xref: /freebsd/sys/sys/mbuf.h (revision a98ff317388a00b992f1bf8404dee596f9383f5e)
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  * MLEN is data length in a normal mbuf.
57  * MHLEN is data length in an mbuf with pktheader.
58  * MINCLSIZE is a smallest amount of data that should be put into cluster.
59  */
60 #define	MLEN		((int)(MSIZE - sizeof(struct m_hdr)))
61 #define	MHLEN		((int)(MLEN - sizeof(struct pkthdr)))
62 #define	MINCLSIZE	(MHLEN + 1)
63 
64 #ifdef _KERNEL
65 /*-
66  * Macro for type conversion: convert mbuf pointer to data pointer of correct
67  * type:
68  *
69  * mtod(m, t)	-- Convert mbuf pointer to data pointer of correct type.
70  */
71 #define	mtod(m, t)	((t)((m)->m_data))
72 
73 /*
74  * Argument structure passed to UMA routines during mbuf and packet
75  * allocations.
76  */
77 struct mb_args {
78 	int	flags;	/* Flags for mbuf being allocated */
79 	short	type;	/* Type of mbuf being allocated */
80 };
81 #endif /* _KERNEL */
82 
83 #if defined(__LP64__)
84 #define M_HDR_PAD    6
85 #else
86 #define M_HDR_PAD    2
87 #endif
88 
89 /*
90  * Header present at the beginning of every mbuf.
91  */
92 struct m_hdr {
93 	struct mbuf	*mh_next;	/* next buffer in chain */
94 	struct mbuf	*mh_nextpkt;	/* next chain in queue/record */
95 	caddr_t		 mh_data;	/* location of data */
96 	int		 mh_len;	/* amount of data in this mbuf */
97 	int		 mh_flags;	/* flags; see below */
98 	short		 mh_type;	/* type of data in this mbuf */
99 	uint8_t          pad[M_HDR_PAD];/* word align                  */
100 };
101 
102 /*
103  * Packet tag structure (see below for details).
104  */
105 struct m_tag {
106 	SLIST_ENTRY(m_tag)	m_tag_link;	/* List of packet tags */
107 	u_int16_t		m_tag_id;	/* Tag ID */
108 	u_int16_t		m_tag_len;	/* Length of data */
109 	u_int32_t		m_tag_cookie;	/* ABI/Module ID */
110 	void			(*m_tag_free)(struct m_tag *);
111 };
112 
113 /*
114  * Record/packet header in first mbuf of chain; valid only if M_PKTHDR is set.
115  */
116 struct pkthdr {
117 	struct ifnet	*rcvif;		/* rcv interface */
118 	/* variables for ip and tcp reassembly */
119 	void		*header;	/* pointer to packet header */
120 	int		 len;		/* total packet length */
121 	uint32_t	 flowid;	/* packet's 4-tuple system
122 					 * flow identifier
123 					 */
124 	/* variables for hardware checksum */
125 	int		 csum_flags;	/* flags regarding checksum */
126 	int		 csum_data;	/* data field used by csum routines */
127 	u_int16_t	 tso_segsz;	/* TSO segment size */
128 	union {
129 		u_int16_t vt_vtag;	/* Ethernet 802.1p+q vlan tag */
130 		u_int16_t vt_nrecs;	/* # of IGMPv3 records in this chain */
131 	} PH_vt;
132 	u_int16_t	 fibnum;	/* this packet should use this fib */
133 	u_int16_t	 pad2;		/* align to 32 bits */
134 	SLIST_HEAD(packet_tags, m_tag) tags; /* list of packet tags */
135 };
136 #define ether_vtag	PH_vt.vt_vtag
137 
138 /*
139  * Description of external storage mapped into mbuf; valid only if M_EXT is
140  * set.
141  */
142 struct m_ext {
143 	caddr_t		 ext_buf;	/* start of buffer */
144 	void		(*ext_free)	/* free routine if not the usual */
145 			    (void *, void *);
146 	void		*ext_arg1;	/* optional argument pointer */
147 	void		*ext_arg2;	/* optional argument pointer */
148 	u_int		 ext_size;	/* size of buffer, for ext_free */
149 	volatile u_int	*ref_cnt;	/* pointer to ref count info */
150 	int		 ext_type;	/* type of external storage */
151 };
152 
153 /*
154  * The core of the mbuf object along with some shortcut defines for practical
155  * purposes.
156  */
157 struct mbuf {
158 	struct m_hdr	m_hdr;
159 	union {
160 		struct {
161 			struct pkthdr	MH_pkthdr;	/* M_PKTHDR set */
162 			union {
163 				struct m_ext	MH_ext;	/* M_EXT set */
164 				char		MH_databuf[MHLEN];
165 			} MH_dat;
166 		} MH;
167 		char	M_databuf[MLEN];		/* !M_PKTHDR, !M_EXT */
168 	} M_dat;
169 };
170 #define	m_next		m_hdr.mh_next
171 #define	m_len		m_hdr.mh_len
172 #define	m_data		m_hdr.mh_data
173 #define	m_type		m_hdr.mh_type
174 #define	m_flags		m_hdr.mh_flags
175 #define	m_nextpkt	m_hdr.mh_nextpkt
176 #define	m_act		m_nextpkt
177 #define	m_pkthdr	M_dat.MH.MH_pkthdr
178 #define	m_ext		M_dat.MH.MH_dat.MH_ext
179 #define	m_pktdat	M_dat.MH.MH_dat.MH_databuf
180 #define	m_dat		M_dat.M_databuf
181 
182 /*
183  * mbuf flags.
