xref: /freebsd/sys/sys/mbuf.h (revision 5abd6fdcfa8cd6bdce99672b3a421d9313870472)
1 /*-
2  * Copyright (c) 1982, 1986, 1988, 1993
3  *	The Regents of the University of California.
4  * All rights reserved.
5  *
6  * Redistribution and use in source and binary forms, with or without
7  * modification, are permitted provided that the following conditions
8  * are met:
9  * 1. Redistributions of source code must retain the above copyright
10  *    notice, this list of conditions and the following disclaimer.
11  * 2. Redistributions in binary form must reproduce the above copyright
12  *    notice, this list of conditions and the following disclaimer in the
13  *    documentation and/or other materials provided with the distribution.
14  * 3. Neither the name of the University nor the names of its contributors
15  *    may be used to endorse or promote products derived from this software
16  *    without specific prior written permission.
17  *
18  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
19  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
20  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
21  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
22  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
23  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
24  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
25  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
26  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
27  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
28  * SUCH DAMAGE.
29  *
30  *	@(#)mbuf.h	8.5 (Berkeley) 2/19/95
31  * $FreeBSD$
32  */
33 
34 #ifndef _SYS_MBUF_H_
35 #define	_SYS_MBUF_H_
36 
37 /* XXX: These includes suck. Sorry! */
38 #include <sys/queue.h>
39 #ifdef _KERNEL
40 #include <sys/systm.h>
41 #include <vm/uma.h>
42 #ifdef WITNESS
43 #include <sys/lock.h>
44 #endif
45 #endif
46 
47 /*
48  * Mbufs are of a single size, MSIZE (sys/param.h), which includes overhead.
49  * An mbuf may add a single "mbuf cluster" of size MCLBYTES (also in
50  * sys/param.h), which has no additional overhead and is used instead of the
51  * internal data area; this is done when at least MINCLSIZE of data must be
52  * stored.  Additionally, it is possible to allocate a separate buffer
53  * externally and attach it to the mbuf in a way similar to that of mbuf
54  * clusters.
55  *
56  * 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  */
83 struct 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  */
94 struct 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  */
109 struct 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  */
122 struct 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  */
169 struct 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  */
185 struct 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_act		m_nextpkt
205 #define	m_pkthdr	M_dat.MH.MH_pkthdr
206 #define	m_ext		M_dat.MH.MH_dat.MH_ext
207 #define	m_pktdat	M_dat.MH.MH_dat.MH_databuf
208 #define	m_dat		M_dat.M_databuf
209 
210 /*
211  * mbuf flags of global significance and layer crossing.
212  * Those of only protocol/layer specific significance are to be mapped
213  * to M_PROTO[1-12] and cleared at layer handoff boundaries.
214  * NB: Limited to the lower 24 bits.
215  */
216 #define	M_EXT		0x00000001 /* has associated external storage */
217 #define	M_PKTHDR	0x00000002 /* start of record */
218 #define	M_EOR		0x00000004 /* end of record */
219 #define	M_RDONLY	0x00000008 /* associated data is marked read-only */
220 #define	M_BCAST		0x00000010 /* send/received as link-level broadcast */
221 #define	M_MCAST		0x00000020 /* send/received as link-level multicast */
222 #define	M_PROMISC	0x00000040 /* packet was not for us */
223 #define	M_VLANTAG	0x00000080 /* ether_vtag is valid */
224 #define	M_FLOWID	0x00000100 /* deprecated: flowid is valid */
225 #define	M_NOFREE	0x00000200 /* do not free mbuf, embedded in cluster */
226 
227 #define	M_PROTO1	0x00001000 /* protocol-specific */
228 #define	M_PROTO2	0x00002000 /* protocol-specific */
229 #define	M_PROTO3	0x00004000 /* protocol-specific */
230 #define	M_PROTO4	0x00008000 /* protocol-specific */
231 #define	M_PROTO5	0x00010000 /* protocol-specific */
232 #define	M_PROTO6	0x00020000 /* protocol-specific */
233 #define	M_PROTO7	0x00040000 /* protocol-specific */
234 #define	M_PROTO8	0x00080000 /* protocol-specific */
235 #define	M_PROTO9	0x00100000 /* protocol-specific */
236 #define	M_PROTO10	0x00200000 /* protocol-specific */
237 #define	M_PROTO11	0x00400000 /* protocol-specific */
238 #define	M_PROTO12	0x00800000 /* protocol-specific */
239 
240 /*
241  * Flags to purge when crossing layers.
242  */
243 #define	M_PROTOFLAGS \
244     (M_PROTO1|M_PROTO2|M_PROTO3|M_PROTO4|M_PROTO5|M_PROTO6|M_PROTO7|M_PROTO8|\
245      M_PROTO9|M_PROTO10|M_PROTO11|M_PROTO12)
246 
247 /*
248  * Flags preserved when copying m_pkthdr.
249  */
250 #define M_COPYFLAGS \
251     (M_PKTHDR|M_EOR|M_RDONLY|M_BCAST|M_MCAST|M_VLANTAG|M_PROMISC| \
252      M_PROTOFLAGS)
253 
254 /*
255  * Mbuf flag description for use with printf(9) %b identifier.
256  */
257 #define	M_FLAG_BITS \
258     "\20\1M_EXT\2M_PKTHDR\3M_EOR\4M_RDONLY\5M_BCAST\6M_MCAST" \
259     "\7M_PROMISC\10M_VLANTAG\11M_FLOWID"
260 #define	M_FLAG_PROTOBITS \
261     "\15M_PROTO1\16M_PROTO2\17M_PROTO3\20M_PROTO4\21M_PROTO5" \
262     "\22M_PROTO6\23M_PROTO7\24M_PROTO8\25M_PROTO9\26M_PROTO10" \
263     "\27M_PROTO11\30M_PROTO12"
264 #define	M_FLAG_PRINTF (M_FLAG_BITS M_FLAG_PROTOBITS)
265 
266 /*
267  * Network interface cards are able to hash protocol fields (such as IPv4
268  * addresses and TCP port numbers) classify packets into flows.  These flows
269  * can then be used to maintain ordering while delivering packets to the OS
270  * via parallel input queues, as well as to provide a stateless affinity
271  * model.  NIC drivers can pass up the hash via m->m_pkthdr.flowid, and set
272  * m_flag fields to indicate how the hash should be interpreted by the
273  * network stack.
