xref: /freebsd/sys/sys/mbuf.h (revision 5bd73b51076b5cb5a2c9810f76c1d7ed20c4460e)
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  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  */
170 struct m_ext {
171 	volatile u_int	*ext_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  */
186 struct 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_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_PROMISC|M_VLANTAG|M_FLOWID| \
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 /* Microsoft RSS standard hash types */
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 /* Non-standard RSS hash types */
290 #define	M_HASHTYPE_RSS_UDP_IPV4		7	/* IPv4 UDP 4-tuple */
291 #define	M_HASHTYPE_RSS_UDP_IPV4_EX	8	/* IPv4 UDP 4-tuple + ext hdrs */
292 #define	M_HASHTYPE_RSS_UDP_IPV6		9	/* IPv6 UDP 4-tuple */
293 #define	M_HASHTYPE_RSS_UDP_IPV6_EX	10	/* IPv6 UDP 4-tuple + ext hdrs */
294 
295 #define	M_HASHTYPE_OPAQUE		255	/* ordering, not affinity */
296 
297 #define	M_HASHTYPE_CLEAR(m)	((m)->m_pkthdr.rsstype = 0)
298 #define	M_HASHTYPE_GET(m)	((m)->m_pkthdr.rsstype)
299 #define	M_HASHTYPE_SET(m, v)	((m)->m_pkthdr.rsstype = (v))
300 #define	M_HASHTYPE_TEST(m, v)	(M_HASHTYPE_GET(m) == (v))
301 
302 /*
303  * COS/QOS class and quality of service tags.
304  * It uses DSCP code points as base.
305  */
306 #define	QOS_DSCP_CS0		0x00
307 #define	QOS_DSCP_DEF		QOS_DSCP_CS0
308 #define	QOS_DSCP_CS1		0x20
309 #define	QOS_DSCP_AF11		0x28
310 #define	QOS_DSCP_AF12		0x30
311 #define	QOS_DSCP_AF13		0x38
312 #define	QOS_DSCP_CS2		0x40
313 #define	QOS_DSCP_AF21		0x48
314 #define	QOS_DSCP_AF22		0x50
315 #define	QOS_DSCP_AF23		0x58
316 #define	QOS_DSCP_CS3		0x60
317 #define	QOS_DSCP_AF31		0x68
318 #define	QOS_DSCP_AF32		0x70
319 #define	QOS_DSCP_AF33		0x78
320 #define	QOS_DSCP_CS4		0x80
321 #define	QOS_DSCP_AF41		0x88
322 #define	QOS_DSCP_AF42		0x90
323 #define	QOS_DSCP_AF43		0x98
324 #define	QOS_DSCP_CS5		0xa0
325 #define	QOS_DSCP_EF		0xb8
326 #define	QOS_DSCP_CS6		0xc0
327 #define	QOS_DSCP_CS7		0xe0
328 
329 /*
330  * External mbuf storage buffer types.
331  */
332 #define	EXT_CLUSTER	1	/* mbuf cluster */
333 #define	EXT_SFBUF	2	/* sendfile(2)'s sf_bufs */
334 #define	EXT_JUMBOP	3	/* jumbo cluster 4096 bytes */
335 #define	EXT_JUMBO9	4	/* jumbo cluster 9216 bytes */
336 #define	EXT_JUMBO16	5	/* jumbo cluster 16184 bytes */
337 #define	EXT_PACKET	6	/* mbuf+cluster from packet zone */
338 #define	EXT_MBUF	7	/* external mbuf reference (M_IOVEC) */
339 
340 #define	EXT_VENDOR1	224	/* for vendor-internal use */
341 #define	EXT_VENDOR2	225	/* for vendor-internal use */
342 #define	EXT_VENDOR3	226	/* for vendor-internal use */
343 #define	EXT_VENDOR4	227	/* for vendor-internal use */
344 
345 #define	EXT_EXP1	244	/* for experimental use */
346 #define	EXT_EXP2	245	/* for experimental use */
347 #define	EXT_EXP3	246	/* for experimental use */
348 #define	EXT_EXP4	247	/* for experimental use */
349 
350 #define	EXT_NET_DRV	252	/* custom ext_buf provided by net driver(s) */
351 #define	EXT_MOD_TYPE	253	/* custom module's ext_buf type */
352 #define	EXT_DISPOSABLE	254	/* can throw this buffer away w/page flipping */
353 #define	EXT_EXTREF	255	/* has externally maintained ext_cnt ptr */
354 
355 /*
356  * Flags for external mbuf buffer types.
357  * NB: limited to the lower 24 bits.
358  */
359 #define	EXT_FLAG_EMBREF		0x000001	/* embedded ext_cnt, notyet */
360 #define	EXT_FLAG_EXTREF		0x000002	/* external ext_cnt, notyet */
361 #define	EXT_FLAG_NOFREE		0x000010	/* don't free mbuf to pool, notyet */
362 
363 #define	EXT_FLAG_VENDOR1	0x010000	/* for vendor-internal use */
364 #define	EXT_FLAG_VENDOR2	0x020000	/* for vendor-internal use */
365 #define	EXT_FLAG_VENDOR3	0x040000	/* for vendor-internal use */
366 #define	EXT_FLAG_VENDOR4	0x080000	/* for vendor-internal use */
367 
368 #define	EXT_FLAG_EXP1		0x100000	/* for experimental use */
369 #define	EXT_FLAG_EXP2		0x200000	/* for experimental use */
370 #define	EXT_FLAG_EXP3		0x400000	/* for experimental use */
371 #define	EXT_FLAG_EXP4		0x800000	/* for experimental use */
372 
373 /*
374  * EXT flag description for use with printf(9) %b identifier.
