xref: /freebsd/sys/sys/mbuf.h (revision 1f4bcc459a76b7aa664f3fd557684cd0ba6da352)
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)
69 #define	MPKTHSIZE	offsetof(struct mbuf, m_pktdat)
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 
154 /*
155  * Description of external storage mapped into mbuf; valid only if M_EXT is
156  * set.
157  * Size ILP32: 28
158  *	 LP64: 48
159  * Compile-time assertions in uipc_mbuf.c test these values to ensure that
160  * they are correct.
161  */
162 struct m_ext {
163 	volatile u_int	*ext_cnt;	/* pointer to ref count info */
164 	caddr_t		 ext_buf;	/* start of buffer */
165 	uint32_t	 ext_size;	/* size of buffer, for ext_free */
166 	uint32_t	 ext_type:8,	/* type of external storage */
167 			 ext_flags:24;	/* external storage mbuf flags */
168 	void		(*ext_free)	/* free routine if not the usual */
169 			    (struct mbuf *, void *, void *);
170 	void		*ext_arg1;	/* optional argument pointer */
171 	void		*ext_arg2;	/* optional argument pointer */
172 };
173 
174 /*
175  * The core of the mbuf object along with some shortcut defines for practical
176  * purposes.
177  */
178 struct mbuf {
179 	/*
180 	 * Header present at the beginning of every mbuf.
181 	 * Size ILP32: 24
182 	 *      LP64: 32
183 	 * Compile-time assertions in uipc_mbuf.c test these values to ensure
184 	 * that they are correct.
185 	 */
186 	union {	/* next buffer in chain */
187 		struct mbuf		*m_next;
188 		SLIST_ENTRY(mbuf)	m_slist;
189 		STAILQ_ENTRY(mbuf)	m_stailq;
190 	};
191 	union {	/* next chain in queue/record */
192 		struct mbuf		*m_nextpkt;
193 		SLIST_ENTRY(mbuf)	m_slistpkt;
194 		STAILQ_ENTRY(mbuf)	m_stailqpkt;
195 	};
196 	caddr_t		 m_data;	/* location of data */
197 	int32_t		 m_len;		/* amount of data in this mbuf */
198 	uint32_t	 m_type:8,	/* type of data in this mbuf */
199 			 m_flags:24;	/* flags; see below */
200 #if !defined(__LP64__)
201 	uint32_t	 m_pad;		/* pad for 64bit alignment */
202 #endif
203 
204 	/*
205 	 * A set of optional headers (packet header, external storage header)
206 	 * and internal data storage.  Historically, these arrays were sized
207 	 * to MHLEN (space left after a packet header) and MLEN (space left
208 	 * after only a regular mbuf header); they are now variable size in
209 	 * order to support future work on variable-size mbufs.
210 	 */
211 	union {
212 		struct {
213 			struct pkthdr	m_pkthdr;	/* M_PKTHDR set */
214 			union {
215 				struct m_ext	m_ext;	/* M_EXT set */
216 				char		m_pktdat[0];
217 			};
218 		};
219 		char	m_dat[0];			/* !M_PKTHDR, !M_EXT */
220 	};
221 };
222 
223 /*
224  * mbuf flags of global significance and layer crossing.
225  * Those of only protocol/layer specific significance are to be mapped
226  * to M_PROTO[1-12] and cleared at layer handoff boundaries.
227  * NB: Limited to the lower 24 bits.
228  */
229 #define	M_EXT		0x00000001 /* has associated external storage */
230 #define	M_PKTHDR	0x00000002 /* start of record */
231 #define	M_EOR		0x00000004 /* end of record */
232 #define	M_RDONLY	0x00000008 /* associated data is marked read-only */
233 #define	M_BCAST		0x00000010 /* send/received as link-level broadcast */
234 #define	M_MCAST		0x00000020 /* send/received as link-level multicast */
235 #define	M_PROMISC	0x00000040 /* packet was not for us */
236 #define	M_VLANTAG	0x00000080 /* ether_vtag is valid */
237 #define	M_UNUSED_8	0x00000100 /* --available-- */
238 #define	M_NOFREE	0x00000200 /* do not free mbuf, embedded in cluster */
239 
240 #define	M_PROTO1	0x00001000 /* protocol-specific */
241 #define	M_PROTO2	0x00002000 /* protocol-specific */
242 #define	M_PROTO3	0x00004000 /* protocol-specific */
243 #define	M_PROTO4	0x00008000 /* protocol-specific */
244 #define	M_PROTO5	0x00010000 /* protocol-specific */
245 #define	M_PROTO6	0x00020000 /* protocol-specific */
246 #define	M_PROTO7	0x00040000 /* protocol-specific */
247 #define	M_PROTO8	0x00080000 /* protocol-specific */
248 #define	M_PROTO9	0x00100000 /* protocol-specific */
249 #define	M_PROTO10	0x00200000 /* protocol-specific */
250 #define	M_PROTO11	0x00400000 /* protocol-specific */
251 #define	M_PROTO12	0x00800000 /* protocol-specific */
252 
253 /*
254  * Flags to purge when crossing layers.
255  */
256 #define	M_PROTOFLAGS \
257     (M_PROTO1|M_PROTO2|M_PROTO3|M_PROTO4|M_PROTO5|M_PROTO6|M_PROTO7|M_PROTO8|\
258      M_PROTO9|M_PROTO10|M_PROTO11|M_PROTO12)
259 
260 /*
261  * Flags preserved when copying m_pkthdr.
262  */
263 #define M_COPYFLAGS \
264     (M_PKTHDR|M_EOR|M_RDONLY|M_BCAST|M_MCAST|M_PROMISC|M_VLANTAG| \
265      M_PROTOFLAGS)
266 
267 /*
268  * Mbuf flag description for use with printf(9) %b identifier.
269  */
270 #define	M_FLAG_BITS \
271     "\20\1M_EXT\2M_PKTHDR\3M_EOR\4M_RDONLY\5M_BCAST\6M_MCAST" \
272     "\7M_PROMISC\10M_VLANTAG"
273 #define	M_FLAG_PROTOBITS \
274     "\15M_PROTO1\16M_PROTO2\17M_PROTO3\20M_PROTO4\21M_PROTO5" \
275     "\22M_PROTO6\23M_PROTO7\24M_PROTO8\25M_PROTO9\26M_PROTO10" \
276     "\27M_PROTO11\30M_PROTO12"
277 #define	M_FLAG_PRINTF (M_FLAG_BITS M_FLAG_PROTOBITS)
278 
279 /*
280  * Network interface cards are able to hash protocol fields (such as IPv4
281  * addresses and TCP port numbers) classify packets into flows.  These flows
282  * can then be used to maintain ordering while delivering packets to the OS
283  * via parallel input queues, as well as to provide a stateless affinity
284  * model.  NIC drivers can pass up the hash via m->m_pkthdr.flowid, and set
285  * m_flag fields to indicate how the hash should be interpreted by the
286  * network stack.
