xref: /freebsd/sys/sys/mbuf.h (revision 15c433351f54e7cd5bec8d36c8e89e6a7fa55b26)
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 	union {
164 		volatile u_int	 ext_count;	/* value of ref count info */
165 		volatile u_int	*ext_cnt;	/* pointer to ref count info */
166 	};
167 	caddr_t		 ext_buf;	/* start of buffer */
168 	uint32_t	 ext_size;	/* size of buffer, for ext_free */
169 	uint32_t	 ext_type:8,	/* type of external storage */
170 			 ext_flags:24;	/* external storage mbuf flags */
171 	void		(*ext_free)	/* free routine if not the usual */
172 			    (struct mbuf *, void *, void *);
173 	void		*ext_arg1;	/* optional argument pointer */
174 	void		*ext_arg2;	/* optional argument pointer */
175 };
176 
177 /*
178  * The core of the mbuf object along with some shortcut defines for practical
179  * purposes.
180  */
181 struct mbuf {
182 	/*
183 	 * Header present at the beginning of every mbuf.
184 	 * Size ILP32: 24
185 	 *      LP64: 32
186 	 * Compile-time assertions in uipc_mbuf.c test these values to ensure
187 	 * that they are correct.
188 	 */
189 	union {	/* next buffer in chain */
190 		struct mbuf		*m_next;
191 		SLIST_ENTRY(mbuf)	m_slist;
192 		STAILQ_ENTRY(mbuf)	m_stailq;
193 	};
194 	union {	/* next chain in queue/record */
195 		struct mbuf		*m_nextpkt;
196 		SLIST_ENTRY(mbuf)	m_slistpkt;
197 		STAILQ_ENTRY(mbuf)	m_stailqpkt;
198 	};
199 	caddr_t		 m_data;	/* location of data */
200 	int32_t		 m_len;		/* amount of data in this mbuf */
201 	uint32_t	 m_type:8,	/* type of data in this mbuf */
202 			 m_flags:24;	/* flags; see below */
203 #if !defined(__LP64__)
204 	uint32_t	 m_pad;		/* pad for 64bit alignment */
205 #endif
206 
207 	/*
208 	 * A set of optional headers (packet header, external storage header)
209 	 * and internal data storage.  Historically, these arrays were sized
210 	 * to MHLEN (space left after a packet header) and MLEN (space left
211 	 * after only a regular mbuf header); they are now variable size in
212 	 * order to support future work on variable-size mbufs.
213 	 */
214 	union {
215 		struct {
216 			struct pkthdr	m_pkthdr;	/* M_PKTHDR set */
217 			union {
218 				struct m_ext	m_ext;	/* M_EXT set */
219 				char		m_pktdat[0];
220 			};
221 		};
222 		char	m_dat[0];			/* !M_PKTHDR, !M_EXT */
223 	};
224 };
225 
226 /*
227  * mbuf flags of global significance and layer crossing.
228  * Those of only protocol/layer specific significance are to be mapped
229  * to M_PROTO[1-12] and cleared at layer handoff boundaries.
230  * NB: Limited to the lower 24 bits.
231  */
232 #define	M_EXT		0x00000001 /* has associated external storage */
233 #define	M_PKTHDR	0x00000002 /* start of record */
234 #define	M_EOR		0x00000004 /* end of record */
235 #define	M_RDONLY	0x00000008 /* associated data is marked read-only */
236 #define	M_BCAST		0x00000010 /* send/received as link-level broadcast */
237 #define	M_MCAST		0x00000020 /* send/received as link-level multicast */
238 #define	M_PROMISC	0x00000040 /* packet was not for us */
239 #define	M_VLANTAG	0x00000080 /* ether_vtag is valid */
240 #define	M_UNUSED_8	0x00000100 /* --available-- */
241 #define	M_NOFREE	0x00000200 /* do not free mbuf, embedded in cluster */
242 
243 #define	M_PROTO1	0x00001000 /* protocol-specific */
244 #define	M_PROTO2	0x00002000 /* protocol-specific */
245 #define	M_PROTO3	0x00004000 /* protocol-specific */
246 #define	M_PROTO4	0x00008000 /* protocol-specific */
247 #define	M_PROTO5	0x00010000 /* protocol-specific */
248 #define	M_PROTO6	0x00020000 /* protocol-specific */
249 #define	M_PROTO7	0x00040000 /* protocol-specific */
250 #define	M_PROTO8	0x00080000 /* protocol-specific */
251 #define	M_PROTO9	0x00100000 /* protocol-specific */
252 #define	M_PROTO10	0x00200000 /* protocol-specific */
253 #define	M_PROTO11	0x00400000 /* protocol-specific */
254 #define	M_PROTO12	0x00800000 /* protocol-specific */
255 
256 /*
257  * Flags to purge when crossing layers.
258  */
259 #define	M_PROTOFLAGS \
260     (M_PROTO1|M_PROTO2|M_PROTO3|M_PROTO4|M_PROTO5|M_PROTO6|M_PROTO7|M_PROTO8|\
261      M_PROTO9|M_PROTO10|M_PROTO11|M_PROTO12)
262 
263 /*
264  * Flags preserved when copying m_pkthdr.
265  */
266 #define M_COPYFLAGS \
267     (M_PKTHDR|M_EOR|M_RDONLY|M_BCAST|M_MCAST|M_PROMISC|M_VLANTAG| \
268      M_PROTOFLAGS)
269 
270 /*
271  * Mbuf flag description for use with printf(9) %b identifier.
272  */
273 #define	M_FLAG_BITS \
274     "\20\1M_EXT\2M_PKTHDR\3M_EOR\4M_RDONLY\5M_BCAST\6M_MCAST" \
275     "\7M_PROMISC\10M_VLANTAG"
276 #define	M_FLAG_PROTOBITS \
277     "\15M_PROTO1\16M_PROTO2\17M_PROTO3\20M_PROTO4\21M_PROTO5" \
278     "\22M_PROTO6\23M_PROTO7\24M_PROTO8\25M_PROTO9\26M_PROTO10" \
279     "\27M_PROTO11\30M_PROTO12"
280 #define	M_FLAG_PRINTF (M_FLAG_BITS M_FLAG_PROTOBITS)
281 
282 /*
283  * Network interface cards are able to hash protocol fields (such as IPv4
284  * addresses and TCP port numbers) classify packets into flows.  These flows
285  * can then be used to maintain ordering while delivering packets to the OS
286  * via parallel input queues, as well as to provide a stateless affinity
287  * model.  NIC drivers can pass up the hash via m->m_pkthdr.flowid, and set
288  * m_flag fields to indicate how the hash should be interpreted by the
289  * network stack.
