xref: /freebsd/sys/sys/mbuf.h (revision bfe691b2f75de2224c7ceb304ebcdef2b42d4179)
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_MBUF	7	/* external mbuf reference (M_IOVEC) */
199 #define	EXT_NET_DRV	100	/* custom ext_buf provided by net driver(s) */
200 #define	EXT_MOD_TYPE	200	/* custom module's ext_buf type */
201 #define	EXT_DISPOSABLE	300	/* can throw this buffer away w/page flipping */
202 #define	EXT_EXTREF	400	/* has externally maintained ref_cnt ptr */
203 
204 /*
205  * Flags copied when copying m_pkthdr.
206  */
207 #define	M_COPYFLAGS	(M_PKTHDR|M_EOR|M_RDONLY|M_PROTO1|M_PROTO1|M_PROTO2|\
208 			    M_PROTO3|M_PROTO4|M_PROTO5|M_SKIP_FIREWALL|\
209 			    M_BCAST|M_MCAST|M_FRAG|M_FIRSTFRAG|M_LASTFRAG|\
210 			    M_VLANTAG|M_PROMISC)
211 
212 /*
213  * Flags to purge when crossing layers.
214  */
215 #define	M_PROTOFLAGS	(M_PROTO1|M_PROTO2|M_PROTO3|M_PROTO4|M_PROTO5)
216 
217 /*
218  * Flags indicating hw checksum support and sw checksum requirements.  This
219  * field can be directly tested against if_data.ifi_hwassist.
220  */
221 #define	CSUM_IP			0x0001		/* will csum IP */
222 #define	CSUM_TCP		0x0002		/* will csum TCP */
223 #define	CSUM_UDP		0x0004		/* will csum UDP */
224 #define	CSUM_IP_FRAGS		0x0008		/* will csum IP fragments */
225 #define	CSUM_FRAGMENT		0x0010		/* will do IP fragmentation */
226 #define	CSUM_TSO		0x0020		/* will do TSO */
227 
228 #define	CSUM_IP_CHECKED		0x0100		/* did csum IP */
229 #define	CSUM_IP_VALID		0x0200		/*   ... the csum is valid */
230 #define	CSUM_DATA_VALID		0x0400		/* csum_data field is valid */
231 #define	CSUM_PSEUDO_HDR		0x0800		/* csum_data has pseudo hdr */
232 
233 #define	CSUM_DELAY_DATA		(CSUM_TCP | CSUM_UDP)
234 #define	CSUM_DELAY_IP		(CSUM_IP)	/* XXX add ipv6 here too? */
235 
236 /*
237  * mbuf types.
238  */
239 #define	MT_NOTMBUF	0	/* USED INTERNALLY ONLY! Object is not mbuf */
240 #define	MT_DATA		1	/* dynamic (data) allocation */
241 #define	MT_HEADER	MT_DATA	/* packet header, use M_PKTHDR instead */
242 #define	MT_SONAME	8	/* socket name */
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  *
253  * Many of these statistics are no longer used; we instead track many
254  * allocator statistics through UMA's built in statistics mechanism.
255  */
256 struct mbstat {
257 	u_long	m_mbufs;	/* XXX */
258 	u_long	m_mclusts;	/* XXX */
259 
260 	u_long	m_drain;	/* times drained protocols for space */
261 	u_long	m_mcfail;	/* XXX: times m_copym failed */
262 	u_long	m_mpfail;	/* XXX: times m_pullup failed */
263 	u_long	m_msize;	/* length of an mbuf */
264 	u_long	m_mclbytes;	/* length of an mbuf cluster */
265 	u_long	m_minclsize;	/* min length of data to allocate a cluster */
266 	u_long	m_mlen;		/* length of data in an mbuf */
267 	u_long	m_mhlen;	/* length of data in a header mbuf */
268 
269 	/* Number of mbtypes (gives # elems in mbtypes[] array: */
270 	short	m_numtypes;
271 
272 	/* XXX: Sendfile stats should eventually move to their own struct */
273 	u_long	sf_iocnt;	/* times sendfile had to do disk I/O */
274 	u_long	sf_allocfail;	/* times sfbuf allocation failed */
275 	u_long	sf_allocwait;	/* times sfbuf allocation had to wait */
276 };
277 
278 /*
279  * Flags specifying how an allocation should be made.
