xref: /freebsd/sys/sys/mbuf.h (revision ca987d4641cdcd7f27e153db17c5bf064934faf5)
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 #ifdef _KERNEL
48 #include <sys/sdt.h>
49 
50 #define	MBUF_PROBE1(probe, arg0)					\
51 	SDT_PROBE1(sdt, , , probe, arg0)
52 #define	MBUF_PROBE2(probe, arg0, arg1)					\
53 	SDT_PROBE2(sdt, , , probe, arg0, arg1)
54 #define	MBUF_PROBE3(probe, arg0, arg1, arg2)				\
55 	SDT_PROBE3(sdt, , , probe, arg0, arg1, arg2)
56 #define	MBUF_PROBE4(probe, arg0, arg1, arg2, arg3)			\
57 	SDT_PROBE4(sdt, , , probe, arg0, arg1, arg2, arg3)
58 #define	MBUF_PROBE5(probe, arg0, arg1, arg2, arg3, arg4)		\
59 	SDT_PROBE5(sdt, , , probe, arg0, arg1, arg2, arg3, arg4)
60 
61 SDT_PROBE_DECLARE(sdt, , , m__init);
62 SDT_PROBE_DECLARE(sdt, , , m__gethdr);
63 SDT_PROBE_DECLARE(sdt, , , m__get);
64 SDT_PROBE_DECLARE(sdt, , , m__getcl);
65 SDT_PROBE_DECLARE(sdt, , , m__clget);
66 SDT_PROBE_DECLARE(sdt, , , m__cljget);
67 SDT_PROBE_DECLARE(sdt, , , m__cljset);
68 SDT_PROBE_DECLARE(sdt, , , m__free);
69 SDT_PROBE_DECLARE(sdt, , , m__freem);
70 
71 #endif /* _KERNEL */
72 
73 /*
74  * Mbufs are of a single size, MSIZE (sys/param.h), which includes overhead.
75  * An mbuf may add a single "mbuf cluster" of size MCLBYTES (also in
76  * sys/param.h), which has no additional overhead and is used instead of the
77  * internal data area; this is done when at least MINCLSIZE of data must be
78  * stored.  Additionally, it is possible to allocate a separate buffer
79  * externally and attach it to the mbuf in a way similar to that of mbuf
80  * clusters.
81  *
82  * NB: These calculation do not take actual compiler-induced alignment and
83  * padding inside the complete struct mbuf into account.  Appropriate
84  * attention is required when changing members of struct mbuf.
85  *
86  * MLEN is data length in a normal mbuf.
87  * MHLEN is data length in an mbuf with pktheader.
88  * MINCLSIZE is a smallest amount of data that should be put into cluster.
89  *
90  * Compile-time assertions in uipc_mbuf.c test these values to ensure that
91  * they are sensible.
92  */
93 struct mbuf;
94 #define	MHSIZE		offsetof(struct mbuf, m_dat)
95 #define	MPKTHSIZE	offsetof(struct mbuf, m_pktdat)
96 #define	MLEN		((int)(MSIZE - MHSIZE))
97 #define	MHLEN		((int)(MSIZE - MPKTHSIZE))
98 #define	MINCLSIZE	(MHLEN + 1)
99 
100 #ifdef _KERNEL
101 /*-
102  * Macro for type conversion: convert mbuf pointer to data pointer of correct
103  * type:
104  *
105  * mtod(m, t)	-- Convert mbuf pointer to data pointer of correct type.
106  * mtodo(m, o) -- Same as above but with offset 'o' into data.
107  */
108 #define	mtod(m, t)	((t)((m)->m_data))
109 #define	mtodo(m, o)	((void *)(((m)->m_data) + (o)))
110 
111 /*
112  * Argument structure passed to UMA routines during mbuf and packet
113  * allocations.
114  */
115 struct mb_args {
116 	int	flags;	/* Flags for mbuf being allocated */
117 	short	type;	/* Type of mbuf being allocated */
118 };
119 #endif /* _KERNEL */
120 
121 /*
122  * Packet tag structure (see below for details).
123  */
124 struct m_tag {
125 	SLIST_ENTRY(m_tag)	m_tag_link;	/* List of packet tags */
126 	u_int16_t		m_tag_id;	/* Tag ID */
127 	u_int16_t		m_tag_len;	/* Length of data */
128 	u_int32_t		m_tag_cookie;	/* ABI/Module ID */
129 	void			(*m_tag_free)(struct m_tag *);
130 };
131 
132 /*
133  * Static network interface owned tag.
134  * Allocated through ifp->if_snd_tag_alloc().
135  */
136 struct m_snd_tag {
137 	struct ifnet *ifp;		/* network interface tag belongs to */
138 };
139 
140 /*
141  * Record/packet header in first mbuf of chain; valid only if M_PKTHDR is set.
142  * Size ILP32: 48
143  *	 LP64: 56
144  * Compile-time assertions in uipc_mbuf.c test these values to ensure that
145  * they are correct.
146  */
147 struct pkthdr {
148 	union {
149 		struct m_snd_tag *snd_tag;	/* send tag, if any */
150 		struct ifnet	*rcvif;		/* rcv interface */
151 	};
152 	SLIST_HEAD(packet_tags, m_tag) tags; /* list of packet tags */
153 	int32_t		 len;		/* total packet length */
154 
155 	/* Layer crossing persistent information. */
156 	uint32_t	 flowid;	/* packet's 4-tuple system */
157 	uint32_t	 csum_flags;	/* checksum and offload features */
158 	uint16_t	 fibnum;	/* this packet should use this fib */
159 	uint8_t		 cosqos;	/* class/quality of service */
160 	uint8_t		 rsstype;	/* hash type */
161 	union {
162 		uint64_t	rcv_tstmp;	/* timestamp in ns */
163 		struct {
164 			uint8_t		 l2hlen;	/* layer 2 hdr len */
165 			uint8_t		 l3hlen;	/* layer 3 hdr len */
166 			uint8_t		 l4hlen;	/* layer 4 hdr len */
167 			uint8_t		 l5hlen;	/* layer 5 hdr len */
168 			uint32_t	 spare;
169 		};
170 	};
171 	union {
172 		uint8_t  eight[8];
173 		uint16_t sixteen[4];
174 		uint32_t thirtytwo[2];
175 		uint64_t sixtyfour[1];
176 		uintptr_t unintptr[1];
177 		void	*ptr;
178 	} PH_per;
179 
180 	/* Layer specific non-persistent local storage for reassembly, etc. */
181 	union {
182 		uint8_t  eight[8];
183 		uint16_t sixteen[4];
184 		uint32_t thirtytwo[2];
185 		uint64_t sixtyfour[1];
186 		uintptr_t unintptr[1];
187 		void 	*ptr;
188 	} PH_loc;
189 };
190 #define	ether_vtag	PH_per.sixteen[0]
191 #define	PH_vt		PH_per
192 #define	vt_nrecs	sixteen[0]
193 #define	tso_segsz	PH_per.sixteen[1]
194 #define	lro_nsegs	tso_segsz
195 #define	csum_phsum	PH_per.sixteen[2]
196 #define	csum_data	PH_per.thirtytwo[1]
197 
198 /*
199  * Description of external storage mapped into mbuf; valid only if M_EXT is
200  * set.
201  * Size ILP32: 28
202  *	 LP64: 48
203  * Compile-time assertions in uipc_mbuf.c test these values to ensure that
204  * they are correct.
205  */
206 typedef	void m_ext_free_t(struct mbuf *);
207 struct m_ext {
208 	union {
209 		/*
210 		 * If EXT_FLAG_EMBREF is set, then we use refcount in the
211 		 * mbuf, the 'ext_count' member.  Otherwise, we have a
212 		 * shadow copy and we use pointer 'ext_cnt'.  The original
213 		 * mbuf is responsible to carry the pointer to free routine
214 		 * and its arguments.  They aren't copied into shadows in
215 		 * mb_dupcl() to avoid dereferencing next cachelines.
