xref: /freebsd/sys/sys/mbuf.h (revision 63518eccca27064285cf2e680510ba9a4c3e2231)
1 /*-
2  * Copyright (c) 1982, 1986, 1988, 1993
3  *	The Regents of the University of California.
4  * All rights reserved.
5  *
6  * Redistribution and use in source and binary forms, with or without
7  * modification, are permitted provided that the following conditions
8  * are met:
9  * 1. Redistributions of source code must retain the above copyright
10  *    notice, this list of conditions and the following disclaimer.
11  * 2. Redistributions in binary form must reproduce the above copyright
12  *    notice, this list of conditions and the following disclaimer in the
13  *    documentation and/or other materials provided with the distribution.
14  * 3. Neither the name of the University nor the names of its contributors
15  *    may be used to endorse or promote products derived from this software
16  *    without specific prior written permission.
17  *
18  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
19  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
20  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
21  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
22  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
23  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
24  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
25  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
26  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
27  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
28  * SUCH DAMAGE.
29  *
30  *	@(#)mbuf.h	8.5 (Berkeley) 2/19/95
31  * $FreeBSD$
32  */
33 
34 #ifndef _SYS_MBUF_H_
35 #define	_SYS_MBUF_H_
36 
37 /* XXX: These includes suck. Sorry! */
38 #include <sys/queue.h>
39 #ifdef _KERNEL
40 #include <sys/systm.h>
41 #include <vm/uma.h>
42 #ifdef WITNESS
43 #include <sys/lock.h>
44 #endif
45 #endif
46 
47 /*
48  * Mbufs are of a single size, MSIZE (sys/param.h), which includes overhead.
49  * An mbuf may add a single "mbuf cluster" of size MCLBYTES (also in
50  * sys/param.h), which has no additional overhead and is used instead of the
51  * internal data area; this is done when at least MINCLSIZE of data must be
52  * stored.  Additionally, it is possible to allocate a separate buffer
53  * externally and attach it to the mbuf in a way similar to that of mbuf
54  * clusters.
55  */
56 #define	MLEN		(MSIZE - sizeof(struct m_hdr))	/* normal data len */
57 #define	MHLEN		(MLEN - sizeof(struct pkthdr))	/* data len w/pkthdr */
58 #define	MINCLSIZE	(MHLEN + 1)	/* smallest amount to put in cluster */
59 #define	M_MAXCOMPRESS	(MHLEN / 2)	/* max amount to copy for compression */
60 
61 #ifdef _KERNEL
62 /*-
63  * Macros for type conversion:
64  * mtod(m, t)	-- Convert mbuf pointer to data pointer of correct type.
65  * dtom(x)	-- Convert data pointer within mbuf to mbuf pointer (XXX).
66  */
67 #define	mtod(m, t)	((t)((m)->m_data))
68 #define	dtom(x)		((struct mbuf *)((intptr_t)(x) & ~(MSIZE-1)))
69 
70 /*
71  * Argument structure passed to UMA routines during mbuf and packet
72  * allocations.
73  */
74 struct mb_args {
75 	int	flags;	/* Flags for mbuf being allocated */
76 	short	type;	/* Type of mbuf being allocated */
77 };
78 #endif /* _KERNEL */
79 
80 /*
81  * Header present at the beginning of every mbuf.
82  */
83 struct m_hdr {
84 	struct mbuf	*mh_next;	/* next buffer in chain */
85 	struct mbuf	*mh_nextpkt;	/* next chain in queue/record */
86 	caddr_t		 mh_data;	/* location of data */
87 	int		 mh_len;	/* amount of data in this mbuf */
88 	int		 mh_flags;	/* flags; see below */
89 	short		 mh_type;	/* type of data in this mbuf */
90 };
91 
92 /*
93  * Packet tag structure (see below for details).
94  */
95 struct m_tag {
96 	SLIST_ENTRY(m_tag)	m_tag_link;	/* List of packet tags */
97 	u_int16_t		m_tag_id;	/* Tag ID */
98 	u_int16_t		m_tag_len;	/* Length of data */
99 	u_int32_t		m_tag_cookie;	/* ABI/Module ID */
100 	void			(*m_tag_free)(struct m_tag *);
101 };
102 
103 /*
104  * Record/packet header in first mbuf of chain; valid only if M_PKTHDR is set.
105  */
106 struct pkthdr {
107 	struct ifnet	*rcvif;		/* rcv interface */
108 	int		 len;		/* total packet length */
109 	/* variables for ip and tcp reassembly */
110 	void		*header;	/* pointer to packet header */
111 	/* variables for hardware checksum */
112 	int		 csum_flags;	/* flags regarding checksum */
113 	int		 csum_data;	/* data field used by csum routines */
114 	u_int16_t	 tso_segsz;	/* TSO segment size */
115 	u_int16_t	 ether_vtag;	/* Ethernet 802.1p+q vlan tag */
116 	SLIST_HEAD(packet_tags, m_tag) tags; /* list of packet tags */
117 };
118 
119 /*
120  * Description of external storage mapped into mbuf; valid only if M_EXT is
121  * set.
