xref: /freebsd/sys/sys/mbuf.h (revision b197d4b893974c9eb4d7b38704c6d5c486235d6f)
1 /*-
2  * SPDX-License-Identifier: BSD-3-Clause
3  *
4  * Copyright (c) 1982, 1986, 1988, 1993
5  *	The Regents of the University of California.
6  * All rights reserved.
7  *
8  * Redistribution and use in source and binary forms, with or without
9  * modification, are permitted provided that the following conditions
10  * are met:
11  * 1. Redistributions of source code must retain the above copyright
12  *    notice, this list of conditions and the following disclaimer.
13  * 2. Redistributions in binary form must reproduce the above copyright
14  *    notice, this list of conditions and the following disclaimer in the
15  *    documentation and/or other materials provided with the distribution.
16  * 3. Neither the name of the University nor the names of its contributors
17  *    may be used to endorse or promote products derived from this software
18  *    without specific prior written permission.
19  *
20  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
21  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
22  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
23  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
24  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
25  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
26  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
27  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
28  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
29  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
30  * SUCH DAMAGE.
31  *
32  *	@(#)mbuf.h	8.5 (Berkeley) 2/19/95
33  * $FreeBSD$
34  */
35 
36 #ifndef _SYS_MBUF_H_
37 #define	_SYS_MBUF_H_
38 
39 /* XXX: These includes suck. Sorry! */
40 #include <sys/queue.h>
41 #ifdef _KERNEL
42 #include <sys/systm.h>
43 #include <sys/refcount.h>
44 #include <vm/uma.h>
45 #ifdef WITNESS
46 #include <sys/lock.h>
47 #endif
48 #endif
49 
50 #ifdef _KERNEL
51 #include <sys/sdt.h>
52 
53 #define	MBUF_PROBE1(probe, arg0)					\
54 	SDT_PROBE1(sdt, , , probe, arg0)
55 #define	MBUF_PROBE2(probe, arg0, arg1)					\
56 	SDT_PROBE2(sdt, , , probe, arg0, arg1)
57 #define	MBUF_PROBE3(probe, arg0, arg1, arg2)				\
58 	SDT_PROBE3(sdt, , , probe, arg0, arg1, arg2)
59 #define	MBUF_PROBE4(probe, arg0, arg1, arg2, arg3)			\
60 	SDT_PROBE4(sdt, , , probe, arg0, arg1, arg2, arg3)
61 #define	MBUF_PROBE5(probe, arg0, arg1, arg2, arg3, arg4)		\
62 	SDT_PROBE5(sdt, , , probe, arg0, arg1, arg2, arg3, arg4)
63 
64 SDT_PROBE_DECLARE(sdt, , , m__init);
65 SDT_PROBE_DECLARE(sdt, , , m__gethdr_raw);
66 SDT_PROBE_DECLARE(sdt, , , m__gethdr);
67 SDT_PROBE_DECLARE(sdt, , , m__get_raw);
68 SDT_PROBE_DECLARE(sdt, , , m__get);
69 SDT_PROBE_DECLARE(sdt, , , m__getcl);
70 SDT_PROBE_DECLARE(sdt, , , m__getjcl);
71 SDT_PROBE_DECLARE(sdt, , , m__clget);
72 SDT_PROBE_DECLARE(sdt, , , m__cljget);
73 SDT_PROBE_DECLARE(sdt, , , m__cljset);
74 SDT_PROBE_DECLARE(sdt, , , m__free);
75 SDT_PROBE_DECLARE(sdt, , , m__freem);
76 
77 #endif /* _KERNEL */
78 
79 /*
80  * Mbufs are of a single size, MSIZE (sys/param.h), which includes overhead.
81  * An mbuf may add a single "mbuf cluster" of size MCLBYTES (also in
82  * sys/param.h), which has no additional overhead and is used instead of the
83  * internal data area; this is done when at least MINCLSIZE of data must be
84  * stored.  Additionally, it is possible to allocate a separate buffer
85  * externally and attach it to the mbuf in a way similar to that of mbuf
86  * clusters.
87  *
88  * NB: These calculation do not take actual compiler-induced alignment and
89  * padding inside the complete struct mbuf into account.  Appropriate
90  * attention is required when changing members of struct mbuf.
91  *
92  * MLEN is data length in a normal mbuf.
93  * MHLEN is data length in an mbuf with pktheader.
94  * MINCLSIZE is a smallest amount of data that should be put into cluster.
95  *
96  * Compile-time assertions in uipc_mbuf.c test these values to ensure that
97  * they are sensible.
98  */
99 struct mbuf;
100 #define	MHSIZE		offsetof(struct mbuf, m_dat)
101 #define	MPKTHSIZE	offsetof(struct mbuf, m_pktdat)
102 #define	MLEN		((int)(MSIZE - MHSIZE))
103 #define	MHLEN		((int)(MSIZE - MPKTHSIZE))
104 #define	MINCLSIZE	(MHLEN + 1)
105 #define	M_NODOM		255
106 
107 #ifdef _KERNEL
108 /*-
109  * Macro for type conversion: convert mbuf pointer to data pointer of correct
110  * type:
111  *
112  * mtod(m, t)	-- Convert mbuf pointer to data pointer of correct type.
113  * mtodo(m, o) -- Same as above but with offset 'o' into data.
114  */
115 #define	mtod(m, t)	((t)((m)->m_data))
116 #define	mtodo(m, o)	((void *)(((m)->m_data) + (o)))
117 
118 /*
119  * Argument structure passed to UMA routines during mbuf and packet
120  * allocations.
121  */
122 struct mb_args {
123 	int	flags;	/* Flags for mbuf being allocated */
124 	short	type;	/* Type of mbuf being allocated */
125 };
126 #endif /* _KERNEL */
127 
128 /*
129  * Packet tag structure (see below for details).
130  */
131 struct m_tag {
132 	SLIST_ENTRY(m_tag)	m_tag_link;	/* List of packet tags */
133 	u_int16_t		m_tag_id;	/* Tag ID */
134 	u_int16_t		m_tag_len;	/* Length of data */
135 	u_int32_t		m_tag_cookie;	/* ABI/Module ID */
136 	void			(*m_tag_free)(struct m_tag *);
137 };
138 
139 /*
140  * Static network interface owned tag.
141  * Allocated through ifp->if_snd_tag_alloc().
142  */
143 struct if_snd_tag_sw;
144 
145 struct m_snd_tag {
146 	struct ifnet *ifp;		/* network interface tag belongs to */
147 	const struct if_snd_tag_sw *sw;
148 	volatile u_int refcount;
149 };
150 
151 /*
152  * Record/packet header in first mbuf of chain; valid only if M_PKTHDR is set.
153  * Size ILP32: 56
154  *	 LP64: 64
155  * Compile-time assertions in uipc_mbuf.c test these values to ensure that
156  * they are correct.
157  */
158 struct pkthdr {
159 	union {
160 		struct m_snd_tag *snd_tag;	/* send tag, if any */
161 		struct ifnet	*rcvif;		/* rcv interface */
162 		struct {
163 			uint16_t rcvidx;	/* rcv interface index ... */
164 			uint16_t rcvgen;	/* ... and generation count */
165 		};
166 	};
167 	union {
168 		struct ifnet	*leaf_rcvif;	/* leaf rcv interface */
169 		struct {
170 			uint16_t leaf_rcvidx;	/* leaf rcv interface index ... */
171 			uint16_t leaf_rcvgen;	/* ... and generation count */
172 		};
173 	};
174 	SLIST_HEAD(packet_tags, m_tag) tags; /* list of packet tags */
175 	int32_t		 len;		/* total packet length */
176 
177 	/* Layer crossing persistent information. */
178 	uint32_t	 flowid;	/* packet's 4-tuple system */
179 	uint32_t	 csum_flags;	/* checksum and offload features */
180 	uint16_t	 fibnum;	/* this packet should use this fib */
181 	uint8_t		 numa_domain;	/* NUMA domain of recvd pkt */
182 	uint8_t		 rsstype;	/* hash type */
183 #if !defined(__LP64__)
184 	uint32_t	 pad;		/* pad for 64bit alignment */
185 #endif
186 	union {
187 		uint64_t	rcv_tstmp;	/* timestamp in ns */
188 		struct {
189 			uint8_t		 l2hlen;	/* layer 2 hdr len */
190 			uint8_t		 l3hlen;	/* layer 3 hdr len */
191 			uint8_t		 l4hlen;	/* layer 4 hdr len */
192 			uint8_t		 l5hlen;	/* layer 5 hdr len */
193 			uint8_t		 inner_l2hlen;
194 			uint8_t		 inner_l3hlen;
195 			uint8_t		 inner_l4hlen;
196 			uint8_t		 inner_l5hlen;
197 		};
198 	};
199 	union {
200 		uint8_t  eight[8];
201 		uint16_t sixteen[4];
202 		uint32_t thirtytwo[2];
203 		uint64_t sixtyfour[1];
204 		uintptr_t unintptr[1];
205 		void	*ptr;
206 	} PH_per;
207 
208 	/* Layer specific non-persistent local storage for reassembly, etc. */
209 	union {
210 		union {
211 			uint8_t  eight[8];
212 			uint16_t sixteen[4];
213 			uint32_t thirtytwo[2];
214 			uint64_t sixtyfour[1];
215 			uintptr_t unintptr[1];
216 			void 	*ptr;
217 		} PH_loc;
218 		/* Upon allocation: total packet memory consumption. */
219 		u_int	memlen;
220 	};
221 };
222 #define	ether_vtag	PH_per.sixteen[0]
223 #define tcp_tun_port	PH_per.sixteen[0] /* outbound */
224 #define	vt_nrecs	PH_per.sixteen[0]	  /* mld and v6-ND */
225 #define	tso_segsz	PH_per.sixteen[1] /* inbound after LRO */
226 #define	lro_nsegs	tso_segsz	  /* inbound after LRO */
227 #define	csum_data	PH_per.thirtytwo[1] /* inbound from hardware up */
228 #define	lro_tcp_d_len	PH_loc.sixteen[0] /* inbound during LRO (no reassembly) */
229 #define	lro_tcp_d_csum	PH_loc.sixteen[1] /* inbound during LRO (no reassembly) */
230 #define	lro_tcp_h_off	PH_loc.sixteen[2] /* inbound during LRO (no reassembly) */
231 #define	lro_etype	PH_loc.sixteen[3] /* inbound during LRO (no reassembly) */
232 /* Note PH_loc is used during IP reassembly (all 8 bytes as a ptr) */
233 
234 /*
235  * TLS records for TLS 1.0-1.2 can have the following header lengths:
236  * - 5 (AES-CBC with implicit IV)
237  * - 21 (AES-CBC with explicit IV)
238  * - 13 (AES-GCM with 8 byte explicit IV)
239  */
240 #define	MBUF_PEXT_HDR_LEN	23
241 
242 /*
243  * TLS records for TLS 1.0-1.2 can have the following maximum trailer
244  * lengths:
245  * - 16 (AES-GCM)
246  * - 36 (AES-CBC with SHA1 and up to 16 bytes of padding)
247  * - 48 (AES-CBC with SHA2-256 and up to 16 bytes of padding)
248  * - 64 (AES-CBC with SHA2-384 and up to 16 bytes of padding)
249  */
250 #define	MBUF_PEXT_TRAIL_LEN	64
251 
252 #if defined(__LP64__)
253 #define MBUF_PEXT_MAX_PGS (40 / sizeof(vm_paddr_t))
254 #else
255 #define MBUF_PEXT_MAX_PGS (64 / sizeof(vm_paddr_t))
256 #endif
257 
258 #define	MBUF_PEXT_MAX_BYTES						\
259     (MBUF_PEXT_MAX_PGS * PAGE_SIZE + MBUF_PEXT_HDR_LEN + MBUF_PEXT_TRAIL_LEN)
260 
261 struct ktls_session;
262 struct socket;
263 
264 /*
265  * Description of external storage mapped into mbuf; valid only if M_EXT is
266  * set.
