xref: /freebsd/sys/sys/mbuf.h (revision 2e5b60079b7d8c3ca68f1390cd90f305e651f8d3)
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  * Compile-time assertions in uipc_mbuf.c test these values to ensure that
65  * they are sensible.
66  */
67 struct mbuf;
68 #define	MHSIZE		offsetof(struct mbuf, M_dat.M_databuf)
69 #define	MPKTHSIZE	offsetof(struct mbuf, M_dat.MH.MH_dat.MH_databuf)
70 #define	MLEN		((int)(MSIZE - MHSIZE))
71 #define	MHLEN		((int)(MSIZE - MPKTHSIZE))
72 #define	MINCLSIZE	(MHLEN + 1)
73 
74 #ifdef _KERNEL
75 /*-
76  * Macro for type conversion: convert mbuf pointer to data pointer of correct
77  * type:
78  *
79  * mtod(m, t)	-- Convert mbuf pointer to data pointer of correct type.
80  * mtodo(m, o) -- Same as above but with offset 'o' into data.
81  */
82 #define	mtod(m, t)	((t)((m)->m_data))
83 #define	mtodo(m, o)	((void *)(((m)->m_data) + (o)))
84 
85 /*
86  * Argument structure passed to UMA routines during mbuf and packet
87  * allocations.
88  */
89 struct mb_args {
90 	int	flags;	/* Flags for mbuf being allocated */
91 	short	type;	/* Type of mbuf being allocated */
92 };
93 #endif /* _KERNEL */
94 
95 /*
96  * Packet tag structure (see below for details).
97  */
98 struct m_tag {
99 	SLIST_ENTRY(m_tag)	m_tag_link;	/* List of packet tags */
100 	u_int16_t		m_tag_id;	/* Tag ID */
101 	u_int16_t		m_tag_len;	/* Length of data */
102 	u_int32_t		m_tag_cookie;	/* ABI/Module ID */
103 	void			(*m_tag_free)(struct m_tag *);
104 };
105 
106 /*
107  * Record/packet header in first mbuf of chain; valid only if M_PKTHDR is set.
108  * Size ILP32: 48
109  *	 LP64: 56
110  * Compile-time assertions in uipc_mbuf.c test these values to ensure that
111  * they are correct.
112  */
113 struct pkthdr {
114 	struct ifnet	*rcvif;		/* rcv interface */
115 	SLIST_HEAD(packet_tags, m_tag) tags; /* list of packet tags */
116 	int32_t		 len;		/* total packet length */
117 
118 	/* Layer crossing persistent information. */
119 	uint32_t	 flowid;	/* packet's 4-tuple system */
120 	uint64_t	 csum_flags;	/* checksum and offload features */
121 	uint16_t	 fibnum;	/* this packet should use this fib */
122 	uint8_t		 cosqos;	/* class/quality of service */
123 	uint8_t		 rsstype;	/* hash type */
124 	uint8_t		 l2hlen;	/* layer 2 header length */
125 	uint8_t		 l3hlen;	/* layer 3 header length */
126 	uint8_t		 l4hlen;	/* layer 4 header length */
127 	uint8_t		 l5hlen;	/* layer 5 header length */
128 	union {
129 		uint8_t  eight[8];
130 		uint16_t sixteen[4];
131 		uint32_t thirtytwo[2];
132 		uint64_t sixtyfour[1];
133 		uintptr_t unintptr[1];
134 		void	*ptr;
135 	} PH_per;
136 
137 	/* Layer specific non-persistent local storage for reassembly, etc. */
138 	union {
139 		uint8_t  eight[8];
140 		uint16_t sixteen[4];
141 		uint32_t thirtytwo[2];
142 		uint64_t sixtyfour[1];
143 		uintptr_t unintptr[1];
144 		void 	*ptr;
145 	} PH_loc;
146 };
147 #define	ether_vtag	PH_per.sixteen[0]
148 #define	PH_vt		PH_per
149 #define	vt_nrecs	sixteen[0]
150 #define	tso_segsz	PH_per.sixteen[1]
151 #define	csum_phsum	PH_per.sixteen[2]
152 #define	csum_data	PH_per.thirtytwo[1]
153 #define	pkt_tcphdr	PH_loc.ptr
154 
155 /*
156  * Description of external storage mapped into mbuf; valid only if M_EXT is
157  * set.
158  * Size ILP32: 28
159  *	 LP64: 48
160  * Compile-time assertions in uipc_mbuf.c test these values to ensure that
161  * they are correct.
162  */
163 struct struct_m_ext {
164 	volatile u_int	*ext_cnt;	/* pointer to ref count info */
165 	caddr_t		 ext_buf;	/* start of buffer */
166 	uint32_t	 ext_size;	/* size of buffer, for ext_free */
167 	uint32_t	 ext_type:8,	/* type of external storage */
168 			 ext_flags:24;	/* external storage mbuf flags */
169 	void		(*ext_free)	/* free routine if not the usual */
170 			    (struct mbuf *, void *, void *);
171 	void		*ext_arg1;	/* optional argument pointer */
172 	void		*ext_arg2;	/* optional argument pointer */
173 };
174 
175 /*
176  * The core of the mbuf object along with some shortcut defines for practical
177  * purposes.
178  */
179 struct mbuf {
180 	/*
181 	 * Header present at the beginning of every mbuf.
182 	 * Size ILP32: 24
183 	 *      LP64: 32
184 	 * Compile-time assertions in uipc_mbuf.c test these values to ensure
185 	 * that they are correct.
186 	 */
187 	struct mbuf	*m_next;	/* next buffer in chain */
188 	struct mbuf	*m_nextpkt;	/* next chain in queue/record */
189 	caddr_t		 m_data;	/* location of data */
190 	int32_t		 m_len;		/* amount of data in this mbuf */
191 	uint32_t	 m_type:8,	/* type of data in this mbuf */
192 			 m_flags:24;	/* flags; see below */
193 #if !defined(__LP64__)
194 	uint32_t	 m_pad;		/* pad for 64bit alignment */
195 #endif
196 
197 	/*
198 	 * A set of optional headers (packet header, external storage header)
199 	 * and internal data storage.  Historically, these arrays were sized
200 	 * to MHLEN (space left after a packet header) and MLEN (space left
201 	 * after only a regular mbuf header); they are now variable size in
202 	 * order to support future work on variable-size mbufs.
203 	 */
204 	union {
205 		struct {
206 			struct pkthdr	MH_pkthdr;	/* M_PKTHDR set */
207 			union {
208 				struct struct_m_ext	MH_ext;	/* M_EXT set */
209 				char		MH_databuf[0];
210 			} MH_dat;
211 		} MH;
212 		char	M_databuf[0];			/* !M_PKTHDR, !M_EXT */
213 	} M_dat;
214 };
215 #define	m_pkthdr	M_dat.MH.MH_pkthdr
216 #define	m_ext		M_dat.MH.MH_dat.MH_ext
217 #define	m_pktdat	M_dat.MH.MH_dat.MH_databuf
218 #define	m_dat		M_dat.M_databuf
219 
220 /*
221  * mbuf flags of global significance and layer crossing.
