mbuf.h revision 281955
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: stable/10/sys/sys/mbuf.h 281955 2015-04-24 23:26:44Z hiren $
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 * NB: These calculation do not take actual compiler-induced alignment and
57 * padding inside the complete struct mbuf into account.  Appropriate
58 * attention is required when changing members of struct mbuf.
59 *
60 * MLEN is data length in a normal mbuf.
61 * MHLEN is data length in an mbuf with pktheader.
62 * MINCLSIZE is a smallest amount of data that should be put into cluster.
63 */
64#define	MLEN		((int)(MSIZE - sizeof(struct m_hdr)))
65#define	MHLEN		((int)(MLEN - sizeof(struct pkthdr)))
66#define	MINCLSIZE	(MHLEN + 1)
67
68#ifdef _KERNEL
69/*-
70 * Macro for type conversion: convert mbuf pointer to data pointer of correct
71 * type:
72 *
73 * mtod(m, t)	-- Convert mbuf pointer to data pointer of correct type.
74 * mtodo(m, o) -- Same as above but with offset 'o' into data.
75 */
76#define	mtod(m, t)	((t)((m)->m_data))
77#define	mtodo(m, o)	((void *)(((m)->m_data) + (o)))
78
79/*
80 * Argument structure passed to UMA routines during mbuf and packet
81 * allocations.
82 */
83struct mb_args {
84	int	flags;	/* Flags for mbuf being allocated */
85	short	type;	/* Type of mbuf being allocated */
86};
87#endif /* _KERNEL */
88
89/*
90 * Header present at the beginning of every mbuf.
91 * Size ILP32: 24
92 *	 LP64: 32
93 */
94struct m_hdr {
95	struct mbuf	*mh_next;	/* next buffer in chain */
96	struct mbuf	*mh_nextpkt;	/* next chain in queue/record */
97	caddr_t		 mh_data;	/* location of data */
98	int32_t		 mh_len;	/* amount of data in this mbuf */
99	uint32_t	 mh_type:8,	/* type of data in this mbuf */
100			 mh_flags:24;	/* flags; see below */
101#if !defined(__LP64__)
102	uint32_t	 mh_pad;	/* pad for 64bit alignment */
103#endif
104};
105
106/*
107 * Packet tag structure (see below for details).
108 */
109struct m_tag {
110	SLIST_ENTRY(m_tag)	m_tag_link;	/* List of packet tags */
111	u_int16_t		m_tag_id;	/* Tag ID */
112	u_int16_t		m_tag_len;	/* Length of data */
113	u_int32_t		m_tag_cookie;	/* ABI/Module ID */
114	void			(*m_tag_free)(struct m_tag *);
115};
116
117/*
118 * Record/packet header in first mbuf of chain; valid only if M_PKTHDR is set.
119 * Size ILP32: 48
120 *	 LP64: 56
121 */
122struct pkthdr {
123	struct ifnet	*rcvif;		/* rcv interface */
124	SLIST_HEAD(packet_tags, m_tag) tags; /* list of packet tags */
125	int32_t		 len;		/* total packet length */
126
127	/* Layer crossing persistent information. */
128	uint32_t	 flowid;	/* packet's 4-tuple system */
129	uint64_t	 csum_flags;	/* checksum and offload features */
130	uint16_t	 fibnum;	/* this packet should use this fib */
131	uint8_t		 cosqos;	/* class/quality of service */
132	uint8_t		 rsstype;	/* hash type */
133	uint8_t		 l2hlen;	/* layer 2 header length */
134	uint8_t		 l3hlen;	/* layer 3 header length */
135	uint8_t		 l4hlen;	/* layer 4 header length */
136	uint8_t		 l5hlen;	/* layer 5 header length */
137	union {
138		uint8_t  eigth[8];
139		uint16_t sixteen[4];
140		uint32_t thirtytwo[2];
141		uint64_t sixtyfour[1];
142		uintptr_t unintptr[1];
143		void	*ptr;
144	} PH_per;
145
146	/* Layer specific non-persistent local storage for reassembly, etc. */
147	union {
148		uint8_t  eigth[8];
149		uint16_t sixteen[4];
150		uint32_t thirtytwo[2];
151		uint64_t sixtyfour[1];
152		uintptr_t unintptr[1];
153		void 	*ptr;
154	} PH_loc;
155};
156#define	ether_vtag	PH_per.sixteen[0]
157#define	PH_vt		PH_per
158#define	vt_nrecs	sixteen[0]
159#define	tso_segsz	PH_per.sixteen[1]
160#define	csum_phsum	PH_per.sixteen[2]
161#define	csum_data	PH_per.thirtytwo[1]
162
163/*
164 * Description of external storage mapped into mbuf; valid only if M_EXT is
165 * set.
166 * Size ILP32: 28
167 *	 LP64: 48
168 */
169struct m_ext {
170	volatile u_int	*ref_cnt;	/* pointer to ref count info */
171	caddr_t		 ext_buf;	/* start of buffer */
172	uint32_t	 ext_size;	/* size of buffer, for ext_free */
173	uint32_t	 ext_type:8,	/* type of external storage */
174			 ext_flags:24;	/* external storage mbuf flags */
175	int		(*ext_free)	/* free routine if not the usual */
176			    (struct mbuf *, void *, void *);
177	void		*ext_arg1;	/* optional argument pointer */
178	void		*ext_arg2;	/* optional argument pointer */
179};
180
181/*
182 * The core of the mbuf object along with some shortcut defines for practical
183 * purposes.
184 */
185struct mbuf {
186	struct m_hdr	m_hdr;
187	union {
188		struct {
189			struct pkthdr	MH_pkthdr;	/* M_PKTHDR set */
190			union {
191				struct m_ext	MH_ext;	/* M_EXT set */
192				char		MH_databuf[MHLEN];
193			} MH_dat;
194		} MH;
195		char	M_databuf[MLEN];		/* !M_PKTHDR, !M_EXT */
196	} M_dat;
197};
198#define	m_next		m_hdr.mh_next
199#define	m_len		m_hdr.mh_len
200#define	m_data		m_hdr.mh_data
201#define	m_type		m_hdr.mh_type
202#define	m_flags		m_hdr.mh_flags
203#define	m_nextpkt	m_hdr.mh_nextpkt
204#define	m_pkthdr	M_dat.MH.MH_pkthdr
205#define	m_ext		M_dat.MH.MH_dat.MH_ext
206#define	m_pktdat	M_dat.MH.MH_dat.MH_databuf
207#define	m_dat		M_dat.M_databuf
208
209/*
210 * NOTE: forwards compatibility definitions for mbuf(9)
211 *
212 * These aren't 1:1 with the macros in r277203; in particular they're exposed
213 * to both userland and kernel, whereas this is exposed to just _KERNEL -- to
214 * avoid disruption with existing KBI/KPIs
215 */
216#define	MHSIZE		offsetof(struct mbuf, m_dat)
217#define	MPKTHSIZE	offsetof(struct mbuf, m_pktdat)
218
219/*
220 * mbuf flags of global significance and layer crossing.
