * is an ABI/module-specific value that identifies a particular annotation
566 * and is private to the module.  For compatibility with systems
567 * like OpenBSD that define packet tags w/o an ABI/module cookie,
568 * the value PACKET_ABI_COMPAT is used to implement m_tag_get and
569 * m_tag_find compatibility shim functions and several tag types are
570 * defined below.  Users that do not require compatibility should use
571 * a private cookie value so that packet tag-related definitions
572 * can be maintained privately.
573 *
574 * Note that the packet tag returned by m_tag_alloc has the default
575 * memory alignment implemented by malloc.  To reference private data
576 * one can use a construct like:
577 *
578 *	struct m_tag *mtag = m_tag_alloc(...);
579 *	struct foo *p = (struct foo *)(mtag+1);
580 *
581 * if the alignment of struct m_tag is sufficient for referencing members
582 * of struct foo.  Otherwise it is necessary to embed struct m_tag within
583 * the private data structure to insure proper alignment; e.g.,
584 *
585 *	struct foo {
586 *		struct m_tag	tag;
587 *		...
588 *	};
589 *	struct foo *p = (struct foo *) m_tag_alloc(...);
590 *	struct m_tag *mtag = &p->tag;
591 */
592
593/*
594 * Persistent tags stay with an mbuf until the mbuf is reclaimed.
595 * Otherwise tags are expected to ``vanish'' when they pass through
596 * a network interface.  For most interfaces this happens normally
597 * as the tags are reclaimed when the mbuf is free'd.  However in
598 * some special cases reclaiming must be done manually.  An example
599 * is packets that pass through the loopback interface.  Also, one
600 * must be careful to do this when ``turning around'' packets (e.g.,
601 * icmp_reflect).
602 *
603 * To mark a tag persistent bit-or this flag in when defining the
604 * tag id.  The tag will then be treated as described above.
605 */
606#define	MTAG_PERSISTENT				0x800
607
608#define	PACKET_TAG_NONE				0  /* Nadda */
609
610/* Packet tags for use with PACKET_ABI_COMPAT. */
611#define	PACKET_TAG_IPSEC_IN_DONE		1  /* IPsec applied, in */
612#define	PACKET_TAG_IPSEC_OUT_DONE		2  /* IPsec applied, out */
613#define	PACKET_TAG_IPSEC_IN_CRYPTO_DONE		3  /* NIC IPsec crypto done */
614#define	PACKET_TAG_IPSEC_OUT_CRYPTO_NEEDED	4  /* NIC IPsec crypto req'ed */
615#define	PACKET_TAG_IPSEC_IN_COULD_DO_CRYPTO	5  /* NIC notifies IPsec */
616#define	PACKET_TAG_IPSEC_PENDING_TDB		6  /* Reminder to do IPsec */
617#define	PACKET_TAG_BRIDGE			7  /* Bridge processing done */
618#define	PACKET_TAG_GIF				8  /* GIF processing done */
619#define	PACKET_TAG_GRE				9  /* GRE processing done */
620#define	PACKET_TAG_IN_PACKET_CHECKSUM		10 /* NIC checksumming done */
621#define	PACKET_TAG_ENCAP			11 /* Encap.  processing */
622#define	PACKET_TAG_IPSEC_SOCKET			12 /* IPSEC socket ref */
623#define	PACKET_TAG_IPSEC_HISTORY		13 /* IPSEC history */
624#define	PACKET_TAG_IPV6_INPUT			14 /* IPV6 input processing */
625#define	PACKET_TAG_DUMMYNET			15 /* dummynet info */
626#define	PACKET_TAG_DIVERT			17 /* divert info */
627#define	PACKET_TAG_IPFORWARD			18 /* ipforward info */
628#define	PACKET_TAG_MACLABEL	(19 | MTAG_PERSISTENT) /* MAC label */
629#define	PACKET_TAG_PF_GENERATED	(20 | MTAG_PERSISTENT) /* PF, pass always */
630#define	PACKET_TAG_PF_ROUTED			21 /* PF routed, avoid loops */
631#define	PACKET_TAG_PF_FRAGCACHE			22 /* PF fragment cached */
632#define	PACKET_TAG_PF_QID			23 /* PF ALTQ queue id */
633#define	PACKET_TAG_PF_TAG			24 /* PF tagged */
634
635/* Packet tag routines. */
636struct	m_tag	*m_tag_alloc(u_int32_t, int, int, int);
637void		 m_tag_delete(struct mbuf *, struct m_tag *);
638void		 m_tag_delete_chain(struct mbuf *, struct m_tag *);
639struct	m_tag	*m_tag_locate(struct mbuf *, u_int32_t, int, struct m_tag *);
640struct	m_tag	*m_tag_copy(struct m_tag *, int);
641int		 m_tag_copy_chain(struct mbuf *, struct mbuf *, int);
642void		 m_tag_delete_nonpersistent(struct mbuf *);
643
644/*
645 * Initialize the list of tags associated with an mbuf.