287  *
288  * Most NICs support RSS, which provides ordering and explicit affinity, and
289  * use the hash m_flag bits to indicate what header fields were covered by
290  * the hash.  M_HASHTYPE_OPAQUE can be set by non-RSS cards or configurations
291  * that provide an opaque flow identifier, allowing for ordering and
292  * distribution without explicit affinity.
293  */
294 /* Microsoft RSS standard hash types */
295 #define	M_HASHTYPE_NONE			0
296 #define	M_HASHTYPE_RSS_IPV4		1	/* IPv4 2-tuple */
297 #define	M_HASHTYPE_RSS_TCP_IPV4		2	/* TCPv4 4-tuple */
298 #define	M_HASHTYPE_RSS_IPV6		3	/* IPv6 2-tuple */
299 #define	M_HASHTYPE_RSS_TCP_IPV6		4	/* TCPv6 4-tuple */
300 #define	M_HASHTYPE_RSS_IPV6_EX		5	/* IPv6 2-tuple + ext hdrs */
301 #define	M_HASHTYPE_RSS_TCP_IPV6_EX	6	/* TCPv6 4-tiple + ext hdrs */
302 /* Non-standard RSS hash types */
303 #define	M_HASHTYPE_RSS_UDP_IPV4		7	/* IPv4 UDP 4-tuple */
304 #define	M_HASHTYPE_RSS_UDP_IPV4_EX	8	/* IPv4 UDP 4-tuple + ext hdrs */
305 #define	M_HASHTYPE_RSS_UDP_IPV6		9	/* IPv6 UDP 4-tuple */
306 #define	M_HASHTYPE_RSS_UDP_IPV6_EX	10	/* IPv6 UDP 4-tuple + ext hdrs */
307 
308 #define	M_HASHTYPE_OPAQUE		255	/* ordering, not affinity */
309 
310 #define	M_HASHTYPE_CLEAR(m)	((m)->m_pkthdr.rsstype = 0)
311 #define	M_HASHTYPE_GET(m)	((m)->m_pkthdr.rsstype)
312 #define	M_HASHTYPE_SET(m, v)	((m)->m_pkthdr.rsstype = (v))
313 #define	M_HASHTYPE_TEST(m, v)	(M_HASHTYPE_GET(m) == (v))
314 
315 /*
316  * COS/QOS class and quality of service tags.
317  * It uses DSCP code points as base.
318  */
319 #define	QOS_DSCP_CS0		0x00
320 #define	QOS_DSCP_DEF		QOS_DSCP_CS0
321 #define	QOS_DSCP_CS1		0x20
322 #define	QOS_DSCP_AF11		0x28
323 #define	QOS_DSCP_AF12		0x30
324 #define	QOS_DSCP_AF13		0x38
325 #define	QOS_DSCP_CS2		0x40
326 #define	QOS_DSCP_AF21		0x48
327 #define	QOS_DSCP_AF22		0x50
328 #define	QOS_DSCP_AF23		0x58
329 #define	QOS_DSCP_CS3		0x60
330 #define	QOS_DSCP_AF31		0x68
331 #define	QOS_DSCP_AF32		0x70
332 #define	QOS_DSCP_AF33		0x78
333 #define	QOS_DSCP_CS4		0x80
334 #define	QOS_DSCP_AF41		0x88
335 #define	QOS_DSCP_AF42		0x90
336 #define	QOS_DSCP_AF43		0x98
337 #define	QOS_DSCP_CS5		0xa0
338 #define	QOS_DSCP_EF		0xb8
339 #define	QOS_DSCP_CS6		0xc0
340 #define	QOS_DSCP_CS7		0xe0
341 
342 /*
343  * External mbuf storage buffer types.
344  */
345 #define	EXT_CLUSTER	1	/* mbuf cluster */
346 #define	EXT_SFBUF	2	/* sendfile(2)'s sf_buf */
347 #define	EXT_JUMBOP	3	/* jumbo cluster page sized */
348 #define	EXT_JUMBO9	4	/* jumbo cluster 9216 bytes */
349 #define	EXT_JUMBO16	5	/* jumbo cluster 16184 bytes */
350 #define	EXT_PACKET	6	/* mbuf+cluster from packet zone */
351 #define	EXT_MBUF	7	/* external mbuf reference (M_IOVEC) */
352 #define	EXT_SFBUF_NOCACHE 8	/* sendfile(2)'s sf_buf not to be cached */
353 
354 #define	EXT_VENDOR1	224	/* for vendor-internal use */
355 #define	EXT_VENDOR2	225	/* for vendor-internal use */
356 #define	EXT_VENDOR3	226	/* for vendor-internal use */
357 #define	EXT_VENDOR4	227	/* for vendor-internal use */
358 
359 #define	EXT_EXP1	244	/* for experimental use */
360 #define	EXT_EXP2	245	/* for experimental use */
361 #define	EXT_EXP3	246	/* for experimental use */
362 #define	EXT_EXP4	247	/* for experimental use */
363 
364 #define	EXT_NET_DRV	252	/* custom ext_buf provided by net driver(s) */
365 #define	EXT_MOD_TYPE	253	/* custom module's ext_buf type */
366 #define	EXT_DISPOSABLE	254	/* can throw this buffer away w/page flipping */
367 #define	EXT_EXTREF	255	/* has externally maintained ext_cnt ptr */
368 
369 /*
370  * Flags for external mbuf buffer types.
371  * NB: limited to the lower 24 bits.
372  */
373 #define	EXT_FLAG_EMBREF		0x000001	/* embedded ext_cnt, notyet */
374 #define	EXT_FLAG_EXTREF		0x000002	/* external ext_cnt, notyet */
375 #define	EXT_FLAG_NOFREE		0x000010	/* don't free mbuf to pool, notyet */
376 
377 #define	EXT_FLAG_VENDOR1	0x010000	/* for vendor-internal use */
378 #define	EXT_FLAG_VENDOR2	0x020000	/* for vendor-internal use */
379 #define	EXT_FLAG_VENDOR3	0x040000	/* for vendor-internal use */
380 #define	EXT_FLAG_VENDOR4	0x080000	/* for vendor-internal use */
381 
382 #define	EXT_FLAG_EXP1		0x100000	/* for experimental use */
383 #define	EXT_FLAG_EXP2		0x200000	/* for experimental use */
384 #define	EXT_FLAG_EXP3		0x400000	/* for experimental use */
385 #define	EXT_FLAG_EXP4		0x800000	/* for experimental use */
386 
387 /*
388  * EXT flag description for use with printf(9) %b identifier.