216 		 */
217 		volatile u_int	 ext_count;
218 		volatile u_int	*ext_cnt;
219 	};
220 	char		*ext_buf;	/* start of buffer */
221 	uint32_t	 ext_size;	/* size of buffer, for ext_free */
222 	uint32_t	 ext_type:8,	/* type of external storage */
223 			 ext_flags:24;	/* external storage mbuf flags */
224 	/*
225 	 * Fields below store the free context for the external storage.
226 	 * They are valid only in the refcount carrying mbuf, the one with
227 	 * EXT_FLAG_EMBREF flag, with exclusion for EXT_EXTREF type, where
228 	 * the free context is copied into all mbufs that use same external
229 	 * storage.
230 	 */
231 #define	m_ext_copylen	offsetof(struct m_ext, ext_free)
232 	m_ext_free_t	*ext_free;	/* free routine if not the usual */
233 	void		*ext_arg1;	/* optional argument pointer */
234 	void		*ext_arg2;	/* optional argument pointer */
235 };
236 
237 /*
238  * The core of the mbuf object along with some shortcut defines for practical
239  * purposes.
240  */
241 struct mbuf {
242 	/*
243 	 * Header present at the beginning of every mbuf.
244 	 * Size ILP32: 24
245 	 *      LP64: 32
246 	 * Compile-time assertions in uipc_mbuf.c test these values to ensure
247 	 * that they are correct.
248 	 */
249 	union {	/* next buffer in chain */
250 		struct mbuf		*m_next;
251 		SLIST_ENTRY(mbuf)	m_slist;
252 		STAILQ_ENTRY(mbuf)	m_stailq;
253 	};
254 	union {	/* next chain in queue/record */
255 		struct mbuf		*m_nextpkt;
256 		SLIST_ENTRY(mbuf)	m_slistpkt;
257 		STAILQ_ENTRY(mbuf)	m_stailqpkt;
258 	};
259 	caddr_t		 m_data;	/* location of data */
260 	int32_t		 m_len;		/* amount of data in this mbuf */
261 	uint32_t	 m_type:8,	/* type of data in this mbuf */
262 			 m_flags:24;	/* flags; see below */
263 #if !defined(__LP64__)
264 	uint32_t	 m_pad;		/* pad for 64bit alignment */
265 #endif
266 
267 	/*
268 	 * A set of optional headers (packet header, external storage header)
269 	 * and internal data storage.  Historically, these arrays were sized
270 	 * to MHLEN (space left after a packet header) and MLEN (space left
271 	 * after only a regular mbuf header); they are now variable size in
272 	 * order to support future work on variable-size mbufs.
273 	 */
274 	union {
275 		struct {
276 			struct pkthdr	m_pkthdr;	/* M_PKTHDR set */
277 			union {
278 				struct m_ext	m_ext;	/* M_EXT set */
279 				char		m_pktdat[0];
280 			};
281 		};
282 		char	m_dat[0];			/* !M_PKTHDR, !M_EXT */
283 	};
284 };
285 
286 /*
287  * mbuf flags of global significance and layer crossing.
288  * Those of only protocol/layer specific significance are to be mapped
289  * to M_PROTO[1-12] and cleared at layer handoff boundaries.
290  * NB: Limited to the lower 24 bits.
291  */
292 #define	M_EXT		0x00000001 /* has associated external storage */
293 #define	M_PKTHDR	0x00000002 /* start of record */
294 #define	M_EOR		0x00000004 /* end of record */
295 #define	M_RDONLY	0x00000008 /* associated data is marked read-only */
296 #define	M_BCAST		0x00000010 /* send/received as link-level broadcast */
297 #define	M_MCAST		0x00000020 /* send/received as link-level multicast */
298 #define	M_PROMISC	0x00000040 /* packet was not for us */
299 #define	M_VLANTAG	0x00000080 /* ether_vtag is valid */
300 #define	M_UNUSED_8	0x00000100 /* --available-- */
301 #define	M_NOFREE	0x00000200 /* do not free mbuf, embedded in cluster */
302 #define	M_TSTMP		0x00000400 /* rcv_tstmp field is valid */
303 #define	M_TSTMP_HPREC	0x00000800 /* rcv_tstmp is high-prec, typically
304 				      hw-stamped on port (useful for IEEE 1588
305 				      and 802.1AS) */
306 
307 #define	M_PROTO1	0x00001000 /* protocol-specific */
308 #define	M_PROTO2	0x00002000 /* protocol-specific */
309 #define	M_PROTO3	0x00004000 /* protocol-specific */
310 #define	M_PROTO4	0x00008000 /* protocol-specific */
311 #define	M_PROTO5	0x00010000 /* protocol-specific */
312 #define	M_PROTO6	0x00020000 /* protocol-specific */
313 #define	M_PROTO7	0x00040000 /* protocol-specific */
314 #define	M_PROTO8	0x00080000 /* protocol-specific */
315 #define	M_PROTO9	0x00100000 /* protocol-specific */
316 #define	M_PROTO10	0x00200000 /* protocol-specific */
317 #define	M_PROTO11	0x00400000 /* protocol-specific */
318 #define	M_PROTO12	0x00800000 /* protocol-specific */
319 
320 #define MB_DTOR_SKIP	0x1	/* don't pollute the cache by touching a freed mbuf */
321 
322 /*
323  * Flags to purge when crossing layers.
324  */
325 #define	M_PROTOFLAGS \
326     (M_PROTO1|M_PROTO2|M_PROTO3|M_PROTO4|M_PROTO5|M_PROTO6|M_PROTO7|M_PROTO8|\
327      M_PROTO9|M_PROTO10|M_PROTO11|M_PROTO12)
328 
329 /*
330  * Flags preserved when copying m_pkthdr.
331  */
332 #define M_COPYFLAGS \
333     (M_PKTHDR|M_EOR|M_RDONLY|M_BCAST|M_MCAST|M_PROMISC|M_VLANTAG|M_TSTMP| \
334      M_TSTMP_HPREC|M_PROTOFLAGS)
335 
336 /*
337  * Mbuf flag description for use with printf(9) %b identifier.
338  */
339 #define	M_FLAG_BITS \
340     "\20\1M_EXT\2M_PKTHDR\3M_EOR\4M_RDONLY\5M_BCAST\6M_MCAST" \
341     "\7M_PROMISC\10M_VLANTAG\13M_TSTMP\14M_TSTMP_HPREC"
342 #define	M_FLAG_PROTOBITS \
343     "\15M_PROTO1\16M_PROTO2\17M_PROTO3\20M_PROTO4\21M_PROTO5" \
344     "\22M_PROTO6\23M_PROTO7\24M_PROTO8\25M_PROTO9\26M_PROTO10" \
345     "\27M_PROTO11\30M_PROTO12"
346 #define	M_FLAG_PRINTF (M_FLAG_BITS M_FLAG_PROTOBITS)
347 
348 /*
349  * Network interface cards are able to hash protocol fields (such as IPv4
350  * addresses and TCP port numbers) classify packets into flows.  These flows
351  * can then be used to maintain ordering while delivering packets to the OS
352  * via parallel input queues, as well as to provide a stateless affinity
353  * model.  NIC drivers can pass up the hash via m->m_pkthdr.flowid, and set
354  * m_flag fields to indicate how the hash should be interpreted by the
355  * network stack.
356  *
357  * Most NICs support RSS, which provides ordering and explicit affinity, and
358  * use the hash m_flag bits to indicate what header fields were covered by
359  * the hash.  M_HASHTYPE_OPAQUE and M_HASHTYPE_OPAQUE_HASH can be set by non-
360  * RSS cards or configurations that provide an opaque flow identifier, allowing
361  * for ordering and distribution without explicit affinity.  Additionally,
362  * M_HASHTYPE_OPAQUE_HASH indicates that the flow identifier has hash
363  * properties.