221 * Those of only protocol/layer specific significance are to be mapped
222 * to M_PROTO[1-12] and cleared at layer handoff boundaries.
223 * NB: Limited to the lower 24 bits.
224 */
225#define	M_EXT		0x00000001 /* has associated external storage */
226#define	M_PKTHDR	0x00000002 /* start of record */
227#define	M_EOR		0x00000004 /* end of record */
228#define	M_RDONLY	0x00000008 /* associated data is marked read-only */
229#define	M_BCAST		0x00000010 /* send/received as link-level broadcast */
230#define	M_MCAST		0x00000020 /* send/received as link-level multicast */
231#define	M_PROMISC	0x00000040 /* packet was not for us */
232#define	M_VLANTAG	0x00000080 /* ether_vtag is valid */
233#define	M_UNUSED_8	0x00000100 /* --available-- */
234#define	M_NOFREE	0x00000200 /* do not free mbuf, embedded in cluster */
235
236#define	M_PROTO1	0x00001000 /* protocol-specific */
237#define	M_PROTO2	0x00002000 /* protocol-specific */
238#define	M_PROTO3	0x00004000 /* protocol-specific */
239#define	M_PROTO4	0x00008000 /* protocol-specific */
240#define	M_PROTO5	0x00010000 /* protocol-specific */
241#define	M_PROTO6	0x00020000 /* protocol-specific */
242#define	M_PROTO7	0x00040000 /* protocol-specific */
243#define	M_PROTO8	0x00080000 /* protocol-specific */
244#define	M_PROTO9	0x00100000 /* protocol-specific */
245#define	M_PROTO10	0x00200000 /* protocol-specific */
246#define	M_PROTO11	0x00400000 /* protocol-specific */
247#define	M_PROTO12	0x00800000 /* protocol-specific */
248
249/*
250 * Flags to purge when crossing layers.
251 */
252#define	M_PROTOFLAGS \
253    (M_PROTO1|M_PROTO2|M_PROTO3|M_PROTO4|M_PROTO5|M_PROTO6|M_PROTO7|M_PROTO8|\
254     M_PROTO9|M_PROTO10|M_PROTO11|M_PROTO12)
255
256/*
257 * Flags preserved when copying m_pkthdr.
258 */
259#define M_COPYFLAGS \
260    (M_PKTHDR|M_EOR|M_RDONLY|M_BCAST|M_MCAST|M_PROMISC|M_VLANTAG| \
261     M_PROTOFLAGS)
262
263/*
264 * Mbuf flag description for use with printf(9) %b identifier.
265 */
266#define	M_FLAG_BITS \
267    "\20\1M_EXT\2M_PKTHDR\3M_EOR\4M_RDONLY\5M_BCAST\6M_MCAST" \
268    "\7M_PROMISC\10M_VLANTAG"
269#define	M_FLAG_PROTOBITS \
270    "\15M_PROTO1\16M_PROTO2\17M_PROTO3\20M_PROTO4\21M_PROTO5" \
271    "\22M_PROTO6\23M_PROTO7\24M_PROTO8\25M_PROTO9\26M_PROTO10" \
272    "\27M_PROTO11\30M_PROTO12"
273#define	M_FLAG_PRINTF (M_FLAG_BITS M_FLAG_PROTOBITS)
274
275/*
276 * Network interface cards are able to hash protocol fields (such as IPv4
277 * addresses and TCP port numbers) classify packets into flows.  These flows
278 * can then be used to maintain ordering while delivering packets to the OS
279 * via parallel input queues, as well as to provide a stateless affinity
280 * model.  NIC drivers can pass up the hash via m->m_pkthdr.flowid, and set
281 * m_flag fields to indicate how the hash should be interpreted by the
282 * network stack.
283 *
284 * Most NICs support RSS, which provides ordering and explicit affinity, and
285 * use the hash m_flag bits to indicate what header fields were covered by
286 * the hash.  M_HASHTYPE_OPAQUE can be set by non-RSS cards or configurations
287 * that provide an opaque flow identifier, allowing for ordering and
288 * distribution without explicit affinity.
289 */
290#define	M_HASHTYPE_NONE			0
291#define	M_HASHTYPE_RSS_IPV4		1	/* IPv4 2-tuple */
292#define	M_HASHTYPE_RSS_TCP_IPV4		2	/* TCPv4 4-tuple */
293#define	M_HASHTYPE_RSS_IPV6		3	/* IPv6 2-tuple */
294#define	M_HASHTYPE_RSS_TCP_IPV6		4	/* TCPv6 4-tuple */
295#define	M_HASHTYPE_RSS_IPV6_EX		5	/* IPv6 2-tuple + ext hdrs */
296#define	M_HASHTYPE_RSS_TCP_IPV6_EX	6	/* TCPv6 4-tiple + ext hdrs */
297#define	M_HASHTYPE_OPAQUE		255	/* ordering, not affinity */
298
299#define	M_HASHTYPE_CLEAR(m)	((m)->m_pkthdr.rsstype = 0)
300#define	M_HASHTYPE_GET(m)	((m)->m_pkthdr.rsstype)
301#define	M_HASHTYPE_SET(m, v)	((m)->m_pkthdr.rsstype = (v))
302#define	M_HASHTYPE_TEST(m, v)	(M_HASHTYPE_GET(m) == (v))
303
304/*
305 * COS/QOS class and quality of service tags.
306 * It uses DSCP code points as base.
