mbuf.h revision 268529
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 268529 2014-07-11 13:58:48Z glebius $
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  eight[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  eight[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#define	pkt_tcphdr	PH_loc.ptr
163
164/*
165 * Description of external storage mapped into mbuf; valid only if M_EXT is
166 * set.
167 * Size ILP32: 28
168 *	 LP64: 48
169 */
170struct m_ext {
171	volatile u_int	*ref_cnt;	/* pointer to ref count info */
172	caddr_t		 ext_buf;	/* start of buffer */
173	uint32_t	 ext_size;	/* size of buffer, for ext_free */
174	uint32_t	 ext_type:8,	/* type of external storage */
175			 ext_flags:24;	/* external storage mbuf flags */
176	void		(*ext_free)	/* free routine if not the usual */
177			    (struct mbuf *, void *, void *);
178	void		*ext_arg1;	/* optional argument pointer */
179	void		*ext_arg2;	/* optional argument pointer */
180};
181
182/*
183 * The core of the mbuf object along with some shortcut defines for practical
184 * purposes.
185 */
186struct mbuf {
187	struct m_hdr	m_hdr;
188	union {
189		struct {
190			struct pkthdr	MH_pkthdr;	/* M_PKTHDR set */
191			union {
192				struct m_ext	MH_ext;	/* M_EXT set */
193				char		MH_databuf[MHLEN];
194			} MH_dat;
195		} MH;
196		char	M_databuf[MLEN];		/* !M_PKTHDR, !M_EXT */
197	} M_dat;
198};
199#define	m_next		m_hdr.mh_next
200#define	m_len		m_hdr.mh_len
201#define	m_data		m_hdr.mh_data
202#define	m_type		m_hdr.mh_type
203#define	m_flags		m_hdr.mh_flags
204#define	m_nextpkt	m_hdr.mh_nextpkt
205#define	m_act		m_nextpkt
206#define	m_pkthdr	M_dat.MH.MH_pkthdr
207#define	m_ext		M_dat.MH.MH_dat.MH_ext
208#define	m_pktdat	M_dat.MH.MH_dat.MH_databuf
209#define	m_dat		M_dat.M_databuf
210
211/*
212 * mbuf flags of global significance and layer crossing.
213 * Those of only protocol/layer specific significance are to be mapped
214 * to M_PROTO[1-12] and cleared at layer handoff boundaries.
215 * NB: Limited to the lower 24 bits.
216 */
217#define	M_EXT		0x00000001 /* has associated external storage */
218#define	M_PKTHDR	0x00000002 /* start of record */
219#define	M_EOR		0x00000004 /* end of record */
220#define	M_RDONLY	0x00000008 /* associated data is marked read-only */
221#define	M_BCAST		0x00000010 /* send/received as link-level broadcast */
222#define	M_MCAST		0x00000020 /* send/received as link-level multicast */
223#define	M_PROMISC	0x00000040 /* packet was not for us */
224#define	M_VLANTAG	0x00000080 /* ether_vtag is valid */
225#define	M_FLOWID	0x00000100 /* deprecated: flowid is valid */
226#define	M_NOFREE	0x00000200 /* do not free mbuf, embedded in cluster */
227
228#define	M_PROTO1	0x00001000 /* protocol-specific */
229#define	M_PROTO2	0x00002000 /* protocol-specific */
230#define	M_PROTO3	0x00004000 /* protocol-specific */
231#define	M_PROTO4	0x00008000 /* protocol-specific */
232#define	M_PROTO5	0x00010000 /* protocol-specific */
233#define	M_PROTO6	0x00020000 /* protocol-specific */
234#define	M_PROTO7	0x00040000 /* protocol-specific */
235#define	M_PROTO8	0x00080000 /* protocol-specific */
236#define	M_PROTO9	0x00100000 /* protocol-specific */
237#define	M_PROTO10	0x00200000 /* protocol-specific */
238#define	M_PROTO11	0x00400000 /* protocol-specific */
239#define	M_PROTO12	0x00800000 /* protocol-specific */
240
241/*
242 * Flags to purge when crossing layers.
243 */
244#define	M_PROTOFLAGS \
245    (M_PROTO1|M_PROTO2|M_PROTO3|M_PROTO4|M_PROTO5|M_PROTO6|M_PROTO7|M_PROTO8|\
246     M_PROTO9|M_PROTO10|M_PROTO11|M_PROTO12)
247
248/*
249 * Flags preserved when copying m_pkthdr.
250 */
251#define M_COPYFLAGS \
252    (M_PKTHDR|M_EOR|M_RDONLY|M_BCAST|M_MCAST|M_VLANTAG|M_PROMISC| \
253     M_PROTOFLAGS)
254
255/*
256 * Mbuf flag description for use with printf(9) %b identifier.
257 */
258#define	M_FLAG_BITS \
259    "\20\1M_EXT\2M_PKTHDR\3M_EOR\4M_RDONLY\5M_BCAST\6M_MCAST" \
260    "\7M_PROMISC\10M_VLANTAG\11M_FLOWID"
261#define	M_FLAG_PROTOBITS \
262    "\15M_PROTO1\16M_PROTO2\17M_PROTO3\20M_PROTO4\21M_PROTO5" \
263    "\22M_PROTO6\23M_PROTO7\24M_PROTO8\25M_PROTO9\26M_PROTO10" \
264    "\27M_PROTO11\30M_PROTO12"
265#define	M_FLAG_PRINTF (M_FLAG_BITS M_FLAG_PROTOBITS)
266
267/*
268 * Network interface cards are able to hash protocol fields (such as IPv4
269 * addresses and TCP port numbers) classify packets into flows.  These flows
270 * can then be used to maintain ordering while delivering packets to the OS
271 * via parallel input queues, as well as to provide a stateless affinity
272 * model.  NIC drivers can pass up the hash via m->m_pkthdr.flowid, and set
273 * m_flag fields to indicate how the hash should be interpreted by the
274 * network stack.
