mbuf.h revision 254857
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: head/sys/sys/mbuf.h 254857 2013-08-25 13:30:37Z andre $
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 * mtodo(m, o) -- Same as above but with offset 'o' into data.
71 */
72#define	mtod(m, t)	((t)((m)->m_data))
73#define	mtodo(m, o)	((void *)(((m)->m_data) + (o)))
74
75/*
76 * Argument structure passed to UMA routines during mbuf and packet
77 * allocations.
78 */
79struct mb_args {
80	int	flags;	/* Flags for mbuf being allocated */
81	short	type;	/* Type of mbuf being allocated */
82};
83#endif /* _KERNEL */
84
85/*
86 * Header present at the beginning of every mbuf.
87 * Size ILP32: 20
88 *	 LP64: 32
89 */
90struct m_hdr {
91	struct mbuf	*mh_next;	/* next buffer in chain */
92	struct mbuf	*mh_nextpkt;	/* next chain in queue/record */
93	caddr_t		 mh_data;	/* location of data */
94	int32_t		 mh_len;	/* amount of data in this mbuf */
95	uint32_t	 mh_type:8,	/* type of data in this mbuf */
96			 mh_flags:24;	/* flags; see below */
97};
98
99/*
100 * Packet tag structure (see below for details).
101 */
102struct m_tag {
103	SLIST_ENTRY(m_tag)	m_tag_link;	/* List of packet tags */
104	u_int16_t		m_tag_id;	/* Tag ID */
105	u_int16_t		m_tag_len;	/* Length of data */
106	u_int32_t		m_tag_cookie;	/* ABI/Module ID */
107	void			(*m_tag_free)(struct m_tag *);
108};
109
110/*
111 * Record/packet header in first mbuf of chain; valid only if M_PKTHDR is set.
112 * Size ILP32: 48
113 *	 LP64: 56
114 */
115struct pkthdr {
116	struct ifnet	*rcvif;		/* rcv interface */
117	SLIST_HEAD(packet_tags, m_tag) tags; /* list of packet tags */
118	int32_t		 len;		/* total packet length */
119
120	/* Layer crossing persistent information. */
121	uint32_t	 flowid;	/* packet's 4-tuple system */
122	uint64_t	 csum_flags;	/* checksum and offload features */
123	uint16_t	 fibnum;	/* this packet should use this fib */
124	uint8_t		 cosqos;	/* class/quality of service */
125	uint8_t		 rsstype;	/* hash type */
126	uint8_t		 l2hlen;	/* layer 2 header length */
127	uint8_t		 l3hlen;	/* layer 3 header length */
128	uint8_t		 l4hlen;	/* layer 4 header length */
129	uint8_t		 l5hlen;	/* layer 5 header length */
130	union {
131		uint8_t  eigth[8];
132		uint16_t sixteen[4];
133		uint32_t thirtytwo[2];
134		uint64_t sixtyfour[1];
135		uintptr_t unintptr[1];
136		void	*ptr;
137	} PH_per;
138
139	/* Layer specific non-persistent local storage for reassembly, etc. */
140	union {
141		uint8_t  eigth[8];
142		uint16_t sixteen[4];
143		uint32_t thirtytwo[2];
144		uint64_t sixtyfour[1];
145		uintptr_t unintptr[1];
146		void 	*ptr;
147	} PH_loc;
148};
149#define	ether_vtag	PH_per.sixteen[0]
150#define	PH_vt		PH_per
151#define	vt_nrecs	sixteen[0]
152#define	tso_segsz	PH_per.sixteen[1]
153#define	csum_phsum	PH_per.sixteen[2]
154#define	csum_data	PH_per.thirtytwo[1]
155
156/*
157 * Description of external storage mapped into mbuf; valid only if M_EXT is
158 * set.
159 * Size ILP32: 28
160 *	 LP64: 48
161 */
162struct m_ext {
163	volatile u_int	*ref_cnt;	/* pointer to ref count info */
164	caddr_t		 ext_buf;	/* start of buffer */
165	uint32_t	 ext_size;	/* size of buffer, for ext_free */
166	uint32_t	 ext_type:8,	/* type of external storage */
167			 ext_flags:24;	/* external storage mbuf flags */
168	int		(*ext_free)	/* free routine if not the usual */
169			    (struct mbuf *, void *, void *);
170	void		*ext_arg1;	/* optional argument pointer */
171	void		*ext_arg2;	/* optional argument pointer */
172};
173
174/*
175 * The core of the mbuf object along with some shortcut defines for practical
176 * purposes.
177 */
178struct mbuf {
179	struct m_hdr	m_hdr;
180	union {
181		struct {
182			struct pkthdr	MH_pkthdr;	/* M_PKTHDR set */
183			union {
184				struct m_ext	MH_ext;	/* M_EXT set */
185				char		MH_databuf[MHLEN];
186			} MH_dat;
187		} MH;
188		char	M_databuf[MLEN];		/* !M_PKTHDR, !M_EXT */
189	} M_dat;
190};
191#define	m_next		m_hdr.mh_next
192#define	m_len		m_hdr.mh_len
193#define	m_data		m_hdr.mh_data
194#define	m_type		m_hdr.mh_type
195#define	m_flags		m_hdr.mh_flags
196#define	m_nextpkt	m_hdr.mh_nextpkt
197#define	m_act		m_nextpkt
198#define	m_pkthdr	M_dat.MH.MH_pkthdr
199#define	m_ext		M_dat.MH.MH_dat.MH_ext
200#define	m_pktdat	M_dat.MH.MH_dat.MH_databuf
201#define	m_dat		M_dat.M_databuf
202
203/*
204 * mbuf flags of global significance and layer crossing.
205 * Those of only protocol/layer specific significance are to be mapped
206 * to M_PROTO[1-12] and cleared at layer handoff boundaries.
207 * NB: Limited to the lower 24 bits.