364  *
365  * The meaning of the IPV6_EX suffix:
366  * "o  Home address from the home address option in the IPv6 destination
367  *     options header.  If the extension header is not present, use the Source
368  *     IPv6 Address.
369  *  o  IPv6 address that is contained in the Routing-Header-Type-2 from the
370  *     associated extension header.  If the extension header is not present,
371  *     use the Destination IPv6 Address."
372  * Quoted from:
373  * https://docs.microsoft.com/en-us/windows-hardware/drivers/network/rss-hashing-types#ndishashipv6ex
374  */
375 #define	M_HASHTYPE_HASHPROP		0x80	/* has hash properties */
376 #define	M_HASHTYPE_HASH(t)		(M_HASHTYPE_HASHPROP | (t))
377 /* Microsoft RSS standard hash types */
378 #define	M_HASHTYPE_NONE			0
379 #define	M_HASHTYPE_RSS_IPV4		M_HASHTYPE_HASH(1) /* IPv4 2-tuple */
380 #define	M_HASHTYPE_RSS_TCP_IPV4		M_HASHTYPE_HASH(2) /* TCPv4 4-tuple */
381 #define	M_HASHTYPE_RSS_IPV6		M_HASHTYPE_HASH(3) /* IPv6 2-tuple */
382 #define	M_HASHTYPE_RSS_TCP_IPV6		M_HASHTYPE_HASH(4) /* TCPv6 4-tuple */
383 #define	M_HASHTYPE_RSS_IPV6_EX		M_HASHTYPE_HASH(5) /* IPv6 2-tuple +
384 							    * ext hdrs */
385 #define	M_HASHTYPE_RSS_TCP_IPV6_EX	M_HASHTYPE_HASH(6) /* TCPv6 4-tuple +
386 							    * ext hdrs */
387 #define	M_HASHTYPE_RSS_UDP_IPV4		M_HASHTYPE_HASH(7) /* IPv4 UDP 4-tuple*/
388 #define	M_HASHTYPE_RSS_UDP_IPV6		M_HASHTYPE_HASH(9) /* IPv6 UDP 4-tuple*/
389 #define	M_HASHTYPE_RSS_UDP_IPV6_EX	M_HASHTYPE_HASH(10)/* IPv6 UDP 4-tuple +
390 							    * ext hdrs */
391 
392 #define	M_HASHTYPE_OPAQUE		63	/* ordering, not affinity */
393 #define	M_HASHTYPE_OPAQUE_HASH		M_HASHTYPE_HASH(M_HASHTYPE_OPAQUE)
394 						/* ordering+hash, not affinity*/
395 
396 #define	M_HASHTYPE_CLEAR(m)	((m)->m_pkthdr.rsstype = 0)
397 #define	M_HASHTYPE_GET(m)	((m)->m_pkthdr.rsstype)
398 #define	M_HASHTYPE_SET(m, v)	((m)->m_pkthdr.rsstype = (v))
399 #define	M_HASHTYPE_TEST(m, v)	(M_HASHTYPE_GET(m) == (v))
400 #define	M_HASHTYPE_ISHASH(m)	(M_HASHTYPE_GET(m) & M_HASHTYPE_HASHPROP)
401 
402 /*
403  * COS/QOS class and quality of service tags.
404  * It uses DSCP code points as base.
405  */
406 #define	QOS_DSCP_CS0		0x00
407 #define	QOS_DSCP_DEF		QOS_DSCP_CS0
408 #define	QOS_DSCP_CS1		0x20
409 #define	QOS_DSCP_AF11		0x28
410 #define	QOS_DSCP_AF12		0x30
411 #define	QOS_DSCP_AF13		0x38
412 #define	QOS_DSCP_CS2		0x40
413 #define	QOS_DSCP_AF21		0x48
414 #define	QOS_DSCP_AF22		0x50
415 #define	QOS_DSCP_AF23		0x58
416 #define	QOS_DSCP_CS3		0x60
417 #define	QOS_DSCP_AF31		0x68
418 #define	QOS_DSCP_AF32		0x70
419 #define	QOS_DSCP_AF33		0x78
420 #define	QOS_DSCP_CS4		0x80
421 #define	QOS_DSCP_AF41		0x88
422 #define	QOS_DSCP_AF42		0x90
423 #define	QOS_DSCP_AF43		0x98
424 #define	QOS_DSCP_CS5		0xa0
425 #define	QOS_DSCP_EF		0xb8
426 #define	QOS_DSCP_CS6		0xc0
427 #define	QOS_DSCP_CS7		0xe0
428 
429 /*
430  * External mbuf storage buffer types.
431  */
432 #define	EXT_CLUSTER	1	/* mbuf cluster */
433 #define	EXT_SFBUF	2	/* sendfile(2)'s sf_buf */
434 #define	EXT_JUMBOP	3	/* jumbo cluster page sized */
435 #define	EXT_JUMBO9	4	/* jumbo cluster 9216 bytes */
436 #define	EXT_JUMBO16	5	/* jumbo cluster 16184 bytes */
437 #define	EXT_PACKET	6	/* mbuf+cluster from packet zone */
438 #define	EXT_MBUF	7	/* external mbuf reference */
439 
440 #define	EXT_VENDOR1	224	/* for vendor-internal use */
441 #define	EXT_VENDOR2	225	/* for vendor-internal use */
442 #define	EXT_VENDOR3	226	/* for vendor-internal use */
443 #define	EXT_VENDOR4	227	/* for vendor-internal use */
444 
445 #define	EXT_EXP1	244	/* for experimental use */
446 #define	EXT_EXP2	245	/* for experimental use */
447 #define	EXT_EXP3	246	/* for experimental use */
448 #define	EXT_EXP4	247	/* for experimental use */
449 
450 #define	EXT_NET_DRV	252	/* custom ext_buf provided by net driver(s) */
451 #define	EXT_MOD_TYPE	253	/* custom module's ext_buf type */
452 #define	EXT_DISPOSABLE	254	/* can throw this buffer away w/page flipping */
453 #define	EXT_EXTREF	255	/* has externally maintained ext_cnt ptr */
454 
455 /*
456  * Flags for external mbuf buffer types.
457  * NB: limited to the lower 24 bits.
458  */
459 #define	EXT_FLAG_EMBREF		0x000001	/* embedded ext_count */
460 #define	EXT_FLAG_EXTREF		0x000002	/* external ext_cnt, notyet */
461 
462 #define	EXT_FLAG_NOFREE		0x000010	/* don't free mbuf to pool, notyet */
463 
464 #define	EXT_FLAG_VENDOR1	0x010000	/* These flags are vendor */
465 #define	EXT_FLAG_VENDOR2	0x020000	/* or submodule specific, */
466 #define	EXT_FLAG_VENDOR3	0x040000	/* not used by mbuf code. */
467 #define	EXT_FLAG_VENDOR4	0x080000	/* Set/read by submodule. */
468 
469 #define	EXT_FLAG_EXP1		0x100000	/* for experimental use */
470 #define	EXT_FLAG_EXP2		0x200000	/* for experimental use */
471 #define	EXT_FLAG_EXP3		0x400000	/* for experimental use */
472 #define	EXT_FLAG_EXP4		0x800000	/* for experimental use */
473 
474 /*
475  * EXT flag description for use with printf(9) %b identifier.
476  */
477 #define	EXT_FLAG_BITS \
478     "\20\1EXT_FLAG_EMBREF\2EXT_FLAG_EXTREF\5EXT_FLAG_NOFREE" \
479     "\21EXT_FLAG_VENDOR1\22EXT_FLAG_VENDOR2\23EXT_FLAG_VENDOR3" \
480     "\24EXT_FLAG_VENDOR4\25EXT_FLAG_EXP1\26EXT_FLAG_EXP2\27EXT_FLAG_EXP3" \
481     "\30EXT_FLAG_EXP4"
482 
483 /*
484  * Flags indicating checksum, segmentation and other offload work to be
485  * done, or already done, by hardware or lower layers.  It is split into
486  * separate inbound and outbound flags.