184  */
185 #define	M_EXT		0x00000001 /* has associated external storage */
186 #define	M_PKTHDR	0x00000002 /* start of record */
187 #define	M_EOR		0x00000004 /* end of record */
188 #define	M_RDONLY	0x00000008 /* associated data is marked read-only */
189 #define	M_PROTO1	0x00000010 /* protocol-specific */
190 #define	M_PROTO2	0x00000020 /* protocol-specific */
191 #define	M_PROTO3	0x00000040 /* protocol-specific */
192 #define	M_PROTO4	0x00000080 /* protocol-specific */
193 #define	M_PROTO5	0x00000100 /* protocol-specific */
194 #define	M_BCAST		0x00000200 /* send/received as link-level broadcast */
195 #define	M_MCAST		0x00000400 /* send/received as link-level multicast */
196 #define	M_FRAG		0x00000800 /* packet is a fragment of a larger packet */
197 #define	M_FIRSTFRAG	0x00001000 /* packet is first fragment */
198 #define	M_LASTFRAG	0x00002000 /* packet is last fragment */
199 #define	M_SKIP_FIREWALL	0x00004000 /* skip firewall processing */
200 		     /*	0x00008000    free */
201 #define	M_VLANTAG	0x00010000 /* ether_vtag is valid */
202 #define	M_PROMISC	0x00020000 /* packet was not for us */
203 #define	M_NOFREE	0x00040000 /* do not free mbuf, embedded in cluster */
204 #define	M_PROTO6	0x00080000 /* protocol-specific */
205 #define	M_PROTO7	0x00100000 /* protocol-specific */
206 #define	M_PROTO8	0x00200000 /* protocol-specific */
207 #define	M_FLOWID	0x00400000 /* deprecated: flowid is valid */
208 #define	M_HASHTYPEBITS	0x0F000000 /* mask of bits holding flowid hash type */
209 
210 #define	M_NOTIFICATION	M_PROTO5    /* SCTP notification */
211 
212 /*
213  * Flags to purge when crossing layers.
214  */
215 #define	M_PROTOFLAGS \
216     (M_PROTO1|M_PROTO2|M_PROTO3|M_PROTO4|M_PROTO5|M_PROTO6|M_PROTO7|M_PROTO8)
217 
218 /*
219  * Network interface cards are able to hash protocol fields (such as IPv4
220  * addresses and TCP port numbers) classify packets into flows.  These flows
221  * can then be used to maintain ordering while delivering packets to the OS
222  * via parallel input queues, as well as to provide a stateless affinity
223  * model.  NIC drivers can pass up the hash via m->m_pkthdr.flowid, and set
224  * m_flag fields to indicate how the hash should be interpreted by the
225  * network stack.
226  *
227  * Most NICs support RSS, which provides ordering and explicit affinity, and
228  * use the hash m_flag bits to indicate what header fields were covered by
229  * the hash.  M_HASHTYPE_OPAQUE can be set by non-RSS cards or configurations
230  * that provide an opaque flow identifier, allowing for ordering and
231  * distribution without explicit affinity.
232  */
233 #define	M_HASHTYPE_SHIFT		24
234 #define	M_HASHTYPE_NONE			0x0
235 #define	M_HASHTYPE_RSS_IPV4		0x1	/* IPv4 2-tuple */
236 #define	M_HASHTYPE_RSS_TCP_IPV4		0x2	/* TCPv4 4-tuple */
237 #define	M_HASHTYPE_RSS_IPV6		0x3	/* IPv6 2-tuple */
238 #define	M_HASHTYPE_RSS_TCP_IPV6		0x4	/* TCPv6 4-tuple */
239 #define	M_HASHTYPE_RSS_IPV6_EX		0x5	/* IPv6 2-tuple + ext hdrs */
240 #define	M_HASHTYPE_RSS_TCP_IPV6_EX	0x6	/* TCPv6 4-tiple + ext hdrs */
241 #define	M_HASHTYPE_OPAQUE		0xf	/* ordering, not affinity */
242 
243 #define	M_HASHTYPE_CLEAR(m)	(m)->m_flags &= ~(M_HASHTYPEBITS)
244 #define	M_HASHTYPE_GET(m)	(((m)->m_flags & M_HASHTYPEBITS) >> \
245 				    M_HASHTYPE_SHIFT)
246 #define	M_HASHTYPE_SET(m, v)	do {					\
247 	(m)->m_flags &= ~M_HASHTYPEBITS;				\
248 	(m)->m_flags |= ((v) << M_HASHTYPE_SHIFT);			\
249 } while (0)
250 #define	M_HASHTYPE_TEST(m, v)	(M_HASHTYPE_GET(m) == (v))
251 
252 /*
253  * Flags preserved when copying m_pkthdr.
254  */
255 #define	M_COPYFLAGS \
256     (M_PKTHDR|M_EOR|M_RDONLY|M_PROTOFLAGS|M_SKIP_FIREWALL|M_BCAST|M_MCAST|\
257      M_FRAG|M_FIRSTFRAG|M_LASTFRAG|M_VLANTAG|M_PROMISC|M_HASHTYPEBITS)
258 
259 /*
260  * External buffer types: identify ext_buf type.