208 */
209#define	M_EXT		0x00000001 /* has associated external storage */
210#define	M_PKTHDR	0x00000002 /* start of record */
211#define	M_EOR		0x00000004 /* end of record */
212#define	M_RDONLY	0x00000008 /* associated data is marked read-only */
213#define	M_BCAST		0x00000010 /* send/received as link-level broadcast */
214#define	M_MCAST		0x00000020 /* send/received as link-level multicast */
215#define	M_PROMISC	0x00000040 /* packet was not for us */
216#define	M_VLANTAG	0x00000080 /* ether_vtag is valid */
217#define	M_FLOWID	0x00000100 /* deprecated: flowid is valid */
218#define	M_NOFREE	0x00000200 /* do not free mbuf, embedded in cluster */
219
220#define	M_PROTO1	0x00001000 /* protocol-specific */
221#define	M_PROTO2	0x00002000 /* protocol-specific */
222#define	M_PROTO3	0x00004000 /* protocol-specific */
223#define	M_PROTO4	0x00008000 /* protocol-specific */
224#define	M_PROTO5	0x00010000 /* protocol-specific */
225#define	M_PROTO6	0x00020000 /* protocol-specific */
226#define	M_PROTO7	0x00040000 /* protocol-specific */
227#define	M_PROTO8	0x00080000 /* protocol-specific */
228#define	M_PROTO9	0x00100000 /* protocol-specific */
229#define	M_PROTO10	0x00200000 /* protocol-specific */
230#define	M_PROTO11	0x00400000 /* protocol-specific */
231#define	M_PROTO12	0x00800000 /* protocol-specific */
232
233/*
234 * Flags to purge when crossing layers.
235 */
236#define	M_PROTOFLAGS \
237    (M_PROTO1|M_PROTO2|M_PROTO3|M_PROTO4|M_PROTO5|M_PROTO6|M_PROTO7|M_PROTO8|\
238     M_PROTO9|M_PROTO10|M_PROTO11|M_PROTO12)
239
240/*
241 * Flags preserved when copying m_pkthdr.
242 */
243#define M_COPYFLAGS \
244    (M_PKTHDR|M_EOR|M_RDONLY|M_BCAST|M_MCAST|M_VLANTAG|M_PROMISC| \
245     M_PROTOFLAGS)
246
247/*
248 * Mbuf flag description for use with printf(9) %b identifier.
249 */
250#define	M_FLAG_BITS \
251    "\20\1M_EXT\2M_PKTHDR\3M_EOR\4M_RDONLY\5M_BCAST\6M_MCAST" \
252    "\7M_PROMISC\10M_VLANTAG\11M_FLOWID"
253#define	M_FLAG_PROTOBITS \
254    "\15M_PROTO1\16M_PROTO2\17M_PROTO3\20M_PROTO4\21M_PROTO5" \
255    "\22M_PROTO6\23M_PROTO7\24M_PROTO8\25M_PROTO9\26M_PROTO10" \
256    "\27M_PROTO11\30M_PROTO12"
257#define	M_FLAG_PRINTF (M_FLAG_BITS M_FLAG_PROTOBITS)
258
259/*
260 * Network interface cards are able to hash protocol fields (such as IPv4
261 * addresses and TCP port numbers) classify packets into flows.  These flows
262 * can then be used to maintain ordering while delivering packets to the OS
263 * via parallel input queues, as well as to provide a stateless affinity
264 * model.  NIC drivers can pass up the hash via m->m_pkthdr.flowid, and set
265 * m_flag fields to indicate how the hash should be interpreted by the
266 * network stack.
267 *
268 * Most NICs support RSS, which provides ordering and explicit affinity, and
269 * use the hash m_flag bits to indicate what header fields were covered by
270 * the hash.  M_HASHTYPE_OPAQUE can be set by non-RSS cards or configurations
271 * that provide an opaque flow identifier, allowing for ordering and
272 * distribution without explicit affinity.
273 */
274#define	M_HASHTYPE_NONE			0
275#define	M_HASHTYPE_RSS_IPV4		1	/* IPv4 2-tuple */
276#define	M_HASHTYPE_RSS_TCP_IPV4		2	/* TCPv4 4-tuple */
277#define	M_HASHTYPE_RSS_IPV6		3	/* IPv6 2-tuple */
278#define	M_HASHTYPE_RSS_TCP_IPV6		4	/* TCPv6 4-tuple */
279#define	M_HASHTYPE_RSS_IPV6_EX		5	/* IPv6 2-tuple + ext hdrs */
280#define	M_HASHTYPE_RSS_TCP_IPV6_EX	6	/* TCPv6 4-tiple + ext hdrs */
281#define	M_HASHTYPE_OPAQUE		255	/* ordering, not affinity */
282
283#define	M_HASHTYPE_CLEAR(m)	((m)->m_pkthdr.rsstype = 0)
284#define	M_HASHTYPE_GET(m)	((m)->m_pkthdr.rsstype)
285#define	M_HASHTYPE_SET(m, v)	((m)->m_pkthdr.rsstype = (v))
286#define	M_HASHTYPE_TEST(m, v)	(M_HASHTYPE_GET(m) == (v))
287
288/*
289 * COS/QOS class and quality of service tags.
290 * It uses DSCP code points as base.
291 */
292#define	QOS_DSCP_CS0		0x00
293#define	QOS_DSCP_DEF		QOS_DSCP_CS0
294#define	QOS_DSCP_CS1		0x20
295#define	QOS_DSCP_AF11		0x28
296#define	QOS_DSCP_AF12		0x30
297#define	QOS_DSCP_AF13		0x38
298#define	QOS_DSCP_CS2		0x40
299#define	QOS_DSCP_AF21		0x48
300#define	QOS_DSCP_AF22		0x50
301#define	QOS_DSCP_AF23		0x58
302#define	QOS_DSCP_CS3		0x60
303#define	QOS_DSCP_AF31		0x68
304#define	QOS_DSCP_AF32		0x70
305#define	QOS_DSCP_AF33		0x78
306#define	QOS_DSCP_CS4		0x80
307#define	QOS_DSCP_AF41		0x88
308#define	QOS_DSCP_AF42		0x90
309#define	QOS_DSCP_AF43		0x98
310#define	QOS_DSCP_CS5		0xa0
311#define	QOS_DSCP_EF		0xb8
312#define	QOS_DSCP_CS6		0xc0
313#define	QOS_DSCP_CS7		0xe0
314
315/*
316 * External mbuf storage buffer types.