307 */
308#define	QOS_DSCP_CS0		0x00
309#define	QOS_DSCP_DEF		QOS_DSCP_CS0
310#define	QOS_DSCP_CS1		0x20
311#define	QOS_DSCP_AF11		0x28
312#define	QOS_DSCP_AF12		0x30
313#define	QOS_DSCP_AF13		0x38
314#define	QOS_DSCP_CS2		0x40
315#define	QOS_DSCP_AF21		0x48
316#define	QOS_DSCP_AF22		0x50
317#define	QOS_DSCP_AF23		0x58
318#define	QOS_DSCP_CS3		0x60
319#define	QOS_DSCP_AF31		0x68
320#define	QOS_DSCP_AF32		0x70
321#define	QOS_DSCP_AF33		0x78
322#define	QOS_DSCP_CS4		0x80
323#define	QOS_DSCP_AF41		0x88
324#define	QOS_DSCP_AF42		0x90
325#define	QOS_DSCP_AF43		0x98
326#define	QOS_DSCP_CS5		0xa0
327#define	QOS_DSCP_EF		0xb8
328#define	QOS_DSCP_CS6		0xc0
329#define	QOS_DSCP_CS7		0xe0
330
331/*
332 * External mbuf storage buffer types.
333 */
334#define	EXT_CLUSTER	1	/* mbuf cluster */
335#define	EXT_SFBUF	2	/* sendfile(2)'s sf_bufs */
336#define	EXT_JUMBOP	3	/* jumbo cluster 4096 bytes */
337#define	EXT_JUMBO9	4	/* jumbo cluster 9216 bytes */
338#define	EXT_JUMBO16	5	/* jumbo cluster 16184 bytes */
339#define	EXT_PACKET	6	/* mbuf+cluster from packet zone */
340#define	EXT_MBUF	7	/* external mbuf reference (M_IOVEC) */
341
342#define	EXT_VENDOR1	224	/* for vendor-internal use */
343#define	EXT_VENDOR2	225	/* for vendor-internal use */
344#define	EXT_VENDOR3	226	/* for vendor-internal use */
345#define	EXT_VENDOR4	227	/* for vendor-internal use */
346
347#define	EXT_EXP1	244	/* for experimental use */
348#define	EXT_EXP2	245	/* for experimental use */
349#define	EXT_EXP3	246	/* for experimental use */
350#define	EXT_EXP4	247	/* for experimental use */
351
352#define	EXT_NET_DRV	252	/* custom ext_buf provided by net driver(s) */
353#define	EXT_MOD_TYPE	253	/* custom module's ext_buf type */
354#define	EXT_DISPOSABLE	254	/* can throw this buffer away w/page flipping */
355#define	EXT_EXTREF	255	/* has externally maintained ref_cnt ptr */
356
357/*
358 * Flags for external mbuf buffer types.
359 * NB: limited to the lower 24 bits.
360 */
361#define	EXT_FLAG_EMBREF		0x000001	/* embedded ref_cnt, notyet */
362#define	EXT_FLAG_EXTREF		0x000002	/* external ref_cnt, notyet */
363#define	EXT_FLAG_NOFREE		0x000010	/* don't free mbuf to pool, notyet */
364
365#define	EXT_FLAG_VENDOR1	0x010000	/* for vendor-internal use */
366#define	EXT_FLAG_VENDOR2	0x020000	/* for vendor-internal use */
367#define	EXT_FLAG_VENDOR3	0x040000	/* for vendor-internal use */
368#define	EXT_FLAG_VENDOR4	0x080000	/* for vendor-internal use */
369
370#define	EXT_FLAG_EXP1		0x100000	/* for experimental use */
371#define	EXT_FLAG_EXP2		0x200000	/* for experimental use */
372#define	EXT_FLAG_EXP3		0x400000	/* for experimental use */
373#define	EXT_FLAG_EXP4		0x800000	/* for experimental use */
374
375/*
376 * EXT flag description for use with printf(9) %b identifier.
377 */
378#define	EXT_FLAG_BITS \
379    "\20\1EXT_FLAG_EMBREF\2EXT_FLAG_EXTREF\5EXT_FLAG_NOFREE" \
380    "\21EXT_FLAG_VENDOR1\22EXT_FLAG_VENDOR2\23EXT_FLAG_VENDOR3" \
381    "\24EXT_FLAG_VENDOR4\25EXT_FLAG_EXP1\26EXT_FLAG_EXP2\27EXT_FLAG_EXP3" \
382    "\30EXT_FLAG_EXP4"
383
384/*
385 * Return values for (*ext_free).
386 */
387#define	EXT_FREE_OK	0	/* Normal return */
388
389/*
390 * Flags indicating checksum, segmentation and other offload work to be
391 * done, or already done, by hardware or lower layers.  It is split into
392 * separate inbound and outbound flags.
393 *
394 * Outbound flags that are set by upper protocol layers requesting lower
395 * layers, or ideally the hardware, to perform these offloading tasks.
396 * For outbound packets this field and its flags can be directly tested
397 * against if_data.ifi_hwassist.
398 */
399#define	CSUM_IP			0x00000001	/* IP header checksum offload */
400#define	CSUM_IP_UDP		0x00000002	/* UDP checksum offload */
401#define	CSUM_IP_TCP		0x00000004	/* TCP checksum offload */
402#define	CSUM_IP_SCTP		0x00000008	/* SCTP checksum offload */
403#define	CSUM_IP_TSO		0x00000010	/* TCP segmentation offload */
404#define	CSUM_IP_ISCSI		0x00000020	/* iSCSI checksum offload */
405
406#define	CSUM_IP6_UDP		0x00000200	/* UDP checksum offload */
407#define	CSUM_IP6_TCP		0x00000400	/* TCP checksum offload */
408#define	CSUM_IP6_SCTP		0x00000800	/* SCTP checksum offload */
409#define	CSUM_IP6_TSO		0x00001000	/* TCP segmentation offload */
410#define	CSUM_IP6_ISCSI		0x00002000	/* iSCSI checksum offload */
411
412/* Inbound checksum support where the checksum was verified by hardware. */
413#define	CSUM_L3_CALC		0x01000000	/* calculated layer 3 csum */
414#define	CSUM_L3_VALID		0x02000000	/* checksum is correct */
415#define	CSUM_L4_CALC		0x04000000	/* calculated layer 4 csum */
416#define	CSUM_L4_VALID		0x08000000	/* checksum is correct */
417#define	CSUM_L5_CALC		0x10000000	/* calculated layer 5 csum */
418#define	CSUM_L5_VALID		0x20000000	/* checksum is correct */
419#define	CSUM_COALESED		0x40000000	/* contains merged segments */
420
421/*
422 * CSUM flag description for use with printf(9) %b identifier.