275 *
276 * Most NICs support RSS, which provides ordering and explicit affinity, and
277 * use the hash m_flag bits to indicate what header fields were covered by
278 * the hash.  M_HASHTYPE_OPAQUE can be set by non-RSS cards or configurations
279 * that provide an opaque flow identifier, allowing for ordering and
280 * distribution without explicit affinity.
281 */
282#define	M_HASHTYPE_NONE			0
283#define	M_HASHTYPE_RSS_IPV4		1	/* IPv4 2-tuple */
284#define	M_HASHTYPE_RSS_TCP_IPV4		2	/* TCPv4 4-tuple */
285#define	M_HASHTYPE_RSS_IPV6		3	/* IPv6 2-tuple */
286#define	M_HASHTYPE_RSS_TCP_IPV6		4	/* TCPv6 4-tuple */
287#define	M_HASHTYPE_RSS_IPV6_EX		5	/* IPv6 2-tuple + ext hdrs */
288#define	M_HASHTYPE_RSS_TCP_IPV6_EX	6	/* TCPv6 4-tiple + ext hdrs */
289#define	M_HASHTYPE_OPAQUE		255	/* ordering, not affinity */
290
291#define	M_HASHTYPE_CLEAR(m)	((m)->m_pkthdr.rsstype = 0)
292#define	M_HASHTYPE_GET(m)	((m)->m_pkthdr.rsstype)
293#define	M_HASHTYPE_SET(m, v)	((m)->m_pkthdr.rsstype = (v))
294#define	M_HASHTYPE_TEST(m, v)	(M_HASHTYPE_GET(m) == (v))
295
296/*
297 * COS/QOS class and quality of service tags.
298 * It uses DSCP code points as base.
299 */
300#define	QOS_DSCP_CS0		0x00
301#define	QOS_DSCP_DEF		QOS_DSCP_CS0
302#define	QOS_DSCP_CS1		0x20
303#define	QOS_DSCP_AF11		0x28
304#define	QOS_DSCP_AF12		0x30
305#define	QOS_DSCP_AF13		0x38
306#define	QOS_DSCP_CS2		0x40
307#define	QOS_DSCP_AF21		0x48
308#define	QOS_DSCP_AF22		0x50
309#define	QOS_DSCP_AF23		0x58
310#define	QOS_DSCP_CS3		0x60
311#define	QOS_DSCP_AF31		0x68
312#define	QOS_DSCP_AF32		0x70
313#define	QOS_DSCP_AF33		0x78
314#define	QOS_DSCP_CS4		0x80
315#define	QOS_DSCP_AF41		0x88
316#define	QOS_DSCP_AF42		0x90
317#define	QOS_DSCP_AF43		0x98
318#define	QOS_DSCP_CS5		0xa0
319#define	QOS_DSCP_EF		0xb8
320#define	QOS_DSCP_CS6		0xc0
321#define	QOS_DSCP_CS7		0xe0
322
323/*
324 * External mbuf storage buffer types.
325 */
326#define	EXT_CLUSTER	1	/* mbuf cluster */
327#define	EXT_SFBUF	2	/* sendfile(2)'s sf_bufs */
328#define	EXT_JUMBOP	3	/* jumbo cluster 4096 bytes */
329#define	EXT_JUMBO9	4	/* jumbo cluster 9216 bytes */
330#define	EXT_JUMBO16	5	/* jumbo cluster 16184 bytes */
331#define	EXT_PACKET	6	/* mbuf+cluster from packet zone */
332#define	EXT_MBUF	7	/* external mbuf reference (M_IOVEC) */
333
334#define	EXT_VENDOR1	224	/* for vendor-internal use */
335#define	EXT_VENDOR2	225	/* for vendor-internal use */
336#define	EXT_VENDOR3	226	/* for vendor-internal use */
337#define	EXT_VENDOR4	227	/* for vendor-internal use */
338
339#define	EXT_EXP1	244	/* for experimental use */
340#define	EXT_EXP2	245	/* for experimental use */
341#define	EXT_EXP3	246	/* for experimental use */
342#define	EXT_EXP4	247	/* for experimental use */
343
344#define	EXT_NET_DRV	252	/* custom ext_buf provided by net driver(s) */
345#define	EXT_MOD_TYPE	253	/* custom module's ext_buf type */
346#define	EXT_DISPOSABLE	254	/* can throw this buffer away w/page flipping */
347#define	EXT_EXTREF	255	/* has externally maintained ref_cnt ptr */
348
349/*
350 * Flags for external mbuf buffer types.
351 * NB: limited to the lower 24 bits.
352 */
353#define	EXT_FLAG_EMBREF		0x000001	/* embedded ref_cnt, notyet */
354#define	EXT_FLAG_EXTREF		0x000002	/* external ref_cnt, notyet */
355#define	EXT_FLAG_NOFREE		0x000010	/* don't free mbuf to pool, notyet */
356
357#define	EXT_FLAG_VENDOR1	0x010000	/* for vendor-internal use */
358#define	EXT_FLAG_VENDOR2	0x020000	/* for vendor-internal use */
359#define	EXT_FLAG_VENDOR3	0x040000	/* for vendor-internal use */
360#define	EXT_FLAG_VENDOR4	0x080000	/* for vendor-internal use */
361
362#define	EXT_FLAG_EXP1		0x100000	/* for experimental use */
363#define	EXT_FLAG_EXP2		0x200000	/* for experimental use */
364#define	EXT_FLAG_EXP3		0x400000	/* for experimental use */
365#define	EXT_FLAG_EXP4		0x800000	/* for experimental use */
366
367/*
368 * EXT flag description for use with printf(9) %b identifier.