267  * Size ILP32: 28
268  *	 LP64: 48
269  * Compile-time assertions in uipc_mbuf.c test these values to ensure that
270  * they are correct.
271  */
272 typedef	void m_ext_free_t(struct mbuf *);
273 struct m_ext {
274 	union {
275 		/*
276 		 * If EXT_FLAG_EMBREF is set, then we use refcount in the
277 		 * mbuf, the 'ext_count' member.  Otherwise, we have a
278 		 * shadow copy and we use pointer 'ext_cnt'.  The original
279 		 * mbuf is responsible to carry the pointer to free routine
280 		 * and its arguments.  They aren't copied into shadows in
281 		 * mb_dupcl() to avoid dereferencing next cachelines.
282 		 */
283 		volatile u_int	 ext_count;
284 		volatile u_int	*ext_cnt;
285 	};
286 	uint32_t	 ext_size;	/* size of buffer, for ext_free */
287 	uint32_t	 ext_type:8,	/* type of external storage */
288 			 ext_flags:24;	/* external storage mbuf flags */
289 	union {
290 		struct {
291 			/*
292 			 * Regular M_EXT mbuf:
293 			 * o ext_buf always points to the external buffer.
294 			 * o ext_free (below) and two optional arguments
295 			 *   ext_arg1 and ext_arg2 store the free context for
296 			 *   the external storage.  They are set only in the
297 			 *   refcount carrying mbuf, the one with
298 			 *   EXT_FLAG_EMBREF flag, with exclusion for
299 			 *   EXT_EXTREF type, where the free context is copied
300 			 *   into all mbufs that use same external storage.
301 			 */
302 			char 	*ext_buf;	/* start of buffer */
303 #define	m_ext_copylen	offsetof(struct m_ext, ext_arg2)
304 			void	*ext_arg2;
305 		};
306 		struct {
307 			/*
308 			 * Multi-page M_EXTPG mbuf:
309 			 * o extpg_pa - page vector.
310 			 * o extpg_trail and extpg_hdr - TLS trailer and
311 			 *   header.
312 			 * Uses ext_free and may also use ext_arg1.
313 			 */
314 			vm_paddr_t	extpg_pa[MBUF_PEXT_MAX_PGS];
315 			char		extpg_trail[MBUF_PEXT_TRAIL_LEN];
316 			char		extpg_hdr[MBUF_PEXT_HDR_LEN];
317 			/* Pretend these 3 fields are part of mbuf itself. */
318 #define	m_epg_pa	m_ext.extpg_pa
319 #define	m_epg_trail	m_ext.extpg_trail
320 #define	m_epg_hdr	m_ext.extpg_hdr
321 #define	m_epg_ext_copylen	offsetof(struct m_ext, ext_free)
322 		};
323 	};
324 	/*
325 	 * Free method and optional argument pointer, both
326 	 * used by M_EXT and M_EXTPG.
327 	 */
328 	m_ext_free_t	*ext_free;
329 	void		*ext_arg1;
330 };
331 
332 /*
333  * The core of the mbuf object along with some shortcut defines for practical
334  * purposes.
335  */
336 struct mbuf {
337 	/*
338 	 * Header present at the beginning of every mbuf.
339 	 * Size ILP32: 24
340 	 *      LP64: 32
341 	 * Compile-time assertions in uipc_mbuf.c test these values to ensure
342 	 * that they are correct.
343 	 */
344 	union {	/* next buffer in chain */
345 		struct mbuf		*m_next;
346 		SLIST_ENTRY(mbuf)	m_slist;
347 		STAILQ_ENTRY(mbuf)	m_stailq;
348 	};
349 	union {	/* next chain in queue/record */
350 		struct mbuf		*m_nextpkt;
351 		SLIST_ENTRY(mbuf)	m_slistpkt;
352 		STAILQ_ENTRY(mbuf)	m_stailqpkt;
353 	};
354 	caddr_t		 m_data;	/* location of data */
355 	int32_t		 m_len;		/* amount of data in this mbuf */
356 	uint32_t	 m_type:8,	/* type of data in this mbuf */
357 			 m_flags:24;	/* flags; see below */
358 #if !defined(__LP64__)
359 	uint32_t	 m_pad;		/* pad for 64bit alignment */
360 #endif
361 
362 	/*
363 	 * A set of optional headers (packet header, external storage header)
364 	 * and internal data storage.  Historically, these arrays were sized
365 	 * to MHLEN (space left after a packet header) and MLEN (space left
366 	 * after only a regular mbuf header); they are now variable size in
367 	 * order to support future work on variable-size mbufs.
368 	 */
369 	union {
370 		struct {
371 			union {
372 				/* M_PKTHDR set. */
373 				struct pkthdr	m_pkthdr;
374 
375 				/* M_EXTPG set.
376 				 * Multi-page M_EXTPG mbuf has its meta data
377 				 * split between the below anonymous structure
378 				 * and m_ext.  It carries vector of pages,
379 				 * optional header and trailer char vectors
380 				 * and pointers to socket/TLS data.
381 				 */
382 #define	m_epg_startcopy		m_epg_npgs
383 #define	m_epg_endcopy		m_epg_stailq
384 				struct {
385 					/* Overall count of pages and count of
386 					 * pages with I/O pending. */
387 					uint8_t	m_epg_npgs;
388 					uint8_t	m_epg_nrdy;
389 					/* TLS header and trailer lengths.
390 					 * The data itself resides in m_ext. */
391 					uint8_t	m_epg_hdrlen;
392 					uint8_t	m_epg_trllen;
393 					/* Offset into 1st page and length of
394 					 * data in the last page. */
395 					uint16_t m_epg_1st_off;
396 					uint16_t m_epg_last_len;
397 					uint8_t	m_epg_flags;
398 #define	EPG_FLAG_ANON	0x1	/* Data can be encrypted in place. */
399 #define	EPG_FLAG_2FREE	0x2	/* Scheduled for free. */
400 					uint8_t	m_epg_record_type;
401 					uint8_t	__spare[2];
402 					int	m_epg_enc_cnt;
403 					struct ktls_session *m_epg_tls;
404 					struct socket	*m_epg_so;
405 					uint64_t	m_epg_seqno;
406 					STAILQ_ENTRY(mbuf) m_epg_stailq;
407 				};
408 			};
409 			union {
410 				/* M_EXT or M_EXTPG set. */
411 				struct m_ext	m_ext;
412 				/* M_PKTHDR set, neither M_EXT nor M_EXTPG. */
413 				char		m_pktdat[0];
414 			};
415 		};
416 		char	m_dat[0];			/* !M_PKTHDR, !M_EXT */
417 	};
418 };
419 
420 #ifdef _KERNEL
421 static inline int
422 m_epg_pagelen(const struct mbuf *m, int pidx, int pgoff)
423 {
424 
425 	KASSERT(pgoff == 0 || pidx == 0,
426 	    ("page %d with non-zero offset %d in %p", pidx, pgoff, m));
427 
428 	if (pidx == m->m_epg_npgs - 1) {
429 		return (m->m_epg_last_len);
430 	} else {
431 		return (PAGE_SIZE - pgoff);
432 	}
433 }
434 
435 #ifdef INVARIANTS
436 #define	MCHECK(ex, msg)	KASSERT((ex),				\
437 	    ("Multi page mbuf %p with " #msg " at %s:%d",	\
438 	    m, __FILE__, __LINE__))
439 /*
440  * NB: This expects a non-empty buffer (npgs > 0 and
441  * last_pg_len > 0).
442  */
443 #define	MBUF_EXT_PGS_ASSERT_SANITY(m)	do {				\
444 	MCHECK(m->m_epg_npgs > 0, "no valid pages");		\
445 	MCHECK(m->m_epg_npgs <= nitems(m->m_epg_pa),		\
446 	    "too many pages");						\
447 	MCHECK(m->m_epg_nrdy <= m->m_epg_npgs,			\
448 	    "too many ready pages");					\
449 	MCHECK(m->m_epg_1st_off < PAGE_SIZE,			\
450 		"too large page offset");				\
451 	MCHECK(m->m_epg_last_len > 0, "zero last page length");	\
452 	MCHECK(m->m_epg_last_len <= PAGE_SIZE,			\
453 	    "too large last page length");				\
454 	if (m->m_epg_npgs == 1)					\
455 		MCHECK(m->m_epg_1st_off +			\
456 		    m->m_epg_last_len <=	 PAGE_SIZE,		\
457 		    "single page too large");				\
458 	MCHECK(m->m_epg_hdrlen <= sizeof(m->m_epg_hdr),		\
459 	    "too large header length");					\
460 	MCHECK(m->m_epg_trllen <= sizeof(m->m_epg_trail),	\
461 	    "too large header length");					\
462 } while (0)
463 #else
464 #define	MBUF_EXT_PGS_ASSERT_SANITY(m)	do {} while (0)
465 #endif
466 #endif
467 
468 /*
469  * mbuf flags of global significance and layer crossing.