423 */
424#define	CSUM_BITS \
425    "\20\1CSUM_IP\2CSUM_IP_UDP\3CSUM_IP_TCP\4CSUM_IP_SCTP\5CSUM_IP_TSO" \
426    "\6CSUM_IP_ISCSI" \
427    "\12CSUM_IP6_UDP\13CSUM_IP6_TCP\14CSUM_IP6_SCTP\15CSUM_IP6_TSO" \
428    "\16CSUM_IP6_ISCSI" \
429    "\31CSUM_L3_CALC\32CSUM_L3_VALID\33CSUM_L4_CALC\34CSUM_L4_VALID" \
430    "\35CSUM_L5_CALC\36CSUM_L5_VALID\37CSUM_COALESED"
431
432/* CSUM flags compatibility mappings. */
433#define	CSUM_IP_CHECKED		CSUM_L3_CALC
434#define	CSUM_IP_VALID		CSUM_L3_VALID
435#define	CSUM_DATA_VALID		CSUM_L4_VALID
436#define	CSUM_PSEUDO_HDR		CSUM_L4_CALC
437#define	CSUM_SCTP_VALID		CSUM_L4_VALID
438#define	CSUM_DELAY_DATA		(CSUM_TCP|CSUM_UDP)
439#define	CSUM_DELAY_IP		CSUM_IP		/* Only v4, no v6 IP hdr csum */
440#define	CSUM_DELAY_DATA_IPV6	(CSUM_TCP_IPV6|CSUM_UDP_IPV6)
441#define	CSUM_DATA_VALID_IPV6	CSUM_DATA_VALID
442#define	CSUM_TCP		CSUM_IP_TCP
443#define	CSUM_UDP		CSUM_IP_UDP
444#define	CSUM_SCTP		CSUM_IP_SCTP
445#define	CSUM_TSO		(CSUM_IP_TSO|CSUM_IP6_TSO)
446#define	CSUM_UDP_IPV6		CSUM_IP6_UDP
447#define	CSUM_TCP_IPV6		CSUM_IP6_TCP
448#define	CSUM_SCTP_IPV6		CSUM_IP6_SCTP
449#define	CSUM_FRAGMENT		0x0		/* Unused */
450
451/*
452 * mbuf types describing the content of the mbuf (including external storage).
453 */
454#define	MT_NOTMBUF	0	/* USED INTERNALLY ONLY! Object is not mbuf */
455#define	MT_DATA		1	/* dynamic (data) allocation */
456#define	MT_HEADER	MT_DATA	/* packet header, use M_PKTHDR instead */
457
458#define	MT_VENDOR1	4	/* for vendor-internal use */
459#define	MT_VENDOR2	5	/* for vendor-internal use */
460#define	MT_VENDOR3	6	/* for vendor-internal use */
461#define	MT_VENDOR4	7	/* for vendor-internal use */
462
463#define	MT_SONAME	8	/* socket name */
464
465#define	MT_EXP1		9	/* for experimental use */
466#define	MT_EXP2		10	/* for experimental use */
467#define	MT_EXP3		11	/* for experimental use */
468#define	MT_EXP4		12	/* for experimental use */
469
470#define	MT_CONTROL	14	/* extra-data protocol message */
471#define	MT_OOBDATA	15	/* expedited data  */
472#define	MT_NTYPES	16	/* number of mbuf types for mbtypes[] */
473
474#define	MT_NOINIT	255	/* Not a type but a flag to allocate
475				   a non-initialized mbuf */
476
477/*
478 * Compatibility with historic mbuf allocator.
479 */
480#define	MBTOM(how)	(how)
481#define	M_DONTWAIT	M_NOWAIT
482#define	M_TRYWAIT	M_WAITOK
483#define	M_WAIT		M_WAITOK
484
485/*
486 * String names of mbuf-related UMA(9) and malloc(9) types.  Exposed to
487 * !_KERNEL so that monitoring tools can look up the zones with
488 * libmemstat(3).
489 */
490#define	MBUF_MEM_NAME		"mbuf"
491#define	MBUF_CLUSTER_MEM_NAME	"mbuf_cluster"
492#define	MBUF_PACKET_MEM_NAME	"mbuf_packet"
493#define	MBUF_JUMBOP_MEM_NAME	"mbuf_jumbo_page"
494#define	MBUF_JUMBO9_MEM_NAME	"mbuf_jumbo_9k"
495#define	MBUF_JUMBO16_MEM_NAME	"mbuf_jumbo_16k"
496#define	MBUF_TAG_MEM_NAME	"mbuf_tag"
497#define	MBUF_EXTREFCNT_MEM_NAME	"mbuf_ext_refcnt"
498
499#ifdef _KERNEL
500
501#ifdef WITNESS
502#define	MBUF_CHECKSLEEP(how) do {					\
503	if (how == M_WAITOK)						\
504		WITNESS_WARN(WARN_GIANTOK | WARN_SLEEPOK, NULL,		\
505		    "Sleeping in \"%s\"", __func__);			\
506} while (0)
507#else
508#define	MBUF_CHECKSLEEP(how)
509#endif
510
511/*
512 * Network buffer allocation API
513 *
514 * The rest of it is defined in kern/kern_mbuf.c
515 */
516extern uma_zone_t	zone_mbuf;
517extern uma_zone_t	zone_clust;
518extern uma_zone_t	zone_pack;
519extern uma_zone_t	zone_jumbop;
520extern uma_zone_t	zone_jumbo9;
521extern uma_zone_t	zone_jumbo16;
522extern uma_zone_t	zone_ext_refcnt;
523
524void		 mb_free_ext(struct mbuf *);
525int		 m_pkthdr_init(struct mbuf *, int);
526
527static __inline int
528m_gettype(int size)
529{
530	int type;
531
532	switch (size) {
533	case MSIZE:
534		type = EXT_MBUF;
535		break;
536	case MCLBYTES:
537		type = EXT_CLUSTER;
538		break;
539#if MJUMPAGESIZE != MCLBYTES
540	case MJUMPAGESIZE:
541		type = EXT_JUMBOP;
542		break;
543#endif
544	case MJUM9BYTES:
545		type = EXT_JUMBO9;
546		break;
547	case MJUM16BYTES:
548		type = EXT_JUMBO16;
549		break;
550	default:
551		panic("%s: invalid cluster size %d", __func__, size);
552	}
553
554	return (type);
555}
556
557/*
558 * Associated an external reference counted buffer with an mbuf.