389  */
390 #define	EXT_FLAG_BITS \
391     "\20\1EXT_FLAG_EMBREF\2EXT_FLAG_EXTREF\5EXT_FLAG_NOFREE" \
392     "\21EXT_FLAG_VENDOR1\22EXT_FLAG_VENDOR2\23EXT_FLAG_VENDOR3" \
393     "\24EXT_FLAG_VENDOR4\25EXT_FLAG_EXP1\26EXT_FLAG_EXP2\27EXT_FLAG_EXP3" \
394     "\30EXT_FLAG_EXP4"
395 
396 /*
397  * External reference/free functions.
398  */
399 void sf_ext_ref(void *, void *);
400 void sf_ext_free(void *, void *);
401 void sf_ext_free_nocache(void *, void *);
402 
403 /*
404  * Flags indicating checksum, segmentation and other offload work to be
405  * done, or already done, by hardware or lower layers.  It is split into
406  * separate inbound and outbound flags.
407  *
408  * Outbound flags that are set by upper protocol layers requesting lower
409  * layers, or ideally the hardware, to perform these offloading tasks.
410  * For outbound packets this field and its flags can be directly tested
411  * against ifnet if_hwassist.
412  */
413 #define	CSUM_IP			0x00000001	/* IP header checksum offload */
414 #define	CSUM_IP_UDP		0x00000002	/* UDP checksum offload */
415 #define	CSUM_IP_TCP		0x00000004	/* TCP checksum offload */
416 #define	CSUM_IP_SCTP		0x00000008	/* SCTP checksum offload */
417 #define	CSUM_IP_TSO		0x00000010	/* TCP segmentation offload */
418 #define	CSUM_IP_ISCSI		0x00000020	/* iSCSI checksum offload */
419 
420 #define	CSUM_IP6_UDP		0x00000200	/* UDP checksum offload */
421 #define	CSUM_IP6_TCP		0x00000400	/* TCP checksum offload */
422 #define	CSUM_IP6_SCTP		0x00000800	/* SCTP checksum offload */
423 #define	CSUM_IP6_TSO		0x00001000	/* TCP segmentation offload */
424 #define	CSUM_IP6_ISCSI		0x00002000	/* iSCSI checksum offload */
425 
426 /* Inbound checksum support where the checksum was verified by hardware. */
427 #define	CSUM_L3_CALC		0x01000000	/* calculated layer 3 csum */
428 #define	CSUM_L3_VALID		0x02000000	/* checksum is correct */
429 #define	CSUM_L4_CALC		0x04000000	/* calculated layer 4 csum */
430 #define	CSUM_L4_VALID		0x08000000	/* checksum is correct */
431 #define	CSUM_L5_CALC		0x10000000	/* calculated layer 5 csum */
432 #define	CSUM_L5_VALID		0x20000000	/* checksum is correct */
433 #define	CSUM_COALESED		0x40000000	/* contains merged segments */
434 
435 /*
436  * CSUM flag description for use with printf(9) %b identifier.
437  */
438 #define	CSUM_BITS \
439     "\20\1CSUM_IP\2CSUM_IP_UDP\3CSUM_IP_TCP\4CSUM_IP_SCTP\5CSUM_IP_TSO" \
440     "\6CSUM_IP_ISCSI" \
441     "\12CSUM_IP6_UDP\13CSUM_IP6_TCP\14CSUM_IP6_SCTP\15CSUM_IP6_TSO" \
442     "\16CSUM_IP6_ISCSI" \
443     "\31CSUM_L3_CALC\32CSUM_L3_VALID\33CSUM_L4_CALC\34CSUM_L4_VALID" \
444     "\35CSUM_L5_CALC\36CSUM_L5_VALID\37CSUM_COALESED"
445 
446 /* CSUM flags compatibility mappings. */
447 #define	CSUM_IP_CHECKED		CSUM_L3_CALC
448 #define	CSUM_IP_VALID		CSUM_L3_VALID
449 #define	CSUM_DATA_VALID		CSUM_L4_VALID
450 #define	CSUM_PSEUDO_HDR		CSUM_L4_CALC
451 #define	CSUM_SCTP_VALID		CSUM_L4_VALID
452 #define	CSUM_DELAY_DATA		(CSUM_TCP|CSUM_UDP)
453 #define	CSUM_DELAY_IP		CSUM_IP		/* Only v4, no v6 IP hdr csum */
454 #define	CSUM_DELAY_DATA_IPV6	(CSUM_TCP_IPV6|CSUM_UDP_IPV6)
455 #define	CSUM_DATA_VALID_IPV6	CSUM_DATA_VALID
456 #define	CSUM_TCP		CSUM_IP_TCP
457 #define	CSUM_UDP		CSUM_IP_UDP
458 #define	CSUM_SCTP		CSUM_IP_SCTP
459 #define	CSUM_TSO		(CSUM_IP_TSO|CSUM_IP6_TSO)
460 #define	CSUM_UDP_IPV6		CSUM_IP6_UDP
461 #define	CSUM_TCP_IPV6		CSUM_IP6_TCP
462 #define	CSUM_SCTP_IPV6		CSUM_IP6_SCTP
463 
464 /*
465  * mbuf types describing the content of the mbuf (including external storage).
466  */
467 #define	MT_NOTMBUF	0	/* USED INTERNALLY ONLY! Object is not mbuf */
468 #define	MT_DATA		1	/* dynamic (data) allocation */
469 #define	MT_HEADER	MT_DATA	/* packet header, use M_PKTHDR instead */
470 
471 #define	MT_VENDOR1	4	/* for vendor-internal use */
472 #define	MT_VENDOR2	5	/* for vendor-internal use */
473 #define	MT_VENDOR3	6	/* for vendor-internal use */
474 #define	MT_VENDOR4	7	/* for vendor-internal use */
475 
476 #define	MT_SONAME	8	/* socket name */
477 
478 #define	MT_EXP1		9	/* for experimental use */
479 #define	MT_EXP2		10	/* for experimental use */
480 #define	MT_EXP3		11	/* for experimental use */
481 #define	MT_EXP4		12	/* for experimental use */
482 
483 #define	MT_CONTROL	14	/* extra-data protocol message */
484 #define	MT_OOBDATA	15	/* expedited data  */
485 #define	MT_NTYPES	16	/* number of mbuf types for mbtypes[] */
486 
487 #define	MT_NOINIT	255	/* Not a type but a flag to allocate
488 				   a non-initialized mbuf */
489 
490 /*
491  * String names of mbuf-related UMA(9) and malloc(9) types.  Exposed to
492  * !_KERNEL so that monitoring tools can look up the zones with
493  * libmemstat(3).