222  * Those of only protocol/layer specific significance are to be mapped
223  * to M_PROTO[1-12] and cleared at layer handoff boundaries.
224  * NB: Limited to the lower 24 bits.
225  */
226 #define	M_EXT		0x00000001 /* has associated external storage */
227 #define	M_PKTHDR	0x00000002 /* start of record */
228 #define	M_EOR		0x00000004 /* end of record */
229 #define	M_RDONLY	0x00000008 /* associated data is marked read-only */
230 #define	M_BCAST		0x00000010 /* send/received as link-level broadcast */
231 #define	M_MCAST		0x00000020 /* send/received as link-level multicast */
232 #define	M_PROMISC	0x00000040 /* packet was not for us */
233 #define	M_VLANTAG	0x00000080 /* ether_vtag is valid */
234 #define	M_UNUSED_8	0x00000100 /* --available-- */
235 #define	M_NOFREE	0x00000200 /* do not free mbuf, embedded in cluster */
236 
237 #define	M_PROTO1	0x00001000 /* protocol-specific */
238 #define	M_PROTO2	0x00002000 /* protocol-specific */
239 #define	M_PROTO3	0x00004000 /* protocol-specific */
240 #define	M_PROTO4	0x00008000 /* protocol-specific */
241 #define	M_PROTO5	0x00010000 /* protocol-specific */
242 #define	M_PROTO6	0x00020000 /* protocol-specific */
243 #define	M_PROTO7	0x00040000 /* protocol-specific */
244 #define	M_PROTO8	0x00080000 /* protocol-specific */
245 #define	M_PROTO9	0x00100000 /* protocol-specific */
246 #define	M_PROTO10	0x00200000 /* protocol-specific */
247 #define	M_PROTO11	0x00400000 /* protocol-specific */
248 #define	M_PROTO12	0x00800000 /* protocol-specific */
249 
250 /*
251  * Flags to purge when crossing layers.
252  */
253 #define	M_PROTOFLAGS \
254     (M_PROTO1|M_PROTO2|M_PROTO3|M_PROTO4|M_PROTO5|M_PROTO6|M_PROTO7|M_PROTO8|\
255      M_PROTO9|M_PROTO10|M_PROTO11|M_PROTO12)
256 
257 /*
258  * Flags preserved when copying m_pkthdr.
259  */
260 #define M_COPYFLAGS \
261     (M_PKTHDR|M_EOR|M_RDONLY|M_BCAST|M_MCAST|M_PROMISC|M_VLANTAG| \
262      M_PROTOFLAGS)
263 
264 /*
265  * Mbuf flag description for use with printf(9) %b identifier.
266  */
267 #define	M_FLAG_BITS \
268     "\20\1M_EXT\2M_PKTHDR\3M_EOR\4M_RDONLY\5M_BCAST\6M_MCAST" \
269     "\7M_PROMISC\10M_VLANTAG"
270 #define	M_FLAG_PROTOBITS \
271     "\15M_PROTO1\16M_PROTO2\17M_PROTO3\20M_PROTO4\21M_PROTO5" \
272     "\22M_PROTO6\23M_PROTO7\24M_PROTO8\25M_PROTO9\26M_PROTO10" \
273     "\27M_PROTO11\30M_PROTO12"
274 #define	M_FLAG_PRINTF (M_FLAG_BITS M_FLAG_PROTOBITS)
275 
276 /*
277  * Network interface cards are able to hash protocol fields (such as IPv4
278  * addresses and TCP port numbers) classify packets into flows.  These flows
279  * can then be used to maintain ordering while delivering packets to the OS
280  * via parallel input queues, as well as to provide a stateless affinity
281  * model.  NIC drivers can pass up the hash via m->m_pkthdr.flowid, and set
282  * m_flag fields to indicate how the hash should be interpreted by the
283  * network stack.
284  *
285  * Most NICs support RSS, which provides ordering and explicit affinity, and
286  * use the hash m_flag bits to indicate what header fields were covered by
287  * the hash.  M_HASHTYPE_OPAQUE can be set by non-RSS cards or configurations
288  * that provide an opaque flow identifier, allowing for ordering and
289  * distribution without explicit affinity.
290  */
291 /* Microsoft RSS standard hash types */
292 #define	M_HASHTYPE_NONE			0
293 #define	M_HASHTYPE_RSS_IPV4		1	/* IPv4 2-tuple */
294 #define	M_HASHTYPE_RSS_TCP_IPV4		2	/* TCPv4 4-tuple */
295 #define	M_HASHTYPE_RSS_IPV6		3	/* IPv6 2-tuple */
296 #define	M_HASHTYPE_RSS_TCP_IPV6		4	/* TCPv6 4-tuple */
297 #define	M_HASHTYPE_RSS_IPV6_EX		5	/* IPv6 2-tuple + ext hdrs */
298 #define	M_HASHTYPE_RSS_TCP_IPV6_EX	6	/* TCPv6 4-tiple + ext hdrs */
299 /* Non-standard RSS hash types */
300 #define	M_HASHTYPE_RSS_UDP_IPV4		7	/* IPv4 UDP 4-tuple */
301 #define	M_HASHTYPE_RSS_UDP_IPV4_EX	8	/* IPv4 UDP 4-tuple + ext hdrs */
302 #define	M_HASHTYPE_RSS_UDP_IPV6		9	/* IPv6 UDP 4-tuple */
303 #define	M_HASHTYPE_RSS_UDP_IPV6_EX	10	/* IPv6 UDP 4-tuple + ext hdrs */
304 
305 #define	M_HASHTYPE_OPAQUE		255	/* ordering, not affinity */
306 
307 #define	M_HASHTYPE_CLEAR(m)	((m)->m_pkthdr.rsstype = 0)
308 #define	M_HASHTYPE_GET(m)	((m)->m_pkthdr.rsstype)
309 #define	M_HASHTYPE_SET(m, v)	((m)->m_pkthdr.rsstype = (v))
310 #define	M_HASHTYPE_TEST(m, v)	(M_HASHTYPE_GET(m) == (v))
311 
312 /*
313  * COS/QOS class and quality of service tags.
314  * It uses DSCP code points as base.