274  *
275  * Most NICs support RSS, which provides ordering and explicit affinity, and
276  * use the hash m_flag bits to indicate what header fields were covered by
277  * the hash.  M_HASHTYPE_OPAQUE can be set by non-RSS cards or configurations
278  * that provide an opaque flow identifier, allowing for ordering and
279  * distribution without explicit affinity.
280  */
281 #define	M_HASHTYPE_NONE			0
282 #define	M_HASHTYPE_RSS_IPV4		1	/* IPv4 2-tuple */
283 #define	M_HASHTYPE_RSS_TCP_IPV4		2	/* TCPv4 4-tuple */
284 #define	M_HASHTYPE_RSS_IPV6		3	/* IPv6 2-tuple */
285 #define	M_HASHTYPE_RSS_TCP_IPV6		4	/* TCPv6 4-tuple */
286 #define	M_HASHTYPE_RSS_IPV6_EX		5	/* IPv6 2-tuple + ext hdrs */
287 #define	M_HASHTYPE_RSS_TCP_IPV6_EX	6	/* TCPv6 4-tiple + ext hdrs */
288 #define	M_HASHTYPE_OPAQUE		255	/* ordering, not affinity */
289 
290 #define	M_HASHTYPE_CLEAR(m)	((m)->m_pkthdr.rsstype = 0)
291 #define	M_HASHTYPE_GET(m)	((m)->m_pkthdr.rsstype)
292 #define	M_HASHTYPE_SET(m, v)	((m)->m_pkthdr.rsstype = (v))
293 #define	M_HASHTYPE_TEST(m, v)	(M_HASHTYPE_GET(m) == (v))
294 
295 /*
296  * COS/QOS class and quality of service tags.
297  * It uses DSCP code points as base.
298  */
299 #define	QOS_DSCP_CS0		0x00
300 #define	QOS_DSCP_DEF		QOS_DSCP_CS0
301 #define	QOS_DSCP_CS1		0x20
302 #define	QOS_DSCP_AF11		0x28
303 #define	QOS_DSCP_AF12		0x30
304 #define	QOS_DSCP_AF13		0x38
305 #define	QOS_DSCP_CS2		0x40
306 #define	QOS_DSCP_AF21		0x48
307 #define	QOS_DSCP_AF22		0x50
308 #define	QOS_DSCP_AF23		0x58
309 #define	QOS_DSCP_CS3		0x60
310 #define	QOS_DSCP_AF31		0x68
311 #define	QOS_DSCP_AF32		0x70
312 #define	QOS_DSCP_AF33		0x78
313 #define	QOS_DSCP_CS4		0x80
314 #define	QOS_DSCP_AF41		0x88
315 #define	QOS_DSCP_AF42		0x90
316 #define	QOS_DSCP_AF43		0x98
317 #define	QOS_DSCP_CS5		0xa0
318 #define	QOS_DSCP_EF		0xb8
319 #define	QOS_DSCP_CS6		0xc0
320 #define	QOS_DSCP_CS7		0xe0
321 
322 /*
323  * External mbuf storage buffer types.
324  */
325 #define	EXT_CLUSTER	1	/* mbuf cluster */
326 #define	EXT_SFBUF	2	/* sendfile(2)'s sf_bufs */
327 #define	EXT_JUMBOP	3	/* jumbo cluster 4096 bytes */
328 #define	EXT_JUMBO9	4	/* jumbo cluster 9216 bytes */
329 #define	EXT_JUMBO16	5	/* jumbo cluster 16184 bytes */
330 #define	EXT_PACKET	6	/* mbuf+cluster from packet zone */
331 #define	EXT_MBUF	7	/* external mbuf reference (M_IOVEC) */
332 
333 #define	EXT_VENDOR1	224	/* for vendor-internal use */
334 #define	EXT_VENDOR2	225	/* for vendor-internal use */
335 #define	EXT_VENDOR3	226	/* for vendor-internal use */
336 #define	EXT_VENDOR4	227	/* for vendor-internal use */
337 
338 #define	EXT_EXP1	244	/* for experimental use */
339 #define	EXT_EXP2	245	/* for experimental use */
340 #define	EXT_EXP3	246	/* for experimental use */
341 #define	EXT_EXP4	247	/* for experimental use */
342 
343 #define	EXT_NET_DRV	252	/* custom ext_buf provided by net driver(s) */
344 #define	EXT_MOD_TYPE	253	/* custom module's ext_buf type */
345 #define	EXT_DISPOSABLE	254	/* can throw this buffer away w/page flipping */
346 #define	EXT_EXTREF	255	/* has externally maintained ref_cnt ptr */
347 
348 /*
349  * Flags for external mbuf buffer types.
350  * NB: limited to the lower 24 bits.
351  */
352 #define	EXT_FLAG_EMBREF		0x000001	/* embedded ref_cnt, notyet */
353 #define	EXT_FLAG_EXTREF		0x000002	/* external ref_cnt, notyet */
354 #define	EXT_FLAG_NOFREE		0x000010	/* don't free mbuf to pool, notyet */
355 
356 #define	EXT_FLAG_VENDOR1	0x010000	/* for vendor-internal use */
357 #define	EXT_FLAG_VENDOR2	0x020000	/* for vendor-internal use */
358 #define	EXT_FLAG_VENDOR3	0x040000	/* for vendor-internal use */
359 #define	EXT_FLAG_VENDOR4	0x080000	/* for vendor-internal use */
360 
361 #define	EXT_FLAG_EXP1		0x100000	/* for experimental use */
362 #define	EXT_FLAG_EXP2		0x200000	/* for experimental use */
363 #define	EXT_FLAG_EXP3		0x400000	/* for experimental use */
364 #define	EXT_FLAG_EXP4		0x800000	/* for experimental use */
365 
366 /*
367  * EXT flag description for use with printf(9) %b identifier.