494  */
495 #define	MBUF_MEM_NAME		"mbuf"
496 #define	MBUF_CLUSTER_MEM_NAME	"mbuf_cluster"
497 #define	MBUF_PACKET_MEM_NAME	"mbuf_packet"
498 #define	MBUF_JUMBOP_MEM_NAME	"mbuf_jumbo_page"
499 #define	MBUF_JUMBO9_MEM_NAME	"mbuf_jumbo_9k"
500 #define	MBUF_JUMBO16_MEM_NAME	"mbuf_jumbo_16k"
501 #define	MBUF_TAG_MEM_NAME	"mbuf_tag"
502 #define	MBUF_EXTREFCNT_MEM_NAME	"mbuf_ext_refcnt"
503 
504 #ifdef _KERNEL
505 
506 #ifdef WITNESS
507 #define	MBUF_CHECKSLEEP(how) do {					\
508 	if (how == M_WAITOK)						\
509 		WITNESS_WARN(WARN_GIANTOK | WARN_SLEEPOK, NULL,		\
510 		    "Sleeping in \"%s\"", __func__);			\
511 } while (0)
512 #else
513 #define	MBUF_CHECKSLEEP(how)
514 #endif
515 
516 /*
517  * Network buffer allocation API
518  *
519  * The rest of it is defined in kern/kern_mbuf.c
520  */
521 extern uma_zone_t	zone_mbuf;
522 extern uma_zone_t	zone_clust;
523 extern uma_zone_t	zone_pack;
524 extern uma_zone_t	zone_jumbop;
525 extern uma_zone_t	zone_jumbo9;
526 extern uma_zone_t	zone_jumbo16;
527 extern uma_zone_t	zone_ext_refcnt;
528 
529 void		 mb_dupcl(struct mbuf *, const struct mbuf *);
530 void		 mb_free_ext(struct mbuf *);
531 int		 m_pkthdr_init(struct mbuf *, int);
532 
533 static __inline int
534 m_gettype(int size)
535 {
536 	int type;
537 
538 	switch (size) {
539 	case MSIZE:
540 		type = EXT_MBUF;
541 		break;
542 	case MCLBYTES:
543 		type = EXT_CLUSTER;
544 		break;
545 #if MJUMPAGESIZE != MCLBYTES
546 	case MJUMPAGESIZE:
547 		type = EXT_JUMBOP;
548 		break;
549 #endif
550 	case MJUM9BYTES:
551 		type = EXT_JUMBO9;
552 		break;
553 	case MJUM16BYTES:
554 		type = EXT_JUMBO16;
555 		break;
556 	default:
557 		panic("%s: invalid cluster size %d", __func__, size);
558 	}
559 
560 	return (type);
561 }
562 
563 /*
564  * Associated an external reference counted buffer with an mbuf.
565  */
566 static __inline void
567 m_extaddref(struct mbuf *m, caddr_t buf, u_int size, u_int *ref_cnt,
568     void (*freef)(struct mbuf *, void *, void *), void *arg1, void *arg2)
569 {
570 
571 	KASSERT(ref_cnt != NULL, ("%s: ref_cnt not provided", __func__));
572 
573 	atomic_add_int(ref_cnt, 1);
574 	m->m_flags |= M_EXT;
575 	m->m_ext.ext_buf = buf;
576 	m->m_ext.ext_cnt = ref_cnt;
577 	m->m_data = m->m_ext.ext_buf;
578 	m->m_ext.ext_size = size;
579 	m->m_ext.ext_free = freef;
580 	m->m_ext.ext_arg1 = arg1;
581 	m->m_ext.ext_arg2 = arg2;
582 	m->m_ext.ext_type = EXT_EXTREF;
583 	m->m_ext.ext_flags = 0;
584 }
585 
586 static __inline uma_zone_t
587 m_getzone(int size)
588 {
589 	uma_zone_t zone;
590 
591 	switch (size) {
592 	case MCLBYTES:
593 		zone = zone_clust;
594 		break;
595 #if MJUMPAGESIZE != MCLBYTES
596 	case MJUMPAGESIZE:
597 		zone = zone_jumbop;
598 		break;
599 #endif
600 	case MJUM9BYTES:
601 		zone = zone_jumbo9;
602 		break;
603 	case MJUM16BYTES:
604 		zone = zone_jumbo16;
605 		break;
606 	default:
607 		panic("%s: invalid cluster size %d", __func__, size);
608 	}
609 
610 	return (zone);
611 }
612 
613 /*
614  * Initialize an mbuf with linear storage.
615  *
616  * Inline because the consumer text overhead will be roughly the same to
617  * initialize or call a function with this many parameters and M_PKTHDR
618  * should go away with constant propagation for !MGETHDR.
619  */
620 static __inline int
621 m_init(struct mbuf *m, uma_zone_t zone __unused, int size __unused, int how,
622     short type, int flags)
623 {
624 	int error;
625 
626 	m->m_next = NULL;
627 	m->m_nextpkt = NULL;
628 	m->m_data = m->m_dat;
629 	m->m_len = 0;
630 	m->m_flags = flags;
631 	m->m_type = type;
632 	if (flags & M_PKTHDR) {
633 		if ((error = m_pkthdr_init(m, how)) != 0)
634 			return (error);
635 	}
636 
637 	return (0);
638 }
639 
640 static __inline struct mbuf *
641 m_get(int how, short type)
642 {
643 	struct mb_args args;
644 
645 	args.flags = 0;
646 	args.type = type;
647 	return (uma_zalloc_arg(zone_mbuf, &args, how));
648 }
649 
650 /*
651  * XXX This should be deprecated, very little use.
652  */
653 static __inline struct mbuf *
654 m_getclr(int how, short type)
655 {
656 	struct mbuf *m;
657 	struct mb_args args;
658 
659 	args.flags = 0;
660 	args.type = type;
661 	m = uma_zalloc_arg(zone_mbuf, &args, how);
662 	if (m != NULL)
663 		bzero(m->m_data, MLEN);
664 	return (m);
665 }
666 
667 static __inline struct mbuf *
668 m_gethdr(int how, short type)
669 {
670 	struct mb_args args;
671 
672 	args.flags = M_PKTHDR;
673 	args.type = type;
674 	return (uma_zalloc_arg(zone_mbuf, &args, how));
675 }
676 
677 static __inline struct mbuf *
678 m_getcl(int how, short type, int flags)
679 {
680 	struct mb_args args;
681 
682 	args.flags = flags;
683 	args.type = type;
684 	return (uma_zalloc_arg(zone_pack, &args, how));
685 }
686 
687 static __inline int
688 m_clget(struct mbuf *m, int how)
689 {
690 
691 	KASSERT((m->m_flags & M_EXT) == 0, ("%s: mbuf %p has M_EXT",
692 	    __func__, m));
693 	m->m_ext.ext_buf = (char *)NULL;
694 	uma_zalloc_arg(zone_clust, m, how);
695 	/*
696 	 * On a cluster allocation failure, drain the packet zone and retry,
697 	 * we might be able to loosen a few clusters up on the drain.
