xref: /freebsd/sys/sys/mbuf.h (revision 243e928310d073338c5ec089f0dce238a80b9866)
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  * Record/packet header in first mbuf of chain; valid only if M_PKTHDR is set.
134  * Size ILP32: 48
135  *	 LP64: 56
136  * Compile-time assertions in uipc_mbuf.c test these values to ensure that
137  * they are correct.
138  */
139 struct pkthdr {
140 	struct ifnet	*rcvif;		/* rcv interface */
141 	SLIST_HEAD(packet_tags, m_tag) tags; /* list of packet tags */
142 	int32_t		 len;		/* total packet length */
143 
144 	/* Layer crossing persistent information. */
145 	uint32_t	 flowid;	/* packet's 4-tuple system */
146 	uint64_t	 csum_flags;	/* checksum and offload features */
147 	uint16_t	 fibnum;	/* this packet should use this fib */
148 	uint8_t		 cosqos;	/* class/quality of service */
149 	uint8_t		 rsstype;	/* hash type */
150 	uint8_t		 l2hlen;	/* layer 2 header length */
151 	uint8_t		 l3hlen;	/* layer 3 header length */
152 	uint8_t		 l4hlen;	/* layer 4 header length */
153 	uint8_t		 l5hlen;	/* layer 5 header length */
154 	union {
155 		uint8_t  eight[8];
156 		uint16_t sixteen[4];
157 		uint32_t thirtytwo[2];
158 		uint64_t sixtyfour[1];
159 		uintptr_t unintptr[1];
160 		void	*ptr;
161 	} PH_per;
162 
163 	/* Layer specific non-persistent local storage for reassembly, etc. */
164 	union {
165 		uint8_t  eight[8];
166 		uint16_t sixteen[4];
167 		uint32_t thirtytwo[2];
168 		uint64_t sixtyfour[1];
169 		uintptr_t unintptr[1];
170 		void 	*ptr;
171 	} PH_loc;
172 };
173 #define	ether_vtag	PH_per.sixteen[0]
174 #define	PH_vt		PH_per
175 #define	vt_nrecs	sixteen[0]
176 #define	tso_segsz	PH_per.sixteen[1]
177 #define	csum_phsum	PH_per.sixteen[2]
178 #define	csum_data	PH_per.thirtytwo[1]
179 
180 /*
181  * Description of external storage mapped into mbuf; valid only if M_EXT is
182  * set.
183  * Size ILP32: 28
184  *	 LP64: 48
185  * Compile-time assertions in uipc_mbuf.c test these values to ensure that
186  * they are correct.
187  */
188 struct m_ext {
189 	union {
190 		volatile u_int	 ext_count;	/* value of ref count info */
191 		volatile u_int	*ext_cnt;	/* pointer to ref count info */
192 	};
193 	caddr_t		 ext_buf;	/* start of buffer */
194 	uint32_t	 ext_size;	/* size of buffer, for ext_free */
195 	uint32_t	 ext_type:8,	/* type of external storage */
196 			 ext_flags:24;	/* external storage mbuf flags */
197 	void		(*ext_free)	/* free routine if not the usual */
198 			    (struct mbuf *, void *, void *);
199 	void		*ext_arg1;	/* optional argument pointer */
200 	void		*ext_arg2;	/* optional argument pointer */
201 };
202 
203 /*
204  * The core of the mbuf object along with some shortcut defines for practical
205  * purposes.
206  */
207 struct mbuf {
208 	/*
209 	 * Header present at the beginning of every mbuf.
210 	 * Size ILP32: 24
211 	 *      LP64: 32
212 	 * Compile-time assertions in uipc_mbuf.c test these values to ensure
213 	 * that they are correct.
214 	 */
215 	union {	/* next buffer in chain */
216 		struct mbuf		*m_next;
217 		SLIST_ENTRY(mbuf)	m_slist;
218 		STAILQ_ENTRY(mbuf)	m_stailq;
219 	};
220 	union {	/* next chain in queue/record */
221 		struct mbuf		*m_nextpkt;
222 		SLIST_ENTRY(mbuf)	m_slistpkt;
223 		STAILQ_ENTRY(mbuf)	m_stailqpkt;
224 	};
225 	caddr_t		 m_data;	/* location of data */
226 	int32_t		 m_len;		/* amount of data in this mbuf */
227 	uint32_t	 m_type:8,	/* type of data in this mbuf */
228 			 m_flags:24;	/* flags; see below */
229 #if !defined(__LP64__)
230 	uint32_t	 m_pad;		/* pad for 64bit alignment */
231 #endif
232 
233 	/*
234 	 * A set of optional headers (packet header, external storage header)
235 	 * and internal data storage.  Historically, these arrays were sized
236 	 * to MHLEN (space left after a packet header) and MLEN (space left
237 	 * after only a regular mbuf header); they are now variable size in
238 	 * order to support future work on variable-size mbufs.
239 	 */
240 	union {
241 		struct {
242 			struct pkthdr	m_pkthdr;	/* M_PKTHDR set */
243 			union {
244 				struct m_ext	m_ext;	/* M_EXT set */
245 				char		m_pktdat[0];
246 			};
247 		};
248 		char	m_dat[0];			/* !M_PKTHDR, !M_EXT */
249 	};
250 };
251 
252 /*
253  * mbuf flags of global significance and layer crossing.
254  * Those of only protocol/layer specific significance are to be mapped
255  * to M_PROTO[1-12] and cleared at layer handoff boundaries.
256  * NB: Limited to the lower 24 bits.
257  */
258 #define	M_EXT		0x00000001 /* has associated external storage */
259 #define	M_PKTHDR	0x00000002 /* start of record */
260 #define	M_EOR		0x00000004 /* end of record */
261 #define	M_RDONLY	0x00000008 /* associated data is marked read-only */
262 #define	M_BCAST		0x00000010 /* send/received as link-level broadcast */
263 #define	M_MCAST		0x00000020 /* send/received as link-level multicast */
264 #define	M_PROMISC	0x00000040 /* packet was not for us */
265 #define	M_VLANTAG	0x00000080 /* ether_vtag is valid */
266 #define	M_UNUSED_8	0x00000100 /* --available-- */
267 #define	M_NOFREE	0x00000200 /* do not free mbuf, embedded in cluster */
268 
269 #define	M_PROTO1	0x00001000 /* protocol-specific */
270 #define	M_PROTO2	0x00002000 /* protocol-specific */
271 #define	M_PROTO3	0x00004000 /* protocol-specific */
272 #define	M_PROTO4	0x00008000 /* protocol-specific */
273 #define	M_PROTO5	0x00010000 /* protocol-specific */
274 #define	M_PROTO6	0x00020000 /* protocol-specific */
275 #define	M_PROTO7	0x00040000 /* protocol-specific */
276 #define	M_PROTO8	0x00080000 /* protocol-specific */
277 #define	M_PROTO9	0x00100000 /* protocol-specific */
278 #define	M_PROTO10	0x00200000 /* protocol-specific */
279 #define	M_PROTO11	0x00400000 /* protocol-specific */
280 #define	M_PROTO12	0x00800000 /* protocol-specific */
281 
282 #define MB_DTOR_SKIP	0x1	/* don't pollute the cache by touching a freed mbuf */
283 
284 /*
285  * Flags to purge when crossing layers.