122  */
123 struct m_ext {
124 	caddr_t		 ext_buf;	/* start of buffer */
125 	void		(*ext_free)	/* free routine if not the usual */
126 			    (void *, void *);
127 	void		*ext_args;	/* optional argument pointer */
128 	u_int		 ext_size;	/* size of buffer, for ext_free */
129 	volatile u_int	*ref_cnt;	/* pointer to ref count info */
130 	int		 ext_type;	/* type of external storage */
131 };
132 
133 /*
134  * The core of the mbuf object along with some shortcut defines for practical
135  * purposes.
136  */
137 struct mbuf {
138 	struct m_hdr	m_hdr;
139 	union {
140 		struct {
141 			struct pkthdr	MH_pkthdr;	/* M_PKTHDR set */
142 			union {
143 				struct m_ext	MH_ext;	/* M_EXT set */
144 				char		MH_databuf[MHLEN];
145 			} MH_dat;
146 		} MH;
147 		char	M_databuf[MLEN];		/* !M_PKTHDR, !M_EXT */
148 	} M_dat;
149 };
150 #define	m_next		m_hdr.mh_next
151 #define	m_len		m_hdr.mh_len
152 #define	m_data		m_hdr.mh_data
153 #define	m_type		m_hdr.mh_type
154 #define	m_flags		m_hdr.mh_flags
155 #define	m_nextpkt	m_hdr.mh_nextpkt
156 #define	m_act		m_nextpkt
157 #define	m_pkthdr	M_dat.MH.MH_pkthdr
158 #define	m_ext		M_dat.MH.MH_dat.MH_ext
159 #define	m_pktdat	M_dat.MH.MH_dat.MH_databuf
160 #define	m_dat		M_dat.M_databuf
161 
162 /*
163  * mbuf flags.
164  */
165 #define	M_EXT		0x0001	/* has associated external storage */
166 #define	M_PKTHDR	0x0002	/* start of record */
167 #define	M_EOR		0x0004	/* end of record */
168 #define	M_RDONLY	0x0008	/* associated data is marked read-only */
169 #define	M_PROTO1	0x0010	/* protocol-specific */
170 #define	M_PROTO2	0x0020	/* protocol-specific */
171 #define	M_PROTO3	0x0040	/* protocol-specific */
172 #define	M_PROTO4	0x0080	/* protocol-specific */
173 #define	M_PROTO5	0x0100	/* protocol-specific */
174 #define	M_NOTIFICATION	0x2000	/* SCTP notification */
175 #define	M_SKIP_FIREWALL	0x4000	/* skip firewall processing */
176 #define	M_FREELIST	0x8000	/* mbuf is on the free list */
177 
178 /*
179  * mbuf pkthdr flags (also stored in m_flags).
180  */
181 #define	M_BCAST		0x0200	/* send/received as link-level broadcast */
182 #define	M_MCAST		0x0400	/* send/received as link-level multicast */
183 #define	M_FRAG		0x0800	/* packet is a fragment of a larger packet */
184 #define	M_FIRSTFRAG	0x1000	/* packet is first fragment */
185 #define	M_LASTFRAG	0x2000	/* packet is last fragment */
186 #define	M_VLANTAG	0x10000	/* ether_vtag is valid */
187 #define	M_PROMISC	0x20000	/* packet was not for us */
188 
189 /*
190  * External buffer types: identify ext_buf type.
191  */
192 #define	EXT_CLUSTER	1	/* mbuf cluster */
193 #define	EXT_SFBUF	2	/* sendfile(2)'s sf_bufs */
194 #define	EXT_JUMBOP	3	/* jumbo cluster 4096 bytes */
195 #define	EXT_JUMBO9	4	/* jumbo cluster 9216 bytes */
196 #define	EXT_JUMBO16	5	/* jumbo cluster 16184 bytes */
197 #define	EXT_PACKET	6	/* mbuf+cluster from packet zone */
198 #define	EXT_NET_DRV	100	/* custom ext_buf provided by net driver(s) */
199 #define	EXT_MOD_TYPE	200	/* custom module's ext_buf type */
200 #define	EXT_DISPOSABLE	300	/* can throw this buffer away w/page flipping */
201 #define	EXT_EXTREF	400	/* has externally maintained ref_cnt ptr */
202 
203 /*
204  * Flags copied when copying m_pkthdr.
205  */
206 #define	M_COPYFLAGS	(M_PKTHDR|M_EOR|M_RDONLY|M_PROTO1|M_PROTO1|M_PROTO2|\
207 			    M_PROTO3|M_PROTO4|M_PROTO5|M_SKIP_FIREWALL|\
208 			    M_BCAST|M_MCAST|M_FRAG|M_FIRSTFRAG|M_LASTFRAG|\
209 			    M_VLANTAG|M_PROMISC)
210 
211 /*
212  * Flags to purge when crossing layers.
213  */
214 #define	M_PROTOFLAGS	(M_PROTO1|M_PROTO2|M_PROTO3|M_PROTO4|M_PROTO5)
215 
216 /*
217  * Flags indicating hw checksum support and sw checksum requirements.  This
218  * field can be directly tested against if_data.ifi_hwassist.