375  */
376 #define	EXT_FLAG_BITS \
377     "\20\1EXT_FLAG_EMBREF\2EXT_FLAG_EXTREF\5EXT_FLAG_NOFREE" \
378     "\21EXT_FLAG_VENDOR1\22EXT_FLAG_VENDOR2\23EXT_FLAG_VENDOR3" \
379     "\24EXT_FLAG_VENDOR4\25EXT_FLAG_EXP1\26EXT_FLAG_EXP2\27EXT_FLAG_EXP3" \
380     "\30EXT_FLAG_EXP4"
381 
382 /*
383  * External reference/free functions.
384  */
385 void sf_ext_ref(void *, void *);
386 void sf_ext_free(void *, void *);
387 
388 /*
389  * Flags indicating checksum, segmentation and other offload work to be
390  * done, or already done, by hardware or lower layers.  It is split into
391  * separate inbound and outbound flags.
392  *
393  * Outbound flags that are set by upper protocol layers requesting lower
394  * layers, or ideally the hardware, to perform these offloading tasks.
395  * For outbound packets this field and its flags can be directly tested
396  * against if_data.ifi_hwassist.
397  */
398 #define	CSUM_IP			0x00000001	/* IP header checksum offload */
399 #define	CSUM_IP_UDP		0x00000002	/* UDP checksum offload */
400 #define	CSUM_IP_TCP		0x00000004	/* TCP checksum offload */
401 #define	CSUM_IP_SCTP		0x00000008	/* SCTP checksum offload */
402 #define	CSUM_IP_TSO		0x00000010	/* TCP segmentation offload */
403 #define	CSUM_IP_ISCSI		0x00000020	/* iSCSI checksum offload */
404 
405 #define	CSUM_IP6_UDP		0x00000200	/* UDP checksum offload */
406 #define	CSUM_IP6_TCP		0x00000400	/* TCP checksum offload */
407 #define	CSUM_IP6_SCTP		0x00000800	/* SCTP checksum offload */
408 #define	CSUM_IP6_TSO		0x00001000	/* TCP segmentation offload */
409 #define	CSUM_IP6_ISCSI		0x00002000	/* iSCSI checksum offload */
410 
411 /* Inbound checksum support where the checksum was verified by hardware. */
412 #define	CSUM_L3_CALC		0x01000000	/* calculated layer 3 csum */
413 #define	CSUM_L3_VALID		0x02000000	/* checksum is correct */
414 #define	CSUM_L4_CALC		0x04000000	/* calculated layer 4 csum */
415 #define	CSUM_L4_VALID		0x08000000	/* checksum is correct */
416 #define	CSUM_L5_CALC		0x10000000	/* calculated layer 5 csum */
417 #define	CSUM_L5_VALID		0x20000000	/* checksum is correct */
418 #define	CSUM_COALESED		0x40000000	/* contains merged segments */
419 
420 /*
421  * CSUM flag description for use with printf(9) %b identifier.
422  */
423 #define	CSUM_BITS \
424     "\20\1CSUM_IP\2CSUM_IP_UDP\3CSUM_IP_TCP\4CSUM_IP_SCTP\5CSUM_IP_TSO" \
425     "\6CSUM_IP_ISCSI" \
426     "\12CSUM_IP6_UDP\13CSUM_IP6_TCP\14CSUM_IP6_SCTP\15CSUM_IP6_TSO" \
427     "\16CSUM_IP6_ISCSI" \
428     "\31CSUM_L3_CALC\32CSUM_L3_VALID\33CSUM_L4_CALC\34CSUM_L4_VALID" \
429     "\35CSUM_L5_CALC\36CSUM_L5_VALID\37CSUM_COALESED"
430 
431 /* CSUM flags compatibility mappings. */
432 #define	CSUM_IP_CHECKED		CSUM_L3_CALC
433 #define	CSUM_IP_VALID		CSUM_L3_VALID
434 #define	CSUM_DATA_VALID		CSUM_L4_VALID
435 #define	CSUM_PSEUDO_HDR		CSUM_L4_CALC
436 #define	CSUM_SCTP_VALID		CSUM_L4_VALID
437 #define	CSUM_DELAY_DATA		(CSUM_TCP|CSUM_UDP)
438 #define	CSUM_DELAY_IP		CSUM_IP		/* Only v4, no v6 IP hdr csum */
439 #define	CSUM_DELAY_DATA_IPV6	(CSUM_TCP_IPV6|CSUM_UDP_IPV6)
440 #define	CSUM_DATA_VALID_IPV6	CSUM_DATA_VALID
441 #define	CSUM_TCP		CSUM_IP_TCP
442 #define	CSUM_UDP		CSUM_IP_UDP
443 #define	CSUM_SCTP		CSUM_IP_SCTP
444 #define	CSUM_TSO		(CSUM_IP_TSO|CSUM_IP6_TSO)
445 #define	CSUM_UDP_IPV6		CSUM_IP6_UDP
446 #define	CSUM_TCP_IPV6		CSUM_IP6_TCP
447 #define	CSUM_SCTP_IPV6		CSUM_IP6_SCTP
448 
449 /*
450  * mbuf types describing the content of the mbuf (including external storage).
