xref: /freebsd/sys/sys/mbuf.h (revision fe6060f10f634930ff71b7c50291ddc610da2475)
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 	};
163 	SLIST_HEAD(packet_tags, m_tag) tags; /* list of packet tags */
164 	int32_t		 len;		/* total packet length */
165 
166 	/* Layer crossing persistent information. */
167 	uint32_t	 flowid;	/* packet's 4-tuple system */
168 	uint32_t	 csum_flags;	/* checksum and offload features */
169 	uint16_t	 fibnum;	/* this packet should use this fib */
170 	uint8_t		 numa_domain;	/* NUMA domain of recvd pkt */
171 	uint8_t		 rsstype;	/* hash type */
172 	union {
173 		uint64_t	rcv_tstmp;	/* timestamp in ns */
174 		struct {
175 			uint8_t		 l2hlen;	/* layer 2 hdr len */
176 			uint8_t		 l3hlen;	/* layer 3 hdr len */
177 			uint8_t		 l4hlen;	/* layer 4 hdr len */
178 			uint8_t		 l5hlen;	/* layer 5 hdr len */
179 			uint8_t		 inner_l2hlen;
180 			uint8_t		 inner_l3hlen;
181 			uint8_t		 inner_l4hlen;
182 			uint8_t		 inner_l5hlen;
183 		};
184 	};
185 	union {
186 		uint8_t  eight[8];
187 		uint16_t sixteen[4];
188 		uint32_t thirtytwo[2];
189 		uint64_t sixtyfour[1];
190 		uintptr_t unintptr[1];
191 		void	*ptr;
192 	} PH_per;
193 
194 	/* Layer specific non-persistent local storage for reassembly, etc. */
195 	union {
196 		uint8_t  eight[8];
197 		uint16_t sixteen[4];
198 		uint32_t thirtytwo[2];
199 		uint64_t sixtyfour[1];
200 		uintptr_t unintptr[1];
201 		void 	*ptr;
202 	} PH_loc;
203 };
204 #define	ether_vtag	PH_per.sixteen[0]
205 #define tcp_tun_port	PH_per.sixteen[0] /* outbound */
206 #define	PH_vt		PH_per
207 #define	vt_nrecs	sixteen[0]	  /* mld and v6-ND */
208 #define	tso_segsz	PH_per.sixteen[1] /* inbound after LRO */
209 #define	lro_nsegs	tso_segsz	  /* inbound after LRO */
210 #define	csum_data	PH_per.thirtytwo[1] /* inbound from hardware up */
211 #define	lro_tcp_d_len	PH_loc.sixteen[0] /* inbound during LRO (no reassembly) */
212 #define	lro_tcp_d_csum	PH_loc.sixteen[1] /* inbound during LRO (no reassembly) */
213 #define	lro_tcp_h_off	PH_loc.sixteen[2] /* inbound during LRO (no reassembly) */
214 #define	lro_etype	PH_loc.sixteen[3] /* inbound during LRO (no reassembly) */
215 /* Note PH_loc is used during IP reassembly (all 8 bytes as a ptr) */
216 
217 /*
218  * TLS records for TLS 1.0-1.2 can have the following header lengths:
219  * - 5 (AES-CBC with implicit IV)
220  * - 21 (AES-CBC with explicit IV)
221  * - 13 (AES-GCM with 8 byte explicit IV)
222  */
223 #define	MBUF_PEXT_HDR_LEN	23
224 
225 /*
226  * TLS records for TLS 1.0-1.2 can have the following maximum trailer
227  * lengths:
228  * - 16 (AES-GCM)
229  * - 36 (AES-CBC with SHA1 and up to 16 bytes of padding)
230  * - 48 (AES-CBC with SHA2-256 and up to 16 bytes of padding)
231  * - 64 (AES-CBC with SHA2-384 and up to 16 bytes of padding)
232  */
233 #define	MBUF_PEXT_TRAIL_LEN	64
234 
235 #if defined(__LP64__)
236 #define MBUF_PEXT_MAX_PGS (40 / sizeof(vm_paddr_t))
237 #else
238 #define MBUF_PEXT_MAX_PGS (72 / sizeof(vm_paddr_t))
239 #endif
240 
241 #define	MBUF_PEXT_MAX_BYTES						\
242     (MBUF_PEXT_MAX_PGS * PAGE_SIZE + MBUF_PEXT_HDR_LEN + MBUF_PEXT_TRAIL_LEN)
243 
244 struct ktls_session;
245 struct socket;
246 
247 /*
248  * Description of external storage mapped into mbuf; valid only if M_EXT is
249  * set.