219  */
220 #define	CSUM_IP			0x0001		/* will csum IP */
221 #define	CSUM_TCP		0x0002		/* will csum TCP */
222 #define	CSUM_UDP		0x0004		/* will csum UDP */
223 #define	CSUM_IP_FRAGS		0x0008		/* will csum IP fragments */
224 #define	CSUM_FRAGMENT		0x0010		/* will do IP fragmentation */
225 #define	CSUM_TSO		0x0020		/* will do TSO */
226 
227 #define	CSUM_IP_CHECKED		0x0100		/* did csum IP */
228 #define	CSUM_IP_VALID		0x0200		/*   ... the csum is valid */
229 #define	CSUM_DATA_VALID		0x0400		/* csum_data field is valid */
230 #define	CSUM_PSEUDO_HDR		0x0800		/* csum_data has pseudo hdr */
231 
232 #define	CSUM_DELAY_DATA		(CSUM_TCP | CSUM_UDP)
233 #define	CSUM_DELAY_IP		(CSUM_IP)	/* XXX add ipv6 here too? */
234 
235 /*
236  * mbuf types.
237  */
238 #define	MT_NOTMBUF	0	/* USED INTERNALLY ONLY! Object is not mbuf */
239 #define	MT_DATA		1	/* dynamic (data) allocation */
240 #define	MT_HEADER	MT_DATA	/* packet header, use M_PKTHDR instead */
241 #define	MT_SONAME	8	/* socket name */
242 #define	MT_CONTROL	14	/* extra-data protocol message */
243 #define	MT_OOBDATA	15	/* expedited data  */
244 #define	MT_NTYPES	16	/* number of mbuf types for mbtypes[] */
245 
246 #define	MT_NOINIT	255	/* Not a type but a flag to allocate
247 				   a non-initialized mbuf */
248 
249 /*
250  * General mbuf allocator statistics structure.
251  *
252  * Many of these statistics are no longer used; we instead track many
253  * allocator statistics through UMA's built in statistics mechanism.
254  */
255 struct mbstat {
256 	u_long	m_mbufs;	/* XXX */
257 	u_long	m_mclusts;	/* XXX */
258 
259 	u_long	m_drain;	/* times drained protocols for space */
260 	u_long	m_mcfail;	/* XXX: times m_copym failed */
261 	u_long	m_mpfail;	/* XXX: times m_pullup failed */
262 	u_long	m_msize;	/* length of an mbuf */
263 	u_long	m_mclbytes;	/* length of an mbuf cluster */
264 	u_long	m_minclsize;	/* min length of data to allocate a cluster */
265 	u_long	m_mlen;		/* length of data in an mbuf */
266 	u_long	m_mhlen;	/* length of data in a header mbuf */
267 
268 	/* Number of mbtypes (gives # elems in mbtypes[] array: */
269 	short	m_numtypes;
270 
271 	/* XXX: Sendfile stats should eventually move to their own struct */
272 	u_long	sf_iocnt;	/* times sendfile had to do disk I/O */
273 	u_long	sf_allocfail;	/* times sfbuf allocation failed */
274 	u_long	sf_allocwait;	/* times sfbuf allocation had to wait */
275 };
276 
277 /*
278  * Flags specifying how an allocation should be made.
279  *
280  * The flag to use is as follows:
281  * - M_DONTWAIT or M_NOWAIT from an interrupt handler to not block allocation.
282  * - M_WAIT or M_WAITOK or M_TRYWAIT from wherever it is safe to block.
283  *
284  * M_DONTWAIT/M_NOWAIT means that we will not block the thread explicitly and
285  * if we cannot allocate immediately we may return NULL, whereas
286  * M_WAIT/M_WAITOK/M_TRYWAIT means that if we cannot allocate resources we
287  * will block until they are available, and thus never return NULL.
288  *
289  * XXX Eventually just phase this out to use M_WAITOK/M_NOWAIT.
290  */
291 #define	MBTOM(how)	(how)
292 #define	M_DONTWAIT	M_NOWAIT
293 #define	M_TRYWAIT	M_WAITOK
294 #define	M_WAIT		M_WAITOK
295 
296 /*
297  * String names of mbuf-related UMA(9) and malloc(9) types.  Exposed to
298  * !_KERNEL so that monitoring tools can look up the zones with
299  * libmemstat(3).