487  *
488  * Outbound flags that are set by upper protocol layers requesting lower
489  * layers, or ideally the hardware, to perform these offloading tasks.
490  * For outbound packets this field and its flags can be directly tested
491  * against ifnet if_hwassist.
492  */
493 #define	CSUM_IP			0x00000001	/* IP header checksum offload */
494 #define	CSUM_IP_UDP		0x00000002	/* UDP checksum offload */
495 #define	CSUM_IP_TCP		0x00000004	/* TCP checksum offload */
496 #define	CSUM_IP_SCTP		0x00000008	/* SCTP checksum offload */
497 #define	CSUM_IP_TSO		0x00000010	/* TCP segmentation offload */
498 #define	CSUM_IP_ISCSI		0x00000020	/* iSCSI checksum offload */
499 
500 #define	CSUM_IP6_UDP		0x00000200	/* UDP checksum offload */
501 #define	CSUM_IP6_TCP		0x00000400	/* TCP checksum offload */
502 #define	CSUM_IP6_SCTP		0x00000800	/* SCTP checksum offload */
503 #define	CSUM_IP6_TSO		0x00001000	/* TCP segmentation offload */
504 #define	CSUM_IP6_ISCSI		0x00002000	/* iSCSI checksum offload */
505 
506 /* Inbound checksum support where the checksum was verified by hardware. */
507 #define	CSUM_L3_CALC		0x01000000	/* calculated layer 3 csum */
508 #define	CSUM_L3_VALID		0x02000000	/* checksum is correct */
509 #define	CSUM_L4_CALC		0x04000000	/* calculated layer 4 csum */
510 #define	CSUM_L4_VALID		0x08000000	/* checksum is correct */
511 #define	CSUM_L5_CALC		0x10000000	/* calculated layer 5 csum */
512 #define	CSUM_L5_VALID		0x20000000	/* checksum is correct */
513 #define	CSUM_COALESCED		0x40000000	/* contains merged segments */
514 
515 /*
516  * CSUM flag description for use with printf(9) %b identifier.
517  */
518 #define	CSUM_BITS \
519     "\20\1CSUM_IP\2CSUM_IP_UDP\3CSUM_IP_TCP\4CSUM_IP_SCTP\5CSUM_IP_TSO" \
520     "\6CSUM_IP_ISCSI" \
521     "\12CSUM_IP6_UDP\13CSUM_IP6_TCP\14CSUM_IP6_SCTP\15CSUM_IP6_TSO" \
522     "\16CSUM_IP6_ISCSI" \
523     "\31CSUM_L3_CALC\32CSUM_L3_VALID\33CSUM_L4_CALC\34CSUM_L4_VALID" \
524     "\35CSUM_L5_CALC\36CSUM_L5_VALID\37CSUM_COALESCED"
525 
526 /* CSUM flags compatibility mappings. */
527 #define	CSUM_IP_CHECKED		CSUM_L3_CALC
528 #define	CSUM_IP_VALID		CSUM_L3_VALID
529 #define	CSUM_DATA_VALID		CSUM_L4_VALID
530 #define	CSUM_PSEUDO_HDR		CSUM_L4_CALC
531 #define	CSUM_SCTP_VALID		CSUM_L4_VALID
532 #define	CSUM_DELAY_DATA		(CSUM_TCP|CSUM_UDP)
533 #define	CSUM_DELAY_IP		CSUM_IP		/* Only v4, no v6 IP hdr csum */
534 #define	CSUM_DELAY_DATA_IPV6	(CSUM_TCP_IPV6|CSUM_UDP_IPV6)
535 #define	CSUM_DATA_VALID_IPV6	CSUM_DATA_VALID
536 #define	CSUM_TCP		CSUM_IP_TCP
537 #define	CSUM_UDP		CSUM_IP_UDP
538 #define	CSUM_SCTP		CSUM_IP_SCTP
539 #define	CSUM_TSO		(CSUM_IP_TSO|CSUM_IP6_TSO)
540 #define	CSUM_UDP_IPV6		CSUM_IP6_UDP
541 #define	CSUM_TCP_IPV6		CSUM_IP6_TCP
542 #define	CSUM_SCTP_IPV6		CSUM_IP6_SCTP
543 
544 /*
545  * mbuf types describing the content of the mbuf (including external storage).
546  */
547 #define	MT_NOTMBUF	0	/* USED INTERNALLY ONLY! Object is not mbuf */
548 #define	MT_DATA		1	/* dynamic (data) allocation */
549 #define	MT_HEADER	MT_DATA	/* packet header, use M_PKTHDR instead */
550 
551 #define	MT_VENDOR1	4	/* for vendor-internal use */
552 #define	MT_VENDOR2	5	/* for vendor-internal use */
553 #define	MT_VENDOR3	6	/* for vendor-internal use */
554 #define	MT_VENDOR4	7	/* for vendor-internal use */
555 
556 #define	MT_SONAME	8	/* socket name */
557 
558 #define	MT_EXP1		9	/* for experimental use */
559 #define	MT_EXP2		10	/* for experimental use */
560 #define	MT_EXP3		11	/* for experimental use */
561 #define	MT_EXP4		12	/* for experimental use */
562 
563 #define	MT_CONTROL	14	/* extra-data protocol message */
564 #define	MT_OOBDATA	15	/* expedited data  */
565 #define	MT_NTYPES	16	/* number of mbuf types for mbtypes[] */
566 
567 #define	MT_NOINIT	255	/* Not a type but a flag to allocate
568 				   a non-initialized mbuf */
569 
570 /*
571  * String names of mbuf-related UMA(9) and malloc(9) types.  Exposed to
572  * !_KERNEL so that monitoring tools can look up the zones with
573  * libmemstat(3).