451  */
452 #define	MT_NOTMBUF	0	/* USED INTERNALLY ONLY! Object is not mbuf */
453 #define	MT_DATA		1	/* dynamic (data) allocation */
454 #define	MT_HEADER	MT_DATA	/* packet header, use M_PKTHDR instead */
455 
456 #define	MT_VENDOR1	4	/* for vendor-internal use */
457 #define	MT_VENDOR2	5	/* for vendor-internal use */
458 #define	MT_VENDOR3	6	/* for vendor-internal use */
459 #define	MT_VENDOR4	7	/* for vendor-internal use */
460 
461 #define	MT_SONAME	8	/* socket name */
462 
463 #define	MT_EXP1		9	/* for experimental use */
464 #define	MT_EXP2		10	/* for experimental use */
465 #define	MT_EXP3		11	/* for experimental use */
466 #define	MT_EXP4		12	/* for experimental use */
467 
468 #define	MT_CONTROL	14	/* extra-data protocol message */
469 #define	MT_OOBDATA	15	/* expedited data  */
470 #define	MT_NTYPES	16	/* number of mbuf types for mbtypes[] */
471 
472 #define	MT_NOINIT	255	/* Not a type but a flag to allocate
473 				   a non-initialized mbuf */
474 
475 /*
476  * String names of mbuf-related UMA(9) and malloc(9) types.  Exposed to
477  * !_KERNEL so that monitoring tools can look up the zones with
478  * libmemstat(3).
479  */
480 #define	MBUF_MEM_NAME		"mbuf"
481 #define	MBUF_CLUSTER_MEM_NAME	"mbuf_cluster"
482 #define	MBUF_PACKET_MEM_NAME	"mbuf_packet"
483 #define	MBUF_JUMBOP_MEM_NAME	"mbuf_jumbo_page"
484 #define	MBUF_JUMBO9_MEM_NAME	"mbuf_jumbo_9k"
485 #define	MBUF_JUMBO16_MEM_NAME	"mbuf_jumbo_16k"
486 #define	MBUF_TAG_MEM_NAME	"mbuf_tag"
487 #define	MBUF_EXTREFCNT_MEM_NAME	"mbuf_ext_refcnt"
488 
489 #ifdef _KERNEL
490 
491 #ifdef WITNESS
492 #define	MBUF_CHECKSLEEP(how) do {					\
493 	if (how == M_WAITOK)						\
494 		WITNESS_WARN(WARN_GIANTOK | WARN_SLEEPOK, NULL,		\
495 		    "Sleeping in \"%s\"", __func__);			\
496 } while (0)
497 #else
498 #define	MBUF_CHECKSLEEP(how)
499 #endif
500 
501 /*
502  * Network buffer allocation API
503  *
504  * The rest of it is defined in kern/kern_mbuf.c
505  */
506 extern uma_zone_t	zone_mbuf;
507 extern uma_zone_t	zone_clust;
508 extern uma_zone_t	zone_pack;
509 extern uma_zone_t	zone_jumbop;
510 extern uma_zone_t	zone_jumbo9;
511 extern uma_zone_t	zone_jumbo16;
512 extern uma_zone_t	zone_ext_refcnt;
513 
514 void		 mb_free_ext(struct mbuf *);
515 int		 m_pkthdr_init(struct mbuf *, int);
516 
517 static __inline int
518 m_gettype(int size)
519 {
520 	int type;
521 
522 	switch (size) {
523 	case MSIZE:
524 		type = EXT_MBUF;
525 		break;
526 	case MCLBYTES:
527 		type = EXT_CLUSTER;
528 		break;
529 #if MJUMPAGESIZE != MCLBYTES
530 	case MJUMPAGESIZE:
531 		type = EXT_JUMBOP;
532 		break;
533 #endif
534 	case MJUM9BYTES:
535 		type = EXT_JUMBO9;
536 		break;
537 	case MJUM16BYTES:
538 		type = EXT_JUMBO16;
539 		break;
540 	default:
541 		panic("%s: invalid cluster size %d", __func__, size);
542 	}
543 
544 	return (type);
545 }
546 
547 /*
548  * Associated an external reference counted buffer with an mbuf.
549  */
550 static __inline void
551 m_extaddref(struct mbuf *m, caddr_t buf, u_int size, u_int *ref_cnt,
552     void (*freef)(struct mbuf *, void *, void *), void *arg1, void *arg2)
553 {
554 
555 	KASSERT(ref_cnt != NULL, ("%s: ref_cnt not provided", __func__));
556 
557 	atomic_add_int(ref_cnt, 1);
558 	m->m_flags |= M_EXT;
559 	m->m_ext.ext_buf = buf;
560 	m->m_ext.ext_cnt = ref_cnt;
561 	m->m_data = m->m_ext.ext_buf;
562 	m->m_ext.ext_size = size;
563 	m->m_ext.ext_free = freef;
564 	m->m_ext.ext_arg1 = arg1;
565 	m->m_ext.ext_arg2 = arg2;
566 	m->m_ext.ext_type = EXT_EXTREF;
567 }
568 
569 static __inline uma_zone_t
570 m_getzone(int size)
571 {
572 	uma_zone_t zone;
573 
574 	switch (size) {
575 	case MCLBYTES:
576 		zone = zone_clust;
577 		break;
578 #if MJUMPAGESIZE != MCLBYTES
579 	case MJUMPAGESIZE:
580 		zone = zone_jumbop;
581 		break;
582 #endif
583 	case MJUM9BYTES:
584 		zone = zone_jumbo9;
585 		break;
586 	case MJUM16BYTES:
587 		zone = zone_jumbo16;
588 		break;
589 	default:
590 		panic("%s: invalid cluster size %d", __func__, size);
591 	}
592 
593 	return (zone);
594 }
595 
596 /*
597  * Initialize an mbuf with linear storage.
