xref: /freebsd/sys/sys/mbuf.h (revision 8e9740b62e950b40dcf7dbe729855be14450d40d)
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: 48
154  *	 LP64: 56
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 	SLIST_HEAD(packet_tags, m_tag) tags; /* list of packet tags */
168 	int32_t		 len;		/* total packet length */
169 
170 	/* Layer crossing persistent information. */
171 	uint32_t	 flowid;	/* packet's 4-tuple system */
172 	uint32_t	 csum_flags;	/* checksum and offload features */
173 	uint16_t	 fibnum;	/* this packet should use this fib */
174 	uint8_t		 numa_domain;	/* NUMA domain of recvd pkt */
175 	uint8_t		 rsstype;	/* hash type */
176 	union {
177 		uint64_t	rcv_tstmp;	/* timestamp in ns */
178 		struct {
179 			uint8_t		 l2hlen;	/* layer 2 hdr len */
180 			uint8_t		 l3hlen;	/* layer 3 hdr len */
181 			uint8_t		 l4hlen;	/* layer 4 hdr len */
182 			uint8_t		 l5hlen;	/* layer 5 hdr len */
183 			uint8_t		 inner_l2hlen;
184 			uint8_t		 inner_l3hlen;
185 			uint8_t		 inner_l4hlen;
186 			uint8_t		 inner_l5hlen;
187 		};
188 	};
189 	union {
190 		uint8_t  eight[8];
191 		uint16_t sixteen[4];
192 		uint32_t thirtytwo[2];
193 		uint64_t sixtyfour[1];
194 		uintptr_t unintptr[1];
195 		void	*ptr;
196 	} PH_per;
197 
198 	/* Layer specific non-persistent local storage for reassembly, etc. */
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_loc;
207 };
208 #define	ether_vtag	PH_per.sixteen[0]
209 #define tcp_tun_port	PH_per.sixteen[0] /* outbound */
210 #define	PH_vt		PH_per
211 #define	vt_nrecs	sixteen[0]	  /* mld and v6-ND */
212 #define	tso_segsz	PH_per.sixteen[1] /* inbound after LRO */
213 #define	lro_nsegs	tso_segsz	  /* inbound after LRO */
214 #define	csum_data	PH_per.thirtytwo[1] /* inbound from hardware up */
215 #define	lro_tcp_d_len	PH_loc.sixteen[0] /* inbound during LRO (no reassembly) */
216 #define	lro_tcp_d_csum	PH_loc.sixteen[1] /* inbound during LRO (no reassembly) */
217 #define	lro_tcp_h_off	PH_loc.sixteen[2] /* inbound during LRO (no reassembly) */
218 #define	lro_etype	PH_loc.sixteen[3] /* inbound during LRO (no reassembly) */
219 /* Note PH_loc is used during IP reassembly (all 8 bytes as a ptr) */
220 
221 /*
222  * TLS records for TLS 1.0-1.2 can have the following header lengths:
223  * - 5 (AES-CBC with implicit IV)
224  * - 21 (AES-CBC with explicit IV)
225  * - 13 (AES-GCM with 8 byte explicit IV)
226  */
227 #define	MBUF_PEXT_HDR_LEN	23
228 
229 /*
230  * TLS records for TLS 1.0-1.2 can have the following maximum trailer
231  * lengths:
232  * - 16 (AES-GCM)
233  * - 36 (AES-CBC with SHA1 and up to 16 bytes of padding)
234  * - 48 (AES-CBC with SHA2-256 and up to 16 bytes of padding)
235  * - 64 (AES-CBC with SHA2-384 and up to 16 bytes of padding)
236  */
237 #define	MBUF_PEXT_TRAIL_LEN	64
238 
239 #if defined(__LP64__)
240 #define MBUF_PEXT_MAX_PGS (40 / sizeof(vm_paddr_t))
241 #else
242 #define MBUF_PEXT_MAX_PGS (72 / sizeof(vm_paddr_t))
243 #endif
244 
245 #define	MBUF_PEXT_MAX_BYTES						\
246     (MBUF_PEXT_MAX_PGS * PAGE_SIZE + MBUF_PEXT_HDR_LEN + MBUF_PEXT_TRAIL_LEN)
247 
248 struct ktls_session;
249 struct socket;
250 
251 /*
252  * Description of external storage mapped into mbuf; valid only if M_EXT is
253  * set.
254  * Size ILP32: 28
255  *	 LP64: 48
256  * Compile-time assertions in uipc_mbuf.c test these values to ensure that
257  * they are correct.
258  */
259 typedef	void m_ext_free_t(struct mbuf *);
260 struct m_ext {
261 	union {
262 		/*
263 		 * If EXT_FLAG_EMBREF is set, then we use refcount in the
264 		 * mbuf, the 'ext_count' member.  Otherwise, we have a
265 		 * shadow copy and we use pointer 'ext_cnt'.  The original
266 		 * mbuf is responsible to carry the pointer to free routine
267 		 * and its arguments.  They aren't copied into shadows in
268 		 * mb_dupcl() to avoid dereferencing next cachelines.
269 		 */
270 		volatile u_int	 ext_count;
271 		volatile u_int	*ext_cnt;
272 	};
273 	uint32_t	 ext_size;	/* size of buffer, for ext_free */
274 	uint32_t	 ext_type:8,	/* type of external storage */
275 			 ext_flags:24;	/* external storage mbuf flags */
276 	union {
277 		struct {
278 			/*
279 			 * Regular M_EXT mbuf:
280 			 * o ext_buf always points to the external buffer.
281 			 * o ext_free (below) and two optional arguments
282 			 *   ext_arg1 and ext_arg2 store the free context for
283 			 *   the external storage.  They are set only in the
284 			 *   refcount carrying mbuf, the one with
285 			 *   EXT_FLAG_EMBREF flag, with exclusion for
286 			 *   EXT_EXTREF type, where the free context is copied
287 			 *   into all mbufs that use same external storage.
288 			 */
289 			char 	*ext_buf;	/* start of buffer */
290 #define	m_ext_copylen	offsetof(struct m_ext, ext_arg2)
291 			void	*ext_arg2;
292 		};
293 		struct {
294 			/*
295 			 * Multi-page M_EXTPG mbuf:
296 			 * o extpg_pa - page vector.
297 			 * o extpg_trail and extpg_hdr - TLS trailer and
298 			 *   header.
299 			 * Uses ext_free and may also use ext_arg1.
300 			 */
301 			vm_paddr_t	extpg_pa[MBUF_PEXT_MAX_PGS];
302 			char		extpg_trail[MBUF_PEXT_TRAIL_LEN];
303 			char		extpg_hdr[MBUF_PEXT_HDR_LEN];
304 			/* Pretend these 3 fields are part of mbuf itself. */
305 #define	m_epg_pa	m_ext.extpg_pa
306 #define	m_epg_trail	m_ext.extpg_trail
307 #define	m_epg_hdr	m_ext.extpg_hdr
308 #define	m_epg_ext_copylen	offsetof(struct m_ext, ext_free)
309 		};
310 	};
311 	/*
312 	 * Free method and optional argument pointer, both
313 	 * used by M_EXT and M_EXTPG.
314 	 */
315 	m_ext_free_t	*ext_free;
316 	void		*ext_arg1;
317 };
318 
319 /*
320  * The core of the mbuf object along with some shortcut defines for practical
321  * purposes.
322  */
323 struct mbuf {
324 	/*
325 	 * Header present at the beginning of every mbuf.
326 	 * Size ILP32: 24
327 	 *      LP64: 32
328 	 * Compile-time assertions in uipc_mbuf.c test these values to ensure
329 	 * that they are correct.
330 	 */
331 	union {	/* next buffer in chain */
332 		struct mbuf		*m_next;
333 		SLIST_ENTRY(mbuf)	m_slist;
334 		STAILQ_ENTRY(mbuf)	m_stailq;
335 	};
336 	union {	/* next chain in queue/record */
337 		struct mbuf		*m_nextpkt;
338 		SLIST_ENTRY(mbuf)	m_slistpkt;
339 		STAILQ_ENTRY(mbuf)	m_stailqpkt;
340 	};
341 	caddr_t		 m_data;	/* location of data */
342 	int32_t		 m_len;		/* amount of data in this mbuf */
343 	uint32_t	 m_type:8,	/* type of data in this mbuf */
344 			 m_flags:24;	/* flags; see below */
345 #if !defined(__LP64__)
346 	uint32_t	 m_pad;		/* pad for 64bit alignment */
347 #endif
348 
349 	/*
350 	 * A set of optional headers (packet header, external storage header)
351 	 * and internal data storage.  Historically, these arrays were sized
352 	 * to MHLEN (space left after a packet header) and MLEN (space left
353 	 * after only a regular mbuf header); they are now variable size in
354 	 * order to support future work on variable-size mbufs.
355 	 */
356 	union {
357 		struct {
358 			union {
359 				/* M_PKTHDR set. */
360 				struct pkthdr	m_pkthdr;
361 
362 				/* M_EXTPG set.
363 				 * Multi-page M_EXTPG mbuf has its meta data
364 				 * split between the below anonymous structure
365 				 * and m_ext.  It carries vector of pages,
366 				 * optional header and trailer char vectors
367 				 * and pointers to socket/TLS data.