574  */
575 #define	MBUF_MEM_NAME		"mbuf"
576 #define	MBUF_CLUSTER_MEM_NAME	"mbuf_cluster"
577 #define	MBUF_PACKET_MEM_NAME	"mbuf_packet"
578 #define	MBUF_JUMBOP_MEM_NAME	"mbuf_jumbo_page"
579 #define	MBUF_JUMBO9_MEM_NAME	"mbuf_jumbo_9k"
580 #define	MBUF_JUMBO16_MEM_NAME	"mbuf_jumbo_16k"
581 #define	MBUF_TAG_MEM_NAME	"mbuf_tag"
582 #define	MBUF_EXTREFCNT_MEM_NAME	"mbuf_ext_refcnt"
583 
584 #ifdef _KERNEL
585 
586 #ifdef WITNESS
587 #define	MBUF_CHECKSLEEP(how) do {					\
588 	if (how == M_WAITOK)						\
589 		WITNESS_WARN(WARN_GIANTOK | WARN_SLEEPOK, NULL,		\
590 		    "Sleeping in \"%s\"", __func__);			\
591 } while (0)
592 #else
593 #define	MBUF_CHECKSLEEP(how)
594 #endif
595 
596 /*
597  * Network buffer allocation API
598  *
599  * The rest of it is defined in kern/kern_mbuf.c
600  */
601 extern uma_zone_t	zone_mbuf;
602 extern uma_zone_t	zone_clust;
603 extern uma_zone_t	zone_pack;
604 extern uma_zone_t	zone_jumbop;
605 extern uma_zone_t	zone_jumbo9;
606 extern uma_zone_t	zone_jumbo16;
607 
608 void		 mb_dupcl(struct mbuf *, struct mbuf *);
609 void		 mb_free_ext(struct mbuf *);
610 void		 m_adj(struct mbuf *, int);
611 int		 m_apply(struct mbuf *, int, int,
612 		    int (*)(void *, void *, u_int), void *);
613 int		 m_append(struct mbuf *, int, c_caddr_t);
614 void		 m_cat(struct mbuf *, struct mbuf *);
615 void		 m_catpkt(struct mbuf *, struct mbuf *);
616 int		 m_clget(struct mbuf *m, int how);
617 void 		*m_cljget(struct mbuf *m, int how, int size);
618 struct mbuf	*m_collapse(struct mbuf *, int, int);
619 void		 m_copyback(struct mbuf *, int, int, c_caddr_t);
620 void		 m_copydata(const struct mbuf *, int, int, caddr_t);
621 struct mbuf	*m_copym(struct mbuf *, int, int, int);
622 struct mbuf	*m_copypacket(struct mbuf *, int);
623 void		 m_copy_pkthdr(struct mbuf *, struct mbuf *);
624 struct mbuf	*m_copyup(struct mbuf *, int, int);
625 struct mbuf	*m_defrag(struct mbuf *, int);
626 void		 m_demote_pkthdr(struct mbuf *);
627 void		 m_demote(struct mbuf *, int, int);
628 struct mbuf	*m_devget(char *, int, int, struct ifnet *,
629 		    void (*)(char *, caddr_t, u_int));
630 struct mbuf	*m_dup(const struct mbuf *, int);
631 int		 m_dup_pkthdr(struct mbuf *, const struct mbuf *, int);
632 void		 m_extadd(struct mbuf *, char *, u_int, m_ext_free_t,
633 		    void *, void *, int, int);
634 u_int		 m_fixhdr(struct mbuf *);
635 struct mbuf	*m_fragment(struct mbuf *, int, int);
636 void		 m_freem(struct mbuf *);
637 struct mbuf	*m_get2(int, int, short, int);
638 struct mbuf	*m_getjcl(int, short, int, int);
639 struct mbuf	*m_getm2(struct mbuf *, int, int, short, int);
640 struct mbuf	*m_getptr(struct mbuf *, int, int *);
641 u_int		 m_length(struct mbuf *, struct mbuf **);
642 int		 m_mbuftouio(struct uio *, const struct mbuf *, int);
643 void		 m_move_pkthdr(struct mbuf *, struct mbuf *);
644 int		 m_pkthdr_init(struct mbuf *, int);
645 struct mbuf	*m_prepend(struct mbuf *, int, int);
646 void		 m_print(const struct mbuf *, int);
647 struct mbuf	*m_pulldown(struct mbuf *, int, int, int *);
648 struct mbuf	*m_pullup(struct mbuf *, int);
649 int		 m_sanity(struct mbuf *, int);
650 struct mbuf	*m_split(struct mbuf *, int, int);
651 struct mbuf	*m_uiotombuf(struct uio *, int, int, int, int);
652 struct mbuf	*m_unshare(struct mbuf *, int);
653 
654 static __inline int
655 m_gettype(int size)
656 {
657 	int type;
658 
659 	switch (size) {
660 	case MSIZE:
661 		type = EXT_MBUF;
662 		break;
663 	case MCLBYTES:
664 		type = EXT_CLUSTER;
665 		break;
666 #if MJUMPAGESIZE != MCLBYTES
667 	case MJUMPAGESIZE:
668 		type = EXT_JUMBOP;
669 		break;
670 #endif
671 	case MJUM9BYTES:
672 		type = EXT_JUMBO9;
673 		break;
674 	case MJUM16BYTES:
675 		type = EXT_JUMBO16;
676 		break;
677 	default:
678 		panic("%s: invalid cluster size %d", __func__, size);
679 	}
680 
681 	return (type);
682 }
683 
684 /*
685  * Associated an external reference counted buffer with an mbuf.
686  */
687 static __inline void
688 m_extaddref(struct mbuf *m, char *buf, u_int size, u_int *ref_cnt,
689     m_ext_free_t freef, void *arg1, void *arg2)
690 {
691 
692 	KASSERT(ref_cnt != NULL, ("%s: ref_cnt not provided", __func__));
693 
694 	atomic_add_int(ref_cnt, 1);
695 	m->m_flags |= M_EXT;
696 	m->m_ext.ext_buf = buf;
697 	m->m_ext.ext_cnt = ref_cnt;
698 	m->m_data = m->m_ext.ext_buf;
699 	m->m_ext.ext_size = size;
700 	m->m_ext.ext_free = freef;
701 	m->m_ext.ext_arg1 = arg1;
702 	m->m_ext.ext_arg2 = arg2;
703 	m->m_ext.ext_type = EXT_EXTREF;
704 	m->m_ext.ext_flags = 0;
705 }
706 
707 static __inline uma_zone_t
708 m_getzone(int size)
709 {
710 	uma_zone_t zone;
711 
712 	switch (size) {
713 	case MCLBYTES:
714 		zone = zone_clust;
715 		break;
716 #if MJUMPAGESIZE != MCLBYTES
717 	case MJUMPAGESIZE:
718 		zone = zone_jumbop;
719 		break;
720 #endif
721 	case MJUM9BYTES:
722 		zone = zone_jumbo9;
723 		break;
724 	case MJUM16BYTES:
725 		zone = zone_jumbo16;
726 		break;
727 	default:
728 		panic("%s: invalid cluster size %d", __func__, size);
729 	}
730 
731 	return (zone);
732 }
733 
734 /*
735  * Initialize an mbuf with linear storage.
736  *
737  * Inline because the consumer text overhead will be roughly the same to
738  * initialize or call a function with this many parameters and M_PKTHDR
739  * should go away with constant propagation for !MGETHDR.
740  */
741 static __inline int
742 m_init(struct mbuf *m, int how, short type, int flags)
743 {
744 	int error;
745 
746 	m->m_next = NULL;
747 	m->m_nextpkt = NULL;
748 	m->m_data = m->m_dat;
749 	m->m_len = 0;
750 	m->m_flags = flags;
751 	m->m_type = type;
752 	if (flags & M_PKTHDR)
753 		error = m_pkthdr_init(m, how);
754 	else
755 		error = 0;
756 
757 	MBUF_PROBE5(m__init, m, how, type, flags, error);
758 	return (error);
759 }
760 
761 static __inline struct mbuf *
762 m_get(int how, short type)
763 {
764 	struct mbuf *m;
765 	struct mb_args args;
766 
767 	args.flags = 0;
768 	args.type = type;
769 	m = uma_zalloc_arg(zone_mbuf, &args, how);
770 	MBUF_PROBE3(m__get, how, type, m);
771 	return (m);
772 }
773 
774 static __inline struct mbuf *
775 m_gethdr(int how, short type)
776 {
777 	struct mbuf *m;
778 	struct mb_args args;
779 
780 	args.flags = M_PKTHDR;
781 	args.type = type;
782 	m = uma_zalloc_arg(zone_mbuf, &args, how);
783 	MBUF_PROBE3(m__gethdr, how, type, m);
784 	return (m);
785 }
786 
787 static __inline struct mbuf *
788 m_getcl(int how, short type, int flags)
789 {
790 	struct mbuf *m;
791 	struct mb_args args;
792 
793 	args.flags = flags;
794 	args.type = type;
795 	m = uma_zalloc_arg(zone_pack, &args, how);
796 	MBUF_PROBE4(m__getcl, how, type, flags, m);
797 	return (m);
798 }
799 
800 /*
801  * XXX: m_cljset() is a dangerous API.  One must attach only a new,
802  * unreferenced cluster to an mbuf(9).  It is not possible to assert
803  * that, so care can be taken only by users of the API.