646 */
647static __inline void
648m_tag_init(struct mbuf *m)
649{
650	SLIST_INIT(&m->m_pkthdr.tags);
651}
652
653/*
654 * Set up the contents of a tag.  Note that this does not
655 * fill in the free method; the caller is expected to do that.
656 *
657 * XXX probably should be called m_tag_init, but that was
658 * already taken.
659 */
660static __inline void
661m_tag_setup(struct m_tag *t, u_int32_t cookie, int type, int len)
662{
663	t->m_tag_id = type;
664	t->m_tag_len = len;
665	t->m_tag_cookie = cookie;
666}
667
668/*
669 * Reclaim resources associated with a tag.
670 */
671static __inline void
672m_tag_free(struct m_tag *t)
673{
674	(*t->m_tag_free)(t);
675}
676
677/*
678 * Return the first tag associated with an mbuf.
679 */
680static __inline struct m_tag *
681m_tag_first(struct mbuf *m)
682{
683	return (SLIST_FIRST(&m->m_pkthdr.tags));
684}
685
686/*
687 * Return the next tag in the list of tags associated with an mbuf.
688 */
689static __inline struct m_tag *
690m_tag_next(struct mbuf *m, struct m_tag *t)
691{
692	return (SLIST_NEXT(t, m_tag_link));
693}
694
695/*
696 * Prepend a tag to the list of tags associated with an mbuf.
697 */
698static __inline void
699m_tag_prepend(struct mbuf *m, struct m_tag *t)
700{
701	SLIST_INSERT_HEAD(&m->m_pkthdr.tags, t, m_tag_link);
702}
703
704/*
705 * Unlink a tag from the list of tags associated with an mbuf.
706 */
707static __inline void
708m_tag_unlink(struct mbuf *m, struct m_tag *t)
709{
710	SLIST_REMOVE(&m->m_pkthdr.tags, t, m_tag, m_tag_link);
711}
712
713/* These are for OpenBSD compatibility. */
714#define	MTAG_ABI_COMPAT		0		/* compatibility ABI */
715
716static __inline struct m_tag *
717m_tag_get(int type, int length, int wait)
718{
719	return (m_tag_alloc(MTAG_ABI_COMPAT, type, length, wait));
720}
721
722static __inline struct m_tag *
723m_tag_find(struct mbuf *m, int type, struct m_tag *start)
724{
725	return (SLIST_EMPTY(&m->m_pkthdr.tags) ?
726	    NULL : m_tag_locate(m, MTAG_ABI_COMPAT, type, start));
727}
728
729/*
730 * Obtain next_hop information associated with the mbuf, if any.
731 * If a tag is present devalidate it also.
732 */
733static __inline struct sockaddr_in *
734m_claim_next(struct mbuf *m, int type)
735{
736	struct m_tag *mtag = m_tag_find(m, type, NULL);
737	if (mtag) {
738		struct sockaddr_in *sin = *(struct sockaddr_in **)(mtag + 1);
739		mtag->m_tag_id = PACKET_TAG_NONE;
740		return (sin);
741	}
742	return (NULL);
743}
744#endif /* _KERNEL */
745
746#endif /* !_SYS_MBUF_H_ */
747