315  */
316 #define	QOS_DSCP_CS0		0x00
317 #define	QOS_DSCP_DEF		QOS_DSCP_CS0
318 #define	QOS_DSCP_CS1		0x20
319 #define	QOS_DSCP_AF11		0x28
320 #define	QOS_DSCP_AF12		0x30
321 #define	QOS_DSCP_AF13		0x38
322 #define	QOS_DSCP_CS2		0x40
323 #define	QOS_DSCP_AF21		0x48
324 #define	QOS_DSCP_AF22		0x50
325 #define	QOS_DSCP_AF23		0x58
326 #define	QOS_DSCP_CS3		0x60
327 #define	QOS_DSCP_AF31		0x68
328 #define	QOS_DSCP_AF32		0x70
329 #define	QOS_DSCP_AF33		0x78
330 #define	QOS_DSCP_CS4		0x80
331 #define	QOS_DSCP_AF41		0x88
332 #define	QOS_DSCP_AF42		0x90
333 #define	QOS_DSCP_AF43		0x98
334 #define	QOS_DSCP_CS5		0xa0
335 #define	QOS_DSCP_EF		0xb8
336 #define	QOS_DSCP_CS6		0xc0
337 #define	QOS_DSCP_CS7		0xe0
338 
339 /*
340  * External mbuf storage buffer types.
341  */
342 #define	EXT_CLUSTER	1	/* mbuf cluster */
343 #define	EXT_SFBUF	2	/* sendfile(2)'s sf_bufs */
344 #define	EXT_JUMBOP	3	/* jumbo cluster 4096 bytes */
345 #define	EXT_JUMBO9	4	/* jumbo cluster 9216 bytes */
346 #define	EXT_JUMBO16	5	/* jumbo cluster 16184 bytes */
347 #define	EXT_PACKET	6	/* mbuf+cluster from packet zone */
348 #define	EXT_MBUF	7	/* external mbuf reference (M_IOVEC) */
349 
350 #define	EXT_VENDOR1	224	/* for vendor-internal use */
351 #define	EXT_VENDOR2	225	/* for vendor-internal use */
352 #define	EXT_VENDOR3	226	/* for vendor-internal use */
353 #define	EXT_VENDOR4	227	/* for vendor-internal use */
354 
355 #define	EXT_EXP1	244	/* for experimental use */
356 #define	EXT_EXP2	245	/* for experimental use */
357 #define	EXT_EXP3	246	/* for experimental use */
358 #define	EXT_EXP4	247	/* for experimental use */
359 
360 #define	EXT_NET_DRV	252	/* custom ext_buf provided by net driver(s) */
361 #define	EXT_MOD_TYPE	253	/* custom module's ext_buf type */
362 #define	EXT_DISPOSABLE	254	/* can throw this buffer away w/page flipping */
363 #define	EXT_EXTREF	255	/* has externally maintained ext_cnt ptr */
364 
365 /*
366  * Flags for external mbuf buffer types.
367  * NB: limited to the lower 24 bits.
368  */
369 #define	EXT_FLAG_EMBREF		0x000001	/* embedded ext_cnt, notyet */
370 #define	EXT_FLAG_EXTREF		0x000002	/* external ext_cnt, notyet */
371 #define	EXT_FLAG_NOFREE		0x000010	/* don't free mbuf to pool, notyet */
372 
373 #define	EXT_FLAG_VENDOR1	0x010000	/* for vendor-internal use */
374 #define	EXT_FLAG_VENDOR2	0x020000	/* for vendor-internal use */
375 #define	EXT_FLAG_VENDOR3	0x040000	/* for vendor-internal use */
376 #define	EXT_FLAG_VENDOR4	0x080000	/* for vendor-internal use */
377 
378 #define	EXT_FLAG_EXP1		0x100000	/* for experimental use */
379 #define	EXT_FLAG_EXP2		0x200000	/* for experimental use */
380 #define	EXT_FLAG_EXP3		0x400000	/* for experimental use */
381 #define	EXT_FLAG_EXP4		0x800000	/* for experimental use */
382 
383 /*
384  * EXT flag description for use with printf(9) %b identifier.
385  */
386 #define	EXT_FLAG_BITS \
387     "\20\1EXT_FLAG_EMBREF\2EXT_FLAG_EXTREF\5EXT_FLAG_NOFREE" \
388     "\21EXT_FLAG_VENDOR1\22EXT_FLAG_VENDOR2\23EXT_FLAG_VENDOR3" \
389     "\24EXT_FLAG_VENDOR4\25EXT_FLAG_EXP1\26EXT_FLAG_EXP2\27EXT_FLAG_EXP3" \
390     "\30EXT_FLAG_EXP4"
391 
392 /*
393  * External reference/free functions.
394  */
395 void sf_ext_ref(void *, void *);
396 void sf_ext_free(void *, void *);
397 
398 /*
399  * Flags indicating checksum, segmentation and other offload work to be
400  * done, or already done, by hardware or lower layers.  It is split into
401  * separate inbound and outbound flags.
402  *
403  * Outbound flags that are set by upper protocol layers requesting lower
404  * layers, or ideally the hardware, to perform these offloading tasks.
405  * For outbound packets this field and its flags can be directly tested
406  * against ifnet if_hwassist.
407  */
408 #define	CSUM_IP			0x00000001	/* IP header checksum offload */
409 #define	CSUM_IP_UDP		0x00000002	/* UDP checksum offload */
410 #define	CSUM_IP_TCP		0x00000004	/* TCP checksum offload */
411 #define	CSUM_IP_SCTP		0x00000008	/* SCTP checksum offload */
412 #define	CSUM_IP_TSO		0x00000010	/* TCP segmentation offload */
413 #define	CSUM_IP_ISCSI		0x00000020	/* iSCSI checksum offload */
414 
415 #define	CSUM_IP6_UDP		0x00000200	/* UDP checksum offload */
416 #define	CSUM_IP6_TCP		0x00000400	/* TCP checksum offload */
417 #define	CSUM_IP6_SCTP		0x00000800	/* SCTP checksum offload */
418 #define	CSUM_IP6_TSO		0x00001000	/* TCP segmentation offload */
419 #define	CSUM_IP6_ISCSI		0x00002000	/* iSCSI checksum offload */
420 
421 /* Inbound checksum support where the checksum was verified by hardware. */
422 #define	CSUM_L3_CALC		0x01000000	/* calculated layer 3 csum */
423 #define	CSUM_L3_VALID		0x02000000	/* checksum is correct */
424 #define	CSUM_L4_CALC		0x04000000	/* calculated layer 4 csum */
425 #define	CSUM_L4_VALID		0x08000000	/* checksum is correct */
426 #define	CSUM_L5_CALC		0x10000000	/* calculated layer 5 csum */
427 #define	CSUM_L5_VALID		0x20000000	/* checksum is correct */
428 #define	CSUM_COALESED		0x40000000	/* contains merged segments */
429 
430 /*
431  * CSUM flag description for use with printf(9) %b identifier.