300  */
301 #define	MBUF_MEM_NAME		"mbuf"
302 #define	MBUF_CLUSTER_MEM_NAME	"mbuf_cluster"
303 #define	MBUF_PACKET_MEM_NAME	"mbuf_packet"
304 #define	MBUF_JUMBOP_MEM_NAME	"mbuf_jumbo_pagesize"
305 #define	MBUF_JUMBO9_MEM_NAME	"mbuf_jumbo_9k"
306 #define	MBUF_JUMBO16_MEM_NAME	"mbuf_jumbo_16k"
307 #define	MBUF_TAG_MEM_NAME	"mbuf_tag"
308 #define	MBUF_EXTREFCNT_MEM_NAME	"mbuf_ext_refcnt"
309 
310 #ifdef _KERNEL
311 
312 #ifdef WITNESS
313 #define	MBUF_CHECKSLEEP(how) do {					\
314 	if (how == M_WAITOK)						\
315 		WITNESS_WARN(WARN_GIANTOK | WARN_SLEEPOK, NULL,		\
316 		    "Sleeping in \"%s\"", __func__);			\
317 } while (0)
318 #else
319 #define	MBUF_CHECKSLEEP(how)
320 #endif
321 
322 /*
323  * Network buffer allocation API
324  *
325  * The rest of it is defined in kern/kern_mbuf.c
326  */
327 
328 extern uma_zone_t	zone_mbuf;
329 extern uma_zone_t	zone_clust;
330 extern uma_zone_t	zone_pack;
331 extern uma_zone_t	zone_jumbop;
332 extern uma_zone_t	zone_jumbo9;
333 extern uma_zone_t	zone_jumbo16;
334 extern uma_zone_t	zone_ext_refcnt;
335 
336 static __inline struct mbuf	*m_get(int how, short type);
337 static __inline struct mbuf	*m_gethdr(int how, short type);
338 static __inline struct mbuf	*m_getcl(int how, short type, int flags);
339 static __inline struct mbuf	*m_getjcl(int how, short type, int flags,
340 				    int size);
341 static __inline struct mbuf	*m_getclr(int how, short type);	/* XXX */
342 static __inline struct mbuf	*m_free(struct mbuf *m);
343 static __inline void		 m_clget(struct mbuf *m, int how);
344 static __inline void		*m_cljget(struct mbuf *m, int how, int size);
345 static __inline void		 m_chtype(struct mbuf *m, short new_type);
346 void				 mb_free_ext(struct mbuf *);
347 
348 static __inline struct mbuf *
349 m_get(int how, short type)
350 {
351 	struct mb_args args;
352 
353 	args.flags = 0;
354 	args.type = type;
355 	return ((struct mbuf *)(uma_zalloc_arg(zone_mbuf, &args, how)));
356 }
357 
358 /*
359  * XXX This should be deprecated, very little use.
360  */
361 static __inline struct mbuf *
362 m_getclr(int how, short type)
363 {
364 	struct mbuf *m;
365 	struct mb_args args;
366 
367 	args.flags = 0;
368 	args.type = type;
369 	m = uma_zalloc_arg(zone_mbuf, &args, how);
370 	if (m != NULL)
371 		bzero(m->m_data, MLEN);
372 	return (m);
373 }
374 
375 static __inline struct mbuf *
376 m_gethdr(int how, short type)
377 {
378 	struct mb_args args;
379 
380 	args.flags = M_PKTHDR;
381 	args.type = type;
382 	return ((struct mbuf *)(uma_zalloc_arg(zone_mbuf, &args, how)));
383 }
384 
385 static __inline struct mbuf *
386 m_getcl(int how, short type, int flags)
387 {
388 	struct mb_args args;
389 
390 	args.flags = flags;
391 	args.type = type;
392 	return ((struct mbuf *)(uma_zalloc_arg(zone_pack, &args, how)));
393 }
394 
395 /*
396  * m_getjcl() returns an mbuf with a cluster of the specified size attached.
397  * For size it takes MCLBYTES, MJUMPAGESIZE, MJUM9BYTES, MJUM16BYTES.
398  *
399  * XXX: This is rather large, should be real function maybe.
400  */
401 static __inline struct mbuf *
402 m_getjcl(int how, short type, int flags, int size)
403 {
404 	struct mb_args args;
405 	struct mbuf *m, *n;
406 	uma_zone_t zone;
407 
408 	args.flags = flags;
409 	args.type = type;
410 
411 	m = uma_zalloc_arg(zone_mbuf, &args, how);
412 	if (m == NULL)
413 		return (NULL);
414 
415 	switch (size) {
416 	case MCLBYTES:
417 		zone = zone_clust;
418 		break;
419 #if MJUMPAGESIZE != MCLBYTES
420 	case MJUMPAGESIZE:
421 		zone = zone_jumbop;
422 		break;
423 #endif
424 	case MJUM9BYTES:
425 		zone = zone_jumbo9;
426 		break;
427 	case MJUM16BYTES:
428 		zone = zone_jumbo16;
429 		break;
430 	default:
431 		panic("%s: m_getjcl: invalid cluster type", __func__);
432 	}
433 	n = uma_zalloc_arg(zone, m, how);
434 	if (n == NULL) {
435 		uma_zfree(zone_mbuf, m);
436 		return (NULL);
437 	}
438 	return (m);
439 }
440 
441 static __inline struct mbuf *
442 m_free(struct mbuf *m)
443 {
444 	struct mbuf *n = m->m_next;
445 
446 	if (m->m_flags & M_EXT)
447 		mb_free_ext(m);
448 	else
449 		uma_zfree(zone_mbuf, m);
450 	return (n);
451 }
452 
453 static __inline void
454 m_clget(struct mbuf *m, int how)
455 {
456 
457 	if (m->m_flags & M_EXT)
458 		printf("%s: %p mbuf already has cluster\n", __func__, m);
459 	m->m_ext.ext_buf = (char *)NULL;
460 	uma_zalloc_arg(zone_clust, m, how);
461 	/*
462 	 * On a cluster allocation failure, drain the packet zone and retry,
463 	 * we might be able to loosen a few clusters up on the drain.