368  */
369 #define	EXT_FLAG_BITS \
370     "\20\1EXT_FLAG_EMBREF\2EXT_FLAG_EXTREF\5EXT_FLAG_NOFREE" \
371     "\21EXT_FLAG_VENDOR1\22EXT_FLAG_VENDOR2\23EXT_FLAG_VENDOR3" \
372     "\24EXT_FLAG_VENDOR4\25EXT_FLAG_EXP1\26EXT_FLAG_EXP2\27EXT_FLAG_EXP3" \
373     "\30EXT_FLAG_EXP4"
374 
375 /*
376  * Return values for (*ext_free).
377  */
378 #define	EXT_FREE_OK	0	/* Normal return */
379 
380 /*
381  * Flags indicating checksum, segmentation and other offload work to be
382  * done, or already done, by hardware or lower layers.  It is split into
383  * separate inbound and outbound flags.
384  *
385  * Outbound flags that are set by upper protocol layers requesting lower
386  * layers, or ideally the hardware, to perform these offloading tasks.
387  * For outbound packets this field and its flags can be directly tested
388  * against if_data.ifi_hwassist.
389  */
390 #define	CSUM_IP			0x00000001	/* IP header checksum offload */
391 #define	CSUM_IP_UDP		0x00000002	/* UDP checksum offload */
392 #define	CSUM_IP_TCP		0x00000004	/* TCP checksum offload */
393 #define	CSUM_IP_SCTP		0x00000008	/* SCTP checksum offload */
394 #define	CSUM_IP_TSO		0x00000010	/* TCP segmentation offload */
395 #define	CSUM_IP_ISCSI		0x00000020	/* iSCSI checksum offload */
396 
397 #define	CSUM_IP6_UDP		0x00000200	/* UDP checksum offload */
398 #define	CSUM_IP6_TCP		0x00000400	/* TCP checksum offload */
399 #define	CSUM_IP6_SCTP		0x00000800	/* SCTP checksum offload */
400 #define	CSUM_IP6_TSO		0x00001000	/* TCP segmentation offload */
401 #define	CSUM_IP6_ISCSI		0x00002000	/* iSCSI checksum offload */
402 
403 /* Inbound checksum support where the checksum was verified by hardware. */
404 #define	CSUM_L3_CALC		0x01000000	/* calculated layer 3 csum */
405 #define	CSUM_L3_VALID		0x02000000	/* checksum is correct */
406 #define	CSUM_L4_CALC		0x04000000	/* calculated layer 4 csum */
407 #define	CSUM_L4_VALID		0x08000000	/* checksum is correct */
408 #define	CSUM_L5_CALC		0x10000000	/* calculated layer 5 csum */
409 #define	CSUM_L5_VALID		0x20000000	/* checksum is correct */
410 #define	CSUM_COALESED		0x40000000	/* contains merged segments */
411 
412 /*
413  * CSUM flag description for use with printf(9) %b identifier.
414  */
415 #define	CSUM_BITS \
416     "\20\1CSUM_IP\2CSUM_IP_UDP\3CSUM_IP_TCP\4CSUM_IP_SCTP\5CSUM_IP_TSO" \
417     "\6CSUM_IP_ISCSI" \
418     "\12CSUM_IP6_UDP\13CSUM_IP6_TCP\14CSUM_IP6_SCTP\15CSUM_IP6_TSO" \
419     "\16CSUM_IP6_ISCSI" \
420     "\31CSUM_L3_CALC\32CSUM_L3_VALID\33CSUM_L4_CALC\34CSUM_L4_VALID" \
421     "\35CSUM_L5_CALC\36CSUM_L5_VALID\37CSUM_COALESED"
422 
423 /* CSUM flags compatibility mappings. */
424 #define	CSUM_IP_CHECKED		CSUM_L3_CALC
425 #define	CSUM_IP_VALID		CSUM_L3_VALID
426 #define	CSUM_DATA_VALID		CSUM_L4_VALID
427 #define	CSUM_PSEUDO_HDR		CSUM_L4_CALC
428 #define	CSUM_SCTP_VALID		CSUM_L4_VALID
429 #define	CSUM_DELAY_DATA		(CSUM_TCP|CSUM_UDP)
430 #define	CSUM_DELAY_IP		CSUM_IP		/* Only v4, no v6 IP hdr csum */
431 #define	CSUM_DELAY_DATA_IPV6	(CSUM_TCP_IPV6|CSUM_UDP_IPV6)
432 #define	CSUM_DATA_VALID_IPV6	CSUM_DATA_VALID
433 #define	CSUM_TCP		CSUM_IP_TCP
434 #define	CSUM_UDP		CSUM_IP_UDP
435 #define	CSUM_SCTP		CSUM_IP_SCTP
436 #define	CSUM_TSO		(CSUM_IP_TSO|CSUM_IP6_TSO)
437 #define	CSUM_UDP_IPV6		CSUM_IP6_UDP
438 #define	CSUM_TCP_IPV6		CSUM_IP6_TCP
439 #define	CSUM_SCTP_IPV6		CSUM_IP6_SCTP
440 #define	CSUM_FRAGMENT		0x0		/* Unused */
441 
442 /*
443  * mbuf types describing the content of the mbuf (including external storage).
444  */
445 #define	MT_NOTMBUF	0	/* USED INTERNALLY ONLY! Object is not mbuf */
446 #define	MT_DATA		1	/* dynamic (data) allocation */
447 #define	MT_HEADER	MT_DATA	/* packet header, use M_PKTHDR instead */
448 
449 #define	MT_VENDOR1	4	/* for vendor-internal use */
450 #define	MT_VENDOR2	5	/* for vendor-internal use */
451 #define	MT_VENDOR3	6	/* for vendor-internal use */
452 #define	MT_VENDOR4	7	/* for vendor-internal use */
453 
454 #define	MT_SONAME	8	/* socket name */
455 
456 #define	MT_EXP1		9	/* for experimental use */
457 #define	MT_EXP2		10	/* for experimental use */
458 #define	MT_EXP3		11	/* for experimental use */
459 #define	MT_EXP4		12	/* for experimental use */
460 
461 #define	MT_CONTROL	14	/* extra-data protocol message */
462 #define	MT_OOBDATA	15	/* expedited data  */
463 #define	MT_NTYPES	16	/* number of mbuf types for mbtypes[] */
464 
465 #define	MT_NOINIT	255	/* Not a type but a flag to allocate
466 				   a non-initialized mbuf */
467 
468 /*
469  * Compatibility with historic mbuf allocator.