432  */
433 #define	CSUM_BITS \
434     "\20\1CSUM_IP\2CSUM_IP_UDP\3CSUM_IP_TCP\4CSUM_IP_SCTP\5CSUM_IP_TSO" \
435     "\6CSUM_IP_ISCSI" \
436     "\12CSUM_IP6_UDP\13CSUM_IP6_TCP\14CSUM_IP6_SCTP\15CSUM_IP6_TSO" \
437     "\16CSUM_IP6_ISCSI" \
438     "\31CSUM_L3_CALC\32CSUM_L3_VALID\33CSUM_L4_CALC\34CSUM_L4_VALID" \
439     "\35CSUM_L5_CALC\36CSUM_L5_VALID\37CSUM_COALESED"
440 
441 /* CSUM flags compatibility mappings. */
442 #define	CSUM_IP_CHECKED		CSUM_L3_CALC
443 #define	CSUM_IP_VALID		CSUM_L3_VALID
444 #define	CSUM_DATA_VALID		CSUM_L4_VALID
445 #define	CSUM_PSEUDO_HDR		CSUM_L4_CALC
446 #define	CSUM_SCTP_VALID		CSUM_L4_VALID
447 #define	CSUM_DELAY_DATA		(CSUM_TCP|CSUM_UDP)
448 #define	CSUM_DELAY_IP		CSUM_IP		/* Only v4, no v6 IP hdr csum */
449 #define	CSUM_DELAY_DATA_IPV6	(CSUM_TCP_IPV6|CSUM_UDP_IPV6)
450 #define	CSUM_DATA_VALID_IPV6	CSUM_DATA_VALID
451 #define	CSUM_TCP		CSUM_IP_TCP
452 #define	CSUM_UDP		CSUM_IP_UDP
453 #define	CSUM_SCTP		CSUM_IP_SCTP
454 #define	CSUM_TSO		(CSUM_IP_TSO|CSUM_IP6_TSO)
455 #define	CSUM_UDP_IPV6		CSUM_IP6_UDP
456 #define	CSUM_TCP_IPV6		CSUM_IP6_TCP
457 #define	CSUM_SCTP_IPV6		CSUM_IP6_SCTP
458 
459 /*
460  * mbuf types describing the content of the mbuf (including external storage).
461  */
462 #define	MT_NOTMBUF	0	/* USED INTERNALLY ONLY! Object is not mbuf */
463 #define	MT_DATA		1	/* dynamic (data) allocation */
464 #define	MT_HEADER	MT_DATA	/* packet header, use M_PKTHDR instead */
465 
466 #define	MT_VENDOR1	4	/* for vendor-internal use */
467 #define	MT_VENDOR2	5	/* for vendor-internal use */
468 #define	MT_VENDOR3	6	/* for vendor-internal use */
469 #define	MT_VENDOR4	7	/* for vendor-internal use */
470 
471 #define	MT_SONAME	8	/* socket name */
472 
473 #define	MT_EXP1		9	/* for experimental use */
474 #define	MT_EXP2		10	/* for experimental use */
475 #define	MT_EXP3		11	/* for experimental use */
476 #define	MT_EXP4		12	/* for experimental use */
477 
478 #define	MT_CONTROL	14	/* extra-data protocol message */
479 #define	MT_OOBDATA	15	/* expedited data  */
480 #define	MT_NTYPES	16	/* number of mbuf types for mbtypes[] */
481 
482 #define	MT_NOINIT	255	/* Not a type but a flag to allocate
483 				   a non-initialized mbuf */
484 
485 /*
486  * String names of mbuf-related UMA(9) and malloc(9) types.  Exposed to
487  * !_KERNEL so that monitoring tools can look up the zones with
488  * libmemstat(3).
489  */
490 #define	MBUF_MEM_NAME		"mbuf"
491 #define	MBUF_CLUSTER_MEM_NAME	"mbuf_cluster"
492 #define	MBUF_PACKET_MEM_NAME	"mbuf_packet"
493 #define	MBUF_JUMBOP_MEM_NAME	"mbuf_jumbo_page"
494 #define	MBUF_JUMBO9_MEM_NAME	"mbuf_jumbo_9k"
495 #define	MBUF_JUMBO16_MEM_NAME	"mbuf_jumbo_16k"
496 #define	MBUF_TAG_MEM_NAME	"mbuf_tag"
497 #define	MBUF_EXTREFCNT_MEM_NAME	"mbuf_ext_refcnt"
498 
499 #ifdef _KERNEL
500 
501 #ifdef WITNESS
502 #define	MBUF_CHECKSLEEP(how) do {					\
503 	if (how == M_WAITOK)						\
504 		WITNESS_WARN(WARN_GIANTOK | WARN_SLEEPOK, NULL,		\
505 		    "Sleeping in \"%s\"", __func__);			\
506 } while (0)
507 #else
508 #define	MBUF_CHECKSLEEP(how)
509 #endif
510 
511 /*
512  * Network buffer allocation API
513  *
514  * The rest of it is defined in kern/kern_mbuf.c
515  */
516 extern uma_zone_t	zone_mbuf;
517 extern uma_zone_t	zone_clust;
518 extern uma_zone_t	zone_pack;
519 extern uma_zone_t	zone_jumbop;
520 extern uma_zone_t	zone_jumbo9;
521 extern uma_zone_t	zone_jumbo16;
522 extern uma_zone_t	zone_ext_refcnt;
523 
524 void		 mb_free_ext(struct mbuf *);
525 int		 m_pkthdr_init(struct mbuf *, int);
526 
527 static __inline int
528 m_gettype(int size)
529 {
530 	int type;
531 
532 	switch (size) {
533 	case MSIZE:
534 		type = EXT_MBUF;
535 		break;
536 	case MCLBYTES:
537 		type = EXT_CLUSTER;
538 		break;
539 #if MJUMPAGESIZE != MCLBYTES
540 	case MJUMPAGESIZE:
541 		type = EXT_JUMBOP;
542 		break;
543 #endif
544 	case MJUM9BYTES:
545 		type = EXT_JUMBO9;
546 		break;
547 	case MJUM16BYTES:
548 		type = EXT_JUMBO16;
549 		break;
550 	default:
551 		panic("%s: invalid cluster size %d", __func__, size);
552 	}
553 
554 	return (type);
555 }
556 
557 /*
558  * Associated an external reference counted buffer with an mbuf.