559 */
560static __inline void
561m_extaddref(struct mbuf *m, caddr_t buf, u_int size, u_int *ref_cnt,
562    int (*freef)(struct mbuf *, void *, void *), void *arg1, void *arg2)
563{
564
565	KASSERT(ref_cnt != NULL, ("%s: ref_cnt not provided", __func__));
566
567	atomic_add_int(ref_cnt, 1);
568	m->m_flags |= M_EXT;
569	m->m_ext.ext_buf = buf;
570	m->m_ext.ref_cnt = ref_cnt;
571	m->m_data = m->m_ext.ext_buf;
572	m->m_ext.ext_size = size;
573	m->m_ext.ext_free = freef;
574	m->m_ext.ext_arg1 = arg1;
575	m->m_ext.ext_arg2 = arg2;
576	m->m_ext.ext_type = EXT_EXTREF;
577	m->m_ext.ext_flags = 0;
578}
579
580static __inline uma_zone_t
581m_getzone(int size)
582{
583	uma_zone_t zone;
584
585	switch (size) {
586	case MCLBYTES:
587		zone = zone_clust;
588		break;
589#if MJUMPAGESIZE != MCLBYTES
590	case MJUMPAGESIZE:
591		zone = zone_jumbop;
592		break;
593#endif
594	case MJUM9BYTES:
595		zone = zone_jumbo9;
596		break;
597	case MJUM16BYTES:
598		zone = zone_jumbo16;
599		break;
600	default:
601		panic("%s: invalid cluster size %d", __func__, size);
602	}
603
604	return (zone);
605}
606
607/*
608 * Initialize an mbuf with linear storage.
609 *
610 * Inline because the consumer text overhead will be roughly the same to
611 * initialize or call a function with this many parameters and M_PKTHDR
612 * should go away with constant propagation for !MGETHDR.
613 */
614static __inline int
615m_init(struct mbuf *m, uma_zone_t zone, int size, int how, short type,
616    int flags)
617{
618	int error;
619
620	m->m_next = NULL;
621	m->m_nextpkt = NULL;
622	m->m_data = m->m_dat;
623	m->m_len = 0;
624	m->m_flags = flags;
625	m->m_type = type;
626	if (flags & M_PKTHDR) {
627		if ((error = m_pkthdr_init(m, how)) != 0)
628			return (error);
629	}
630
631	return (0);
632}
633
634static __inline struct mbuf *
635m_get(int how, short type)
636{
637	struct mb_args args;
638
639	args.flags = 0;
640	args.type = type;
641	return (uma_zalloc_arg(zone_mbuf, &args, how));
642}
643
644/*
645 * XXX This should be deprecated, very little use.
646 */
647static __inline struct mbuf *
648m_getclr(int how, short type)
649{
650	struct mbuf *m;
651	struct mb_args args;
652
653	args.flags = 0;
654	args.type = type;
655	m = uma_zalloc_arg(zone_mbuf, &args, how);
656	if (m != NULL)
657		bzero(m->m_data, MLEN);
658	return (m);
659}
660
661static __inline struct mbuf *
662m_gethdr(int how, short type)
663{
664	struct mb_args args;
665
666	args.flags = M_PKTHDR;
667	args.type = type;
668	return (uma_zalloc_arg(zone_mbuf, &args, how));
669}
670
671static __inline struct mbuf *
672m_getcl(int how, short type, int flags)
673{
674	struct mb_args args;
675
676	args.flags = flags;
677	args.type = type;
678	return (uma_zalloc_arg(zone_pack, &args, how));
679}
680
681static __inline void
682m_clget(struct mbuf *m, int how)
683{
684
685	if (m->m_flags & M_EXT)
686		printf("%s: %p mbuf already has cluster\n", __func__, m);
687	m->m_ext.ext_buf = (char *)NULL;
688	uma_zalloc_arg(zone_clust, m, how);
689	/*
690	 * On a cluster allocation failure, drain the packet zone and retry,
691	 * we might be able to loosen a few clusters up on the drain.
692	 */
693	if ((how & M_NOWAIT) && (m->m_ext.ext_buf == NULL)) {
694		zone_drain(zone_pack);
695		uma_zalloc_arg(zone_clust, m, how);
696	}
697}
698
699/*
700 * m_cljget() is different from m_clget() as it can allocate clusters without
701 * attaching them to an mbuf.  In that case the return value is the pointer
702 * to the cluster of the requested size.  If an mbuf was specified, it gets
703 * the cluster attached to it and the return value can be safely ignored.
704 * For size it takes MCLBYTES, MJUMPAGESIZE, MJUM9BYTES, MJUM16BYTES.
705 */
706static __inline void *
707m_cljget(struct mbuf *m, int how, int size)
708{
709	uma_zone_t zone;
710
711	if (m && m->m_flags & M_EXT)
712		printf("%s: %p mbuf already has cluster\n", __func__, m);
713	if (m != NULL)
714		m->m_ext.ext_buf = NULL;
715
716	zone = m_getzone(size);
717	return (uma_zalloc_arg(zone, m, how));
718}
719
720static __inline void
721m_cljset(struct mbuf *m, void *cl, int type)
722{
723	uma_zone_t zone;
724	int size;
725
726	switch (type) {
727	case EXT_CLUSTER:
728		size = MCLBYTES;
729		zone = zone_clust;
730		break;
731#if MJUMPAGESIZE != MCLBYTES
732	case EXT_JUMBOP:
733		size = MJUMPAGESIZE;
734		zone = zone_jumbop;
735		break;
736#endif
737	case EXT_JUMBO9:
738		size = MJUM9BYTES;
739		zone = zone_jumbo9;
740		break;
741	case EXT_JUMBO16:
742		size = MJUM16BYTES;
743		zone = zone_jumbo16;
744		break;
745	default:
746		panic("%s: unknown cluster type %d", __func__, type);
747		break;
748	}
749
750	m->m_data = m->m_ext.ext_buf = cl;
751	m->m_ext.ext_free = m->m_ext.ext_arg1 = m->m_ext.ext_arg2 = NULL;
752	m->m_ext.ext_size = size;
753	m->m_ext.ext_type = type;
754	m->m_ext.ext_flags = 0;
755	m->m_ext.ref_cnt = uma_find_refcnt(zone, cl);
756	m->m_flags |= M_EXT;
757
758}
759
760static __inline void
761m_chtype(struct mbuf *m, short new_type)
762{
763
764	m->m_type = new_type;
765}
766
767static __inline void
768m_clrprotoflags(struct mbuf *m)
769{
770
771	m->m_flags &= ~M_PROTOFLAGS;
772}
773
774static __inline struct mbuf *
775m_last(struct mbuf *m)
776{
777
778	while (m->m_next)
779		m = m->m_next;
780	return (m);
781}
782
783/*
784 * mbuf, cluster, and external object allocation macros (for compatibility
785 * purposes).