470  */
471 #define	MBTOM(how)	(how)
472 #define	M_DONTWAIT	M_NOWAIT
473 #define	M_TRYWAIT	M_WAITOK
474 #define	M_WAIT		M_WAITOK
475 
476 /*
477  * String names of mbuf-related UMA(9) and malloc(9) types.  Exposed to
478  * !_KERNEL so that monitoring tools can look up the zones with
479  * libmemstat(3).
480  */
481 #define	MBUF_MEM_NAME		"mbuf"
482 #define	MBUF_CLUSTER_MEM_NAME	"mbuf_cluster"
483 #define	MBUF_PACKET_MEM_NAME	"mbuf_packet"
484 #define	MBUF_JUMBOP_MEM_NAME	"mbuf_jumbo_page"
485 #define	MBUF_JUMBO9_MEM_NAME	"mbuf_jumbo_9k"
486 #define	MBUF_JUMBO16_MEM_NAME	"mbuf_jumbo_16k"
487 #define	MBUF_TAG_MEM_NAME	"mbuf_tag"
488 #define	MBUF_EXTREFCNT_MEM_NAME	"mbuf_ext_refcnt"
489 
490 #ifdef _KERNEL
491 
492 #ifdef WITNESS
493 #define	MBUF_CHECKSLEEP(how) do {					\
494 	if (how == M_WAITOK)						\
495 		WITNESS_WARN(WARN_GIANTOK | WARN_SLEEPOK, NULL,		\
496 		    "Sleeping in \"%s\"", __func__);			\
497 } while (0)
498 #else
499 #define	MBUF_CHECKSLEEP(how)
500 #endif
501 
502 /*
503  * Network buffer allocation API
504  *
505  * The rest of it is defined in kern/kern_mbuf.c
506  */
507 extern uma_zone_t	zone_mbuf;
508 extern uma_zone_t	zone_clust;
509 extern uma_zone_t	zone_pack;
510 extern uma_zone_t	zone_jumbop;
511 extern uma_zone_t	zone_jumbo9;
512 extern uma_zone_t	zone_jumbo16;
513 extern uma_zone_t	zone_ext_refcnt;
514 
515 void		 mb_free_ext(struct mbuf *);
516 int		 m_pkthdr_init(struct mbuf *, int);
517 
518 static __inline int
519 m_gettype(int size)
520 {
521 	int type;
522 
523 	switch (size) {
524 	case MSIZE:
525 		type = EXT_MBUF;
526 		break;
527 	case MCLBYTES:
528 		type = EXT_CLUSTER;
529 		break;
530 #if MJUMPAGESIZE != MCLBYTES
531 	case MJUMPAGESIZE:
532 		type = EXT_JUMBOP;
533 		break;
534 #endif
535 	case MJUM9BYTES:
536 		type = EXT_JUMBO9;
537 		break;
538 	case MJUM16BYTES:
539 		type = EXT_JUMBO16;
540 		break;
541 	default:
542 		panic("%s: invalid cluster size", __func__);
543 	}
544 
545 	return (type);
546 }
547 
548 /*
549  * Associated an external reference counted buffer with an mbuf.
550  */
551 static __inline void
552 m_extaddref(struct mbuf *m, caddr_t buf, u_int size, u_int *ref_cnt,
553     int (*freef)(struct mbuf *, void *, void *), void *arg1, void *arg2)
554 {
555 
556 	KASSERT(ref_cnt != NULL, ("%s: ref_cnt not provided", __func__));
557 
558 	atomic_add_int(ref_cnt, 1);
559 	m->m_flags |= M_EXT;
560 	m->m_ext.ext_buf = buf;
561 	m->m_ext.ref_cnt = ref_cnt;
562 	m->m_data = m->m_ext.ext_buf;
563 	m->m_ext.ext_size = size;
564 	m->m_ext.ext_free = freef;
565 	m->m_ext.ext_arg1 = arg1;
566 	m->m_ext.ext_arg2 = arg2;
567 	m->m_ext.ext_type = EXT_EXTREF;
568 }
569 
570 static __inline uma_zone_t
571 m_getzone(int size)
572 {
573 	uma_zone_t zone;
574 
575 	switch (size) {
576 	case MCLBYTES:
577 		zone = zone_clust;
578 		break;
579 #if MJUMPAGESIZE != MCLBYTES
580 	case MJUMPAGESIZE:
581 		zone = zone_jumbop;
582 		break;
583 #endif
584 	case MJUM9BYTES:
585 		zone = zone_jumbo9;
586 		break;
587 	case MJUM16BYTES:
588 		zone = zone_jumbo16;
589 		break;
590 	default:
591 		panic("%s: invalid cluster size", __func__);
592 	}
593 
594 	return (zone);
595 }
596 
597 /*
598  * Initialize an mbuf with linear storage.
599  *
600  * Inline because the consumer text overhead will be roughly the same to
601  * initialize or call a function with this many parameters and M_PKTHDR
602  * should go away with constant propagation for !MGETHDR.
603  */
604 static __inline int
605 m_init(struct mbuf *m, uma_zone_t zone, int size, int how, short type,
606     int flags)
607 {
608 	int error;
609 
610 	m->m_next = NULL;
611 	m->m_nextpkt = NULL;
612 	m->m_data = m->m_dat;
613 	m->m_len = 0;
614 	m->m_flags = flags;
615 	m->m_type = type;
616 	if (flags & M_PKTHDR) {
617 		if ((error = m_pkthdr_init(m, how)) != 0)
618 			return (error);
619 	}
620 
621 	return (0);
622 }
623 
624 static __inline struct mbuf *
625 m_get(int how, short type)
626 {
627 	struct mb_args args;
628 
629 	args.flags = 0;
630 	args.type = type;
631 	return (uma_zalloc_arg(zone_mbuf, &args, how));
632 }
633 
634 /*
635  * XXX This should be deprecated, very little use.