598  *
599  * Inline because the consumer text overhead will be roughly the same to
600  * initialize or call a function with this many parameters and M_PKTHDR
601  * should go away with constant propagation for !MGETHDR.
602  */
603 static __inline int
604 m_init(struct mbuf *m, uma_zone_t zone, int size, int how, short type,
605     int flags)
606 {
607 	int error;
608 
609 	m->m_next = NULL;
610 	m->m_nextpkt = NULL;
611 	m->m_data = m->m_dat;
612 	m->m_len = 0;
613 	m->m_flags = flags;
614 	m->m_type = type;
615 	if (flags & M_PKTHDR) {
616 		if ((error = m_pkthdr_init(m, how)) != 0)
617 			return (error);
618 	}
619 
620 	return (0);
621 }
622 
623 static __inline struct mbuf *
624 m_get(int how, short type)
625 {
626 	struct mb_args args;
627 
628 	args.flags = 0;
629 	args.type = type;
630 	return (uma_zalloc_arg(zone_mbuf, &args, how));
631 }
632 
633 /*
634  * XXX This should be deprecated, very little use.
635  */
636 static __inline struct mbuf *
637 m_getclr(int how, short type)
638 {
639 	struct mbuf *m;
640 	struct mb_args args;
641 
642 	args.flags = 0;
643 	args.type = type;
644 	m = uma_zalloc_arg(zone_mbuf, &args, how);
645 	if (m != NULL)
646 		bzero(m->m_data, MLEN);
647 	return (m);
648 }
649 
650 static __inline struct mbuf *
651 m_gethdr(int how, short type)
652 {
653 	struct mb_args args;
654 
655 	args.flags = M_PKTHDR;
656 	args.type = type;
657 	return (uma_zalloc_arg(zone_mbuf, &args, how));
658 }
659 
660 static __inline struct mbuf *
661 m_getcl(int how, short type, int flags)
662 {
663 	struct mb_args args;
664 
665 	args.flags = flags;
666 	args.type = type;
667 	return (uma_zalloc_arg(zone_pack, &args, how));
668 }
669 
670 static __inline void
671 m_clget(struct mbuf *m, int how)
672 {
673 
674 	if (m->m_flags & M_EXT)
675 		printf("%s: %p mbuf already has external storage\n", __func__, m);
676 	m->m_ext.ext_buf = (char *)NULL;
677 	uma_zalloc_arg(zone_clust, m, how);
678 	/*
679 	 * On a cluster allocation failure, drain the packet zone and retry,
680 	 * we might be able to loosen a few clusters up on the drain.
681 	 */
682 	if ((how & M_NOWAIT) && (m->m_ext.ext_buf == NULL)) {
683 		zone_drain(zone_pack);
684 		uma_zalloc_arg(zone_clust, m, how);
685 	}
686 }
687 
688 /*
689  * m_cljget() is different from m_clget() as it can allocate clusters without
690  * attaching them to an mbuf.  In that case the return value is the pointer
691  * to the cluster of the requested size.  If an mbuf was specified, it gets
692  * the cluster attached to it and the return value can be safely ignored.
693  * For size it takes MCLBYTES, MJUMPAGESIZE, MJUM9BYTES, MJUM16BYTES.
694  */
695 static __inline void *
696 m_cljget(struct mbuf *m, int how, int size)
697 {
698 	uma_zone_t zone;
699 
700 	if (m && m->m_flags & M_EXT)
701 		printf("%s: %p mbuf already has external storage\n", __func__, m);
702 	if (m != NULL)
703 		m->m_ext.ext_buf = NULL;
704 
705 	zone = m_getzone(size);
706 	return (uma_zalloc_arg(zone, m, how));
707 }
708 
709 static __inline void
710 m_cljset(struct mbuf *m, void *cl, int type)
711 {
712 	uma_zone_t zone;
713 	int size;
714 
715 	switch (type) {
716 	case EXT_CLUSTER:
717 		size = MCLBYTES;
718 		zone = zone_clust;
719 		break;
720 #if MJUMPAGESIZE != MCLBYTES
721 	case EXT_JUMBOP:
722 		size = MJUMPAGESIZE;
723 		zone = zone_jumbop;
724 		break;
725 #endif
726 	case EXT_JUMBO9:
727 		size = MJUM9BYTES;
728 		zone = zone_jumbo9;
729 		break;
730 	case EXT_JUMBO16:
731 		size = MJUM16BYTES;
732 		zone = zone_jumbo16;
733 		break;
734 	default:
735 		panic("%s: unknown cluster type %d", __func__, type);
736 		break;
737 	}
738 
739 	m->m_data = m->m_ext.ext_buf = cl;
740 	m->m_ext.ext_free = m->m_ext.ext_arg1 = m->m_ext.ext_arg2 = NULL;
741 	m->m_ext.ext_size = size;
742 	m->m_ext.ext_type = type;
743 	m->m_ext.ext_flags = 0;
744 	m->m_ext.ext_cnt = uma_find_refcnt(zone, cl);
745 	m->m_flags |= M_EXT;
746 
747 }
748 
749 static __inline void
750 m_chtype(struct mbuf *m, short new_type)
751 {
752 
753 	m->m_type = new_type;
754 }
755 
756 static __inline void
757 m_clrprotoflags(struct mbuf *m)
758 {
759 
760 	m->m_flags &= ~M_PROTOFLAGS;
761 }
762 
763 static __inline struct mbuf *
764 m_last(struct mbuf *m)
765 {
766 
767 	while (m->m_next)
768 		m = m->m_next;
769 	return (m);
770 }
771 
772 /*
773  * mbuf, cluster, and external object allocation macros (for compatibility
774  * purposes).