698 	 */
699 	if ((how & M_NOWAIT) && (m->m_ext.ext_buf == NULL)) {
700 		zone_drain(zone_pack);
701 		uma_zalloc_arg(zone_clust, m, how);
702 	}
703 	return (m->m_flags & M_EXT);
704 }
705 
706 /*
707  * m_cljget() is different from m_clget() as it can allocate clusters without
708  * attaching them to an mbuf.  In that case the return value is the pointer
709  * to the cluster of the requested size.  If an mbuf was specified, it gets
710  * the cluster attached to it and the return value can be safely ignored.
711  * For size it takes MCLBYTES, MJUMPAGESIZE, MJUM9BYTES, MJUM16BYTES.
712  */
713 static __inline void *
714 m_cljget(struct mbuf *m, int how, int size)
715 {
716 	uma_zone_t zone;
717 
718 	if (m != NULL) {
719 		KASSERT((m->m_flags & M_EXT) == 0, ("%s: mbuf %p has M_EXT",
720 		    __func__, m));
721 		m->m_ext.ext_buf = NULL;
722 	}
723 
724 	zone = m_getzone(size);
725 	return (uma_zalloc_arg(zone, m, how));
726 }
727 
728 static __inline void
729 m_cljset(struct mbuf *m, void *cl, int type)
730 {
731 	uma_zone_t zone;
732 	int size;
733 
734 	switch (type) {
735 	case EXT_CLUSTER:
736 		size = MCLBYTES;
737 		zone = zone_clust;
738 		break;
739 #if MJUMPAGESIZE != MCLBYTES
740 	case EXT_JUMBOP:
741 		size = MJUMPAGESIZE;
742 		zone = zone_jumbop;
743 		break;
744 #endif
745 	case EXT_JUMBO9:
746 		size = MJUM9BYTES;
747 		zone = zone_jumbo9;
748 		break;
749 	case EXT_JUMBO16:
750 		size = MJUM16BYTES;
751 		zone = zone_jumbo16;
752 		break;
753 	default:
754 		panic("%s: unknown cluster type %d", __func__, type);
755 		break;
756 	}
757 
758 	m->m_data = m->m_ext.ext_buf = cl;
759 	m->m_ext.ext_free = m->m_ext.ext_arg1 = m->m_ext.ext_arg2 = NULL;
760 	m->m_ext.ext_size = size;
761 	m->m_ext.ext_type = type;
762 	m->m_ext.ext_flags = 0;
763 	m->m_ext.ext_cnt = uma_find_refcnt(zone, cl);
764 	m->m_flags |= M_EXT;
765 
766 }
767 
768 static __inline void
769 m_chtype(struct mbuf *m, short new_type)
770 {
771 
772 	m->m_type = new_type;
773 }
774 
775 static __inline void
776 m_clrprotoflags(struct mbuf *m)
777 {
778 
779 	while (m) {
780 		m->m_flags &= ~M_PROTOFLAGS;
781 		m = m->m_next;
782 	}
783 }
784 
785 static __inline struct mbuf *
786 m_last(struct mbuf *m)
787 {
788 
789 	while (m->m_next)
790 		m = m->m_next;
791 	return (m);
792 }
793 
794 /*
795  * mbuf, cluster, and external object allocation macros (for compatibility
796  * purposes).
797  */
798 #define	M_MOVE_PKTHDR(to, from)	m_move_pkthdr((to), (from))
799 #define	MGET(m, how, type)	((m) = m_get((how), (type)))
800 #define	MGETHDR(m, how, type)	((m) = m_gethdr((how), (type)))
801 #define	MCLGET(m, how)		m_clget((m), (how))
802 #define	MEXTADD(m, buf, size, free, arg1, arg2, flags, type)		\
803     (void )m_extadd((m), (caddr_t)(buf), (size), (free), (arg1), (arg2),\
804     (flags), (type), M_NOWAIT)
805 #define	m_getm(m, len, how, type)					\
806     m_getm2((m), (len), (how), (type), M_PKTHDR)
807 
808 /*
809  * Evaluate TRUE if it's safe to write to the mbuf m's data region (this can
810  * be both the local data payload, or an external buffer area, depending on
811  * whether M_EXT is set).
812  */
813 #define	M_WRITABLE(m)	(!((m)->m_flags & M_RDONLY) &&			\
814 			 (!(((m)->m_flags & M_EXT)) ||			\
815 			 (*((m)->m_ext.ext_cnt) == 1)) )		\
816 
817 /* Check if the supplied mbuf has a packet header, or else panic. */
818 #define	M_ASSERTPKTHDR(m)						\
819 	KASSERT((m) != NULL && (m)->m_flags & M_PKTHDR,			\
820 	    ("%s: no mbuf packet header!", __func__))
821 
822 /*
823  * Ensure that the supplied mbuf is a valid, non-free mbuf.
824  *
825  * XXX: Broken at the moment.  Need some UMA magic to make it work again.
826  */
827 #define	M_ASSERTVALID(m)						\
828 	KASSERT((((struct mbuf *)m)->m_flags & 0) == 0,			\
829 	    ("%s: attempted use of a free mbuf!", __func__))
830 
831 /*
832  * Return the address of the start of the buffer associated with an mbuf,
833  * handling external storage, packet-header mbufs, and regular data mbufs.
834  */
835 #define	M_START(m)							\
836 	(((m)->m_flags & M_EXT) ? (m)->m_ext.ext_buf :			\
837 	 ((m)->m_flags & M_PKTHDR) ? &(m)->m_pktdat[0] :		\
838 	 &(m)->m_dat[0])
839 
840 /*
841  * Return the size of the buffer associated with an mbuf, handling external
842  * storage, packet-header mbufs, and regular data mbufs.
843  */
844 #define	M_SIZE(m)							\
845 	(((m)->m_flags & M_EXT) ? (m)->m_ext.ext_size :			\
846 	 ((m)->m_flags & M_PKTHDR) ? MHLEN :				\
847 	 MLEN)
848 
849 /*
850  * Set the m_data pointer of a newly allocated mbuf to place an object of the
851  * specified size at the end of the mbuf, longword aligned.
852  *
853  * NB: Historically, we had M_ALIGN(), MH_ALIGN(), and MEXT_ALIGN() as
854  * separate macros, each asserting that it was called at the proper moment.
855  * This required callers to themselves test the storage type and call the
856  * right one.  Rather than require callers to be aware of those layout
857  * decisions, we centralize here.