368 				 */
369 #define	m_epg_startcopy		m_epg_npgs
370 #define	m_epg_endcopy		m_epg_stailq
371 				struct {
372 					/* Overall count of pages and count of
373 					 * pages with I/O pending. */
374 					uint8_t	m_epg_npgs;
375 					uint8_t	m_epg_nrdy;
376 					/* TLS header and trailer lengths.
377 					 * The data itself resides in m_ext. */
378 					uint8_t	m_epg_hdrlen;
379 					uint8_t	m_epg_trllen;
380 					/* Offset into 1st page and length of
381 					 * data in the last page. */
382 					uint16_t m_epg_1st_off;
383 					uint16_t m_epg_last_len;
384 					uint8_t	m_epg_flags;
385 #define	EPG_FLAG_ANON	0x1	/* Data can be encrypted in place. */
386 #define	EPG_FLAG_2FREE	0x2	/* Scheduled for free. */
387 					uint8_t	m_epg_record_type;
388 					uint8_t	__spare[2];
389 					int	m_epg_enc_cnt;
390 					struct ktls_session *m_epg_tls;
391 					struct socket	*m_epg_so;
392 					uint64_t	m_epg_seqno;
393 					STAILQ_ENTRY(mbuf) m_epg_stailq;
394 				};
395 			};
396 			union {
397 				/* M_EXT or M_EXTPG set. */
398 				struct m_ext	m_ext;
399 				/* M_PKTHDR set, neither M_EXT nor M_EXTPG. */
400 				char		m_pktdat[0];
401 			};
402 		};
403 		char	m_dat[0];			/* !M_PKTHDR, !M_EXT */
404 	};
405 };
406 
407 #ifdef _KERNEL
408 static inline int
409 m_epg_pagelen(const struct mbuf *m, int pidx, int pgoff)
410 {
411 
412 	KASSERT(pgoff == 0 || pidx == 0,
413 	    ("page %d with non-zero offset %d in %p", pidx, pgoff, m));
414 
415 	if (pidx == m->m_epg_npgs - 1) {
416 		return (m->m_epg_last_len);
417 	} else {
418 		return (PAGE_SIZE - pgoff);
419 	}
420 }
421 
422 #ifdef INVARIANTS
423 #define	MCHECK(ex, msg)	KASSERT((ex),				\
424 	    ("Multi page mbuf %p with " #msg " at %s:%d",	\
425 	    m, __FILE__, __LINE__))
426 /*
427  * NB: This expects a non-empty buffer (npgs > 0 and
428  * last_pg_len > 0).
429  */
430 #define	MBUF_EXT_PGS_ASSERT_SANITY(m)	do {				\
431 	MCHECK(m->m_epg_npgs > 0, "no valid pages");		\
432 	MCHECK(m->m_epg_npgs <= nitems(m->m_epg_pa),		\
433 	    "too many pages");						\
434 	MCHECK(m->m_epg_nrdy <= m->m_epg_npgs,			\
435 	    "too many ready pages");					\
436 	MCHECK(m->m_epg_1st_off < PAGE_SIZE,			\
437 		"too large page offset");				\
438 	MCHECK(m->m_epg_last_len > 0, "zero last page length");	\
439 	MCHECK(m->m_epg_last_len <= PAGE_SIZE,			\
440 	    "too large last page length");				\
441 	if (m->m_epg_npgs == 1)					\
442 		MCHECK(m->m_epg_1st_off +			\
443 		    m->m_epg_last_len <=	 PAGE_SIZE,		\
444 		    "single page too large");				\
445 	MCHECK(m->m_epg_hdrlen <= sizeof(m->m_epg_hdr),		\
446 	    "too large header length");					\
447 	MCHECK(m->m_epg_trllen <= sizeof(m->m_epg_trail),	\
448 	    "too large header length");					\
449 } while (0)
450 #else
451 #define	MBUF_EXT_PGS_ASSERT_SANITY(m)	do {} while (0)
452 #endif
453 #endif
454 
455 /*
456  * mbuf flags of global significance and layer crossing.
457  * Those of only protocol/layer specific significance are to be mapped
458  * to M_PROTO[1-11] and cleared at layer handoff boundaries.
459  * NB: Limited to the lower 24 bits.
460  */
461 #define	M_EXT		0x00000001 /* has associated external storage */
462 #define	M_PKTHDR	0x00000002 /* start of record */
463 #define	M_EOR		0x00000004 /* end of record */
464 #define	M_RDONLY	0x00000008 /* associated data is marked read-only */
465 #define	M_BCAST		0x00000010 /* send/received as link-level broadcast */
466 #define	M_MCAST		0x00000020 /* send/received as link-level multicast */
467 #define	M_PROMISC	0x00000040 /* packet was not for us */
468 #define	M_VLANTAG	0x00000080 /* ether_vtag is valid */
469 #define	M_EXTPG		0x00000100 /* has array of unmapped pages and TLS */
470 #define	M_NOFREE	0x00000200 /* do not free mbuf, embedded in cluster */
471 #define	M_TSTMP		0x00000400 /* rcv_tstmp field is valid */
472 #define	M_TSTMP_HPREC	0x00000800 /* rcv_tstmp is high-prec, typically
473 				      hw-stamped on port (useful for IEEE 1588
474 				      and 802.1AS) */
475 #define M_TSTMP_LRO	0x00001000 /* Time LRO pushed in pkt is valid in (PH_loc) */
476 
477 #define	M_PROTO1	0x00002000 /* protocol-specific */
478 #define	M_PROTO2	0x00004000 /* protocol-specific */
479 #define	M_PROTO3	0x00008000 /* protocol-specific */
480 #define	M_PROTO4	0x00010000 /* protocol-specific */
481 #define	M_PROTO5	0x00020000 /* protocol-specific */
482 #define	M_PROTO6	0x00040000 /* protocol-specific */
483 #define	M_PROTO7	0x00080000 /* protocol-specific */
484 #define	M_PROTO8	0x00100000 /* protocol-specific */
485 #define	M_PROTO9	0x00200000 /* protocol-specific */
486 #define	M_PROTO10	0x00400000 /* protocol-specific */
487 #define	M_PROTO11	0x00800000 /* protocol-specific */
488 
489 /*
490  * Flags to purge when crossing layers.
491  */
492 #define	M_PROTOFLAGS \
493     (M_PROTO1|M_PROTO2|M_PROTO3|M_PROTO4|M_PROTO5|M_PROTO6|M_PROTO7|M_PROTO8|\
494      M_PROTO9|M_PROTO10|M_PROTO11)
495 
496 /*
497  * Flags preserved when copying m_pkthdr.
498  */
499 #define M_COPYFLAGS \
500     (M_PKTHDR|M_EOR|M_RDONLY|M_BCAST|M_MCAST|M_PROMISC|M_VLANTAG|M_TSTMP| \
501      M_TSTMP_HPREC|M_TSTMP_LRO|M_PROTOFLAGS)
502 
503 /*
504  * Flags preserved during demote.
505  */
506 #define	M_DEMOTEFLAGS \
507     (M_EXT | M_RDONLY | M_NOFREE | M_EXTPG)
508 
509 /*
510  * Mbuf flag description for use with printf(9) %b identifier.
511  */
512 #define	M_FLAG_BITS \
513     "\20\1M_EXT\2M_PKTHDR\3M_EOR\4M_RDONLY\5M_BCAST\6M_MCAST" \
514     "\7M_PROMISC\10M_VLANTAG\11M_EXTPG\12M_NOFREE\13M_TSTMP\14M_TSTMP_HPREC\15M_TSTMP_LRO"
515 #define	M_FLAG_PROTOBITS \
516     "\16M_PROTO1\17M_PROTO2\20M_PROTO3\21M_PROTO4" \
517     "\22M_PROTO5\23M_PROTO6\24M_PROTO7\25M_PROTO8\26M_PROTO9" \
518     "\27M_PROTO10\28M_PROTO11"
519 #define	M_FLAG_PRINTF (M_FLAG_BITS M_FLAG_PROTOBITS)
520 
521 /*
522  * Network interface cards are able to hash protocol fields (such as IPv4
523  * addresses and TCP port numbers) classify packets into flows.  These flows
524  * can then be used to maintain ordering while delivering packets to the OS
525  * via parallel input queues, as well as to provide a stateless affinity
526  * model.  NIC drivers can pass up the hash via m->m_pkthdr.flowid, and set
527  * m_flag fields to indicate how the hash should be interpreted by the
528  * network stack.
529  *
530  * Most NICs support RSS, which provides ordering and explicit affinity, and
531  * use the hash m_flag bits to indicate what header fields were covered by
532  * the hash.  M_HASHTYPE_OPAQUE and M_HASHTYPE_OPAQUE_HASH can be set by non-
533  * RSS cards or configurations that provide an opaque flow identifier, allowing
534  * for ordering and distribution without explicit affinity.  Additionally,
535  * M_HASHTYPE_OPAQUE_HASH indicates that the flow identifier has hash
536  * properties.
537  *
538  * The meaning of the IPV6_EX suffix:
539  * "o  Home address from the home address option in the IPv6 destination
540  *     options header.  If the extension header is not present, use the Source
541  *     IPv6 Address.
542  *  o  IPv6 address that is contained in the Routing-Header-Type-2 from the
543  *     associated extension header.  If the extension header is not present,
544  *     use the Destination IPv6 Address."