261  */
262 #define	EXT_CLUSTER	1	/* mbuf cluster */
263 #define	EXT_SFBUF	2	/* sendfile(2)'s sf_bufs */
264 #define	EXT_JUMBOP	3	/* jumbo cluster 4096 bytes */
265 #define	EXT_JUMBO9	4	/* jumbo cluster 9216 bytes */
266 #define	EXT_JUMBO16	5	/* jumbo cluster 16184 bytes */
267 #define	EXT_PACKET	6	/* mbuf+cluster from packet zone */
268 #define	EXT_MBUF	7	/* external mbuf reference (M_IOVEC) */
269 #define	EXT_NET_DRV	100	/* custom ext_buf provided by net driver(s) */
270 #define	EXT_MOD_TYPE	200	/* custom module's ext_buf type */
271 #define	EXT_DISPOSABLE	300	/* can throw this buffer away w/page flipping */
272 #define	EXT_EXTREF	400	/* has externally maintained ref_cnt ptr */
273 
274 /*
275  * Flags indicating hw checksum support and sw checksum requirements.  This
276  * field can be directly tested against if_data.ifi_hwassist.
277  */
278 #define	CSUM_IP			0x0001		/* will csum IP */
279 #define	CSUM_TCP		0x0002		/* will csum TCP */
280 #define	CSUM_UDP		0x0004		/* will csum UDP */
281 #define	CSUM_FRAGMENT		0x0010		/* will do IP fragmentation */
282 #define	CSUM_TSO		0x0020		/* will do TSO */
283 #define	CSUM_SCTP		0x0040		/* will csum SCTP */
284 #define CSUM_SCTP_IPV6		0x0080		/* will csum IPv6/SCTP */
285 
286 #define	CSUM_IP_CHECKED		0x0100		/* did csum IP */
287 #define	CSUM_IP_VALID		0x0200		/*   ... the csum is valid */
288 #define	CSUM_DATA_VALID		0x0400		/* csum_data field is valid */
289 #define	CSUM_PSEUDO_HDR		0x0800		/* csum_data has pseudo hdr */
290 #define	CSUM_SCTP_VALID		0x1000		/* SCTP checksum is valid */
291 #define	CSUM_UDP_IPV6		0x2000		/* will csum IPv6/UDP */
292 #define	CSUM_TCP_IPV6		0x4000		/* will csum IPv6/TCP */
293 /*	CSUM_TSO_IPV6		0x8000		will do IPv6/TSO */
294 
295 /*	CSUM_FRAGMENT_IPV6	0x10000		will do IPv6 fragementation */
296 
297 #define	CSUM_DELAY_DATA_IPV6	(CSUM_TCP_IPV6 | CSUM_UDP_IPV6)
298 #define	CSUM_DATA_VALID_IPV6	CSUM_DATA_VALID
299 
300 #define	CSUM_DELAY_DATA		(CSUM_TCP | CSUM_UDP)
301 #define	CSUM_DELAY_IP		(CSUM_IP)	/* Only v4, no v6 IP hdr csum */
302 
303 /*
304  * mbuf types.
305  */
306 #define	MT_NOTMBUF	0	/* USED INTERNALLY ONLY! Object is not mbuf */
307 #define	MT_DATA		1	/* dynamic (data) allocation */
308 #define	MT_HEADER	MT_DATA	/* packet header, use M_PKTHDR instead */
309 #define	MT_SONAME	8	/* socket name */
310 #define	MT_CONTROL	14	/* extra-data protocol message */
311 #define	MT_OOBDATA	15	/* expedited data  */
312 #define	MT_NTYPES	16	/* number of mbuf types for mbtypes[] */
313 
314 #define	MT_NOINIT	255	/* Not a type but a flag to allocate
315 				   a non-initialized mbuf */
316 
317 #define MB_NOTAGS	0x1UL	/* no tags attached to mbuf */
318 
319 /*
320  * Compatibility with historic mbuf allocator.
321  */
322 #define	MBTOM(how)	(how)
323 #define	M_DONTWAIT	M_NOWAIT
324 #define	M_TRYWAIT	M_WAITOK
325 #define	M_WAIT		M_WAITOK
326 
327 /*
328  * String names of mbuf-related UMA(9) and malloc(9) types.  Exposed to
329  * !_KERNEL so that monitoring tools can look up the zones with
330  * libmemstat(3).