286  */
287 #define	M_PROTOFLAGS \
288     (M_PROTO1|M_PROTO2|M_PROTO3|M_PROTO4|M_PROTO5|M_PROTO6|M_PROTO7|M_PROTO8|\
289      M_PROTO9|M_PROTO10|M_PROTO11|M_PROTO12)
290 
291 /*
292  * Flags preserved when copying m_pkthdr.
293  */
294 #define M_COPYFLAGS \
295     (M_PKTHDR|M_EOR|M_RDONLY|M_BCAST|M_MCAST|M_PROMISC|M_VLANTAG| \
296      M_PROTOFLAGS)
297 
298 /*
299  * Mbuf flag description for use with printf(9) %b identifier.
300  */
301 #define	M_FLAG_BITS \
302     "\20\1M_EXT\2M_PKTHDR\3M_EOR\4M_RDONLY\5M_BCAST\6M_MCAST" \
303     "\7M_PROMISC\10M_VLANTAG"
304 #define	M_FLAG_PROTOBITS \
305     "\15M_PROTO1\16M_PROTO2\17M_PROTO3\20M_PROTO4\21M_PROTO5" \
306     "\22M_PROTO6\23M_PROTO7\24M_PROTO8\25M_PROTO9\26M_PROTO10" \
307     "\27M_PROTO11\30M_PROTO12"
308 #define	M_FLAG_PRINTF (M_FLAG_BITS M_FLAG_PROTOBITS)
309 
310 /*
311  * Network interface cards are able to hash protocol fields (such as IPv4
312  * addresses and TCP port numbers) classify packets into flows.  These flows
313  * can then be used to maintain ordering while delivering packets to the OS
314  * via parallel input queues, as well as to provide a stateless affinity
315  * model.  NIC drivers can pass up the hash via m->m_pkthdr.flowid, and set
316  * m_flag fields to indicate how the hash should be interpreted by the
317  * network stack.
318  *
319  * Most NICs support RSS, which provides ordering and explicit affinity, and
320  * use the hash m_flag bits to indicate what header fields were covered by
321  * the hash.  M_HASHTYPE_OPAQUE can be set by non-RSS cards or configurations
322  * that provide an opaque flow identifier, allowing for ordering and
323  * distribution without explicit affinity.
324  */
325 /* Microsoft RSS standard hash types */
326 #define	M_HASHTYPE_NONE			0
327 #define	M_HASHTYPE_RSS_IPV4		1	/* IPv4 2-tuple */
328 #define	M_HASHTYPE_RSS_TCP_IPV4		2	/* TCPv4 4-tuple */
329 #define	M_HASHTYPE_RSS_IPV6		3	/* IPv6 2-tuple */
330 #define	M_HASHTYPE_RSS_TCP_IPV6		4	/* TCPv6 4-tuple */
331 #define	M_HASHTYPE_RSS_IPV6_EX		5	/* IPv6 2-tuple + ext hdrs */
332 #define	M_HASHTYPE_RSS_TCP_IPV6_EX	6	/* TCPv6 4-tiple + ext hdrs */
333 /* Non-standard RSS hash types */
334 #define	M_HASHTYPE_RSS_UDP_IPV4		7	/* IPv4 UDP 4-tuple */
335 #define	M_HASHTYPE_RSS_UDP_IPV4_EX	8	/* IPv4 UDP 4-tuple + ext hdrs */
336 #define	M_HASHTYPE_RSS_UDP_IPV6		9	/* IPv6 UDP 4-tuple */
337 #define	M_HASHTYPE_RSS_UDP_IPV6_EX	10	/* IPv6 UDP 4-tuple + ext hdrs */
338 
339 #define	M_HASHTYPE_OPAQUE		255	/* ordering, not affinity */
340 
341 #define	M_HASHTYPE_CLEAR(m)	((m)->m_pkthdr.rsstype = 0)
342 #define	M_HASHTYPE_GET(m)	((m)->m_pkthdr.rsstype)
343 #define	M_HASHTYPE_SET(m, v)	((m)->m_pkthdr.rsstype = (v))
344 #define	M_HASHTYPE_TEST(m, v)	(M_HASHTYPE_GET(m) == (v))
345 
346 /*
347  * COS/QOS class and quality of service tags.
348  * It uses DSCP code points as base.
349  */
350 #define	QOS_DSCP_CS0		0x00
351 #define	QOS_DSCP_DEF		QOS_DSCP_CS0
352 #define	QOS_DSCP_CS1		0x20
353 #define	QOS_DSCP_AF11		0x28
354 #define	QOS_DSCP_AF12		0x30
355 #define	QOS_DSCP_AF13		0x38
356 #define	QOS_DSCP_CS2		0x40
357 #define	QOS_DSCP_AF21		0x48
358 #define	QOS_DSCP_AF22		0x50
359 #define	QOS_DSCP_AF23		0x58
360 #define	QOS_DSCP_CS3		0x60
361 #define	QOS_DSCP_AF31		0x68
362 #define	QOS_DSCP_AF32		0x70
363 #define	QOS_DSCP_AF33		0x78
364 #define	QOS_DSCP_CS4		0x80
365 #define	QOS_DSCP_AF41		0x88
366 #define	QOS_DSCP_AF42		0x90
367 #define	QOS_DSCP_AF43		0x98
368 #define	QOS_DSCP_CS5		0xa0
369 #define	QOS_DSCP_EF		0xb8
370 #define	QOS_DSCP_CS6		0xc0
371 #define	QOS_DSCP_CS7		0xe0
372 
373 /*
374  * External mbuf storage buffer types.
375  */
376 #define	EXT_CLUSTER	1	/* mbuf cluster */
377 #define	EXT_SFBUF	2	/* sendfile(2)'s sf_buf */
378 #define	EXT_JUMBOP	3	/* jumbo cluster page sized */
379 #define	EXT_JUMBO9	4	/* jumbo cluster 9216 bytes */
380 #define	EXT_JUMBO16	5	/* jumbo cluster 16184 bytes */
381 #define	EXT_PACKET	6	/* mbuf+cluster from packet zone */
382 #define	EXT_MBUF	7	/* external mbuf reference (M_IOVEC) */
383 #define	EXT_SFBUF_NOCACHE 8	/* sendfile(2)'s sf_buf not to be cached */
384 
385 #define	EXT_VENDOR1	224	/* for vendor-internal use */
386 #define	EXT_VENDOR2	225	/* for vendor-internal use */
387 #define	EXT_VENDOR3	226	/* for vendor-internal use */
388 #define	EXT_VENDOR4	227	/* for vendor-internal use */
389 
390 #define	EXT_EXP1	244	/* for experimental use */
391 #define	EXT_EXP2	245	/* for experimental use */
392 #define	EXT_EXP3	246	/* for experimental use */
393 #define	EXT_EXP4	247	/* for experimental use */
394 
395 #define	EXT_NET_DRV	252	/* custom ext_buf provided by net driver(s) */
396 #define	EXT_MOD_TYPE	253	/* custom module's ext_buf type */
397 #define	EXT_DISPOSABLE	254	/* can throw this buffer away w/page flipping */
398 #define	EXT_EXTREF	255	/* has externally maintained ext_cnt ptr */
399 
400 /*
401  * Flags for external mbuf buffer types.