470  * Those of only protocol/layer specific significance are to be mapped
471  * to M_PROTO[1-11] and cleared at layer handoff boundaries.
472  * NB: Limited to the lower 24 bits.
473  */
474 #define	M_EXT		0x00000001 /* has associated external storage */
475 #define	M_PKTHDR	0x00000002 /* start of record */
476 #define	M_EOR		0x00000004 /* end of record */
477 #define	M_RDONLY	0x00000008 /* associated data is marked read-only */
478 #define	M_BCAST		0x00000010 /* send/received as link-level broadcast */
479 #define	M_MCAST		0x00000020 /* send/received as link-level multicast */
480 #define	M_PROMISC	0x00000040 /* packet was not for us */
481 #define	M_VLANTAG	0x00000080 /* ether_vtag is valid */
482 #define	M_EXTPG		0x00000100 /* has array of unmapped pages and TLS */
483 #define	M_NOFREE	0x00000200 /* do not free mbuf, embedded in cluster */
484 #define	M_TSTMP		0x00000400 /* rcv_tstmp field is valid */
485 #define	M_TSTMP_HPREC	0x00000800 /* rcv_tstmp is high-prec, typically
486 				      hw-stamped on port (useful for IEEE 1588
487 				      and 802.1AS) */
488 #define M_TSTMP_LRO	0x00001000 /* Time LRO pushed in pkt is valid in (PH_loc) */
489 
490 #define	M_PROTO1	0x00002000 /* protocol-specific */
491 #define	M_PROTO2	0x00004000 /* protocol-specific */
492 #define	M_PROTO3	0x00008000 /* protocol-specific */
493 #define	M_PROTO4	0x00010000 /* protocol-specific */
494 #define	M_PROTO5	0x00020000 /* protocol-specific */
495 #define	M_PROTO6	0x00040000 /* protocol-specific */
496 #define	M_PROTO7	0x00080000 /* protocol-specific */
497 #define	M_PROTO8	0x00100000 /* protocol-specific */
498 #define	M_PROTO9	0x00200000 /* protocol-specific */
499 #define	M_PROTO10	0x00400000 /* protocol-specific */
500 #define	M_PROTO11	0x00800000 /* protocol-specific */
501 
502 /*
503  * Flags to purge when crossing layers.
504  */
505 #define	M_PROTOFLAGS \
506     (M_PROTO1|M_PROTO2|M_PROTO3|M_PROTO4|M_PROTO5|M_PROTO6|M_PROTO7|M_PROTO8|\
507      M_PROTO9|M_PROTO10|M_PROTO11)
508 
509 /*
510  * Flags preserved when copying m_pkthdr.
511  */
512 #define M_COPYFLAGS \
513     (M_PKTHDR|M_EOR|M_RDONLY|M_BCAST|M_MCAST|M_PROMISC|M_VLANTAG|M_TSTMP| \
514      M_TSTMP_HPREC|M_TSTMP_LRO|M_PROTOFLAGS)
515 
516 /*
517  * Flags preserved during demote.
518  */
519 #define	M_DEMOTEFLAGS \
520     (M_EXT | M_RDONLY | M_NOFREE | M_EXTPG)
521 
522 /*
523  * Mbuf flag description for use with printf(9) %b identifier.
524  */
525 #define	M_FLAG_BITS \
526     "\20\1M_EXT\2M_PKTHDR\3M_EOR\4M_RDONLY\5M_BCAST\6M_MCAST" \
527     "\7M_PROMISC\10M_VLANTAG\11M_EXTPG\12M_NOFREE\13M_TSTMP\14M_TSTMP_HPREC\15M_TSTMP_LRO"
528 #define	M_FLAG_PROTOBITS \
529     "\16M_PROTO1\17M_PROTO2\20M_PROTO3\21M_PROTO4" \
530     "\22M_PROTO5\23M_PROTO6\24M_PROTO7\25M_PROTO8\26M_PROTO9" \
531     "\27M_PROTO10\28M_PROTO11"
532 #define	M_FLAG_PRINTF (M_FLAG_BITS M_FLAG_PROTOBITS)
533 
534 /*
535  * Network interface cards are able to hash protocol fields (such as IPv4
536  * addresses and TCP port numbers) classify packets into flows.  These flows
537  * can then be used to maintain ordering while delivering packets to the OS
538  * via parallel input queues, as well as to provide a stateless affinity
539  * model.  NIC drivers can pass up the hash via m->m_pkthdr.flowid, and set
540  * m_flag fields to indicate how the hash should be interpreted by the
541  * network stack.
542  *
543  * Most NICs support RSS, which provides ordering and explicit affinity, and
544  * use the hash m_flag bits to indicate what header fields were covered by
545  * the hash.  M_HASHTYPE_OPAQUE and M_HASHTYPE_OPAQUE_HASH can be set by non-
546  * RSS cards or configurations that provide an opaque flow identifier, allowing
547  * for ordering and distribution without explicit affinity.  Additionally,
548  * M_HASHTYPE_OPAQUE_HASH indicates that the flow identifier has hash
549  * properties.
550  *
551  * The meaning of the IPV6_EX suffix:
552  * "o  Home address from the home address option in the IPv6 destination
553  *     options header.  If the extension header is not present, use the Source
554  *     IPv6 Address.
555  *  o  IPv6 address that is contained in the Routing-Header-Type-2 from the
556  *     associated extension header.  If the extension header is not present,
557  *     use the Destination IPv6 Address."
558  * Quoted from:
559  * https://docs.microsoft.com/en-us/windows-hardware/drivers/network/rss-hashing-types#ndishashipv6ex
560  */
561 #define	M_HASHTYPE_HASHPROP		0x80	/* has hash properties */
562 #define	M_HASHTYPE_INNER		0x40	/* calculated from inner headers */
563 #define	M_HASHTYPE_HASH(t)		(M_HASHTYPE_HASHPROP | (t))
564 /* Microsoft RSS standard hash types */
565 #define	M_HASHTYPE_NONE			0
566 #define	M_HASHTYPE_RSS_IPV4		M_HASHTYPE_HASH(1) /* IPv4 2-tuple */
567 #define	M_HASHTYPE_RSS_TCP_IPV4		M_HASHTYPE_HASH(2) /* TCPv4 4-tuple */
568 #define	M_HASHTYPE_RSS_IPV6		M_HASHTYPE_HASH(3) /* IPv6 2-tuple */
569 #define	M_HASHTYPE_RSS_TCP_IPV6		M_HASHTYPE_HASH(4) /* TCPv6 4-tuple */
570 #define	M_HASHTYPE_RSS_IPV6_EX		M_HASHTYPE_HASH(5) /* IPv6 2-tuple +
571 							    * ext hdrs */
572 #define	M_HASHTYPE_RSS_TCP_IPV6_EX	M_HASHTYPE_HASH(6) /* TCPv6 4-tuple +
573 							    * ext hdrs */
574 #define	M_HASHTYPE_RSS_UDP_IPV4		M_HASHTYPE_HASH(7) /* IPv4 UDP 4-tuple*/
575 #define	M_HASHTYPE_RSS_UDP_IPV6		M_HASHTYPE_HASH(9) /* IPv6 UDP 4-tuple*/
576 #define	M_HASHTYPE_RSS_UDP_IPV6_EX	M_HASHTYPE_HASH(10)/* IPv6 UDP 4-tuple +
577 							    * ext hdrs */
578 
579 #define	M_HASHTYPE_OPAQUE		0x3f	/* ordering, not affinity */
580 #define	M_HASHTYPE_OPAQUE_HASH		M_HASHTYPE_HASH(M_HASHTYPE_OPAQUE)
581 						/* ordering+hash, not affinity*/
582 
583 #define	M_HASHTYPE_CLEAR(m)	((m)->m_pkthdr.rsstype = 0)
584 #define	M_HASHTYPE_GET(m)	((m)->m_pkthdr.rsstype & ~M_HASHTYPE_INNER)
585 #define	M_HASHTYPE_SET(m, v)	((m)->m_pkthdr.rsstype = (v))
586 #define	M_HASHTYPE_TEST(m, v)	(M_HASHTYPE_GET(m) == (v))
587 #define	M_HASHTYPE_ISHASH(m)	\
588     (((m)->m_pkthdr.rsstype & M_HASHTYPE_HASHPROP) != 0)
589 #define	M_HASHTYPE_SETINNER(m)	do {			\
590 	(m)->m_pkthdr.rsstype |= M_HASHTYPE_INNER;	\
591     } while (0)
592 
593 /*
594  * External mbuf storage buffer types.
595  */
596 #define	EXT_CLUSTER	1	/* mbuf cluster */
597 #define	EXT_SFBUF	2	/* sendfile(2)'s sf_buf */
598 #define	EXT_JUMBOP	3	/* jumbo cluster page sized */
599 #define	EXT_JUMBO9	4	/* jumbo cluster 9216 bytes */
600 #define	EXT_JUMBO16	5	/* jumbo cluster 16184 bytes */
601 #define	EXT_PACKET	6	/* mbuf+cluster from packet zone */
602 #define	EXT_MBUF	7	/* external mbuf reference */
603 #define	EXT_RXRING	8	/* data in NIC receive ring */
604 
605 #define	EXT_VENDOR1	224	/* for vendor-internal use */
606 #define	EXT_VENDOR2	225	/* for vendor-internal use */
607 #define	EXT_VENDOR3	226	/* for vendor-internal use */
608 #define	EXT_VENDOR4	227	/* for vendor-internal use */
609 
610 #define	EXT_EXP1	244	/* for experimental use */
611 #define	EXT_EXP2	245	/* for experimental use */
612 #define	EXT_EXP3	246	/* for experimental use */
613 #define	EXT_EXP4	247	/* for experimental use */
614 
615 #define	EXT_NET_DRV	252	/* custom ext_buf provided by net driver(s) */
616 #define	EXT_MOD_TYPE	253	/* custom module's ext_buf type */
617 #define	EXT_DISPOSABLE	254	/* can throw this buffer away w/page flipping */
618 #define	EXT_EXTREF	255	/* has externally maintained ext_cnt ptr */
619 
620 /*
621  * Flags for external mbuf buffer types.