464 	 */
465 	if ((how & M_NOWAIT) && (m->m_ext.ext_buf == NULL)) {
466 		zone_drain(zone_pack);
467 		uma_zalloc_arg(zone_clust, m, how);
468 	}
469 }
470 
471 /*
472  * m_cljget() is different from m_clget() as it can allocate clusters without
473  * attaching them to an mbuf.  In that case the return value is the pointer
474  * to the cluster of the requested size.  If an mbuf was specified, it gets
475  * the cluster attached to it and the return value can be safely ignored.
476  * For size it takes MCLBYTES, MJUMPAGESIZE, MJUM9BYTES, MJUM16BYTES.
477  */
478 static __inline void *
479 m_cljget(struct mbuf *m, int how, int size)
480 {
481 	uma_zone_t zone;
482 
483 	if (m && m->m_flags & M_EXT)
484 		printf("%s: %p mbuf already has cluster\n", __func__, m);
485 	if (m != NULL)
486 		m->m_ext.ext_buf = NULL;
487 
488 	switch (size) {
489 	case MCLBYTES:
490 		zone = zone_clust;
491 		break;
492 #if MJUMPAGESIZE != MCLBYTES
493 	case MJUMPAGESIZE:
494 		zone = zone_jumbop;
495 		break;
496 #endif
497 	case MJUM9BYTES:
498 		zone = zone_jumbo9;
499 		break;
500 	case MJUM16BYTES:
501 		zone = zone_jumbo16;
502 		break;
503 	default:
504 		panic("%s: m_getjcl: invalid cluster type", __func__);
505 	}
506 
507 	return (uma_zalloc_arg(zone, m, how));
508 }
509 
510 static __inline void
511 m_chtype(struct mbuf *m, short new_type)
512 {
513 
514 	m->m_type = new_type;
515 }
516 
517 /*
518  * mbuf, cluster, and external object allocation macros (for compatibility
519  * purposes).
520  */
521 #define	M_MOVE_PKTHDR(to, from)	m_move_pkthdr((to), (from))
522 #define	MGET(m, how, type)	((m) = m_get((how), (type)))
523 #define	MGETHDR(m, how, type)	((m) = m_gethdr((how), (type)))
524 #define	MCLGET(m, how)		m_clget((m), (how))
525 #define	MEXTADD(m, buf, size, free, args, flags, type) 			\
526     m_extadd((m), (caddr_t)(buf), (size), (free), (args), (flags), (type))
527 #define	m_getm(m, len, how, type)					\
528     m_getm2((m), (len), (how), (type), M_PKTHDR)
529 
530 /*
531  * Evaluate TRUE if it's safe to write to the mbuf m's data region (this can
532  * be both the local data payload, or an external buffer area, depending on
533  * whether M_EXT is set).
534  */
535 #define	M_WRITABLE(m)	(!((m)->m_flags & M_RDONLY) &&			\
536 			 (!(((m)->m_flags & M_EXT)) ||			\
537 			 (*((m)->m_ext.ref_cnt) == 1)) )		\
538 
539 /* Check if the supplied mbuf has a packet header, or else panic. */
540 #define	M_ASSERTPKTHDR(m)						\
541 	KASSERT(m != NULL && m->m_flags & M_PKTHDR,			\
542 	    ("%s: no mbuf packet header!", __func__))
543 
544 /*
545  * Ensure that the supplied mbuf is a valid, non-free mbuf.
546  *
547  * XXX: Broken at the moment.  Need some UMA magic to make it work again.
548  */
549 #define	M_ASSERTVALID(m)						\
550 	KASSERT((((struct mbuf *)m)->m_flags & 0) == 0,			\
551 	    ("%s: attempted use of a free mbuf!", __func__))
552 
553 /*
554  * Set the m_data pointer of a newly-allocated mbuf (m_get/MGET) to place an
555  * object of the specified size at the end of the mbuf, longword aligned.
556  */
557 #define	M_ALIGN(m, len) do {						\
558 	KASSERT(!((m)->m_flags & (M_PKTHDR|M_EXT)),			\
559 		("%s: M_ALIGN not normal mbuf", __func__));		\
560 	KASSERT((m)->m_data == (m)->m_dat,				\
561 		("%s: M_ALIGN not a virgin mbuf", __func__));		\
562 	(m)->m_data += (MLEN - (len)) & ~(sizeof(long) - 1);		\
563 } while (0)
564 
565 /*
566  * As above, for mbufs allocated with m_gethdr/MGETHDR or initialized by
567  * M_DUP/MOVE_PKTHDR.