402  * NB: limited to the lower 24 bits.
403  */
404 #define	EXT_FLAG_EMBREF		0x000001	/* embedded ext_count */
405 #define	EXT_FLAG_EXTREF		0x000002	/* external ext_cnt, notyet */
406 
407 #define	EXT_FLAG_NOFREE		0x000010	/* don't free mbuf to pool, notyet */
408 
409 #define	EXT_FLAG_VENDOR1	0x010000	/* for vendor-internal use */
410 #define	EXT_FLAG_VENDOR2	0x020000	/* for vendor-internal use */
411 #define	EXT_FLAG_VENDOR3	0x040000	/* for vendor-internal use */
412 #define	EXT_FLAG_VENDOR4	0x080000	/* for vendor-internal use */
413 
414 #define	EXT_FLAG_EXP1		0x100000	/* for experimental use */
415 #define	EXT_FLAG_EXP2		0x200000	/* for experimental use */
416 #define	EXT_FLAG_EXP3		0x400000	/* for experimental use */
417 #define	EXT_FLAG_EXP4		0x800000	/* for experimental use */
418 
419 /*
420  * EXT flag description for use with printf(9) %b identifier.
421  */
422 #define	EXT_FLAG_BITS \
423     "\20\1EXT_FLAG_EMBREF\2EXT_FLAG_EXTREF\5EXT_FLAG_NOFREE" \
424     "\21EXT_FLAG_VENDOR1\22EXT_FLAG_VENDOR2\23EXT_FLAG_VENDOR3" \
425     "\24EXT_FLAG_VENDOR4\25EXT_FLAG_EXP1\26EXT_FLAG_EXP2\27EXT_FLAG_EXP3" \
426     "\30EXT_FLAG_EXP4"
427 
428 /*
429  * External reference/free functions.
430  */
431 void sf_ext_free(void *, void *);
432 void sf_ext_free_nocache(void *, void *);
433 
434 /*
435  * Flags indicating checksum, segmentation and other offload work to be
436  * done, or already done, by hardware or lower layers.  It is split into
437  * separate inbound and outbound flags.
438  *
439  * Outbound flags that are set by upper protocol layers requesting lower
440  * layers, or ideally the hardware, to perform these offloading tasks.
441  * For outbound packets this field and its flags can be directly tested
442  * against ifnet if_hwassist.
443  */
444 #define	CSUM_IP			0x00000001	/* IP header checksum offload */
445 #define	CSUM_IP_UDP		0x00000002	/* UDP checksum offload */
446 #define	CSUM_IP_TCP		0x00000004	/* TCP checksum offload */
447 #define	CSUM_IP_SCTP		0x00000008	/* SCTP checksum offload */
448 #define	CSUM_IP_TSO		0x00000010	/* TCP segmentation offload */
449 #define	CSUM_IP_ISCSI		0x00000020	/* iSCSI checksum offload */
450 
451 #define	CSUM_IP6_UDP		0x00000200	/* UDP checksum offload */
452 #define	CSUM_IP6_TCP		0x00000400	/* TCP checksum offload */
453 #define	CSUM_IP6_SCTP		0x00000800	/* SCTP checksum offload */
454 #define	CSUM_IP6_TSO		0x00001000	/* TCP segmentation offload */
455 #define	CSUM_IP6_ISCSI		0x00002000	/* iSCSI checksum offload */
456 
457 /* Inbound checksum support where the checksum was verified by hardware. */
458 #define	CSUM_L3_CALC		0x01000000	/* calculated layer 3 csum */
459 #define	CSUM_L3_VALID		0x02000000	/* checksum is correct */
460 #define	CSUM_L4_CALC		0x04000000	/* calculated layer 4 csum */
461 #define	CSUM_L4_VALID		0x08000000	/* checksum is correct */
462 #define	CSUM_L5_CALC		0x10000000	/* calculated layer 5 csum */
463 #define	CSUM_L5_VALID		0x20000000	/* checksum is correct */
464 #define	CSUM_COALESED		0x40000000	/* contains merged segments */
465 
466 /*
467  * CSUM flag description for use with printf(9) %b identifier.
468  */
469 #define	CSUM_BITS \
470     "\20\1CSUM_IP\2CSUM_IP_UDP\3CSUM_IP_TCP\4CSUM_IP_SCTP\5CSUM_IP_TSO" \
471     "\6CSUM_IP_ISCSI" \
472     "\12CSUM_IP6_UDP\13CSUM_IP6_TCP\14CSUM_IP6_SCTP\15CSUM_IP6_TSO" \
473     "\16CSUM_IP6_ISCSI" \
474     "\31CSUM_L3_CALC\32CSUM_L3_VALID\33CSUM_L4_CALC\34CSUM_L4_VALID" \
475     "\35CSUM_L5_CALC\36CSUM_L5_VALID\37CSUM_COALESED"
476 
477 /* CSUM flags compatibility mappings. */
478 #define	CSUM_IP_CHECKED		CSUM_L3_CALC
479 #define	CSUM_IP_VALID		CSUM_L3_VALID
480 #define	CSUM_DATA_VALID		CSUM_L4_VALID
481 #define	CSUM_PSEUDO_HDR		CSUM_L4_CALC
482 #define	CSUM_SCTP_VALID		CSUM_L4_VALID
483 #define	CSUM_DELAY_DATA		(CSUM_TCP|CSUM_UDP)
484 #define	CSUM_DELAY_IP		CSUM_IP		/* Only v4, no v6 IP hdr csum */
485 #define	CSUM_DELAY_DATA_IPV6	(CSUM_TCP_IPV6|CSUM_UDP_IPV6)
486 #define	CSUM_DATA_VALID_IPV6	CSUM_DATA_VALID
487 #define	CSUM_TCP		CSUM_IP_TCP
488 #define	CSUM_UDP		CSUM_IP_UDP
489 #define	CSUM_SCTP		CSUM_IP_SCTP
490 #define	CSUM_TSO		(CSUM_IP_TSO|CSUM_IP6_TSO)
491 #define	CSUM_UDP_IPV6		CSUM_IP6_UDP
492 #define	CSUM_TCP_IPV6		CSUM_IP6_TCP
493 #define	CSUM_SCTP_IPV6		CSUM_IP6_SCTP
494 
495 /*
496  * mbuf types describing the content of the mbuf (including external storage).