317 */
318#define	EXT_CLUSTER	1	/* mbuf cluster */
319#define	EXT_SFBUF	2	/* sendfile(2)'s sf_bufs */
320#define	EXT_JUMBOP	3	/* jumbo cluster 4096 bytes */
321#define	EXT_JUMBO9	4	/* jumbo cluster 9216 bytes */
322#define	EXT_JUMBO16	5	/* jumbo cluster 16184 bytes */
323#define	EXT_PACKET	6	/* mbuf+cluster from packet zone */
324#define	EXT_MBUF	7	/* external mbuf reference (M_IOVEC) */
325
326#define	EXT_VENDOR1	224	/* for vendor-internal use */
327#define	EXT_VENDOR2	225	/* for vendor-internal use */
328#define	EXT_VENDOR3	226	/* for vendor-internal use */
329#define	EXT_VENDOR4	227	/* for vendor-internal use */
330
331#define	EXT_EXP1	244	/* for experimental use */
332#define	EXT_EXP2	245	/* for experimental use */
333#define	EXT_EXP3	246	/* for experimental use */
334#define	EXT_EXP4	247	/* for experimental use */
335
336#define	EXT_NET_DRV	252	/* custom ext_buf provided by net driver(s) */
337#define	EXT_MOD_TYPE	253	/* custom module's ext_buf type */
338#define	EXT_DISPOSABLE	254	/* can throw this buffer away w/page flipping */
339#define	EXT_EXTREF	255	/* has externally maintained ref_cnt ptr */
340
341/*
342 * Flags for external mbuf buffer types.
343 * NB: limited to the lower 24 bits.
344 */
345#define	EXT_FLAG_EMBREF		0x000001	/* embedded ref_cnt, notyet */
346#define	EXT_FLAG_EXTREF		0x000002	/* external ref_cnt, notyet */
347#define	EXT_FLAG_NOFREE		0x000010	/* don't free mbuf to pool, notyet */
348
349#define	EXT_FLAG_VENDOR1	0x010000	/* for vendor-internal use */
350#define	EXT_FLAG_VENDOR2	0x020000	/* for vendor-internal use */
351#define	EXT_FLAG_VENDOR3	0x040000	/* for vendor-internal use */
352#define	EXT_FLAG_VENDOR4	0x080000	/* for vendor-internal use */
353
354#define	EXT_FLAG_EXP1		0x100000	/* for experimental use */
355#define	EXT_FLAG_EXP2		0x200000	/* for experimental use */
356#define	EXT_FLAG_EXP3		0x400000	/* for experimental use */
357#define	EXT_FLAG_EXP4		0x800000	/* for experimental use */
358
359/*
360 * EXT flag description for use with printf(9) %b identifier.
361 */
362#define	EXT_FLAG_BITS \
363    "\20\1EXT_FLAG_EMBREF\2EXT_FLAG_EXTREF\5EXT_FLAG_NOFREE" \
364    "\21EXT_FLAG_VENDOR1\22EXT_FLAG_VENDOR2\23EXT_FLAG_VENDOR3" \
365    "\24EXT_FLAG_VENDOR4\25EXT_FLAG_EXP1\26EXT_FLAG_EXP2\27EXT_FLAG_EXP3" \
366    "\30EXT_FLAG_EXP4"
367
368/*
369 * Return values for (*ext_free).
370 */
371#define	EXT_FREE_OK	0	/* Normal return */
372
373/*
374 * Flags indicating checksum, segmentation and other offload work to be
375 * done, or already done, by hardware or lower layers.  It is split into
376 * separate inbound and outbound flags.
377 *
378 * Outbound flags that are set by upper protocol layers requesting lower
379 * layers, or ideally the hardware, to perform these offloading tasks.
380 * For outbound packets this field and its flags can be directly tested
381 * against if_data.ifi_hwassist.
382 */
383#define	CSUM_IP			0x00000001	/* IP header checksum offload */
384#define	CSUM_IP_UDP		0x00000002	/* UDP checksum offload */
385#define	CSUM_IP_TCP		0x00000004	/* TCP checksum offload */
386#define	CSUM_IP_SCTP		0x00000008	/* SCTP checksum offload */
387#define	CSUM_IP_TSO		0x00000010	/* TCP segmentation offload */
388#define	CSUM_IP_ISCSI		0x00000020	/* iSCSI checksum offload */
389
390#define	CSUM_IP6_UDP		0x00000200	/* UDP checksum offload */
391#define	CSUM_IP6_TCP		0x00000400	/* TCP checksum offload */
392#define	CSUM_IP6_SCTP		0x00000800	/* SCTP checksum offload */
393#define	CSUM_IP6_TSO		0x00001000	/* TCP segmentation offload */
394#define	CSUM_IP6_ISCSI		0x00002000	/* iSCSI checksum offload */
395
396/* Inbound checksum support where the checksum was verified by hardware. */
397#define	CSUM_L3_CALC		0x01000000	/* calculated layer 3 csum */
398#define	CSUM_L3_VALID		0x02000000	/* checksum is correct */
399#define	CSUM_L4_CALC		0x04000000	/* calculated layer 4 csum */
400#define	CSUM_L4_VALID		0x08000000	/* checksum is correct */
401#define	CSUM_L5_CALC		0x10000000	/* calculated layer 5 csum */
402#define	CSUM_L5_VALID		0x20000000	/* checksum is correct */
403#define	CSUM_COALESED		0x40000000	/* contains merged segments */
404
405/*
406 * CSUM flag description for use with printf(9) %b identifier.