775  */
776 #define	M_MOVE_PKTHDR(to, from)	m_move_pkthdr((to), (from))
777 #define	MGET(m, how, type)	((m) = m_get((how), (type)))
778 #define	MGETHDR(m, how, type)	((m) = m_gethdr((how), (type)))
779 #define	MCLGET(m, how)		m_clget((m), (how))
780 #define	MEXTADD(m, buf, size, free, arg1, arg2, flags, type)		\
781     (void )m_extadd((m), (caddr_t)(buf), (size), (free), (arg1), (arg2),\
782     (flags), (type), M_NOWAIT)
783 #define	m_getm(m, len, how, type)					\
784     m_getm2((m), (len), (how), (type), M_PKTHDR)
785 
786 /*
787  * Evaluate TRUE if it's safe to write to the mbuf m's data region (this can
788  * be both the local data payload, or an external buffer area, depending on
789  * whether M_EXT is set).
790  */
791 #define	M_WRITABLE(m)	(!((m)->m_flags & M_RDONLY) &&			\
792 			 (!(((m)->m_flags & M_EXT)) ||			\
793 			 (*((m)->m_ext.ext_cnt) == 1)) )		\
794 
795 /* Check if the supplied mbuf has a packet header, or else panic. */
796 #define	M_ASSERTPKTHDR(m)						\
797 	KASSERT((m) != NULL && (m)->m_flags & M_PKTHDR,			\
798 	    ("%s: no mbuf packet header!", __func__))
799 
800 /*
801  * Ensure that the supplied mbuf is a valid, non-free mbuf.
802  *
803  * XXX: Broken at the moment.  Need some UMA magic to make it work again.
804  */
805 #define	M_ASSERTVALID(m)						\
806 	KASSERT((((struct mbuf *)m)->m_flags & 0) == 0,			\
807 	    ("%s: attempted use of a free mbuf!", __func__))
808 
809 /*
810  * Set the m_data pointer of a newly-allocated mbuf (m_get/MGET) to place an
811  * object of the specified size at the end of the mbuf, longword aligned.
812  */
813 #define	M_ALIGN(m, len) do {						\
814 	KASSERT(!((m)->m_flags & (M_PKTHDR|M_EXT)),			\
815 		("%s: M_ALIGN not normal mbuf", __func__));		\
816 	KASSERT((m)->m_data == (m)->m_dat,				\
817 		("%s: M_ALIGN not a virgin mbuf", __func__));		\
818 	(m)->m_data += (MLEN - (len)) & ~(sizeof(long) - 1);		\
819 } while (0)
820 
821 /*
822  * As above, for mbufs allocated with m_gethdr/MGETHDR or initialized by
823  * M_DUP/MOVE_PKTHDR.
824  */
825 #define	MH_ALIGN(m, len) do {						\
826 	KASSERT((m)->m_flags & M_PKTHDR && !((m)->m_flags & M_EXT),	\
827 		("%s: MH_ALIGN not PKTHDR mbuf", __func__));		\
828 	KASSERT((m)->m_data == (m)->m_pktdat,				\
829 		("%s: MH_ALIGN not a virgin mbuf", __func__));		\
830 	(m)->m_data += (MHLEN - (len)) & ~(sizeof(long) - 1);		\
831 } while (0)
832 
833 /*
834  * As above, for mbuf with external storage.
835  */
836 #define	MEXT_ALIGN(m, len) do {						\
837 	KASSERT((m)->m_flags & M_EXT,					\
838 		("%s: MEXT_ALIGN not an M_EXT mbuf", __func__));	\
839 	KASSERT((m)->m_data == (m)->m_ext.ext_buf,			\
840 		("%s: MEXT_ALIGN not a virgin mbuf", __func__));	\
841 	(m)->m_data += ((m)->m_ext.ext_size - (len)) &			\
842 	    ~(sizeof(long) - 1); 					\
843 } while (0)
844 
845 /*
846  * Return the address of the start of the buffer associated with an mbuf,
847  * handling external storage, packet-header mbufs, and regular data mbufs.
848  */
849 #define	M_START(m)							\
850 	(((m)->m_flags & M_EXT) ? (m)->m_ext.ext_buf :			\
851 	 ((m)->m_flags & M_PKTHDR) ? &(m)->m_pktdat[0] :		\
852 	 &(m)->m_dat[0])
853 
854 /*
855  * Return the size of the buffer associated with an mbuf, handling external
856  * storage, packet-header mbufs, and regular data mbufs.
857  */
858 #define	M_SIZE(m)							\
859 	(((m)->m_flags & M_EXT) ? (m)->m_ext.ext_size :			\
860 	 ((m)->m_flags & M_PKTHDR) ? MHLEN :				\
861 	 MLEN)
862 
863 /*
864  * Compute the amount of space available before the current start of data in
865  * an mbuf.
866  *
867  * The M_WRITABLE() is a temporary, conservative safety measure: the burden
868  * of checking writability of the mbuf data area rests solely with the caller.