497  */
498 #define	MT_NOTMBUF	0	/* USED INTERNALLY ONLY! Object is not mbuf */
499 #define	MT_DATA		1	/* dynamic (data) allocation */
500 #define	MT_HEADER	MT_DATA	/* packet header, use M_PKTHDR instead */
501 
502 #define	MT_VENDOR1	4	/* for vendor-internal use */
503 #define	MT_VENDOR2	5	/* for vendor-internal use */
504 #define	MT_VENDOR3	6	/* for vendor-internal use */
505 #define	MT_VENDOR4	7	/* for vendor-internal use */
506 
507 #define	MT_SONAME	8	/* socket name */
508 
509 #define	MT_EXP1		9	/* for experimental use */
510 #define	MT_EXP2		10	/* for experimental use */
511 #define	MT_EXP3		11	/* for experimental use */
512 #define	MT_EXP4		12	/* for experimental use */
513 
514 #define	MT_CONTROL	14	/* extra-data protocol message */
515 #define	MT_OOBDATA	15	/* expedited data  */
516 #define	MT_NTYPES	16	/* number of mbuf types for mbtypes[] */
517 
518 #define	MT_NOINIT	255	/* Not a type but a flag to allocate
519 				   a non-initialized mbuf */
520 
521 /*
522  * String names of mbuf-related UMA(9) and malloc(9) types.  Exposed to
523  * !_KERNEL so that monitoring tools can look up the zones with
524  * libmemstat(3).
525  */
526 #define	MBUF_MEM_NAME		"mbuf"
527 #define	MBUF_CLUSTER_MEM_NAME	"mbuf_cluster"
528 #define	MBUF_PACKET_MEM_NAME	"mbuf_packet"
529 #define	MBUF_JUMBOP_MEM_NAME	"mbuf_jumbo_page"
530 #define	MBUF_JUMBO9_MEM_NAME	"mbuf_jumbo_9k"
531 #define	MBUF_JUMBO16_MEM_NAME	"mbuf_jumbo_16k"
532 #define	MBUF_TAG_MEM_NAME	"mbuf_tag"
533 #define	MBUF_EXTREFCNT_MEM_NAME	"mbuf_ext_refcnt"
534 
535 #ifdef _KERNEL
536 
537 #ifdef WITNESS
538 #define	MBUF_CHECKSLEEP(how) do {					\
539 	if (how == M_WAITOK)						\
540 		WITNESS_WARN(WARN_GIANTOK | WARN_SLEEPOK, NULL,		\
541 		    "Sleeping in \"%s\"", __func__);			\
542 } while (0)
543 #else
544 #define	MBUF_CHECKSLEEP(how)
545 #endif
546 
547 /*
548  * Network buffer allocation API
549  *
550  * The rest of it is defined in kern/kern_mbuf.c
551  */
552 extern uma_zone_t	zone_mbuf;
553 extern uma_zone_t	zone_clust;
554 extern uma_zone_t	zone_pack;
555 extern uma_zone_t	zone_jumbop;
556 extern uma_zone_t	zone_jumbo9;
557 extern uma_zone_t	zone_jumbo16;
558 
559 void		 mb_dupcl(struct mbuf *, struct mbuf *);
560 void		 mb_free_ext(struct mbuf *);
561 void		 m_adj(struct mbuf *, int);
562 int		 m_apply(struct mbuf *, int, int,
563 		    int (*)(void *, void *, u_int), void *);
564 int		 m_append(struct mbuf *, int, c_caddr_t);
565 void		 m_cat(struct mbuf *, struct mbuf *);
566 void		 m_catpkt(struct mbuf *, struct mbuf *);
567 int		 m_clget(struct mbuf *m, int how);
568 void 		*m_cljget(struct mbuf *m, int how, int size);
569 struct mbuf	*m_collapse(struct mbuf *, int, int);
570 void		 m_copyback(struct mbuf *, int, int, c_caddr_t);
571 void		 m_copydata(const struct mbuf *, int, int, caddr_t);
572 struct mbuf	*m_copym(struct mbuf *, int, int, int);
573 struct mbuf	*m_copypacket(struct mbuf *, int);
574 void		 m_copy_pkthdr(struct mbuf *, struct mbuf *);
575 struct mbuf	*m_copyup(struct mbuf *, int, int);
576 struct mbuf	*m_defrag(struct mbuf *, int);
577 void		 m_demote_pkthdr(struct mbuf *);
578 void		 m_demote(struct mbuf *, int, int);
579 struct mbuf	*m_devget(char *, int, int, struct ifnet *,
580 		    void (*)(char *, caddr_t, u_int));
581 struct mbuf	*m_dup(const struct mbuf *, int);
582 int		 m_dup_pkthdr(struct mbuf *, const struct mbuf *, int);
583 void		 m_extadd(struct mbuf *, caddr_t, u_int,
584 		    void (*)(struct mbuf *, void *, void *), void *, void *,
585 		    int, int);
586 u_int		 m_fixhdr(struct mbuf *);
587 struct mbuf	*m_fragment(struct mbuf *, int, int);
588 void		 m_freem(struct mbuf *);
589 struct mbuf	*m_get2(int, int, short, int);
590 struct mbuf	*m_getjcl(int, short, int, int);
591 struct mbuf	*m_getm2(struct mbuf *, int, int, short, int);
592 struct mbuf	*m_getptr(struct mbuf *, int, int *);
593 u_int		 m_length(struct mbuf *, struct mbuf **);
594 int		 m_mbuftouio(struct uio *, struct mbuf *, int);
595 void		 m_move_pkthdr(struct mbuf *, struct mbuf *);
596 int		 m_pkthdr_init(struct mbuf *, int);
597 struct mbuf	*m_prepend(struct mbuf *, int, int);
598 void		 m_print(const struct mbuf *, int);
599 struct mbuf	*m_pulldown(struct mbuf *, int, int, int *);
600 struct mbuf	*m_pullup(struct mbuf *, int);
601 int		 m_sanity(struct mbuf *, int);
602 struct mbuf	*m_split(struct mbuf *, int, int);
603 struct mbuf	*m_uiotombuf(struct uio *, int, int, int, int);
604 struct mbuf	*m_unshare(struct mbuf *, int);
605 
606 static __inline int
607 m_gettype(int size)
608 {
609 	int type;
610 
611 	switch (size) {
612 	case MSIZE:
613 		type = EXT_MBUF;
614 		break;
615 	case MCLBYTES:
616 		type = EXT_CLUSTER;
617 		break;
618 #if MJUMPAGESIZE != MCLBYTES
619 	case MJUMPAGESIZE:
620 		type = EXT_JUMBOP;
621 		break;
622 #endif
623 	case MJUM9BYTES:
624 		type = EXT_JUMBO9;
625 		break;
626 	case MJUM16BYTES:
627 		type = EXT_JUMBO16;
628 		break;
629 	default:
630 		panic("%s: invalid cluster size %d", __func__, size);
631 	}
632 
633 	return (type);
634 }
635 
636 /*
637  * Associated an external reference counted buffer with an mbuf.