559  */
560 static __inline void
561 m_extaddref(struct mbuf *m, caddr_t buf, u_int size, u_int *ref_cnt,
562     void (*freef)(struct mbuf *, void *, void *), void *arg1, void *arg2)
563 {
564 
565 	KASSERT(ref_cnt != NULL, ("%s: ref_cnt not provided", __func__));
566 
567 	atomic_add_int(ref_cnt, 1);
568 	m->m_flags |= M_EXT;
569 	m->m_ext.ext_buf = buf;
570 	m->m_ext.ext_cnt = ref_cnt;
571 	m->m_data = m->m_ext.ext_buf;
572 	m->m_ext.ext_size = size;
573 	m->m_ext.ext_free = freef;
574 	m->m_ext.ext_arg1 = arg1;
575 	m->m_ext.ext_arg2 = arg2;
576 	m->m_ext.ext_type = EXT_EXTREF;
577 }
578 
579 static __inline uma_zone_t
580 m_getzone(int size)
581 {
582 	uma_zone_t zone;
583 
584 	switch (size) {
585 	case MCLBYTES:
586 		zone = zone_clust;
587 		break;
588 #if MJUMPAGESIZE != MCLBYTES
589 	case MJUMPAGESIZE:
590 		zone = zone_jumbop;
591 		break;
592 #endif
593 	case MJUM9BYTES:
594 		zone = zone_jumbo9;
595 		break;
596 	case MJUM16BYTES:
597 		zone = zone_jumbo16;
598 		break;
599 	default:
600 		panic("%s: invalid cluster size %d", __func__, size);
601 	}
602 
603 	return (zone);
604 }
605 
606 /*
607  * Initialize an mbuf with linear storage.
608  *
609  * Inline because the consumer text overhead will be roughly the same to
610  * initialize or call a function with this many parameters and M_PKTHDR
611  * should go away with constant propagation for !MGETHDR.
612  */
613 static __inline int
614 m_init(struct mbuf *m, uma_zone_t zone, int size, int how, short type,
615     int flags)
616 {
617 	int error;
618 
619 	m->m_next = NULL;
620 	m->m_nextpkt = NULL;
621 	m->m_data = m->m_dat;
622 	m->m_len = 0;
623 	m->m_flags = flags;
624 	m->m_type = type;
625 	if (flags & M_PKTHDR) {
626 		if ((error = m_pkthdr_init(m, how)) != 0)
627 			return (error);
628 	}
629 
630 	return (0);
631 }
632 
633 static __inline struct mbuf *
634 m_get(int how, short type)
635 {
636 	struct mb_args args;
637 
638 	args.flags = 0;
639 	args.type = type;
640 	return (uma_zalloc_arg(zone_mbuf, &args, how));
641 }
642 
643 /*
644  * XXX This should be deprecated, very little use.
645  */
646 static __inline struct mbuf *
647 m_getclr(int how, short type)
648 {
649 	struct mbuf *m;
650 	struct mb_args args;
651 
652 	args.flags = 0;
653 	args.type = type;
654 	m = uma_zalloc_arg(zone_mbuf, &args, how);
655 	if (m != NULL)
656 		bzero(m->m_data, MLEN);
657 	return (m);
658 }
659 
660 static __inline struct mbuf *
661 m_gethdr(int how, short type)
662 {
663 	struct mb_args args;
664 
665 	args.flags = M_PKTHDR;
666 	args.type = type;
667 	return (uma_zalloc_arg(zone_mbuf, &args, how));
668 }
669 
670 static __inline struct mbuf *
671 m_getcl(int how, short type, int flags)
672 {
673 	struct mb_args args;
674 
675 	args.flags = flags;
676 	args.type = type;
677 	return (uma_zalloc_arg(zone_pack, &args, how));
678 }
679 
680 static __inline int
681 m_clget(struct mbuf *m, int how)
682 {
683 
684 	if (m->m_flags & M_EXT)
685 		printf("%s: %p mbuf already has external storage\n", __func__, m);
686 	m->m_ext.ext_buf = (char *)NULL;
687 	uma_zalloc_arg(zone_clust, m, how);
688 	/*
689 	 * On a cluster allocation failure, drain the packet zone and retry,
690 	 * we might be able to loosen a few clusters up on the drain.
691 	 */
692 	if ((how & M_NOWAIT) && (m->m_ext.ext_buf == NULL)) {
693 		zone_drain(zone_pack);
694 		uma_zalloc_arg(zone_clust, m, how);
695 	}
696 	return (m->m_flags & M_EXT);
697 }
698 
699 /*
700  * m_cljget() is different from m_clget() as it can allocate clusters without
701  * attaching them to an mbuf.  In that case the return value is the pointer
702  * to the cluster of the requested size.  If an mbuf was specified, it gets
703  * the cluster attached to it and the return value can be safely ignored.
704  * For size it takes MCLBYTES, MJUMPAGESIZE, MJUM9BYTES, MJUM16BYTES.
705  */
706 static __inline void *
707 m_cljget(struct mbuf *m, int how, int size)
708 {
709 	uma_zone_t zone;
710 
711 	if (m && m->m_flags & M_EXT)
712 		printf("%s: %p mbuf already has external storage\n", __func__, m);
713 	if (m != NULL)
714 		m->m_ext.ext_buf = NULL;
715 
716 	zone = m_getzone(size);
717 	return (uma_zalloc_arg(zone, m, how));
718 }
719 
720 static __inline void
721 m_cljset(struct mbuf *m, void *cl, int type)
722 {
723 	uma_zone_t zone;
724 	int size;
725 
726 	switch (type) {
727 	case EXT_CLUSTER:
728 		size = MCLBYTES;
729 		zone = zone_clust;
730 		break;
731 #if MJUMPAGESIZE != MCLBYTES
732 	case EXT_JUMBOP:
733 		size = MJUMPAGESIZE;
734 		zone = zone_jumbop;
735 		break;
736 #endif
737 	case EXT_JUMBO9:
738 		size = MJUM9BYTES;
739 		zone = zone_jumbo9;
740 		break;
741 	case EXT_JUMBO16:
742 		size = MJUM16BYTES;
743 		zone = zone_jumbo16;
744 		break;
745 	default:
746 		panic("%s: unknown cluster type %d", __func__, type);
747 		break;
748 	}
749 
750 	m->m_data = m->m_ext.ext_buf = cl;
751 	m->m_ext.ext_free = m->m_ext.ext_arg1 = m->m_ext.ext_arg2 = NULL;
752 	m->m_ext.ext_size = size;
753 	m->m_ext.ext_type = type;
754 	m->m_ext.ext_flags = 0;
755 	m->m_ext.ext_cnt = uma_find_refcnt(zone, cl);
756 	m->m_flags |= M_EXT;
757 
758 }
759 
760 static __inline void
761 m_chtype(struct mbuf *m, short new_type)
762 {
763 
764 	m->m_type = new_type;
765 }
766 
767 static __inline void
768 m_clrprotoflags(struct mbuf *m)
769 {
770 
771 	while (m) {
772 		m->m_flags &= ~M_PROTOFLAGS;
773 		m = m->m_next;
774 	}
775 }
776 
777 static __inline struct mbuf *
778 m_last(struct mbuf *m)
779 {
780 
781 	while (m->m_next)
782 		m = m->m_next;
783 	return (m);
784 }
785 
786 /*
787  * mbuf, cluster, and external object allocation macros (for compatibility
788  * purposes).