786 */
787#define	M_MOVE_PKTHDR(to, from)	m_move_pkthdr((to), (from))
788#define	MGET(m, how, type)	((m) = m_get((how), (type)))
789#define	MGETHDR(m, how, type)	((m) = m_gethdr((how), (type)))
790#define	MCLGET(m, how)		m_clget((m), (how))
791#define	MEXTADD(m, buf, size, free, arg1, arg2, flags, type)		\
792    (void )m_extadd((m), (caddr_t)(buf), (size), (free), (arg1), (arg2),\
793    (flags), (type), M_NOWAIT)
794#define	m_getm(m, len, how, type)					\
795    m_getm2((m), (len), (how), (type), M_PKTHDR)
796
797/*
798 * Evaluate TRUE if it's safe to write to the mbuf m's data region (this can
799 * be both the local data payload, or an external buffer area, depending on
800 * whether M_EXT is set).
801 */
802#define	M_WRITABLE(m)	(!((m)->m_flags & M_RDONLY) &&			\
803			 (!(((m)->m_flags & M_EXT)) ||			\
804			 (*((m)->m_ext.ref_cnt) == 1)) )		\
805
806/* Check if the supplied mbuf has a packet header, or else panic. */
807#define	M_ASSERTPKTHDR(m)						\
808	KASSERT((m) != NULL && (m)->m_flags & M_PKTHDR,			\
809	    ("%s: no mbuf packet header!", __func__))
810
811/*
812 * Ensure that the supplied mbuf is a valid, non-free mbuf.
813 *
814 * XXX: Broken at the moment.  Need some UMA magic to make it work again.
815 */
816#define	M_ASSERTVALID(m)						\
817	KASSERT((((struct mbuf *)m)->m_flags & 0) == 0,			\
818	    ("%s: attempted use of a free mbuf!", __func__))
819
820/*
821 * Set the m_data pointer of a newly-allocated mbuf (m_get/MGET) to place an
822 * object of the specified size at the end of the mbuf, longword aligned.
823 */
824#define	M_ALIGN(m, len) do {						\
825	KASSERT(!((m)->m_flags & (M_PKTHDR|M_EXT)),			\
826		("%s: M_ALIGN not normal mbuf", __func__));		\
827	KASSERT((m)->m_data == (m)->m_dat,				\
828		("%s: M_ALIGN not a virgin mbuf", __func__));		\
829	(m)->m_data += (MLEN - (len)) & ~(sizeof(long) - 1);		\
830} while (0)
831
832/*
833 * As above, for mbufs allocated with m_gethdr/MGETHDR or initialized by
834 * M_DUP/MOVE_PKTHDR.
835 */
836#define	MH_ALIGN(m, len) do {						\
837	KASSERT((m)->m_flags & M_PKTHDR && !((m)->m_flags & M_EXT),	\
838		("%s: MH_ALIGN not PKTHDR mbuf", __func__));		\
839	KASSERT((m)->m_data == (m)->m_pktdat,				\
840		("%s: MH_ALIGN not a virgin mbuf", __func__));		\
841	(m)->m_data += (MHLEN - (len)) & ~(sizeof(long) - 1);		\
842} while (0)
843
844/*
845 * As above, for mbuf with external storage.
846 */
847#define	MEXT_ALIGN(m, len) do {						\
848	KASSERT((m)->m_flags & M_EXT,					\
849		("%s: MEXT_ALIGN not an M_EXT mbuf", __func__));	\
850	KASSERT((m)->m_data == (m)->m_ext.ext_buf,			\
851		("%s: MEXT_ALIGN not a virgin mbuf", __func__));	\
852	(m)->m_data += ((m)->m_ext.ext_size - (len)) &			\
853	    ~(sizeof(long) - 1); 					\
854} while (0)
855
856/*
857 * Compute the amount of space available before the current start of data in
858 * an mbuf.
859 *
860 * The M_WRITABLE() is a temporary, conservative safety measure: the burden
861 * of checking writability of the mbuf data area rests solely with the caller.
862 */
863#define	M_LEADINGSPACE(m)						\
864	((m)->m_flags & M_EXT ?						\
865	    (M_WRITABLE(m) ? (m)->m_data - (m)->m_ext.ext_buf : 0):	\
866	    (m)->m_flags & M_PKTHDR ? (m)->m_data - (m)->m_pktdat :	\
867	    (m)->m_data - (m)->m_dat)
868
869/*
870 * Compute the amount of space available after the end of data in an mbuf.
871 *
872 * The M_WRITABLE() is a temporary, conservative safety measure: the burden
873 * of checking writability of the mbuf data area rests solely with the caller.
874 */
875#define	M_TRAILINGSPACE(m)						\
876	((m)->m_flags & M_EXT ?						\
877	    (M_WRITABLE(m) ? (m)->m_ext.ext_buf + (m)->m_ext.ext_size	\
878		- ((m)->m_data + (m)->m_len) : 0) :			\
879	    &(m)->m_dat[MLEN] - ((m)->m_data + (m)->m_len))
880
881/*
882 * Arrange to prepend space of size plen to mbuf m.  If a new mbuf must be
883 * allocated, how specifies whether to wait.  If the allocation fails, the
884 * original mbuf chain is freed and m is set to NULL.
885 */
886#define	M_PREPEND(m, plen, how) do {					\
887	struct mbuf **_mmp = &(m);					\
888	struct mbuf *_mm = *_mmp;					\
889	int _mplen = (plen);						\
890	int __mhow = (how);						\
891									\
892	MBUF_CHECKSLEEP(how);						\
893	if (M_LEADINGSPACE(_mm) >= _mplen) {				\
894		_mm->m_data -= _mplen;					\
895		_mm->m_len += _mplen;					\
896	} else								\
897		_mm = m_prepend(_mm, _mplen, __mhow);			\
898	if (_mm != NULL && _mm->m_flags & M_PKTHDR)			\
899		_mm->m_pkthdr.len += _mplen;				\
900	*_mmp = _mm;							\
901} while (0)
902
903/*
904 * Change mbuf to new type.  This is a relatively expensive operation and
905 * should be avoided.