638  */
639 static __inline void
640 m_extaddref(struct mbuf *m, caddr_t buf, u_int size, u_int *ref_cnt,
641     void (*freef)(struct mbuf *, void *, void *), void *arg1, void *arg2)
642 {
643 
644 	KASSERT(ref_cnt != NULL, ("%s: ref_cnt not provided", __func__));
645 
646 	atomic_add_int(ref_cnt, 1);
647 	m->m_flags |= M_EXT;
648 	m->m_ext.ext_buf = buf;
649 	m->m_ext.ext_cnt = ref_cnt;
650 	m->m_data = m->m_ext.ext_buf;
651 	m->m_ext.ext_size = size;
652 	m->m_ext.ext_free = freef;
653 	m->m_ext.ext_arg1 = arg1;
654 	m->m_ext.ext_arg2 = arg2;
655 	m->m_ext.ext_type = EXT_EXTREF;
656 	m->m_ext.ext_flags = 0;
657 }
658 
659 static __inline uma_zone_t
660 m_getzone(int size)
661 {
662 	uma_zone_t zone;
663 
664 	switch (size) {
665 	case MCLBYTES:
666 		zone = zone_clust;
667 		break;
668 #if MJUMPAGESIZE != MCLBYTES
669 	case MJUMPAGESIZE:
670 		zone = zone_jumbop;
671 		break;
672 #endif
673 	case MJUM9BYTES:
674 		zone = zone_jumbo9;
675 		break;
676 	case MJUM16BYTES:
677 		zone = zone_jumbo16;
678 		break;
679 	default:
680 		panic("%s: invalid cluster size %d", __func__, size);
681 	}
682 
683 	return (zone);
684 }
685 
686 /*
687  * Initialize an mbuf with linear storage.
688  *
689  * Inline because the consumer text overhead will be roughly the same to
690  * initialize or call a function with this many parameters and M_PKTHDR
691  * should go away with constant propagation for !MGETHDR.
692  */
693 static __inline int
694 m_init(struct mbuf *m, int how, short type, int flags)
695 {
696 	int error;
697 
698 	m->m_next = NULL;
699 	m->m_nextpkt = NULL;
700 	m->m_data = m->m_dat;
701 	m->m_len = 0;
702 	m->m_flags = flags;
703 	m->m_type = type;
704 	if (flags & M_PKTHDR)
705 		error = m_pkthdr_init(m, how);
706 	else
707 		error = 0;
708 
709 	MBUF_PROBE5(m__init, m, how, type, flags, error);
710 	return (error);
711 }
712 
713 static __inline struct mbuf *
714 m_get(int how, short type)
715 {
716 	struct mbuf *m;
717 	struct mb_args args;
718 
719 	args.flags = 0;
720 	args.type = type;
721 	m = uma_zalloc_arg(zone_mbuf, &args, how);
722 	MBUF_PROBE3(m__get, how, type, m);
723 	return (m);
724 }
725 
726 static __inline struct mbuf *
727 m_gethdr(int how, short type)
728 {
729 	struct mbuf *m;
730 	struct mb_args args;
731 
732 	args.flags = M_PKTHDR;
733 	args.type = type;
734 	m = uma_zalloc_arg(zone_mbuf, &args, how);
735 	MBUF_PROBE3(m__gethdr, how, type, m);
736 	return (m);
737 }
738 
739 static __inline struct mbuf *
740 m_getcl(int how, short type, int flags)
741 {
742 	struct mbuf *m;
743 	struct mb_args args;
744 
745 	args.flags = flags;
746 	args.type = type;
747 	m = uma_zalloc_arg(zone_pack, &args, how);
748 	MBUF_PROBE4(m__getcl, how, type, flags, m);
749 	return (m);
750 }
751 
752 /*
753  * XXX: m_cljset() is a dangerous API.  One must attach only a new,
754  * unreferenced cluster to an mbuf(9).  It is not possible to assert
755  * that, so care can be taken only by users of the API.
756  */
757 static __inline void
758 m_cljset(struct mbuf *m, void *cl, int type)
759 {
760 	int size;
761 
762 	switch (type) {
763 	case EXT_CLUSTER:
764 		size = MCLBYTES;
765 		break;
766 #if MJUMPAGESIZE != MCLBYTES
767 	case EXT_JUMBOP:
768 		size = MJUMPAGESIZE;
769 		break;
770 #endif
771 	case EXT_JUMBO9:
772 		size = MJUM9BYTES;
773 		break;
774 	case EXT_JUMBO16:
775 		size = MJUM16BYTES;
776 		break;
777 	default:
778 		panic("%s: unknown cluster type %d", __func__, type);
779 		break;
780 	}
781 
782 	m->m_data = m->m_ext.ext_buf = cl;
783 	m->m_ext.ext_free = m->m_ext.ext_arg1 = m->m_ext.ext_arg2 = NULL;
784 	m->m_ext.ext_size = size;
785 	m->m_ext.ext_type = type;
786 	m->m_ext.ext_flags = EXT_FLAG_EMBREF;
787 	m->m_ext.ext_count = 1;
788 	m->m_flags |= M_EXT;
789 	MBUF_PROBE3(m__cljset, m, cl, type);
790 }
791 
792 static __inline void
793 m_chtype(struct mbuf *m, short new_type)
794 {
795 
796 	m->m_type = new_type;
797 }
798 
799 static __inline void
800 m_clrprotoflags(struct mbuf *m)
801 {
802 
803 	while (m) {
804 		m->m_flags &= ~M_PROTOFLAGS;
805 		m = m->m_next;
806 	}
807 }
808 
809 static __inline struct mbuf *
810 m_last(struct mbuf *m)
811 {
812 
813 	while (m->m_next)
814 		m = m->m_next;
815 	return (m);
816 }
817 
818 static inline u_int
819 m_extrefcnt(struct mbuf *m)
820 {
821 
822 	KASSERT(m->m_flags & M_EXT, ("%s: M_EXT missing", __func__));
823 
824 	return ((m->m_ext.ext_flags & EXT_FLAG_EMBREF) ? m->m_ext.ext_count :
825 	    *m->m_ext.ext_cnt);
826 }
827 
828 /*
829  * mbuf, cluster, and external object allocation macros (for compatibility
830  * purposes).