789  */
790 #define	M_MOVE_PKTHDR(to, from)	m_move_pkthdr((to), (from))
791 #define	MGET(m, how, type)	((m) = m_get((how), (type)))
792 #define	MGETHDR(m, how, type)	((m) = m_gethdr((how), (type)))
793 #define	MCLGET(m, how)		m_clget((m), (how))
794 #define	MEXTADD(m, buf, size, free, arg1, arg2, flags, type)		\
795     (void )m_extadd((m), (caddr_t)(buf), (size), (free), (arg1), (arg2),\
796     (flags), (type), M_NOWAIT)
797 #define	m_getm(m, len, how, type)					\
798     m_getm2((m), (len), (how), (type), M_PKTHDR)
799 
800 /*
801  * Evaluate TRUE if it's safe to write to the mbuf m's data region (this can
802  * be both the local data payload, or an external buffer area, depending on
803  * whether M_EXT is set).
804  */
805 #define	M_WRITABLE(m)	(!((m)->m_flags & M_RDONLY) &&			\
806 			 (!(((m)->m_flags & M_EXT)) ||			\
807 			 (*((m)->m_ext.ext_cnt) == 1)) )		\
808 
809 /* Check if the supplied mbuf has a packet header, or else panic. */
810 #define	M_ASSERTPKTHDR(m)						\
811 	KASSERT((m) != NULL && (m)->m_flags & M_PKTHDR,			\
812 	    ("%s: no mbuf packet header!", __func__))
813 
814 /*
815  * Ensure that the supplied mbuf is a valid, non-free mbuf.
816  *
817  * XXX: Broken at the moment.  Need some UMA magic to make it work again.
818  */
819 #define	M_ASSERTVALID(m)						\
820 	KASSERT((((struct mbuf *)m)->m_flags & 0) == 0,			\
821 	    ("%s: attempted use of a free mbuf!", __func__))
822 
823 /*
824  * Return the address of the start of the buffer associated with an mbuf,
825  * handling external storage, packet-header mbufs, and regular data mbufs.
826  */
827 #define	M_START(m)							\
828 	(((m)->m_flags & M_EXT) ? (m)->m_ext.ext_buf :			\
829 	 ((m)->m_flags & M_PKTHDR) ? &(m)->m_pktdat[0] :		\
830 	 &(m)->m_dat[0])
831 
832 /*
833  * Return the size of the buffer associated with an mbuf, handling external
834  * storage, packet-header mbufs, and regular data mbufs.
835  */
836 #define	M_SIZE(m)							\
837 	(((m)->m_flags & M_EXT) ? (m)->m_ext.ext_size :			\
838 	 ((m)->m_flags & M_PKTHDR) ? MHLEN :				\
839 	 MLEN)
840 
841 /*
842  * Set the m_data pointer of a newly allocated mbuf to place an object of the
843  * specified size at the end of the mbuf, longword aligned.
844  *
845  * NB: Historically, we had M_ALIGN(), MH_ALIGN(), and MEXT_ALIGN() as
846  * separate macros, each asserting that it was called at the proper moment.
847  * This required callers to themselves test the storage type and call the
848  * right one.  Rather than require callers to be aware of those layout
849  * decisions, we centralize here.
850  */
851 static __inline void
852 m_align(struct mbuf *m, int len)
853 {
854 #ifdef INVARIANTS
855 	const char *msg = "%s: not a virgin mbuf";
856 #endif
857 	int adjust;
858 
859 	KASSERT(m->m_data == M_START(m), (msg, __func__));
860 
861 	adjust = M_SIZE(m) - len;
862 	m->m_data += adjust &~ (sizeof(long)-1);
863 }
864 
865 #define	M_ALIGN(m, len)		m_align(m, len)
866 #define	MH_ALIGN(m, len)	m_align(m, len)
867 #define	MEXT_ALIGN(m, len)	m_align(m, len)
868 
869 /*
870  * Compute the amount of space available before the current start of data in
871  * an mbuf.
872  *
873  * The M_WRITABLE() is a temporary, conservative safety measure: the burden
874  * of checking writability of the mbuf data area rests solely with the caller.
875  *
876  * NB: In previous versions, M_LEADINGSPACE() would only check M_WRITABLE()
877  * for mbufs with external storage.  We now allow mbuf-embedded data to be
878  * read-only as well.
879  */
880 #define	M_LEADINGSPACE(m)						\
881 	(M_WRITABLE(m) ? ((m)->m_data - M_START(m)) : 0)
882 
883 /*
884  * Compute the amount of space available after the end of data in an mbuf.
885  *
886  * The M_WRITABLE() is a temporary, conservative safety measure: the burden
887  * of checking writability of the mbuf data area rests solely with the caller.
888  *
889  * NB: In previous versions, M_TRAILINGSPACE() would only check M_WRITABLE()
890  * for mbufs with external storage.  We now allow mbuf-embedded data to be
891  * read-only as well.
892  */
893 #define	M_TRAILINGSPACE(m)						\
894 	(M_WRITABLE(m) ?						\
895 	    ((M_START(m) + M_SIZE(m)) - ((m)->m_data + (m)->m_len)) : 0)
896 
897 /*
898  * Arrange to prepend space of size plen to mbuf m.  If a new mbuf must be
899  * allocated, how specifies whether to wait.  If the allocation fails, the
900  * original mbuf chain is freed and m is set to NULL.
901  */
902 #define	M_PREPEND(m, plen, how) do {					\
903 	struct mbuf **_mmp = &(m);					\
904 	struct mbuf *_mm = *_mmp;					\
905 	int _mplen = (plen);						\
906 	int __mhow = (how);						\
907 									\
908 	MBUF_CHECKSLEEP(how);						\
909 	if (M_LEADINGSPACE(_mm) >= _mplen) {				\
910 		_mm->m_data -= _mplen;					\
911 		_mm->m_len += _mplen;					\
912 	} else								\
913 		_mm = m_prepend(_mm, _mplen, __mhow);			\
914 	if (_mm != NULL && _mm->m_flags & M_PKTHDR)			\
915 		_mm->m_pkthdr.len += _mplen;				\
916 	*_mmp = _mm;							\
917 } while (0)
918 
919 /*
920  * Change mbuf to new type.  This is a relatively expensive operation and
921  * should be avoided.