545  * Quoted from:
546  * https://docs.microsoft.com/en-us/windows-hardware/drivers/network/rss-hashing-types#ndishashipv6ex
547  */
548 #define	M_HASHTYPE_HASHPROP		0x80	/* has hash properties */
549 #define	M_HASHTYPE_INNER		0x40	/* calculated from inner headers */
550 #define	M_HASHTYPE_HASH(t)		(M_HASHTYPE_HASHPROP | (t))
551 /* Microsoft RSS standard hash types */
552 #define	M_HASHTYPE_NONE			0
553 #define	M_HASHTYPE_RSS_IPV4		M_HASHTYPE_HASH(1) /* IPv4 2-tuple */
554 #define	M_HASHTYPE_RSS_TCP_IPV4		M_HASHTYPE_HASH(2) /* TCPv4 4-tuple */
555 #define	M_HASHTYPE_RSS_IPV6		M_HASHTYPE_HASH(3) /* IPv6 2-tuple */
556 #define	M_HASHTYPE_RSS_TCP_IPV6		M_HASHTYPE_HASH(4) /* TCPv6 4-tuple */
557 #define	M_HASHTYPE_RSS_IPV6_EX		M_HASHTYPE_HASH(5) /* IPv6 2-tuple +
558 							    * ext hdrs */
559 #define	M_HASHTYPE_RSS_TCP_IPV6_EX	M_HASHTYPE_HASH(6) /* TCPv6 4-tuple +
560 							    * ext hdrs */
561 #define	M_HASHTYPE_RSS_UDP_IPV4		M_HASHTYPE_HASH(7) /* IPv4 UDP 4-tuple*/
562 #define	M_HASHTYPE_RSS_UDP_IPV6		M_HASHTYPE_HASH(9) /* IPv6 UDP 4-tuple*/
563 #define	M_HASHTYPE_RSS_UDP_IPV6_EX	M_HASHTYPE_HASH(10)/* IPv6 UDP 4-tuple +
564 							    * ext hdrs */
565 
566 #define	M_HASHTYPE_OPAQUE		0x3f	/* ordering, not affinity */
567 #define	M_HASHTYPE_OPAQUE_HASH		M_HASHTYPE_HASH(M_HASHTYPE_OPAQUE)
568 						/* ordering+hash, not affinity*/
569 
570 #define	M_HASHTYPE_CLEAR(m)	((m)->m_pkthdr.rsstype = 0)
571 #define	M_HASHTYPE_GET(m)	((m)->m_pkthdr.rsstype & ~M_HASHTYPE_INNER)
572 #define	M_HASHTYPE_SET(m, v)	((m)->m_pkthdr.rsstype = (v))
573 #define	M_HASHTYPE_TEST(m, v)	(M_HASHTYPE_GET(m) == (v))
574 #define	M_HASHTYPE_ISHASH(m)	\
575     (((m)->m_pkthdr.rsstype & M_HASHTYPE_HASHPROP) != 0)
576 #define	M_HASHTYPE_SETINNER(m)	do {			\
577 	(m)->m_pkthdr.rsstype |= M_HASHTYPE_INNER;	\
578     } while (0)
579 
580 /*
581  * External mbuf storage buffer types.
582  */
583 #define	EXT_CLUSTER	1	/* mbuf cluster */
584 #define	EXT_SFBUF	2	/* sendfile(2)'s sf_buf */
585 #define	EXT_JUMBOP	3	/* jumbo cluster page sized */
586 #define	EXT_JUMBO9	4	/* jumbo cluster 9216 bytes */
587 #define	EXT_JUMBO16	5	/* jumbo cluster 16184 bytes */
588 #define	EXT_PACKET	6	/* mbuf+cluster from packet zone */
589 #define	EXT_MBUF	7	/* external mbuf reference */
590 #define	EXT_RXRING	8	/* data in NIC receive ring */
591 
592 #define	EXT_VENDOR1	224	/* for vendor-internal use */
593 #define	EXT_VENDOR2	225	/* for vendor-internal use */
594 #define	EXT_VENDOR3	226	/* for vendor-internal use */
595 #define	EXT_VENDOR4	227	/* for vendor-internal use */
596 
597 #define	EXT_EXP1	244	/* for experimental use */
598 #define	EXT_EXP2	245	/* for experimental use */
599 #define	EXT_EXP3	246	/* for experimental use */
600 #define	EXT_EXP4	247	/* for experimental use */
601 
602 #define	EXT_NET_DRV	252	/* custom ext_buf provided by net driver(s) */
603 #define	EXT_MOD_TYPE	253	/* custom module's ext_buf type */
604 #define	EXT_DISPOSABLE	254	/* can throw this buffer away w/page flipping */
605 #define	EXT_EXTREF	255	/* has externally maintained ext_cnt ptr */
606 
607 /*
608  * Flags for external mbuf buffer types.
609  * NB: limited to the lower 24 bits.
610  */
611 #define	EXT_FLAG_EMBREF		0x000001	/* embedded ext_count */
612 #define	EXT_FLAG_EXTREF		0x000002	/* external ext_cnt, notyet */
613 
614 #define	EXT_FLAG_NOFREE		0x000010	/* don't free mbuf to pool, notyet */
615 
616 #define	EXT_FLAG_VENDOR1	0x010000	/* These flags are vendor */
617 #define	EXT_FLAG_VENDOR2	0x020000	/* or submodule specific, */
618 #define	EXT_FLAG_VENDOR3	0x040000	/* not used by mbuf code. */
619 #define	EXT_FLAG_VENDOR4	0x080000	/* Set/read by submodule. */
620 
621 #define	EXT_FLAG_EXP1		0x100000	/* for experimental use */
622 #define	EXT_FLAG_EXP2		0x200000	/* for experimental use */
623 #define	EXT_FLAG_EXP3		0x400000	/* for experimental use */
624 #define	EXT_FLAG_EXP4		0x800000	/* for experimental use */
625 
626 /*
627  * EXT flag description for use with printf(9) %b identifier.
628  */
629 #define	EXT_FLAG_BITS \
630     "\20\1EXT_FLAG_EMBREF\2EXT_FLAG_EXTREF\5EXT_FLAG_NOFREE" \
631     "\21EXT_FLAG_VENDOR1\22EXT_FLAG_VENDOR2\23EXT_FLAG_VENDOR3" \
632     "\24EXT_FLAG_VENDOR4\25EXT_FLAG_EXP1\26EXT_FLAG_EXP2\27EXT_FLAG_EXP3" \
633     "\30EXT_FLAG_EXP4"
634 
635 /*
636  * Flags indicating checksum, segmentation and other offload work to be
637  * done, or already done, by hardware or lower layers.  It is split into
638  * separate inbound and outbound flags.
639  *
640  * Outbound flags that are set by upper protocol layers requesting lower
641  * layers, or ideally the hardware, to perform these offloading tasks.
642  * For outbound packets this field and its flags can be directly tested
643  * against ifnet if_hwassist.  Note that the outbound and the inbound flags do
644  * not collide right now but they could be allowed to (as long as the flags are
645  * scrubbed appropriately when the direction of an mbuf changes).  CSUM_BITS
646  * would also have to split into CSUM_BITS_TX and CSUM_BITS_RX.
647  *
648  * CSUM_INNER_<x> is the same as CSUM_<x> but it applies to the inner frame.
649  * The CSUM_ENCAP_<x> bits identify the outer encapsulation.
650  */
651 #define	CSUM_IP			0x00000001	/* IP header checksum offload */
652 #define	CSUM_IP_UDP		0x00000002	/* UDP checksum offload */
653 #define	CSUM_IP_TCP		0x00000004	/* TCP checksum offload */
654 #define	CSUM_IP_SCTP		0x00000008	/* SCTP checksum offload */
655 #define	CSUM_IP_TSO		0x00000010	/* TCP segmentation offload */
656 #define	CSUM_IP_ISCSI		0x00000020	/* iSCSI checksum offload */
657 
658 #define	CSUM_INNER_IP6_UDP	0x00000040
659 #define	CSUM_INNER_IP6_TCP	0x00000080
660 #define	CSUM_INNER_IP6_TSO	0x00000100
661 #define	CSUM_IP6_UDP		0x00000200	/* UDP checksum offload */
662 #define	CSUM_IP6_TCP		0x00000400	/* TCP checksum offload */
663 #define	CSUM_IP6_SCTP		0x00000800	/* SCTP checksum offload */
664 #define	CSUM_IP6_TSO		0x00001000	/* TCP segmentation offload */
665 #define	CSUM_IP6_ISCSI		0x00002000	/* iSCSI checksum offload */
666 
667 #define	CSUM_INNER_IP		0x00004000
668 #define	CSUM_INNER_IP_UDP	0x00008000
669 #define	CSUM_INNER_IP_TCP	0x00010000
670 #define	CSUM_INNER_IP_TSO	0x00020000
671 
672 #define	CSUM_ENCAP_VXLAN	0x00040000	/* VXLAN outer encapsulation */
673 #define	CSUM_ENCAP_RSVD1	0x00080000
674 
675 /* Inbound checksum support where the checksum was verified by hardware. */
676 #define	CSUM_INNER_L3_CALC	0x00100000
677 #define	CSUM_INNER_L3_VALID	0x00200000
678 #define	CSUM_INNER_L4_CALC	0x00400000
679 #define	CSUM_INNER_L4_VALID	0x00800000
680 #define	CSUM_L3_CALC		0x01000000	/* calculated layer 3 csum */
681 #define	CSUM_L3_VALID		0x02000000	/* checksum is correct */
682 #define	CSUM_L4_CALC		0x04000000	/* calculated layer 4 csum */
683 #define	CSUM_L4_VALID		0x08000000	/* checksum is correct */
684 #define	CSUM_L5_CALC		0x10000000	/* calculated layer 5 csum */
685 #define	CSUM_L5_VALID		0x20000000	/* checksum is correct */
686 #define	CSUM_COALESCED		0x40000000	/* contains merged segments */
687 
688 #define	CSUM_SND_TAG		0x80000000	/* Packet header has send tag */
689 
690 #define CSUM_FLAGS_TX (CSUM_IP | CSUM_IP_UDP | CSUM_IP_TCP | CSUM_IP_SCTP | \
691     CSUM_IP_TSO | CSUM_IP_ISCSI | CSUM_INNER_IP6_UDP | CSUM_INNER_IP6_TCP | \
692     CSUM_INNER_IP6_TSO | CSUM_IP6_UDP | CSUM_IP6_TCP | CSUM_IP6_SCTP | \
693     CSUM_IP6_TSO | CSUM_IP6_ISCSI | CSUM_INNER_IP | CSUM_INNER_IP_UDP | \
694     CSUM_INNER_IP_TCP | CSUM_INNER_IP_TSO | CSUM_ENCAP_VXLAN | \
695     CSUM_ENCAP_RSVD1 | CSUM_SND_TAG)
696 
697 #define CSUM_FLAGS_RX (CSUM_INNER_L3_CALC | CSUM_INNER_L3_VALID | \
698     CSUM_INNER_L4_CALC | CSUM_INNER_L4_VALID | CSUM_L3_CALC | CSUM_L3_VALID | \
699     CSUM_L4_CALC | CSUM_L4_VALID | CSUM_L5_CALC | CSUM_L5_VALID | \
700     CSUM_COALESCED)
701 
702 /*
703  * CSUM flag description for use with printf(9) %b identifier.