290  *
291  * Most NICs support RSS, which provides ordering and explicit affinity, and
292  * use the hash m_flag bits to indicate what header fields were covered by
293  * the hash.  M_HASHTYPE_OPAQUE can be set by non-RSS cards or configurations
294  * that provide an opaque flow identifier, allowing for ordering and
295  * distribution without explicit affinity.
296  */
297 /* Microsoft RSS standard hash types */
298 #define	M_HASHTYPE_NONE			0
299 #define	M_HASHTYPE_RSS_IPV4		1	/* IPv4 2-tuple */
300 #define	M_HASHTYPE_RSS_TCP_IPV4		2	/* TCPv4 4-tuple */
301 #define	M_HASHTYPE_RSS_IPV6		3	/* IPv6 2-tuple */
302 #define	M_HASHTYPE_RSS_TCP_IPV6		4	/* TCPv6 4-tuple */
303 #define	M_HASHTYPE_RSS_IPV6_EX		5	/* IPv6 2-tuple + ext hdrs */
304 #define	M_HASHTYPE_RSS_TCP_IPV6_EX	6	/* TCPv6 4-tiple + ext hdrs */
305 /* Non-standard RSS hash types */
306 #define	M_HASHTYPE_RSS_UDP_IPV4		7	/* IPv4 UDP 4-tuple */
307 #define	M_HASHTYPE_RSS_UDP_IPV4_EX	8	/* IPv4 UDP 4-tuple + ext hdrs */
308 #define	M_HASHTYPE_RSS_UDP_IPV6		9	/* IPv6 UDP 4-tuple */
309 #define	M_HASHTYPE_RSS_UDP_IPV6_EX	10	/* IPv6 UDP 4-tuple + ext hdrs */
310 
311 #define	M_HASHTYPE_OPAQUE		255	/* ordering, not affinity */
312 
313 #define	M_HASHTYPE_CLEAR(m)	((m)->m_pkthdr.rsstype = 0)
314 #define	M_HASHTYPE_GET(m)	((m)->m_pkthdr.rsstype)
315 #define	M_HASHTYPE_SET(m, v)	((m)->m_pkthdr.rsstype = (v))
316 #define	M_HASHTYPE_TEST(m, v)	(M_HASHTYPE_GET(m) == (v))
317 
318 /*
319  * COS/QOS class and quality of service tags.
320  * It uses DSCP code points as base.
321  */
322 #define	QOS_DSCP_CS0		0x00
323 #define	QOS_DSCP_DEF		QOS_DSCP_CS0
324 #define	QOS_DSCP_CS1		0x20
325 #define	QOS_DSCP_AF11		0x28
326 #define	QOS_DSCP_AF12		0x30
327 #define	QOS_DSCP_AF13		0x38
328 #define	QOS_DSCP_CS2		0x40
329 #define	QOS_DSCP_AF21		0x48
330 #define	QOS_DSCP_AF22		0x50
331 #define	QOS_DSCP_AF23		0x58
332 #define	QOS_DSCP_CS3		0x60
333 #define	QOS_DSCP_AF31		0x68
334 #define	QOS_DSCP_AF32		0x70
335 #define	QOS_DSCP_AF33		0x78
336 #define	QOS_DSCP_CS4		0x80
337 #define	QOS_DSCP_AF41		0x88
338 #define	QOS_DSCP_AF42		0x90
339 #define	QOS_DSCP_AF43		0x98
340 #define	QOS_DSCP_CS5		0xa0
341 #define	QOS_DSCP_EF		0xb8
342 #define	QOS_DSCP_CS6		0xc0
343 #define	QOS_DSCP_CS7		0xe0
344 
345 /*
346  * External mbuf storage buffer types.
347  */
348 #define	EXT_CLUSTER	1	/* mbuf cluster */
349 #define	EXT_SFBUF	2	/* sendfile(2)'s sf_buf */
350 #define	EXT_JUMBOP	3	/* jumbo cluster page sized */
351 #define	EXT_JUMBO9	4	/* jumbo cluster 9216 bytes */
352 #define	EXT_JUMBO16	5	/* jumbo cluster 16184 bytes */
353 #define	EXT_PACKET	6	/* mbuf+cluster from packet zone */
354 #define	EXT_MBUF	7	/* external mbuf reference (M_IOVEC) */
355 #define	EXT_SFBUF_NOCACHE 8	/* sendfile(2)'s sf_buf not to be cached */
356 
357 #define	EXT_VENDOR1	224	/* for vendor-internal use */
358 #define	EXT_VENDOR2	225	/* for vendor-internal use */
359 #define	EXT_VENDOR3	226	/* for vendor-internal use */
360 #define	EXT_VENDOR4	227	/* for vendor-internal use */
361 
362 #define	EXT_EXP1	244	/* for experimental use */
363 #define	EXT_EXP2	245	/* for experimental use */
364 #define	EXT_EXP3	246	/* for experimental use */
365 #define	EXT_EXP4	247	/* for experimental use */
366 
367 #define	EXT_NET_DRV	252	/* custom ext_buf provided by net driver(s) */
368 #define	EXT_MOD_TYPE	253	/* custom module's ext_buf type */
369 #define	EXT_DISPOSABLE	254	/* can throw this buffer away w/page flipping */
370 #define	EXT_EXTREF	255	/* has externally maintained ext_cnt ptr */
371 
372 /*
373  * Flags for external mbuf buffer types.
374  * NB: limited to the lower 24 bits.
375  */
376 #define	EXT_FLAG_EMBREF		0x000001	/* embedded ext_count */
377 #define	EXT_FLAG_EXTREF		0x000002	/* external ext_cnt, notyet */
378 #define	EXT_FLAG_NOFREE		0x000010	/* don't free mbuf to pool, notyet */
379 
380 #define	EXT_FLAG_VENDOR1	0x010000	/* for vendor-internal use */
381 #define	EXT_FLAG_VENDOR2	0x020000	/* for vendor-internal use */
382 #define	EXT_FLAG_VENDOR3	0x040000	/* for vendor-internal use */
383 #define	EXT_FLAG_VENDOR4	0x080000	/* for vendor-internal use */
384 
385 #define	EXT_FLAG_EXP1		0x100000	/* for experimental use */
386 #define	EXT_FLAG_EXP2		0x200000	/* for experimental use */
387 #define	EXT_FLAG_EXP3		0x400000	/* for experimental use */
388 #define	EXT_FLAG_EXP4		0x800000	/* for experimental use */
389 
390 /*
391  * EXT flag description for use with printf(9) %b identifier.