369 */
370#define	EXT_FLAG_BITS \
371    "\20\1EXT_FLAG_EMBREF\2EXT_FLAG_EXTREF\5EXT_FLAG_NOFREE" \
372    "\21EXT_FLAG_VENDOR1\22EXT_FLAG_VENDOR2\23EXT_FLAG_VENDOR3" \
373    "\24EXT_FLAG_VENDOR4\25EXT_FLAG_EXP1\26EXT_FLAG_EXP2\27EXT_FLAG_EXP3" \
374    "\30EXT_FLAG_EXP4"
375
376/*
377 * Flags indicating checksum, segmentation and other offload work to be
378 * done, or already done, by hardware or lower layers.  It is split into
379 * separate inbound and outbound flags.
380 *
381 * Outbound flags that are set by upper protocol layers requesting lower
382 * layers, or ideally the hardware, to perform these offloading tasks.
383 * For outbound packets this field and its flags can be directly tested
384 * against if_data.ifi_hwassist.
385 */
386#define	CSUM_IP			0x00000001	/* IP header checksum offload */
387#define	CSUM_IP_UDP		0x00000002	/* UDP checksum offload */
388#define	CSUM_IP_TCP		0x00000004	/* TCP checksum offload */
389#define	CSUM_IP_SCTP		0x00000008	/* SCTP checksum offload */
390#define	CSUM_IP_TSO		0x00000010	/* TCP segmentation offload */
391#define	CSUM_IP_ISCSI		0x00000020	/* iSCSI checksum offload */
392
393#define	CSUM_IP6_UDP		0x00000200	/* UDP checksum offload */
394#define	CSUM_IP6_TCP		0x00000400	/* TCP checksum offload */
395#define	CSUM_IP6_SCTP		0x00000800	/* SCTP checksum offload */
396#define	CSUM_IP6_TSO		0x00001000	/* TCP segmentation offload */
397#define	CSUM_IP6_ISCSI		0x00002000	/* iSCSI checksum offload */
398
399/* Inbound checksum support where the checksum was verified by hardware. */
400#define	CSUM_L3_CALC		0x01000000	/* calculated layer 3 csum */
401#define	CSUM_L3_VALID		0x02000000	/* checksum is correct */
402#define	CSUM_L4_CALC		0x04000000	/* calculated layer 4 csum */
403#define	CSUM_L4_VALID		0x08000000	/* checksum is correct */
404#define	CSUM_L5_CALC		0x10000000	/* calculated layer 5 csum */
405#define	CSUM_L5_VALID		0x20000000	/* checksum is correct */
406#define	CSUM_COALESED		0x40000000	/* contains merged segments */
407
408/*
409 * CSUM flag description for use with printf(9) %b identifier.
410 */
411#define	CSUM_BITS \
412    "\20\1CSUM_IP\2CSUM_IP_UDP\3CSUM_IP_TCP\4CSUM_IP_SCTP\5CSUM_IP_TSO" \
413    "\6CSUM_IP_ISCSI" \
414    "\12CSUM_IP6_UDP\13CSUM_IP6_TCP\14CSUM_IP6_SCTP\15CSUM_IP6_TSO" \
415    "\16CSUM_IP6_ISCSI" \
416    "\31CSUM_L3_CALC\32CSUM_L3_VALID\33CSUM_L4_CALC\34CSUM_L4_VALID" \
417    "\35CSUM_L5_CALC\36CSUM_L5_VALID\37CSUM_COALESED"
418
419/* CSUM flags compatibility mappings. */
420#define	CSUM_IP_CHECKED		CSUM_L3_CALC
421#define	CSUM_IP_VALID		CSUM_L3_VALID
422#define	CSUM_DATA_VALID		CSUM_L4_VALID
423#define	CSUM_PSEUDO_HDR		CSUM_L4_CALC
424#define	CSUM_SCTP_VALID		CSUM_L4_VALID
425#define	CSUM_DELAY_DATA		(CSUM_TCP|CSUM_UDP)
426#define	CSUM_DELAY_IP		CSUM_IP		/* Only v4, no v6 IP hdr csum */
427#define	CSUM_DELAY_DATA_IPV6	(CSUM_TCP_IPV6|CSUM_UDP_IPV6)
428#define	CSUM_DATA_VALID_IPV6	CSUM_DATA_VALID
429#define	CSUM_TCP		CSUM_IP_TCP
430#define	CSUM_UDP		CSUM_IP_UDP
431#define	CSUM_SCTP		CSUM_IP_SCTP
432#define	CSUM_TSO		(CSUM_IP_TSO|CSUM_IP6_TSO)
433#define	CSUM_UDP_IPV6		CSUM_IP6_UDP
434#define	CSUM_TCP_IPV6		CSUM_IP6_TCP
435#define	CSUM_SCTP_IPV6		CSUM_IP6_SCTP
436#define	CSUM_FRAGMENT		0x0		/* Unused */
437
438/*
439 * mbuf types describing the content of the mbuf (including external storage).
440 */
441#define	MT_NOTMBUF	0	/* USED INTERNALLY ONLY! Object is not mbuf */
442#define	MT_DATA		1	/* dynamic (data) allocation */
443#define	MT_HEADER	MT_DATA	/* packet header, use M_PKTHDR instead */
444
445#define	MT_VENDOR1	4	/* for vendor-internal use */
446#define	MT_VENDOR2	5	/* for vendor-internal use */
447#define	MT_VENDOR3	6	/* for vendor-internal use */
448#define	MT_VENDOR4	7	/* for vendor-internal use */
449
450#define	MT_SONAME	8	/* socket name */
451
452#define	MT_EXP1		9	/* for experimental use */
453#define	MT_EXP2		10	/* for experimental use */
454#define	MT_EXP3		11	/* for experimental use */
455#define	MT_EXP4		12	/* for experimental use */
456
457#define	MT_CONTROL	14	/* extra-data protocol message */
458#define	MT_OOBDATA	15	/* expedited data  */
459#define	MT_NTYPES	16	/* number of mbuf types for mbtypes[] */
460
461#define	MT_NOINIT	255	/* Not a type but a flag to allocate
462				   a non-initialized mbuf */
463
464/*
465 * String names of mbuf-related UMA(9) and malloc(9) types.  Exposed to
466 * !_KERNEL so that monitoring tools can look up the zones with
467 * libmemstat(3).