636  */
637 static __inline struct mbuf *
638 m_getclr(int how, short type)
639 {
640 	struct mbuf *m;
641 	struct mb_args args;
642 
643 	args.flags = 0;
644 	args.type = type;
645 	m = uma_zalloc_arg(zone_mbuf, &args, how);
646 	if (m != NULL)
647 		bzero(m->m_data, MLEN);
648 	return (m);
649 }
650 
651 static __inline struct mbuf *
652 m_gethdr(int how, short type)
653 {
654 	struct mb_args args;
655 
656 	args.flags = M_PKTHDR;
657 	args.type = type;
658 	return (uma_zalloc_arg(zone_mbuf, &args, how));
659 }
660 
661 static __inline struct mbuf *
662 m_getcl(int how, short type, int flags)
663 {
664 	struct mb_args args;
665 
666 	args.flags = flags;
667 	args.type = type;
668 	return (uma_zalloc_arg(zone_pack, &args, how));
669 }
670 
671 static __inline void
672 m_clget(struct mbuf *m, int how)
673 {
674 
675 	if (m->m_flags & M_EXT)
676 		printf("%s: %p mbuf already has cluster\n", __func__, m);
677 	m->m_ext.ext_buf = (char *)NULL;
678 	uma_zalloc_arg(zone_clust, m, how);
679 	/*
680 	 * On a cluster allocation failure, drain the packet zone and retry,
681 	 * we might be able to loosen a few clusters up on the drain.
682 	 */
683 	if ((how & M_NOWAIT) && (m->m_ext.ext_buf == NULL)) {
684 		zone_drain(zone_pack);
685 		uma_zalloc_arg(zone_clust, m, how);
686 	}
687 }
688 
689 /*
690  * m_cljget() is different from m_clget() as it can allocate clusters without
691  * attaching them to an mbuf.  In that case the return value is the pointer
692  * to the cluster of the requested size.  If an mbuf was specified, it gets
693  * the cluster attached to it and the return value can be safely ignored.
694  * For size it takes MCLBYTES, MJUMPAGESIZE, MJUM9BYTES, MJUM16BYTES.
695  */
696 static __inline void *
697 m_cljget(struct mbuf *m, int how, int size)
698 {
699 	uma_zone_t zone;
700 
701 	if (m && m->m_flags & M_EXT)
702 		printf("%s: %p mbuf already has cluster\n", __func__, m);
703 	if (m != NULL)
704 		m->m_ext.ext_buf = NULL;
705 
706 	zone = m_getzone(size);
707 	return (uma_zalloc_arg(zone, m, how));
708 }
709 
710 static __inline void
711 m_cljset(struct mbuf *m, void *cl, int type)
712 {
713 	uma_zone_t zone;
714 	int size;
715 
716 	switch (type) {
717 	case EXT_CLUSTER:
718 		size = MCLBYTES;
719 		zone = zone_clust;
720 		break;
721 #if MJUMPAGESIZE != MCLBYTES
722 	case EXT_JUMBOP:
723 		size = MJUMPAGESIZE;
724 		zone = zone_jumbop;
725 		break;
726 #endif
727 	case EXT_JUMBO9:
728 		size = MJUM9BYTES;
729 		zone = zone_jumbo9;
730 		break;
731 	case EXT_JUMBO16:
732 		size = MJUM16BYTES;
733 		zone = zone_jumbo16;
734 		break;
735 	default:
736 		panic("%s: unknown cluster type", __func__);
737 		break;
738 	}
739 
740 	m->m_data = m->m_ext.ext_buf = cl;
741 	m->m_ext.ext_free = m->m_ext.ext_arg1 = m->m_ext.ext_arg2 = NULL;
742 	m->m_ext.ext_size = size;
743 	m->m_ext.ext_type = type;
744 	m->m_ext.ext_flags = 0;
745 	m->m_ext.ref_cnt = uma_find_refcnt(zone, cl);
746 	m->m_flags |= M_EXT;
747 
748 }
749 
750 static __inline void
751 m_chtype(struct mbuf *m, short new_type)
752 {
753 
754 	m->m_type = new_type;
755 }
756 
757 static __inline void
758 m_clrprotoflags(struct mbuf *m)
759 {
760 
761 	m->m_flags &= ~M_PROTOFLAGS;
762 }
763 
764 static __inline struct mbuf *
765 m_last(struct mbuf *m)
766 {
767 
768 	while (m->m_next)
769 		m = m->m_next;
770 	return (m);
771 }
772 
773 /*
774  * mbuf, cluster, and external object allocation macros (for compatibility
775  * purposes).
776  */
777 #define	M_MOVE_PKTHDR(to, from)	m_move_pkthdr((to), (from))
778 #define	MGET(m, how, type)	((m) = m_get((how), (type)))
779 #define	MGETHDR(m, how, type)	((m) = m_gethdr((how), (type)))
780 #define	MCLGET(m, how)		m_clget((m), (how))
781 #define	MEXTADD(m, buf, size, free, arg1, arg2, flags, type)		\
782     (void )m_extadd((m), (caddr_t)(buf), (size), (free), (arg1), (arg2),\
783     (flags), (type), M_NOWAIT)
784 #define	m_getm(m, len, how, type)					\
785     m_getm2((m), (len), (how), (type), M_PKTHDR)
786 
787 /*
788  * Evaluate TRUE if it's safe to write to the mbuf m's data region (this can
789  * be both the local data payload, or an external buffer area, depending on
790  * whether M_EXT is set).
791  */
792 #define	M_WRITABLE(m)	(!((m)->m_flags & M_RDONLY) &&			\
793 			 (!(((m)->m_flags & M_EXT)) ||			\
794 			 (*((m)->m_ext.ref_cnt) == 1)) )		\
795 
796 /* Check if the supplied mbuf has a packet header, or else panic. */
797 #define	M_ASSERTPKTHDR(m)						\
798 	KASSERT((m) != NULL && (m)->m_flags & M_PKTHDR,			\
799 	    ("%s: no mbuf packet header!", __func__))
800 
801 /*
802  * Ensure that the supplied mbuf is a valid, non-free mbuf.
803  *
804  * XXX: Broken at the moment.  Need some UMA magic to make it work again.