331  */
332 #define	MBUF_MEM_NAME		"mbuf"
333 #define	MBUF_CLUSTER_MEM_NAME	"mbuf_cluster"
334 #define	MBUF_PACKET_MEM_NAME	"mbuf_packet"
335 #define	MBUF_JUMBOP_MEM_NAME	"mbuf_jumbo_page"
336 #define	MBUF_JUMBO9_MEM_NAME	"mbuf_jumbo_9k"
337 #define	MBUF_JUMBO16_MEM_NAME	"mbuf_jumbo_16k"
338 #define	MBUF_TAG_MEM_NAME	"mbuf_tag"
339 #define	MBUF_EXTREFCNT_MEM_NAME	"mbuf_ext_refcnt"
340 
341 #ifdef _KERNEL
342 
343 #ifdef WITNESS
344 #define	MBUF_CHECKSLEEP(how) do {					\
345 	if (how == M_WAITOK)						\
346 		WITNESS_WARN(WARN_GIANTOK | WARN_SLEEPOK, NULL,		\
347 		    "Sleeping in \"%s\"", __func__);			\
348 } while (0)
349 #else
350 #define	MBUF_CHECKSLEEP(how)
351 #endif
352 
353 /*
354  * Network buffer allocation API
355  *
356  * The rest of it is defined in kern/kern_mbuf.c
357  */
358 extern uma_zone_t	zone_mbuf;
359 extern uma_zone_t	zone_clust;
360 extern uma_zone_t	zone_pack;
361 extern uma_zone_t	zone_jumbop;
362 extern uma_zone_t	zone_jumbo9;
363 extern uma_zone_t	zone_jumbo16;
364 extern uma_zone_t	zone_ext_refcnt;
365 
366 void		 mb_free_ext(struct mbuf *);
367 int		 m_pkthdr_init(struct mbuf *, int);
368 
369 static __inline int
370 m_gettype(int size)
371 {
372 	int type;
373 
374 	switch (size) {
375 	case MSIZE:
376 		type = EXT_MBUF;
377 		break;
378 	case MCLBYTES:
379 		type = EXT_CLUSTER;
380 		break;
381 #if MJUMPAGESIZE != MCLBYTES
382 	case MJUMPAGESIZE:
383 		type = EXT_JUMBOP;
384 		break;
385 #endif
386 	case MJUM9BYTES:
387 		type = EXT_JUMBO9;
388 		break;
389 	case MJUM16BYTES:
390 		type = EXT_JUMBO16;
391 		break;
392 	default:
393 		panic("%s: invalid cluster size", __func__);
394 	}
395 
396 	return (type);
397 }
398 
399 static __inline uma_zone_t
400 m_getzone(int size)
401 {
402 	uma_zone_t zone;
403 
404 	switch (size) {
405 	case MCLBYTES:
406 		zone = zone_clust;
407 		break;
408 #if MJUMPAGESIZE != MCLBYTES
409 	case MJUMPAGESIZE:
410 		zone = zone_jumbop;
411 		break;
412 #endif
413 	case MJUM9BYTES:
414 		zone = zone_jumbo9;
415 		break;
416 	case MJUM16BYTES:
417 		zone = zone_jumbo16;
418 		break;
419 	default:
420 		panic("%s: invalid cluster size", __func__);
421 	}
422 
423 	return (zone);
424 }
425 
426 /*
427  * Initialize an mbuf with linear storage.
428  *
429  * Inline because the consumer text overhead will be roughly the same to
430  * initialize or call a function with this many parameters and M_PKTHDR
431  * should go away with constant propagation for !MGETHDR.
432  */
433 static __inline int
434 m_init(struct mbuf *m, uma_zone_t zone, int size, int how, short type,
435     int flags)
436 {
437 	int error;
438 
439 	m->m_next = NULL;
440 	m->m_nextpkt = NULL;
441 	m->m_data = m->m_dat;
442 	m->m_len = 0;
443 	m->m_flags = flags;
444 	m->m_type = type;
445 	if (flags & M_PKTHDR) {
446 		if ((error = m_pkthdr_init(m, how)) != 0)
447 			return (error);
448 	}
449 
450 	return (0);
451 }
452 
453 static __inline struct mbuf *
454 m_get(int how, short type)
455 {
456 	struct mb_args args;
457 
458 	args.flags = 0;
459 	args.type = type;
460 	return (uma_zalloc_arg(zone_mbuf, &args, how));
461 }
462 
463 /*
464  * XXX This should be deprecated, very little use.
465  */
466 static __inline struct mbuf *
467 m_getclr(int how, short type)
468 {
469 	struct mbuf *m;
470 	struct mb_args args;
471 
472 	args.flags = 0;
473 	args.type = type;
474 	m = uma_zalloc_arg(zone_mbuf, &args, how);
475 	if (m != NULL)
476 		bzero(m->m_data, MLEN);
477 	return (m);
478 }
479 
480 static __inline struct mbuf *
481 m_gethdr(int how, short type)
482 {
483 	struct mb_args args;
484 
485 	args.flags = M_PKTHDR;
486 	args.type = type;
487 	return (uma_zalloc_arg(zone_mbuf, &args, how));
488 }
489 
490 static __inline struct mbuf *
491 m_getcl(int how, short type, int flags)
492 {
493 	struct mb_args args;
494 
495 	args.flags = flags;
496 	args.type = type;
497 	return (uma_zalloc_arg(zone_pack, &args, how));
498 }
499 
500 static __inline void
501 m_free_fast(struct mbuf *m)
502 {
503 #ifdef INVARIANTS
504 	if (m->m_flags & M_PKTHDR)
505 		KASSERT(SLIST_EMPTY(&m->m_pkthdr.tags), ("doing fast free of mbuf with tags"));
506 #endif
507 
508 	uma_zfree_arg(zone_mbuf, m, (void *)MB_NOTAGS);
509 }
510 
511 static __inline struct mbuf *
512 m_free(struct mbuf *m)
513 {
514 	struct mbuf *n = m->m_next;
515 
516 	if (m->m_flags & M_EXT)
517 		mb_free_ext(m);
518 	else if ((m->m_flags & M_NOFREE) == 0)
519 		uma_zfree(zone_mbuf, m);
520 	return (n);
521 }
522 
523 static __inline void
524 m_clget(struct mbuf *m, int how)
525 {
526 
527 	if (m->m_flags & M_EXT)
528 		printf("%s: %p mbuf already has cluster\n", __func__, m);
529 	m->m_ext.ext_buf = (char *)NULL;
530 	uma_zalloc_arg(zone_clust, m, how);
531 	/*
532 	 * On a cluster allocation failure, drain the packet zone and retry,
533 	 * we might be able to loosen a few clusters up on the drain.