704  */
705 #define	CSUM_BITS \
706     "\20\1CSUM_IP\2CSUM_IP_UDP\3CSUM_IP_TCP\4CSUM_IP_SCTP\5CSUM_IP_TSO" \
707     "\6CSUM_IP_ISCSI\7CSUM_INNER_IP6_UDP\10CSUM_INNER_IP6_TCP" \
708     "\11CSUM_INNER_IP6_TSO\12CSUM_IP6_UDP\13CSUM_IP6_TCP\14CSUM_IP6_SCTP" \
709     "\15CSUM_IP6_TSO\16CSUM_IP6_ISCSI\17CSUM_INNER_IP\20CSUM_INNER_IP_UDP" \
710     "\21CSUM_INNER_IP_TCP\22CSUM_INNER_IP_TSO\23CSUM_ENCAP_VXLAN" \
711     "\24CSUM_ENCAP_RSVD1\25CSUM_INNER_L3_CALC\26CSUM_INNER_L3_VALID" \
712     "\27CSUM_INNER_L4_CALC\30CSUM_INNER_L4_VALID\31CSUM_L3_CALC" \
713     "\32CSUM_L3_VALID\33CSUM_L4_CALC\34CSUM_L4_VALID\35CSUM_L5_CALC" \
714     "\36CSUM_L5_VALID\37CSUM_COALESCED\40CSUM_SND_TAG"
715 
716 /* CSUM flags compatibility mappings. */
717 #define	CSUM_IP_CHECKED		CSUM_L3_CALC
718 #define	CSUM_IP_VALID		CSUM_L3_VALID
719 #define	CSUM_DATA_VALID		CSUM_L4_VALID
720 #define	CSUM_PSEUDO_HDR		CSUM_L4_CALC
721 #define	CSUM_SCTP_VALID		CSUM_L4_VALID
722 #define	CSUM_DELAY_DATA		(CSUM_TCP|CSUM_UDP)
723 #define	CSUM_DELAY_IP		CSUM_IP		/* Only v4, no v6 IP hdr csum */
724 #define	CSUM_DELAY_DATA_IPV6	(CSUM_TCP_IPV6|CSUM_UDP_IPV6)
725 #define	CSUM_DATA_VALID_IPV6	CSUM_DATA_VALID
726 #define	CSUM_TCP		CSUM_IP_TCP
727 #define	CSUM_UDP		CSUM_IP_UDP
728 #define	CSUM_SCTP		CSUM_IP_SCTP
729 #define	CSUM_TSO		(CSUM_IP_TSO|CSUM_IP6_TSO)
730 #define	CSUM_INNER_TSO		(CSUM_INNER_IP_TSO|CSUM_INNER_IP6_TSO)
731 #define	CSUM_UDP_IPV6		CSUM_IP6_UDP
732 #define	CSUM_TCP_IPV6		CSUM_IP6_TCP
733 #define	CSUM_SCTP_IPV6		CSUM_IP6_SCTP
734 #define	CSUM_TLS_MASK		(CSUM_L5_CALC|CSUM_L5_VALID)
735 #define	CSUM_TLS_DECRYPTED	CSUM_L5_CALC
736 
737 /*
738  * mbuf types describing the content of the mbuf (including external storage).
739  */
740 #define	MT_NOTMBUF	0	/* USED INTERNALLY ONLY! Object is not mbuf */
741 #define	MT_DATA		1	/* dynamic (data) allocation */
742 #define	MT_HEADER	MT_DATA	/* packet header, use M_PKTHDR instead */
743 
744 #define	MT_VENDOR1	4	/* for vendor-internal use */
745 #define	MT_VENDOR2	5	/* for vendor-internal use */
746 #define	MT_VENDOR3	6	/* for vendor-internal use */
747 #define	MT_VENDOR4	7	/* for vendor-internal use */
748 
749 #define	MT_SONAME	8	/* socket name */
750 
751 #define	MT_EXP1		9	/* for experimental use */
752 #define	MT_EXP2		10	/* for experimental use */
753 #define	MT_EXP3		11	/* for experimental use */
754 #define	MT_EXP4		12	/* for experimental use */
755 
756 #define	MT_CONTROL	14	/* extra-data protocol message */
757 #define	MT_EXTCONTROL	15	/* control message with externalized contents */
758 #define	MT_OOBDATA	16	/* expedited data  */
759 
760 #define	MT_NOINIT	255	/* Not a type but a flag to allocate
761 				   a non-initialized mbuf */
762 
763 /*
764  * String names of mbuf-related UMA(9) and malloc(9) types.  Exposed to
765  * !_KERNEL so that monitoring tools can look up the zones with
766  * libmemstat(3).
767  */
768 #define	MBUF_MEM_NAME		"mbuf"
769 #define	MBUF_CLUSTER_MEM_NAME	"mbuf_cluster"
770 #define	MBUF_PACKET_MEM_NAME	"mbuf_packet"
771 #define	MBUF_JUMBOP_MEM_NAME	"mbuf_jumbo_page"
772 #define	MBUF_JUMBO9_MEM_NAME	"mbuf_jumbo_9k"
773 #define	MBUF_JUMBO16_MEM_NAME	"mbuf_jumbo_16k"
774 #define	MBUF_TAG_MEM_NAME	"mbuf_tag"
775 #define	MBUF_EXTREFCNT_MEM_NAME	"mbuf_ext_refcnt"
776 #define	MBUF_EXTPGS_MEM_NAME	"mbuf_extpgs"
777 
778 #ifdef _KERNEL
779 union if_snd_tag_alloc_params;
780 
781 #ifdef WITNESS
782 #define	MBUF_CHECKSLEEP(how) do {					\
783 	if (how == M_WAITOK)						\
784 		WITNESS_WARN(WARN_GIANTOK | WARN_SLEEPOK, NULL,		\
785 		    "Sleeping in \"%s\"", __func__);			\
786 } while (0)
787 #else
788 #define	MBUF_CHECKSLEEP(how) do {} while (0)
789 #endif
790 
791 /*
792  * Network buffer allocation API
793  *
794  * The rest of it is defined in kern/kern_mbuf.c
795  */
796 extern uma_zone_t	zone_mbuf;
797 extern uma_zone_t	zone_clust;
798 extern uma_zone_t	zone_pack;
799 extern uma_zone_t	zone_jumbop;
800 extern uma_zone_t	zone_jumbo9;
801 extern uma_zone_t	zone_jumbo16;
802 extern uma_zone_t	zone_extpgs;
803 
804 void		 mb_dupcl(struct mbuf *, struct mbuf *);
805 void		 mb_free_ext(struct mbuf *);
806 void		 mb_free_extpg(struct mbuf *);
807 void		 mb_free_mext_pgs(struct mbuf *);
808 struct mbuf	*mb_alloc_ext_pgs(int, m_ext_free_t);
809 struct mbuf	*mb_alloc_ext_plus_pages(int, int);
810 struct mbuf	*mb_mapped_to_unmapped(struct mbuf *, int, int, int,
811 		    struct mbuf **);
812 int		 mb_unmapped_compress(struct mbuf *m);
813 struct mbuf 	*mb_unmapped_to_ext(struct mbuf *m);
814 void		 mb_free_notready(struct mbuf *m, int count);
815 void		 m_adj(struct mbuf *, int);
816 void		 m_adj_decap(struct mbuf *, int);
817 int		 m_apply(struct mbuf *, int, int,
818 		    int (*)(void *, void *, u_int), void *);
819 int		 m_append(struct mbuf *, int, c_caddr_t);
820 void		 m_cat(struct mbuf *, struct mbuf *);
821 void		 m_catpkt(struct mbuf *, struct mbuf *);
822 int		 m_clget(struct mbuf *m, int how);
823 void 		*m_cljget(struct mbuf *m, int how, int size);
824 struct mbuf	*m_collapse(struct mbuf *, int, int);
825 void		 m_copyback(struct mbuf *, int, int, c_caddr_t);
826 void		 m_copydata(const struct mbuf *, int, int, caddr_t);
827 struct mbuf	*m_copym(struct mbuf *, int, int, int);
828 struct mbuf	*m_copypacket(struct mbuf *, int);
829 void		 m_copy_pkthdr(struct mbuf *, struct mbuf *);
830 struct mbuf	*m_copyup(struct mbuf *, int, int);
831 struct mbuf	*m_defrag(struct mbuf *, int);
832 void		 m_demote_pkthdr(struct mbuf *);
833 void		 m_demote(struct mbuf *, int, int);
834 struct mbuf	*m_devget(char *, int, int, struct ifnet *,
835 		    void (*)(char *, caddr_t, u_int));
836 void		 m_dispose_extcontrolm(struct mbuf *m);
837 struct mbuf	*m_dup(const struct mbuf *, int);
838 int		 m_dup_pkthdr(struct mbuf *, const struct mbuf *, int);