622  * NB: limited to the lower 24 bits.
623  */
624 #define	EXT_FLAG_EMBREF		0x000001	/* embedded ext_count */
625 #define	EXT_FLAG_EXTREF		0x000002	/* external ext_cnt, notyet */
626 
627 #define	EXT_FLAG_NOFREE		0x000010	/* don't free mbuf to pool, notyet */
628 
629 #define	EXT_FLAG_VENDOR1	0x010000	/* These flags are vendor */
630 #define	EXT_FLAG_VENDOR2	0x020000	/* or submodule specific, */
631 #define	EXT_FLAG_VENDOR3	0x040000	/* not used by mbuf code. */
632 #define	EXT_FLAG_VENDOR4	0x080000	/* Set/read by submodule. */
633 
634 #define	EXT_FLAG_EXP1		0x100000	/* for experimental use */
635 #define	EXT_FLAG_EXP2		0x200000	/* for experimental use */
636 #define	EXT_FLAG_EXP3		0x400000	/* for experimental use */
637 #define	EXT_FLAG_EXP4		0x800000	/* for experimental use */
638 
639 /*
640  * EXT flag description for use with printf(9) %b identifier.
641  */
642 #define	EXT_FLAG_BITS \
643     "\20\1EXT_FLAG_EMBREF\2EXT_FLAG_EXTREF\5EXT_FLAG_NOFREE" \
644     "\21EXT_FLAG_VENDOR1\22EXT_FLAG_VENDOR2\23EXT_FLAG_VENDOR3" \
645     "\24EXT_FLAG_VENDOR4\25EXT_FLAG_EXP1\26EXT_FLAG_EXP2\27EXT_FLAG_EXP3" \
646     "\30EXT_FLAG_EXP4"
647 
648 /*
649  * Flags indicating checksum, segmentation and other offload work to be
650  * done, or already done, by hardware or lower layers.  It is split into
651  * separate inbound and outbound flags.
652  *
653  * Outbound flags that are set by upper protocol layers requesting lower
654  * layers, or ideally the hardware, to perform these offloading tasks.
655  * For outbound packets this field and its flags can be directly tested
656  * against ifnet if_hwassist.  Note that the outbound and the inbound flags do
657  * not collide right now but they could be allowed to (as long as the flags are
658  * scrubbed appropriately when the direction of an mbuf changes).  CSUM_BITS
659  * would also have to split into CSUM_BITS_TX and CSUM_BITS_RX.
660  *
661  * CSUM_INNER_<x> is the same as CSUM_<x> but it applies to the inner frame.
662  * The CSUM_ENCAP_<x> bits identify the outer encapsulation.
663  */
664 #define	CSUM_IP			0x00000001	/* IP header checksum offload */
665 #define	CSUM_IP_UDP		0x00000002	/* UDP checksum offload */
666 #define	CSUM_IP_TCP		0x00000004	/* TCP checksum offload */
667 #define	CSUM_IP_SCTP		0x00000008	/* SCTP checksum offload */
668 #define	CSUM_IP_TSO		0x00000010	/* TCP segmentation offload */
669 #define	CSUM_IP_ISCSI		0x00000020	/* iSCSI checksum offload */
670 
671 #define	CSUM_INNER_IP6_UDP	0x00000040
672 #define	CSUM_INNER_IP6_TCP	0x00000080
673 #define	CSUM_INNER_IP6_TSO	0x00000100
674 #define	CSUM_IP6_UDP		0x00000200	/* UDP checksum offload */
675 #define	CSUM_IP6_TCP		0x00000400	/* TCP checksum offload */
676 #define	CSUM_IP6_SCTP		0x00000800	/* SCTP checksum offload */
677 #define	CSUM_IP6_TSO		0x00001000	/* TCP segmentation offload */
678 #define	CSUM_IP6_ISCSI		0x00002000	/* iSCSI checksum offload */
679 
680 #define	CSUM_INNER_IP		0x00004000
681 #define	CSUM_INNER_IP_UDP	0x00008000
682 #define	CSUM_INNER_IP_TCP	0x00010000
683 #define	CSUM_INNER_IP_TSO	0x00020000
684 
685 #define	CSUM_ENCAP_VXLAN	0x00040000	/* VXLAN outer encapsulation */
686 #define	CSUM_ENCAP_RSVD1	0x00080000
687 
688 /* Inbound checksum support where the checksum was verified by hardware. */
689 #define	CSUM_INNER_L3_CALC	0x00100000
690 #define	CSUM_INNER_L3_VALID	0x00200000
691 #define	CSUM_INNER_L4_CALC	0x00400000
692 #define	CSUM_INNER_L4_VALID	0x00800000
693 #define	CSUM_L3_CALC		0x01000000	/* calculated layer 3 csum */
694 #define	CSUM_L3_VALID		0x02000000	/* checksum is correct */
695 #define	CSUM_L4_CALC		0x04000000	/* calculated layer 4 csum */
696 #define	CSUM_L4_VALID		0x08000000	/* checksum is correct */
697 #define	CSUM_L5_CALC		0x10000000	/* calculated layer 5 csum */
698 #define	CSUM_L5_VALID		0x20000000	/* checksum is correct */
699 #define	CSUM_COALESCED		0x40000000	/* contains merged segments */
700 
701 #define	CSUM_SND_TAG		0x80000000	/* Packet header has send tag */
702 
703 #define CSUM_FLAGS_TX (CSUM_IP | CSUM_IP_UDP | CSUM_IP_TCP | CSUM_IP_SCTP | \
704     CSUM_IP_TSO | CSUM_IP_ISCSI | CSUM_INNER_IP6_UDP | CSUM_INNER_IP6_TCP | \
705     CSUM_INNER_IP6_TSO | CSUM_IP6_UDP | CSUM_IP6_TCP | CSUM_IP6_SCTP | \
706     CSUM_IP6_TSO | CSUM_IP6_ISCSI | CSUM_INNER_IP | CSUM_INNER_IP_UDP | \
707     CSUM_INNER_IP_TCP | CSUM_INNER_IP_TSO | CSUM_ENCAP_VXLAN | \
708     CSUM_ENCAP_RSVD1 | CSUM_SND_TAG)
709 
710 #define CSUM_FLAGS_RX (CSUM_INNER_L3_CALC | CSUM_INNER_L3_VALID | \
711     CSUM_INNER_L4_CALC | CSUM_INNER_L4_VALID | CSUM_L3_CALC | CSUM_L3_VALID | \
712     CSUM_L4_CALC | CSUM_L4_VALID | CSUM_L5_CALC | CSUM_L5_VALID | \
713     CSUM_COALESCED)
714 
715 /*
716  * CSUM flag description for use with printf(9) %b identifier.
717  */
718 #define	CSUM_BITS \
719     "\20\1CSUM_IP\2CSUM_IP_UDP\3CSUM_IP_TCP\4CSUM_IP_SCTP\5CSUM_IP_TSO" \
720     "\6CSUM_IP_ISCSI\7CSUM_INNER_IP6_UDP\10CSUM_INNER_IP6_TCP" \
721     "\11CSUM_INNER_IP6_TSO\12CSUM_IP6_UDP\13CSUM_IP6_TCP\14CSUM_IP6_SCTP" \
722     "\15CSUM_IP6_TSO\16CSUM_IP6_ISCSI\17CSUM_INNER_IP\20CSUM_INNER_IP_UDP" \
723     "\21CSUM_INNER_IP_TCP\22CSUM_INNER_IP_TSO\23CSUM_ENCAP_VXLAN" \
724     "\24CSUM_ENCAP_RSVD1\25CSUM_INNER_L3_CALC\26CSUM_INNER_L3_VALID" \
725     "\27CSUM_INNER_L4_CALC\30CSUM_INNER_L4_VALID\31CSUM_L3_CALC" \
726     "\32CSUM_L3_VALID\33CSUM_L4_CALC\34CSUM_L4_VALID\35CSUM_L5_CALC" \
727     "\36CSUM_L5_VALID\37CSUM_COALESCED\40CSUM_SND_TAG"
728 
729 /* CSUM flags compatibility mappings. */
730 #define	CSUM_IP_CHECKED		CSUM_L3_CALC
731 #define	CSUM_IP_VALID		CSUM_L3_VALID
732 #define	CSUM_DATA_VALID		CSUM_L4_VALID
733 #define	CSUM_PSEUDO_HDR		CSUM_L4_CALC
734 #define	CSUM_SCTP_VALID		CSUM_L4_VALID
735 #define	CSUM_DELAY_DATA		(CSUM_TCP|CSUM_UDP)
736 #define	CSUM_DELAY_IP		CSUM_IP		/* Only v4, no v6 IP hdr csum */
737 #define	CSUM_DELAY_DATA_IPV6	(CSUM_TCP_IPV6|CSUM_UDP_IPV6)
738 #define	CSUM_DATA_VALID_IPV6	CSUM_DATA_VALID
739 #define	CSUM_TCP		CSUM_IP_TCP
740 #define	CSUM_UDP		CSUM_IP_UDP
741 #define	CSUM_SCTP		CSUM_IP_SCTP
742 #define	CSUM_TSO		(CSUM_IP_TSO|CSUM_IP6_TSO)
743 #define	CSUM_INNER_TSO		(CSUM_INNER_IP_TSO|CSUM_INNER_IP6_TSO)
744 #define	CSUM_UDP_IPV6		CSUM_IP6_UDP
745 #define	CSUM_TCP_IPV6		CSUM_IP6_TCP
746 #define	CSUM_SCTP_IPV6		CSUM_IP6_SCTP
747 #define	CSUM_TLS_MASK		(CSUM_L5_CALC|CSUM_L5_VALID)
748 #define	CSUM_TLS_DECRYPTED	CSUM_L5_CALC
749 
750 /*
751  * mbuf types describing the content of the mbuf (including external storage).