250  * Size ILP32: 28
251  *	 LP64: 48
252  * Compile-time assertions in uipc_mbuf.c test these values to ensure that
253  * they are correct.
254  */
255 typedef	void m_ext_free_t(struct mbuf *);
256 struct m_ext {
257 	union {
258 		/*
259 		 * If EXT_FLAG_EMBREF is set, then we use refcount in the
260 		 * mbuf, the 'ext_count' member.  Otherwise, we have a
261 		 * shadow copy and we use pointer 'ext_cnt'.  The original
262 		 * mbuf is responsible to carry the pointer to free routine
263 		 * and its arguments.  They aren't copied into shadows in
264 		 * mb_dupcl() to avoid dereferencing next cachelines.
265 		 */
266 		volatile u_int	 ext_count;
267 		volatile u_int	*ext_cnt;
268 	};
269 	uint32_t	 ext_size;	/* size of buffer, for ext_free */
270 	uint32_t	 ext_type:8,	/* type of external storage */
271 			 ext_flags:24;	/* external storage mbuf flags */
272 	union {
273 		struct {
274 			/*
275 			 * Regular M_EXT mbuf:
276 			 * o ext_buf always points to the external buffer.
277 			 * o ext_free (below) and two optional arguments
278 			 *   ext_arg1 and ext_arg2 store the free context for
279 			 *   the external storage.  They are set only in the
280 			 *   refcount carrying mbuf, the one with
281 			 *   EXT_FLAG_EMBREF flag, with exclusion for
282 			 *   EXT_EXTREF type, where the free context is copied
283 			 *   into all mbufs that use same external storage.
284 			 */
285 			char 	*ext_buf;	/* start of buffer */
286 #define	m_ext_copylen	offsetof(struct m_ext, ext_arg2)
287 			void	*ext_arg2;
288 		};
289 		struct {
290 			/*
291 			 * Multi-page M_EXTPG mbuf:
292 			 * o extpg_pa - page vector.
293 			 * o extpg_trail and extpg_hdr - TLS trailer and
294 			 *   header.
295 			 * Uses ext_free and may also use ext_arg1.
296 			 */
297 			vm_paddr_t	extpg_pa[MBUF_PEXT_MAX_PGS];
298 			char		extpg_trail[MBUF_PEXT_TRAIL_LEN];
299 			char		extpg_hdr[MBUF_PEXT_HDR_LEN];
300 			/* Pretend these 3 fields are part of mbuf itself. */
301 #define	m_epg_pa	m_ext.extpg_pa
302 #define	m_epg_trail	m_ext.extpg_trail
303 #define	m_epg_hdr	m_ext.extpg_hdr
304 #define	m_epg_ext_copylen	offsetof(struct m_ext, ext_free)
305 		};
306 	};
307 	/*
308 	 * Free method and optional argument pointer, both
309 	 * used by M_EXT and M_EXTPG.
310 	 */
311 	m_ext_free_t	*ext_free;
312 	void		*ext_arg1;
313 };
314 
315 /*
316  * The core of the mbuf object along with some shortcut defines for practical
317  * purposes.
318  */
319 struct mbuf {
320 	/*
321 	 * Header present at the beginning of every mbuf.
322 	 * Size ILP32: 24
323 	 *      LP64: 32
324 	 * Compile-time assertions in uipc_mbuf.c test these values to ensure
325 	 * that they are correct.
326 	 */
327 	union {	/* next buffer in chain */
328 		struct mbuf		*m_next;
329 		SLIST_ENTRY(mbuf)	m_slist;
330 		STAILQ_ENTRY(mbuf)	m_stailq;
331 	};
332 	union {	/* next chain in queue/record */
333 		struct mbuf		*m_nextpkt;
334 		SLIST_ENTRY(mbuf)	m_slistpkt;
335 		STAILQ_ENTRY(mbuf)	m_stailqpkt;
336 	};
337 	caddr_t		 m_data;	/* location of data */
338 	int32_t		 m_len;		/* amount of data in this mbuf */
339 	uint32_t	 m_type:8,	/* type of data in this mbuf */
340 			 m_flags:24;	/* flags; see below */
341 #if !defined(__LP64__)
342 	uint32_t	 m_pad;		/* pad for 64bit alignment */
343 #endif
344 
345 	/*
346 	 * A set of optional headers (packet header, external storage header)
347 	 * and internal data storage.  Historically, these arrays were sized
348 	 * to MHLEN (space left after a packet header) and MLEN (space left
349 	 * after only a regular mbuf header); they are now variable size in
350 	 * order to support future work on variable-size mbufs.