407 */
408#define	CSUM_BITS \
409    "\20\1CSUM_IP\2CSUM_IP_UDP\3CSUM_IP_TCP\4CSUM_IP_SCTP\5CSUM_IP_TSO" \
410    "\6CSUM_IP_ISCSI" \
411    "\12CSUM_IP6_UDP\13CSUM_IP6_TCP\14CSUM_IP6_SCTP\15CSUM_IP6_TSO" \
412    "\16CSUM_IP6_ISCSI" \
413    "\31CSUM_L3_CALC\32CSUM_L3_VALID\33CSUM_L4_CALC\34CSUM_L4_VALID" \
414    "\35CSUM_L5_CALC\36CSUM_L5_VALID\37CSUM_COALESED"
415
416/* CSUM flags compatibility mappings. */
417#define	CSUM_IP_CHECKED		CSUM_L3_CALC
418#define	CSUM_IP_VALID		CSUM_L3_VALID
419#define	CSUM_DATA_VALID		CSUM_L4_VALID
420#define	CSUM_PSEUDO_HDR		CSUM_L4_CALC
421#define	CSUM_SCTP_VALID		CSUM_L4_VALID
422#define	CSUM_DELAY_DATA		(CSUM_TCP|CSUM_UDP)
423#define	CSUM_DELAY_IP		CSUM_IP		/* Only v4, no v6 IP hdr csum */
424#define	CSUM_DELAY_DATA_IPV6	(CSUM_TCP_IPV6|CSUM_UDP_IPV6)
425#define	CSUM_DATA_VALID_IPV6	CSUM_DATA_VALID
426#define	CSUM_TCP		CSUM_IP_TCP
427#define	CSUM_UDP		CSUM_IP_UDP
428#define	CSUM_SCTP		CSUM_IP_SCTP
429#define	CSUM_TSO		(CSUM_IP_TSO|CSUM_IP6_TSO)
430#define	CSUM_UDP_IPV6		CSUM_IP6_UDP
431#define	CSUM_TCP_IPV6		CSUM_IP6_TCP
432#define	CSUM_SCTP_IPV6		CSUM_IP6_SCTP
433#define	CSUM_FRAGMENT		0x0		/* Unused */
434
435/*
436 * mbuf types describing the content of the mbuf (including external storage).
437 */
438#define	MT_NOTMBUF	0	/* USED INTERNALLY ONLY! Object is not mbuf */
439#define	MT_DATA		1	/* dynamic (data) allocation */
440#define	MT_HEADER	MT_DATA	/* packet header, use M_PKTHDR instead */
441
442#define	MT_VENDOR1	4	/* for vendor-internal use */
443#define	MT_VENDOR2	5	/* for vendor-internal use */
444#define	MT_VENDOR3	6	/* for vendor-internal use */
445#define	MT_VENDOR4	7	/* for vendor-internal use */
446
447#define	MT_SONAME	8	/* socket name */
448
449#define	MT_EXP1		9	/* for experimental use */
450#define	MT_EXP2		10	/* for experimental use */
451#define	MT_EXP3		11	/* for experimental use */
452#define	MT_EXP4		12	/* for experimental use */
453
454#define	MT_CONTROL	14	/* extra-data protocol message */
455#define	MT_OOBDATA	15	/* expedited data  */
456#define	MT_NTYPES	16	/* number of mbuf types for mbtypes[] */
457
458#define	MT_NOINIT	255	/* Not a type but a flag to allocate
459				   a non-initialized mbuf */
460
461/*
462 * Compatibility with historic mbuf allocator.
463 */
464#define	MBTOM(how)	(how)
465#define	M_DONTWAIT	M_NOWAIT
466#define	M_TRYWAIT	M_WAITOK
467#define	M_WAIT		M_WAITOK
468
469/*
470 * String names of mbuf-related UMA(9) and malloc(9) types.  Exposed to
471 * !_KERNEL so that monitoring tools can look up the zones with
472 * libmemstat(3).
473 */
474#define	MBUF_MEM_NAME		"mbuf"
475#define	MBUF_CLUSTER_MEM_NAME	"mbuf_cluster"
476#define	MBUF_PACKET_MEM_NAME	"mbuf_packet"
477#define	MBUF_JUMBOP_MEM_NAME	"mbuf_jumbo_page"
478#define	MBUF_JUMBO9_MEM_NAME	"mbuf_jumbo_9k"
479#define	MBUF_JUMBO16_MEM_NAME	"mbuf_jumbo_16k"
480#define	MBUF_TAG_MEM_NAME	"mbuf_tag"
481#define	MBUF_EXTREFCNT_MEM_NAME	"mbuf_ext_refcnt"
482
483#ifdef _KERNEL
484
485#ifdef WITNESS
486#define	MBUF_CHECKSLEEP(how) do {					\
487	if (how == M_WAITOK)						\
488		WITNESS_WARN(WARN_GIANTOK | WARN_SLEEPOK, NULL,		\
489		    "Sleeping in \"%s\"", __func__);			\
490} while (0)
491#else
492#define	MBUF_CHECKSLEEP(how)
493#endif
494
495/*
496 * Network buffer allocation API
497 *
498 * The rest of it is defined in kern/kern_mbuf.c
499 */
500extern uma_zone_t	zone_mbuf;
501extern uma_zone_t	zone_clust;
502extern uma_zone_t	zone_pack;
503extern uma_zone_t	zone_jumbop;
504extern uma_zone_t	zone_jumbo9;
505extern uma_zone_t	zone_jumbo16;
506extern uma_zone_t	zone_ext_refcnt;
507
508void		 mb_free_ext(struct mbuf *);
509int		 m_pkthdr_init(struct mbuf *, int);
510
511static __inline int
512m_gettype(int size)
513{
514	int type;
515
516	switch (size) {
517	case MSIZE:
518		type = EXT_MBUF;
519		break;
520	case MCLBYTES:
521		type = EXT_CLUSTER;
522		break;
523#if MJUMPAGESIZE != MCLBYTES
524	case MJUMPAGESIZE:
525		type = EXT_JUMBOP;
526		break;
527#endif
528	case MJUM9BYTES:
529		type = EXT_JUMBO9;
530		break;
531	case MJUM16BYTES:
532		type = EXT_JUMBO16;
533		break;
534	default:
535		panic("%s: invalid cluster size", __func__);
536	}
537
538	return (type);
539}
540
541static __inline uma_zone_t
542m_getzone(int size)
543{
544	uma_zone_t zone;
545
546	switch (size) {
547	case MCLBYTES:
548		zone = zone_clust;
549		break;
550#if MJUMPAGESIZE != MCLBYTES
551	case MJUMPAGESIZE:
552		zone = zone_jumbop;
553		break;
554#endif
555	case MJUM9BYTES:
556		zone = zone_jumbo9;
557		break;
558	case MJUM16BYTES:
559		zone = zone_jumbo16;
560		break;
561	default:
562		panic("%s: invalid cluster size", __func__);
563	}
564
565	return (zone);
566}
567
568/*
569 * Initialize an mbuf with linear storage.