392  */
393 #define	EXT_FLAG_BITS \
394     "\20\1EXT_FLAG_EMBREF\2EXT_FLAG_EXTREF\5EXT_FLAG_NOFREE" \
395     "\21EXT_FLAG_VENDOR1\22EXT_FLAG_VENDOR2\23EXT_FLAG_VENDOR3" \
396     "\24EXT_FLAG_VENDOR4\25EXT_FLAG_EXP1\26EXT_FLAG_EXP2\27EXT_FLAG_EXP3" \
397     "\30EXT_FLAG_EXP4"
398 
399 /*
400  * External reference/free functions.
401  */
402 void sf_ext_free(void *, void *);
403 void sf_ext_free_nocache(void *, void *);
404 
405 /*
406  * Flags indicating checksum, segmentation and other offload work to be
407  * done, or already done, by hardware or lower layers.  It is split into
408  * separate inbound and outbound flags.
409  *
410  * Outbound flags that are set by upper protocol layers requesting lower
411  * layers, or ideally the hardware, to perform these offloading tasks.
412  * For outbound packets this field and its flags can be directly tested
413  * against ifnet if_hwassist.
414  */
415 #define	CSUM_IP			0x00000001	/* IP header checksum offload */
416 #define	CSUM_IP_UDP		0x00000002	/* UDP checksum offload */
417 #define	CSUM_IP_TCP		0x00000004	/* TCP checksum offload */
418 #define	CSUM_IP_SCTP		0x00000008	/* SCTP checksum offload */
419 #define	CSUM_IP_TSO		0x00000010	/* TCP segmentation offload */
420 #define	CSUM_IP_ISCSI		0x00000020	/* iSCSI checksum offload */
421 
422 #define	CSUM_IP6_UDP		0x00000200	/* UDP checksum offload */
423 #define	CSUM_IP6_TCP		0x00000400	/* TCP checksum offload */
424 #define	CSUM_IP6_SCTP		0x00000800	/* SCTP checksum offload */
425 #define	CSUM_IP6_TSO		0x00001000	/* TCP segmentation offload */
426 #define	CSUM_IP6_ISCSI		0x00002000	/* iSCSI checksum offload */
427 
428 /* Inbound checksum support where the checksum was verified by hardware. */
429 #define	CSUM_L3_CALC		0x01000000	/* calculated layer 3 csum */
430 #define	CSUM_L3_VALID		0x02000000	/* checksum is correct */
431 #define	CSUM_L4_CALC		0x04000000	/* calculated layer 4 csum */
432 #define	CSUM_L4_VALID		0x08000000	/* checksum is correct */
433 #define	CSUM_L5_CALC		0x10000000	/* calculated layer 5 csum */
434 #define	CSUM_L5_VALID		0x20000000	/* checksum is correct */
435 #define	CSUM_COALESED		0x40000000	/* contains merged segments */
436 
437 /*
438  * CSUM flag description for use with printf(9) %b identifier.
439  */
440 #define	CSUM_BITS \
441     "\20\1CSUM_IP\2CSUM_IP_UDP\3CSUM_IP_TCP\4CSUM_IP_SCTP\5CSUM_IP_TSO" \
442     "\6CSUM_IP_ISCSI" \
443     "\12CSUM_IP6_UDP\13CSUM_IP6_TCP\14CSUM_IP6_SCTP\15CSUM_IP6_TSO" \
444     "\16CSUM_IP6_ISCSI" \
445     "\31CSUM_L3_CALC\32CSUM_L3_VALID\33CSUM_L4_CALC\34CSUM_L4_VALID" \
446     "\35CSUM_L5_CALC\36CSUM_L5_VALID\37CSUM_COALESED"
447 
448 /* CSUM flags compatibility mappings. */
449 #define	CSUM_IP_CHECKED		CSUM_L3_CALC
450 #define	CSUM_IP_VALID		CSUM_L3_VALID
451 #define	CSUM_DATA_VALID		CSUM_L4_VALID
452 #define	CSUM_PSEUDO_HDR		CSUM_L4_CALC
453 #define	CSUM_SCTP_VALID		CSUM_L4_VALID
454 #define	CSUM_DELAY_DATA		(CSUM_TCP|CSUM_UDP)
455 #define	CSUM_DELAY_IP		CSUM_IP		/* Only v4, no v6 IP hdr csum */
456 #define	CSUM_DELAY_DATA_IPV6	(CSUM_TCP_IPV6|CSUM_UDP_IPV6)
457 #define	CSUM_DATA_VALID_IPV6	CSUM_DATA_VALID
458 #define	CSUM_TCP		CSUM_IP_TCP
459 #define	CSUM_UDP		CSUM_IP_UDP
460 #define	CSUM_SCTP		CSUM_IP_SCTP
461 #define	CSUM_TSO		(CSUM_IP_TSO|CSUM_IP6_TSO)
462 #define	CSUM_UDP_IPV6		CSUM_IP6_UDP
463 #define	CSUM_TCP_IPV6		CSUM_IP6_TCP
464 #define	CSUM_SCTP_IPV6		CSUM_IP6_SCTP
465 
466 /*
467  * mbuf types describing the content of the mbuf (including external storage).
468  */
469 #define	MT_NOTMBUF	0	/* USED INTERNALLY ONLY! Object is not mbuf */
470 #define	MT_DATA		1	/* dynamic (data) allocation */
471 #define	MT_HEADER	MT_DATA	/* packet header, use M_PKTHDR instead */
472 
473 #define	MT_VENDOR1	4	/* for vendor-internal use */
474 #define	MT_VENDOR2	5	/* for vendor-internal use */
475 #define	MT_VENDOR3	6	/* for vendor-internal use */
476 #define	MT_VENDOR4	7	/* for vendor-internal use */
477 
478 #define	MT_SONAME	8	/* socket name */
479 
480 #define	MT_EXP1		9	/* for experimental use */
481 #define	MT_EXP2		10	/* for experimental use */
482 #define	MT_EXP3		11	/* for experimental use */
483 #define	MT_EXP4		12	/* for experimental use */
484 
485 #define	MT_CONTROL	14	/* extra-data protocol message */
486 #define	MT_OOBDATA	15	/* expedited data  */
487 #define	MT_NTYPES	16	/* number of mbuf types for mbtypes[] */
488 
489 #define	MT_NOINIT	255	/* Not a type but a flag to allocate
490 				   a non-initialized mbuf */
491 
492 /*
493  * String names of mbuf-related UMA(9) and malloc(9) types.  Exposed to
494  * !_KERNEL so that monitoring tools can look up the zones with
495  * libmemstat(3).