858  */
859 static __inline void
860 m_align(struct mbuf *m, int len)
861 {
862 #ifdef INVARIANTS
863 	const char *msg = "%s: not a virgin mbuf";
864 #endif
865 	int adjust;
866 
867 	KASSERT(m->m_data == M_START(m), (msg, __func__));
868 
869 	adjust = M_SIZE(m) - len;
870 	m->m_data += adjust &~ (sizeof(long)-1);
871 }
872 
873 #define	M_ALIGN(m, len)		m_align(m, len)
874 #define	MH_ALIGN(m, len)	m_align(m, len)
875 #define	MEXT_ALIGN(m, len)	m_align(m, len)
876 
877 /*
878  * Compute the amount of space available before the current start of data in
879  * an mbuf.
880  *
881  * The M_WRITABLE() is a temporary, conservative safety measure: the burden
882  * of checking writability of the mbuf data area rests solely with the caller.
883  *
884  * NB: In previous versions, M_LEADINGSPACE() would only check M_WRITABLE()
885  * for mbufs with external storage.  We now allow mbuf-embedded data to be
886  * read-only as well.
887  */
888 #define	M_LEADINGSPACE(m)						\
889 	(M_WRITABLE(m) ? ((m)->m_data - M_START(m)) : 0)
890 
891 /*
892  * Compute the amount of space available after the end of data in an mbuf.
893  *
894  * The M_WRITABLE() is a temporary, conservative safety measure: the burden
895  * of checking writability of the mbuf data area rests solely with the caller.
896  *
897  * NB: In previous versions, M_TRAILINGSPACE() would only check M_WRITABLE()
898  * for mbufs with external storage.  We now allow mbuf-embedded data to be
899  * read-only as well.
900  */
901 #define	M_TRAILINGSPACE(m)						\
902 	(M_WRITABLE(m) ?						\
903 	    ((M_START(m) + M_SIZE(m)) - ((m)->m_data + (m)->m_len)) : 0)
904 
905 /*
906  * Arrange to prepend space of size plen to mbuf m.  If a new mbuf must be
907  * allocated, how specifies whether to wait.  If the allocation fails, the
908  * original mbuf chain is freed and m is set to NULL.
909  */
910 #define	M_PREPEND(m, plen, how) do {					\
911 	struct mbuf **_mmp = &(m);					\
912 	struct mbuf *_mm = *_mmp;					\
913 	int _mplen = (plen);						\
914 	int __mhow = (how);						\
915 									\
916 	MBUF_CHECKSLEEP(how);						\
917 	if (M_LEADINGSPACE(_mm) >= _mplen) {				\
918 		_mm->m_data -= _mplen;					\
919 		_mm->m_len += _mplen;					\
920 	} else								\
921 		_mm = m_prepend(_mm, _mplen, __mhow);			\
922 	if (_mm != NULL && _mm->m_flags & M_PKTHDR)			\
923 		_mm->m_pkthdr.len += _mplen;				\
924 	*_mmp = _mm;							\
925 } while (0)
926 
927 /*
928  * Change mbuf to new type.  This is a relatively expensive operation and
929  * should be avoided.
930  */
931 #define	MCHTYPE(m, t)	m_chtype((m), (t))
932 
933 /* Length to m_copy to copy all. */
934 #define	M_COPYALL	1000000000
935 
936 /* Compatibility with 4.3. */
937 #define	m_copy(m, o, l)	m_copym((m), (o), (l), M_NOWAIT)
938 
939 extern int		max_datalen;	/* MHLEN - max_hdr */
940 extern int		max_hdr;	/* Largest link + protocol header */
941 extern int		max_linkhdr;	/* Largest link-level header */
942 extern int		max_protohdr;	/* Largest protocol header */
943 extern int		nmbclusters;	/* Maximum number of clusters */
944 
945 struct uio;
946 
947 void		 m_adj(struct mbuf *, int);
948 int		 m_apply(struct mbuf *, int, int,
949 		    int (*)(void *, void *, u_int), void *);
950 int		 m_append(struct mbuf *, int, c_caddr_t);
951 void		 m_cat(struct mbuf *, struct mbuf *);
952 void		 m_catpkt(struct mbuf *, struct mbuf *);
953 int		 m_extadd(struct mbuf *, caddr_t, u_int,
954 		    void (*)(struct mbuf *, void *, void *), void *, void *,
955 		    int, int, int);
956 struct mbuf	*m_collapse(struct mbuf *, int, int);
957 void		 m_copyback(struct mbuf *, int, int, c_caddr_t);
958 void		 m_copydata(const struct mbuf *, int, int, caddr_t);
959 struct mbuf	*m_copym(const struct mbuf *, int, int, int);
960 struct mbuf	*m_copypacket(struct mbuf *, int);
961 void		 m_copy_pkthdr(struct mbuf *, struct mbuf *);
962 struct mbuf	*m_copyup(struct mbuf *, int, int);
963 struct mbuf	*m_defrag(struct mbuf *, int);
964 void		 m_demote_pkthdr(struct mbuf *);
965 void		 m_demote(struct mbuf *, int, int);
966 struct mbuf	*m_devget(char *, int, int, struct ifnet *,
967 		    void (*)(char *, caddr_t, u_int));
968 struct mbuf	*m_dup(const struct mbuf *, int);
969 int		 m_dup_pkthdr(struct mbuf *, const struct mbuf *, int);
970 u_int		 m_fixhdr(struct mbuf *);
971 struct mbuf	*m_fragment(struct mbuf *, int, int);
972 void		 m_freem(struct mbuf *);
973 struct mbuf	*m_get2(int, int, short, int);
974 struct mbuf	*m_getjcl(int, short, int, int);
975 struct mbuf	*m_getm2(struct mbuf *, int, int, short, int);
976 struct mbuf	*m_getptr(struct mbuf *, int, int *);
977 u_int		 m_length(struct mbuf *, struct mbuf **);
978 int		 m_mbuftouio(struct uio *, struct mbuf *, int);
979 void		 m_move_pkthdr(struct mbuf *, struct mbuf *);
980 struct mbuf	*m_prepend(struct mbuf *, int, int);
981 void		 m_print(const struct mbuf *, int);
982 struct mbuf	*m_pulldown(struct mbuf *, int, int, int *);
983 struct mbuf	*m_pullup(struct mbuf *, int);
984 int		 m_sanity(struct mbuf *, int);
985 struct mbuf	*m_split(struct mbuf *, int, int);
986 struct mbuf	*m_uiotombuf(struct uio *, int, int, int, int);
987 struct mbuf	*m_unshare(struct mbuf *, int);
988 
989 /*-
990  * Network packets may have annotations attached by affixing a list of
991  * "packet tags" to the pkthdr structure.  Packet tags are dynamically
992  * allocated semi-opaque data structures that have a fixed header
993  * (struct m_tag) that specifies the size of the memory block and a
994  * <cookie,type> pair that identifies it.  The cookie is a 32-bit unique
995  * unsigned value used to identify a module or ABI.  By convention this value
996  * is chosen as the date+time that the module is created, expressed as the
997  * number of seconds since the epoch (e.g., using date -u +'%s').  The type
998  * value is an ABI/module-specific value that identifies a particular
999  * annotation and is private to the module.  For compatibility with systems
1000  * like OpenBSD that define packet tags w/o an ABI/module cookie, the value
1001  * PACKET_ABI_COMPAT is used to implement m_tag_get and m_tag_find
1002  * compatibility shim functions and several tag types are defined below.