570 *
571 * Inline because the consumer text overhead will be roughly the same to
572 * initialize or call a function with this many parameters and M_PKTHDR
573 * should go away with constant propagation for !MGETHDR.
574 */
575static __inline int
576m_init(struct mbuf *m, uma_zone_t zone, int size, int how, short type,
577    int flags)
578{
579	int error;
580
581	m->m_next = NULL;
582	m->m_nextpkt = NULL;
583	m->m_data = m->m_dat;
584	m->m_len = 0;
585	m->m_flags = flags;
586	m->m_type = type;
587	if (flags & M_PKTHDR) {
588		if ((error = m_pkthdr_init(m, how)) != 0)
589			return (error);
590	}
591
592	return (0);
593}
594
595static __inline struct mbuf *
596m_get(int how, short type)
597{
598	struct mb_args args;
599
600	args.flags = 0;
601	args.type = type;
602	return (uma_zalloc_arg(zone_mbuf, &args, how));
603}
604
605/*
606 * XXX This should be deprecated, very little use.
607 */
608static __inline struct mbuf *
609m_getclr(int how, short type)
610{
611	struct mbuf *m;
612	struct mb_args args;
613
614	args.flags = 0;
615	args.type = type;
616	m = uma_zalloc_arg(zone_mbuf, &args, how);
617	if (m != NULL)
618		bzero(m->m_data, MLEN);
619	return (m);
620}
621
622static __inline struct mbuf *
623m_gethdr(int how, short type)
624{
625	struct mb_args args;
626
627	args.flags = M_PKTHDR;
628	args.type = type;
629	return (uma_zalloc_arg(zone_mbuf, &args, how));
630}
631
632static __inline struct mbuf *
633m_getcl(int how, short type, int flags)
634{
635	struct mb_args args;
636
637	args.flags = flags;
638	args.type = type;
639	return (uma_zalloc_arg(zone_pack, &args, how));
640}
641
642static __inline struct mbuf *
643m_free(struct mbuf *m)
644{
645	struct mbuf *n = m->m_next;
646
647	if (m->m_flags & M_EXT)
648		mb_free_ext(m);
649	else if ((m->m_flags & M_NOFREE) == 0)
650		uma_zfree(zone_mbuf, m);
651	return (n);
652}
653
654static __inline void
655m_clget(struct mbuf *m, int how)
656{
657
658	if (m->m_flags & M_EXT)
659		printf("%s: %p mbuf already has cluster\n", __func__, m);
660	m->m_ext.ext_buf = (char *)NULL;
661	uma_zalloc_arg(zone_clust, m, how);
662	/*
663	 * On a cluster allocation failure, drain the packet zone and retry,
664	 * we might be able to loosen a few clusters up on the drain.
665	 */
666	if ((how & M_NOWAIT) && (m->m_ext.ext_buf == NULL)) {
667		zone_drain(zone_pack);
668		uma_zalloc_arg(zone_clust, m, how);
669	}
670}
671
672/*
673 * m_cljget() is different from m_clget() as it can allocate clusters without
674 * attaching them to an mbuf.  In that case the return value is the pointer
675 * to the cluster of the requested size.  If an mbuf was specified, it gets
676 * the cluster attached to it and the return value can be safely ignored.
677 * For size it takes MCLBYTES, MJUMPAGESIZE, MJUM9BYTES, MJUM16BYTES.
678 */
679static __inline void *
680m_cljget(struct mbuf *m, int how, int size)
681{
682	uma_zone_t zone;
683
684	if (m && m->m_flags & M_EXT)
685		printf("%s: %p mbuf already has cluster\n", __func__, m);
686	if (m != NULL)
687		m->m_ext.ext_buf = NULL;
688
689	zone = m_getzone(size);
690	return (uma_zalloc_arg(zone, m, how));
691}
692
693static __inline void
694m_cljset(struct mbuf *m, void *cl, int type)
695{
696	uma_zone_t zone;
697	int size;
698
699	switch (type) {
700	case EXT_CLUSTER:
701		size = MCLBYTES;
702		zone = zone_clust;
703		break;
704#if MJUMPAGESIZE != MCLBYTES
705	case EXT_JUMBOP:
706		size = MJUMPAGESIZE;
707		zone = zone_jumbop;
708		break;
709#endif
710	case EXT_JUMBO9:
711		size = MJUM9BYTES;
712		zone = zone_jumbo9;
713		break;
714	case EXT_JUMBO16:
715		size = MJUM16BYTES;
716		zone = zone_jumbo16;
717		break;
718	default:
719		panic("%s: unknown cluster type", __func__);
720		break;
721	}
722
723	m->m_data = m->m_ext.ext_buf = cl;
724	m->m_ext.ext_free = m->m_ext.ext_arg1 = m->m_ext.ext_arg2 = NULL;
725	m->m_ext.ext_size = size;
726	m->m_ext.ext_type = type;
727	m->m_ext.ext_flags = 0;
728	m->m_ext.ref_cnt = uma_find_refcnt(zone, cl);
729	m->m_flags |= M_EXT;
730
731}
732
733static __inline void
734m_chtype(struct mbuf *m, short new_type)
735{
736
737	m->m_type = new_type;
738}
739
740static __inline void
741m_clrprotoflags(struct mbuf *m)
742{
743
744	m->m_flags &= ~M_PROTOFLAGS;
745}
746
747static __inline struct mbuf *
748m_last(struct mbuf *m)
749{
750
751	while (m->m_next)
752		m = m->m_next;
753	return (m);
754}
755
756/*
757 * mbuf, cluster, and external object allocation macros (for compatibility
758 * purposes).