468 */
469#define	MBUF_MEM_NAME		"mbuf"
470#define	MBUF_CLUSTER_MEM_NAME	"mbuf_cluster"
471#define	MBUF_PACKET_MEM_NAME	"mbuf_packet"
472#define	MBUF_JUMBOP_MEM_NAME	"mbuf_jumbo_page"
473#define	MBUF_JUMBO9_MEM_NAME	"mbuf_jumbo_9k"
474#define	MBUF_JUMBO16_MEM_NAME	"mbuf_jumbo_16k"
475#define	MBUF_TAG_MEM_NAME	"mbuf_tag"
476#define	MBUF_EXTREFCNT_MEM_NAME	"mbuf_ext_refcnt"
477
478#ifdef _KERNEL
479
480#ifdef WITNESS
481#define	MBUF_CHECKSLEEP(how) do {					\
482	if (how == M_WAITOK)						\
483		WITNESS_WARN(WARN_GIANTOK | WARN_SLEEPOK, NULL,		\
484		    "Sleeping in \"%s\"", __func__);			\
485} while (0)
486#else
487#define	MBUF_CHECKSLEEP(how)
488#endif
489
490/*
491 * Network buffer allocation API
492 *
493 * The rest of it is defined in kern/kern_mbuf.c
494 */
495extern uma_zone_t	zone_mbuf;
496extern uma_zone_t	zone_clust;
497extern uma_zone_t	zone_pack;
498extern uma_zone_t	zone_jumbop;
499extern uma_zone_t	zone_jumbo9;
500extern uma_zone_t	zone_jumbo16;
501extern uma_zone_t	zone_ext_refcnt;
502
503void		 mb_free_ext(struct mbuf *);
504int		 m_pkthdr_init(struct mbuf *, int);
505
506static __inline int
507m_gettype(int size)
508{
509	int type;
510
511	switch (size) {
512	case MSIZE:
513		type = EXT_MBUF;
514		break;
515	case MCLBYTES:
516		type = EXT_CLUSTER;
517		break;
518#if MJUMPAGESIZE != MCLBYTES
519	case MJUMPAGESIZE:
520		type = EXT_JUMBOP;
521		break;
522#endif
523	case MJUM9BYTES:
524		type = EXT_JUMBO9;
525		break;
526	case MJUM16BYTES:
527		type = EXT_JUMBO16;
528		break;
529	default:
530		panic("%s: invalid cluster size %d", __func__, size);
531	}
532
533	return (type);
534}
535
536/*
537 * Associated an external reference counted buffer with an mbuf.
538 */
539static __inline void
540m_extaddref(struct mbuf *m, caddr_t buf, u_int size, u_int *ref_cnt,
541    void (*freef)(struct mbuf *, void *, void *), void *arg1, void *arg2)
542{
543
544	KASSERT(ref_cnt != NULL, ("%s: ref_cnt not provided", __func__));
545
546	atomic_add_int(ref_cnt, 1);
547	m->m_flags |= M_EXT;
548	m->m_ext.ext_buf = buf;
549	m->m_ext.ref_cnt = ref_cnt;
550	m->m_data = m->m_ext.ext_buf;
551	m->m_ext.ext_size = size;
552	m->m_ext.ext_free = freef;
553	m->m_ext.ext_arg1 = arg1;
554	m->m_ext.ext_arg2 = arg2;
555	m->m_ext.ext_type = EXT_EXTREF;
556}
557
558static __inline uma_zone_t
559m_getzone(int size)
560{
561	uma_zone_t zone;
562
563	switch (size) {
564	case MCLBYTES:
565		zone = zone_clust;
566		break;
567#if MJUMPAGESIZE != MCLBYTES
568	case MJUMPAGESIZE:
569		zone = zone_jumbop;
570		break;
571#endif
572	case MJUM9BYTES:
573		zone = zone_jumbo9;
574		break;
575	case MJUM16BYTES:
576		zone = zone_jumbo16;
577		break;
578	default:
579		panic("%s: invalid cluster size %d", __func__, size);
580	}
581
582	return (zone);
583}
584
585/*
586 * Initialize an mbuf with linear storage.
587 *
588 * Inline because the consumer text overhead will be roughly the same to
589 * initialize or call a function with this many parameters and M_PKTHDR
590 * should go away with constant propagation for !MGETHDR.
591 */
592static __inline int
593m_init(struct mbuf *m, uma_zone_t zone, int size, int how, short type,
594    int flags)
595{
596	int error;
597
598	m->m_next = NULL;
599	m->m_nextpkt = NULL;
600	m->m_data = m->m_dat;
601	m->m_len = 0;
602	m->m_flags = flags;
603	m->m_type = type;
604	if (flags & M_PKTHDR) {
605		if ((error = m_pkthdr_init(m, how)) != 0)
606			return (error);
607	}
608
609	return (0);
610}
611
612static __inline struct mbuf *
613m_get(int how, short type)
614{
615	struct mb_args args;
616
617	args.flags = 0;
618	args.type = type;
619	return (uma_zalloc_arg(zone_mbuf, &args, how));
620}
621
622/*
623 * XXX This should be deprecated, very little use.