752  */
753 #define	MT_NOTMBUF	0	/* USED INTERNALLY ONLY! Object is not mbuf */
754 #define	MT_DATA		1	/* dynamic (data) allocation */
755 #define	MT_HEADER	MT_DATA	/* packet header, use M_PKTHDR instead */
756 
757 #define	MT_VENDOR1	4	/* for vendor-internal use */
758 #define	MT_VENDOR2	5	/* for vendor-internal use */
759 #define	MT_VENDOR3	6	/* for vendor-internal use */
760 #define	MT_VENDOR4	7	/* for vendor-internal use */
761 
762 #define	MT_SONAME	8	/* socket name */
763 
764 #define	MT_EXP1		9	/* for experimental use */
765 #define	MT_EXP2		10	/* for experimental use */
766 #define	MT_EXP3		11	/* for experimental use */
767 #define	MT_EXP4		12	/* for experimental use */
768 
769 #define	MT_CONTROL	14	/* extra-data protocol message */
770 #define	MT_EXTCONTROL	15	/* control message with externalized contents */
771 #define	MT_OOBDATA	16	/* expedited data  */
772 
773 #define	MT_NOINIT	255	/* Not a type but a flag to allocate
774 				   a non-initialized mbuf */
775 
776 /*
777  * String names of mbuf-related UMA(9) and malloc(9) types.  Exposed to
778  * !_KERNEL so that monitoring tools can look up the zones with
779  * libmemstat(3).
780  */
781 #define	MBUF_MEM_NAME		"mbuf"
782 #define	MBUF_CLUSTER_MEM_NAME	"mbuf_cluster"
783 #define	MBUF_PACKET_MEM_NAME	"mbuf_packet"
784 #define	MBUF_JUMBOP_MEM_NAME	"mbuf_jumbo_page"
785 #define	MBUF_JUMBO9_MEM_NAME	"mbuf_jumbo_9k"
786 #define	MBUF_JUMBO16_MEM_NAME	"mbuf_jumbo_16k"
787 #define	MBUF_TAG_MEM_NAME	"mbuf_tag"
788 #define	MBUF_EXTREFCNT_MEM_NAME	"mbuf_ext_refcnt"
789 #define	MBUF_EXTPGS_MEM_NAME	"mbuf_extpgs"
790 
791 #ifdef _KERNEL
792 union if_snd_tag_alloc_params;
793 
794 #ifdef WITNESS
795 #define	MBUF_CHECKSLEEP(how) do {					\
796 	if (how == M_WAITOK)						\
797 		WITNESS_WARN(WARN_GIANTOK | WARN_SLEEPOK, NULL,		\
798 		    "Sleeping in \"%s\"", __func__);			\
799 } while (0)
800 #else
801 #define	MBUF_CHECKSLEEP(how) do {} while (0)
802 #endif
803 
804 /*
805  * Network buffer allocation API
806  *
807  * The rest of it is defined in kern/kern_mbuf.c
808  */
809 extern uma_zone_t	zone_mbuf;
810 extern uma_zone_t	zone_clust;
811 extern uma_zone_t	zone_pack;
812 extern uma_zone_t	zone_jumbop;
813 extern uma_zone_t	zone_jumbo9;
814 extern uma_zone_t	zone_jumbo16;
815 extern uma_zone_t	zone_extpgs;
816 
817 void		 mb_dupcl(struct mbuf *, struct mbuf *);
818 void		 mb_free_ext(struct mbuf *);
819 void		 mb_free_extpg(struct mbuf *);
820 void		 mb_free_mext_pgs(struct mbuf *);
821 struct mbuf	*mb_alloc_ext_pgs(int, m_ext_free_t);
822 struct mbuf	*mb_alloc_ext_plus_pages(int, int);
823 struct mbuf	*mb_mapped_to_unmapped(struct mbuf *, int, int, int,
824 		    struct mbuf **);
825 int		 mb_unmapped_compress(struct mbuf *m);
826 struct mbuf 	*mb_unmapped_to_ext(struct mbuf *m);
827 void		 mb_free_notready(struct mbuf *m, int count);
828 void		 m_adj(struct mbuf *, int);
829 void		 m_adj_decap(struct mbuf *, int);
830 int		 m_apply(struct mbuf *, int, int,
831 		    int (*)(void *, void *, u_int), void *);
832 int		 m_append(struct mbuf *, int, c_caddr_t);
833 void		 m_cat(struct mbuf *, struct mbuf *);
834 void		 m_catpkt(struct mbuf *, struct mbuf *);
835 int		 m_clget(struct mbuf *m, int how);
836 void 		*m_cljget(struct mbuf *m, int how, int size);
837 struct mbuf	*m_collapse(struct mbuf *, int, int);
838 void		 m_copyback(struct mbuf *, int, int, c_caddr_t);
839 void		 m_copydata(const struct mbuf *, int, int, caddr_t);
840 struct mbuf	*m_copym(struct mbuf *, int, int, int);
841 struct mbuf	*m_copypacket(struct mbuf *, int);
842 void		 m_copy_pkthdr(struct mbuf *, struct mbuf *);
843 struct mbuf	*m_copyup(struct mbuf *, int, int);
844 struct mbuf	*m_defrag(struct mbuf *, int);
845 void		 m_demote_pkthdr(struct mbuf *);
846 void		 m_demote(struct mbuf *, int, int);
847 struct mbuf	*m_devget(char *, int, int, struct ifnet *,
848 		    void (*)(char *, caddr_t, u_int));
849 void		 m_dispose_extcontrolm(struct mbuf *m);
850 struct mbuf	*m_dup(const struct mbuf *, int);
851 int		 m_dup_pkthdr(struct mbuf *, const struct mbuf *, int);
852 void		 m_extadd(struct mbuf *, char *, u_int, m_ext_free_t,
853 		    void *, void *, int, int);
854 u_int		 m_fixhdr(struct mbuf *);
855 struct mbuf	*m_fragment(struct mbuf *, int, int);
856 void		 m_freem(struct mbuf *);
857 void		 m_free_raw(struct mbuf *);
858 struct mbuf	*m_get2(int, int, short, int);
859 struct mbuf	*m_get3(int, int, short, int);
860 struct mbuf	*m_getjcl(int, short, int, int);
861 struct mbuf	*m_getm2(struct mbuf *, int, int, short, int);
862 struct mbuf	*m_getptr(struct mbuf *, int, int *);
863 u_int		 m_length(struct mbuf *, struct mbuf **);
864 int		 m_mbuftouio(struct uio *, const struct mbuf *, int);
865 void		 m_move_pkthdr(struct mbuf *, struct mbuf *);
866 int		 m_pkthdr_init(struct mbuf *, int);
867 struct mbuf	*m_prepend(struct mbuf *, int, int);
868 void		 m_print(const struct mbuf *, int);
869 struct mbuf	*m_pulldown(struct mbuf *, int, int, int *);
870 struct mbuf	*m_pullup(struct mbuf *, int);
871 int		 m_sanity(struct mbuf *, int);
872 struct mbuf	*m_split(struct mbuf *, int, int);
873 struct mbuf	*m_uiotombuf(struct uio *, int, int, int, int);
874 int		 m_unmapped_uiomove(const struct mbuf *, int, struct uio *,
875 		    int);
876 struct mbuf	*m_unshare(struct mbuf *, int);
877 int		 m_snd_tag_alloc(struct ifnet *,
878 		    union if_snd_tag_alloc_params *, struct m_snd_tag **);
879 void		 m_snd_tag_init(struct m_snd_tag *, struct ifnet *,
880 		    const struct if_snd_tag_sw *);
881 void		 m_snd_tag_destroy(struct m_snd_tag *);
882 void		 m_rcvif_serialize(struct mbuf *);
883 struct ifnet	*m_rcvif_restore(struct mbuf *);
884 
885 static __inline int
886 m_gettype(int size)
887 {
888 	int type;
889 
890 	switch (size) {
891 	case MSIZE:
892 		type = EXT_MBUF;
893 		break;
894 	case MCLBYTES:
895 		type = EXT_CLUSTER;
896 		break;
897 	case MJUMPAGESIZE:
898 		type = EXT_JUMBOP;
899 		break;
900 	case MJUM9BYTES:
901 		type = EXT_JUMBO9;
902 		break;
903 	case MJUM16BYTES:
904 		type = EXT_JUMBO16;
905 		break;
906 	default:
907 		panic("%s: invalid cluster size %d", __func__, size);
908 	}
909 
910 	return (type);
911 }
912 
913 /*
914  * Associated an external reference counted buffer with an mbuf.
915  */
916 static __inline void
917 m_extaddref(struct mbuf *m, char *buf, u_int size, u_int *ref_cnt,
918     m_ext_free_t freef, void *arg1, void *arg2)
919 {
920 
921 	KASSERT(ref_cnt != NULL, ("%s: ref_cnt not provided", __func__));
922 
923 	atomic_add_int(ref_cnt, 1);
924 	m->m_flags |= M_EXT;
925 	m->m_ext.ext_buf = buf;
926 	m->m_ext.ext_cnt = ref_cnt;
927 	m->m_data = m->m_ext.ext_buf;
928 	m->m_ext.ext_size = size;
929 	m->m_ext.ext_free = freef;
930 	m->m_ext.ext_arg1 = arg1;
931 	m->m_ext.ext_arg2 = arg2;
932 	m->m_ext.ext_type = EXT_EXTREF;
933 	m->m_ext.ext_flags = 0;
934 }
935 
936 static __inline uma_zone_t
937 m_getzone(int size)
938 {
939 	uma_zone_t zone;
940 
941 	switch (size) {
942 	case MCLBYTES:
943 		zone = zone_clust;
944 		break;
945 	case MJUMPAGESIZE:
946 		zone = zone_jumbop;
947 		break;
948 	case MJUM9BYTES:
949 		zone = zone_jumbo9;
950 		break;
951 	case MJUM16BYTES:
952 		zone = zone_jumbo16;
953 		break;
954 	default:
955 		panic("%s: invalid cluster size %d", __func__, size);
956 	}
957 
958 	return (zone);
959 }
960 
961 /*
962  * Initialize an mbuf with linear storage.
963  *
964  * Inline because the consumer text overhead will be roughly the same to
965  * initialize or call a function with this many parameters and M_PKTHDR
966  * should go away with constant propagation for !MGETHDR.