351 	 */
352 	union {
353 		struct {
354 			union {
355 				/* M_PKTHDR set. */
356 				struct pkthdr	m_pkthdr;
357 
358 				/* M_EXTPG set.
359 				 * Multi-page M_EXTPG mbuf has its meta data
360 				 * split between the below anonymous structure
361 				 * and m_ext.  It carries vector of pages,
362 				 * optional header and trailer char vectors
363 				 * and pointers to socket/TLS data.
364 				 */
365 #define	m_epg_startcopy		m_epg_npgs
366 #define	m_epg_endcopy		m_epg_stailq
367 				struct {
368 					/* Overall count of pages and count of
369 					 * pages with I/O pending. */
370 					uint8_t	m_epg_npgs;
371 					uint8_t	m_epg_nrdy;
372 					/* TLS header and trailer lengths.
373 					 * The data itself resides in m_ext. */
374 					uint8_t	m_epg_hdrlen;
375 					uint8_t	m_epg_trllen;
376 					/* Offset into 1st page and length of
377 					 * data in the last page. */
378 					uint16_t m_epg_1st_off;
379 					uint16_t m_epg_last_len;
380 					uint8_t	m_epg_flags;
381 #define	EPG_FLAG_ANON	0x1	/* Data can be encrypted in place. */
382 #define	EPG_FLAG_2FREE	0x2	/* Scheduled for free. */
383 					uint8_t	m_epg_record_type;
384 					uint8_t	__spare[2];
385 					int	m_epg_enc_cnt;
386 					struct ktls_session *m_epg_tls;
387 					struct socket	*m_epg_so;
388 					uint64_t	m_epg_seqno;
389 					STAILQ_ENTRY(mbuf) m_epg_stailq;
390 				};
391 			};
392 			union {
393 				/* M_EXT or M_EXTPG set. */
394 				struct m_ext	m_ext;
395 				/* M_PKTHDR set, neither M_EXT nor M_EXTPG. */
396 				char		m_pktdat[0];
397 			};
398 		};
399 		char	m_dat[0];			/* !M_PKTHDR, !M_EXT */
400 	};
401 };
402 
403 #ifdef _KERNEL
404 static inline int
405 m_epg_pagelen(const struct mbuf *m, int pidx, int pgoff)
406 {
407 
408 	KASSERT(pgoff == 0 || pidx == 0,
409 	    ("page %d with non-zero offset %d in %p", pidx, pgoff, m));
410 
411 	if (pidx == m->m_epg_npgs - 1) {
412 		return (m->m_epg_last_len);
413 	} else {
414 		return (PAGE_SIZE - pgoff);
415 	}
416 }
417 
418 #ifdef INVARIANTS
419 #define	MCHECK(ex, msg)	KASSERT((ex),				\
420 	    ("Multi page mbuf %p with " #msg " at %s:%d",	\
421 	    m, __FILE__, __LINE__))
422 /*
423  * NB: This expects a non-empty buffer (npgs > 0 and
424  * last_pg_len > 0).
425  */
426 #define	MBUF_EXT_PGS_ASSERT_SANITY(m)	do {				\
427 	MCHECK(m->m_epg_npgs > 0, "no valid pages");		\
428 	MCHECK(m->m_epg_npgs <= nitems(m->m_epg_pa),		\
429 	    "too many pages");						\
430 	MCHECK(m->m_epg_nrdy <= m->m_epg_npgs,			\
431 	    "too many ready pages");					\
432 	MCHECK(m->m_epg_1st_off < PAGE_SIZE,			\
433 		"too large page offset");				\
434 	MCHECK(m->m_epg_last_len > 0, "zero last page length");	\
435 	MCHECK(m->m_epg_last_len <= PAGE_SIZE,			\
436 	    "too large last page length");				\
437 	if (m->m_epg_npgs == 1)					\
438 		MCHECK(m->m_epg_1st_off +			\
439 		    m->m_epg_last_len <=	 PAGE_SIZE,		\
440 		    "single page too large");				\
441 	MCHECK(m->m_epg_hdrlen <= sizeof(m->m_epg_hdr),		\
442 	    "too large header length");					\
443 	MCHECK(m->m_epg_trllen <= sizeof(m->m_epg_trail),	\
444 	    "too large header length");					\
445 } while (0)
446 #else
447 #define	MBUF_EXT_PGS_ASSERT_SANITY(m)	do {} while (0);
448 #endif
449 #endif
450 
451 /*
452  * mbuf flags of global significance and layer crossing.