496  */
497 #define	MBUF_MEM_NAME		"mbuf"
498 #define	MBUF_CLUSTER_MEM_NAME	"mbuf_cluster"
499 #define	MBUF_PACKET_MEM_NAME	"mbuf_packet"
500 #define	MBUF_JUMBOP_MEM_NAME	"mbuf_jumbo_page"
501 #define	MBUF_JUMBO9_MEM_NAME	"mbuf_jumbo_9k"
502 #define	MBUF_JUMBO16_MEM_NAME	"mbuf_jumbo_16k"
503 #define	MBUF_TAG_MEM_NAME	"mbuf_tag"
504 #define	MBUF_EXTREFCNT_MEM_NAME	"mbuf_ext_refcnt"
505 
506 #ifdef _KERNEL
507 
508 #ifdef WITNESS
509 #define	MBUF_CHECKSLEEP(how) do {					\
510 	if (how == M_WAITOK)						\
511 		WITNESS_WARN(WARN_GIANTOK | WARN_SLEEPOK, NULL,		\
512 		    "Sleeping in \"%s\"", __func__);			\
513 } while (0)
514 #else
515 #define	MBUF_CHECKSLEEP(how)
516 #endif
517 
518 /*
519  * Network buffer allocation API
520  *
521  * The rest of it is defined in kern/kern_mbuf.c
522  */
523 extern uma_zone_t	zone_mbuf;
524 extern uma_zone_t	zone_clust;
525 extern uma_zone_t	zone_pack;
526 extern uma_zone_t	zone_jumbop;
527 extern uma_zone_t	zone_jumbo9;
528 extern uma_zone_t	zone_jumbo16;
529 
530 void		 mb_dupcl(struct mbuf *, struct mbuf *);
531 void		 mb_free_ext(struct mbuf *);
532 void		 m_adj(struct mbuf *, int);
533 int		 m_apply(struct mbuf *, int, int,
534 		    int (*)(void *, void *, u_int), void *);
535 int		 m_append(struct mbuf *, int, c_caddr_t);
536 void		 m_cat(struct mbuf *, struct mbuf *);
537 void		 m_catpkt(struct mbuf *, struct mbuf *);
538 int		 m_clget(struct mbuf *m, int how);
539 void 		*m_cljget(struct mbuf *m, int how, int size);
540 struct mbuf	*m_collapse(struct mbuf *, int, int);
541 void		 m_copyback(struct mbuf *, int, int, c_caddr_t);
542 void		 m_copydata(const struct mbuf *, int, int, caddr_t);
543 struct mbuf	*m_copym(struct mbuf *, int, int, int);
544 struct mbuf	*m_copypacket(struct mbuf *, int);
545 void		 m_copy_pkthdr(struct mbuf *, struct mbuf *);
546 struct mbuf	*m_copyup(struct mbuf *, int, int);
547 struct mbuf	*m_defrag(struct mbuf *, int);
548 void		 m_demote_pkthdr(struct mbuf *);
549 void		 m_demote(struct mbuf *, int, int);
550 struct mbuf	*m_devget(char *, int, int, struct ifnet *,
551 		    void (*)(char *, caddr_t, u_int));
552 struct mbuf	*m_dup(const struct mbuf *, int);
553 int		 m_dup_pkthdr(struct mbuf *, const struct mbuf *, int);
554 void		 m_extadd(struct mbuf *, caddr_t, u_int,
555 		    void (*)(struct mbuf *, void *, void *), void *, void *,
556 		    int, int);
557 u_int		 m_fixhdr(struct mbuf *);
558 struct mbuf	*m_fragment(struct mbuf *, int, int);
559 void		 m_freem(struct mbuf *);
560 struct mbuf	*m_get2(int, int, short, int);
561 struct mbuf	*m_getjcl(int, short, int, int);
562 struct mbuf	*m_getm2(struct mbuf *, int, int, short, int);
563 struct mbuf	*m_getptr(struct mbuf *, int, int *);
564 u_int		 m_length(struct mbuf *, struct mbuf **);
565 int		 m_mbuftouio(struct uio *, struct mbuf *, int);
566 void		 m_move_pkthdr(struct mbuf *, struct mbuf *);
567 int		 m_pkthdr_init(struct mbuf *, int);
568 struct mbuf	*m_prepend(struct mbuf *, int, int);
569 void		 m_print(const struct mbuf *, int);
570 struct mbuf	*m_pulldown(struct mbuf *, int, int, int *);
571 struct mbuf	*m_pullup(struct mbuf *, int);
572 int		 m_sanity(struct mbuf *, int);
573 struct mbuf	*m_split(struct mbuf *, int, int);
574 struct mbuf	*m_uiotombuf(struct uio *, int, int, int, int);
575 struct mbuf	*m_unshare(struct mbuf *, int);
576 
577 static __inline int
578 m_gettype(int size)
579 {
580 	int type;
581 
582 	switch (size) {
583 	case MSIZE:
584 		type = EXT_MBUF;
585 		break;
586 	case MCLBYTES:
587 		type = EXT_CLUSTER;
588 		break;
589 #if MJUMPAGESIZE != MCLBYTES
590 	case MJUMPAGESIZE:
591 		type = EXT_JUMBOP;
592 		break;
593 #endif
594 	case MJUM9BYTES:
595 		type = EXT_JUMBO9;
596 		break;
597 	case MJUM16BYTES:
598 		type = EXT_JUMBO16;
599 		break;
600 	default:
601 		panic("%s: invalid cluster size %d", __func__, size);
602 	}
603 
604 	return (type);
605 }
606 
607 /*
608  * Associated an external reference counted buffer with an mbuf.