967  */
968 static __inline int
969 m_init(struct mbuf *m, int how, short type, int flags)
970 {
971 	int error;
972 
973 	m->m_next = NULL;
974 	m->m_nextpkt = NULL;
975 	m->m_data = m->m_dat;
976 	m->m_len = 0;
977 	m->m_flags = flags;
978 	m->m_type = type;
979 	if (flags & M_PKTHDR)
980 		error = m_pkthdr_init(m, how);
981 	else
982 		error = 0;
983 
984 	MBUF_PROBE5(m__init, m, how, type, flags, error);
985 	return (error);
986 }
987 
988 static __inline struct mbuf *
989 m_get_raw(int how, short type)
990 {
991 	struct mbuf *m;
992 	struct mb_args args;
993 
994 	args.flags = 0;
995 	args.type = type | MT_NOINIT;
996 	m = uma_zalloc_arg(zone_mbuf, &args, how);
997 	MBUF_PROBE3(m__get_raw, how, type, m);
998 	return (m);
999 }
1000 
1001 static __inline struct mbuf *
1002 m_get(int how, short type)
1003 {
1004 	struct mbuf *m;
1005 	struct mb_args args;
1006 
1007 	args.flags = 0;
1008 	args.type = type;
1009 	m = uma_zalloc_arg(zone_mbuf, &args, how);
1010 	MBUF_PROBE3(m__get, how, type, m);
1011 	return (m);
1012 }
1013 
1014 static __inline struct mbuf *
1015 m_gethdr_raw(int how, short type)
1016 {
1017 	struct mbuf *m;
1018 	struct mb_args args;
1019 
1020 	args.flags = M_PKTHDR;
1021 	args.type = type | MT_NOINIT;
1022 	m = uma_zalloc_arg(zone_mbuf, &args, how);
1023 	MBUF_PROBE3(m__gethdr_raw, how, type, m);
1024 	return (m);
1025 }
1026 
1027 static __inline struct mbuf *
1028 m_gethdr(int how, short type)
1029 {
1030 	struct mbuf *m;
1031 	struct mb_args args;
1032 
1033 	args.flags = M_PKTHDR;
1034 	args.type = type;
1035 	m = uma_zalloc_arg(zone_mbuf, &args, how);
1036 	MBUF_PROBE3(m__gethdr, how, type, m);
1037 	return (m);
1038 }
1039 
1040 static __inline struct mbuf *
1041 m_getcl(int how, short type, int flags)
1042 {
1043 	struct mbuf *m;
1044 	struct mb_args args;
1045 
1046 	args.flags = flags;
1047 	args.type = type;
1048 	m = uma_zalloc_arg(zone_pack, &args, how);
1049 	MBUF_PROBE4(m__getcl, how, type, flags, m);
1050 	return (m);
1051 }
1052 
1053 /*
1054  * XXX: m_cljset() is a dangerous API.  One must attach only a new,
1055  * unreferenced cluster to an mbuf(9).  It is not possible to assert
1056  * that, so care can be taken only by users of the API.
1057  */
1058 static __inline void
1059 m_cljset(struct mbuf *m, void *cl, int type)
1060 {
1061 	int size;
1062 
1063 	switch (type) {
1064 	case EXT_CLUSTER:
1065 		size = MCLBYTES;
1066 		break;
1067 	case EXT_JUMBOP:
1068 		size = MJUMPAGESIZE;
1069 		break;
1070 	case EXT_JUMBO9:
1071 		size = MJUM9BYTES;
1072 		break;
1073 	case EXT_JUMBO16:
1074 		size = MJUM16BYTES;
1075 		break;
1076 	default:
1077 		panic("%s: unknown cluster type %d", __func__, type);
1078 		break;
1079 	}
1080 
1081 	m->m_data = m->m_ext.ext_buf = cl;
1082 	m->m_ext.ext_free = m->m_ext.ext_arg1 = m->m_ext.ext_arg2 = NULL;
1083 	m->m_ext.ext_size = size;
1084 	m->m_ext.ext_type = type;
1085 	m->m_ext.ext_flags = EXT_FLAG_EMBREF;
1086 	m->m_ext.ext_count = 1;
1087 	m->m_flags |= M_EXT;
1088 	MBUF_PROBE3(m__cljset, m, cl, type);
1089 }
1090 
1091 static __inline void
1092 m_chtype(struct mbuf *m, short new_type)
1093 {
1094 
1095 	m->m_type = new_type;
1096 }
1097 
1098 static __inline void
1099 m_clrprotoflags(struct mbuf *m)
1100 {
1101 
1102 	while (m) {
1103 		m->m_flags &= ~M_PROTOFLAGS;
1104 		m = m->m_next;
1105 	}
1106 }
1107 
1108 static __inline struct mbuf *
1109 m_last(struct mbuf *m)
1110 {
1111 
1112 	while (m->m_next)
1113 		m = m->m_next;
1114 	return (m);
1115 }
1116 
1117 static inline u_int
1118 m_extrefcnt(struct mbuf *m)
1119 {
1120 
1121 	KASSERT(m->m_flags & M_EXT, ("%s: M_EXT missing", __func__));
1122 
1123 	return ((m->m_ext.ext_flags & EXT_FLAG_EMBREF) ? m->m_ext.ext_count :
1124 	    *m->m_ext.ext_cnt);
1125 }
1126 
1127 /*
1128  * mbuf, cluster, and external object allocation macros (for compatibility
1129  * purposes).
1130  */
1131 #define	M_MOVE_PKTHDR(to, from)	m_move_pkthdr((to), (from))
1132 #define	MGET(m, how, type)	((m) = m_get((how), (type)))
1133 #define	MGETHDR(m, how, type)	((m) = m_gethdr((how), (type)))
1134 #define	MCLGET(m, how)		m_clget((m), (how))
1135 #define	MEXTADD(m, buf, size, free, arg1, arg2, flags, type)		\
1136     m_extadd((m), (char *)(buf), (size), (free), (arg1), (arg2),	\
1137     (flags), (type))
1138 #define	m_getm(m, len, how, type)					\
1139     m_getm2((m), (len), (how), (type), M_PKTHDR)
1140 
1141 /*
1142  * Evaluate TRUE if it's safe to write to the mbuf m's data region (this can
1143  * be both the local data payload, or an external buffer area, depending on
1144  * whether M_EXT is set).
1145  */
1146 #define	M_WRITABLE(m)	(((m)->m_flags & (M_RDONLY | M_EXTPG)) == 0 &&	\
1147 			 (!(((m)->m_flags & M_EXT)) ||			\
1148 			 (m_extrefcnt(m) == 1)))
1149 
1150 /* Check if the supplied mbuf has a packet header, or else panic. */
1151 #define	M_ASSERTPKTHDR(m)						\
1152 	KASSERT((m) != NULL && (m)->m_flags & M_PKTHDR,			\
1153 	    ("%s: no mbuf packet header!", __func__))
1154 
1155 /* Check if the supplied mbuf has no send tag, or else panic. */
1156 #define	M_ASSERT_NO_SND_TAG(m)						\
1157 	KASSERT((m) != NULL && (m)->m_flags & M_PKTHDR &&		\
1158 	       ((m)->m_pkthdr.csum_flags & CSUM_SND_TAG) == 0,		\
1159 	    ("%s: receive mbuf has send tag!", __func__))
1160 
1161 /* Check if mbuf is multipage. */
1162 #define M_ASSERTEXTPG(m)						\
1163 	KASSERT(((m)->m_flags & (M_EXTPG|M_PKTHDR)) == M_EXTPG,		\
1164 	    ("%s: m %p is not multipage!", __func__, m))
1165 
1166 /*
1167  * Ensure that the supplied mbuf is a valid, non-free mbuf.
1168  *
1169  * XXX: Broken at the moment.  Need some UMA magic to make it work again.
1170  */
1171 #define	M_ASSERTVALID(m)						\
1172 	KASSERT((((struct mbuf *)m)->m_flags & 0) == 0,			\
1173 	    ("%s: attempted use of a free mbuf!", __func__))
1174 
1175 /* Check whether any mbuf in the chain is unmapped. */
1176 #ifdef INVARIANTS
1177 #define	M_ASSERTMAPPED(m) do {						\
1178 	for (struct mbuf *__m = (m); __m != NULL; __m = __m->m_next)	\
1179 		KASSERT((__m->m_flags & M_EXTPG) == 0,			\
1180 		    ("%s: chain %p contains an unmapped mbuf", __func__, (m)));\
1181 } while (0)
1182 #else
1183 #define	M_ASSERTMAPPED(m) do {} while (0)
1184 #endif
1185 
1186 /*
1187  * Return the address of the start of the buffer associated with an mbuf,
1188  * handling external storage, packet-header mbufs, and regular data mbufs.
1189  */
1190 #define	M_START(m)							\
1191 	(((m)->m_flags & M_EXTPG) ? NULL :				\
1192 	 ((m)->m_flags & M_EXT) ? (m)->m_ext.ext_buf :			\
1193 	 ((m)->m_flags & M_PKTHDR) ? &(m)->m_pktdat[0] :		\
1194 	 &(m)->m_dat[0])
1195 
1196 /*
1197  * Return the size of the buffer associated with an mbuf, handling external
1198  * storage, packet-header mbufs, and regular data mbufs.
1199  */
1200 #define	M_SIZE(m)							\
1201 	(((m)->m_flags & M_EXT) ? (m)->m_ext.ext_size :			\
1202 	 ((m)->m_flags & M_PKTHDR) ? MHLEN :				\
1203 	 MLEN)
1204 
1205 /*
1206  * Set the m_data pointer of a newly allocated mbuf to place an object of the
1207  * specified size at the end of the mbuf, longword aligned.
1208  *
1209  * NB: Historically, we had M_ALIGN(), MH_ALIGN(), and MEXT_ALIGN() as
1210  * separate macros, each asserting that it was called at the proper moment.
1211  * This required callers to themselves test the storage type and call the
1212  * right one.  Rather than require callers to be aware of those layout
1213  * decisions, we centralize here.
1214  */
1215 static __inline void
1216 m_align(struct mbuf *m, int len)
1217 {
1218 #ifdef INVARIANTS
1219 	const char *msg = "%s: not a virgin mbuf";
1220 #endif
1221 	int adjust;
1222 
1223 	KASSERT(m->m_data == M_START(m), (msg, __func__));
1224 
1225 	adjust = M_SIZE(m) - len;
1226 	m->m_data += adjust &~ (sizeof(long)-1);
1227 }
1228 
1229 #define	M_ALIGN(m, len)		m_align(m, len)
1230 #define	MH_ALIGN(m, len)	m_align(m, len)
1231 #define	MEXT_ALIGN(m, len)	m_align(m, len)
1232 
1233 /*
1234  * Compute the amount of space available before the current start of data in
1235  * an mbuf.