906 */
907#define	MCHTYPE(m, t)	m_chtype((m), (t))
908
909/* Length to m_copy to copy all. */
910#define	M_COPYALL	1000000000
911
912/* Compatibility with 4.3. */
913#define	m_copy(m, o, l)	m_copym((m), (o), (l), M_NOWAIT)
914
915extern int		max_datalen;	/* MHLEN - max_hdr */
916extern int		max_hdr;	/* Largest link + protocol header */
917extern int		max_linkhdr;	/* Largest link-level header */
918extern int		max_protohdr;	/* Largest protocol header */
919extern int		nmbclusters;	/* Maximum number of clusters */
920
921struct uio;
922
923void		 m_adj(struct mbuf *, int);
924void		 m_align(struct mbuf *, int);
925int		 m_apply(struct mbuf *, int, int,
926		    int (*)(void *, void *, u_int), void *);
927int		 m_append(struct mbuf *, int, c_caddr_t);
928void		 m_cat(struct mbuf *, struct mbuf *);
929int		 m_extadd(struct mbuf *, caddr_t, u_int,
930		    int (*)(struct mbuf *, void *, void *), void *, void *,
931		    int, int, int);
932struct mbuf	*m_collapse(struct mbuf *, int, int);
933void		 m_copyback(struct mbuf *, int, int, c_caddr_t);
934void		 m_copydata(const struct mbuf *, int, int, caddr_t);
935struct mbuf	*m_copym(struct mbuf *, int, int, int);
936struct mbuf	*m_copymdata(struct mbuf *, struct mbuf *,
937		    int, int, int, int);
938struct mbuf	*m_copypacket(struct mbuf *, int);
939void		 m_copy_pkthdr(struct mbuf *, struct mbuf *);
940struct mbuf	*m_copyup(struct mbuf *, int, int);
941struct mbuf	*m_defrag(struct mbuf *, int);
942void		 m_demote(struct mbuf *, int);
943struct mbuf	*m_devget(char *, int, int, struct ifnet *,
944		    void (*)(char *, caddr_t, u_int));
945struct mbuf	*m_dup(struct mbuf *, int);
946int		 m_dup_pkthdr(struct mbuf *, struct mbuf *, int);
947u_int		 m_fixhdr(struct mbuf *);
948struct mbuf	*m_fragment(struct mbuf *, int, int);
949void		 m_freem(struct mbuf *);
950struct mbuf	*m_get2(int, int, short, int);
951struct mbuf	*m_getjcl(int, short, int, int);
952struct mbuf	*m_getm2(struct mbuf *, int, int, short, int);
953struct mbuf	*m_getptr(struct mbuf *, int, int *);
954u_int		 m_length(struct mbuf *, struct mbuf **);
955int		 m_mbuftouio(struct uio *, struct mbuf *, int);
956void		 m_move_pkthdr(struct mbuf *, struct mbuf *);
957struct mbuf	*m_prepend(struct mbuf *, int, int);
958void		 m_print(const struct mbuf *, int);
959struct mbuf	*m_pulldown(struct mbuf *, int, int, int *);
960struct mbuf	*m_pullup(struct mbuf *, int);
961int		 m_sanity(struct mbuf *, int);
962struct mbuf	*m_split(struct mbuf *, int, int);
963struct mbuf	*m_uiotombuf(struct uio *, int, int, int, int);
964struct mbuf	*m_unshare(struct mbuf *, int);
965
966/*-
967 * Network packets may have annotations attached by affixing a list of
968 * "packet tags" to the pkthdr structure.  Packet tags are dynamically
969 * allocated semi-opaque data structures that have a fixed header
970 * (struct m_tag) that specifies the size of the memory block and a
971 * <cookie,type> pair that identifies it.  The cookie is a 32-bit unique
972 * unsigned value used to identify a module or ABI.  By convention this value
973 * is chosen as the date+time that the module is created, expressed as the
974 * number of seconds since the epoch (e.g., using date -u +'%s').  The type
975 * value is an ABI/module-specific value that identifies a particular
976 * annotation and is private to the module.  For compatibility with systems
977 * like OpenBSD that define packet tags w/o an ABI/module cookie, the value
978 * PACKET_ABI_COMPAT is used to implement m_tag_get and m_tag_find
979 * compatibility shim functions and several tag types are defined below.
980 * Users that do not require compatibility should use a private cookie value
981 * so that packet tag-related definitions can be maintained privately.
982 *
983 * Note that the packet tag returned by m_tag_alloc has the default memory
984 * alignment implemented by malloc.  To reference private data one can use a
985 * construct like:
986 *
987 *	struct m_tag *mtag = m_tag_alloc(...);
988 *	struct foo *p = (struct foo *)(mtag+1);
989 *
990 * if the alignment of struct m_tag is sufficient for referencing members of
991 * struct foo.  Otherwise it is necessary to embed struct m_tag within the
992 * private data structure to insure proper alignment; e.g.,
993 *
994 *	struct foo {
995 *		struct m_tag	tag;
996 *		...
997 *	};
998 *	struct foo *p = (struct foo *) m_tag_alloc(...);
999 *	struct m_tag *mtag = &p->tag;
1000 */
1001
1002/*
1003 * Persistent tags stay with an mbuf until the mbuf is reclaimed.  Otherwise
1004 * tags are expected to ``vanish'' when they pass through a network
1005 * interface.  For most interfaces this happens normally as the tags are
1006 * reclaimed when the mbuf is free'd.  However in some special cases
1007 * reclaiming must be done manually.  An example is packets that pass through
1008 * the loopback interface.  Also, one must be careful to do this when
1009 * ``turning around'' packets (e.g., icmp_reflect).
1010 *
1011 * To mark a tag persistent bit-or this flag in when defining the tag id.
1012 * The tag will then be treated as described above.