869  *
870  * NB: In previous versions, M_LEADINGSPACE() would only check M_WRITABLE()
871  * for mbufs with external storage.  We now allow mbuf-embedded data to be
872  * read-only as well.
873  */
874 #define	M_LEADINGSPACE(m)						\
875 	(M_WRITABLE(m) ? ((m)->m_data - M_START(m)) : 0)
876 
877 /*
878  * Compute the amount of space available after the end of data in an mbuf.
879  *
880  * The M_WRITABLE() is a temporary, conservative safety measure: the burden
881  * of checking writability of the mbuf data area rests solely with the caller.
882  *
883  * NB: In previous versions, M_TRAILINGSPACE() would only check M_WRITABLE()
884  * for mbufs with external storage.  We now allow mbuf-embedded data to be
885  * read-only as well.
886  */
887 #define	M_TRAILINGSPACE(m)						\
888 	(M_WRITABLE(m) ?						\
889 	    ((M_START(m) + M_SIZE(m)) - ((m)->m_data + (m)->m_len)) : 0)
890 
891 /*
892  * Arrange to prepend space of size plen to mbuf m.  If a new mbuf must be
893  * allocated, how specifies whether to wait.  If the allocation fails, the
894  * original mbuf chain is freed and m is set to NULL.
895  */
896 #define	M_PREPEND(m, plen, how) do {					\
897 	struct mbuf **_mmp = &(m);					\
898 	struct mbuf *_mm = *_mmp;					\
899 	int _mplen = (plen);						\
900 	int __mhow = (how);						\
901 									\
902 	MBUF_CHECKSLEEP(how);						\
903 	if (M_LEADINGSPACE(_mm) >= _mplen) {				\
904 		_mm->m_data -= _mplen;					\
905 		_mm->m_len += _mplen;					\
906 	} else								\
907 		_mm = m_prepend(_mm, _mplen, __mhow);			\
908 	if (_mm != NULL && _mm->m_flags & M_PKTHDR)			\
909 		_mm->m_pkthdr.len += _mplen;				\
910 	*_mmp = _mm;							\
911 } while (0)
912 
913 /*
914  * Change mbuf to new type.  This is a relatively expensive operation and
915  * should be avoided.
916  */
917 #define	MCHTYPE(m, t)	m_chtype((m), (t))
918 
919 /* Length to m_copy to copy all. */
920 #define	M_COPYALL	1000000000
921 
922 /* Compatibility with 4.3. */
923 #define	m_copy(m, o, l)	m_copym((m), (o), (l), M_NOWAIT)
924 
925 extern int		max_datalen;	/* MHLEN - max_hdr */
926 extern int		max_hdr;	/* Largest link + protocol header */
927 extern int		max_linkhdr;	/* Largest link-level header */
928 extern int		max_protohdr;	/* Largest protocol header */
929 extern int		nmbclusters;	/* Maximum number of clusters */
930 
931 struct uio;
932 
933 void		 m_adj(struct mbuf *, int);
934 void		 m_align(struct mbuf *, int);
935 int		 m_apply(struct mbuf *, int, int,
936 		    int (*)(void *, void *, u_int), void *);
937 int		 m_append(struct mbuf *, int, c_caddr_t);
938 void		 m_cat(struct mbuf *, struct mbuf *);
939 void		 m_catpkt(struct mbuf *, struct mbuf *);
940 int		 m_extadd(struct mbuf *, caddr_t, u_int,
941 		    void (*)(struct mbuf *, void *, void *), void *, void *,
942 		    int, int, int);
943 struct mbuf	*m_collapse(struct mbuf *, int, int);
944 void		 m_copyback(struct mbuf *, int, int, c_caddr_t);
945 void		 m_copydata(const struct mbuf *, int, int, caddr_t);
946 struct mbuf	*m_copym(struct mbuf *, int, int, int);
947 struct mbuf	*m_copymdata(struct mbuf *, struct mbuf *,
948 		    int, int, int, int);
949 struct mbuf	*m_copypacket(struct mbuf *, int);
950 void		 m_copy_pkthdr(struct mbuf *, struct mbuf *);
951 struct mbuf	*m_copyup(struct mbuf *, int, int);
952 struct mbuf	*m_defrag(struct mbuf *, int);
953 void		 m_demote(struct mbuf *, int);
954 struct mbuf	*m_devget(char *, int, int, struct ifnet *,
955 		    void (*)(char *, caddr_t, u_int));
956 struct mbuf	*m_dup(struct mbuf *, int);
957 int		 m_dup_pkthdr(struct mbuf *, struct mbuf *, int);
958 u_int		 m_fixhdr(struct mbuf *);
959 struct mbuf	*m_fragment(struct mbuf *, int, int);
960 void		 m_freem(struct mbuf *);
961 struct mbuf	*m_get2(int, int, short, int);
962 struct mbuf	*m_getjcl(int, short, int, int);
963 struct mbuf	*m_getm2(struct mbuf *, int, int, short, int);
964 struct mbuf	*m_getptr(struct mbuf *, int, int *);
965 u_int		 m_length(struct mbuf *, struct mbuf **);
966 int		 m_mbuftouio(struct uio *, struct mbuf *, int);
967 void		 m_move_pkthdr(struct mbuf *, struct mbuf *);
968 struct mbuf	*m_prepend(struct mbuf *, int, int);
969 void		 m_print(const struct mbuf *, int);
970 struct mbuf	*m_pulldown(struct mbuf *, int, int, int *);
971 struct mbuf	*m_pullup(struct mbuf *, int);
972 int		 m_sanity(struct mbuf *, int);
973 struct mbuf	*m_split(struct mbuf *, int, int);
974 struct mbuf	*m_uiotombuf(struct uio *, int, int, int, int);
975 struct mbuf	*m_unshare(struct mbuf *, int);
976 
977 /*-
978  * Network packets may have annotations attached by affixing a list of
979  * "packet tags" to the pkthdr structure.  Packet tags are dynamically
980  * allocated semi-opaque data structures that have a fixed header
981  * (struct m_tag) that specifies the size of the memory block and a
982  * <cookie,type> pair that identifies it.  The cookie is a 32-bit unique
983  * unsigned value used to identify a module or ABI.  By convention this value
984  * is chosen as the date+time that the module is created, expressed as the
985  * number of seconds since the epoch (e.g., using date -u +'%s').  The type
986  * value is an ABI/module-specific value that identifies a particular
987  * annotation and is private to the module.  For compatibility with systems
988  * like OpenBSD that define packet tags w/o an ABI/module cookie, the value
989  * PACKET_ABI_COMPAT is used to implement m_tag_get and m_tag_find
990  * compatibility shim functions and several tag types are defined below.