609  */
610 static __inline void
611 m_extaddref(struct mbuf *m, caddr_t buf, u_int size, u_int *ref_cnt,
612     void (*freef)(struct mbuf *, void *, void *), void *arg1, void *arg2)
613 {
614 
615 	KASSERT(ref_cnt != NULL, ("%s: ref_cnt not provided", __func__));
616 
617 	atomic_add_int(ref_cnt, 1);
618 	m->m_flags |= M_EXT;
619 	m->m_ext.ext_buf = buf;
620 	m->m_ext.ext_cnt = ref_cnt;
621 	m->m_data = m->m_ext.ext_buf;
622 	m->m_ext.ext_size = size;
623 	m->m_ext.ext_free = freef;
624 	m->m_ext.ext_arg1 = arg1;
625 	m->m_ext.ext_arg2 = arg2;
626 	m->m_ext.ext_type = EXT_EXTREF;
627 	m->m_ext.ext_flags = 0;
628 }
629 
630 static __inline uma_zone_t
631 m_getzone(int size)
632 {
633 	uma_zone_t zone;
634 
635 	switch (size) {
636 	case MCLBYTES:
637 		zone = zone_clust;
638 		break;
639 #if MJUMPAGESIZE != MCLBYTES
640 	case MJUMPAGESIZE:
641 		zone = zone_jumbop;
642 		break;
643 #endif
644 	case MJUM9BYTES:
645 		zone = zone_jumbo9;
646 		break;
647 	case MJUM16BYTES:
648 		zone = zone_jumbo16;
649 		break;
650 	default:
651 		panic("%s: invalid cluster size %d", __func__, size);
652 	}
653 
654 	return (zone);
655 }
656 
657 /*
658  * Initialize an mbuf with linear storage.
659  *
660  * Inline because the consumer text overhead will be roughly the same to
661  * initialize or call a function with this many parameters and M_PKTHDR
662  * should go away with constant propagation for !MGETHDR.
663  */
664 static __inline int
665 m_init(struct mbuf *m, int how, short type, int flags)
666 {
667 	int error;
668 
669 	m->m_next = NULL;
670 	m->m_nextpkt = NULL;
671 	m->m_data = m->m_dat;
672 	m->m_len = 0;
673 	m->m_flags = flags;
674 	m->m_type = type;
675 	if (flags & M_PKTHDR) {
676 		if ((error = m_pkthdr_init(m, how)) != 0)
677 			return (error);
678 	}
679 
680 	return (0);
681 }
682 
683 static __inline struct mbuf *
684 m_get(int how, short type)
685 {
686 	struct mb_args args;
687 
688 	args.flags = 0;
689 	args.type = type;
690 	return (uma_zalloc_arg(zone_mbuf, &args, how));
691 }
692 
693 static __inline struct mbuf *
694 m_gethdr(int how, short type)
695 {
696 	struct mb_args args;
697 
698 	args.flags = M_PKTHDR;
699 	args.type = type;
700 	return (uma_zalloc_arg(zone_mbuf, &args, how));
701 }
702 
703 static __inline struct mbuf *
704 m_getcl(int how, short type, int flags)
705 {
706 	struct mb_args args;
707 
708 	args.flags = flags;
709 	args.type = type;
710 	return (uma_zalloc_arg(zone_pack, &args, how));
711 }
712 
713 /*
714  * XXX: m_cljset() is a dangerous API.  One must attach only a new,
715  * unreferenced cluster to an mbuf(9).  It is not possible to assert
716  * that, so care can be taken only by users of the API.
717  */
718 static __inline void
719 m_cljset(struct mbuf *m, void *cl, int type)
720 {
721 	int size;
722 
723 	switch (type) {
724 	case EXT_CLUSTER:
725 		size = MCLBYTES;
726 		break;
727 #if MJUMPAGESIZE != MCLBYTES
728 	case EXT_JUMBOP:
729 		size = MJUMPAGESIZE;
730 		break;
731 #endif
732 	case EXT_JUMBO9:
733 		size = MJUM9BYTES;
734 		break;
735 	case EXT_JUMBO16:
736 		size = MJUM16BYTES;
737 		break;
738 	default:
739 		panic("%s: unknown cluster type %d", __func__, type);
740 		break;
741 	}
742 
743 	m->m_data = m->m_ext.ext_buf = cl;
744 	m->m_ext.ext_free = m->m_ext.ext_arg1 = m->m_ext.ext_arg2 = NULL;
745 	m->m_ext.ext_size = size;
746 	m->m_ext.ext_type = type;
747 	m->m_ext.ext_flags = EXT_FLAG_EMBREF;
748 	m->m_ext.ext_count = 1;
749 	m->m_flags |= M_EXT;
750 }
751 
752 static __inline void
753 m_chtype(struct mbuf *m, short new_type)
754 {
755 
756 	m->m_type = new_type;
757 }
758 
759 static __inline void
760 m_clrprotoflags(struct mbuf *m)
761 {
762 
763 	while (m) {
764 		m->m_flags &= ~M_PROTOFLAGS;
765 		m = m->m_next;
766 	}
767 }
768 
769 static __inline struct mbuf *
770 m_last(struct mbuf *m)
771 {
772 
773 	while (m->m_next)
774 		m = m->m_next;
775 	return (m);
776 }
777 
778 static inline u_int
779 m_extrefcnt(struct mbuf *m)
780 {
781 
782 	KASSERT(m->m_flags & M_EXT, ("%s: M_EXT missing", __func__));
783 
784 	return ((m->m_ext.ext_flags & EXT_FLAG_EMBREF) ? m->m_ext.ext_count :
785 	    *m->m_ext.ext_cnt);
786 }
787 
788 /*
789  * mbuf, cluster, and external object allocation macros (for compatibility
790  * purposes).
791  */
792 #define	M_MOVE_PKTHDR(to, from)	m_move_pkthdr((to), (from))
793 #define	MGET(m, how, type)	((m) = m_get((how), (type)))
794 #define	MGETHDR(m, how, type)	((m) = m_gethdr((how), (type)))
795 #define	MCLGET(m, how)		m_clget((m), (how))
796 #define	MEXTADD(m, buf, size, free, arg1, arg2, flags, type)		\
797     m_extadd((m), (caddr_t)(buf), (size), (free), (arg1), (arg2),	\
798     (flags), (type))
799 #define	m_getm(m, len, how, type)					\
800     m_getm2((m), (len), (how), (type), M_PKTHDR)
801 
802 /*
803  * Evaluate TRUE if it's safe to write to the mbuf m's data region (this can
804  * be both the local data payload, or an external buffer area, depending on
805  * whether M_EXT is set).
806  */
807 #define	M_WRITABLE(m)	(!((m)->m_flags & M_RDONLY) &&			\
808 			 (!(((m)->m_flags & M_EXT)) ||			\
809 			 (m_extrefcnt(m) == 1)))
810 
811 /* Check if the supplied mbuf has a packet header, or else panic. */
812 #define	M_ASSERTPKTHDR(m)						\
813 	KASSERT((m) != NULL && (m)->m_flags & M_PKTHDR,			\
814 	    ("%s: no mbuf packet header!", __func__))
815 
816 /*
817  * Ensure that the supplied mbuf is a valid, non-free mbuf.