922  */
923 #define	MCHTYPE(m, t)	m_chtype((m), (t))
924 
925 /* Length to m_copy to copy all. */
926 #define	M_COPYALL	1000000000
927 
928 /* Compatibility with 4.3. */
929 #define	m_copy(m, o, l)	m_copym((m), (o), (l), M_NOWAIT)
930 
931 extern int		max_datalen;	/* MHLEN - max_hdr */
932 extern int		max_hdr;	/* Largest link + protocol header */
933 extern int		max_linkhdr;	/* Largest link-level header */
934 extern int		max_protohdr;	/* Largest protocol header */
935 extern int		nmbclusters;	/* Maximum number of clusters */
936 
937 struct uio;
938 
939 void		 m_adj(struct mbuf *, int);
940 int		 m_apply(struct mbuf *, int, int,
941 		    int (*)(void *, void *, u_int), void *);
942 int		 m_append(struct mbuf *, int, c_caddr_t);
943 void		 m_cat(struct mbuf *, struct mbuf *);
944 void		 m_catpkt(struct mbuf *, struct mbuf *);
945 int		 m_extadd(struct mbuf *, caddr_t, u_int,
946 		    void (*)(struct mbuf *, void *, void *), void *, void *,
947 		    int, int, int);
948 struct mbuf	*m_collapse(struct mbuf *, int, int);
949 void		 m_copyback(struct mbuf *, int, int, c_caddr_t);
950 void		 m_copydata(const struct mbuf *, int, int, caddr_t);
951 struct mbuf	*m_copym(struct mbuf *, int, int, int);
952 struct mbuf	*m_copypacket(struct mbuf *, int);
953 void		 m_copy_pkthdr(struct mbuf *, struct mbuf *);
954 struct mbuf	*m_copyup(struct mbuf *, int, int);
955 struct mbuf	*m_defrag(struct mbuf *, int);
956 void		 m_demote(struct mbuf *, int, int);
957 struct mbuf	*m_devget(char *, int, int, struct ifnet *,
958 		    void (*)(char *, caddr_t, u_int));
959 struct mbuf	*m_dup(struct mbuf *, int);
960 int		 m_dup_pkthdr(struct mbuf *, struct mbuf *, int);
961 u_int		 m_fixhdr(struct mbuf *);
962 struct mbuf	*m_fragment(struct mbuf *, int, int);
963 void		 m_freem(struct mbuf *);
964 struct mbuf	*m_get2(int, int, short, int);
965 struct mbuf	*m_getjcl(int, short, int, int);
966 struct mbuf	*m_getm2(struct mbuf *, int, int, short, int);
967 struct mbuf	*m_getptr(struct mbuf *, int, int *);
968 u_int		 m_length(struct mbuf *, struct mbuf **);
969 int		 m_mbuftouio(struct uio *, struct mbuf *, int);
970 void		 m_move_pkthdr(struct mbuf *, struct mbuf *);
971 struct mbuf	*m_prepend(struct mbuf *, int, int);
972 void		 m_print(const struct mbuf *, int);
973 struct mbuf	*m_pulldown(struct mbuf *, int, int, int *);
974 struct mbuf	*m_pullup(struct mbuf *, int);
975 int		 m_sanity(struct mbuf *, int);
976 struct mbuf	*m_split(struct mbuf *, int, int);
977 struct mbuf	*m_uiotombuf(struct uio *, int, int, int, int);
978 struct mbuf	*m_unshare(struct mbuf *, int);
979 
980 /*-
981  * Network packets may have annotations attached by affixing a list of
982  * "packet tags" to the pkthdr structure.  Packet tags are dynamically
983  * allocated semi-opaque data structures that have a fixed header
984  * (struct m_tag) that specifies the size of the memory block and a
985  * <cookie,type> pair that identifies it.  The cookie is a 32-bit unique
986  * unsigned value used to identify a module or ABI.  By convention this value
987  * is chosen as the date+time that the module is created, expressed as the
988  * number of seconds since the epoch (e.g., using date -u +'%s').  The type
989  * value is an ABI/module-specific value that identifies a particular
990  * annotation and is private to the module.  For compatibility with systems
991  * like OpenBSD that define packet tags w/o an ABI/module cookie, the value
992  * PACKET_ABI_COMPAT is used to implement m_tag_get and m_tag_find
993  * compatibility shim functions and several tag types are defined below.
994  * Users that do not require compatibility should use a private cookie value
995  * so that packet tag-related definitions can be maintained privately.
996  *
997  * Note that the packet tag returned by m_tag_alloc has the default memory
998  * alignment implemented by malloc.  To reference private data one can use a
999  * construct like:
1000  *
1001  *	struct m_tag *mtag = m_tag_alloc(...);
1002  *	struct foo *p = (struct foo *)(mtag+1);
1003  *
1004  * if the alignment of struct m_tag is sufficient for referencing members of
1005  * struct foo.  Otherwise it is necessary to embed struct m_tag within the
1006  * private data structure to insure proper alignment; e.g.,
1007  *
1008  *	struct foo {
1009  *		struct m_tag	tag;
1010  *		...
1011  *	};
1012  *	struct foo *p = (struct foo *) m_tag_alloc(...);
1013  *	struct m_tag *mtag = &p->tag;
1014  */
1015 
1016 /*
1017  * Persistent tags stay with an mbuf until the mbuf is reclaimed.  Otherwise
1018  * tags are expected to ``vanish'' when they pass through a network
1019  * interface.  For most interfaces this happens normally as the tags are
1020  * reclaimed when the mbuf is free'd.  However in some special cases
1021  * reclaiming must be done manually.  An example is packets that pass through
1022  * the loopback interface.  Also, one must be careful to do this when
1023  * ``turning around'' packets (e.g., icmp_reflect).
1024  *
1025  * To mark a tag persistent bit-or this flag in when defining the tag id.
1026  * The tag will then be treated as described above.