1003  * Users that do not require compatibility should use a private cookie value
1004  * so that packet tag-related definitions can be maintained privately.
1005  *
1006  * Note that the packet tag returned by m_tag_alloc has the default memory
1007  * alignment implemented by malloc.  To reference private data one can use a
1008  * construct like:
1009  *
1010  *	struct m_tag *mtag = m_tag_alloc(...);
1011  *	struct foo *p = (struct foo *)(mtag+1);
1012  *
1013  * if the alignment of struct m_tag is sufficient for referencing members of
1014  * struct foo.  Otherwise it is necessary to embed struct m_tag within the
1015  * private data structure to insure proper alignment; e.g.,
1016  *
1017  *	struct foo {
1018  *		struct m_tag	tag;
1019  *		...
1020  *	};
1021  *	struct foo *p = (struct foo *) m_tag_alloc(...);
1022  *	struct m_tag *mtag = &p->tag;
1023  */
1024 
1025 /*
1026  * Persistent tags stay with an mbuf until the mbuf is reclaimed.  Otherwise
1027  * tags are expected to ``vanish'' when they pass through a network
1028  * interface.  For most interfaces this happens normally as the tags are
1029  * reclaimed when the mbuf is free'd.  However in some special cases
1030  * reclaiming must be done manually.  An example is packets that pass through
1031  * the loopback interface.  Also, one must be careful to do this when
1032  * ``turning around'' packets (e.g., icmp_reflect).
1033  *
1034  * To mark a tag persistent bit-or this flag in when defining the tag id.
1035  * The tag will then be treated as described above.
1036  */
1037 #define	MTAG_PERSISTENT				0x800
1038 
1039 #define	PACKET_TAG_NONE				0  /* Nadda */
1040 
1041 /* Packet tags for use with PACKET_ABI_COMPAT. */
1042 #define	PACKET_TAG_IPSEC_IN_DONE		1  /* IPsec applied, in */
1043 #define	PACKET_TAG_IPSEC_OUT_DONE		2  /* IPsec applied, out */
1044 #define	PACKET_TAG_IPSEC_IN_CRYPTO_DONE		3  /* NIC IPsec crypto done */
1045 #define	PACKET_TAG_IPSEC_OUT_CRYPTO_NEEDED	4  /* NIC IPsec crypto req'ed */
1046 #define	PACKET_TAG_IPSEC_IN_COULD_DO_CRYPTO	5  /* NIC notifies IPsec */
1047 #define	PACKET_TAG_IPSEC_PENDING_TDB		6  /* Reminder to do IPsec */
1048 #define	PACKET_TAG_BRIDGE			7  /* Bridge processing done */
1049 #define	PACKET_TAG_GIF				8  /* GIF processing done */
1050 #define	PACKET_TAG_GRE				9  /* GRE processing done */
1051 #define	PACKET_TAG_IN_PACKET_CHECKSUM		10 /* NIC checksumming done */
1052 #define	PACKET_TAG_ENCAP			11 /* Encap.  processing */
1053 #define	PACKET_TAG_IPSEC_SOCKET			12 /* IPSEC socket ref */
1054 #define	PACKET_TAG_IPSEC_HISTORY		13 /* IPSEC history */
1055 #define	PACKET_TAG_IPV6_INPUT			14 /* IPV6 input processing */
1056 #define	PACKET_TAG_DUMMYNET			15 /* dummynet info */
1057 #define	PACKET_TAG_DIVERT			17 /* divert info */
1058 #define	PACKET_TAG_IPFORWARD			18 /* ipforward info */
1059 #define	PACKET_TAG_MACLABEL	(19 | MTAG_PERSISTENT) /* MAC label */
1060 #define	PACKET_TAG_PF		(21 | MTAG_PERSISTENT) /* PF/ALTQ information */
1061 #define	PACKET_TAG_RTSOCKFAM			25 /* rtsock sa family */
1062 #define	PACKET_TAG_IPOPTIONS			27 /* Saved IP options */
1063 #define	PACKET_TAG_CARP				28 /* CARP info */
1064 #define	PACKET_TAG_IPSEC_NAT_T_PORTS		29 /* two uint16_t */
1065 #define	PACKET_TAG_ND_OUTGOING			30 /* ND outgoing */
1066 
1067 /* Specific cookies and tags. */
1068 
1069 /* Packet tag routines. */
1070 struct m_tag	*m_tag_alloc(u_int32_t, int, int, int);
1071 void		 m_tag_delete(struct mbuf *, struct m_tag *);
1072 void		 m_tag_delete_chain(struct mbuf *, struct m_tag *);
1073 void		 m_tag_free_default(struct m_tag *);
1074 struct m_tag	*m_tag_locate(struct mbuf *, u_int32_t, int, struct m_tag *);
1075 struct m_tag	*m_tag_copy(struct m_tag *, int);
1076 int		 m_tag_copy_chain(struct mbuf *, const struct mbuf *, int);
1077 void		 m_tag_delete_nonpersistent(struct mbuf *);
1078 
1079 /*
1080  * Initialize the list of tags associated with an mbuf.
1081  */
1082 static __inline void
1083 m_tag_init(struct mbuf *m)
1084 {
1085 
1086 	SLIST_INIT(&m->m_pkthdr.tags);
1087 }
1088 
1089 /*
1090  * Set up the contents of a tag.  Note that this does not fill in the free
1091  * method; the caller is expected to do that.
1092  *
1093  * XXX probably should be called m_tag_init, but that was already taken.
1094  */
1095 static __inline void
1096 m_tag_setup(struct m_tag *t, u_int32_t cookie, int type, int len)
1097 {
1098 
1099 	t->m_tag_id = type;
1100 	t->m_tag_len = len;
1101 	t->m_tag_cookie = cookie;
1102 }
1103 
1104 /*
1105  * Reclaim resources associated with a tag.
1106  */
1107 static __inline void
1108 m_tag_free(struct m_tag *t)
1109 {
1110 
1111 	(*t->m_tag_free)(t);
1112 }
1113 
1114 /*
1115  * Return the first tag associated with an mbuf.
1116  */
1117 static __inline struct m_tag *
1118 m_tag_first(struct mbuf *m)
1119 {
1120 
1121 	return (SLIST_FIRST(&m->m_pkthdr.tags));
1122 }
1123 
1124 /*
1125  * Return the next tag in the list of tags associated with an mbuf.