453  * Those of only protocol/layer specific significance are to be mapped
454  * to M_PROTO[1-11] and cleared at layer handoff boundaries.
455  * NB: Limited to the lower 24 bits.
456  */
457 #define	M_EXT		0x00000001 /* has associated external storage */
458 #define	M_PKTHDR	0x00000002 /* start of record */
459 #define	M_EOR		0x00000004 /* end of record */
460 #define	M_RDONLY	0x00000008 /* associated data is marked read-only */
461 #define	M_BCAST		0x00000010 /* send/received as link-level broadcast */
462 #define	M_MCAST		0x00000020 /* send/received as link-level multicast */
463 #define	M_PROMISC	0x00000040 /* packet was not for us */
464 #define	M_VLANTAG	0x00000080 /* ether_vtag is valid */
465 #define	M_EXTPG		0x00000100 /* has array of unmapped pages and TLS */
466 #define	M_NOFREE	0x00000200 /* do not free mbuf, embedded in cluster */
467 #define	M_TSTMP		0x00000400 /* rcv_tstmp field is valid */
468 #define	M_TSTMP_HPREC	0x00000800 /* rcv_tstmp is high-prec, typically
469 				      hw-stamped on port (useful for IEEE 1588
470 				      and 802.1AS) */
471 #define M_TSTMP_LRO	0x00001000 /* Time LRO pushed in pkt is valid in (PH_loc) */
472 
473 #define	M_PROTO1	0x00002000 /* protocol-specific */
474 #define	M_PROTO2	0x00004000 /* protocol-specific */
475 #define	M_PROTO3	0x00008000 /* protocol-specific */
476 #define	M_PROTO4	0x00010000 /* protocol-specific */
477 #define	M_PROTO5	0x00020000 /* protocol-specific */
478 #define	M_PROTO6	0x00040000 /* protocol-specific */
479 #define	M_PROTO7	0x00080000 /* protocol-specific */
480 #define	M_PROTO8	0x00100000 /* protocol-specific */
481 #define	M_PROTO9	0x00200000 /* protocol-specific */
482 #define	M_PROTO10	0x00400000 /* protocol-specific */
483 #define	M_PROTO11	0x00800000 /* protocol-specific */
484 
485 /*
486  * Flags to purge when crossing layers.
487  */
488 #define	M_PROTOFLAGS \
489     (M_PROTO1|M_PROTO2|M_PROTO3|M_PROTO4|M_PROTO5|M_PROTO6|M_PROTO7|M_PROTO8|\
490      M_PROTO9|M_PROTO10|M_PROTO11)
491 
492 /*
493  * Flags preserved when copying m_pkthdr.
494  */
495 #define M_COPYFLAGS \
496     (M_PKTHDR|M_EOR|M_RDONLY|M_BCAST|M_MCAST|M_PROMISC|M_VLANTAG|M_TSTMP| \
497      M_TSTMP_HPREC|M_TSTMP_LRO|M_PROTOFLAGS)
498 
499 /*
500  * Flags preserved during demote.
501  */
502 #define	M_DEMOTEFLAGS \
503     (M_EXT | M_RDONLY | M_NOFREE | M_EXTPG)
504 
505 /*
506  * Mbuf flag description for use with printf(9) %b identifier.