624 */
625static __inline struct mbuf *
626m_getclr(int how, short type)
627{
628	struct mbuf *m;
629	struct mb_args args;
630
631	args.flags = 0;
632	args.type = type;
633	m = uma_zalloc_arg(zone_mbuf, &args, how);
634	if (m != NULL)
635		bzero(m->m_data, MLEN);
636	return (m);
637}
638
639static __inline struct mbuf *
640m_gethdr(int how, short type)
641{
642	struct mb_args args;
643
644	args.flags = M_PKTHDR;
645	args.type = type;
646	return (uma_zalloc_arg(zone_mbuf, &args, how));
647}
648
649static __inline struct mbuf *
650m_getcl(int how, short type, int flags)
651{
652	struct mb_args args;
653
654	args.flags = flags;
655	args.type = type;
656	return (uma_zalloc_arg(zone_pack, &args, how));
657}
658
659static __inline void
660m_clget(struct mbuf *m, int how)
661{
662
663	if (m->m_flags & M_EXT)
664		printf("%s: %p mbuf already has cluster\n", __func__, m);
665	m->m_ext.ext_buf = (char *)NULL;
666	uma_zalloc_arg(zone_clust, m, how);
667	/*
668	 * On a cluster allocation failure, drain the packet zone and retry,
669	 * we might be able to loosen a few clusters up on the drain.
670	 */
671	if ((how & M_NOWAIT) && (m->m_ext.ext_buf == NULL)) {
672		zone_drain(zone_pack);
673		uma_zalloc_arg(zone_clust, m, how);
674	}
675}
676
677/*
678 * m_cljget() is different from m_clget() as it can allocate clusters without
679 * attaching them to an mbuf.  In that case the return value is the pointer
680 * to the cluster of the requested size.  If an mbuf was specified, it gets
681 * the cluster attached to it and the return value can be safely ignored.
682 * For size it takes MCLBYTES, MJUMPAGESIZE, MJUM9BYTES, MJUM16BYTES.
683 */
684static __inline void *
685m_cljget(struct mbuf *m, int how, int size)
686{
687	uma_zone_t zone;
688
689	if (m && m->m_flags & M_EXT)
690		printf("%s: %p mbuf already has cluster\n", __func__, m);
691	if (m != NULL)
692		m->m_ext.ext_buf = NULL;
693
694	zone = m_getzone(size);
695	return (uma_zalloc_arg(zone, m, how));
696}
697
698static __inline void
699m_cljset(struct mbuf *m, void *cl, int type)
700{
701	uma_zone_t zone;
702	int size;
703
704	switch (type) {
705	case EXT_CLUSTER:
706		size = MCLBYTES;
707		zone = zone_clust;
708		break;
709#if MJUMPAGESIZE != MCLBYTES
710	case EXT_JUMBOP:
711		size = MJUMPAGESIZE;
712		zone = zone_jumbop;
713		break;
714#endif
715	case EXT_JUMBO9:
716		size = MJUM9BYTES;
717		zone = zone_jumbo9;
718		break;
719	case EXT_JUMBO16:
720		size = MJUM16BYTES;
721		zone = zone_jumbo16;
722		break;
723	default:
724		panic("%s: unknown cluster type %d", __func__, type);
725		break;
726	}
727
728	m->m_data = m->m_ext.ext_buf = cl;
729	m->m_ext.ext_free = m->m_ext.ext_arg1 = m->m_ext.ext_arg2 = NULL;
730	m->m_ext.ext_size = size;
731	m->m_ext.ext_type = type;
732	m->m_ext.ext_flags = 0;
733	m->m_ext.ref_cnt = uma_find_refcnt(zone, cl);
734	m->m_flags |= M_EXT;
735
736}
737
738static __inline void
739m_chtype(struct mbuf *m, short new_type)
740{
741
742	m->m_type = new_type;
743}
744
745static __inline void
746m_clrprotoflags(struct mbuf *m)
747{
748
749	m->m_flags &= ~M_PROTOFLAGS;
750}
751
752static __inline struct mbuf *
753m_last(struct mbuf *m)
754{
755
756	while (m->m_next)
757		m = m->m_next;
758	return (m);
759}
760
761/*
762 * mbuf, cluster, and external object allocation macros (for compatibility
763 * purposes).
764 */
765#define	M_MOVE_PKTHDR(to, from)	m_move_pkthdr((to), (from))
766#define	MGET(m, how, type)	((m) = m_get((how), (type)))
767#define	MGETHDR(m, how, type)	((m) = m_gethdr((how), (type)))
768#define	MCLGET(m, how)		m_clget((m), (how))
769#define	MEXTADD(m, buf, size, free, arg1, arg2, flags, type)		\
770    (void )m_extadd((m), (caddr_t)(buf), (size), (free), (arg1), (arg2),\
771    (flags), (type), M_NOWAIT)
772#define	m_getm(m, len, how, type)					\
773    m_getm2((m), (len), (how), (type), M_PKTHDR)
774
775/*
776 * Evaluate TRUE if it's safe to write to the mbuf m's data region (this can
777 * be both the local data payload, or an external buffer area, depending on
778 * whether M_EXT is set).
779 */
780#define	M_WRITABLE(m)	(!((m)->m_flags & M_RDONLY) &&			\
781			 (!(((m)->m_flags & M_EXT)) ||			\
782			 (*((m)->m_ext.ref_cnt) == 1)) )		\
783
784/* Check if the supplied mbuf has a packet header, or else panic. */
785#define	M_ASSERTPKTHDR(m)						\
786	KASSERT((m) != NULL && (m)->m_flags & M_PKTHDR,			\
787	    ("%s: no mbuf packet header!", __func__))
788
789/*
790 * Ensure that the supplied mbuf is a valid, non-free mbuf.
791 *
792 * XXX: Broken at the moment.  Need some UMA magic to make it work again.
793 */
794#define	M_ASSERTVALID(m)						\
795	KASSERT((((struct mbuf *)m)->m_flags & 0) == 0,			\
796	    ("%s: attempted use of a free mbuf!", __func__))
797
798/*
799 * Set the m_data pointer of a newly-allocated mbuf (m_get/MGET) to place an
800 * object of the specified size at the end of the mbuf, longword aligned.