804  */
805 static __inline void
806 m_cljset(struct mbuf *m, void *cl, int type)
807 {
808 	int size;
809 
810 	switch (type) {
811 	case EXT_CLUSTER:
812 		size = MCLBYTES;
813 		break;
814 #if MJUMPAGESIZE != MCLBYTES
815 	case EXT_JUMBOP:
816 		size = MJUMPAGESIZE;
817 		break;
818 #endif
819 	case EXT_JUMBO9:
820 		size = MJUM9BYTES;
821 		break;
822 	case EXT_JUMBO16:
823 		size = MJUM16BYTES;
824 		break;
825 	default:
826 		panic("%s: unknown cluster type %d", __func__, type);
827 		break;
828 	}
829 
830 	m->m_data = m->m_ext.ext_buf = cl;
831 	m->m_ext.ext_free = m->m_ext.ext_arg1 = m->m_ext.ext_arg2 = NULL;
832 	m->m_ext.ext_size = size;
833 	m->m_ext.ext_type = type;
834 	m->m_ext.ext_flags = EXT_FLAG_EMBREF;
835 	m->m_ext.ext_count = 1;
836 	m->m_flags |= M_EXT;
837 	MBUF_PROBE3(m__cljset, m, cl, type);
838 }
839 
840 static __inline void
841 m_chtype(struct mbuf *m, short new_type)
842 {
843 
844 	m->m_type = new_type;
845 }
846 
847 static __inline void
848 m_clrprotoflags(struct mbuf *m)
849 {
850 
851 	while (m) {
852 		m->m_flags &= ~M_PROTOFLAGS;
853 		m = m->m_next;
854 	}
855 }
856 
857 static __inline struct mbuf *
858 m_last(struct mbuf *m)
859 {
860 
861 	while (m->m_next)
862 		m = m->m_next;
863 	return (m);
864 }
865 
866 static inline u_int
867 m_extrefcnt(struct mbuf *m)
868 {
869 
870 	KASSERT(m->m_flags & M_EXT, ("%s: M_EXT missing", __func__));
871 
872 	return ((m->m_ext.ext_flags & EXT_FLAG_EMBREF) ? m->m_ext.ext_count :
873 	    *m->m_ext.ext_cnt);
874 }
875 
876 /*
877  * mbuf, cluster, and external object allocation macros (for compatibility
878  * purposes).
879  */
880 #define	M_MOVE_PKTHDR(to, from)	m_move_pkthdr((to), (from))
881 #define	MGET(m, how, type)	((m) = m_get((how), (type)))
882 #define	MGETHDR(m, how, type)	((m) = m_gethdr((how), (type)))
883 #define	MCLGET(m, how)		m_clget((m), (how))
884 #define	MEXTADD(m, buf, size, free, arg1, arg2, flags, type)		\
885     m_extadd((m), (char *)(buf), (size), (free), (arg1), (arg2),	\
886     (flags), (type))
887 #define	m_getm(m, len, how, type)					\
888     m_getm2((m), (len), (how), (type), M_PKTHDR)
889 
890 /*
891  * Evaluate TRUE if it's safe to write to the mbuf m's data region (this can
892  * be both the local data payload, or an external buffer area, depending on
893  * whether M_EXT is set).
894  */
895 #define	M_WRITABLE(m)	(!((m)->m_flags & M_RDONLY) &&			\
896 			 (!(((m)->m_flags & M_EXT)) ||			\
897 			 (m_extrefcnt(m) == 1)))
898 
899 /* Check if the supplied mbuf has a packet header, or else panic. */
900 #define	M_ASSERTPKTHDR(m)						\
901 	KASSERT((m) != NULL && (m)->m_flags & M_PKTHDR,			\
902 	    ("%s: no mbuf packet header!", __func__))
903 
904 /*
905  * Ensure that the supplied mbuf is a valid, non-free mbuf.
906  *
907  * XXX: Broken at the moment.  Need some UMA magic to make it work again.
908  */
909 #define	M_ASSERTVALID(m)						\
910 	KASSERT((((struct mbuf *)m)->m_flags & 0) == 0,			\
911 	    ("%s: attempted use of a free mbuf!", __func__))
912 
913 /*
914  * Return the address of the start of the buffer associated with an mbuf,
915  * handling external storage, packet-header mbufs, and regular data mbufs.
916  */
917 #define	M_START(m)							\
918 	(((m)->m_flags & M_EXT) ? (m)->m_ext.ext_buf :			\
919 	 ((m)->m_flags & M_PKTHDR) ? &(m)->m_pktdat[0] :		\
920 	 &(m)->m_dat[0])
921 
922 /*
923  * Return the size of the buffer associated with an mbuf, handling external
924  * storage, packet-header mbufs, and regular data mbufs.
925  */
926 #define	M_SIZE(m)							\
927 	(((m)->m_flags & M_EXT) ? (m)->m_ext.ext_size :			\
928 	 ((m)->m_flags & M_PKTHDR) ? MHLEN :				\
929 	 MLEN)
930 
931 /*
932  * Set the m_data pointer of a newly allocated mbuf to place an object of the
933  * specified size at the end of the mbuf, longword aligned.
934  *
935  * NB: Historically, we had M_ALIGN(), MH_ALIGN(), and MEXT_ALIGN() as
936  * separate macros, each asserting that it was called at the proper moment.
937  * This required callers to themselves test the storage type and call the
938  * right one.  Rather than require callers to be aware of those layout
939  * decisions, we centralize here.
940  */
941 static __inline void
942 m_align(struct mbuf *m, int len)
943 {
944 #ifdef INVARIANTS
945 	const char *msg = "%s: not a virgin mbuf";
946 #endif
947 	int adjust;
948 
949 	KASSERT(m->m_data == M_START(m), (msg, __func__));
950 
951 	adjust = M_SIZE(m) - len;
952 	m->m_data += adjust &~ (sizeof(long)-1);
953 }
954 
955 #define	M_ALIGN(m, len)		m_align(m, len)
956 #define	MH_ALIGN(m, len)	m_align(m, len)
957 #define	MEXT_ALIGN(m, len)	m_align(m, len)
958 
959 /*
960  * Compute the amount of space available before the current start of data in
961  * an mbuf.
962  *
963  * The M_WRITABLE() is a temporary, conservative safety measure: the burden
964  * of checking writability of the mbuf data area rests solely with the caller.
965  *
966  * NB: In previous versions, M_LEADINGSPACE() would only check M_WRITABLE()
967  * for mbufs with external storage.  We now allow mbuf-embedded data to be
968  * read-only as well.
969  */
970 #define	M_LEADINGSPACE(m)						\
971 	(M_WRITABLE(m) ? ((m)->m_data - M_START(m)) : 0)
972 
973 /*
974  * Compute the amount of space available after the end of data in an mbuf.
975  *
976  * The M_WRITABLE() is a temporary, conservative safety measure: the burden
977  * of checking writability of the mbuf data area rests solely with the caller.
978  *
979  * NB: In previous versions, M_TRAILINGSPACE() would only check M_WRITABLE()
980  * for mbufs with external storage.  We now allow mbuf-embedded data to be
981  * read-only as well.
982  */
983 #define	M_TRAILINGSPACE(m)						\
984 	(M_WRITABLE(m) ?						\
985 	    ((M_START(m) + M_SIZE(m)) - ((m)->m_data + (m)->m_len)) : 0)
986 
987 /*
988  * Arrange to prepend space of size plen to mbuf m.  If a new mbuf must be
989  * allocated, how specifies whether to wait.  If the allocation fails, the
990  * original mbuf chain is freed and m is set to NULL.
991  */
992 #define	M_PREPEND(m, plen, how) do {					\
993 	struct mbuf **_mmp = &(m);					\
994 	struct mbuf *_mm = *_mmp;					\
995 	int _mplen = (plen);						\
996 	int __mhow = (how);						\
997 									\
998 	MBUF_CHECKSLEEP(how);						\
999 	if (M_LEADINGSPACE(_mm) >= _mplen) {				\
1000 		_mm->m_data -= _mplen;					\
1001 		_mm->m_len += _mplen;					\
1002 	} else								\
1003 		_mm = m_prepend(_mm, _mplen, __mhow);			\
1004 	if (_mm != NULL && _mm->m_flags & M_PKTHDR)			\
1005 		_mm->m_pkthdr.len += _mplen;				\
1006 	*_mmp = _mm;							\
1007 } while (0)
1008 
1009 /*
1010  * Change mbuf to new type.  This is a relatively expensive operation and
1011  * should be avoided.