818  *
819  * XXX: Broken at the moment.  Need some UMA magic to make it work again.
820  */
821 #define	M_ASSERTVALID(m)						\
822 	KASSERT((((struct mbuf *)m)->m_flags & 0) == 0,			\
823 	    ("%s: attempted use of a free mbuf!", __func__))
824 
825 /*
826  * Return the address of the start of the buffer associated with an mbuf,
827  * handling external storage, packet-header mbufs, and regular data mbufs.
828  */
829 #define	M_START(m)							\
830 	(((m)->m_flags & M_EXT) ? (m)->m_ext.ext_buf :			\
831 	 ((m)->m_flags & M_PKTHDR) ? &(m)->m_pktdat[0] :		\
832 	 &(m)->m_dat[0])
833 
834 /*
835  * Return the size of the buffer associated with an mbuf, handling external
836  * storage, packet-header mbufs, and regular data mbufs.
837  */
838 #define	M_SIZE(m)							\
839 	(((m)->m_flags & M_EXT) ? (m)->m_ext.ext_size :			\
840 	 ((m)->m_flags & M_PKTHDR) ? MHLEN :				\
841 	 MLEN)
842 
843 /*
844  * Set the m_data pointer of a newly allocated mbuf to place an object of the
845  * specified size at the end of the mbuf, longword aligned.
846  *
847  * NB: Historically, we had M_ALIGN(), MH_ALIGN(), and MEXT_ALIGN() as
848  * separate macros, each asserting that it was called at the proper moment.
849  * This required callers to themselves test the storage type and call the
850  * right one.  Rather than require callers to be aware of those layout
851  * decisions, we centralize here.
852  */
853 static __inline void
854 m_align(struct mbuf *m, int len)
855 {
856 #ifdef INVARIANTS
857 	const char *msg = "%s: not a virgin mbuf";
858 #endif
859 	int adjust;
860 
861 	KASSERT(m->m_data == M_START(m), (msg, __func__));
862 
863 	adjust = M_SIZE(m) - len;
864 	m->m_data += adjust &~ (sizeof(long)-1);
865 }
866 
867 #define	M_ALIGN(m, len)		m_align(m, len)
868 #define	MH_ALIGN(m, len)	m_align(m, len)
869 #define	MEXT_ALIGN(m, len)	m_align(m, len)
870 
871 /*
872  * Compute the amount of space available before the current start of data in
873  * an mbuf.
874  *
875  * The M_WRITABLE() is a temporary, conservative safety measure: the burden
876  * of checking writability of the mbuf data area rests solely with the caller.
877  *
878  * NB: In previous versions, M_LEADINGSPACE() would only check M_WRITABLE()
879  * for mbufs with external storage.  We now allow mbuf-embedded data to be
880  * read-only as well.
881  */
882 #define	M_LEADINGSPACE(m)						\
883 	(M_WRITABLE(m) ? ((m)->m_data - M_START(m)) : 0)
884 
885 /*
886  * Compute the amount of space available after the end of data in an mbuf.
887  *
888  * The M_WRITABLE() is a temporary, conservative safety measure: the burden
889  * of checking writability of the mbuf data area rests solely with the caller.
890  *
891  * NB: In previous versions, M_TRAILINGSPACE() would only check M_WRITABLE()
892  * for mbufs with external storage.  We now allow mbuf-embedded data to be
893  * read-only as well.
894  */
895 #define	M_TRAILINGSPACE(m)						\
896 	(M_WRITABLE(m) ?						\
897 	    ((M_START(m) + M_SIZE(m)) - ((m)->m_data + (m)->m_len)) : 0)
898 
899 /*
900  * Arrange to prepend space of size plen to mbuf m.  If a new mbuf must be
901  * allocated, how specifies whether to wait.  If the allocation fails, the
902  * original mbuf chain is freed and m is set to NULL.
903  */
904 #define	M_PREPEND(m, plen, how) do {					\
905 	struct mbuf **_mmp = &(m);					\
906 	struct mbuf *_mm = *_mmp;					\
907 	int _mplen = (plen);						\
908 	int __mhow = (how);						\
909 									\
910 	MBUF_CHECKSLEEP(how);						\
911 	if (M_LEADINGSPACE(_mm) >= _mplen) {				\
912 		_mm->m_data -= _mplen;					\
913 		_mm->m_len += _mplen;					\
914 	} else								\
915 		_mm = m_prepend(_mm, _mplen, __mhow);			\
916 	if (_mm != NULL && _mm->m_flags & M_PKTHDR)			\
917 		_mm->m_pkthdr.len += _mplen;				\
918 	*_mmp = _mm;							\
919 } while (0)
920 
921 /*
922  * Change mbuf to new type.  This is a relatively expensive operation and
923  * should be avoided.
924  */
925 #define	MCHTYPE(m, t)	m_chtype((m), (t))
926 
927 /* Length to m_copy to copy all. */
928 #define	M_COPYALL	1000000000
929 
930 /* Compatibility with 4.3. */
931 #define	m_copy(m, o, l)	m_copym((m), (o), (l), M_NOWAIT)
932 
933 extern int		max_datalen;	/* MHLEN - max_hdr */
934 extern int		max_hdr;	/* Largest link + protocol header */
935 extern int		max_linkhdr;	/* Largest link-level header */
936 extern int		max_protohdr;	/* Largest protocol header */
937 extern int		nmbclusters;	/* Maximum number of clusters */
938 
939 /*-
940  * Network packets may have annotations attached by affixing a list of
941  * "packet tags" to the pkthdr structure.  Packet tags are dynamically
942  * allocated semi-opaque data structures that have a fixed header
943  * (struct m_tag) that specifies the size of the memory block and a
944  * <cookie,type> pair that identifies it.  The cookie is a 32-bit unique
945  * unsigned value used to identify a module or ABI.  By convention this value
946  * is chosen as the date+time that the module is created, expressed as the
947  * number of seconds since the epoch (e.g., using date -u +'%s').  The type
948  * value is an ABI/module-specific value that identifies a particular
949  * annotation and is private to the module.  For compatibility with systems
950  * like OpenBSD that define packet tags w/o an ABI/module cookie, the value
951  * PACKET_ABI_COMPAT is used to implement m_tag_get and m_tag_find
952  * compatibility shim functions and several tag types are defined below.
953  * Users that do not require compatibility should use a private cookie value
954  * so that packet tag-related definitions can be maintained privately.