805  */
806 #define	M_ASSERTVALID(m)						\
807 	KASSERT((((struct mbuf *)m)->m_flags & 0) == 0,			\
808 	    ("%s: attempted use of a free mbuf!", __func__))
809 
810 /*
811  * Set the m_data pointer of a newly-allocated mbuf (m_get/MGET) to place an
812  * object of the specified size at the end of the mbuf, longword aligned.
813  */
814 #define	M_ALIGN(m, len) do {						\
815 	KASSERT(!((m)->m_flags & (M_PKTHDR|M_EXT)),			\
816 		("%s: M_ALIGN not normal mbuf", __func__));		\
817 	KASSERT((m)->m_data == (m)->m_dat,				\
818 		("%s: M_ALIGN not a virgin mbuf", __func__));		\
819 	(m)->m_data += (MLEN - (len)) & ~(sizeof(long) - 1);		\
820 } while (0)
821 
822 /*
823  * As above, for mbufs allocated with m_gethdr/MGETHDR or initialized by
824  * M_DUP/MOVE_PKTHDR.
825  */
826 #define	MH_ALIGN(m, len) do {						\
827 	KASSERT((m)->m_flags & M_PKTHDR && !((m)->m_flags & M_EXT),	\
828 		("%s: MH_ALIGN not PKTHDR mbuf", __func__));		\
829 	KASSERT((m)->m_data == (m)->m_pktdat,				\
830 		("%s: MH_ALIGN not a virgin mbuf", __func__));		\
831 	(m)->m_data += (MHLEN - (len)) & ~(sizeof(long) - 1);		\
832 } while (0)
833 
834 /*
835  * As above, for mbuf with external storage.
836  */
837 #define	MEXT_ALIGN(m, len) do {						\
838 	KASSERT((m)->m_flags & M_EXT,					\
839 		("%s: MEXT_ALIGN not an M_EXT mbuf", __func__));	\
840 	KASSERT((m)->m_data == (m)->m_ext.ext_buf,			\
841 		("%s: MEXT_ALIGN not a virgin mbuf", __func__));	\
842 	(m)->m_data += ((m)->m_ext.ext_size - (len)) &			\
843 	    ~(sizeof(long) - 1); 					\
844 } while (0)
845 
846 /*
847  * Compute the amount of space available before the current start of data in
848  * 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_LEADINGSPACE(m)						\
854 	((m)->m_flags & M_EXT ?						\
855 	    (M_WRITABLE(m) ? (m)->m_data - (m)->m_ext.ext_buf : 0):	\
856 	    (m)->m_flags & M_PKTHDR ? (m)->m_data - (m)->m_pktdat :	\
857 	    (m)->m_data - (m)->m_dat)
858 
859 /*
860  * Compute the amount of space available after the end of data in an mbuf.
861  *
862  * The M_WRITABLE() is a temporary, conservative safety measure: the burden
863  * of checking writability of the mbuf data area rests solely with the caller.
864  */
865 #define	M_TRAILINGSPACE(m)						\
866 	((m)->m_flags & M_EXT ?						\
867 	    (M_WRITABLE(m) ? (m)->m_ext.ext_buf + (m)->m_ext.ext_size	\
868 		- ((m)->m_data + (m)->m_len) : 0) :			\
869 	    &(m)->m_dat[MLEN] - ((m)->m_data + (m)->m_len))
870 
871 /*
872  * Arrange to prepend space of size plen to mbuf m.  If a new mbuf must be
873  * allocated, how specifies whether to wait.  If the allocation fails, the
874  * original mbuf chain is freed and m is set to NULL.
875  */
876 #define	M_PREPEND(m, plen, how) do {					\
877 	struct mbuf **_mmp = &(m);					\
878 	struct mbuf *_mm = *_mmp;					\
879 	int _mplen = (plen);						\
880 	int __mhow = (how);						\
881 									\
882 	MBUF_CHECKSLEEP(how);						\
883 	if (M_LEADINGSPACE(_mm) >= _mplen) {				\
884 		_mm->m_data -= _mplen;					\
885 		_mm->m_len += _mplen;					\
886 	} else								\
887 		_mm = m_prepend(_mm, _mplen, __mhow);			\
888 	if (_mm != NULL && _mm->m_flags & M_PKTHDR)			\
889 		_mm->m_pkthdr.len += _mplen;				\
890 	*_mmp = _mm;							\
891 } while (0)
892 
893 /*
894  * Change mbuf to new type.  This is a relatively expensive operation and
895  * should be avoided.