534 	 */
535 	if ((how & M_NOWAIT) && (m->m_ext.ext_buf == NULL)) {
536 		zone_drain(zone_pack);
537 		uma_zalloc_arg(zone_clust, m, how);
538 	}
539 }
540 
541 /*
542  * m_cljget() is different from m_clget() as it can allocate clusters without
543  * attaching them to an mbuf.  In that case the return value is the pointer
544  * to the cluster of the requested size.  If an mbuf was specified, it gets
545  * the cluster attached to it and the return value can be safely ignored.
546  * For size it takes MCLBYTES, MJUMPAGESIZE, MJUM9BYTES, MJUM16BYTES.
547  */
548 static __inline void *
549 m_cljget(struct mbuf *m, int how, int size)
550 {
551 	uma_zone_t zone;
552 
553 	if (m && m->m_flags & M_EXT)
554 		printf("%s: %p mbuf already has cluster\n", __func__, m);
555 	if (m != NULL)
556 		m->m_ext.ext_buf = NULL;
557 
558 	zone = m_getzone(size);
559 	return (uma_zalloc_arg(zone, m, how));
560 }
561 
562 static __inline void
563 m_cljset(struct mbuf *m, void *cl, int type)
564 {
565 	uma_zone_t zone;
566 	int size;
567 
568 	switch (type) {
569 	case EXT_CLUSTER:
570 		size = MCLBYTES;
571 		zone = zone_clust;
572 		break;
573 #if MJUMPAGESIZE != MCLBYTES
574 	case EXT_JUMBOP:
575 		size = MJUMPAGESIZE;
576 		zone = zone_jumbop;
577 		break;
578 #endif
579 	case EXT_JUMBO9:
580 		size = MJUM9BYTES;
581 		zone = zone_jumbo9;
582 		break;
583 	case EXT_JUMBO16:
584 		size = MJUM16BYTES;
585 		zone = zone_jumbo16;
586 		break;
587 	default:
588 		panic("%s: unknown cluster type", __func__);
589 		break;
590 	}
591 
592 	m->m_data = m->m_ext.ext_buf = cl;
593 	m->m_ext.ext_free = m->m_ext.ext_arg1 = m->m_ext.ext_arg2 = NULL;
594 	m->m_ext.ext_size = size;
595 	m->m_ext.ext_type = type;
596 	m->m_ext.ref_cnt = uma_find_refcnt(zone, cl);
597 	m->m_flags |= M_EXT;
598 
599 }
600 
601 static __inline void
602 m_chtype(struct mbuf *m, short new_type)
603 {
604 
605 	m->m_type = new_type;
606 }
607 
608 static __inline struct mbuf *
609 m_last(struct mbuf *m)
610 {
611 
612 	while (m->m_next)
613 		m = m->m_next;
614 	return (m);
615 }
616 
617 /*
618  * mbuf, cluster, and external object allocation macros (for compatibility
619  * purposes).
620  */
621 #define	M_MOVE_PKTHDR(to, from)	m_move_pkthdr((to), (from))
622 #define	MGET(m, how, type)	((m) = m_get((how), (type)))
623 #define	MGETHDR(m, how, type)	((m) = m_gethdr((how), (type)))
624 #define	MCLGET(m, how)		m_clget((m), (how))
625 #define	MEXTADD(m, buf, size, free, arg1, arg2, flags, type)		\
626     (void )m_extadd((m), (caddr_t)(buf), (size), (free), (arg1), (arg2),\
627     (flags), (type), M_NOWAIT)
628 #define	m_getm(m, len, how, type)					\
629     m_getm2((m), (len), (how), (type), M_PKTHDR)
630 
631 /*
632  * Evaluate TRUE if it's safe to write to the mbuf m's data region (this can
633  * be both the local data payload, or an external buffer area, depending on
634  * whether M_EXT is set).
635  */
636 #define	M_WRITABLE(m)	(!((m)->m_flags & M_RDONLY) &&			\
637 			 (!(((m)->m_flags & M_EXT)) ||			\
638 			 (*((m)->m_ext.ref_cnt) == 1)) )		\
639 
640 /* Check if the supplied mbuf has a packet header, or else panic. */
641 #define	M_ASSERTPKTHDR(m)						\
642 	KASSERT((m) != NULL && (m)->m_flags & M_PKTHDR,			\
643 	    ("%s: no mbuf packet header!", __func__))
644 
645 /*
646  * Ensure that the supplied mbuf is a valid, non-free mbuf.
647  *
648  * XXX: Broken at the moment.  Need some UMA magic to make it work again.
649  */
650 #define	M_ASSERTVALID(m)						\
651 	KASSERT((((struct mbuf *)m)->m_flags & 0) == 0,			\
652 	    ("%s: attempted use of a free mbuf!", __func__))
653 
654 /*
655  * Set the m_data pointer of a newly-allocated mbuf (m_get/MGET) to place an
656  * object of the specified size at the end of the mbuf, longword aligned.
657  */
658 #define	M_ALIGN(m, len) do {						\
659 	KASSERT(!((m)->m_flags & (M_PKTHDR|M_EXT)),			\
660 		("%s: M_ALIGN not normal mbuf", __func__));		\
661 	KASSERT((m)->m_data == (m)->m_dat,				\
662 		("%s: M_ALIGN not a virgin mbuf", __func__));		\
663 	(m)->m_data += (MLEN - (len)) & ~(sizeof(long) - 1);		\
664 } while (0)
665 
666 /*
667  * As above, for mbufs allocated with m_gethdr/MGETHDR or initialized by
668  * M_DUP/MOVE_PKTHDR.