507  */
508 #define	M_FLAG_BITS \
509     "\20\1M_EXT\2M_PKTHDR\3M_EOR\4M_RDONLY\5M_BCAST\6M_MCAST" \
510     "\7M_PROMISC\10M_VLANTAG\11M_EXTPG\12M_NOFREE\13M_TSTMP\14M_TSTMP_HPREC\15M_TSTMP_LRO"
511 #define	M_FLAG_PROTOBITS \
512     "\16M_PROTO1\17M_PROTO2\20M_PROTO3\21M_PROTO4" \
513     "\22M_PROTO5\23M_PROTO6\24M_PROTO7\25M_PROTO8\26M_PROTO9" \
514     "\27M_PROTO10\28M_PROTO11"
515 #define	M_FLAG_PRINTF (M_FLAG_BITS M_FLAG_PROTOBITS)
516 
517 /*
518  * Network interface cards are able to hash protocol fields (such as IPv4
519  * addresses and TCP port numbers) classify packets into flows.  These flows
520  * can then be used to maintain ordering while delivering packets to the OS
521  * via parallel input queues, as well as to provide a stateless affinity
522  * model.  NIC drivers can pass up the hash via m->m_pkthdr.flowid, and set
523  * m_flag fields to indicate how the hash should be interpreted by the
524  * network stack.
525  *
526  * Most NICs support RSS, which provides ordering and explicit affinity, and
527  * use the hash m_flag bits to indicate what header fields were covered by
528  * the hash.  M_HASHTYPE_OPAQUE and M_HASHTYPE_OPAQUE_HASH can be set by non-
529  * RSS cards or configurations that provide an opaque flow identifier, allowing
530  * for ordering and distribution without explicit affinity.  Additionally,
531  * M_HASHTYPE_OPAQUE_HASH indicates that the flow identifier has hash
532  * properties.
533  *
534  * The meaning of the IPV6_EX suffix:
535  * "o  Home address from the home address option in the IPv6 destination
536  *     options header.  If the extension header is not present, use the Source
537  *     IPv6 Address.
538  *  o  IPv6 address that is contained in the Routing-Header-Type-2 from the
539  *     associated extension header.  If the extension header is not present,
540  *     use the Destination IPv6 Address."
541  * Quoted from:
542  * https://docs.microsoft.com/en-us/windows-hardware/drivers/network/rss-hashing-types#ndishashipv6ex
543  */
544 #define	M_HASHTYPE_HASHPROP		0x80	/* has hash properties */
545 #define	M_HASHTYPE_INNER		0x40	/* calculated from inner headers */
546 #define	M_HASHTYPE_HASH(t)		(M_HASHTYPE_HASHPROP | (t))
547 /* Microsoft RSS standard hash types */
548 #define	M_HASHTYPE_NONE			0
549 #define	M_HASHTYPE_RSS_IPV4		M_HASHTYPE_HASH(1) /* IPv4 2-tuple */
550 #define	M_HASHTYPE_RSS_TCP_IPV4		M_HASHTYPE_HASH(2) /* TCPv4 4-tuple */
551 #define	M_HASHTYPE_RSS_IPV6		M_HASHTYPE_HASH(3) /* IPv6 2-tuple */
552 #define	M_HASHTYPE_RSS_TCP_IPV6		M_HASHTYPE_HASH(4) /* TCPv6 4-tuple */
553 #define	M_HASHTYPE_RSS_IPV6_EX		M_HASHTYPE_HASH(5) /* IPv6 2-tuple +
554 							    * ext hdrs */
555 #define	M_HASHTYPE_RSS_TCP_IPV6_EX	M_HASHTYPE_HASH(6) /* TCPv6 4-tuple +
556 							    * ext hdrs */
557 #define	M_HASHTYPE_RSS_UDP_IPV4		M_HASHTYPE_HASH(7) /* IPv4 UDP 4-tuple*/
558 #define	M_HASHTYPE_RSS_UDP_IPV6		M_HASHTYPE_HASH(9) /* IPv6 UDP 4-tuple*/
559 #define	M_HASHTYPE_RSS_UDP_IPV6_EX	M_HASHTYPE_HASH(10)/* IPv6 UDP 4-tuple +
560 							    * ext hdrs */
561 
562 #define	M_HASHTYPE_OPAQUE		0x3f	/* ordering, not affinity */
563 #define	M_HASHTYPE_OPAQUE_HASH		M_HASHTYPE_HASH(M_HASHTYPE_OPAQUE)
564 						/* ordering+hash, not affinity*/
565 
566 #define	M_HASHTYPE_CLEAR(m)	((m)->m_pkthdr.rsstype = 0)
567 #define	M_HASHTYPE_GET(m)	((m)->m_pkthdr.rsstype & ~M_HASHTYPE_INNER)
568 #define	M_HASHTYPE_SET(m, v)	((m)->m_pkthdr.rsstype = (v))
569 #define	M_HASHTYPE_TEST(m, v)	(M_HASHTYPE_GET(m) == (v))
570 #define	M_HASHTYPE_ISHASH(m)	\
571     (((m)->m_pkthdr.rsstype & M_HASHTYPE_HASHPROP) != 0)
572 #define	M_HASHTYPE_SETINNER(m)	do {			\
573 	(m)->m_pkthdr.rsstype |= M_HASHTYPE_INNER;	\
574     } while (0)
575 
576 /*
577  * External mbuf storage buffer types.