1126  */
1127 static __inline struct m_tag *
1128 m_tag_next(struct mbuf *m __unused, struct m_tag *t)
1129 {
1130 
1131 	return (SLIST_NEXT(t, m_tag_link));
1132 }
1133 
1134 /*
1135  * Prepend a tag to the list of tags associated with an mbuf.
1136  */
1137 static __inline void
1138 m_tag_prepend(struct mbuf *m, struct m_tag *t)
1139 {
1140 
1141 	SLIST_INSERT_HEAD(&m->m_pkthdr.tags, t, m_tag_link);
1142 }
1143 
1144 /*
1145  * Unlink a tag from the list of tags associated with an mbuf.
1146  */
1147 static __inline void
1148 m_tag_unlink(struct mbuf *m, struct m_tag *t)
1149 {
1150 
1151 	SLIST_REMOVE(&m->m_pkthdr.tags, t, m_tag, m_tag_link);
1152 }
1153 
1154 /* These are for OpenBSD compatibility. */
1155 #define	MTAG_ABI_COMPAT		0		/* compatibility ABI */
1156 
1157 static __inline struct m_tag *
1158 m_tag_get(int type, int length, int wait)
1159 {
1160 	return (m_tag_alloc(MTAG_ABI_COMPAT, type, length, wait));
1161 }
1162 
1163 static __inline struct m_tag *
1164 m_tag_find(struct mbuf *m, int type, struct m_tag *start)
1165 {
1166 	return (SLIST_EMPTY(&m->m_pkthdr.tags) ? (struct m_tag *)NULL :
1167 	    m_tag_locate(m, MTAG_ABI_COMPAT, type, start));
1168 }
1169 
1170 static __inline struct mbuf *
1171 m_free(struct mbuf *m)
1172 {
1173 	struct mbuf *n = m->m_next;
1174 
1175 	if ((m->m_flags & (M_PKTHDR|M_NOFREE)) == (M_PKTHDR|M_NOFREE))
1176 		m_tag_delete_chain(m, NULL);
1177 	if (m->m_flags & M_EXT)
1178 		mb_free_ext(m);
1179 	else if ((m->m_flags & M_NOFREE) == 0)
1180 		uma_zfree(zone_mbuf, m);
1181 	return (n);
1182 }
1183 
1184 static __inline int
1185 rt_m_getfib(struct mbuf *m)
1186 {
1187 	KASSERT(m->m_flags & M_PKTHDR , ("Attempt to get FIB from non header mbuf."));
1188 	return (m->m_pkthdr.fibnum);
1189 }
1190 
1191 #define M_GETFIB(_m)   rt_m_getfib(_m)
1192 
1193 #define M_SETFIB(_m, _fib) do {						\
1194         KASSERT((_m)->m_flags & M_PKTHDR, ("Attempt to set FIB on non header mbuf."));	\
1195 	((_m)->m_pkthdr.fibnum) = (_fib);				\
1196 } while (0)
1197 
1198 /* flags passed as first argument for "m_ether_tcpip_hash()" */
1199 #define	MBUF_HASHFLAG_L2	(1 << 2)
1200 #define	MBUF_HASHFLAG_L3	(1 << 3)
1201 #define	MBUF_HASHFLAG_L4	(1 << 4)
1202 
1203 /* mbuf hashing helper routines */
1204 uint32_t	m_ether_tcpip_hash_init(void);
1205 uint32_t	m_ether_tcpip_hash(const uint32_t, const struct mbuf *, const uint32_t);
1206 
1207 #ifdef MBUF_PROFILING
1208  void m_profile(struct mbuf *m);
1209  #define M_PROFILE(m) m_profile(m)
1210 #else
1211  #define M_PROFILE(m)
1212 #endif
1213 
1214 struct mbufq {
1215 	STAILQ_HEAD(, mbuf)	mq_head;
1216 	int			mq_len;
1217 	int			mq_maxlen;
1218 };
1219 
1220 static inline void
1221 mbufq_init(struct mbufq *mq, int maxlen)
1222 {
1223 
1224 	STAILQ_INIT(&mq->mq_head);
1225 	mq->mq_maxlen = maxlen;
1226 	mq->mq_len = 0;
1227 }
1228 
1229 static inline struct mbuf *
1230 mbufq_flush(struct mbufq *mq)
1231 {
1232 	struct mbuf *m;
1233 
1234 	m = STAILQ_FIRST(&mq->mq_head);
1235 	STAILQ_INIT(&mq->mq_head);
1236 	mq->mq_len = 0;
1237 	return (m);
1238 }
1239 
1240 static inline void
1241 mbufq_drain(struct mbufq *mq)
1242 {
1243 	struct mbuf *m, *n;
1244 
1245 	n = mbufq_flush(mq);
1246 	while ((m = n) != NULL) {
1247 		n = STAILQ_NEXT(m, m_stailqpkt);
1248 		m_freem(m);
1249 	}
1250 }
1251 
1252 static inline struct mbuf *
1253 mbufq_first(const struct mbufq *mq)
1254 {
1255 
1256 	return (STAILQ_FIRST(&mq->mq_head));
1257 }
1258 
1259 static inline struct mbuf *
1260 mbufq_last(const struct mbufq *mq)
1261 {
1262 
1263 	return (STAILQ_LAST(&mq->mq_head, mbuf, m_stailqpkt));
1264 }
1265 
1266 static inline int
1267 mbufq_full(const struct mbufq *mq)
1268 {
1269 
1270 	return (mq->mq_len >= mq->mq_maxlen);
1271 }
1272 
1273 static inline int
1274 mbufq_len(const struct mbufq *mq)
1275 {
1276 
1277 	return (mq->mq_len);
1278 }
1279 
1280 static inline int
1281 mbufq_enqueue(struct mbufq *mq, struct mbuf *m)
1282 {
1283 
1284 	if (mbufq_full(mq))
1285 		return (ENOBUFS);
1286 	STAILQ_INSERT_TAIL(&mq->mq_head, m, m_stailqpkt);
1287 	mq->mq_len++;
1288 	return (0);
1289 }
1290 
1291 static inline struct mbuf *
1292 mbufq_dequeue(struct mbufq *mq)
1293 {
1294 	struct mbuf *m;
1295 
1296 	m = STAILQ_FIRST(&mq->mq_head);
1297 	if (m) {
1298 		STAILQ_REMOVE_HEAD(&mq->mq_head, m_stailqpkt);
1299 		m->m_nextpkt = NULL;
1300 		mq->mq_len--;
1301 	}
1302 	return (m);
1303 }
1304 
1305 static inline void
1306 mbufq_prepend(struct mbufq *mq, struct mbuf *m)
1307 {
1308 
1309 	STAILQ_INSERT_HEAD(&mq->mq_head, m, m_stailqpkt);
1310 	mq->mq_len++;
1311 }
1312 #endif /* _KERNEL */
1313 #endif /* !_SYS_MBUF_H_ */
1314