955  *
956  * Note that the packet tag returned by m_tag_alloc has the default memory
957  * alignment implemented by malloc.  To reference private data one can use a
958  * construct like:
959  *
960  *	struct m_tag *mtag = m_tag_alloc(...);
961  *	struct foo *p = (struct foo *)(mtag+1);
962  *
963  * if the alignment of struct m_tag is sufficient for referencing members of
964  * struct foo.  Otherwise it is necessary to embed struct m_tag within the
965  * private data structure to insure proper alignment; e.g.,
966  *
967  *	struct foo {
968  *		struct m_tag	tag;
969  *		...
970  *	};
971  *	struct foo *p = (struct foo *) m_tag_alloc(...);
972  *	struct m_tag *mtag = &p->tag;
973  */
974 
975 /*
976  * Persistent tags stay with an mbuf until the mbuf is reclaimed.  Otherwise
977  * tags are expected to ``vanish'' when they pass through a network
978  * interface.  For most interfaces this happens normally as the tags are
979  * reclaimed when the mbuf is free'd.  However in some special cases
980  * reclaiming must be done manually.  An example is packets that pass through
981  * the loopback interface.  Also, one must be careful to do this when
982  * ``turning around'' packets (e.g., icmp_reflect).
983  *
984  * To mark a tag persistent bit-or this flag in when defining the tag id.
985  * The tag will then be treated as described above.
986  */
987 #define	MTAG_PERSISTENT				0x800
988 
989 #define	PACKET_TAG_NONE				0  /* Nadda */
990 
991 /* Packet tags for use with PACKET_ABI_COMPAT. */
992 #define	PACKET_TAG_IPSEC_IN_DONE		1  /* IPsec applied, in */
993 #define	PACKET_TAG_IPSEC_OUT_DONE		2  /* IPsec applied, out */
994 #define	PACKET_TAG_IPSEC_IN_CRYPTO_DONE		3  /* NIC IPsec crypto done */
995 #define	PACKET_TAG_IPSEC_OUT_CRYPTO_NEEDED	4  /* NIC IPsec crypto req'ed */
996 #define	PACKET_TAG_IPSEC_IN_COULD_DO_CRYPTO	5  /* NIC notifies IPsec */
997 #define	PACKET_TAG_IPSEC_PENDING_TDB		6  /* Reminder to do IPsec */
998 #define	PACKET_TAG_BRIDGE			7  /* Bridge processing done */
999 #define	PACKET_TAG_GIF				8  /* GIF processing done */
1000 #define	PACKET_TAG_GRE				9  /* GRE processing done */
1001 #define	PACKET_TAG_IN_PACKET_CHECKSUM		10 /* NIC checksumming done */
1002 #define	PACKET_TAG_ENCAP			11 /* Encap.  processing */
1003 #define	PACKET_TAG_IPSEC_SOCKET			12 /* IPSEC socket ref */
1004 #define	PACKET_TAG_IPSEC_HISTORY		13 /* IPSEC history */
1005 #define	PACKET_TAG_IPV6_INPUT			14 /* IPV6 input processing */
1006 #define	PACKET_TAG_DUMMYNET			15 /* dummynet info */
1007 #define	PACKET_TAG_DIVERT			17 /* divert info */
1008 #define	PACKET_TAG_IPFORWARD			18 /* ipforward info */
1009 #define	PACKET_TAG_MACLABEL	(19 | MTAG_PERSISTENT) /* MAC label */
1010 #define	PACKET_TAG_PF		(21 | MTAG_PERSISTENT) /* PF/ALTQ information */
1011 #define	PACKET_TAG_RTSOCKFAM			25 /* rtsock sa family */
1012 #define	PACKET_TAG_IPOPTIONS			27 /* Saved IP options */
1013 #define	PACKET_TAG_CARP				28 /* CARP info */
1014 #define	PACKET_TAG_IPSEC_NAT_T_PORTS		29 /* two uint16_t */
1015 #define	PACKET_TAG_ND_OUTGOING			30 /* ND outgoing */
1016 
1017 /* Specific cookies and tags. */
1018 
1019 /* Packet tag routines. */
1020 struct m_tag	*m_tag_alloc(u_int32_t, int, int, int);
1021 void		 m_tag_delete(struct mbuf *, struct m_tag *);
1022 void		 m_tag_delete_chain(struct mbuf *, struct m_tag *);
1023 void		 m_tag_free_default(struct m_tag *);
1024 struct m_tag	*m_tag_locate(struct mbuf *, u_int32_t, int, struct m_tag *);
1025 struct m_tag	*m_tag_copy(struct m_tag *, int);
1026 int		 m_tag_copy_chain(struct mbuf *, const struct mbuf *, int);
1027 void		 m_tag_delete_nonpersistent(struct mbuf *);
1028 
1029 /*
1030  * Initialize the list of tags associated with an mbuf.
1031  */
1032 static __inline void
1033 m_tag_init(struct mbuf *m)
1034 {
1035 
1036 	SLIST_INIT(&m->m_pkthdr.tags);
1037 }
1038 
1039 /*
1040  * Set up the contents of a tag.  Note that this does not fill in the free
1041  * method; the caller is expected to do that.
1042  *
1043  * XXX probably should be called m_tag_init, but that was already taken.
1044  */
1045 static __inline void
1046 m_tag_setup(struct m_tag *t, u_int32_t cookie, int type, int len)
1047 {
1048 
1049 	t->m_tag_id = type;
1050 	t->m_tag_len = len;
1051 	t->m_tag_cookie = cookie;
1052 }
1053 
1054 /*
1055  * Reclaim resources associated with a tag.
1056  */
1057 static __inline void
1058 m_tag_free(struct m_tag *t)
1059 {
1060 
1061 	(*t->m_tag_free)(t);
1062 }
1063 
1064 /*
1065  * Return the first tag associated with an mbuf.
1066  */
1067 static __inline struct m_tag *
1068 m_tag_first(struct mbuf *m)
1069 {
1070 
1071 	return (SLIST_FIRST(&m->m_pkthdr.tags));
1072 }
1073 
1074 /*
1075  * Return the next tag in the list of tags associated with an mbuf.
1076  */
1077 static __inline struct m_tag *
1078 m_tag_next(struct mbuf *m __unused, struct m_tag *t)
1079 {
1080 
1081 	return (SLIST_NEXT(t, m_tag_link));
1082 }
1083 
1084 /*
1085  * Prepend a tag to the list of tags associated with an mbuf.