1013 */
1014#define	MTAG_PERSISTENT				0x800
1015
1016#define	PACKET_TAG_NONE				0  /* Nadda */
1017
1018/* Packet tags for use with PACKET_ABI_COMPAT. */
1019#define	PACKET_TAG_IPSEC_IN_DONE		1  /* IPsec applied, in */
1020#define	PACKET_TAG_IPSEC_OUT_DONE		2  /* IPsec applied, out */
1021#define	PACKET_TAG_IPSEC_IN_CRYPTO_DONE		3  /* NIC IPsec crypto done */
1022#define	PACKET_TAG_IPSEC_OUT_CRYPTO_NEEDED	4  /* NIC IPsec crypto req'ed */
1023#define	PACKET_TAG_IPSEC_IN_COULD_DO_CRYPTO	5  /* NIC notifies IPsec */
1024#define	PACKET_TAG_IPSEC_PENDING_TDB		6  /* Reminder to do IPsec */
1025#define	PACKET_TAG_BRIDGE			7  /* Bridge processing done */
1026#define	PACKET_TAG_GIF				8  /* GIF processing done */
1027#define	PACKET_TAG_GRE				9  /* GRE processing done */
1028#define	PACKET_TAG_IN_PACKET_CHECKSUM		10 /* NIC checksumming done */
1029#define	PACKET_TAG_ENCAP			11 /* Encap.  processing */
1030#define	PACKET_TAG_IPSEC_SOCKET			12 /* IPSEC socket ref */
1031#define	PACKET_TAG_IPSEC_HISTORY		13 /* IPSEC history */
1032#define	PACKET_TAG_IPV6_INPUT			14 /* IPV6 input processing */
1033#define	PACKET_TAG_DUMMYNET			15 /* dummynet info */
1034#define	PACKET_TAG_DIVERT			17 /* divert info */
1035#define	PACKET_TAG_IPFORWARD			18 /* ipforward info */
1036#define	PACKET_TAG_MACLABEL	(19 | MTAG_PERSISTENT) /* MAC label */
1037#define	PACKET_TAG_PF		(21 | MTAG_PERSISTENT) /* PF/ALTQ information */
1038#define	PACKET_TAG_RTSOCKFAM			25 /* rtsock sa family */
1039#define	PACKET_TAG_IPOPTIONS			27 /* Saved IP options */
1040#define	PACKET_TAG_CARP				28 /* CARP info */
1041#define	PACKET_TAG_IPSEC_NAT_T_PORTS		29 /* two uint16_t */
1042#define	PACKET_TAG_ND_OUTGOING			30 /* ND outgoing */
1043
1044/* Specific cookies and tags. */
1045
1046/* Packet tag routines. */
1047struct m_tag	*m_tag_alloc(u_int32_t, int, int, int);
1048void		 m_tag_delete(struct mbuf *, struct m_tag *);
1049void		 m_tag_delete_chain(struct mbuf *, struct m_tag *);
1050void		 m_tag_free_default(struct m_tag *);
1051struct m_tag	*m_tag_locate(struct mbuf *, u_int32_t, int, struct m_tag *);
1052struct m_tag	*m_tag_copy(struct m_tag *, int);
1053int		 m_tag_copy_chain(struct mbuf *, struct mbuf *, int);
1054void		 m_tag_delete_nonpersistent(struct mbuf *);
1055
1056/*
1057 * Initialize the list of tags associated with an mbuf.
1058 */
1059static __inline void
1060m_tag_init(struct mbuf *m)
1061{
1062
1063	SLIST_INIT(&m->m_pkthdr.tags);
1064}
1065
1066/*
1067 * Set up the contents of a tag.  Note that this does not fill in the free
1068 * method; the caller is expected to do that.
1069 *
1070 * XXX probably should be called m_tag_init, but that was already taken.
1071 */
1072static __inline void
1073m_tag_setup(struct m_tag *t, u_int32_t cookie, int type, int len)
1074{
1075
1076	t->m_tag_id = type;
1077	t->m_tag_len = len;
1078	t->m_tag_cookie = cookie;
1079}
1080
1081/*
1082 * Reclaim resources associated with a tag.
1083 */
1084static __inline void
1085m_tag_free(struct m_tag *t)
1086{
1087
1088	(*t->m_tag_free)(t);
1089}
1090
1091/*
1092 * Return the first tag associated with an mbuf.
1093 */
1094static __inline struct m_tag *
1095m_tag_first(struct mbuf *m)
1096{
1097
1098	return (SLIST_FIRST(&m->m_pkthdr.tags));
1099}
1100
1101/*
1102 * Return the next tag in the list of tags associated with an mbuf.
1103 */
1104static __inline struct m_tag *
1105m_tag_next(struct mbuf *m, struct m_tag *t)
1106{
1107
1108	return (SLIST_NEXT(t, m_tag_link));
1109}
1110
1111/*
1112 * Prepend a tag to the list of tags associated with an mbuf.
1113 */
1114static __inline void
1115m_tag_prepend(struct mbuf *m, struct m_tag *t)
1116{
1117
1118	SLIST_INSERT_HEAD(&m->m_pkthdr.tags, t, m_tag_link);
1119}
1120
1121/*
1122 * Unlink a tag from the list of tags associated with an mbuf.
1123 */
1124static __inline void
1125m_tag_unlink(struct mbuf *m, struct m_tag *t)
1126{
1127
1128	SLIST_REMOVE(&m->m_pkthdr.tags, t, m_tag, m_tag_link);
1129}
1130
1131/* These are for OpenBSD compatibility. */
1132#define	MTAG_ABI_COMPAT		0		/* compatibility ABI */
1133
1134static __inline struct m_tag *
1135m_tag_get(int type, int length, int wait)
1136{
1137	return (m_tag_alloc(MTAG_ABI_COMPAT, type, length, wait));
1138}
1139
1140static __inline struct m_tag *
1141m_tag_find(struct mbuf *m, int type, struct m_tag *start)
1142{
1143	return (SLIST_EMPTY(&m->m_pkthdr.tags) ? (struct m_tag *)NULL :
1144	    m_tag_locate(m, MTAG_ABI_COMPAT, type, start));
1145}
1146
1147static __inline struct mbuf *
1148m_free(struct mbuf *m)
1149{
1150	struct mbuf *n = m->m_next;
1151
1152	if ((m->m_flags & (M_PKTHDR|M_NOFREE)) == (M_PKTHDR|M_NOFREE))
1153		m_tag_delete_chain(m, NULL);
1154	if (m->m_flags & M_EXT)
1155		mb_free_ext(m);
1156	else if ((m->m_flags & M_NOFREE) == 0)
1157		uma_zfree(zone_mbuf, m);
1158	return (n);
1159}
1160
1161static int inline
1162rt_m_getfib(struct mbuf *m)
1163{
1164	KASSERT(m->m_flags & M_PKTHDR , ("Attempt to get FIB from non header mbuf."));
1165	return (m->m_pkthdr.fibnum);
1166}
1167
1168#define M_GETFIB(_m)   rt_m_getfib(_m)
1169
1170#define M_SETFIB(_m, _fib) do {						\
1171        KASSERT((_m)->m_flags & M_PKTHDR, ("Attempt to set FIB on non header mbuf."));	\
1172	((_m)->m_pkthdr.fibnum) = (_fib);				\
1173} while (0)
1174
1175#endif /* _KERNEL */
1176
1177#ifdef MBUF_PROFILING
1178 void m_profile(struct mbuf *m);
1179 #define M_PROFILE(m) m_profile(m)
1180#else
1181 #define M_PROFILE(m)
1182#endif
1183
1184
1185#endif /* !_SYS_MBUF_H_ */
1186