1236  *
1237  * The M_WRITABLE() is a temporary, conservative safety measure: the burden
1238  * of checking writability of the mbuf data area rests solely with the caller.
1239  *
1240  * NB: In previous versions, M_LEADINGSPACE() would only check M_WRITABLE()
1241  * for mbufs with external storage.  We now allow mbuf-embedded data to be
1242  * read-only as well.
1243  */
1244 #define	M_LEADINGSPACE(m)						\
1245 	(M_WRITABLE(m) ? ((m)->m_data - M_START(m)) : 0)
1246 
1247 /*
1248  * Compute the amount of space available after the end of data in an mbuf.
1249  *
1250  * The M_WRITABLE() is a temporary, conservative safety measure: the burden
1251  * of checking writability of the mbuf data area rests solely with the caller.
1252  *
1253  * NB: In previous versions, M_TRAILINGSPACE() would only check M_WRITABLE()
1254  * for mbufs with external storage.  We now allow mbuf-embedded data to be
1255  * read-only as well.
1256  */
1257 #define	M_TRAILINGSPACE(m)						\
1258 	(M_WRITABLE(m) ?						\
1259 	    ((M_START(m) + M_SIZE(m)) - ((m)->m_data + (m)->m_len)) : 0)
1260 
1261 /*
1262  * Arrange to prepend space of size plen to mbuf m.  If a new mbuf must be
1263  * allocated, how specifies whether to wait.  If the allocation fails, the
1264  * original mbuf chain is freed and m is set to NULL.
1265  */
1266 #define	M_PREPEND(m, plen, how) do {					\
1267 	struct mbuf **_mmp = &(m);					\
1268 	struct mbuf *_mm = *_mmp;					\
1269 	int _mplen = (plen);						\
1270 	int __mhow = (how);						\
1271 									\
1272 	MBUF_CHECKSLEEP(how);						\
1273 	if (M_LEADINGSPACE(_mm) >= _mplen) {				\
1274 		_mm->m_data -= _mplen;					\
1275 		_mm->m_len += _mplen;					\
1276 	} else								\
1277 		_mm = m_prepend(_mm, _mplen, __mhow);			\
1278 	if (_mm != NULL && _mm->m_flags & M_PKTHDR)			\
1279 		_mm->m_pkthdr.len += _mplen;				\
1280 	*_mmp = _mm;							\
1281 } while (0)
1282 
1283 /*
1284  * Change mbuf to new type.  This is a relatively expensive operation and
1285  * should be avoided.
1286  */
1287 #define	MCHTYPE(m, t)	m_chtype((m), (t))
1288 
1289 /* Return the rcvif of a packet header. */
1290 static __inline struct ifnet *
1291 m_rcvif(struct mbuf *m)
1292 {
1293 
1294 	M_ASSERTPKTHDR(m);
1295 	if (m->m_pkthdr.csum_flags & CSUM_SND_TAG)
1296 		return (NULL);
1297 	return (m->m_pkthdr.rcvif);
1298 }
1299 
1300 /* Length to m_copy to copy all. */
1301 #define	M_COPYALL	1000000000
1302 
1303 extern u_int		max_linkhdr;	/* Largest link-level header */
1304 extern u_int		max_hdr;	/* Largest link + protocol header */
1305 extern u_int		max_protohdr;	/* Largest protocol header */
1306 void max_linkhdr_grow(u_int);
1307 void max_protohdr_grow(u_int);
1308 
1309 extern int		nmbclusters;	/* Maximum number of clusters */
1310 extern bool		mb_use_ext_pgs;	/* Use ext_pgs for sendfile */
1311 
1312 /*-
1313  * Network packets may have annotations attached by affixing a list of
1314  * "packet tags" to the pkthdr structure.  Packet tags are dynamically
1315  * allocated semi-opaque data structures that have a fixed header
1316  * (struct m_tag) that specifies the size of the memory block and a
1317  * <cookie,type> pair that identifies it.  The cookie is a 32-bit unique
1318  * unsigned value used to identify a module or ABI.  By convention this value
1319  * is chosen as the date+time that the module is created, expressed as the
1320  * number of seconds since the epoch (e.g., using date -u +'%s').  The type
1321  * value is an ABI/module-specific value that identifies a particular
1322  * annotation and is private to the module.  For compatibility with systems
1323  * like OpenBSD that define packet tags w/o an ABI/module cookie, the value
1324  * PACKET_ABI_COMPAT is used to implement m_tag_get and m_tag_find
1325  * compatibility shim functions and several tag types are defined below.
1326  * Users that do not require compatibility should use a private cookie value
1327  * so that packet tag-related definitions can be maintained privately.
1328  *
1329  * Note that the packet tag returned by m_tag_alloc has the default memory
1330  * alignment implemented by malloc.  To reference private data one can use a
1331  * construct like:
1332  *
1333  *	struct m_tag *mtag = m_tag_alloc(...);
1334  *	struct foo *p = (struct foo *)(mtag+1);
1335  *
1336  * if the alignment of struct m_tag is sufficient for referencing members of
1337  * struct foo.  Otherwise it is necessary to embed struct m_tag within the
1338  * private data structure to insure proper alignment; e.g.,
1339  *
1340  *	struct foo {
1341  *		struct m_tag	tag;
1342  *		...
1343  *	};
1344  *	struct foo *p = (struct foo *) m_tag_alloc(...);
1345  *	struct m_tag *mtag = &p->tag;
1346  */
1347 
1348 /*
1349  * Persistent tags stay with an mbuf until the mbuf is reclaimed.  Otherwise
1350  * tags are expected to ``vanish'' when they pass through a network
1351  * interface.  For most interfaces this happens normally as the tags are
1352  * reclaimed when the mbuf is free'd.  However in some special cases
1353  * reclaiming must be done manually.  An example is packets that pass through
1354  * the loopback interface.  Also, one must be careful to do this when
1355  * ``turning around'' packets (e.g., icmp_reflect).
1356  *
1357  * To mark a tag persistent bit-or this flag in when defining the tag id.
1358  * The tag will then be treated as described above.
1359  */
1360 #define	MTAG_PERSISTENT				0x800
1361 
1362 #define	PACKET_TAG_NONE				0  /* Nadda */
1363 
1364 /* Packet tags for use with PACKET_ABI_COMPAT. */
1365 #define	PACKET_TAG_IPSEC_IN_DONE		1  /* IPsec applied, in */
1366 #define	PACKET_TAG_IPSEC_OUT_DONE		2  /* IPsec applied, out */
1367 #define	PACKET_TAG_IPSEC_IN_CRYPTO_DONE		3  /* NIC IPsec crypto done */
1368 #define	PACKET_TAG_IPSEC_OUT_CRYPTO_NEEDED	4  /* NIC IPsec crypto req'ed */
1369 #define	PACKET_TAG_IPSEC_IN_COULD_DO_CRYPTO	5  /* NIC notifies IPsec */
1370 #define	PACKET_TAG_IPSEC_PENDING_TDB		6  /* Reminder to do IPsec */
1371 #define	PACKET_TAG_BRIDGE			7  /* Bridge processing done */
1372 #define	PACKET_TAG_GIF				8  /* GIF processing done */
1373 #define	PACKET_TAG_GRE				9  /* GRE processing done */
1374 #define	PACKET_TAG_IN_PACKET_CHECKSUM		10 /* NIC checksumming done */
1375 #define	PACKET_TAG_ENCAP			11 /* Encap.  processing */
1376 #define	PACKET_TAG_IPSEC_SOCKET			12 /* IPSEC socket ref */
1377 #define	PACKET_TAG_IPSEC_HISTORY		13 /* IPSEC history */
1378 #define	PACKET_TAG_IPV6_INPUT			14 /* IPV6 input processing */
1379 #define	PACKET_TAG_DUMMYNET			15 /* dummynet info */
1380 #define	PACKET_TAG_DIVERT			17 /* divert info */
1381 #define	PACKET_TAG_IPFORWARD			18 /* ipforward info */
1382 #define	PACKET_TAG_MACLABEL	(19 | MTAG_PERSISTENT) /* MAC label */
1383 #define	PACKET_TAG_PF				21 /* PF/ALTQ information */
1384 /* was	PACKET_TAG_RTSOCKFAM			25    rtsock sa family */
1385 #define	PACKET_TAG_IPOPTIONS			27 /* Saved IP options */
1386 #define	PACKET_TAG_CARP				28 /* CARP info */
1387 #define	PACKET_TAG_IPSEC_NAT_T_PORTS		29 /* two uint16_t */
1388 #define	PACKET_TAG_ND_OUTGOING			30 /* ND outgoing */
1389 
1390 /* Specific cookies and tags. */
1391 
1392 /* Packet tag routines. */
1393 struct m_tag	*m_tag_alloc(uint32_t, uint16_t, int, int);
1394 void		 m_tag_delete(struct mbuf *, struct m_tag *);
1395 void		 m_tag_delete_chain(struct mbuf *, struct m_tag *);
1396 void		 m_tag_free_default(struct m_tag *);
1397 struct m_tag	*m_tag_locate(struct mbuf *, uint32_t, uint16_t,
1398     struct m_tag *);
1399 struct m_tag	*m_tag_copy(struct m_tag *, int);
1400 int		 m_tag_copy_chain(struct mbuf *, const struct mbuf *, int);
1401 void		 m_tag_delete_nonpersistent(struct mbuf *);
1402 
1403 /*
1404  * Initialize the list of tags associated with an mbuf.
1405  */
1406 static __inline void
1407 m_tag_init(struct mbuf *m)
1408 {
1409 
1410 	SLIST_INIT(&m->m_pkthdr.tags);
1411 }
1412 
1413 /*
1414  * Set up the contents of a tag.  Note that this does not fill in the free
1415  * method; the caller is expected to do that.
1416  *
1417  * XXX probably should be called m_tag_init, but that was already taken.
1418  */
1419 static __inline void
1420 m_tag_setup(struct m_tag *t, uint32_t cookie, uint16_t type, int len)
1421 {
1422 
1423 	t->m_tag_id = type;
1424 	t->m_tag_len = len;
1425 	t->m_tag_cookie = cookie;
1426 }
1427 
1428 /*
1429  * Reclaim resources associated with a tag.