839 void		 m_extadd(struct mbuf *, char *, u_int, m_ext_free_t,
840 		    void *, void *, int, int);
841 u_int		 m_fixhdr(struct mbuf *);
842 struct mbuf	*m_fragment(struct mbuf *, int, int);
843 void		 m_freem(struct mbuf *);
844 void		 m_free_raw(struct mbuf *);
845 struct mbuf	*m_get2(int, int, short, int);
846 struct mbuf	*m_get3(int, int, short, int);
847 struct mbuf	*m_getjcl(int, short, int, int);
848 struct mbuf	*m_getm2(struct mbuf *, int, int, short, int);
849 struct mbuf	*m_getptr(struct mbuf *, int, int *);
850 u_int		 m_length(struct mbuf *, struct mbuf **);
851 int		 m_mbuftouio(struct uio *, const struct mbuf *, int);
852 void		 m_move_pkthdr(struct mbuf *, struct mbuf *);
853 int		 m_pkthdr_init(struct mbuf *, int);
854 struct mbuf	*m_prepend(struct mbuf *, int, int);
855 void		 m_print(const struct mbuf *, int);
856 struct mbuf	*m_pulldown(struct mbuf *, int, int, int *);
857 struct mbuf	*m_pullup(struct mbuf *, int);
858 int		 m_sanity(struct mbuf *, int);
859 struct mbuf	*m_split(struct mbuf *, int, int);
860 struct mbuf	*m_uiotombuf(struct uio *, int, int, int, int);
861 int		 m_unmapped_uiomove(const struct mbuf *, int, struct uio *,
862 		    int);
863 struct mbuf	*m_unshare(struct mbuf *, int);
864 int		 m_snd_tag_alloc(struct ifnet *,
865 		    union if_snd_tag_alloc_params *, struct m_snd_tag **);
866 void		 m_snd_tag_init(struct m_snd_tag *, struct ifnet *,
867 		    const struct if_snd_tag_sw *);
868 void		 m_snd_tag_destroy(struct m_snd_tag *);
869 void		 m_rcvif_serialize(struct mbuf *);
870 struct ifnet	*m_rcvif_restore(struct mbuf *);
871 
872 static __inline int
873 m_gettype(int size)
874 {
875 	int type;
876 
877 	switch (size) {
878 	case MSIZE:
879 		type = EXT_MBUF;
880 		break;
881 	case MCLBYTES:
882 		type = EXT_CLUSTER;
883 		break;
884 #if MJUMPAGESIZE != MCLBYTES
885 	case MJUMPAGESIZE:
886 		type = EXT_JUMBOP;
887 		break;
888 #endif
889 	case MJUM9BYTES:
890 		type = EXT_JUMBO9;
891 		break;
892 	case MJUM16BYTES:
893 		type = EXT_JUMBO16;
894 		break;
895 	default:
896 		panic("%s: invalid cluster size %d", __func__, size);
897 	}
898 
899 	return (type);
900 }
901 
902 /*
903  * Associated an external reference counted buffer with an mbuf.
904  */
905 static __inline void
906 m_extaddref(struct mbuf *m, char *buf, u_int size, u_int *ref_cnt,
907     m_ext_free_t freef, void *arg1, void *arg2)
908 {
909 
910 	KASSERT(ref_cnt != NULL, ("%s: ref_cnt not provided", __func__));
911 
912 	atomic_add_int(ref_cnt, 1);
913 	m->m_flags |= M_EXT;
914 	m->m_ext.ext_buf = buf;
915 	m->m_ext.ext_cnt = ref_cnt;
916 	m->m_data = m->m_ext.ext_buf;
917 	m->m_ext.ext_size = size;
918 	m->m_ext.ext_free = freef;
919 	m->m_ext.ext_arg1 = arg1;
920 	m->m_ext.ext_arg2 = arg2;
921 	m->m_ext.ext_type = EXT_EXTREF;
922 	m->m_ext.ext_flags = 0;
923 }
924 
925 static __inline uma_zone_t
926 m_getzone(int size)
927 {
928 	uma_zone_t zone;
929 
930 	switch (size) {
931 	case MCLBYTES:
932 		zone = zone_clust;
933 		break;
934 #if MJUMPAGESIZE != MCLBYTES
935 	case MJUMPAGESIZE:
936 		zone = zone_jumbop;
937 		break;
938 #endif
939 	case MJUM9BYTES:
940 		zone = zone_jumbo9;
941 		break;
942 	case MJUM16BYTES:
943 		zone = zone_jumbo16;
944 		break;
945 	default:
946 		panic("%s: invalid cluster size %d", __func__, size);
947 	}
948 
949 	return (zone);
950 }
951 
952 /*
953  * Initialize an mbuf with linear storage.
954  *
955  * Inline because the consumer text overhead will be roughly the same to
956  * initialize or call a function with this many parameters and M_PKTHDR
957  * should go away with constant propagation for !MGETHDR.
958  */
959 static __inline int
960 m_init(struct mbuf *m, int how, short type, int flags)
961 {
962 	int error;
963 
964 	m->m_next = NULL;
965 	m->m_nextpkt = NULL;
966 	m->m_data = m->m_dat;
967 	m->m_len = 0;
968 	m->m_flags = flags;
969 	m->m_type = type;
970 	if (flags & M_PKTHDR)
971 		error = m_pkthdr_init(m, how);
972 	else
973 		error = 0;
974 
975 	MBUF_PROBE5(m__init, m, how, type, flags, error);
976 	return (error);
977 }
978 
979 static __inline struct mbuf *
980 m_get_raw(int how, short type)
981 {
982 	struct mbuf *m;
983 	struct mb_args args;
984 
985 	args.flags = 0;
986 	args.type = type | MT_NOINIT;
987 	m = uma_zalloc_arg(zone_mbuf, &args, how);
988 	MBUF_PROBE3(m__get_raw, how, type, m);
989 	return (m);
990 }
991 
992 static __inline struct mbuf *
993 m_get(int how, short type)
994 {
995 	struct mbuf *m;
996 	struct mb_args args;
997 
998 	args.flags = 0;
999 	args.type = type;
1000 	m = uma_zalloc_arg(zone_mbuf, &args, how);
1001 	MBUF_PROBE3(m__get, how, type, m);
1002 	return (m);
1003 }
1004 
1005 static __inline struct mbuf *
1006 m_gethdr_raw(int how, short type)
1007 {
1008 	struct mbuf *m;
1009 	struct mb_args args;
1010 
1011 	args.flags = M_PKTHDR;
1012 	args.type = type | MT_NOINIT;
1013 	m = uma_zalloc_arg(zone_mbuf, &args, how);
1014 	MBUF_PROBE3(m__gethdr_raw, how, type, m);
1015 	return (m);
1016 }
1017 
1018 static __inline struct mbuf *
1019 m_gethdr(int how, short type)
1020 {
1021 	struct mbuf *m;
1022 	struct mb_args args;
1023 
1024 	args.flags = M_PKTHDR;
1025 	args.type = type;
1026 	m = uma_zalloc_arg(zone_mbuf, &args, how);
1027 	MBUF_PROBE3(m__gethdr, how, type, m);
1028 	return (m);
1029 }
1030 
1031 static __inline struct mbuf *
1032 m_getcl(int how, short type, int flags)
1033 {
1034 	struct mbuf *m;
1035 	struct mb_args args;
1036 
1037 	args.flags = flags;
1038 	args.type = type;
1039 	m = uma_zalloc_arg(zone_pack, &args, how);
1040 	MBUF_PROBE4(m__getcl, how, type, flags, m);
1041 	return (m);
1042 }
1043 
1044 /*
1045  * XXX: m_cljset() is a dangerous API.  One must attach only a new,
1046  * unreferenced cluster to an mbuf(9).  It is not possible to assert
1047  * that, so care can be taken only by users of the API.