1012  */
1013 #define	MCHTYPE(m, t)	m_chtype((m), (t))
1014 
1015 /* Length to m_copy to copy all. */
1016 #define	M_COPYALL	1000000000
1017 
1018 extern int		max_datalen;	/* MHLEN - max_hdr */
1019 extern int		max_hdr;	/* Largest link + protocol header */
1020 extern int		max_linkhdr;	/* Largest link-level header */
1021 extern int		max_protohdr;	/* Largest protocol header */
1022 extern int		nmbclusters;	/* Maximum number of clusters */
1023 
1024 /*-
1025  * Network packets may have annotations attached by affixing a list of
1026  * "packet tags" to the pkthdr structure.  Packet tags are dynamically
1027  * allocated semi-opaque data structures that have a fixed header
1028  * (struct m_tag) that specifies the size of the memory block and a
1029  * <cookie,type> pair that identifies it.  The cookie is a 32-bit unique
1030  * unsigned value used to identify a module or ABI.  By convention this value
1031  * is chosen as the date+time that the module is created, expressed as the
1032  * number of seconds since the epoch (e.g., using date -u +'%s').  The type
1033  * value is an ABI/module-specific value that identifies a particular
1034  * annotation and is private to the module.  For compatibility with systems
1035  * like OpenBSD that define packet tags w/o an ABI/module cookie, the value
1036  * PACKET_ABI_COMPAT is used to implement m_tag_get and m_tag_find
1037  * compatibility shim functions and several tag types are defined below.
1038  * Users that do not require compatibility should use a private cookie value
1039  * so that packet tag-related definitions can be maintained privately.
1040  *
1041  * Note that the packet tag returned by m_tag_alloc has the default memory
1042  * alignment implemented by malloc.  To reference private data one can use a
1043  * construct like:
1044  *
1045  *	struct m_tag *mtag = m_tag_alloc(...);
1046  *	struct foo *p = (struct foo *)(mtag+1);
1047  *
1048  * if the alignment of struct m_tag is sufficient for referencing members of
1049  * struct foo.  Otherwise it is necessary to embed struct m_tag within the
1050  * private data structure to insure proper alignment; e.g.,
1051  *
1052  *	struct foo {
1053  *		struct m_tag	tag;
1054  *		...
1055  *	};
1056  *	struct foo *p = (struct foo *) m_tag_alloc(...);
1057  *	struct m_tag *mtag = &p->tag;
1058  */
1059 
1060 /*
1061  * Persistent tags stay with an mbuf until the mbuf is reclaimed.  Otherwise
1062  * tags are expected to ``vanish'' when they pass through a network
1063  * interface.  For most interfaces this happens normally as the tags are
1064  * reclaimed when the mbuf is free'd.  However in some special cases
1065  * reclaiming must be done manually.  An example is packets that pass through
1066  * the loopback interface.  Also, one must be careful to do this when
1067  * ``turning around'' packets (e.g., icmp_reflect).
1068  *
1069  * To mark a tag persistent bit-or this flag in when defining the tag id.
1070  * The tag will then be treated as described above.
1071  */
1072 #define	MTAG_PERSISTENT				0x800
1073 
1074 #define	PACKET_TAG_NONE				0  /* Nadda */
1075 
1076 /* Packet tags for use with PACKET_ABI_COMPAT. */
1077 #define	PACKET_TAG_IPSEC_IN_DONE		1  /* IPsec applied, in */
1078 #define	PACKET_TAG_IPSEC_OUT_DONE		2  /* IPsec applied, out */
1079 #define	PACKET_TAG_IPSEC_IN_CRYPTO_DONE		3  /* NIC IPsec crypto done */
1080 #define	PACKET_TAG_IPSEC_OUT_CRYPTO_NEEDED	4  /* NIC IPsec crypto req'ed */
1081 #define	PACKET_TAG_IPSEC_IN_COULD_DO_CRYPTO	5  /* NIC notifies IPsec */
1082 #define	PACKET_TAG_IPSEC_PENDING_TDB		6  /* Reminder to do IPsec */
1083 #define	PACKET_TAG_BRIDGE			7  /* Bridge processing done */
1084 #define	PACKET_TAG_GIF				8  /* GIF processing done */
1085 #define	PACKET_TAG_GRE				9  /* GRE processing done */
1086 #define	PACKET_TAG_IN_PACKET_CHECKSUM		10 /* NIC checksumming done */
1087 #define	PACKET_TAG_ENCAP			11 /* Encap.  processing */
1088 #define	PACKET_TAG_IPSEC_SOCKET			12 /* IPSEC socket ref */
1089 #define	PACKET_TAG_IPSEC_HISTORY		13 /* IPSEC history */
1090 #define	PACKET_TAG_IPV6_INPUT			14 /* IPV6 input processing */
1091 #define	PACKET_TAG_DUMMYNET			15 /* dummynet info */
1092 #define	PACKET_TAG_DIVERT			17 /* divert info */
1093 #define	PACKET_TAG_IPFORWARD			18 /* ipforward info */
1094 #define	PACKET_TAG_MACLABEL	(19 | MTAG_PERSISTENT) /* MAC label */
1095 #define	PACKET_TAG_PF		(21 | MTAG_PERSISTENT) /* PF/ALTQ information */
1096 #define	PACKET_TAG_RTSOCKFAM			25 /* rtsock sa family */
1097 #define	PACKET_TAG_IPOPTIONS			27 /* Saved IP options */
1098 #define	PACKET_TAG_CARP				28 /* CARP info */
1099 #define	PACKET_TAG_IPSEC_NAT_T_PORTS		29 /* two uint16_t */
1100 #define	PACKET_TAG_ND_OUTGOING			30 /* ND outgoing */
1101 
1102 /* Specific cookies and tags. */
1103 
1104 /* Packet tag routines. */
1105 struct m_tag	*m_tag_alloc(u_int32_t, int, int, int);
1106 void		 m_tag_delete(struct mbuf *, struct m_tag *);
1107 void		 m_tag_delete_chain(struct mbuf *, struct m_tag *);
1108 void		 m_tag_free_default(struct m_tag *);
1109 struct m_tag	*m_tag_locate(struct mbuf *, u_int32_t, int, struct m_tag *);
1110 struct m_tag	*m_tag_copy(struct m_tag *, int);
1111 int		 m_tag_copy_chain(struct mbuf *, const struct mbuf *, int);
1112 void		 m_tag_delete_nonpersistent(struct mbuf *);
1113 
1114 /*
1115  * Initialize the list of tags associated with an mbuf.
1116  */
1117 static __inline void
1118 m_tag_init(struct mbuf *m)
1119 {
1120 
1121 	SLIST_INIT(&m->m_pkthdr.tags);
1122 }
1123 
1124 /*
1125  * Set up the contents of a tag.  Note that this does not fill in the free
1126  * method; the caller is expected to do that.
1127  *
1128  * XXX probably should be called m_tag_init, but that was already taken.
1129  */
1130 static __inline void
1131 m_tag_setup(struct m_tag *t, u_int32_t cookie, int type, int len)
1132 {
1133 
1134 	t->m_tag_id = type;
1135 	t->m_tag_len = len;
1136 	t->m_tag_cookie = cookie;
1137 }
1138 
1139 /*
1140  * Reclaim resources associated with a tag.
1141  */
1142 static __inline void
1143 m_tag_free(struct m_tag *t)
1144 {
1145 
1146 	(*t->m_tag_free)(t);
1147 }
1148 
1149 /*
1150  * Return the first tag associated with an mbuf.
1151  */
1152 static __inline struct m_tag *
1153 m_tag_first(struct mbuf *m)
1154 {
1155 
1156 	return (SLIST_FIRST(&m->m_pkthdr.tags));
1157 }
1158 
1159 /*
1160  * Return the next tag in the list of tags associated with an mbuf.
1161  */
1162 static __inline struct m_tag *
1163 m_tag_next(struct mbuf *m __unused, struct m_tag *t)
1164 {
1165 
1166 	return (SLIST_NEXT(t, m_tag_link));
1167 }
1168 
1169 /*
1170  * Prepend a tag to the list of tags associated with an mbuf.