759 */
760#define	M_MOVE_PKTHDR(to, from)	m_move_pkthdr((to), (from))
761#define	MGET(m, how, type)	((m) = m_get((how), (type)))
762#define	MGETHDR(m, how, type)	((m) = m_gethdr((how), (type)))
763#define	MCLGET(m, how)		m_clget((m), (how))
764#define	MEXTADD(m, buf, size, free, arg1, arg2, flags, type)		\
765    (void )m_extadd((m), (caddr_t)(buf), (size), (free), (arg1), (arg2),\
766    (flags), (type), M_NOWAIT)
767#define	m_getm(m, len, how, type)					\
768    m_getm2((m), (len), (how), (type), M_PKTHDR)
769
770/*
771 * Evaluate TRUE if it's safe to write to the mbuf m's data region (this can
772 * be both the local data payload, or an external buffer area, depending on
773 * whether M_EXT is set).
774 */
775#define	M_WRITABLE(m)	(!((m)->m_flags & M_RDONLY) &&			\
776			 (!(((m)->m_flags & M_EXT)) ||			\
777			 (*((m)->m_ext.ref_cnt) == 1)) )		\
778
779/* Check if the supplied mbuf has a packet header, or else panic. */
780#define	M_ASSERTPKTHDR(m)						\
781	KASSERT((m) != NULL && (m)->m_flags & M_PKTHDR,			\
782	    ("%s: no mbuf packet header!", __func__))
783
784/*
785 * Ensure that the supplied mbuf is a valid, non-free mbuf.
786 *
787 * XXX: Broken at the moment.  Need some UMA magic to make it work again.
788 */
789#define	M_ASSERTVALID(m)						\
790	KASSERT((((struct mbuf *)m)->m_flags & 0) == 0,			\
791	    ("%s: attempted use of a free mbuf!", __func__))
792
793/*
794 * Set the m_data pointer of a newly-allocated mbuf (m_get/MGET) to place an
795 * object of the specified size at the end of the mbuf, longword aligned.
796 */
797#define	M_ALIGN(m, len) do {						\
798	KASSERT(!((m)->m_flags & (M_PKTHDR|M_EXT)),			\
799		("%s: M_ALIGN not normal mbuf", __func__));		\
800	KASSERT((m)->m_data == (m)->m_dat,				\
801		("%s: M_ALIGN not a virgin mbuf", __func__));		\
802	(m)->m_data += (MLEN - (len)) & ~(sizeof(long) - 1);		\
803} while (0)
804
805/*
806 * As above, for mbufs allocated with m_gethdr/MGETHDR or initialized by
807 * M_DUP/MOVE_PKTHDR.
808 */
809#define	MH_ALIGN(m, len) do {						\
810	KASSERT((m)->m_flags & M_PKTHDR && !((m)->m_flags & M_EXT),	\
811		("%s: MH_ALIGN not PKTHDR mbuf", __func__));		\
812	KASSERT((m)->m_data == (m)->m_pktdat,				\
813		("%s: MH_ALIGN not a virgin mbuf", __func__));		\
814	(m)->m_data += (MHLEN - (len)) & ~(sizeof(long) - 1);		\
815} while (0)
816
817/*
818 * As above, for mbuf with external storage.
819 */
820#define	MEXT_ALIGN(m, len) do {						\
821	KASSERT((m)->m_flags & M_EXT,					\
822		("%s: MEXT_ALIGN not an M_EXT mbuf", __func__));	\
823	KASSERT((m)->m_data == (m)->m_ext.ext_buf,			\
824		("%s: MEXT_ALIGN not a virgin mbuf", __func__));	\
825	(m)->m_data += ((m)->m_ext.ext_size - (len)) &			\
826	    ~(sizeof(long) - 1); 					\
827} while (0)
828
829/*
830 * Compute the amount of space available before the current start of data in
831 * an mbuf.
832 *
833 * The M_WRITABLE() is a temporary, conservative safety measure: the burden
834 * of checking writability of the mbuf data area rests solely with the caller.
835 */
836#define	M_LEADINGSPACE(m)						\
837	((m)->m_flags & M_EXT ?						\
838	    (M_WRITABLE(m) ? (m)->m_data - (m)->m_ext.ext_buf : 0):	\
839	    (m)->m_flags & M_PKTHDR ? (m)->m_data - (m)->m_pktdat :	\
840	    (m)->m_data - (m)->m_dat)
841
842/*
843 * Compute the amount of space available after the end of data in an mbuf.
844 *
845 * The M_WRITABLE() is a temporary, conservative safety measure: the burden
846 * of checking writability of the mbuf data area rests solely with the caller.
847 */
848#define	M_TRAILINGSPACE(m)						\
849	((m)->m_flags & M_EXT ?						\
850	    (M_WRITABLE(m) ? (m)->m_ext.ext_buf + (m)->m_ext.ext_size	\
851		- ((m)->m_data + (m)->m_len) : 0) :			\
852	    &(m)->m_dat[MLEN] - ((m)->m_data + (m)->m_len))
853
854/*
855 * Arrange to prepend space of size plen to mbuf m.  If a new mbuf must be
856 * allocated, how specifies whether to wait.  If the allocation fails, the
857 * original mbuf chain is freed and m is set to NULL.
858 */
859#define	M_PREPEND(m, plen, how) do {					\
860	struct mbuf **_mmp = &(m);					\
861	struct mbuf *_mm = *_mmp;					\
862	int _mplen = (plen);						\
863	int __mhow = (how);						\
864									\
865	MBUF_CHECKSLEEP(how);						\
866	if (M_LEADINGSPACE(_mm) >= _mplen) {				\
867		_mm->m_data -= _mplen;					\
868		_mm->m_len += _mplen;					\
869	} else								\
870		_mm = m_prepend(_mm, _mplen, __mhow);			\
871	if (_mm != NULL && _mm->m_flags & M_PKTHDR)			\
872		_mm->m_pkthdr.len += _mplen;				\
873	*_mmp = _mm;							\
874} while (0)
875
876/*
877 * Change mbuf to new type.  This is a relatively expensive operation and
878 * should be avoided.