831  */
832 #define	M_MOVE_PKTHDR(to, from)	m_move_pkthdr((to), (from))
833 #define	MGET(m, how, type)	((m) = m_get((how), (type)))
834 #define	MGETHDR(m, how, type)	((m) = m_gethdr((how), (type)))
835 #define	MCLGET(m, how)		m_clget((m), (how))
836 #define	MEXTADD(m, buf, size, free, arg1, arg2, flags, type)		\
837     m_extadd((m), (caddr_t)(buf), (size), (free), (arg1), (arg2),	\
838     (flags), (type))
839 #define	m_getm(m, len, how, type)					\
840     m_getm2((m), (len), (how), (type), M_PKTHDR)
841 
842 /*
843  * Evaluate TRUE if it's safe to write to the mbuf m's data region (this can
844  * be both the local data payload, or an external buffer area, depending on
845  * whether M_EXT is set).
846  */
847 #define	M_WRITABLE(m)	(!((m)->m_flags & M_RDONLY) &&			\
848 			 (!(((m)->m_flags & M_EXT)) ||			\
849 			 (m_extrefcnt(m) == 1)))
850 
851 /* Check if the supplied mbuf has a packet header, or else panic. */
852 #define	M_ASSERTPKTHDR(m)						\
853 	KASSERT((m) != NULL && (m)->m_flags & M_PKTHDR,			\
854 	    ("%s: no mbuf packet header!", __func__))
855 
856 /*
857  * Ensure that the supplied mbuf is a valid, non-free mbuf.
858  *
859  * XXX: Broken at the moment.  Need some UMA magic to make it work again.
860  */
861 #define	M_ASSERTVALID(m)						\
862 	KASSERT((((struct mbuf *)m)->m_flags & 0) == 0,			\
863 	    ("%s: attempted use of a free mbuf!", __func__))
864 
865 /*
866  * Return the address of the start of the buffer associated with an mbuf,
867  * handling external storage, packet-header mbufs, and regular data mbufs.
868  */
869 #define	M_START(m)							\
870 	(((m)->m_flags & M_EXT) ? (m)->m_ext.ext_buf :			\
871 	 ((m)->m_flags & M_PKTHDR) ? &(m)->m_pktdat[0] :		\
872 	 &(m)->m_dat[0])
873 
874 /*
875  * Return the size of the buffer associated with an mbuf, handling external
876  * storage, packet-header mbufs, and regular data mbufs.
877  */
878 #define	M_SIZE(m)							\
879 	(((m)->m_flags & M_EXT) ? (m)->m_ext.ext_size :			\
880 	 ((m)->m_flags & M_PKTHDR) ? MHLEN :				\
881 	 MLEN)
882 
883 /*
884  * Set the m_data pointer of a newly allocated mbuf to place an object of the
885  * specified size at the end of the mbuf, longword aligned.
886  *
887  * NB: Historically, we had M_ALIGN(), MH_ALIGN(), and MEXT_ALIGN() as
888  * separate macros, each asserting that it was called at the proper moment.
889  * This required callers to themselves test the storage type and call the
890  * right one.  Rather than require callers to be aware of those layout
891  * decisions, we centralize here.
892  */
893 static __inline void
894 m_align(struct mbuf *m, int len)
895 {
896 #ifdef INVARIANTS
897 	const char *msg = "%s: not a virgin mbuf";
898 #endif
899 	int adjust;
900 
901 	KASSERT(m->m_data == M_START(m), (msg, __func__));
902 
903 	adjust = M_SIZE(m) - len;
904 	m->m_data += adjust &~ (sizeof(long)-1);
905 }
906 
907 #define	M_ALIGN(m, len)		m_align(m, len)
908 #define	MH_ALIGN(m, len)	m_align(m, len)
909 #define	MEXT_ALIGN(m, len)	m_align(m, len)
910 
911 /*
912  * Compute the amount of space available before the current start of data in
913  * an mbuf.
914  *
915  * The M_WRITABLE() is a temporary, conservative safety measure: the burden
916  * of checking writability of the mbuf data area rests solely with the caller.
917  *
918  * NB: In previous versions, M_LEADINGSPACE() would only check M_WRITABLE()
919  * for mbufs with external storage.  We now allow mbuf-embedded data to be
920  * read-only as well.
921  */
922 #define	M_LEADINGSPACE(m)						\
923 	(M_WRITABLE(m) ? ((m)->m_data - M_START(m)) : 0)
924 
925 /*
926  * Compute the amount of space available after the end of data in an mbuf.
927  *
928  * The M_WRITABLE() is a temporary, conservative safety measure: the burden
929  * of checking writability of the mbuf data area rests solely with the caller.
930  *
931  * NB: In previous versions, M_TRAILINGSPACE() would only check M_WRITABLE()
932  * for mbufs with external storage.  We now allow mbuf-embedded data to be
933  * read-only as well.
934  */
935 #define	M_TRAILINGSPACE(m)						\
936 	(M_WRITABLE(m) ?						\
937 	    ((M_START(m) + M_SIZE(m)) - ((m)->m_data + (m)->m_len)) : 0)
938 
939 /*
940  * Arrange to prepend space of size plen to mbuf m.  If a new mbuf must be
941  * allocated, how specifies whether to wait.  If the allocation fails, the
942  * original mbuf chain is freed and m is set to NULL.
943  */
944 #define	M_PREPEND(m, plen, how) do {					\
945 	struct mbuf **_mmp = &(m);					\
946 	struct mbuf *_mm = *_mmp;					\
947 	int _mplen = (plen);						\
948 	int __mhow = (how);						\
949 									\
950 	MBUF_CHECKSLEEP(how);						\
951 	if (M_LEADINGSPACE(_mm) >= _mplen) {				\
952 		_mm->m_data -= _mplen;					\
953 		_mm->m_len += _mplen;					\
954 	} else								\
955 		_mm = m_prepend(_mm, _mplen, __mhow);			\
956 	if (_mm != NULL && _mm->m_flags & M_PKTHDR)			\
957 		_mm->m_pkthdr.len += _mplen;				\
958 	*_mmp = _mm;							\
959 } while (0)
960 
961 /*
962  * Change mbuf to new type.  This is a relatively expensive operation and
963  * should be avoided.
964  */
965 #define	MCHTYPE(m, t)	m_chtype((m), (t))
966 
967 /* Length to m_copy to copy all. */
968 #define	M_COPYALL	1000000000
969 
970 /* Compatibility with 4.3. */
971 #define	m_copy(m, o, l)	m_copym((m), (o), (l), M_NOWAIT)
972 
973 extern int		max_datalen;	/* MHLEN - max_hdr */
974 extern int		max_hdr;	/* Largest link + protocol header */
975 extern int		max_linkhdr;	/* Largest link-level header */
976 extern int		max_protohdr;	/* Largest protocol header */
977 extern int		nmbclusters;	/* Maximum number of clusters */
978 
979 /*-
980  * Network packets may have annotations attached by affixing a list of
981  * "packet tags" to the pkthdr structure.  Packet tags are dynamically
982  * allocated semi-opaque data structures that have a fixed header
983  * (struct m_tag) that specifies the size of the memory block and a
984  * <cookie,type> pair that identifies it.  The cookie is a 32-bit unique
985  * unsigned value used to identify a module or ABI.  By convention this value
986  * is chosen as the date+time that the module is created, expressed as the
987  * number of seconds since the epoch (e.g., using date -u +'%s').  The type
988  * value is an ABI/module-specific value that identifies a particular
989  * annotation and is private to the module.  For compatibility with systems
990  * like OpenBSD that define packet tags w/o an ABI/module cookie, the value
991  * PACKET_ABI_COMPAT is used to implement m_tag_get and m_tag_find
992  * compatibility shim functions and several tag types are defined below.