578  */
579 #define	EXT_CLUSTER	1	/* mbuf cluster */
580 #define	EXT_SFBUF	2	/* sendfile(2)'s sf_buf */
581 #define	EXT_JUMBOP	3	/* jumbo cluster page sized */
582 #define	EXT_JUMBO9	4	/* jumbo cluster 9216 bytes */
583 #define	EXT_JUMBO16	5	/* jumbo cluster 16184 bytes */
584 #define	EXT_PACKET	6	/* mbuf+cluster from packet zone */
585 #define	EXT_MBUF	7	/* external mbuf reference */
586 #define	EXT_RXRING	8	/* data in NIC receive ring */
587 
588 #define	EXT_VENDOR1	224	/* for vendor-internal use */
589 #define	EXT_VENDOR2	225	/* for vendor-internal use */
590 #define	EXT_VENDOR3	226	/* for vendor-internal use */
591 #define	EXT_VENDOR4	227	/* for vendor-internal use */
592 
593 #define	EXT_EXP1	244	/* for experimental use */
594 #define	EXT_EXP2	245	/* for experimental use */
595 #define	EXT_EXP3	246	/* for experimental use */
596 #define	EXT_EXP4	247	/* for experimental use */
597 
598 #define	EXT_NET_DRV	252	/* custom ext_buf provided by net driver(s) */
599 #define	EXT_MOD_TYPE	253	/* custom module's ext_buf type */
600 #define	EXT_DISPOSABLE	254	/* can throw this buffer away w/page flipping */
601 #define	EXT_EXTREF	255	/* has externally maintained ext_cnt ptr */
602 
603 /*
604  * Flags for external mbuf buffer types.
605  * NB: limited to the lower 24 bits.
606  */
607 #define	EXT_FLAG_EMBREF		0x000001	/* embedded ext_count */
608 #define	EXT_FLAG_EXTREF		0x000002	/* external ext_cnt, notyet */
609 
610 #define	EXT_FLAG_NOFREE		0x000010	/* don't free mbuf to pool, notyet */
611 
612 #define	EXT_FLAG_VENDOR1	0x010000	/* These flags are vendor */
613 #define	EXT_FLAG_VENDOR2	0x020000	/* or submodule specific, */
614 #define	EXT_FLAG_VENDOR3	0x040000	/* not used by mbuf code. */
615 #define	EXT_FLAG_VENDOR4	0x080000	/* Set/read by submodule. */
616 
617 #define	EXT_FLAG_EXP1		0x100000	/* for experimental use */
618 #define	EXT_FLAG_EXP2		0x200000	/* for experimental use */
619 #define	EXT_FLAG_EXP3		0x400000	/* for experimental use */
620 #define	EXT_FLAG_EXP4		0x800000	/* for experimental use */
621 
622 /*
623  * EXT flag description for use with printf(9) %b identifier.
624  */
625 #define	EXT_FLAG_BITS \
626     "\20\1EXT_FLAG_EMBREF\2EXT_FLAG_EXTREF\5EXT_FLAG_NOFREE" \
627     "\21EXT_FLAG_VENDOR1\22EXT_FLAG_VENDOR2\23EXT_FLAG_VENDOR3" \
628     "\24EXT_FLAG_VENDOR4\25EXT_FLAG_EXP1\26EXT_FLAG_EXP2\27EXT_FLAG_EXP3" \
629     "\30EXT_FLAG_EXP4"
630 
631 /*
632  * Flags indicating checksum, segmentation and other offload work to be
633  * done, or already done, by hardware or lower layers.  It is split into
634  * separate inbound and outbound flags.
635  *
636  * Outbound flags that are set by upper protocol layers requesting lower
637  * layers, or ideally the hardware, to perform these offloading tasks.