879 */
880#define	MCHTYPE(m, t)	m_chtype((m), (t))
881
882/* Length to m_copy to copy all. */
883#define	M_COPYALL	1000000000
884
885/* Compatibility with 4.3. */
886#define	m_copy(m, o, l)	m_copym((m), (o), (l), M_NOWAIT)
887
888extern int		max_datalen;	/* MHLEN - max_hdr */
889extern int		max_hdr;	/* Largest link + protocol header */
890extern int		max_linkhdr;	/* Largest link-level header */
891extern int		max_protohdr;	/* Largest protocol header */
892extern int		nmbclusters;	/* Maximum number of clusters */
893
894struct uio;
895
896void		 m_adj(struct mbuf *, int);
897void		 m_align(struct mbuf *, int);
898int		 m_apply(struct mbuf *, int, int,
899		    int (*)(void *, void *, u_int), void *);
900int		 m_append(struct mbuf *, int, c_caddr_t);
901void		 m_cat(struct mbuf *, struct mbuf *);
902int		 m_extadd(struct mbuf *, caddr_t, u_int,
903		    int (*)(struct mbuf *, void *, void *), void *, void *,
904		    int, int, int);
905struct mbuf	*m_collapse(struct mbuf *, int, int);
906void		 m_copyback(struct mbuf *, int, int, c_caddr_t);
907void		 m_copydata(const struct mbuf *, int, int, caddr_t);
908struct mbuf	*m_copym(struct mbuf *, int, int, int);
909struct mbuf	*m_copymdata(struct mbuf *, struct mbuf *,
910		    int, int, int, int);
911struct mbuf	*m_copypacket(struct mbuf *, int);
912void		 m_copy_pkthdr(struct mbuf *, struct mbuf *);
913struct mbuf	*m_copyup(struct mbuf *, int, int);
914struct mbuf	*m_defrag(struct mbuf *, int);
915void		 m_demote(struct mbuf *, int);
916struct mbuf	*m_devget(char *, int, int, struct ifnet *,
917		    void (*)(char *, caddr_t, u_int));
918struct mbuf	*m_dup(struct mbuf *, int);
919int		 m_dup_pkthdr(struct mbuf *, struct mbuf *, int);
920u_int		 m_fixhdr(struct mbuf *);
921struct mbuf	*m_fragment(struct mbuf *, int, int);
922void		 m_freem(struct mbuf *);
923struct mbuf	*m_get2(int, int, short, int);
924struct mbuf	*m_getjcl(int, short, int, int);
925struct mbuf	*m_getm2(struct mbuf *, int, int, short, int);
926struct mbuf	*m_getptr(struct mbuf *, int, int *);
927u_int		 m_length(struct mbuf *, struct mbuf **);
928int		 m_mbuftouio(struct uio *, struct mbuf *, int);
929void		 m_move_pkthdr(struct mbuf *, struct mbuf *);
930struct mbuf	*m_prepend(struct mbuf *, int, int);
931void		 m_print(const struct mbuf *, int);
932struct mbuf	*m_pulldown(struct mbuf *, int, int, int *);
933struct mbuf	*m_pullup(struct mbuf *, int);
934int		 m_sanity(struct mbuf *, int);
935struct mbuf	*m_split(struct mbuf *, int, int);
936struct mbuf	*m_uiotombuf(struct uio *, int, int, int, int);
937struct mbuf	*m_unshare(struct mbuf *, int);
938
939/*-
940 * Network packets may have annotations attached by affixing a list of
941 * "packet tags" to the pkthdr structure.  Packet tags are dynamically
942 * allocated semi-opaque data structures that have a fixed header
943 * (struct m_tag) that specifies the size of the memory block and a
944 * <cookie,type> pair that identifies it.  The cookie is a 32-bit unique
945 * unsigned value used to identify a module or ABI.  By convention this value
946 * is chosen as the date+time that the module is created, expressed as the
947 * number of seconds since the epoch (e.g., using date -u +'%s').  The type
948 * value is an ABI/module-specific value that identifies a particular
949 * annotation and is private to the module.  For compatibility with systems
950 * like OpenBSD that define packet tags w/o an ABI/module cookie, the value
951 * PACKET_ABI_COMPAT is used to implement m_tag_get and m_tag_find
952 * compatibility shim functions and several tag types are defined below.
953 * Users that do not require compatibility should use a private cookie value
954 * so that packet tag-related definitions can be maintained privately.
955 *
956 * Note that the packet tag returned by m_tag_alloc has the default memory
957 * alignment implemented by malloc.  To reference private data one can use a
958 * construct like:
959 *
960 *	struct m_tag *mtag = m_tag_alloc(...);
961 *	struct foo *p = (struct foo *)(mtag+1);
962 *
963 * if the alignment of struct m_tag is sufficient for referencing members of
964 * struct foo.  Otherwise it is necessary to embed struct m_tag within the
965 * private data structure to insure proper alignment; e.g.,
966 *
967 *	struct foo {
968 *		struct m_tag	tag;
969 *		...
970 *	};
971 *	struct foo *p = (struct foo *) m_tag_alloc(...);
972 *	struct m_tag *mtag = &p->tag;
973 */
974
975/*
976 * Persistent tags stay with an mbuf until the mbuf is reclaimed.  Otherwise
977 * tags are expected to ``vanish'' when they pass through a network
978 * interface.  For most interfaces this happens normally as the tags are
979 * reclaimed when the mbuf is free'd.  However in some special cases
980 * reclaiming must be done manually.  An example is packets that pass through
981 * the loopback interface.  Also, one must be careful to do this when
982 * ``turning around'' packets (e.g., icmp_reflect).
983 *
984 * To mark a tag persistent bit-or this flag in when defining the tag id.
985 * The tag will then be treated as described above.