896  */
897 #define	MCHTYPE(m, t)	m_chtype((m), (t))
898 
899 /* Length to m_copy to copy all. */
900 #define	M_COPYALL	1000000000
901 
902 /* Compatibility with 4.3. */
903 #define	m_copy(m, o, l)	m_copym((m), (o), (l), M_NOWAIT)
904 
905 extern int		max_datalen;	/* MHLEN - max_hdr */
906 extern int		max_hdr;	/* Largest link + protocol header */
907 extern int		max_linkhdr;	/* Largest link-level header */
908 extern int		max_protohdr;	/* Largest protocol header */
909 extern int		nmbclusters;	/* Maximum number of clusters */
910 
911 struct uio;
912 
913 void		 m_adj(struct mbuf *, int);
914 void		 m_align(struct mbuf *, int);
915 int		 m_apply(struct mbuf *, int, int,
916 		    int (*)(void *, void *, u_int), void *);
917 int		 m_append(struct mbuf *, int, c_caddr_t);
918 void		 m_cat(struct mbuf *, struct mbuf *);
919 int		 m_extadd(struct mbuf *, caddr_t, u_int,
920 		    int (*)(struct mbuf *, void *, void *), void *, void *,
921 		    int, int, int);
922 struct mbuf	*m_collapse(struct mbuf *, int, int);
923 void		 m_copyback(struct mbuf *, int, int, c_caddr_t);
924 void		 m_copydata(const struct mbuf *, int, int, caddr_t);
925 struct mbuf	*m_copym(struct mbuf *, int, int, int);
926 struct mbuf	*m_copymdata(struct mbuf *, struct mbuf *,
927 		    int, int, int, int);
928 struct mbuf	*m_copypacket(struct mbuf *, int);
929 void		 m_copy_pkthdr(struct mbuf *, struct mbuf *);
930 struct mbuf	*m_copyup(struct mbuf *, int, int);
931 struct mbuf	*m_defrag(struct mbuf *, int);
932 void		 m_demote(struct mbuf *, int);
933 struct mbuf	*m_devget(char *, int, int, struct ifnet *,
934 		    void (*)(char *, caddr_t, u_int));
935 struct mbuf	*m_dup(struct mbuf *, int);
936 int		 m_dup_pkthdr(struct mbuf *, struct mbuf *, int);
937 u_int		 m_fixhdr(struct mbuf *);
938 struct mbuf	*m_fragment(struct mbuf *, int, int);
939 void		 m_freem(struct mbuf *);
940 struct mbuf	*m_get2(int, int, short, int);
941 struct mbuf	*m_getjcl(int, short, int, int);
942 struct mbuf	*m_getm2(struct mbuf *, int, int, short, int);
943 struct mbuf	*m_getptr(struct mbuf *, int, int *);
944 u_int		 m_length(struct mbuf *, struct mbuf **);
945 int		 m_mbuftouio(struct uio *, struct mbuf *, int);
946 void		 m_move_pkthdr(struct mbuf *, struct mbuf *);
947 struct mbuf	*m_prepend(struct mbuf *, int, int);
948 void		 m_print(const struct mbuf *, int);
949 struct mbuf	*m_pulldown(struct mbuf *, int, int, int *);
950 struct mbuf	*m_pullup(struct mbuf *, int);
951 int		 m_sanity(struct mbuf *, int);
952 struct mbuf	*m_split(struct mbuf *, int, int);
953 struct mbuf	*m_uiotombuf(struct uio *, int, int, int, int);
954 struct mbuf	*m_unshare(struct mbuf *, int);
955 
956 /*-
957  * Network packets may have annotations attached by affixing a list of
958  * "packet tags" to the pkthdr structure.  Packet tags are dynamically
959  * allocated semi-opaque data structures that have a fixed header
960  * (struct m_tag) that specifies the size of the memory block and a
961  * <cookie,type> pair that identifies it.  The cookie is a 32-bit unique
962  * unsigned value used to identify a module or ABI.  By convention this value
963  * is chosen as the date+time that the module is created, expressed as the
964  * number of seconds since the epoch (e.g., using date -u +'%s').  The type
965  * value is an ABI/module-specific value that identifies a particular
966  * annotation and is private to the module.  For compatibility with systems
967  * like OpenBSD that define packet tags w/o an ABI/module cookie, the value
968  * PACKET_ABI_COMPAT is used to implement m_tag_get and m_tag_find
969  * compatibility shim functions and several tag types are defined below.
970  * Users that do not require compatibility should use a private cookie value
971  * so that packet tag-related definitions can be maintained privately.
972  *
973  * Note that the packet tag returned by m_tag_alloc has the default memory
974  * alignment implemented by malloc.  To reference private data one can use a
975  * construct like:
976  *
977  *	struct m_tag *mtag = m_tag_alloc(...);
978  *	struct foo *p = (struct foo *)(mtag+1);
979  *
980  * if the alignment of struct m_tag is sufficient for referencing members of
981  * struct foo.  Otherwise it is necessary to embed struct m_tag within the
982  * private data structure to insure proper alignment; e.g.,
983  *
984  *	struct foo {
985  *		struct m_tag	tag;
986  *		...
987  *	};
988  *	struct foo *p = (struct foo *) m_tag_alloc(...);
989  *	struct m_tag *mtag = &p->tag;
990  */
991 
992 /*
993  * Persistent tags stay with an mbuf until the mbuf is reclaimed.  Otherwise
994  * tags are expected to ``vanish'' when they pass through a network
995  * interface.  For most interfaces this happens normally as the tags are
996  * reclaimed when the mbuf is free'd.  However in some special cases
997  * reclaiming must be done manually.  An example is packets that pass through
998  * the loopback interface.  Also, one must be careful to do this when
999  * ``turning around'' packets (e.g., icmp_reflect).
1000  *
1001  * To mark a tag persistent bit-or this flag in when defining the tag id.
1002  * The tag will then be treated as described above.
1003  */
1004 #define	MTAG_PERSISTENT				0x800
1005 
1006 #define	PACKET_TAG_NONE				0  /* Nadda */
1007 
1008 /* Packet tags for use with PACKET_ABI_COMPAT. */
1009 #define	PACKET_TAG_IPSEC_IN_DONE		1  /* IPsec applied, in */
1010 #define	PACKET_TAG_IPSEC_OUT_DONE		2  /* IPsec applied, out */
1011 #define	PACKET_TAG_IPSEC_IN_CRYPTO_DONE		3  /* NIC IPsec crypto done */
1012 #define	PACKET_TAG_IPSEC_OUT_CRYPTO_NEEDED	4  /* NIC IPsec crypto req'ed */
1013 #define	PACKET_TAG_IPSEC_IN_COULD_DO_CRYPTO	5  /* NIC notifies IPsec */
1014 #define	PACKET_TAG_IPSEC_PENDING_TDB		6  /* Reminder to do IPsec */
1015 #define	PACKET_TAG_BRIDGE			7  /* Bridge processing done */
1016 #define	PACKET_TAG_GIF				8  /* GIF processing done */
1017 #define	PACKET_TAG_GRE				9  /* GRE processing done */
1018 #define	PACKET_TAG_IN_PACKET_CHECKSUM		10 /* NIC checksumming done */
1019 #define	PACKET_TAG_ENCAP			11 /* Encap.  processing */
1020 #define	PACKET_TAG_IPSEC_SOCKET			12 /* IPSEC socket ref */
1021 #define	PACKET_TAG_IPSEC_HISTORY		13 /* IPSEC history */
1022 #define	PACKET_TAG_IPV6_INPUT			14 /* IPV6 input processing */
1023 #define	PACKET_TAG_DUMMYNET			15 /* dummynet info */
1024 #define	PACKET_TAG_DIVERT			17 /* divert info */
1025 #define	PACKET_TAG_IPFORWARD			18 /* ipforward info */
1026 #define	PACKET_TAG_MACLABEL	(19 | MTAG_PERSISTENT) /* MAC label */
1027 #define	PACKET_TAG_PF		(21 | MTAG_PERSISTENT) /* PF/ALTQ information */
1028 #define	PACKET_TAG_RTSOCKFAM			25 /* rtsock sa family */
1029 #define	PACKET_TAG_IPOPTIONS			27 /* Saved IP options */
1030 #define	PACKET_TAG_CARP				28 /* CARP info */
1031 #define	PACKET_TAG_IPSEC_NAT_T_PORTS		29 /* two uint16_t */
1032 #define	PACKET_TAG_ND_OUTGOING			30 /* ND outgoing */
1033 
1034 /* Specific cookies and tags. */
1035 
1036 /* Packet tag routines. */
1037 struct m_tag	*m_tag_alloc(u_int32_t, int, int, int);
1038 void		 m_tag_delete(struct mbuf *, struct m_tag *);
1039 void		 m_tag_delete_chain(struct mbuf *, struct m_tag *);
1040 void		 m_tag_free_default(struct m_tag *);
1041 struct m_tag	*m_tag_locate(struct mbuf *, u_int32_t, int, struct m_tag *);
1042 struct m_tag	*m_tag_copy(struct m_tag *, int);
1043 int		 m_tag_copy_chain(struct mbuf *, struct mbuf *, int);
1044 void		 m_tag_delete_nonpersistent(struct mbuf *);
1045 
1046 /*
1047  * Initialize the list of tags associated with an mbuf.