1430  */
1431 static __inline void
1432 m_tag_free(struct m_tag *t)
1433 {
1434 
1435 	(*t->m_tag_free)(t);
1436 }
1437 
1438 /*
1439  * Return the first tag associated with an mbuf.
1440  */
1441 static __inline struct m_tag *
1442 m_tag_first(struct mbuf *m)
1443 {
1444 
1445 	return (SLIST_FIRST(&m->m_pkthdr.tags));
1446 }
1447 
1448 /*
1449  * Return the next tag in the list of tags associated with an mbuf.
1450  */
1451 static __inline struct m_tag *
1452 m_tag_next(struct mbuf *m __unused, struct m_tag *t)
1453 {
1454 
1455 	return (SLIST_NEXT(t, m_tag_link));
1456 }
1457 
1458 /*
1459  * Prepend a tag to the list of tags associated with an mbuf.
1460  */
1461 static __inline void
1462 m_tag_prepend(struct mbuf *m, struct m_tag *t)
1463 {
1464 
1465 	SLIST_INSERT_HEAD(&m->m_pkthdr.tags, t, m_tag_link);
1466 }
1467 
1468 /*
1469  * Unlink a tag from the list of tags associated with an mbuf.
1470  */
1471 static __inline void
1472 m_tag_unlink(struct mbuf *m, struct m_tag *t)
1473 {
1474 
1475 	SLIST_REMOVE(&m->m_pkthdr.tags, t, m_tag, m_tag_link);
1476 }
1477 
1478 /* These are for OpenBSD compatibility. */
1479 #define	MTAG_ABI_COMPAT		0		/* compatibility ABI */
1480 
1481 static __inline struct m_tag *
1482 m_tag_get(uint16_t type, int length, int wait)
1483 {
1484 	return (m_tag_alloc(MTAG_ABI_COMPAT, type, length, wait));
1485 }
1486 
1487 static __inline struct m_tag *
1488 m_tag_find(struct mbuf *m, uint16_t type, struct m_tag *start)
1489 {
1490 	return (SLIST_EMPTY(&m->m_pkthdr.tags) ? (struct m_tag *)NULL :
1491 	    m_tag_locate(m, MTAG_ABI_COMPAT, type, start));
1492 }
1493 
1494 static inline struct m_snd_tag *
1495 m_snd_tag_ref(struct m_snd_tag *mst)
1496 {
1497 
1498 	refcount_acquire(&mst->refcount);
1499 	return (mst);
1500 }
1501 
1502 static inline void
1503 m_snd_tag_rele(struct m_snd_tag *mst)
1504 {
1505 
1506 	if (refcount_release(&mst->refcount))
1507 		m_snd_tag_destroy(mst);
1508 }
1509 
1510 static __inline struct mbuf *
1511 m_free(struct mbuf *m)
1512 {
1513 	struct mbuf *n = m->m_next;
1514 
1515 	MBUF_PROBE1(m__free, m);
1516 	if ((m->m_flags & (M_PKTHDR|M_NOFREE)) == (M_PKTHDR|M_NOFREE))
1517 		m_tag_delete_chain(m, NULL);
1518 	if (m->m_flags & M_PKTHDR && m->m_pkthdr.csum_flags & CSUM_SND_TAG)
1519 		m_snd_tag_rele(m->m_pkthdr.snd_tag);
1520 	if (m->m_flags & M_EXTPG)
1521 		mb_free_extpg(m);
1522 	else if (m->m_flags & M_EXT)
1523 		mb_free_ext(m);
1524 	else if ((m->m_flags & M_NOFREE) == 0)
1525 		uma_zfree(zone_mbuf, m);
1526 	return (n);
1527 }
1528 
1529 static __inline int
1530 rt_m_getfib(struct mbuf *m)
1531 {
1532 	KASSERT(m->m_flags & M_PKTHDR , ("Attempt to get FIB from non header mbuf."));
1533 	return (m->m_pkthdr.fibnum);
1534 }
1535 
1536 #define M_GETFIB(_m)   rt_m_getfib(_m)
1537 
1538 #define M_SETFIB(_m, _fib) do {						\
1539         KASSERT((_m)->m_flags & M_PKTHDR, ("Attempt to set FIB on non header mbuf."));	\
1540 	((_m)->m_pkthdr.fibnum) = (_fib);				\
1541 } while (0)
1542 
1543 /* flags passed as first argument for "m_xxx_tcpip_hash()" */
1544 #define	MBUF_HASHFLAG_L2	(1 << 2)
1545 #define	MBUF_HASHFLAG_L3	(1 << 3)
1546 #define	MBUF_HASHFLAG_L4	(1 << 4)
1547 
1548 /* mbuf hashing helper routines */
1549 uint32_t	m_ether_tcpip_hash_init(void);
1550 uint32_t	m_ether_tcpip_hash(const uint32_t, const struct mbuf *, uint32_t);
1551 uint32_t	m_infiniband_tcpip_hash_init(void);
1552 uint32_t	m_infiniband_tcpip_hash(const uint32_t, const struct mbuf *, uint32_t);
1553 
1554 #ifdef MBUF_PROFILING
1555  void m_profile(struct mbuf *m);
1556  #define M_PROFILE(m) m_profile(m)
1557 #else
1558  #define M_PROFILE(m)
1559 #endif
1560 
1561 struct mbufq {
1562 	STAILQ_HEAD(, mbuf)	mq_head;
1563 	int			mq_len;
1564 	int			mq_maxlen;
1565 };
1566 
1567 static inline void
1568 mbufq_init(struct mbufq *mq, int maxlen)
1569 {
1570 
1571 	STAILQ_INIT(&mq->mq_head);
1572 	mq->mq_maxlen = maxlen;
1573 	mq->mq_len = 0;
1574 }
1575 
1576 static inline struct mbuf *
1577 mbufq_flush(struct mbufq *mq)
1578 {
1579 	struct mbuf *m;
1580 
1581 	m = STAILQ_FIRST(&mq->mq_head);
1582 	STAILQ_INIT(&mq->mq_head);
1583 	mq->mq_len = 0;
1584 	return (m);
1585 }
1586 
1587 static inline void
1588 mbufq_drain(struct mbufq *mq)
1589 {
1590 	struct mbuf *m, *n;
1591 
1592 	n = mbufq_flush(mq);
1593 	while ((m = n) != NULL) {
1594 		n = STAILQ_NEXT(m, m_stailqpkt);
1595 		m_freem(m);
1596 	}
1597 }
1598 
1599 static inline struct mbuf *
1600 mbufq_first(const struct mbufq *mq)
1601 {
1602 
1603 	return (STAILQ_FIRST(&mq->mq_head));
1604 }
1605 
1606 static inline struct mbuf *
1607 mbufq_last(const struct mbufq *mq)
1608 {
1609 
1610 	return (STAILQ_LAST(&mq->mq_head, mbuf, m_stailqpkt));
1611 }
1612 
1613 static inline int
1614 mbufq_full(const struct mbufq *mq)
1615 {
1616 
1617 	return (mq->mq_maxlen > 0 && mq->mq_len >= mq->mq_maxlen);
1618 }
1619 
1620 static inline int
1621 mbufq_len(const struct mbufq *mq)
1622 {
1623 
1624 	return (mq->mq_len);
1625 }
1626 
1627 static inline int
1628 mbufq_enqueue(struct mbufq *mq, struct mbuf *m)
1629 {
1630 
1631 	if (mbufq_full(mq))
1632 		return (ENOBUFS);
1633 	STAILQ_INSERT_TAIL(&mq->mq_head, m, m_stailqpkt);
1634 	mq->mq_len++;
1635 	return (0);
1636 }
1637 
1638 static inline struct mbuf *
1639 mbufq_dequeue(struct mbufq *mq)
1640 {
1641 	struct mbuf *m;
1642 
1643 	m = STAILQ_FIRST(&mq->mq_head);
1644 	if (m) {
1645 		STAILQ_REMOVE_HEAD(&mq->mq_head, m_stailqpkt);
1646 		m->m_nextpkt = NULL;
1647 		mq->mq_len--;
1648 	}
1649 	return (m);
1650 }
1651 
1652 static inline void
1653 mbufq_prepend(struct mbufq *mq, struct mbuf *m)
1654 {
1655 
1656 	STAILQ_INSERT_HEAD(&mq->mq_head, m, m_stailqpkt);
1657 	mq->mq_len++;
1658 }
1659 
1660 /*
1661  * Note: this doesn't enforce the maximum list size for dst.
1662  */
1663 static inline void
1664 mbufq_concat(struct mbufq *mq_dst, struct mbufq *mq_src)
1665 {
1666 
1667 	mq_dst->mq_len += mq_src->mq_len;
1668 	STAILQ_CONCAT(&mq_dst->mq_head, &mq_src->mq_head);
1669 	mq_src->mq_len = 0;
1670 }
1671 
1672 #ifdef _SYS_TIMESPEC_H_
1673 static inline void
1674 mbuf_tstmp2timespec(struct mbuf *m, struct timespec *ts)
1675 {
1676 
1677 	KASSERT((m->m_flags & M_PKTHDR) != 0, ("mbuf %p no M_PKTHDR", m));
1678 	KASSERT((m->m_flags & (M_TSTMP|M_TSTMP_LRO)) != 0, ("mbuf %p no M_TSTMP or M_TSTMP_LRO", m));
1679 	ts->tv_sec = m->m_pkthdr.rcv_tstmp / 1000000000;
1680 	ts->tv_nsec = m->m_pkthdr.rcv_tstmp % 1000000000;
1681 }
1682 #endif
1683 
1684 #ifdef DEBUGNET
1685 /* Invoked from the debugnet client code. */
1686 void	debugnet_mbuf_drain(void);
1687 void	debugnet_mbuf_start(void);
1688 void	debugnet_mbuf_finish(void);
1689 void	debugnet_mbuf_reinit(int nmbuf, int nclust, int clsize);
1690 #endif
1691 
1692 static inline bool
1693 mbuf_has_tls_session(struct mbuf *m)
1694 {
1695 
1696 	if (m->m_flags & M_EXTPG) {
1697 		if (m->m_epg_tls != NULL) {
1698 			return (true);
1699 		}
1700 	}
1701 	return (false);
1702 }
1703 
1704 #endif /* _KERNEL */
1705 #endif /* !_SYS_MBUF_H_ */
1706