568  */
569 #define	MH_ALIGN(m, len) do {						\
570 	KASSERT((m)->m_flags & M_PKTHDR && !((m)->m_flags & M_EXT),	\
571 		("%s: MH_ALIGN not PKTHDR mbuf", __func__));		\
572 	KASSERT((m)->m_data == (m)->m_pktdat,				\
573 		("%s: MH_ALIGN not a virgin mbuf", __func__));		\
574 	(m)->m_data += (MHLEN - (len)) & ~(sizeof(long) - 1);		\
575 } while (0)
576 
577 /*
578  * Compute the amount of space available before the current start of data in
579  * an mbuf.
580  *
581  * The M_WRITABLE() is a temporary, conservative safety measure: the burden
582  * of checking writability of the mbuf data area rests solely with the caller.
583  */
584 #define	M_LEADINGSPACE(m)						\
585 	((m)->m_flags & M_EXT ?						\
586 	    (M_WRITABLE(m) ? (m)->m_data - (m)->m_ext.ext_buf : 0):	\
587 	    (m)->m_flags & M_PKTHDR ? (m)->m_data - (m)->m_pktdat :	\
588 	    (m)->m_data - (m)->m_dat)
589 
590 /*
591  * Compute the amount of space available after the end of data in an mbuf.
592  *
593  * The M_WRITABLE() is a temporary, conservative safety measure: the burden
594  * of checking writability of the mbuf data area rests solely with the caller.
595  */
596 #define	M_TRAILINGSPACE(m)						\
597 	((m)->m_flags & M_EXT ?						\
598 	    (M_WRITABLE(m) ? (m)->m_ext.ext_buf + (m)->m_ext.ext_size	\
599 		- ((m)->m_data + (m)->m_len) : 0) :			\
600 	    &(m)->m_dat[MLEN] - ((m)->m_data + (m)->m_len))
601 
602 /*
603  * Arrange to prepend space of size plen to mbuf m.  If a new mbuf must be
604  * allocated, how specifies whether to wait.  If the allocation fails, the
605  * original mbuf chain is freed and m is set to NULL.
606  */
607 #define	M_PREPEND(m, plen, how) do {					\
608 	struct mbuf **_mmp = &(m);					\
609 	struct mbuf *_mm = *_mmp;					\
610 	int _mplen = (plen);						\
611 	int __mhow = (how);						\
612 									\
613 	MBUF_CHECKSLEEP(how);						\
614 	if (M_LEADINGSPACE(_mm) >= _mplen) {				\
615 		_mm->m_data -= _mplen;					\
616 		_mm->m_len += _mplen;					\
617 	} else								\
618 		_mm = m_prepend(_mm, _mplen, __mhow);			\
619 	if (_mm != NULL && _mm->m_flags & M_PKTHDR)			\
620 		_mm->m_pkthdr.len += _mplen;				\
621 	*_mmp = _mm;							\
622 } while (0)
623 
624 /*
625  * Change mbuf to new type.  This is a relatively expensive operation and
626  * should be avoided.
627  */
628 #define	MCHTYPE(m, t)	m_chtype((m), (t))
629 
630 /* Length to m_copy to copy all. */
631 #define	M_COPYALL	1000000000
632 
633 /* Compatibility with 4.3. */
634 #define	m_copy(m, o, l)	m_copym((m), (o), (l), M_DONTWAIT)
635 
636 extern int		max_datalen;	/* MHLEN - max_hdr */
637 extern int		max_hdr;	/* Largest link + protocol header */
638 extern int		max_linkhdr;	/* Largest link-level header */
639 extern int		max_protohdr;	/* Largest protocol header */
640 extern struct mbstat	mbstat;		/* General mbuf stats/infos */
641 extern int		nmbclusters;	/* Maximum number of clusters */
642 
643 struct uio;
644 
645 void		 m_adj(struct mbuf *, int);
646 void		 m_align(struct mbuf *, int);
647 int		 m_apply(struct mbuf *, int, int,
648 		    int (*)(void *, void *, u_int), void *);
649 int		 m_append(struct mbuf *, int, c_caddr_t);
650 void		 m_cat(struct mbuf *, struct mbuf *);
651 void		 m_extadd(struct mbuf *, caddr_t, u_int,
652 		    void (*)(void *, void *), void *, int, int);
653 void		 m_copyback(struct mbuf *, int, int, c_caddr_t);
654 void		 m_copydata(const struct mbuf *, int, int, caddr_t);
655 struct mbuf	*m_copym(struct mbuf *, int, int, int);
656 struct mbuf	*m_copymdata(struct mbuf *, struct mbuf *,
657 		    int, int, int, int);
658 struct mbuf	*m_copypacket(struct mbuf *, int);
659 void		 m_copy_pkthdr(struct mbuf *, struct mbuf *);