669  */
670 #define	MH_ALIGN(m, len) do {						\
671 	KASSERT((m)->m_flags & M_PKTHDR && !((m)->m_flags & M_EXT),	\
672 		("%s: MH_ALIGN not PKTHDR mbuf", __func__));		\
673 	KASSERT((m)->m_data == (m)->m_pktdat,				\
674 		("%s: MH_ALIGN not a virgin mbuf", __func__));		\
675 	(m)->m_data += (MHLEN - (len)) & ~(sizeof(long) - 1);		\
676 } while (0)
677 
678 /*
679  * As above, for mbuf with external storage.
680  */
681 #define	MEXT_ALIGN(m, len) do {						\
682 	KASSERT((m)->m_flags & M_EXT,					\
683 		("%s: MEXT_ALIGN not an M_EXT mbuf", __func__));	\
684 	KASSERT((m)->m_data == (m)->m_ext.ext_buf,			\
685 		("%s: MEXT_ALIGN not a virgin mbuf", __func__));	\
686 	(m)->m_data += ((m)->m_ext.ext_size - (len)) &			\
687 	    ~(sizeof(long) - 1); 					\
688 } while (0)
689 
690 /*
691  * Compute the amount of space available before the current start of data in
692  * an mbuf.
693  *
694  * The M_WRITABLE() is a temporary, conservative safety measure: the burden
695  * of checking writability of the mbuf data area rests solely with the caller.
696  */
697 #define	M_LEADINGSPACE(m)						\
698 	((m)->m_flags & M_EXT ?						\
699 	    (M_WRITABLE(m) ? (m)->m_data - (m)->m_ext.ext_buf : 0):	\
700 	    (m)->m_flags & M_PKTHDR ? (m)->m_data - (m)->m_pktdat :	\
701 	    (m)->m_data - (m)->m_dat)
702 
703 /*
704  * Compute the amount of space available after the end of data in an mbuf.
705  *
706  * The M_WRITABLE() is a temporary, conservative safety measure: the burden
707  * of checking writability of the mbuf data area rests solely with the caller.
708  */
709 #define	M_TRAILINGSPACE(m)						\
710 	((m)->m_flags & M_EXT ?						\
711 	    (M_WRITABLE(m) ? (m)->m_ext.ext_buf + (m)->m_ext.ext_size	\
712 		- ((m)->m_data + (m)->m_len) : 0) :			\
713 	    &(m)->m_dat[MLEN] - ((m)->m_data + (m)->m_len))
714 
715 /*
716  * Arrange to prepend space of size plen to mbuf m.  If a new mbuf must be
717  * allocated, how specifies whether to wait.  If the allocation fails, the
718  * original mbuf chain is freed and m is set to NULL.
719  */
720 #define	M_PREPEND(m, plen, how) do {					\
721 	struct mbuf **_mmp = &(m);					\
722 	struct mbuf *_mm = *_mmp;					\
723 	int _mplen = (plen);						\
724 	int __mhow = (how);						\
725 									\
726 	MBUF_CHECKSLEEP(how);						\
727 	if (M_LEADINGSPACE(_mm) >= _mplen) {				\
728 		_mm->m_data -= _mplen;					\
729 		_mm->m_len += _mplen;					\
730 	} else								\
731 		_mm = m_prepend(_mm, _mplen, __mhow);			\
732 	if (_mm != NULL && _mm->m_flags & M_PKTHDR)			\
733 		_mm->m_pkthdr.len += _mplen;				\
734 	*_mmp = _mm;							\
735 } while (0)
736 
737 /*
738  * Change mbuf to new type.  This is a relatively expensive operation and
739  * should be avoided.
740  */
741 #define	MCHTYPE(m, t)	m_chtype((m), (t))
742 
743 /* Length to m_copy to copy all. */
744 #define	M_COPYALL	1000000000
745 
746 /* Compatibility with 4.3. */
747 #define	m_copy(m, o, l)	m_copym((m), (o), (l), M_NOWAIT)
748 
749 extern int		max_datalen;	/* MHLEN - max_hdr */
750 extern int		max_hdr;	/* Largest link + protocol header */
751 extern int		max_linkhdr;	/* Largest link-level header */
752 extern int		max_protohdr;	/* Largest protocol header */
753 extern int		nmbclusters;	/* Maximum number of clusters */
754 
755 struct uio;
756 
757 void		 m_adj(struct mbuf *, int);
758 void		 m_align(struct mbuf *, int);
759 int		 m_apply(struct mbuf *, int, int,
760 		    int (*)(void *, void *, u_int), void *);
761 int		 m_append(struct mbuf *, int, c_caddr_t);
762 void		 m_cat(struct mbuf *, struct mbuf *);
763 int		 m_extadd(struct mbuf *, caddr_t, u_int,
764 		    void (*)(void *, void *), void *, void *, int, int, int);
765 struct mbuf	*m_collapse(struct mbuf *, int, int);
766 void		 m_copyback(struct mbuf *, int, int, c_caddr_t);
767 void		 m_copydata(const struct mbuf *, int, int, caddr_t);
768 struct mbuf	*m_copym(struct mbuf *, int, int, int);
769 struct mbuf	*m_copymdata(struct mbuf *, struct mbuf *,