1086  */
1087 static __inline void
1088 m_tag_prepend(struct mbuf *m, struct m_tag *t)
1089 {
1090 
1091 	SLIST_INSERT_HEAD(&m->m_pkthdr.tags, t, m_tag_link);
1092 }
1093 
1094 /*
1095  * Unlink a tag from the list of tags associated with an mbuf.
1096  */
1097 static __inline void
1098 m_tag_unlink(struct mbuf *m, struct m_tag *t)
1099 {
1100 
1101 	SLIST_REMOVE(&m->m_pkthdr.tags, t, m_tag, m_tag_link);
1102 }
1103 
1104 /* These are for OpenBSD compatibility. */
1105 #define	MTAG_ABI_COMPAT		0		/* compatibility ABI */
1106 
1107 static __inline struct m_tag *
1108 m_tag_get(int type, int length, int wait)
1109 {
1110 	return (m_tag_alloc(MTAG_ABI_COMPAT, type, length, wait));
1111 }
1112 
1113 static __inline struct m_tag *
1114 m_tag_find(struct mbuf *m, int type, struct m_tag *start)
1115 {
1116 	return (SLIST_EMPTY(&m->m_pkthdr.tags) ? (struct m_tag *)NULL :
1117 	    m_tag_locate(m, MTAG_ABI_COMPAT, type, start));
1118 }
1119 
1120 static __inline struct mbuf *
1121 m_free(struct mbuf *m)
1122 {
1123 	struct mbuf *n = m->m_next;
1124 
1125 	if ((m->m_flags & (M_PKTHDR|M_NOFREE)) == (M_PKTHDR|M_NOFREE))
1126 		m_tag_delete_chain(m, NULL);
1127 	if (m->m_flags & M_EXT)
1128 		mb_free_ext(m);
1129 	else if ((m->m_flags & M_NOFREE) == 0)
1130 		uma_zfree(zone_mbuf, m);
1131 	return (n);
1132 }
1133 
1134 static __inline int
1135 rt_m_getfib(struct mbuf *m)
1136 {
1137 	KASSERT(m->m_flags & M_PKTHDR , ("Attempt to get FIB from non header mbuf."));
1138 	return (m->m_pkthdr.fibnum);
1139 }
1140 
1141 #define M_GETFIB(_m)   rt_m_getfib(_m)
1142 
1143 #define M_SETFIB(_m, _fib) do {						\
1144         KASSERT((_m)->m_flags & M_PKTHDR, ("Attempt to set FIB on non header mbuf."));	\
1145 	((_m)->m_pkthdr.fibnum) = (_fib);				\
1146 } while (0)
1147 
1148 /* flags passed as first argument for "m_ether_tcpip_hash()" */
1149 #define	MBUF_HASHFLAG_L2	(1 << 2)
1150 #define	MBUF_HASHFLAG_L3	(1 << 3)
1151 #define	MBUF_HASHFLAG_L4	(1 << 4)
1152 
1153 /* mbuf hashing helper routines */
1154 uint32_t	m_ether_tcpip_hash_init(void);
1155 uint32_t	m_ether_tcpip_hash(const uint32_t, const struct mbuf *, const uint32_t);
1156 
1157 #ifdef MBUF_PROFILING
1158  void m_profile(struct mbuf *m);
1159  #define M_PROFILE(m) m_profile(m)
1160 #else
1161  #define M_PROFILE(m)
1162 #endif
1163 
1164 struct mbufq {
1165 	STAILQ_HEAD(, mbuf)	mq_head;
1166 	int			mq_len;
1167 	int			mq_maxlen;
1168 };
1169 
1170 static inline void
1171 mbufq_init(struct mbufq *mq, int maxlen)
1172 {
1173 
1174 	STAILQ_INIT(&mq->mq_head);
1175 	mq->mq_maxlen = maxlen;
1176 	mq->mq_len = 0;
1177 }
1178 
1179 static inline struct mbuf *
1180 mbufq_flush(struct mbufq *mq)
1181 {
1182 	struct mbuf *m;
1183 
1184 	m = STAILQ_FIRST(&mq->mq_head);
1185 	STAILQ_INIT(&mq->mq_head);
1186 	mq->mq_len = 0;
1187 	return (m);
1188 }
1189 
1190 static inline void
1191 mbufq_drain(struct mbufq *mq)
1192 {
1193 	struct mbuf *m, *n;
1194 
1195 	n = mbufq_flush(mq);
1196 	while ((m = n) != NULL) {
1197 		n = STAILQ_NEXT(m, m_stailqpkt);
1198 		m_freem(m);
1199 	}
1200 }
1201 
1202 static inline struct mbuf *
1203 mbufq_first(const struct mbufq *mq)
1204 {
1205 
1206 	return (STAILQ_FIRST(&mq->mq_head));
1207 }
1208 
1209 static inline struct mbuf *
1210 mbufq_last(const struct mbufq *mq)
1211 {
1212 
1213 	return (STAILQ_LAST(&mq->mq_head, mbuf, m_stailqpkt));
1214 }
1215 
1216 static inline int
1217 mbufq_full(const struct mbufq *mq)
1218 {
1219 
1220 	return (mq->mq_len >= mq->mq_maxlen);
1221 }
1222 
1223 static inline int
1224 mbufq_len(const struct mbufq *mq)
1225 {
1226 
1227 	return (mq->mq_len);
1228 }
1229 
1230 static inline int
1231 mbufq_enqueue(struct mbufq *mq, struct mbuf *m)
1232 {
1233 
1234 	if (mbufq_full(mq))
1235 		return (ENOBUFS);
1236 	STAILQ_INSERT_TAIL(&mq->mq_head, m, m_stailqpkt);
1237 	mq->mq_len++;
1238 	return (0);
1239 }
1240 
1241 static inline struct mbuf *
1242 mbufq_dequeue(struct mbufq *mq)
1243 {
1244 	struct mbuf *m;
1245 
1246 	m = STAILQ_FIRST(&mq->mq_head);
1247 	if (m) {
1248 		STAILQ_REMOVE_HEAD(&mq->mq_head, m_stailqpkt);
1249 		m->m_nextpkt = NULL;
1250 		mq->mq_len--;
1251 	}
1252 	return (m);
1253 }
1254 
1255 static inline void
1256 mbufq_prepend(struct mbufq *mq, struct mbuf *m)
1257 {
1258 
1259 	STAILQ_INSERT_HEAD(&mq->mq_head, m, m_stailqpkt);
1260 	mq->mq_len++;
1261 }
1262 #endif /* _KERNEL */
1263 #endif /* !_SYS_MBUF_H_ */
1264