1048  */
1049 static __inline void
1050 m_cljset(struct mbuf *m, void *cl, int type)
1051 {
1052 	int size;
1053 
1054 	switch (type) {
1055 	case EXT_CLUSTER:
1056 		size = MCLBYTES;
1057 		break;
1058 #if MJUMPAGESIZE != MCLBYTES
1059 	case EXT_JUMBOP:
1060 		size = MJUMPAGESIZE;
1061 		break;
1062 #endif
1063 	case EXT_JUMBO9:
1064 		size = MJUM9BYTES;
1065 		break;
1066 	case EXT_JUMBO16:
1067 		size = MJUM16BYTES;
1068 		break;
1069 	default:
1070 		panic("%s: unknown cluster type %d", __func__, type);
1071 		break;
1072 	}
1073 
1074 	m->m_data = m->m_ext.ext_buf = cl;
1075 	m->m_ext.ext_free = m->m_ext.ext_arg1 = m->m_ext.ext_arg2 = NULL;
1076 	m->m_ext.ext_size = size;
1077 	m->m_ext.ext_type = type;
1078 	m->m_ext.ext_flags = EXT_FLAG_EMBREF;
1079 	m->m_ext.ext_count = 1;
1080 	m->m_flags |= M_EXT;
1081 	MBUF_PROBE3(m__cljset, m, cl, type);
1082 }
1083 
1084 static __inline void
1085 m_chtype(struct mbuf *m, short new_type)
1086 {
1087 
1088 	m->m_type = new_type;
1089 }
1090 
1091 static __inline void
1092 m_clrprotoflags(struct mbuf *m)
1093 {
1094 
1095 	while (m) {
1096 		m->m_flags &= ~M_PROTOFLAGS;
1097 		m = m->m_next;
1098 	}
1099 }
1100 
1101 static __inline struct mbuf *
1102 m_last(struct mbuf *m)
1103 {
1104 
1105 	while (m->m_next)
1106 		m = m->m_next;
1107 	return (m);
1108 }
1109 
1110 static inline u_int
1111 m_extrefcnt(struct mbuf *m)
1112 {
1113 
1114 	KASSERT(m->m_flags & M_EXT, ("%s: M_EXT missing", __func__));
1115 
1116 	return ((m->m_ext.ext_flags & EXT_FLAG_EMBREF) ? m->m_ext.ext_count :
1117 	    *m->m_ext.ext_cnt);
1118 }
1119 
1120 /*
1121  * mbuf, cluster, and external object allocation macros (for compatibility
1122  * purposes).
1123  */
1124 #define	M_MOVE_PKTHDR(to, from)	m_move_pkthdr((to), (from))
1125 #define	MGET(m, how, type)	((m) = m_get((how), (type)))
1126 #define	MGETHDR(m, how, type)	((m) = m_gethdr((how), (type)))
1127 #define	MCLGET(m, how)		m_clget((m), (how))
1128 #define	MEXTADD(m, buf, size, free, arg1, arg2, flags, type)		\
1129     m_extadd((m), (char *)(buf), (size), (free), (arg1), (arg2),	\
1130     (flags), (type))
1131 #define	m_getm(m, len, how, type)					\
1132     m_getm2((m), (len), (how), (type), M_PKTHDR)
1133 
1134 /*
1135  * Evaluate TRUE if it's safe to write to the mbuf m's data region (this can
1136  * be both the local data payload, or an external buffer area, depending on
1137  * whether M_EXT is set).
1138  */
1139 #define	M_WRITABLE(m)	(((m)->m_flags & (M_RDONLY | M_EXTPG)) == 0 &&	\
1140 			 (!(((m)->m_flags & M_EXT)) ||			\
1141 			 (m_extrefcnt(m) == 1)))
1142 
1143 /* Check if the supplied mbuf has a packet header, or else panic. */
1144 #define	M_ASSERTPKTHDR(m)						\
1145 	KASSERT((m) != NULL && (m)->m_flags & M_PKTHDR,			\
1146 	    ("%s: no mbuf packet header!", __func__))
1147 
1148 /* Check if the supplied mbuf has no send tag, or else panic. */
1149 #define	M_ASSERT_NO_SND_TAG(m)						\
1150 	KASSERT((m) != NULL && (m)->m_flags & M_PKTHDR &&		\
1151 	       ((m)->m_pkthdr.csum_flags & CSUM_SND_TAG) == 0,		\
1152 	    ("%s: receive mbuf has send tag!", __func__))
1153 
1154 /* Check if mbuf is multipage. */
1155 #define M_ASSERTEXTPG(m)						\
1156 	KASSERT(((m)->m_flags & (M_EXTPG|M_PKTHDR)) == M_EXTPG,		\
1157 	    ("%s: m %p is not multipage!", __func__, m))
1158 
1159 /*
1160  * Ensure that the supplied mbuf is a valid, non-free mbuf.
1161  *
1162  * XXX: Broken at the moment.  Need some UMA magic to make it work again.
1163  */
1164 #define	M_ASSERTVALID(m)						\
1165 	KASSERT((((struct mbuf *)m)->m_flags & 0) == 0,			\
1166 	    ("%s: attempted use of a free mbuf!", __func__))
1167 
1168 /* Check whether any mbuf in the chain is unmapped. */
1169 #ifdef INVARIANTS
1170 #define	M_ASSERTMAPPED(m) do {						\
1171 	for (struct mbuf *__m = (m); __m != NULL; __m = __m->m_next)	\
1172 		KASSERT((__m->m_flags & M_EXTPG) == 0,			\
1173 		    ("%s: chain %p contains an unmapped mbuf", __func__, (m)));\
1174 } while (0)
1175 #else
1176 #define	M_ASSERTMAPPED(m) do {} while (0)
1177 #endif
1178 
1179 /*
1180  * Return the address of the start of the buffer associated with an mbuf,
1181  * handling external storage, packet-header mbufs, and regular data mbufs.
1182  */
1183 #define	M_START(m)							\
1184 	(((m)->m_flags & M_EXTPG) ? NULL :				\
1185 	 ((m)->m_flags & M_EXT) ? (m)->m_ext.ext_buf :			\
1186 	 ((m)->m_flags & M_PKTHDR) ? &(m)->m_pktdat[0] :		\
1187 	 &(m)->m_dat[0])
1188 
1189 /*
1190  * Return the size of the buffer associated with an mbuf, handling external
1191  * storage, packet-header mbufs, and regular data mbufs.
1192  */
1193 #define	M_SIZE(m)							\
1194 	(((m)->m_flags & M_EXT) ? (m)->m_ext.ext_size :			\
1195 	 ((m)->m_flags & M_PKTHDR) ? MHLEN :				\
1196 	 MLEN)
1197 
1198 /*
1199  * Set the m_data pointer of a newly allocated mbuf to place an object of the
1200  * specified size at the end of the mbuf, longword aligned.
1201  *
1202  * NB: Historically, we had M_ALIGN(), MH_ALIGN(), and MEXT_ALIGN() as
1203  * separate macros, each asserting that it was called at the proper moment.
1204  * This required callers to themselves test the storage type and call the
1205  * right one.  Rather than require callers to be aware of those layout
1206  * decisions, we centralize here.
1207  */
1208 static __inline void
1209 m_align(struct mbuf *m, int len)
1210 {
1211 #ifdef INVARIANTS
1212 	const char *msg = "%s: not a virgin mbuf";
1213 #endif
1214 	int adjust;
1215 
1216 	KASSERT(m->m_data == M_START(m), (msg, __func__));
1217 
1218 	adjust = M_SIZE(m) - len;
1219 	m->m_data += adjust &~ (sizeof(long)-1);
1220 }
1221 
1222 #define	M_ALIGN(m, len)		m_align(m, len)
1223 #define	MH_ALIGN(m, len)	m_align(m, len)
1224 #define	MEXT_ALIGN(m, len)	m_align(m, len)
1225 
1226 /*
1227  * Compute the amount of space available before the current start of data in
1228  * an mbuf.
1229  *
1230  * The M_WRITABLE() is a temporary, conservative safety measure: the burden
1231  * of checking writability of the mbuf data area rests solely with the caller.
1232  *
1233  * NB: In previous versions, M_LEADINGSPACE() would only check M_WRITABLE()
1234  * for mbufs with external storage.  We now allow mbuf-embedded data to be
1235  * read-only as well.
1236  */
1237 #define	M_LEADINGSPACE(m)						\
1238 	(M_WRITABLE(m) ? ((m)->m_data - M_START(m)) : 0)
1239 
1240 /*
1241  * Compute the amount of space available after the end of data in an mbuf.
1242  *
1243  * The M_WRITABLE() is a temporary, conservative safety measure: the burden
1244  * of checking writability of the mbuf data area rests solely with the caller.
1245  *
1246  * NB: In previous versions, M_TRAILINGSPACE() would only check M_WRITABLE()
1247  * for mbufs with external storage.  We now allow mbuf-embedded data to be
1248  * read-only as well.
1249  */
1250 #define	M_TRAILINGSPACE(m)						\
1251 	(M_WRITABLE(m) ?						\
1252 	    ((M_START(m) + M_SIZE(m)) - ((m)->m_data + (m)->m_len)) : 0)
1253 
1254 /*
1255  * Arrange to prepend space of size plen to mbuf m.  If a new mbuf must be
1256  * allocated, how specifies whether to wait.  If the allocation fails, the
1257  * original mbuf chain is freed and m is set to NULL.