1171  */
1172 static __inline void
1173 m_tag_prepend(struct mbuf *m, struct m_tag *t)
1174 {
1175 
1176 	SLIST_INSERT_HEAD(&m->m_pkthdr.tags, t, m_tag_link);
1177 }
1178 
1179 /*
1180  * Unlink a tag from the list of tags associated with an mbuf.
1181  */
1182 static __inline void
1183 m_tag_unlink(struct mbuf *m, struct m_tag *t)
1184 {
1185 
1186 	SLIST_REMOVE(&m->m_pkthdr.tags, t, m_tag, m_tag_link);
1187 }
1188 
1189 /* These are for OpenBSD compatibility. */
1190 #define	MTAG_ABI_COMPAT		0		/* compatibility ABI */
1191 
1192 static __inline struct m_tag *
1193 m_tag_get(int type, int length, int wait)
1194 {
1195 	return (m_tag_alloc(MTAG_ABI_COMPAT, type, length, wait));
1196 }
1197 
1198 static __inline struct m_tag *
1199 m_tag_find(struct mbuf *m, int type, struct m_tag *start)
1200 {
1201 	return (SLIST_EMPTY(&m->m_pkthdr.tags) ? (struct m_tag *)NULL :
1202 	    m_tag_locate(m, MTAG_ABI_COMPAT, type, start));
1203 }
1204 
1205 static __inline struct mbuf *
1206 m_free(struct mbuf *m)
1207 {
1208 	struct mbuf *n = m->m_next;
1209 
1210 	MBUF_PROBE1(m__free, m);
1211 	if ((m->m_flags & (M_PKTHDR|M_NOFREE)) == (M_PKTHDR|M_NOFREE))
1212 		m_tag_delete_chain(m, NULL);
1213 	if (m->m_flags & M_EXT)
1214 		mb_free_ext(m);
1215 	else if ((m->m_flags & M_NOFREE) == 0)
1216 		uma_zfree(zone_mbuf, m);
1217 	return (n);
1218 }
1219 
1220 static __inline int
1221 rt_m_getfib(struct mbuf *m)
1222 {
1223 	KASSERT(m->m_flags & M_PKTHDR , ("Attempt to get FIB from non header mbuf."));
1224 	return (m->m_pkthdr.fibnum);
1225 }
1226 
1227 #define M_GETFIB(_m)   rt_m_getfib(_m)
1228 
1229 #define M_SETFIB(_m, _fib) do {						\
1230         KASSERT((_m)->m_flags & M_PKTHDR, ("Attempt to set FIB on non header mbuf."));	\
1231 	((_m)->m_pkthdr.fibnum) = (_fib);				\
1232 } while (0)
1233 
1234 /* flags passed as first argument for "m_ether_tcpip_hash()" */
1235 #define	MBUF_HASHFLAG_L2	(1 << 2)
1236 #define	MBUF_HASHFLAG_L3	(1 << 3)
1237 #define	MBUF_HASHFLAG_L4	(1 << 4)
1238 
1239 /* mbuf hashing helper routines */
1240 uint32_t	m_ether_tcpip_hash_init(void);
1241 uint32_t	m_ether_tcpip_hash(const uint32_t, const struct mbuf *, const uint32_t);
1242 
1243 #ifdef MBUF_PROFILING
1244  void m_profile(struct mbuf *m);
1245  #define M_PROFILE(m) m_profile(m)
1246 #else
1247  #define M_PROFILE(m)
1248 #endif
1249 
1250 struct mbufq {
1251 	STAILQ_HEAD(, mbuf)	mq_head;
1252 	int			mq_len;
1253 	int			mq_maxlen;
1254 };
1255 
1256 static inline void
1257 mbufq_init(struct mbufq *mq, int maxlen)
1258 {
1259 
1260 	STAILQ_INIT(&mq->mq_head);
1261 	mq->mq_maxlen = maxlen;
1262 	mq->mq_len = 0;
1263 }
1264 
1265 static inline struct mbuf *
1266 mbufq_flush(struct mbufq *mq)
1267 {
1268 	struct mbuf *m;
1269 
1270 	m = STAILQ_FIRST(&mq->mq_head);
1271 	STAILQ_INIT(&mq->mq_head);
1272 	mq->mq_len = 0;
1273 	return (m);
1274 }
1275 
1276 static inline void
1277 mbufq_drain(struct mbufq *mq)
1278 {
1279 	struct mbuf *m, *n;
1280 
1281 	n = mbufq_flush(mq);
1282 	while ((m = n) != NULL) {
1283 		n = STAILQ_NEXT(m, m_stailqpkt);
1284 		m_freem(m);
1285 	}
1286 }
1287 
1288 static inline struct mbuf *
1289 mbufq_first(const struct mbufq *mq)
1290 {
1291 
1292 	return (STAILQ_FIRST(&mq->mq_head));
1293 }
1294 
1295 static inline struct mbuf *
1296 mbufq_last(const struct mbufq *mq)
1297 {
1298 
1299 	return (STAILQ_LAST(&mq->mq_head, mbuf, m_stailqpkt));
1300 }
1301 
1302 static inline int
1303 mbufq_full(const struct mbufq *mq)
1304 {
1305 
1306 	return (mq->mq_len >= mq->mq_maxlen);
1307 }
1308 
1309 static inline int
1310 mbufq_len(const struct mbufq *mq)
1311 {
1312 
1313 	return (mq->mq_len);
1314 }
1315 
1316 static inline int
1317 mbufq_enqueue(struct mbufq *mq, struct mbuf *m)
1318 {
1319 
1320 	if (mbufq_full(mq))
1321 		return (ENOBUFS);
1322 	STAILQ_INSERT_TAIL(&mq->mq_head, m, m_stailqpkt);
1323 	mq->mq_len++;
1324 	return (0);
1325 }
1326 
1327 static inline struct mbuf *
1328 mbufq_dequeue(struct mbufq *mq)
1329 {
1330 	struct mbuf *m;
1331 
1332 	m = STAILQ_FIRST(&mq->mq_head);
1333 	if (m) {
1334 		STAILQ_REMOVE_HEAD(&mq->mq_head, m_stailqpkt);
1335 		m->m_nextpkt = NULL;
1336 		mq->mq_len--;
1337 	}
1338 	return (m);
1339 }
1340 
1341 static inline void
1342 mbufq_prepend(struct mbufq *mq, struct mbuf *m)
1343 {
1344 
1345 	STAILQ_INSERT_HEAD(&mq->mq_head, m, m_stailqpkt);
1346 	mq->mq_len++;
1347 }
1348 
1349 /*
1350  * Note: this doesn't enforce the maximum list size for dst.
1351  */
1352 static inline void
1353 mbufq_concat(struct mbufq *mq_dst, struct mbufq *mq_src)
1354 {
1355 
1356 	mq_dst->mq_len += mq_src->mq_len;
1357 	STAILQ_CONCAT(&mq_dst->mq_head, &mq_src->mq_head);
1358 	mq_src->mq_len = 0;
1359 }
1360 
1361 #ifdef _SYS_TIMESPEC_H_
1362 static inline void
1363 mbuf_tstmp2timespec(struct mbuf *m, struct timespec *ts)
1364 {
1365 
1366 	KASSERT((m->m_flags & M_PKTHDR) != 0, ("mbuf %p no M_PKTHDR", m));
1367 	KASSERT((m->m_flags & M_TSTMP) != 0, ("mbuf %p no M_TSTMP", m));
1368 	ts->tv_sec = m->m_pkthdr.rcv_tstmp / 1000000000;
1369 	ts->tv_nsec = m->m_pkthdr.rcv_tstmp % 1000000000;
1370 }
1371 #endif
1372 
1373 #endif /* _KERNEL */
1374 #endif /* !_SYS_MBUF_H_ */
1375