660 struct mbuf	*m_copyup(struct mbuf *n, int len, int dstoff);
661 struct mbuf	*m_defrag(struct mbuf *, int);
662 void		 m_demote(struct mbuf *, int);
663 struct mbuf	*m_devget(char *, int, int, struct ifnet *,
664 		    void (*)(char *, caddr_t, u_int));
665 struct mbuf	*m_dup(struct mbuf *, int);
666 int		 m_dup_pkthdr(struct mbuf *, struct mbuf *, int);
667 u_int		 m_fixhdr(struct mbuf *);
668 struct mbuf	*m_fragment(struct mbuf *, int, int);
669 void		 m_freem(struct mbuf *);
670 struct mbuf	*m_getm2(struct mbuf *, int, int, short, int);
671 struct mbuf	*m_getptr(struct mbuf *, int, int *);
672 u_int		 m_length(struct mbuf *, struct mbuf **);
673 void		 m_move_pkthdr(struct mbuf *, struct mbuf *);
674 struct mbuf	*m_prepend(struct mbuf *, int, int);
675 void		 m_print(const struct mbuf *, int);
676 struct mbuf	*m_pulldown(struct mbuf *, int, int, int *);
677 struct mbuf	*m_pullup(struct mbuf *, int);
678 int		m_sanity(struct mbuf *, int);
679 struct mbuf	*m_split(struct mbuf *, int, int);
680 struct mbuf	*m_uiotombuf(struct uio *, int, int, int, int);
681 struct mbuf	*m_unshare(struct mbuf *, int how);
682 
683 /*-
684  * Network packets may have annotations attached by affixing a list of
685  * "packet tags" to the pkthdr structure.  Packet tags are dynamically
686  * allocated semi-opaque data structures that have a fixed header
687  * (struct m_tag) that specifies the size of the memory block and a
688  * <cookie,type> pair that identifies it.  The cookie is a 32-bit unique
689  * unsigned value used to identify a module or ABI.  By convention this value
690  * is chosen as the date+time that the module is created, expressed as the
691  * number of seconds since the epoch (e.g., using date -u +'%s').  The type
692  * value is an ABI/module-specific value that identifies a particular
693  * annotation and is private to the module.  For compatibility with systems
694  * like OpenBSD that define packet tags w/o an ABI/module cookie, the value
695  * PACKET_ABI_COMPAT is used to implement m_tag_get and m_tag_find
696  * compatibility shim functions and several tag types are defined below.
697  * Users that do not require compatibility should use a private cookie value
698  * so that packet tag-related definitions can be maintained privately.
699  *
700  * Note that the packet tag returned by m_tag_alloc has the default memory
701  * alignment implemented by malloc.  To reference private data one can use a
702  * construct like:
703  *
704  *	struct m_tag *mtag = m_tag_alloc(...);
705  *	struct foo *p = (struct foo *)(mtag+1);
706  *
707  * if the alignment of struct m_tag is sufficient for referencing members of
708  * struct foo.  Otherwise it is necessary to embed struct m_tag within the
709  * private data structure to insure proper alignment; e.g.,
710  *
711  *	struct foo {
712  *		struct m_tag	tag;
713  *		...
714  *	};
715  *	struct foo *p = (struct foo *) m_tag_alloc(...);
716  *	struct m_tag *mtag = &p->tag;
717  */
718 
719 /*
720  * Persistent tags stay with an mbuf until the mbuf is reclaimed.  Otherwise
721  * tags are expected to ``vanish'' when they pass through a network
722  * interface.  For most interfaces this happens normally as the tags are
723  * reclaimed when the mbuf is free'd.  However in some special cases
724  * reclaiming must be done manually.  An example is packets that pass through
725  * the loopback interface.  Also, one must be careful to do this when
726  * ``turning around'' packets (e.g., icmp_reflect).
727  *
728  * To mark a tag persistent bit-or this flag in when defining the tag id.
729  * The tag will then be treated as described above.