1258  */
1259 #define	M_PREPEND(m, plen, how) do {					\
1260 	struct mbuf **_mmp = &(m);					\
1261 	struct mbuf *_mm = *_mmp;					\
1262 	int _mplen = (plen);						\
1263 	int __mhow = (how);						\
1264 									\
1265 	MBUF_CHECKSLEEP(how);						\
1266 	if (M_LEADINGSPACE(_mm) >= _mplen) {				\
1267 		_mm->m_data -= _mplen;					\
1268 		_mm->m_len += _mplen;					\
1269 	} else								\
1270 		_mm = m_prepend(_mm, _mplen, __mhow);			\
1271 	if (_mm != NULL && _mm->m_flags & M_PKTHDR)			\
1272 		_mm->m_pkthdr.len += _mplen;				\
1273 	*_mmp = _mm;							\
1274 } while (0)
1275 
1276 /*
1277  * Change mbuf to new type.  This is a relatively expensive operation and
1278  * should be avoided.
1279  */
1280 #define	MCHTYPE(m, t)	m_chtype((m), (t))
1281 
1282 /* Return the rcvif of a packet header. */
1283 static __inline struct ifnet *
1284 m_rcvif(struct mbuf *m)
1285 {
1286 
1287 	M_ASSERTPKTHDR(m);
1288 	if (m->m_pkthdr.csum_flags & CSUM_SND_TAG)
1289 		return (NULL);
1290 	return (m->m_pkthdr.rcvif);
1291 }
1292 
1293 /* Length to m_copy to copy all. */
1294 #define	M_COPYALL	1000000000
1295 
1296 extern int		max_datalen;	/* MHLEN - max_hdr */
1297 extern int		max_hdr;	/* Largest link + protocol header */
1298 extern int		max_linkhdr;	/* Largest link-level header */
1299 extern int		max_protohdr;	/* Largest protocol header */
1300 extern int		nmbclusters;	/* Maximum number of clusters */
1301 extern bool		mb_use_ext_pgs;	/* Use ext_pgs for sendfile */
1302 
1303 /*-
1304  * Network packets may have annotations attached by affixing a list of
1305  * "packet tags" to the pkthdr structure.  Packet tags are dynamically
1306  * allocated semi-opaque data structures that have a fixed header
1307  * (struct m_tag) that specifies the size of the memory block and a
1308  * <cookie,type> pair that identifies it.  The cookie is a 32-bit unique
1309  * unsigned value used to identify a module or ABI.  By convention this value
1310  * is chosen as the date+time that the module is created, expressed as the
1311  * number of seconds since the epoch (e.g., using date -u +'%s').  The type
1312  * value is an ABI/module-specific value that identifies a particular
1313  * annotation and is private to the module.  For compatibility with systems
1314  * like OpenBSD that define packet tags w/o an ABI/module cookie, the value
1315  * PACKET_ABI_COMPAT is used to implement m_tag_get and m_tag_find
1316  * compatibility shim functions and several tag types are defined below.
1317  * Users that do not require compatibility should use a private cookie value
1318  * so that packet tag-related definitions can be maintained privately.
1319  *
1320  * Note that the packet tag returned by m_tag_alloc has the default memory
1321  * alignment implemented by malloc.  To reference private data one can use a
1322  * construct like:
1323  *
1324  *	struct m_tag *mtag = m_tag_alloc(...);
1325  *	struct foo *p = (struct foo *)(mtag+1);
1326  *
1327  * if the alignment of struct m_tag is sufficient for referencing members of
1328  * struct foo.  Otherwise it is necessary to embed struct m_tag within the
1329  * private data structure to insure proper alignment; e.g.,
1330  *
1331  *	struct foo {
1332  *		struct m_tag	tag;
1333  *		...
1334  *	};
1335  *	struct foo *p = (struct foo *) m_tag_alloc(...);
1336  *	struct m_tag *mtag = &p->tag;
1337  */
1338 
1339 /*
1340  * Persistent tags stay with an mbuf until the mbuf is reclaimed.  Otherwise
1341  * tags are expected to ``vanish'' when they pass through a network
1342  * interface.  For most interfaces this happens normally as the tags are
1343  * reclaimed when the mbuf is free'd.  However in some special cases
1344  * reclaiming must be done manually.  An example is packets that pass through
1345  * the loopback interface.  Also, one must be careful to do this when
1346  * ``turning around'' packets (e.g., icmp_reflect).
1347  *
1348  * To mark a tag persistent bit-or this flag in when defining the tag id.
1349  * The tag will then be treated as described above.
1350  */
1351 #define	MTAG_PERSISTENT				0x800
1352 
1353 #define	PACKET_TAG_NONE				0  /* Nadda */
1354 
1355 /* Packet tags for use with PACKET_ABI_COMPAT. */
1356 #define	PACKET_TAG_IPSEC_IN_DONE		1  /* IPsec applied, in */
1357 #define	PACKET_TAG_IPSEC_OUT_DONE		2  /* IPsec applied, out */
1358 #define	PACKET_TAG_IPSEC_IN_CRYPTO_DONE		3  /* NIC IPsec crypto done */
1359 #define	PACKET_TAG_IPSEC_OUT_CRYPTO_NEEDED	4  /* NIC IPsec crypto req'ed */
1360 #define	PACKET_TAG_IPSEC_IN_COULD_DO_CRYPTO	5  /* NIC notifies IPsec */
1361 #define	PACKET_TAG_IPSEC_PENDING_TDB		6  /* Reminder to do IPsec */
1362 #define	PACKET_TAG_BRIDGE			7  /* Bridge processing done */
1363 #define	PACKET_TAG_GIF				8  /* GIF processing done */
1364 #define	PACKET_TAG_GRE				9  /* GRE processing done */
1365 #define	PACKET_TAG_IN_PACKET_CHECKSUM		10 /* NIC checksumming done */
1366 #define	PACKET_TAG_ENCAP			11 /* Encap.  processing */
1367 #define	PACKET_TAG_IPSEC_SOCKET			12 /* IPSEC socket ref */
1368 #define	PACKET_TAG_IPSEC_HISTORY		13 /* IPSEC history */
1369 #define	PACKET_TAG_IPV6_INPUT			14 /* IPV6 input processing */
1370 #define	PACKET_TAG_DUMMYNET			15 /* dummynet info */
1371 #define	PACKET_TAG_DIVERT			17 /* divert info */
1372 #define	PACKET_TAG_IPFORWARD			18 /* ipforward info */
1373 #define	PACKET_TAG_MACLABEL	(19 | MTAG_PERSISTENT) /* MAC label */
1374 #define	PACKET_TAG_PF				21 /* PF/ALTQ information */
1375 #define	PACKET_TAG_RTSOCKFAM			25 /* rtsock sa family */
1376 #define	PACKET_TAG_IPOPTIONS			27 /* Saved IP options */
1377 #define	PACKET_TAG_CARP				28 /* CARP info */
1378 #define	PACKET_TAG_IPSEC_NAT_T_PORTS		29 /* two uint16_t */
1379 #define	PACKET_TAG_ND_OUTGOING			30 /* ND outgoing */
1380 
1381 /* Specific cookies and tags. */
1382 
1383 /* Packet tag routines. */
1384 struct m_tag	*m_tag_alloc(uint32_t, uint16_t, int, int);
1385 void		 m_tag_delete(struct mbuf *, struct m_tag *);
1386 void		 m_tag_delete_chain(struct mbuf *, struct m_tag *);
1387 void		 m_tag_free_default(struct m_tag *);
1388 struct m_tag	*m_tag_locate(struct mbuf *, uint32_t, uint16_t,
1389     struct m_tag *);
1390 struct m_tag	*m_tag_copy(struct m_tag *, int);
1391 int		 m_tag_copy_chain(struct mbuf *, const struct mbuf *, int);
1392 void		 m_tag_delete_nonpersistent(struct mbuf *);
1393 
1394 /*
1395  * Initialize the list of tags associated with an mbuf.
1396  */
1397 static __inline void
1398 m_tag_init(struct mbuf *m)
1399 {
1400 
1401 	SLIST_INIT(&m->m_pkthdr.tags);
1402 }
1403 
1404 /*
1405  * Set up the contents of a tag.  Note that this does not fill in the free
1406  * method; the caller is expected to do that.
1407  *
1408  * XXX probably should be called m_tag_init, but that was already taken.
1409  */
1410 static __inline void
1411 m_tag_setup(struct m_tag *t, uint32_t cookie, uint16_t type, int len)
1412 {
1413 
1414 	t->m_tag_id = type;
1415 	t->m_tag_len = len;
1416 	t->m_tag_cookie = cookie;
1417 }
1418 
1419 /*
1420  * Reclaim resources associated with a tag.
1421  */
1422 static __inline void
1423 m_tag_free(struct m_tag *t)
1424 {
1425 
1426 	(*t->m_tag_free)(t);
1427 }
1428 
1429 /*
1430  * Return the first tag associated with an mbuf.
1431  */
1432 static __inline struct m_tag *
1433 m_tag_first(struct mbuf *m)
1434 {
1435 
1436 	return (SLIST_FIRST(&m->m_pkthdr.tags));
1437 }
1438 
1439 /*
1440  * Return the next tag in the list of tags associated with an mbuf.
1441  */
1442 static __inline struct m_tag *
1443 m_tag_next(struct mbuf *m __unused, struct m_tag *t)
1444 {
1445 
1446 	return (SLIST_NEXT(t, m_tag_link));
1447 }
1448 
1449 /*
1450  * Prepend a tag to the list of tags associated with an mbuf.