280  *
281  * The flag to use is as follows:
282  * - M_DONTWAIT or M_NOWAIT from an interrupt handler to not block allocation.
283  * - M_WAIT or M_WAITOK or M_TRYWAIT from wherever it is safe to block.
284  *
285  * M_DONTWAIT/M_NOWAIT means that we will not block the thread explicitly and
286  * if we cannot allocate immediately we may return NULL, whereas
287  * M_WAIT/M_WAITOK/M_TRYWAIT means that if we cannot allocate resources we
288  * will block until they are available, and thus never return NULL.
289  *
290  * XXX Eventually just phase this out to use M_WAITOK/M_NOWAIT.
291  */
292 #define	MBTOM(how)	(how)
293 #define	M_DONTWAIT	M_NOWAIT
294 #define	M_TRYWAIT	M_WAITOK
295 #define	M_WAIT		M_WAITOK
296 
297 /*
298  * String names of mbuf-related UMA(9) and malloc(9) types.  Exposed to
299  * !_KERNEL so that monitoring tools can look up the zones with
300  * libmemstat(3).
301  */
302 #define	MBUF_MEM_NAME		"mbuf"
303 #define	MBUF_CLUSTER_MEM_NAME	"mbuf_cluster"
304 #define	MBUF_PACKET_MEM_NAME	"mbuf_packet"
305 #define	MBUF_JUMBOP_MEM_NAME	"mbuf_jumbo_pagesize"
306 #define	MBUF_JUMBO9_MEM_NAME	"mbuf_jumbo_9k"
307 #define	MBUF_JUMBO16_MEM_NAME	"mbuf_jumbo_16k"
308 #define	MBUF_TAG_MEM_NAME	"mbuf_tag"
309 #define	MBUF_EXTREFCNT_MEM_NAME	"mbuf_ext_refcnt"
310 
311 #ifdef _KERNEL
312 
313 #ifdef WITNESS
314 #define	MBUF_CHECKSLEEP(how) do {					\
315 	if (how == M_WAITOK)						\
316 		WITNESS_WARN(WARN_GIANTOK | WARN_SLEEPOK, NULL,		\
317 		    "Sleeping in \"%s\"", __func__);			\
318 } while (0)
319 #else
320 #define	MBUF_CHECKSLEEP(how)
321 #endif
322 
323 /*
324  * Network buffer allocation API
325  *
326  * The rest of it is defined in kern/kern_mbuf.c
327  */
328 
329 extern uma_zone_t	zone_mbuf;
330 extern uma_zone_t	zone_clust;
331 extern uma_zone_t	zone_pack;
332 extern uma_zone_t	zone_jumbop;
333 extern uma_zone_t	zone_jumbo9;
334 extern uma_zone_t	zone_jumbo16;
335 extern uma_zone_t	zone_ext_refcnt;
336 
337 static __inline struct mbuf	*m_get(int how, short type);
338 static __inline struct mbuf	*m_gethdr(int how, short type);
339 static __inline struct mbuf	*m_getcl(int how, short type, int flags);
340 static __inline struct mbuf	*m_getjcl(int how, short type, int flags,
341 				    int size);
342 static __inline struct mbuf	*m_getclr(int how, short type);	/* XXX */
343 static __inline struct mbuf	*m_free(struct mbuf *m);
344 static __inline void		 m_clget(struct mbuf *m, int how);
345 static __inline void		*m_cljget(struct mbuf *m, int how, int size);
346 static __inline void		 m_chtype(struct mbuf *m, short new_type);
347 void				 mb_free_ext(struct mbuf *);
348 
349 static __inline int
350 m_gettype(int size)
351 {
352 	int type;
353 
354 	switch (size) {
355 	case MSIZE:
356 		type = EXT_MBUF;
357 		break;
358 	case MCLBYTES:
359 		type = EXT_CLUSTER;
360 		break;
361 #if MJUMPAGESIZE != MCLBYTES
362 	case MJUMPAGESIZE:
363 		type = EXT_JUMBOP;
364 		break;
365 #endif
366 	case MJUM9BYTES:
367 		type = EXT_JUMBO9;
368 		break;
369 	case MJUM16BYTES:
370 		type = EXT_JUMBO16;
371 		break;
372 	default:
373 		panic("%s: m_getjcl: invalid cluster size", __func__);
374 	}
375 
376 	return (type);
377 }
378 
379 static __inline uma_zone_t
380 m_getzone(int size)
381 {
382 	uma_zone_t zone;
383 
384 	switch (size) {
385 	case MSIZE:
386 		zone = zone_mbuf;
387 		break;
388 	case MCLBYTES:
389 		zone = zone_clust;
390 		break;
391 #if MJUMPAGESIZE != MCLBYTES
392 	case MJUMPAGESIZE:
393 		zone = zone_jumbop;
394 		break;
395 #endif
396 	case MJUM9BYTES:
397 		zone = zone_jumbo9;
398 		break;
399 	case MJUM16BYTES:
400 		zone = zone_jumbo16;
401 		break;
402 	default:
403 		panic("%s: m_getjcl: invalid cluster type", __func__);
404 	}
405 
406 	return (zone);
407 }
408 
409 static __inline struct mbuf *
410 m_get(int how, short type)
411 {
412 	struct mb_args args;
413 
414 	args.flags = 0;
415 	args.type = type;
416 	return ((struct mbuf *)(uma_zalloc_arg(zone_mbuf, &args, how)));
417 }
418 
419 /*
420  * XXX This should be deprecated, very little use.
421  */
422 static __inline struct mbuf *
423 m_getclr(int how, short type)
424 {
425 	struct mbuf *m;
426 	struct mb_args args;
427 
428 	args.flags = 0;
429 	args.type = type;
430 	m = uma_zalloc_arg(zone_mbuf, &args, how);
431 	if (m != NULL)
432 		bzero(m->m_data, MLEN);
433 	return (m);
434 }
435 
436 static __inline struct mbuf *
437 m_gethdr(int how, short type)
438 {
439 	struct mb_args args;
440 
441 	args.flags = M_PKTHDR;
442 	args.type = type;
443 	return ((struct mbuf *)(uma_zalloc_arg(zone_mbuf, &args, how)));
444 }
445 
446 static __inline struct mbuf *
447 m_getcl(int how, short type, int flags)
448 {
449 	struct mb_args args;
450 
451 	args.flags = flags;
452 	args.type = type;
453 	return ((struct mbuf *)(uma_zalloc_arg(zone_pack, &args, how)));
454 }
455 
456 /*
457  * m_getjcl() returns an mbuf with a cluster of the specified size attached.
458  * For size it takes MCLBYTES, MJUMPAGESIZE, MJUM9BYTES, MJUM16BYTES.
459  *
460  * XXX: This is rather large, should be real function maybe.
461  */
462 static __inline struct mbuf *
463 m_getjcl(int how, short type, int flags, int size)
464 {
465 	struct mb_args args;
466 	struct mbuf *m, *n;
467 	uma_zone_t zone;
468 
469 	args.flags = flags;
470 	args.type = type;
471 
472 	m = uma_zalloc_arg(zone_mbuf, &args, how);
473 	if (m == NULL)
474 		return (NULL);
475 
476 	zone = m_getzone(size);
477 	n = uma_zalloc_arg(zone, m, how);
478 	if (n == NULL) {
479 		uma_zfree(zone_mbuf, m);
480 		return (NULL);
481 	}
482 	return (m);
483 }
484 
485 static __inline struct mbuf *
486 m_free(struct mbuf *m)
487 {
488 	struct mbuf *n = m->m_next;
489 
490 	if (m->m_flags & M_EXT)
491 		mb_free_ext(m);
492 	else
493 		uma_zfree(zone_mbuf, m);
494 	return (n);
495 }
496 
497 static __inline void
498 m_clget(struct mbuf *m, int how)
499 {
500 
501 	if (m->m_flags & M_EXT)
502 		printf("%s: %p mbuf already has cluster\n", __func__, m);
503 	m->m_ext.ext_buf = (char *)NULL;
504 	uma_zalloc_arg(zone_clust, m, how);
505 	/*
506 	 * On a cluster allocation failure, drain the packet zone and retry,
507 	 * we might be able to loosen a few clusters up on the drain.