991  * Users that do not require compatibility should use a private cookie value
992  * so that packet tag-related definitions can be maintained privately.
993  *
994  * Note that the packet tag returned by m_tag_alloc has the default memory
995  * alignment implemented by malloc.  To reference private data one can use a
996  * construct like:
997  *
998  *	struct m_tag *mtag = m_tag_alloc(...);
999  *	struct foo *p = (struct foo *)(mtag+1);
1000  *
1001  * if the alignment of struct m_tag is sufficient for referencing members of
1002  * struct foo.  Otherwise it is necessary to embed struct m_tag within the
1003  * private data structure to insure proper alignment; e.g.,
1004  *
1005  *	struct foo {
1006  *		struct m_tag	tag;
1007  *		...
1008  *	};
1009  *	struct foo *p = (struct foo *) m_tag_alloc(...);
1010  *	struct m_tag *mtag = &p->tag;
1011  */
1012 
1013 /*
1014  * Persistent tags stay with an mbuf until the mbuf is reclaimed.  Otherwise
1015  * tags are expected to ``vanish'' when they pass through a network
1016  * interface.  For most interfaces this happens normally as the tags are
1017  * reclaimed when the mbuf is free'd.  However in some special cases
1018  * reclaiming must be done manually.  An example is packets that pass through
1019  * the loopback interface.  Also, one must be careful to do this when
1020  * ``turning around'' packets (e.g., icmp_reflect).
1021  *
1022  * To mark a tag persistent bit-or this flag in when defining the tag id.
1023  * The tag will then be treated as described above.
1024  */
1025 #define	MTAG_PERSISTENT				0x800
1026 
1027 #define	PACKET_TAG_NONE				0  /* Nadda */
1028 
1029 /* Packet tags for use with PACKET_ABI_COMPAT. */
1030 #define	PACKET_TAG_IPSEC_IN_DONE		1  /* IPsec applied, in */
1031 #define	PACKET_TAG_IPSEC_OUT_DONE		2  /* IPsec applied, out */
1032 #define	PACKET_TAG_IPSEC_IN_CRYPTO_DONE		3  /* NIC IPsec crypto done */
1033 #define	PACKET_TAG_IPSEC_OUT_CRYPTO_NEEDED	4  /* NIC IPsec crypto req'ed */
1034 #define	PACKET_TAG_IPSEC_IN_COULD_DO_CRYPTO	5  /* NIC notifies IPsec */
1035 #define	PACKET_TAG_IPSEC_PENDING_TDB		6  /* Reminder to do IPsec */
1036 #define	PACKET_TAG_BRIDGE			7  /* Bridge processing done */
1037 #define	PACKET_TAG_GIF				8  /* GIF processing done */
1038 #define	PACKET_TAG_GRE				9  /* GRE processing done */
1039 #define	PACKET_TAG_IN_PACKET_CHECKSUM		10 /* NIC checksumming done */
1040 #define	PACKET_TAG_ENCAP			11 /* Encap.  processing */
1041 #define	PACKET_TAG_IPSEC_SOCKET			12 /* IPSEC socket ref */
1042 #define	PACKET_TAG_IPSEC_HISTORY		13 /* IPSEC history */
1043 #define	PACKET_TAG_IPV6_INPUT			14 /* IPV6 input processing */
1044 #define	PACKET_TAG_DUMMYNET			15 /* dummynet info */
1045 #define	PACKET_TAG_DIVERT			17 /* divert info */
1046 #define	PACKET_TAG_IPFORWARD			18 /* ipforward info */
1047 #define	PACKET_TAG_MACLABEL	(19 | MTAG_PERSISTENT) /* MAC label */
1048 #define	PACKET_TAG_PF		(21 | MTAG_PERSISTENT) /* PF/ALTQ information */
1049 #define	PACKET_TAG_RTSOCKFAM			25 /* rtsock sa family */
1050 #define	PACKET_TAG_IPOPTIONS			27 /* Saved IP options */
1051 #define	PACKET_TAG_CARP				28 /* CARP info */
1052 #define	PACKET_TAG_IPSEC_NAT_T_PORTS		29 /* two uint16_t */
1053 #define	PACKET_TAG_ND_OUTGOING			30 /* ND outgoing */
1054 
1055 /* Specific cookies and tags. */
1056 
1057 /* Packet tag routines. */
1058 struct m_tag	*m_tag_alloc(u_int32_t, int, int, int);
1059 void		 m_tag_delete(struct mbuf *, struct m_tag *);
1060 void		 m_tag_delete_chain(struct mbuf *, struct m_tag *);
1061 void		 m_tag_free_default(struct m_tag *);
1062 struct m_tag	*m_tag_locate(struct mbuf *, u_int32_t, int, struct m_tag *);
1063 struct m_tag	*m_tag_copy(struct m_tag *, int);
1064 int		 m_tag_copy_chain(struct mbuf *, struct mbuf *, int);
1065 void		 m_tag_delete_nonpersistent(struct mbuf *);
1066 
1067 /*
1068  * Initialize the list of tags associated with an mbuf.