993  * Users that do not require compatibility should use a private cookie value
994  * so that packet tag-related definitions can be maintained privately.
995  *
996  * Note that the packet tag returned by m_tag_alloc has the default memory
997  * alignment implemented by malloc.  To reference private data one can use a
998  * construct like:
999  *
1000  *	struct m_tag *mtag = m_tag_alloc(...);
1001  *	struct foo *p = (struct foo *)(mtag+1);
1002  *
1003  * if the alignment of struct m_tag is sufficient for referencing members of
1004  * struct foo.  Otherwise it is necessary to embed struct m_tag within the
1005  * private data structure to insure proper alignment; e.g.,
1006  *
1007  *	struct foo {
1008  *		struct m_tag	tag;
1009  *		...
1010  *	};
1011  *	struct foo *p = (struct foo *) m_tag_alloc(...);
1012  *	struct m_tag *mtag = &p->tag;
1013  */
1014 
1015 /*
1016  * Persistent tags stay with an mbuf until the mbuf is reclaimed.  Otherwise
1017  * tags are expected to ``vanish'' when they pass through a network
1018  * interface.  For most interfaces this happens normally as the tags are
1019  * reclaimed when the mbuf is free'd.  However in some special cases
1020  * reclaiming must be done manually.  An example is packets that pass through
1021  * the loopback interface.  Also, one must be careful to do this when
1022  * ``turning around'' packets (e.g., icmp_reflect).
1023  *
1024  * To mark a tag persistent bit-or this flag in when defining the tag id.
1025  * The tag will then be treated as described above.
1026  */
1027 #define	MTAG_PERSISTENT				0x800
1028 
1029 #define	PACKET_TAG_NONE				0  /* Nadda */
1030 
1031 /* Packet tags for use with PACKET_ABI_COMPAT. */
1032 #define	PACKET_TAG_IPSEC_IN_DONE		1  /* IPsec applied, in */
1033 #define	PACKET_TAG_IPSEC_OUT_DONE		2  /* IPsec applied, out */
1034 #define	PACKET_TAG_IPSEC_IN_CRYPTO_DONE		3  /* NIC IPsec crypto done */
1035 #define	PACKET_TAG_IPSEC_OUT_CRYPTO_NEEDED	4  /* NIC IPsec crypto req'ed */
1036 #define	PACKET_TAG_IPSEC_IN_COULD_DO_CRYPTO	5  /* NIC notifies IPsec */
1037 #define	PACKET_TAG_IPSEC_PENDING_TDB		6  /* Reminder to do IPsec */
1038 #define	PACKET_TAG_BRIDGE			7  /* Bridge processing done */
1039 #define	PACKET_TAG_GIF				8  /* GIF processing done */
1040 #define	PACKET_TAG_GRE				9  /* GRE processing done */
1041 #define	PACKET_TAG_IN_PACKET_CHECKSUM		10 /* NIC checksumming done */
1042 #define	PACKET_TAG_ENCAP			11 /* Encap.  processing */
1043 #define	PACKET_TAG_IPSEC_SOCKET			12 /* IPSEC socket ref */
1044 #define	PACKET_TAG_IPSEC_HISTORY		13 /* IPSEC history */
1045 #define	PACKET_TAG_IPV6_INPUT			14 /* IPV6 input processing */
1046 #define	PACKET_TAG_DUMMYNET			15 /* dummynet info */
1047 #define	PACKET_TAG_DIVERT			17 /* divert info */
1048 #define	PACKET_TAG_IPFORWARD			18 /* ipforward info */
1049 #define	PACKET_TAG_MACLABEL	(19 | MTAG_PERSISTENT) /* MAC label */
1050 #define	PACKET_TAG_PF		(21 | MTAG_PERSISTENT) /* PF/ALTQ information */
1051 #define	PACKET_TAG_RTSOCKFAM			25 /* rtsock sa family */
1052 #define	PACKET_TAG_IPOPTIONS			27 /* Saved IP options */
1053 #define	PACKET_TAG_CARP				28 /* CARP info */
1054 #define	PACKET_TAG_IPSEC_NAT_T_PORTS		29 /* two uint16_t */
1055 #define	PACKET_TAG_ND_OUTGOING			30 /* ND outgoing */
1056 
1057 /* Specific cookies and tags. */
1058 
1059 /* Packet tag routines. */
1060 struct m_tag	*m_tag_alloc(u_int32_t, int, int, int);
1061 void		 m_tag_delete(struct mbuf *, struct m_tag *);
1062 void		 m_tag_delete_chain(struct mbuf *, struct m_tag *);
1063 void		 m_tag_free_default(struct m_tag *);
1064 struct m_tag	*m_tag_locate(struct mbuf *, u_int32_t, int, struct m_tag *);
1065 struct m_tag	*m_tag_copy(struct m_tag *, int);
1066 int		 m_tag_copy_chain(struct mbuf *, const struct mbuf *, int);
1067 void		 m_tag_delete_nonpersistent(struct mbuf *);
1068 
1069 /*
1070  * Initialize the list of tags associated with an mbuf.
1071  */
1072 static __inline void
1073 m_tag_init(struct mbuf *m)
1074 {
1075 
1076 	SLIST_INIT(&m->m_pkthdr.tags);
1077 }
1078 
1079 /*
1080  * Set up the contents of a tag.  Note that this does not fill in the free
1081  * method; the caller is expected to do that.
1082  *
1083  * XXX probably should be called m_tag_init, but that was already taken.
1084  */
1085 static __inline void
1086 m_tag_setup(struct m_tag *t, u_int32_t cookie, int type, int len)
1087 {
1088 
1089 	t->m_tag_id = type;
1090 	t->m_tag_len = len;
1091 	t->m_tag_cookie = cookie;
1092 }
1093 
1094 /*
1095  * Reclaim resources associated with a tag.
1096  */
1097 static __inline void
1098 m_tag_free(struct m_tag *t)
1099 {
1100 
1101 	(*t->m_tag_free)(t);
1102 }
1103 
1104 /*
1105  * Return the first tag associated with an mbuf.
1106  */
1107 static __inline struct m_tag *
1108 m_tag_first(struct mbuf *m)
1109 {
1110 
1111 	return (SLIST_FIRST(&m->m_pkthdr.tags));
1112 }
1113 
1114 /*
1115  * Return the next tag in the list of tags associated with an mbuf.