508 	 */
509 	if ((how & M_NOWAIT) && (m->m_ext.ext_buf == NULL)) {
510 		zone_drain(zone_pack);
511 		uma_zalloc_arg(zone_clust, m, how);
512 	}
513 }
514 
515 /*
516  * m_cljget() is different from m_clget() as it can allocate clusters without
517  * attaching them to an mbuf.  In that case the return value is the pointer
518  * to the cluster of the requested size.  If an mbuf was specified, it gets
519  * the cluster attached to it and the return value can be safely ignored.
520  * For size it takes MCLBYTES, MJUMPAGESIZE, MJUM9BYTES, MJUM16BYTES.
521  */
522 static __inline void *
523 m_cljget(struct mbuf *m, int how, int size)
524 {
525 	uma_zone_t zone;
526 
527 	if (m && m->m_flags & M_EXT)
528 		printf("%s: %p mbuf already has cluster\n", __func__, m);
529 	if (m != NULL)
530 		m->m_ext.ext_buf = NULL;
531 
532 	zone = m_getzone(size);
533 	return (uma_zalloc_arg(zone, m, how));
534 }
535 
536 static __inline void
537 m_cljset(struct mbuf *m, void *cl, int type)
538 {
539 	uma_zone_t zone;
540 	int size;
541 
542 	switch (type) {
543 	case EXT_CLUSTER:
544 		size = MCLBYTES;
545 		zone = zone_clust;
546 		break;
547 #if MJUMPAGESIZE != MCLBYTES
548 	case EXT_JUMBOP:
549 		size = MJUMPAGESIZE;
550 		zone = zone_jumbop;
551 		break;
552 #endif
553 	case EXT_JUMBO9:
554 		size = MJUM9BYTES;
555 		zone = zone_jumbo9;
556 		break;
557 	case EXT_JUMBO16:
558 		size = MJUM16BYTES;
559 		zone = zone_jumbo16;
560 		break;
561 	default:
562 		panic("unknown cluster type");
563 		break;
564 	}
565 
566 	m->m_data = m->m_ext.ext_buf = cl;
567 	m->m_ext.ext_free = m->m_ext.ext_args = NULL;
568 	m->m_ext.ext_size = size;
569 	m->m_ext.ext_type = type;
570 	m->m_ext.ref_cnt = uma_find_refcnt(zone, cl);
571 	m->m_flags |= M_EXT;
572 
573 }
574 
575 static __inline void
576 m_chtype(struct mbuf *m, short new_type)
577 {
578 
579 	m->m_type = new_type;
580 }
581 
582 /*
583  * mbuf, cluster, and external object allocation macros (for compatibility
584  * purposes).
585  */
586 #define	M_MOVE_PKTHDR(to, from)	m_move_pkthdr((to), (from))
587 #define	MGET(m, how, type)	((m) = m_get((how), (type)))
588 #define	MGETHDR(m, how, type)	((m) = m_gethdr((how), (type)))
589 #define	MCLGET(m, how)		m_clget((m), (how))
590 #define	MEXTADD(m, buf, size, free, args, flags, type) 			\
591     m_extadd((m), (caddr_t)(buf), (size), (free), (args), (flags), (type))
592 #define	m_getm(m, len, how, type)					\
593     m_getm2((m), (len), (how), (type), M_PKTHDR)
594 
595 /*
596  * Evaluate TRUE if it's safe to write to the mbuf m's data region (this can
597  * be both the local data payload, or an external buffer area, depending on
598  * whether M_EXT is set).
599  */
600 #define	M_WRITABLE(m)	(!((m)->m_flags & M_RDONLY) &&			\
601 			 (!(((m)->m_flags & M_EXT)) ||			\
602 			 (*((m)->m_ext.ref_cnt) == 1)) )		\
603 
604 /* Check if the supplied mbuf has a packet header, or else panic. */
605 #define	M_ASSERTPKTHDR(m)						\
606 	KASSERT(m != NULL && m->m_flags & M_PKTHDR,			\
607 	    ("%s: no mbuf packet header!", __func__))
608 
609 /*
610  * Ensure that the supplied mbuf is a valid, non-free mbuf.