1027  */
1028 #define	MTAG_PERSISTENT				0x800
1029 
1030 #define	PACKET_TAG_NONE				0  /* Nadda */
1031 
1032 /* Packet tags for use with PACKET_ABI_COMPAT. */
1033 #define	PACKET_TAG_IPSEC_IN_DONE		1  /* IPsec applied, in */
1034 #define	PACKET_TAG_IPSEC_OUT_DONE		2  /* IPsec applied, out */
1035 #define	PACKET_TAG_IPSEC_IN_CRYPTO_DONE		3  /* NIC IPsec crypto done */
1036 #define	PACKET_TAG_IPSEC_OUT_CRYPTO_NEEDED	4  /* NIC IPsec crypto req'ed */
1037 #define	PACKET_TAG_IPSEC_IN_COULD_DO_CRYPTO	5  /* NIC notifies IPsec */
1038 #define	PACKET_TAG_IPSEC_PENDING_TDB		6  /* Reminder to do IPsec */
1039 #define	PACKET_TAG_BRIDGE			7  /* Bridge processing done */
1040 #define	PACKET_TAG_GIF				8  /* GIF processing done */
1041 #define	PACKET_TAG_GRE				9  /* GRE processing done */
1042 #define	PACKET_TAG_IN_PACKET_CHECKSUM		10 /* NIC checksumming done */
1043 #define	PACKET_TAG_ENCAP			11 /* Encap.  processing */
1044 #define	PACKET_TAG_IPSEC_SOCKET			12 /* IPSEC socket ref */
1045 #define	PACKET_TAG_IPSEC_HISTORY		13 /* IPSEC history */
1046 #define	PACKET_TAG_IPV6_INPUT			14 /* IPV6 input processing */
1047 #define	PACKET_TAG_DUMMYNET			15 /* dummynet info */
1048 #define	PACKET_TAG_DIVERT			17 /* divert info */
1049 #define	PACKET_TAG_IPFORWARD			18 /* ipforward info */
1050 #define	PACKET_TAG_MACLABEL	(19 | MTAG_PERSISTENT) /* MAC label */
1051 #define	PACKET_TAG_PF		(21 | MTAG_PERSISTENT) /* PF/ALTQ information */
1052 #define	PACKET_TAG_RTSOCKFAM			25 /* rtsock sa family */
1053 #define	PACKET_TAG_IPOPTIONS			27 /* Saved IP options */
1054 #define	PACKET_TAG_CARP				28 /* CARP info */
1055 #define	PACKET_TAG_IPSEC_NAT_T_PORTS		29 /* two uint16_t */
1056 #define	PACKET_TAG_ND_OUTGOING			30 /* ND outgoing */
1057 
1058 /* Specific cookies and tags. */
1059 
1060 /* Packet tag routines. */
1061 struct m_tag	*m_tag_alloc(u_int32_t, int, int, int);
1062 void		 m_tag_delete(struct mbuf *, struct m_tag *);
1063 void		 m_tag_delete_chain(struct mbuf *, struct m_tag *);
1064 void		 m_tag_free_default(struct m_tag *);
1065 struct m_tag	*m_tag_locate(struct mbuf *, u_int32_t, int, struct m_tag *);
1066 struct m_tag	*m_tag_copy(struct m_tag *, int);
1067 int		 m_tag_copy_chain(struct mbuf *, struct mbuf *, int);
1068 void		 m_tag_delete_nonpersistent(struct mbuf *);
1069 
1070 /*
1071  * Initialize the list of tags associated with an mbuf.
1072  */
1073 static __inline void
1074 m_tag_init(struct mbuf *m)
1075 {
1076 
1077 	SLIST_INIT(&m->m_pkthdr.tags);
1078 }
1079 
1080 /*
1081  * Set up the contents of a tag.  Note that this does not fill in the free
1082  * method; the caller is expected to do that.
1083  *
1084  * XXX probably should be called m_tag_init, but that was already taken.
1085  */
1086 static __inline void
1087 m_tag_setup(struct m_tag *t, u_int32_t cookie, int type, int len)
1088 {
1089 
1090 	t->m_tag_id = type;
1091 	t->m_tag_len = len;
1092 	t->m_tag_cookie = cookie;
1093 }
1094 
1095 /*
1096  * Reclaim resources associated with a tag.
1097  */
1098 static __inline void
1099 m_tag_free(struct m_tag *t)
1100 {
1101 
1102 	(*t->m_tag_free)(t);
1103 }
1104 
1105 /*
1106  * Return the first tag associated with an mbuf.
1107  */
1108 static __inline struct m_tag *
1109 m_tag_first(struct mbuf *m)
1110 {
1111 
1112 	return (SLIST_FIRST(&m->m_pkthdr.tags));
1113 }
1114 
1115 /*
1116  * Return the next tag in the list of tags associated with an mbuf.
1117  */
1118 static __inline struct m_tag *
1119 m_tag_next(struct mbuf *m, struct m_tag *t)
1120 {
1121 
1122 	return (SLIST_NEXT(t, m_tag_link));
1123 }
1124 
1125 /*
1126  * Prepend a tag to the list of tags associated with an mbuf.
1127  */
1128 static __inline void
1129 m_tag_prepend(struct mbuf *m, struct m_tag *t)
1130 {
1131 
1132 	SLIST_INSERT_HEAD(&m->m_pkthdr.tags, t, m_tag_link);
1133 }
1134 
1135 /*
1136  * Unlink a tag from the list of tags associated with an mbuf.
1137  */
1138 static __inline void
1139 m_tag_unlink(struct mbuf *m, struct m_tag *t)
1140 {
1141 
1142 	SLIST_REMOVE(&m->m_pkthdr.tags, t, m_tag, m_tag_link);
1143 }
1144 
1145 /* These are for OpenBSD compatibility. */
1146 #define	MTAG_ABI_COMPAT		0		/* compatibility ABI */
1147 
1148 static __inline struct m_tag *
1149 m_tag_get(int type, int length, int wait)
1150 {
1151 	return (m_tag_alloc(MTAG_ABI_COMPAT, type, length, wait));
1152 }
1153 
1154 static __inline struct m_tag *
1155 m_tag_find(struct mbuf *m, int type, struct m_tag *start)
1156 {
1157 	return (SLIST_EMPTY(&m->m_pkthdr.tags) ? (struct m_tag *)NULL :
1158 	    m_tag_locate(m, MTAG_ABI_COMPAT, type, start));
1159 }
1160 
1161 static __inline struct mbuf *
1162 m_free(struct mbuf *m)
1163 {
1164 	struct mbuf *n = m->m_next;
1165 
1166 	if ((m->m_flags & (M_PKTHDR|M_NOFREE)) == (M_PKTHDR|M_NOFREE))
1167 		m_tag_delete_chain(m, NULL);
1168 	if (m->m_flags & M_EXT)
1169 		mb_free_ext(m);
1170 	else if ((m->m_flags & M_NOFREE) == 0)
1171 		uma_zfree(zone_mbuf, m);
1172 	return (n);
1173 }
1174 
1175 static int inline
1176 rt_m_getfib(struct mbuf *m)
1177 {
1178 	KASSERT(m->m_flags & M_PKTHDR , ("Attempt to get FIB from non header mbuf."));
1179 	return (m->m_pkthdr.fibnum);
1180 }
1181 
1182 #define M_GETFIB(_m)   rt_m_getfib(_m)
1183 
1184 #define M_SETFIB(_m, _fib) do {						\
1185         KASSERT((_m)->m_flags & M_PKTHDR, ("Attempt to set FIB on non header mbuf."));	\
1186 	((_m)->m_pkthdr.fibnum) = (_fib);				\
1187 } while (0)
1188 
1189 #endif /* _KERNEL */
1190 
1191 #ifdef MBUF_PROFILING
1192  void m_profile(struct mbuf *m);
1193  #define M_PROFILE(m) m_profile(m)
1194 #else
1195  #define M_PROFILE(m)
1196 #endif
1197 
1198 
1199 #endif /* !_SYS_MBUF_H_ */
1200