986 */
987#define	MTAG_PERSISTENT				0x800
988
989#define	PACKET_TAG_NONE				0  /* Nadda */
990
991/* Packet tags for use with PACKET_ABI_COMPAT. */
992#define	PACKET_TAG_IPSEC_IN_DONE		1  /* IPsec applied, in */
993#define	PACKET_TAG_IPSEC_OUT_DONE		2  /* IPsec applied, out */
994#define	PACKET_TAG_IPSEC_IN_CRYPTO_DONE		3  /* NIC IPsec crypto done */
995#define	PACKET_TAG_IPSEC_OUT_CRYPTO_NEEDED	4  /* NIC IPsec crypto req'ed */
996#define	PACKET_TAG_IPSEC_IN_COULD_DO_CRYPTO	5  /* NIC notifies IPsec */
997#define	PACKET_TAG_IPSEC_PENDING_TDB		6  /* Reminder to do IPsec */
998#define	PACKET_TAG_BRIDGE			7  /* Bridge processing done */
999#define	PACKET_TAG_GIF				8  /* GIF processing done */
1000#define	PACKET_TAG_GRE				9  /* GRE processing done */
1001#define	PACKET_TAG_IN_PACKET_CHECKSUM		10 /* NIC checksumming done */
1002#define	PACKET_TAG_ENCAP			11 /* Encap.  processing */
1003#define	PACKET_TAG_IPSEC_SOCKET			12 /* IPSEC socket ref */
1004#define	PACKET_TAG_IPSEC_HISTORY		13 /* IPSEC history */
1005#define	PACKET_TAG_IPV6_INPUT			14 /* IPV6 input processing */
1006#define	PACKET_TAG_DUMMYNET			15 /* dummynet info */
1007#define	PACKET_TAG_DIVERT			17 /* divert info */
1008#define	PACKET_TAG_IPFORWARD			18 /* ipforward info */
1009#define	PACKET_TAG_MACLABEL	(19 | MTAG_PERSISTENT) /* MAC label */
1010#define	PACKET_TAG_PF		(21 | MTAG_PERSISTENT) /* PF/ALTQ information */
1011#define	PACKET_TAG_RTSOCKFAM			25 /* rtsock sa family */
1012#define	PACKET_TAG_IPOPTIONS			27 /* Saved IP options */
1013#define	PACKET_TAG_CARP				28 /* CARP info */
1014#define	PACKET_TAG_IPSEC_NAT_T_PORTS		29 /* two uint16_t */
1015#define	PACKET_TAG_ND_OUTGOING			30 /* ND outgoing */
1016
1017/* Specific cookies and tags. */
1018
1019/* Packet tag routines. */
1020struct m_tag	*m_tag_alloc(u_int32_t, int, int, int);
1021void		 m_tag_delete(struct mbuf *, struct m_tag *);
1022void		 m_tag_delete_chain(struct mbuf *, struct m_tag *);
1023void		 m_tag_free_default(struct m_tag *);
1024struct m_tag	*m_tag_locate(struct mbuf *, u_int32_t, int, struct m_tag *);
1025struct m_tag	*m_tag_copy(struct m_tag *, int);
1026int		 m_tag_copy_chain(struct mbuf *, struct mbuf *, int);
1027void		 m_tag_delete_nonpersistent(struct mbuf *);
1028
1029/*
1030 * Initialize the list of tags associated with an mbuf.
1031 */
1032static __inline void
1033m_tag_init(struct mbuf *m)
1034{
1035
1036	SLIST_INIT(&m->m_pkthdr.tags);
1037}
1038
1039/*
1040 * Set up the contents of a tag.  Note that this does not fill in the free
1041 * method; the caller is expected to do that.
1042 *
1043 * XXX probably should be called m_tag_init, but that was already taken.
1044 */
1045static __inline void
1046m_tag_setup(struct m_tag *t, u_int32_t cookie, int type, int len)
1047{
1048
1049	t->m_tag_id = type;
1050	t->m_tag_len = len;
1051	t->m_tag_cookie = cookie;
1052}
1053
1054/*
1055 * Reclaim resources associated with a tag.
1056 */
1057static __inline void
1058m_tag_free(struct m_tag *t)
1059{
1060
1061	(*t->m_tag_free)(t);
1062}
1063
1064/*
1065 * Return the first tag associated with an mbuf.
1066 */
1067static __inline struct m_tag *
1068m_tag_first(struct mbuf *m)
1069{
1070
1071	return (SLIST_FIRST(&m->m_pkthdr.tags));
1072}
1073
1074/*
1075 * Return the next tag in the list of tags associated with an mbuf.
1076 */
1077static __inline struct m_tag *
1078m_tag_next(struct mbuf *m, struct m_tag *t)
1079{
1080
1081	return (SLIST_NEXT(t, m_tag_link));
1082}
1083
1084/*
1085 * Prepend a tag to the list of tags associated with an mbuf.
1086 */
1087static __inline void
1088m_tag_prepend(struct mbuf *m, struct m_tag *t)
1089{
1090
1091	SLIST_INSERT_HEAD(&m->m_pkthdr.tags, t, m_tag_link);
1092}
1093
1094/*
1095 * Unlink a tag from the list of tags associated with an mbuf.
1096 */
1097static __inline void
1098m_tag_unlink(struct mbuf *m, struct m_tag *t)
1099{
1100
1101	SLIST_REMOVE(&m->m_pkthdr.tags, t, m_tag, m_tag_link);
1102}
1103
1104/* These are for OpenBSD compatibility. */
1105#define	MTAG_ABI_COMPAT		0		/* compatibility ABI */
1106
1107static __inline struct m_tag *
1108m_tag_get(int type, int length, int wait)
1109{
1110	return (m_tag_alloc(MTAG_ABI_COMPAT, type, length, wait));
1111}
1112
1113static __inline struct m_tag *
1114m_tag_find(struct mbuf *m, int type, struct m_tag *start)
1115{
1116	return (SLIST_EMPTY(&m->m_pkthdr.tags) ? (struct m_tag *)NULL :
1117	    m_tag_locate(m, MTAG_ABI_COMPAT, type, start));
1118}
1119
1120static int inline
1121rt_m_getfib(struct mbuf *m)
1122{
1123	KASSERT(m->m_flags & M_PKTHDR , ("Attempt to get FIB from non header mbuf."));
1124	return (m->m_pkthdr.fibnum);
1125}
1126
1127#define M_GETFIB(_m)   rt_m_getfib(_m)
1128
1129#define M_SETFIB(_m, _fib) do {						\
1130        KASSERT((_m)->m_flags & M_PKTHDR, ("Attempt to set FIB on non header mbuf."));	\
1131	((_m)->m_pkthdr.fibnum) = (_fib);				\
1132} while (0)
1133
1134#endif /* _KERNEL */
1135
1136#ifdef MBUF_PROFILING
1137 void m_profile(struct mbuf *m);
1138 #define M_PROFILE(m) m_profile(m)
1139#else
1140 #define M_PROFILE(m)
1141#endif
1142
1143
1144#endif /* !_SYS_MBUF_H_ */
1145