611  *
612  * XXX: Broken at the moment.  Need some UMA magic to make it work again.
613  */
614 #define	M_ASSERTVALID(m)						\
615 	KASSERT((((struct mbuf *)m)->m_flags & 0) == 0,			\
616 	    ("%s: attempted use of a free mbuf!", __func__))
617 
618 /*
619  * Set the m_data pointer of a newly-allocated mbuf (m_get/MGET) to place an
620  * object of the specified size at the end of the mbuf, longword aligned.
621  */
622 #define	M_ALIGN(m, len) do {						\
623 	KASSERT(!((m)->m_flags & (M_PKTHDR|M_EXT)),			\
624 		("%s: M_ALIGN not normal mbuf", __func__));		\
625 	KASSERT((m)->m_data == (m)->m_dat,				\
626 		("%s: M_ALIGN not a virgin mbuf", __func__));		\
627 	(m)->m_data += (MLEN - (len)) & ~(sizeof(long) - 1);		\
628 } while (0)
629 
630 /*
631  * As above, for mbufs allocated with m_gethdr/MGETHDR or initialized by
632  * M_DUP/MOVE_PKTHDR.
633  */
634 #define	MH_ALIGN(m, len) do {						\
635 	KASSERT((m)->m_flags & M_PKTHDR && !((m)->m_flags & M_EXT),	\
636 		("%s: MH_ALIGN not PKTHDR mbuf", __func__));		\
637 	KASSERT((m)->m_data == (m)->m_pktdat,				\
638 		("%s: MH_ALIGN not a virgin mbuf", __func__));		\
639 	(m)->m_data += (MHLEN - (len)) & ~(sizeof(long) - 1);		\
640 } while (0)
641 
642 /*
643  * Compute the amount of space available before the current start of data in
644  * an mbuf.
645  *
646  * The M_WRITABLE() is a temporary, conservative safety measure: the burden
647  * of checking writability of the mbuf data area rests solely with the caller.
648  */
649 #define	M_LEADINGSPACE(m)						\
650 	((m)->m_flags & M_EXT ?						\
651 	    (M_WRITABLE(m) ? (m)->m_data - (m)->m_ext.ext_buf : 0):	\
652 	    (m)->m_flags & M_PKTHDR ? (m)->m_data - (m)->m_pktdat :	\
653 	    (m)->m_data - (m)->m_dat)
654 
655 /*
656  * Compute the amount of space available after the end of data in an mbuf.
657  *
658  * The M_WRITABLE() is a temporary, conservative safety measure: the burden
659  * of checking writability of the mbuf data area rests solely with the caller.
660  */
661 #define	M_TRAILINGSPACE(m)						\
662 	((m)->m_flags & M_EXT ?						\
663 	    (M_WRITABLE(m) ? (m)->m_ext.ext_buf + (m)->m_ext.ext_size	\
664 		- ((m)->m_data + (m)->m_len) : 0) :			\
665 	    &(m)->m_dat[MLEN] - ((m)->m_data + (m)->m_len))
666 
667 /*
668  * Arrange to prepend space of size plen to mbuf m.  If a new mbuf must be
669  * allocated, how specifies whether to wait.  If the allocation fails, the
670  * original mbuf chain is freed and m is set to NULL.
671  */
672 #define	M_PREPEND(m, plen, how) do {					\
673 	struct mbuf **_mmp = &(m);					\
674 	struct mbuf *_mm = *_mmp;					\
675 	int _mplen = (plen);						\
676 	int __mhow = (how);						\
677 									\
678 	MBUF_CHECKSLEEP(how);						\
679 	if (M_LEADINGSPACE(_mm) >= _mplen) {				\
680 		_mm->m_data -= _mplen;					\
681 		_mm->m_len += _mplen;					\
682 	} else								\
683 		_mm = m_prepend(_mm, _mplen, __mhow);			\
684 	if (_mm != NULL && _mm->m_flags & M_PKTHDR)			\
685 		_mm->m_pkthdr.len += _mplen;				\
686 	*_mmp = _mm;							\
687 } while (0)
688 
689 /*
690  * Change mbuf to new type.  This is a relatively expensive operation and
691  * should be avoided.