1069  */
1070 static __inline void
1071 m_tag_init(struct mbuf *m)
1072 {
1073 
1074 	SLIST_INIT(&m->m_pkthdr.tags);
1075 }
1076 
1077 /*
1078  * Set up the contents of a tag.  Note that this does not fill in the free
1079  * method; the caller is expected to do that.
1080  *
1081  * XXX probably should be called m_tag_init, but that was already taken.
1082  */
1083 static __inline void
1084 m_tag_setup(struct m_tag *t, u_int32_t cookie, int type, int len)
1085 {
1086 
1087 	t->m_tag_id = type;
1088 	t->m_tag_len = len;
1089 	t->m_tag_cookie = cookie;
1090 }
1091 
1092 /*
1093  * Reclaim resources associated with a tag.
1094  */
1095 static __inline void
1096 m_tag_free(struct m_tag *t)
1097 {
1098 
1099 	(*t->m_tag_free)(t);
1100 }
1101 
1102 /*
1103  * Return the first tag associated with an mbuf.
1104  */
1105 static __inline struct m_tag *
1106 m_tag_first(struct mbuf *m)
1107 {
1108 
1109 	return (SLIST_FIRST(&m->m_pkthdr.tags));
1110 }
1111 
1112 /*
1113  * Return the next tag in the list of tags associated with an mbuf.
1114  */
1115 static __inline struct m_tag *
1116 m_tag_next(struct mbuf *m, struct m_tag *t)
1117 {
1118 
1119 	return (SLIST_NEXT(t, m_tag_link));
1120 }
1121 
1122 /*
1123  * Prepend a tag to the list of tags associated with an mbuf.
1124  */
1125 static __inline void
1126 m_tag_prepend(struct mbuf *m, struct m_tag *t)
1127 {
1128 
1129 	SLIST_INSERT_HEAD(&m->m_pkthdr.tags, t, m_tag_link);
1130 }
1131 
1132 /*
1133  * Unlink a tag from the list of tags associated with an mbuf.
1134  */
1135 static __inline void
1136 m_tag_unlink(struct mbuf *m, struct m_tag *t)
1137 {
1138 
1139 	SLIST_REMOVE(&m->m_pkthdr.tags, t, m_tag, m_tag_link);
1140 }
1141 
1142 /* These are for OpenBSD compatibility. */
1143 #define	MTAG_ABI_COMPAT		0		/* compatibility ABI */
1144 
1145 static __inline struct m_tag *
1146 m_tag_get(int type, int length, int wait)
1147 {
1148 	return (m_tag_alloc(MTAG_ABI_COMPAT, type, length, wait));
1149 }
1150 
1151 static __inline struct m_tag *
1152 m_tag_find(struct mbuf *m, int type, struct m_tag *start)
1153 {
1154 	return (SLIST_EMPTY(&m->m_pkthdr.tags) ? (struct m_tag *)NULL :
1155 	    m_tag_locate(m, MTAG_ABI_COMPAT, type, start));
1156 }
1157 
1158 static __inline struct mbuf *
1159 m_free(struct mbuf *m)
1160 {
1161 	struct mbuf *n = m->m_next;
1162 
1163 	if ((m->m_flags & (M_PKTHDR|M_NOFREE)) == (M_PKTHDR|M_NOFREE))
1164 		m_tag_delete_chain(m, NULL);
1165 	if (m->m_flags & M_EXT)
1166 		mb_free_ext(m);
1167 	else if ((m->m_flags & M_NOFREE) == 0)
1168 		uma_zfree(zone_mbuf, m);
1169 	return (n);
1170 }
1171 
1172 static int inline
1173 rt_m_getfib(struct mbuf *m)
1174 {
1175 	KASSERT(m->m_flags & M_PKTHDR , ("Attempt to get FIB from non header mbuf."));
1176 	return (m->m_pkthdr.fibnum);
1177 }
1178 
1179 #define M_GETFIB(_m)   rt_m_getfib(_m)
1180 
1181 #define M_SETFIB(_m, _fib) do {						\
1182         KASSERT((_m)->m_flags & M_PKTHDR, ("Attempt to set FIB on non header mbuf."));	\
1183 	((_m)->m_pkthdr.fibnum) = (_fib);				\
1184 } while (0)
1185 
1186 #endif /* _KERNEL */
1187 
1188 #ifdef MBUF_PROFILING
1189  void m_profile(struct mbuf *m);
1190  #define M_PROFILE(m) m_profile(m)
1191 #else
1192  #define M_PROFILE(m)
1193 #endif
1194 
1195 
1196 #endif /* !_SYS_MBUF_H_ */
1197