730  */
731 #define	MTAG_PERSISTENT				0x800
732 
733 #define	PACKET_TAG_NONE				0  /* Nadda */
734 
735 /* Packet tags for use with PACKET_ABI_COMPAT. */
736 #define	PACKET_TAG_IPSEC_IN_DONE		1  /* IPsec applied, in */
737 #define	PACKET_TAG_IPSEC_OUT_DONE		2  /* IPsec applied, out */
738 #define	PACKET_TAG_IPSEC_IN_CRYPTO_DONE		3  /* NIC IPsec crypto done */
739 #define	PACKET_TAG_IPSEC_OUT_CRYPTO_NEEDED	4  /* NIC IPsec crypto req'ed */
740 #define	PACKET_TAG_IPSEC_IN_COULD_DO_CRYPTO	5  /* NIC notifies IPsec */
741 #define	PACKET_TAG_IPSEC_PENDING_TDB		6  /* Reminder to do IPsec */
742 #define	PACKET_TAG_BRIDGE			7  /* Bridge processing done */
743 #define	PACKET_TAG_GIF				8  /* GIF processing done */
744 #define	PACKET_TAG_GRE				9  /* GRE processing done */
745 #define	PACKET_TAG_IN_PACKET_CHECKSUM		10 /* NIC checksumming done */
746 #define	PACKET_TAG_ENCAP			11 /* Encap.  processing */
747 #define	PACKET_TAG_IPSEC_SOCKET			12 /* IPSEC socket ref */
748 #define	PACKET_TAG_IPSEC_HISTORY		13 /* IPSEC history */
749 #define	PACKET_TAG_IPV6_INPUT			14 /* IPV6 input processing */
750 #define	PACKET_TAG_DUMMYNET			15 /* dummynet info */
751 #define	PACKET_TAG_DIVERT			17 /* divert info */
752 #define	PACKET_TAG_IPFORWARD			18 /* ipforward info */
753 #define	PACKET_TAG_MACLABEL	(19 | MTAG_PERSISTENT) /* MAC label */
754 #define	PACKET_TAG_PF_ROUTED			21 /* PF routed, avoid loops */
755 #define	PACKET_TAG_PF_FRAGCACHE			22 /* PF fragment cached */
756 #define	PACKET_TAG_PF_QID			23 /* PF ALTQ queue id */
757 #define	PACKET_TAG_PF_TAG			24 /* PF tagged */
758 #define	PACKET_TAG_RTSOCKFAM			25 /* rtsock sa family */
759 #define	PACKET_TAG_PF_TRANSLATE_LOCALHOST	26 /* PF translate localhost */
760 #define	PACKET_TAG_IPOPTIONS			27 /* Saved IP options */
761 #define	PACKET_TAG_CARP                         28 /* CARP info */
762 
763 /* Specific cookies and tags. */
764 
765 /* Packet tag routines. */
766 struct m_tag	*m_tag_alloc(u_int32_t, int, int, int);
767 void		 m_tag_delete(struct mbuf *, struct m_tag *);
768 void		 m_tag_delete_chain(struct mbuf *, struct m_tag *);
769 void		 m_tag_free_default(struct m_tag *);
770 struct m_tag	*m_tag_locate(struct mbuf *, u_int32_t, int, struct m_tag *);
771 struct m_tag	*m_tag_copy(struct m_tag *, int);
772 int		 m_tag_copy_chain(struct mbuf *, struct mbuf *, int);
773 void		 m_tag_delete_nonpersistent(struct mbuf *);
774 
775 /*
776  * Initialize the list of tags associated with an mbuf.
777  */
778 static __inline void
779 m_tag_init(struct mbuf *m)
780 {
781 
782 	SLIST_INIT(&m->m_pkthdr.tags);
783 }
784 
785 /*
786  * Set up the contents of a tag.  Note that this does not fill in the free
787  * method; the caller is expected to do that.
788  *
789  * XXX probably should be called m_tag_init, but that was already taken.
790  */
791 static __inline void
792 m_tag_setup(struct m_tag *t, u_int32_t cookie, int type, int len)
793 {
794 
795 	t->m_tag_id = type;
796 	t->m_tag_len = len;
797 	t->m_tag_cookie = cookie;
798 }
799 
800 /*
801  * Reclaim resources associated with a tag.
802  */
803 static __inline void
804 m_tag_free(struct m_tag *t)
805 {
806 
807 	(*t->m_tag_free)(t);
808 }
809 
810 /*
811  * Return the first tag associated with an mbuf.
812  */
813 static __inline struct m_tag *
814 m_tag_first(struct mbuf *m)
815 {
816 
817 	return (SLIST_FIRST(&m->m_pkthdr.tags));
818 }
819 
820 /*
821  * Return the next tag in the list of tags associated with an mbuf.
822  */
823 static __inline struct m_tag *
824 m_tag_next(struct mbuf *m, struct m_tag *t)
825 {
826 
827 	return (SLIST_NEXT(t, m_tag_link));
828 }
829 
830 /*
831  * Prepend a tag to the list of tags associated with an mbuf.
832  */
833 static __inline void
834 m_tag_prepend(struct mbuf *m, struct m_tag *t)
835 {
836 
837 	SLIST_INSERT_HEAD(&m->m_pkthdr.tags, t, m_tag_link);
838 }
839 
840 /*
841  * Unlink a tag from the list of tags associated with an mbuf.
842  */
843 static __inline void
844 m_tag_unlink(struct mbuf *m, struct m_tag *t)
845 {
846 
847 	SLIST_REMOVE(&m->m_pkthdr.tags, t, m_tag, m_tag_link);
848 }
849 
850 /* These are for OpenBSD compatibility. */
851 #define	MTAG_ABI_COMPAT		0		/* compatibility ABI */
852 
853 static __inline struct m_tag *
854 m_tag_get(int type, int length, int wait)
855 {
856 	return (m_tag_alloc(MTAG_ABI_COMPAT, type, length, wait));
857 }
858 
859 static __inline struct m_tag *
860 m_tag_find(struct mbuf *m, int type, struct m_tag *start)
861 {
862 	return (SLIST_EMPTY(&m->m_pkthdr.tags) ? (struct m_tag *)NULL :
863 	    m_tag_locate(m, MTAG_ABI_COMPAT, type, start));
864 }
865 
866 #endif /* _KERNEL */
867 
868 #endif /* !_SYS_MBUF_H_ */
869