1048  */
1049 static __inline void
1050 m_tag_init(struct mbuf *m)
1051 {
1052 
1053 	SLIST_INIT(&m->m_pkthdr.tags);
1054 }
1055 
1056 /*
1057  * Set up the contents of a tag.  Note that this does not fill in the free
1058  * method; the caller is expected to do that.
1059  *
1060  * XXX probably should be called m_tag_init, but that was already taken.
1061  */
1062 static __inline void
1063 m_tag_setup(struct m_tag *t, u_int32_t cookie, int type, int len)
1064 {
1065 
1066 	t->m_tag_id = type;
1067 	t->m_tag_len = len;
1068 	t->m_tag_cookie = cookie;
1069 }
1070 
1071 /*
1072  * Reclaim resources associated with a tag.
1073  */
1074 static __inline void
1075 m_tag_free(struct m_tag *t)
1076 {
1077 
1078 	(*t->m_tag_free)(t);
1079 }
1080 
1081 /*
1082  * Return the first tag associated with an mbuf.
1083  */
1084 static __inline struct m_tag *
1085 m_tag_first(struct mbuf *m)
1086 {
1087 
1088 	return (SLIST_FIRST(&m->m_pkthdr.tags));
1089 }
1090 
1091 /*
1092  * Return the next tag in the list of tags associated with an mbuf.
1093  */
1094 static __inline struct m_tag *
1095 m_tag_next(struct mbuf *m, struct m_tag *t)
1096 {
1097 
1098 	return (SLIST_NEXT(t, m_tag_link));
1099 }
1100 
1101 /*
1102  * Prepend a tag to the list of tags associated with an mbuf.
1103  */
1104 static __inline void
1105 m_tag_prepend(struct mbuf *m, struct m_tag *t)
1106 {
1107 
1108 	SLIST_INSERT_HEAD(&m->m_pkthdr.tags, t, m_tag_link);
1109 }
1110 
1111 /*
1112  * Unlink a tag from the list of tags associated with an mbuf.
1113  */
1114 static __inline void
1115 m_tag_unlink(struct mbuf *m, struct m_tag *t)
1116 {
1117 
1118 	SLIST_REMOVE(&m->m_pkthdr.tags, t, m_tag, m_tag_link);
1119 }
1120 
1121 /* These are for OpenBSD compatibility. */
1122 #define	MTAG_ABI_COMPAT		0		/* compatibility ABI */
1123 
1124 static __inline struct m_tag *
1125 m_tag_get(int type, int length, int wait)
1126 {
1127 	return (m_tag_alloc(MTAG_ABI_COMPAT, type, length, wait));
1128 }
1129 
1130 static __inline struct m_tag *
1131 m_tag_find(struct mbuf *m, int type, struct m_tag *start)
1132 {
1133 	return (SLIST_EMPTY(&m->m_pkthdr.tags) ? (struct m_tag *)NULL :
1134 	    m_tag_locate(m, MTAG_ABI_COMPAT, type, start));
1135 }
1136 
1137 static __inline struct mbuf *
1138 m_free(struct mbuf *m)
1139 {
1140 	struct mbuf *n = m->m_next;
1141 
1142 	if ((m->m_flags & (M_PKTHDR|M_NOFREE)) == (M_PKTHDR|M_NOFREE))
1143 		m_tag_delete_chain(m, NULL);
1144 	if (m->m_flags & M_EXT)
1145 		mb_free_ext(m);
1146 	else if ((m->m_flags & M_NOFREE) == 0)
1147 		uma_zfree(zone_mbuf, m);
1148 	return (n);
1149 }
1150 
1151 static int inline
1152 rt_m_getfib(struct mbuf *m)
1153 {
1154 	KASSERT(m->m_flags & M_PKTHDR , ("Attempt to get FIB from non header mbuf."));
1155 	return (m->m_pkthdr.fibnum);
1156 }
1157 
1158 #define M_GETFIB(_m)   rt_m_getfib(_m)
1159 
1160 #define M_SETFIB(_m, _fib) do {						\
1161         KASSERT((_m)->m_flags & M_PKTHDR, ("Attempt to set FIB on non header mbuf."));	\
1162 	((_m)->m_pkthdr.fibnum) = (_fib);				\
1163 } while (0)
1164 
1165 #endif /* _KERNEL */
1166 
1167 #ifdef MBUF_PROFILING
1168  void m_profile(struct mbuf *m);
1169  #define M_PROFILE(m) m_profile(m)
1170 #else
1171  #define M_PROFILE(m)
1172 #endif
1173 
1174 
1175 #endif /* !_SYS_MBUF_H_ */
1176