692  */
693 #define	MCHTYPE(m, t)	m_chtype((m), (t))
694 
695 /* Length to m_copy to copy all. */
696 #define	M_COPYALL	1000000000
697 
698 /* Compatibility with 4.3. */
699 #define	m_copy(m, o, l)	m_copym((m), (o), (l), M_DONTWAIT)
700 
701 extern int		max_datalen;	/* MHLEN - max_hdr */
702 extern int		max_hdr;	/* Largest link + protocol header */
703 extern int		max_linkhdr;	/* Largest link-level header */
704 extern int		max_protohdr;	/* Largest protocol header */
705 extern struct mbstat	mbstat;		/* General mbuf stats/infos */
706 extern int		nmbclusters;	/* Maximum number of clusters */
707 
708 struct uio;
709 
710 void		 m_adj(struct mbuf *, int);
711 void		 m_align(struct mbuf *, int);
712 int		 m_apply(struct mbuf *, int, int,
713 		    int (*)(void *, void *, u_int), void *);
714 int		 m_append(struct mbuf *, int, c_caddr_t);
715 void		 m_cat(struct mbuf *, struct mbuf *);
716 void		 m_extadd(struct mbuf *, caddr_t, u_int,
717 		    void (*)(void *, void *), void *, int, int);
718 void		 m_copyback(struct mbuf *, int, int, c_caddr_t);
719 void		 m_copydata(const struct mbuf *, int, int, caddr_t);
720 struct mbuf	*m_copym(struct mbuf *, int, int, int);
721 struct mbuf	*m_copymdata(struct mbuf *, struct mbuf *,
722 		    int, int, int, int);
723 struct mbuf	*m_copypacket(struct mbuf *, int);
724 void		 m_copy_pkthdr(struct mbuf *, struct mbuf *);
725 struct mbuf	*m_copyup(struct mbuf *n, int len, int dstoff);
726 struct mbuf	*m_defrag(struct mbuf *, int);
727 void		 m_demote(struct mbuf *, int);
728 struct mbuf	*m_devget(char *, int, int, struct ifnet *,
729 		    void (*)(char *, caddr_t, u_int));
730 struct mbuf	*m_dup(struct mbuf *, int);
731 int		 m_dup_pkthdr(struct mbuf *, struct mbuf *, int);
732 u_int		 m_fixhdr(struct mbuf *);
733 struct mbuf	*m_fragment(struct mbuf *, int, int);
734 void		 m_freem(struct mbuf *);
735 struct mbuf	*m_getm2(struct mbuf *, int, int, short, int);
736 struct mbuf	*m_getptr(struct mbuf *, int, int *);
737 u_int		 m_length(struct mbuf *, struct mbuf **);
738 void		 m_move_pkthdr(struct mbuf *, struct mbuf *);
739 struct mbuf	*m_prepend(struct mbuf *, int, int);
740 void		 m_print(const struct mbuf *, int);
741 struct mbuf	*m_pulldown(struct mbuf *, int, int, int *);
742 struct mbuf	*m_pullup(struct mbuf *, int);
743 int		m_sanity(struct mbuf *, int);
744 struct mbuf	*m_split(struct mbuf *, int, int);
745 struct mbuf	*m_uiotombuf(struct uio *, int, int, int, int);
746 struct mbuf	*m_unshare(struct mbuf *, int how);
747 
748 /*-
749  * Network packets may have annotations attached by affixing a list of
750  * "packet tags" to the pkthdr structure.  Packet tags are dynamically
751  * allocated semi-opaque data structures that have a fixed header
752  * (struct m_tag) that specifies the size of the memory block and a
753  * <cookie,type> pair that identifies it.  The cookie is a 32-bit unique
754  * unsigned value used to identify a module or ABI.  By convention this value
755  * is chosen as the date+time that the module is created, expressed as the
756  * number of seconds since the epoch (e.g., using date -u +'%s').  The type
757  * value is an ABI/module-specific value that identifies a particular
758  * annotation and is private to the module.  For compatibility with systems
759  * like OpenBSD that define packet tags w/o an ABI/module cookie, the value
760  * PACKET_ABI_COMPAT is used to implement m_tag_get and m_tag_find
761  * compatibility shim functions and several tag types are defined below.