801 */
802#define	M_ALIGN(m, len) do {						\
803	KASSERT(!((m)->m_flags & (M_PKTHDR|M_EXT)),			\
804		("%s: M_ALIGN not normal mbuf", __func__));		\
805	KASSERT((m)->m_data == (m)->m_dat,				\
806		("%s: M_ALIGN not a virgin mbuf", __func__));		\
807	(m)->m_data += (MLEN - (len)) & ~(sizeof(long) - 1);		\
808} while (0)
809
810/*
811 * As above, for mbufs allocated with m_gethdr/MGETHDR or initialized by
812 * M_DUP/MOVE_PKTHDR.
813 */
814#define	MH_ALIGN(m, len) do {						\
815	KASSERT((m)->m_flags & M_PKTHDR && !((m)->m_flags & M_EXT),	\
816		("%s: MH_ALIGN not PKTHDR mbuf", __func__));		\
817	KASSERT((m)->m_data == (m)->m_pktdat,				\
818		("%s: MH_ALIGN not a virgin mbuf", __func__));		\
819	(m)->m_data += (MHLEN - (len)) & ~(sizeof(long) - 1);		\
820} while (0)
821
822/*
823 * As above, for mbuf with external storage.
824 */
825#define	MEXT_ALIGN(m, len) do {						\
826	KASSERT((m)->m_flags & M_EXT,					\
827		("%s: MEXT_ALIGN not an M_EXT mbuf", __func__));	\
828	KASSERT((m)->m_data == (m)->m_ext.ext_buf,			\
829		("%s: MEXT_ALIGN not a virgin mbuf", __func__));	\
830	(m)->m_data += ((m)->m_ext.ext_size - (len)) &			\
831	    ~(sizeof(long) - 1); 					\
832} while (0)
833
834/*
835 * Compute the amount of space available before the current start of data in
836 * an mbuf.
837 *
838 * The M_WRITABLE() is a temporary, conservative safety measure: the burden
839 * of checking writability of the mbuf data area rests solely with the caller.
840 */
841#define	M_LEADINGSPACE(m)						\
842	((m)->m_flags & M_EXT ?						\
843	    (M_WRITABLE(m) ? (m)->m_data - (m)->m_ext.ext_buf : 0):	\
844	    (m)->m_flags & M_PKTHDR ? (m)->m_data - (m)->m_pktdat :	\
845	    (m)->m_data - (m)->m_dat)
846
847/*
848 * Compute the amount of space available after the end of data in an mbuf.
849 *
850 * The M_WRITABLE() is a temporary, conservative safety measure: the burden
851 * of checking writability of the mbuf data area rests solely with the caller.
852 */
853#define	M_TRAILINGSPACE(m)						\
854	((m)->m_flags & M_EXT ?						\
855	    (M_WRITABLE(m) ? (m)->m_ext.ext_buf + (m)->m_ext.ext_size	\
856		- ((m)->m_data + (m)->m_len) : 0) :			\
857	    &(m)->m_dat[MLEN] - ((m)->m_data + (m)->m_len))
858
859/*
860 * Arrange to prepend space of size plen to mbuf m.  If a new mbuf must be
861 * allocated, how specifies whether to wait.  If the allocation fails, the
862 * original mbuf chain is freed and m is set to NULL.
863 */
864#define	M_PREPEND(m, plen, how) do {					\
865	struct mbuf **_mmp = &(m);					\
866	struct mbuf *_mm = *_mmp;					\
867	int _mplen = (plen);						\
868	int __mhow = (how);						\
869									\
870	MBUF_CHECKSLEEP(how);						\
871	if (M_LEADINGSPACE(_mm) >= _mplen) {				\
872		_mm->m_data -= _mplen;					\
873		_mm->m_len += _mplen;					\
874	} else								\
875		_mm = m_prepend(_mm, _mplen, __mhow);			\
876	if (_mm != NULL && _mm->m_flags & M_PKTHDR)			\
877		_mm->m_pkthdr.len += _mplen;				\
878	*_mmp = _mm;							\
879} while (0)
880
881/*
882 * Change mbuf to new type.  This is a relatively expensive operation and
883 * should be avoided.