1116  */
1117 static __inline struct m_tag *
1118 m_tag_next(struct mbuf *m __unused, struct m_tag *t)
1119 {
1120 
1121 	return (SLIST_NEXT(t, m_tag_link));
1122 }
1123 
1124 /*
1125  * Prepend a tag to the list of tags associated with an mbuf.
1126  */
1127 static __inline void
1128 m_tag_prepend(struct mbuf *m, struct m_tag *t)
1129 {
1130 
1131 	SLIST_INSERT_HEAD(&m->m_pkthdr.tags, t, m_tag_link);
1132 }
1133 
1134 /*
1135  * Unlink a tag from the list of tags associated with an mbuf.
1136  */
1137 static __inline void
1138 m_tag_unlink(struct mbuf *m, struct m_tag *t)
1139 {
1140 
1141 	SLIST_REMOVE(&m->m_pkthdr.tags, t, m_tag, m_tag_link);
1142 }
1143 
1144 /* These are for OpenBSD compatibility. */
1145 #define	MTAG_ABI_COMPAT		0		/* compatibility ABI */
1146 
1147 static __inline struct m_tag *
1148 m_tag_get(int type, int length, int wait)
1149 {
1150 	return (m_tag_alloc(MTAG_ABI_COMPAT, type, length, wait));
1151 }
1152 
1153 static __inline struct m_tag *
1154 m_tag_find(struct mbuf *m, int type, struct m_tag *start)
1155 {
1156 	return (SLIST_EMPTY(&m->m_pkthdr.tags) ? (struct m_tag *)NULL :
1157 	    m_tag_locate(m, MTAG_ABI_COMPAT, type, start));
1158 }
1159 
1160 static __inline struct mbuf *
1161 m_free(struct mbuf *m)
1162 {
1163 	struct mbuf *n = m->m_next;
1164 
1165 	MBUF_PROBE1(m__free, m);
1166 	if ((m->m_flags & (M_PKTHDR|M_NOFREE)) == (M_PKTHDR|M_NOFREE))
1167 		m_tag_delete_chain(m, NULL);
1168 	if (m->m_flags & M_EXT)
1169 		mb_free_ext(m);
1170 	else if ((m->m_flags & M_NOFREE) == 0)
1171 		uma_zfree(zone_mbuf, m);
1172 	return (n);
1173 }
1174 
1175 static __inline int
1176 rt_m_getfib(struct mbuf *m)
1177 {
1178 	KASSERT(m->m_flags & M_PKTHDR , ("Attempt to get FIB from non header mbuf."));
1179 	return (m->m_pkthdr.fibnum);
1180 }
1181 
1182 #define M_GETFIB(_m)   rt_m_getfib(_m)
1183 
1184 #define M_SETFIB(_m, _fib) do {						\
1185         KASSERT((_m)->m_flags & M_PKTHDR, ("Attempt to set FIB on non header mbuf."));	\
1186 	((_m)->m_pkthdr.fibnum) = (_fib);				\
1187 } while (0)
1188 
1189 /* flags passed as first argument for "m_ether_tcpip_hash()" */
1190 #define	MBUF_HASHFLAG_L2	(1 << 2)
1191 #define	MBUF_HASHFLAG_L3	(1 << 3)
1192 #define	MBUF_HASHFLAG_L4	(1 << 4)
1193 
1194 /* mbuf hashing helper routines */
1195 uint32_t	m_ether_tcpip_hash_init(void);
1196 uint32_t	m_ether_tcpip_hash(const uint32_t, const struct mbuf *, const uint32_t);
1197 
1198 #ifdef MBUF_PROFILING
1199  void m_profile(struct mbuf *m);
1200  #define M_PROFILE(m) m_profile(m)
1201 #else
1202  #define M_PROFILE(m)
1203 #endif
1204 
1205 struct mbufq {
1206 	STAILQ_HEAD(, mbuf)	mq_head;
1207 	int			mq_len;
1208 	int			mq_maxlen;
1209 };
1210 
1211 static inline void
1212 mbufq_init(struct mbufq *mq, int maxlen)
1213 {
1214 
1215 	STAILQ_INIT(&mq->mq_head);
1216 	mq->mq_maxlen = maxlen;
1217 	mq->mq_len = 0;
1218 }
1219 
1220 static inline struct mbuf *
1221 mbufq_flush(struct mbufq *mq)
1222 {
1223 	struct mbuf *m;
1224 
1225 	m = STAILQ_FIRST(&mq->mq_head);
1226 	STAILQ_INIT(&mq->mq_head);
1227 	mq->mq_len = 0;
1228 	return (m);
1229 }
1230 
1231 static inline void
1232 mbufq_drain(struct mbufq *mq)
1233 {
1234 	struct mbuf *m, *n;
1235 
1236 	n = mbufq_flush(mq);
1237 	while ((m = n) != NULL) {
1238 		n = STAILQ_NEXT(m, m_stailqpkt);
1239 		m_freem(m);
1240 	}
1241 }
1242 
1243 static inline struct mbuf *
1244 mbufq_first(const struct mbufq *mq)
1245 {
1246 
1247 	return (STAILQ_FIRST(&mq->mq_head));
1248 }
1249 
1250 static inline struct mbuf *
1251 mbufq_last(const struct mbufq *mq)
1252 {
1253 
1254 	return (STAILQ_LAST(&mq->mq_head, mbuf, m_stailqpkt));
1255 }
1256 
1257 static inline int
1258 mbufq_full(const struct mbufq *mq)
1259 {
1260 
1261 	return (mq->mq_len >= mq->mq_maxlen);
1262 }
1263 
1264 static inline int
1265 mbufq_len(const struct mbufq *mq)
1266 {
1267 
1268 	return (mq->mq_len);
1269 }
1270 
1271 static inline int
1272 mbufq_enqueue(struct mbufq *mq, struct mbuf *m)
1273 {
1274 
1275 	if (mbufq_full(mq))
1276 		return (ENOBUFS);
1277 	STAILQ_INSERT_TAIL(&mq->mq_head, m, m_stailqpkt);
1278 	mq->mq_len++;
1279 	return (0);
1280 }
1281 
1282 static inline struct mbuf *
1283 mbufq_dequeue(struct mbufq *mq)
1284 {
1285 	struct mbuf *m;
1286 
1287 	m = STAILQ_FIRST(&mq->mq_head);
1288 	if (m) {
1289 		STAILQ_REMOVE_HEAD(&mq->mq_head, m_stailqpkt);
1290 		m->m_nextpkt = NULL;
1291 		mq->mq_len--;
1292 	}
1293 	return (m);
1294 }
1295 
1296 static inline void
1297 mbufq_prepend(struct mbufq *mq, struct mbuf *m)
1298 {
1299 
1300 	STAILQ_INSERT_HEAD(&mq->mq_head, m, m_stailqpkt);
1301 	mq->mq_len++;
1302 }
1303 #endif /* _KERNEL */
1304 #endif /* !_SYS_MBUF_H_ */
1305