1451  */
1452 static __inline void
1453 m_tag_prepend(struct mbuf *m, struct m_tag *t)
1454 {
1455 
1456 	SLIST_INSERT_HEAD(&m->m_pkthdr.tags, t, m_tag_link);
1457 }
1458 
1459 /*
1460  * Unlink a tag from the list of tags associated with an mbuf.
1461  */
1462 static __inline void
1463 m_tag_unlink(struct mbuf *m, struct m_tag *t)
1464 {
1465 
1466 	SLIST_REMOVE(&m->m_pkthdr.tags, t, m_tag, m_tag_link);
1467 }
1468 
1469 /* These are for OpenBSD compatibility. */
1470 #define	MTAG_ABI_COMPAT		0		/* compatibility ABI */
1471 
1472 static __inline struct m_tag *
1473 m_tag_get(uint16_t type, int length, int wait)
1474 {
1475 	return (m_tag_alloc(MTAG_ABI_COMPAT, type, length, wait));
1476 }
1477 
1478 static __inline struct m_tag *
1479 m_tag_find(struct mbuf *m, uint16_t type, struct m_tag *start)
1480 {
1481 	return (SLIST_EMPTY(&m->m_pkthdr.tags) ? (struct m_tag *)NULL :
1482 	    m_tag_locate(m, MTAG_ABI_COMPAT, type, start));
1483 }
1484 
1485 static inline struct m_snd_tag *
1486 m_snd_tag_ref(struct m_snd_tag *mst)
1487 {
1488 
1489 	refcount_acquire(&mst->refcount);
1490 	return (mst);
1491 }
1492 
1493 static inline void
1494 m_snd_tag_rele(struct m_snd_tag *mst)
1495 {
1496 
1497 	if (refcount_release(&mst->refcount))
1498 		m_snd_tag_destroy(mst);
1499 }
1500 
1501 static __inline struct mbuf *
1502 m_free(struct mbuf *m)
1503 {
1504 	struct mbuf *n = m->m_next;
1505 
1506 	MBUF_PROBE1(m__free, m);
1507 	if ((m->m_flags & (M_PKTHDR|M_NOFREE)) == (M_PKTHDR|M_NOFREE))
1508 		m_tag_delete_chain(m, NULL);
1509 	if (m->m_flags & M_PKTHDR && m->m_pkthdr.csum_flags & CSUM_SND_TAG)
1510 		m_snd_tag_rele(m->m_pkthdr.snd_tag);
1511 	if (m->m_flags & M_EXTPG)
1512 		mb_free_extpg(m);
1513 	else if (m->m_flags & M_EXT)
1514 		mb_free_ext(m);
1515 	else if ((m->m_flags & M_NOFREE) == 0)
1516 		uma_zfree(zone_mbuf, m);
1517 	return (n);
1518 }
1519 
1520 static __inline int
1521 rt_m_getfib(struct mbuf *m)
1522 {
1523 	KASSERT(m->m_flags & M_PKTHDR , ("Attempt to get FIB from non header mbuf."));
1524 	return (m->m_pkthdr.fibnum);
1525 }
1526 
1527 #define M_GETFIB(_m)   rt_m_getfib(_m)
1528 
1529 #define M_SETFIB(_m, _fib) do {						\
1530         KASSERT((_m)->m_flags & M_PKTHDR, ("Attempt to set FIB on non header mbuf."));	\
1531 	((_m)->m_pkthdr.fibnum) = (_fib);				\
1532 } while (0)
1533 
1534 /* flags passed as first argument for "m_xxx_tcpip_hash()" */
1535 #define	MBUF_HASHFLAG_L2	(1 << 2)
1536 #define	MBUF_HASHFLAG_L3	(1 << 3)
1537 #define	MBUF_HASHFLAG_L4	(1 << 4)
1538 
1539 /* mbuf hashing helper routines */
1540 uint32_t	m_ether_tcpip_hash_init(void);
1541 uint32_t	m_ether_tcpip_hash(const uint32_t, const struct mbuf *, uint32_t);
1542 uint32_t	m_infiniband_tcpip_hash_init(void);
1543 uint32_t	m_infiniband_tcpip_hash(const uint32_t, const struct mbuf *, uint32_t);
1544 
1545 #ifdef MBUF_PROFILING
1546  void m_profile(struct mbuf *m);
1547  #define M_PROFILE(m) m_profile(m)
1548 #else
1549  #define M_PROFILE(m)
1550 #endif
1551 
1552 struct mbufq {
1553 	STAILQ_HEAD(, mbuf)	mq_head;
1554 	int			mq_len;
1555 	int			mq_maxlen;
1556 };
1557 
1558 static inline void
1559 mbufq_init(struct mbufq *mq, int maxlen)
1560 {
1561 
1562 	STAILQ_INIT(&mq->mq_head);
1563 	mq->mq_maxlen = maxlen;
1564 	mq->mq_len = 0;
1565 }
1566 
1567 static inline struct mbuf *
1568 mbufq_flush(struct mbufq *mq)
1569 {
1570 	struct mbuf *m;
1571 
1572 	m = STAILQ_FIRST(&mq->mq_head);
1573 	STAILQ_INIT(&mq->mq_head);
1574 	mq->mq_len = 0;
1575 	return (m);
1576 }
1577 
1578 static inline void
1579 mbufq_drain(struct mbufq *mq)
1580 {
1581 	struct mbuf *m, *n;
1582 
1583 	n = mbufq_flush(mq);
1584 	while ((m = n) != NULL) {
1585 		n = STAILQ_NEXT(m, m_stailqpkt);
1586 		m_freem(m);
1587 	}
1588 }
1589 
1590 static inline struct mbuf *
1591 mbufq_first(const struct mbufq *mq)
1592 {
1593 
1594 	return (STAILQ_FIRST(&mq->mq_head));
1595 }
1596 
1597 static inline struct mbuf *
1598 mbufq_last(const struct mbufq *mq)
1599 {
1600 
1601 	return (STAILQ_LAST(&mq->mq_head, mbuf, m_stailqpkt));
1602 }
1603 
1604 static inline int
1605 mbufq_full(const struct mbufq *mq)
1606 {
1607 
1608 	return (mq->mq_maxlen > 0 && mq->mq_len >= mq->mq_maxlen);
1609 }
1610 
1611 static inline int
1612 mbufq_len(const struct mbufq *mq)
1613 {
1614 
1615 	return (mq->mq_len);
1616 }
1617 
1618 static inline int
1619 mbufq_enqueue(struct mbufq *mq, struct mbuf *m)
1620 {
1621 
1622 	if (mbufq_full(mq))
1623 		return (ENOBUFS);
1624 	STAILQ_INSERT_TAIL(&mq->mq_head, m, m_stailqpkt);
1625 	mq->mq_len++;
1626 	return (0);
1627 }
1628 
1629 static inline struct mbuf *
1630 mbufq_dequeue(struct mbufq *mq)
1631 {
1632 	struct mbuf *m;
1633 
1634 	m = STAILQ_FIRST(&mq->mq_head);
1635 	if (m) {
1636 		STAILQ_REMOVE_HEAD(&mq->mq_head, m_stailqpkt);
1637 		m->m_nextpkt = NULL;
1638 		mq->mq_len--;
1639 	}
1640 	return (m);
1641 }
1642 
1643 static inline void
1644 mbufq_prepend(struct mbufq *mq, struct mbuf *m)
1645 {
1646 
1647 	STAILQ_INSERT_HEAD(&mq->mq_head, m, m_stailqpkt);
1648 	mq->mq_len++;
1649 }
1650 
1651 /*
1652  * Note: this doesn't enforce the maximum list size for dst.
1653  */
1654 static inline void
1655 mbufq_concat(struct mbufq *mq_dst, struct mbufq *mq_src)
1656 {
1657 
1658 	mq_dst->mq_len += mq_src->mq_len;
1659 	STAILQ_CONCAT(&mq_dst->mq_head, &mq_src->mq_head);
1660 	mq_src->mq_len = 0;
1661 }
1662 
1663 #ifdef _SYS_TIMESPEC_H_
1664 static inline void
1665 mbuf_tstmp2timespec(struct mbuf *m, struct timespec *ts)
1666 {
1667 
1668 	KASSERT((m->m_flags & M_PKTHDR) != 0, ("mbuf %p no M_PKTHDR", m));
1669 	KASSERT((m->m_flags & (M_TSTMP|M_TSTMP_LRO)) != 0, ("mbuf %p no M_TSTMP or M_TSTMP_LRO", m));
1670 	ts->tv_sec = m->m_pkthdr.rcv_tstmp / 1000000000;
1671 	ts->tv_nsec = m->m_pkthdr.rcv_tstmp % 1000000000;
1672 }
1673 #endif
1674 
1675 #ifdef DEBUGNET
1676 /* Invoked from the debugnet client code. */
1677 void	debugnet_mbuf_drain(void);
1678 void	debugnet_mbuf_start(void);
1679 void	debugnet_mbuf_finish(void);
1680 void	debugnet_mbuf_reinit(int nmbuf, int nclust, int clsize);
1681 #endif
1682 
1683 static inline bool
1684 mbuf_has_tls_session(struct mbuf *m)
1685 {
1686 
1687 	if (m->m_flags & M_EXTPG) {
1688 		if (m->m_epg_tls != NULL) {
1689 			return (true);
1690 		}
1691 	}
1692 	return (false);
1693 }
1694 
1695 #endif /* _KERNEL */
1696 #endif /* !_SYS_MBUF_H_ */
1697