884 */
885#define	MCHTYPE(m, t)	m_chtype((m), (t))
886
887/* Length to m_copy to copy all. */
888#define	M_COPYALL	1000000000
889
890/* Compatibility with 4.3. */
891#define	m_copy(m, o, l)	m_copym((m), (o), (l), M_NOWAIT)
892
893extern int		max_datalen;	/* MHLEN - max_hdr */
894extern int		max_hdr;	/* Largest link + protocol header */
895extern int		max_linkhdr;	/* Largest link-level header */
896extern int		max_protohdr;	/* Largest protocol header */
897extern int		nmbclusters;	/* Maximum number of clusters */
898
899struct uio;
900
901void		 m_adj(struct mbuf *, int);
902void		 m_align(struct mbuf *, int);
903int		 m_apply(struct mbuf *, int, int,
904		    int (*)(void *, void *, u_int), void *);
905int		 m_append(struct mbuf *, int, c_caddr_t);
906void		 m_cat(struct mbuf *, struct mbuf *);
907int		 m_extadd(struct mbuf *, caddr_t, u_int,
908		    void (*)(struct mbuf *, void *, void *), void *, void *,
909		    int, int, int);
910struct mbuf	*m_collapse(struct mbuf *, int, int);
911void		 m_copyback(struct mbuf *, int, int, c_caddr_t);
912void		 m_copydata(const struct mbuf *, int, int, caddr_t);
913struct mbuf	*m_copym(struct mbuf *, int, int, int);
914struct mbuf	*m_copymdata(struct mbuf *, struct mbuf *,
915		    int, int, int, int);
916struct mbuf	*m_copypacket(struct mbuf *, int);
917void		 m_copy_pkthdr(struct mbuf *, struct mbuf *);
918struct mbuf	*m_copyup(struct mbuf *, int, int);
919struct mbuf	*m_defrag(struct mbuf *, int);
920void		 m_demote(struct mbuf *, int);
921struct mbuf	*m_devget(char *, int, int, struct ifnet *,
922		    void (*)(char *, caddr_t, u_int));
923struct mbuf	*m_dup(struct mbuf *, int);
924int		 m_dup_pkthdr(struct mbuf *, struct mbuf *, int);
925u_int		 m_fixhdr(struct mbuf *);
926struct mbuf	*m_fragment(struct mbuf *, int, int);
927void		 m_freem(struct mbuf *);
928struct mbuf	*m_get2(int, int, short, int);
929struct mbuf	*m_getjcl(int, short, int, int);
930struct mbuf	*m_getm2(struct mbuf *, int, int, short, int);
931struct mbuf	*m_getptr(struct mbuf *, int, int *);
932u_int		 m_length(struct mbuf *, struct mbuf **);
933int		 m_mbuftouio(struct uio *, struct mbuf *, int);
934void		 m_move_pkthdr(struct mbuf *, struct mbuf *);
935struct mbuf	*m_prepend(struct mbuf *, int, int);
936void		 m_print(const struct mbuf *, int);
937struct mbuf	*m_pulldown(struct mbuf *, int, int, int *);
938struct mbuf	*m_pullup(struct mbuf *, int);
939int		 m_sanity(struct mbuf *, int);
940struct mbuf	*m_split(struct mbuf *, int, int);
941struct mbuf	*m_uiotombuf(struct uio *, int, int, int, int);
942struct mbuf	*m_unshare(struct mbuf *, int);
943
944/*-
945 * Network packets may have annotations attached by affixing a list of
946 * "packet tags" to the pkthdr structure.  Packet tags are dynamically
947 * allocated semi-opaque data structures that have a fixed header
948 * (struct m_tag) that specifies the size of the memory block and a
949 * <cookie,type> pair that identifies it.  The cookie is a 32-bit unique
950 * unsigned value used to identify a module or ABI.  By convention this value
951 * is chosen as the date+time that the module is created, expressed as the
952 * number of seconds since the epoch (e.g., using date -u +'%s').  The type
953 * value is an ABI/module-specific value that identifies a particular
954 * annotation and is private to the module.  For compatibility with systems
955 * like OpenBSD that define packet tags w/o an ABI/module cookie, the value
956 * PACKET_ABI_COMPAT is used to implement m_tag_get and m_tag_find
957 * compatibility shim functions and several tag types are defined below.
958 * Users that do not require compatibility should use a private cookie value
959 * so that packet tag-related definitions can be maintained privately.
960 *
961 * Note that the packet tag returned by m_tag_alloc has the default memory
962 * alignment implemented by malloc.  To reference private data one can use a
963 * construct like:
964 *
965 *	struct m_tag *mtag = m_tag_alloc(...);
966 *	struct foo *p = (struct foo *)(mtag+1);
967 *
968 * if the alignment of struct m_tag is sufficient for referencing members of
969 * struct foo.  Otherwise it is necessary to embed struct m_tag within the
970 * private data structure to insure proper alignment; e.g.,
971 *
972 *	struct foo {
973 *		struct m_tag	tag;
974 *		...
975 *	};
976 *	struct foo *p = (struct foo *) m_tag_alloc(...);
977 *	struct m_tag *mtag = &p->tag;
978 */
979
980/*
981 * Persistent tags stay with an mbuf until the mbuf is reclaimed.  Otherwise
982 * tags are expected to ``vanish'' when they pass through a network
983 * interface.  For most interfaces this happens normally as the tags are
984 * reclaimed when the mbuf is free'd.  However in some special cases
985 * reclaiming must be done manually.  An example is packets that pass through
986 * the loopback interface.  Also, one must be careful to do this when
987 * ``turning around'' packets (e.g., icmp_reflect).
988 *
989 * To mark a tag persistent bit-or this flag in when defining the tag id.
990 * The tag will then be treated as described above.
991 */
992#define	MTAG_PERSISTENT				0x800
993
994#define	PACKET_TAG_NONE				0  /* Nadda */
995
996/* Packet tags for use with PACKET_ABI_COMPAT. */
997#define	PACKET_TAG_IPSEC_IN_DONE		1  /* IPsec applied, in */
998#define	PACKET_TAG_IPSEC_OUT_DONE		2  /* IPsec applied, out */
999#define	PACKET_TAG_IPSEC_IN_CRYPTO_DONE		3  /* NIC IPsec crypto done */
1000#define	PACKET_TAG_IPSEC_OUT_CRYPTO_NEEDED	4  /* NIC IPsec crypto req'ed */
1001#define	PACKET_TAG_IPSEC_IN_COULD_DO_CRYPTO	5  /* NIC notifies IPsec */
1002#define	PACKET_TAG_IPSEC_PENDING_TDB		6  /* Reminder to do IPsec */
1003#define	PACKET_TAG_BRIDGE			7  /* Bridge processing done */
1004#define	PACKET_TAG_GIF				8  /* GIF processing done */
1005#define	PACKET_TAG_GRE				9  /* GRE processing done */
1006#define	PACKET_TAG_IN_PACKET_CHECKSUM		10 /* NIC checksumming done */
1007#define	PACKET_TAG_ENCAP			11 /* Encap.  processing */
1008#define	PACKET_TAG_IPSEC_SOCKET			12 /* IPSEC socket ref */
1009#define	PACKET_TAG_IPSEC_HISTORY		13 /* IPSEC history */
1010#define	PACKET_TAG_IPV6_INPUT			14 /* IPV6 input processing */
1011#define	PACKET_TAG_DUMMYNET			15 /* dummynet info */
1012#define	PACKET_TAG_DIVERT			17 /* divert info */
1013#define	PACKET_TAG_IPFORWARD			18 /* ipforward info */
1014#define	PACKET_TAG_MACLABEL	(19 | MTAG_PERSISTENT) /* MAC label */
1015#define	PACKET_TAG_PF		(21 | MTAG_PERSISTENT) /* PF/ALTQ information */
1016#define	PACKET_TAG_RTSOCKFAM			25 /* rtsock sa family */
1017#define	PACKET_TAG_IPOPTIONS			27 /* Saved IP options */
1018#define	PACKET_TAG_CARP				28 /* CARP info */
1019#define	PACKET_TAG_IPSEC_NAT_T_PORTS		29 /* two uint16_t */
1020#define	PACKET_TAG_ND_OUTGOING			30 /* ND outgoing */
1021
1022/* Specific cookies and tags. */
1023
1024/* Packet tag routines. */
1025struct m_tag	*m_tag_alloc(u_int32_t, int, int, int);
1026void		 m_tag_delete(struct mbuf *, struct m_tag *);
1027void		 m_tag_delete_chain(struct mbuf *, struct m_tag *);
1028void		 m_tag_free_default(struct m_tag *);
1029struct m_tag	*m_tag_locate(struct mbuf *, u_int32_t, int, struct m_tag *);
1030struct m_tag	*m_tag_copy(struct m_tag *, int);
1031int		 m_tag_copy_chain(struct mbuf *, struct mbuf *, int);
1032void		 m_tag_delete_nonpersistent(struct mbuf *);
1033
1034/*
1035 * Initialize the list of tags associated with an mbuf.