762  * Users that do not require compatibility should use a private cookie value
763  * so that packet tag-related definitions can be maintained privately.
764  *
765  * Note that the packet tag returned by m_tag_alloc has the default memory
766  * alignment implemented by malloc.  To reference private data one can use a
767  * construct like:
768  *
769  *	struct m_tag *mtag = m_tag_alloc(...);
770  *	struct foo *p = (struct foo *)(mtag+1);
771  *
772  * if the alignment of struct m_tag is sufficient for referencing members of
773  * struct foo.  Otherwise it is necessary to embed struct m_tag within the
774  * private data structure to insure proper alignment; e.g.,
775  *
776  *	struct foo {
777  *		struct m_tag	tag;
778  *		...
779  *	};
780  *	struct foo *p = (struct foo *) m_tag_alloc(...);
781  *	struct m_tag *mtag = &p->tag;
782  */
783 
784 /*
785  * Persistent tags stay with an mbuf until the mbuf is reclaimed.  Otherwise
786  * tags are expected to ``vanish'' when they pass through a network
787  * interface.  For most interfaces this happens normally as the tags are
788  * reclaimed when the mbuf is free'd.  However in some special cases
789  * reclaiming must be done manually.  An example is packets that pass through
790  * the loopback interface.  Also, one must be careful to do this when
791  * ``turning around'' packets (e.g., icmp_reflect).
792  *
793  * To mark a tag persistent bit-or this flag in when defining the tag id.
794  * The tag will then be treated as described above.
795  */
796 #define	MTAG_PERSISTENT				0x800
797 
798 #define	PACKET_TAG_NONE				0  /* Nadda */
799 
800 /* Packet tags for use with PACKET_ABI_COMPAT. */
801 #define	PACKET_TAG_IPSEC_IN_DONE		1  /* IPsec applied, in */
802 #define	PACKET_TAG_IPSEC_OUT_DONE		2  /* IPsec applied, out */
803 #define	PACKET_TAG_IPSEC_IN_CRYPTO_DONE		3  /* NIC IPsec crypto done */
804 #define	PACKET_TAG_IPSEC_OUT_CRYPTO_NEEDED	4  /* NIC IPsec crypto req'ed */
805 #define	PACKET_TAG_IPSEC_IN_COULD_DO_CRYPTO	5  /* NIC notifies IPsec */
806 #define	PACKET_TAG_IPSEC_PENDING_TDB		6  /* Reminder to do IPsec */
807 #define	PACKET_TAG_BRIDGE			7  /* Bridge processing done */
808 #define	PACKET_TAG_GIF				8  /* GIF processing done */
809 #define	PACKET_TAG_GRE				9  /* GRE processing done */
810 #define	PACKET_TAG_IN_PACKET_CHECKSUM		10 /* NIC checksumming done */
811 #define	PACKET_TAG_ENCAP			11 /* Encap.  processing */
812 #define	PACKET_TAG_IPSEC_SOCKET			12 /* IPSEC socket ref */
813 #define	PACKET_TAG_IPSEC_HISTORY		13 /* IPSEC history */
814 #define	PACKET_TAG_IPV6_INPUT			14 /* IPV6 input processing */
815 #define	PACKET_TAG_DUMMYNET			15 /* dummynet info */
816 #define	PACKET_TAG_DIVERT			17 /* divert info */
817 #define	PACKET_TAG_IPFORWARD			18 /* ipforward info */
818 #define	PACKET_TAG_MACLABEL	(19 | MTAG_PERSISTENT) /* MAC label */
819 #define	PACKET_TAG_PF_ROUTED			21 /* PF routed, avoid loops */
820 #define	PACKET_TAG_PF_FRAGCACHE			22 /* PF fragment cached */
821 #define	PACKET_TAG_PF_QID			23 /* PF ALTQ queue id */
822 #define	PACKET_TAG_PF_TAG			24 /* PF tagged */