770 		    int, int, int, int);
771 struct mbuf	*m_copypacket(struct mbuf *, int);
772 void		 m_copy_pkthdr(struct mbuf *, struct mbuf *);
773 struct mbuf	*m_copyup(struct mbuf *, int, int);
774 struct mbuf	*m_defrag(struct mbuf *, int);
775 void		 m_demote(struct mbuf *, int);
776 struct mbuf	*m_devget(char *, int, int, struct ifnet *,
777 		    void (*)(char *, caddr_t, u_int));
778 struct mbuf	*m_dup(struct mbuf *, int);
779 int		 m_dup_pkthdr(struct mbuf *, struct mbuf *, int);
780 u_int		 m_fixhdr(struct mbuf *);
781 struct mbuf	*m_fragment(struct mbuf *, int, int);
782 void		 m_freem(struct mbuf *);
783 struct mbuf	*m_get2(int, int, short, int);
784 struct mbuf	*m_getjcl(int, short, int, int);
785 struct mbuf	*m_getm2(struct mbuf *, int, int, short, int);
786 struct mbuf	*m_getptr(struct mbuf *, int, int *);
787 u_int		 m_length(struct mbuf *, struct mbuf **);
788 int		 m_mbuftouio(struct uio *, struct mbuf *, int);
789 void		 m_move_pkthdr(struct mbuf *, struct mbuf *);
790 struct mbuf	*m_prepend(struct mbuf *, int, int);
791 void		 m_print(const struct mbuf *, int);
792 struct mbuf	*m_pulldown(struct mbuf *, int, int, int *);
793 struct mbuf	*m_pullup(struct mbuf *, int);
794 int		 m_sanity(struct mbuf *, int);
795 struct mbuf	*m_split(struct mbuf *, int, int);
796 struct mbuf	*m_uiotombuf(struct uio *, int, int, int, int);
797 struct mbuf	*m_unshare(struct mbuf *, int);
798 
799 /*-
800  * Network packets may have annotations attached by affixing a list of
801  * "packet tags" to the pkthdr structure.  Packet tags are dynamically
802  * allocated semi-opaque data structures that have a fixed header
803  * (struct m_tag) that specifies the size of the memory block and a
804  * <cookie,type> pair that identifies it.  The cookie is a 32-bit unique
805  * unsigned value used to identify a module or ABI.  By convention this value
806  * is chosen as the date+time that the module is created, expressed as the
807  * number of seconds since the epoch (e.g., using date -u +'%s').  The type
808  * value is an ABI/module-specific value that identifies a particular
809  * annotation and is private to the module.  For compatibility with systems
810  * like OpenBSD that define packet tags w/o an ABI/module cookie, the value
811  * PACKET_ABI_COMPAT is used to implement m_tag_get and m_tag_find
812  * compatibility shim functions and several tag types are defined below.
813  * Users that do not require compatibility should use a private cookie value
814  * so that packet tag-related definitions can be maintained privately.
815  *
816  * Note that the packet tag returned by m_tag_alloc has the default memory
817  * alignment implemented by malloc.  To reference private data one can use a
818  * construct like:
819  *
820  *	struct m_tag *mtag = m_tag_alloc(...);
821  *	struct foo *p = (struct foo *)(mtag+1);
822  *
823  * if the alignment of struct m_tag is sufficient for referencing members of
824  * struct foo.  Otherwise it is necessary to embed struct m_tag within the
825  * private data structure to insure proper alignment; e.g.,
826  *
827  *	struct foo {
828  *		struct m_tag	tag;
829  *		...
830  *	};
831  *	struct foo *p = (struct foo *) m_tag_alloc(...);
832  *	struct m_tag *mtag = &p->tag;
833  */
834 
835 /*
836  * Persistent tags stay with an mbuf until the mbuf is reclaimed.  Otherwise
837  * tags are expected to ``vanish'' when they pass through a network
838  * interface.  For most interfaces this happens normally as the tags are
839  * reclaimed when the mbuf is free'd.  However in some special cases
840  * reclaiming must be done manually.  An example is packets that pass through
841  * the loopback interface.  Also, one must be careful to do this when
842  * ``turning around'' packets (e.g., icmp_reflect).
843  *
844  * To mark a tag persistent bit-or this flag in when defining the tag id.
845  * The tag will then be treated as described above.