1036 */
1037static __inline void
1038m_tag_init(struct mbuf *m)
1039{
1040
1041	SLIST_INIT(&m->m_pkthdr.tags);
1042}
1043
1044/*
1045 * Set up the contents of a tag.  Note that this does not fill in the free
1046 * method; the caller is expected to do that.
1047 *
1048 * XXX probably should be called m_tag_init, but that was already taken.
1049 */
1050static __inline void
1051m_tag_setup(struct m_tag *t, u_int32_t cookie, int type, int len)
1052{
1053
1054	t->m_tag_id = type;
1055	t->m_tag_len = len;
1056	t->m_tag_cookie = cookie;
1057}
1058
1059/*
1060 * Reclaim resources associated with a tag.
1061 */
1062static __inline void
1063m_tag_free(struct m_tag *t)
1064{
1065
1066	(*t->m_tag_free)(t);
1067}
1068
1069/*
1070 * Return the first tag associated with an mbuf.
1071 */
1072static __inline struct m_tag *
1073m_tag_first(struct mbuf *m)
1074{
1075
1076	return (SLIST_FIRST(&m->m_pkthdr.tags));
1077}
1078
1079/*
1080 * Return the next tag in the list of tags associated with an mbuf.
1081 */
1082static __inline struct m_tag *
1083m_tag_next(struct mbuf *m, struct m_tag *t)
1084{
1085
1086	return (SLIST_NEXT(t, m_tag_link));
1087}
1088
1089/*
1090 * Prepend a tag to the list of tags associated with an mbuf.
1091 */
1092static __inline void
1093m_tag_prepend(struct mbuf *m, struct m_tag *t)
1094{
1095
1096	SLIST_INSERT_HEAD(&m->m_pkthdr.tags, t, m_tag_link);
1097}
1098
1099/*
1100 * Unlink a tag from the list of tags associated with an mbuf.
1101 */
1102static __inline void
1103m_tag_unlink(struct mbuf *m, struct m_tag *t)
1104{
1105
1106	SLIST_REMOVE(&m->m_pkthdr.tags, t, m_tag, m_tag_link);
1107}
1108
1109/* These are for OpenBSD compatibility. */
1110#define	MTAG_ABI_COMPAT		0		/* compatibility ABI */
1111
1112static __inline struct m_tag *
1113m_tag_get(int type, int length, int wait)
1114{
1115	return (m_tag_alloc(MTAG_ABI_COMPAT, type, length, wait));
1116}
1117
1118static __inline struct m_tag *
1119m_tag_find(struct mbuf *m, int type, struct m_tag *start)
1120{
1121	return (SLIST_EMPTY(&m->m_pkthdr.tags) ? (struct m_tag *)NULL :
1122	    m_tag_locate(m, MTAG_ABI_COMPAT, type, start));
1123}
1124
1125static __inline struct mbuf *
1126m_free(struct mbuf *m)
1127{
1128	struct mbuf *n = m->m_next;
1129
1130	if ((m->m_flags & (M_PKTHDR|M_NOFREE)) == (M_PKTHDR|M_NOFREE))
1131		m_tag_delete_chain(m, NULL);
1132	if (m->m_flags & M_EXT)
1133		mb_free_ext(m);
1134	else if ((m->m_flags & M_NOFREE) == 0)
1135		uma_zfree(zone_mbuf, m);
1136	return (n);
1137}
1138
1139static int inline
1140rt_m_getfib(struct mbuf *m)
1141{
1142	KASSERT(m->m_flags & M_PKTHDR , ("Attempt to get FIB from non header mbuf."));
1143	return (m->m_pkthdr.fibnum);
1144}
1145
1146#define M_GETFIB(_m)   rt_m_getfib(_m)
1147
1148#define M_SETFIB(_m, _fib) do {						\
1149        KASSERT((_m)->m_flags & M_PKTHDR, ("Attempt to set FIB on non header mbuf."));	\
1150	((_m)->m_pkthdr.fibnum) = (_fib);				\
1151} while (0)
1152
1153#endif /* _KERNEL */
1154
1155#ifdef MBUF_PROFILING
1156 void m_profile(struct mbuf *m);
1157 #define M_PROFILE(m) m_profile(m)
1158#else
1159 #define M_PROFILE(m)
1160#endif
1161
1162
1163#endif /* !_SYS_MBUF_H_ */
1164