823 #define	PACKET_TAG_RTSOCKFAM			25 /* rtsock sa family */
824 #define	PACKET_TAG_PF_TRANSLATE_LOCALHOST	26 /* PF translate localhost */
825 #define	PACKET_TAG_IPOPTIONS			27 /* Saved IP options */
826 #define	PACKET_TAG_CARP                         28 /* CARP info */
827 
828 /* Specific cookies and tags. */
829 
830 /* Packet tag routines. */
831 struct m_tag	*m_tag_alloc(u_int32_t, int, int, int);
832 void		 m_tag_delete(struct mbuf *, struct m_tag *);
833 void		 m_tag_delete_chain(struct mbuf *, struct m_tag *);
834 void		 m_tag_free_default(struct m_tag *);
835 struct m_tag	*m_tag_locate(struct mbuf *, u_int32_t, int, struct m_tag *);
836 struct m_tag	*m_tag_copy(struct m_tag *, int);
837 int		 m_tag_copy_chain(struct mbuf *, struct mbuf *, int);
838 void		 m_tag_delete_nonpersistent(struct mbuf *);
839 
840 /*
841  * Initialize the list of tags associated with an mbuf.
842  */
843 static __inline void
844 m_tag_init(struct mbuf *m)
845 {
846 
847 	SLIST_INIT(&m->m_pkthdr.tags);
848 }
849 
850 /*
851  * Set up the contents of a tag.  Note that this does not fill in the free
852  * method; the caller is expected to do that.
853  *
854  * XXX probably should be called m_tag_init, but that was already taken.
855  */
856 static __inline void
857 m_tag_setup(struct m_tag *t, u_int32_t cookie, int type, int len)
858 {
859 
860 	t->m_tag_id = type;
861 	t->m_tag_len = len;
862 	t->m_tag_cookie = cookie;
863 }
864 
865 /*
866  * Reclaim resources associated with a tag.
867  */
868 static __inline void
869 m_tag_free(struct m_tag *t)
870 {
871 
872 	(*t->m_tag_free)(t);
873 }
874 
875 /*
876  * Return the first tag associated with an mbuf.
877  */
878 static __inline struct m_tag *
879 m_tag_first(struct mbuf *m)
880 {
881 
882 	return (SLIST_FIRST(&m->m_pkthdr.tags));
883 }
884 
885 /*
886  * Return the next tag in the list of tags associated with an mbuf.
887  */
888 static __inline struct m_tag *
889 m_tag_next(struct mbuf *m, struct m_tag *t)
890 {
891 
892 	return (SLIST_NEXT(t, m_tag_link));
893 }
894 
895 /*
896  * Prepend a tag to the list of tags associated with an mbuf.
897  */
898 static __inline void
899 m_tag_prepend(struct mbuf *m, struct m_tag *t)
900 {
901 
902 	SLIST_INSERT_HEAD(&m->m_pkthdr.tags, t, m_tag_link);
903 }
904 
905 /*
906  * Unlink a tag from the list of tags associated with an mbuf.
907  */
908 static __inline void
909 m_tag_unlink(struct mbuf *m, struct m_tag *t)
910 {
911 
912 	SLIST_REMOVE(&m->m_pkthdr.tags, t, m_tag, m_tag_link);
913 }
914 
915 /* These are for OpenBSD compatibility. */
916 #define	MTAG_ABI_COMPAT		0		/* compatibility ABI */
917 
918 static __inline struct m_tag *
919 m_tag_get(int type, int length, int wait)
920 {
921 	return (m_tag_alloc(MTAG_ABI_COMPAT, type, length, wait));
922 }
923 
924 static __inline struct m_tag *
925 m_tag_find(struct mbuf *m, int type, struct m_tag *start)
926 {
927 	return (SLIST_EMPTY(&m->m_pkthdr.tags) ? (struct m_tag *)NULL :
928 	    m_tag_locate(m, MTAG_ABI_COMPAT, type, start));
929 }
930 
931 #endif /* _KERNEL */
932 
933 #endif /* !_SYS_MBUF_H_ */
934