846  */
847 #define	MTAG_PERSISTENT				0x800
848 
849 #define	PACKET_TAG_NONE				0  /* Nadda */
850 
851 /* Packet tags for use with PACKET_ABI_COMPAT. */
852 #define	PACKET_TAG_IPSEC_IN_DONE		1  /* IPsec applied, in */
853 #define	PACKET_TAG_IPSEC_OUT_DONE		2  /* IPsec applied, out */
854 #define	PACKET_TAG_IPSEC_IN_CRYPTO_DONE		3  /* NIC IPsec crypto done */
855 #define	PACKET_TAG_IPSEC_OUT_CRYPTO_NEEDED	4  /* NIC IPsec crypto req'ed */
856 #define	PACKET_TAG_IPSEC_IN_COULD_DO_CRYPTO	5  /* NIC notifies IPsec */
857 #define	PACKET_TAG_IPSEC_PENDING_TDB		6  /* Reminder to do IPsec */
858 #define	PACKET_TAG_BRIDGE			7  /* Bridge processing done */
859 #define	PACKET_TAG_GIF				8  /* GIF processing done */
860 #define	PACKET_TAG_GRE				9  /* GRE processing done */
861 #define	PACKET_TAG_IN_PACKET_CHECKSUM		10 /* NIC checksumming done */
862 #define	PACKET_TAG_ENCAP			11 /* Encap.  processing */
863 #define	PACKET_TAG_IPSEC_SOCKET			12 /* IPSEC socket ref */
864 #define	PACKET_TAG_IPSEC_HISTORY		13 /* IPSEC history */
865 #define	PACKET_TAG_IPV6_INPUT			14 /* IPV6 input processing */
866 #define	PACKET_TAG_DUMMYNET			15 /* dummynet info */
867 #define	PACKET_TAG_DIVERT			17 /* divert info */
868 #define	PACKET_TAG_IPFORWARD			18 /* ipforward info */
869 #define	PACKET_TAG_MACLABEL	(19 | MTAG_PERSISTENT) /* MAC label */
870 #define	PACKET_TAG_PF		(21 | MTAG_PERSISTENT) /* PF/ALTQ information */
871 #define	PACKET_TAG_RTSOCKFAM			25 /* rtsock sa family */
872 #define	PACKET_TAG_IPOPTIONS			27 /* Saved IP options */
873 #define	PACKET_TAG_CARP				28 /* CARP info */
874 #define	PACKET_TAG_IPSEC_NAT_T_PORTS		29 /* two uint16_t */
875 #define	PACKET_TAG_ND_OUTGOING			30 /* ND outgoing */
876 
877 /* Specific cookies and tags. */
878 
879 /* Packet tag routines. */
880 struct m_tag	*m_tag_alloc(u_int32_t, int, int, int);
881 void		 m_tag_delete(struct mbuf *, struct m_tag *);
882 void		 m_tag_delete_chain(struct mbuf *, struct m_tag *);
883 void		 m_tag_free_default(struct m_tag *);
884 struct m_tag	*m_tag_locate(struct mbuf *, u_int32_t, int, struct m_tag *);
885 struct m_tag	*m_tag_copy(struct m_tag *, int);
886 int		 m_tag_copy_chain(struct mbuf *, struct mbuf *, int);
887 void		 m_tag_delete_nonpersistent(struct mbuf *);
888 
889 /*
890  * Initialize the list of tags associated with an mbuf.
891  */
892 static __inline void
893 m_tag_init(struct mbuf *m)
894 {
895 
896 	SLIST_INIT(&m->m_pkthdr.tags);
897 }
898 
899 /*
900  * Set up the contents of a tag.  Note that this does not fill in the free
901  * method; the caller is expected to do that.
902  *
903  * XXX probably should be called m_tag_init, but that was already taken.
904  */
905 static __inline void
906 m_tag_setup(struct m_tag *t, u_int32_t cookie, int type, int len)
907 {
908 
909 	t->m_tag_id = type;
910 	t->m_tag_len = len;
911 	t->m_tag_cookie = cookie;
912 }
913 
914 /*
915  * Reclaim resources associated with a tag.
916  */
917 static __inline void
918 m_tag_free(struct m_tag *t)
919 {
920 
921 	(*t->m_tag_free)(t);
922 }
923 
924 /*
925  * Return the first tag associated with an mbuf.
926  */
927 static __inline struct m_tag *
928 m_tag_first(struct mbuf *m)
929 {
930 
931 	return (SLIST_FIRST(&m->m_pkthdr.tags));
932 }
933 
934 /*
935  * Return the next tag in the list of tags associated with an mbuf.
936  */
937 static __inline struct m_tag *
938 m_tag_next(struct mbuf *m, struct m_tag *t)
939 {
940 
941 	return (SLIST_NEXT(t, m_tag_link));
942 }
943 
944 /*
945  * Prepend a tag to the list of tags associated with an mbuf.
946  */
947 static __inline void
948 m_tag_prepend(struct mbuf *m, struct m_tag *t)
949 {
950 
951 	SLIST_INSERT_HEAD(&m->m_pkthdr.tags, t, m_tag_link);
952 }
953 
954 /*
955  * Unlink a tag from the list of tags associated with an mbuf.
956  */
957 static __inline void
958 m_tag_unlink(struct mbuf *m, struct m_tag *t)
959 {
960 
961 	SLIST_REMOVE(&m->m_pkthdr.tags, t, m_tag, m_tag_link);
962 }
963 
964 /* These are for OpenBSD compatibility. */
965 #define	MTAG_ABI_COMPAT		0		/* compatibility ABI */
966 
967 static __inline struct m_tag *
968 m_tag_get(int type, int length, int wait)
969 {
970 	return (m_tag_alloc(MTAG_ABI_COMPAT, type, length, wait));
971 }
972 
973 static __inline struct m_tag *
974 m_tag_find(struct mbuf *m, int type, struct m_tag *start)
975 {
976 	return (SLIST_EMPTY(&m->m_pkthdr.tags) ? (struct m_tag *)NULL :
977 	    m_tag_locate(m, MTAG_ABI_COMPAT, type, start));
978 }
979 
980 static int inline
981 rt_m_getfib(struct mbuf *m)
982 {
983 	KASSERT(m->m_flags & M_PKTHDR , ("Attempt to get FIB from non header mbuf."));
984 	return (m->m_pkthdr.fibnum);
985 }
986 
987 #define M_GETFIB(_m)   rt_m_getfib(_m)
988 
989 #define M_SETFIB(_m, _fib) do {						\
990         KASSERT((_m)->m_flags & M_PKTHDR, ("Attempt to set FIB on non header mbuf."));	\
991 	((_m)->m_pkthdr.fibnum) = (_fib);				\
992 } while (0)
993 
994 #endif /* _KERNEL */
995 
996 #ifdef MBUF_PROFILING
997  void m_profile(struct mbuf *m);
998  #define M_PROFILE(m) m_profile(m)
999 #else
1000  #define M_PROFILE(m)
1001 #endif
1002 
1003 
1004 #endif /* !_SYS_MBUF_H_ */
1005