638  * For outbound packets this field and its flags can be directly tested
639  * against ifnet if_hwassist.  Note that the outbound and the inbound flags do
640  * not collide right now but they could be allowed to (as long as the flags are
641  * scrubbed appropriately when the direction of an mbuf changes).  CSUM_BITS
642  * would also have to split into CSUM_BITS_TX and CSUM_BITS_RX.
643  *
644  * CSUM_INNER_<x> is the same as CSUM_<x> but it applies to the inner frame.
645  * The CSUM_ENCAP_<x> bits identify the outer encapsulation.
646  */
647 #define	CSUM_IP			0x00000001	/* IP header checksum offload */
648 #define	CSUM_IP_UDP		0x00000002	/* UDP checksum offload */
649 #define	CSUM_IP_TCP		0x00000004	/* TCP checksum offload */
650 #define	CSUM_IP_SCTP		0x00000008	/* SCTP checksum offload */
651 #define	CSUM_IP_TSO		0x00000010	/* TCP segmentation offload */
652 #define	CSUM_IP_ISCSI		0x00000020	/* iSCSI checksum offload */
653 
654 #define	CSUM_INNER_IP6_UDP	0x00000040
655 #define	CSUM_INNER_IP6_TCP	0x00000080
656 #define	CSUM_INNER_IP6_TSO	0x00000100
657 #define	CSUM_IP6_UDP		0x00000200	/* UDP checksum offload */
658 #define	CSUM_IP6_TCP		0x00000400	/* TCP checksum offload */
659 #define	CSUM_IP6_SCTP		0x00000800	/* SCTP checksum offload */
660 #define	CSUM_IP6_TSO		0x00001000	/* TCP segmentation offload */
661 #define	CSUM_IP6_ISCSI		0x00002000	/* iSCSI checksum offload */
662 
663 #define	CSUM_INNER_IP		0x00004000
664 #define	CSUM_INNER_IP_UDP	0x00008000
665 #define	CSUM_INNER_IP_TCP	0x00010000
666 #define	CSUM_INNER_IP_TSO	0x00020000
667 
668 #define	CSUM_ENCAP_VXLAN	0x00040000	/* VXLAN outer encapsulation */
669 #define	CSUM_ENCAP_RSVD1	0x00080000
670 
671 /* Inbound checksum support where the checksum was verified by hardware. */
672 #define	CSUM_INNER_L3_CALC	0x00100000
673 #define	CSUM_INNER_L3_VALID	0x00200000
674 #define	CSUM_INNER_L4_CALC	0x00400000
675 #define	CSUM_INNER_L4_VALID	0x00800000
676 #define	CSUM_L3_CALC		0x01000000	/* calculated layer 3 csum */
677 #define	CSUM_L3_VALID		0x02000000	/* checksum is correct */
678 #define	CSUM_L4_CALC		0x04000000	/* calculated layer 4 csum */
679 #define	CSUM_L4_VALID		0x08000000	/* checksum is correct */
680 #define	CSUM_L5_CALC		0x10000000	/* calculated layer 5 csum */
681 #define	CSUM_L5_VALID		0x20000000	/* checksum is correct */
682 #define	CSUM_COALESCED		0x40000000	/* contains merged segments */
683 
684 #define	CSUM_SND_TAG		0x80000000	/* Packet header has send tag */
685 
686 #define CSUM_FLAGS_TX (CSUM_IP | CSUM_IP_UDP | CSUM_IP_TCP | CSUM_IP_SCTP | \
687     CSUM_IP_TSO | CSUM_IP_ISCSI | CSUM_INNER_IP6_UDP | CSUM_INNER_IP6_TCP | \
688     CSUM_INNER_IP6_TSO | CSUM_IP6_UDP | CSUM_IP6_TCP | CSUM_IP6_SCTP | \
689     CSUM_IP6_TSO | CSUM_IP6_ISCSI | CSUM_INNER_IP | CSUM_INNER_IP_UDP | \
690     CSUM_INNER_IP_TCP | CSUM_INNER_IP_TSO | CSUM_ENCAP_VXLAN | \
691     CSUM_ENCAP_RSVD1 | CSUM_SND_TAG)
692 
693 #define CSUM_FLAGS_RX (CSUM_INNER_L3_CALC | CSUM_INNER_L3_VALID | \
694     CSUM_INNER_L4_CALC | CSUM_INNER_L4_VALID | CSUM_L3_CALC | CSUM_L3_VALID | \
695     CSUM_L4_CALC | CSUM_L4_VALID | CSUM_L5_CALC | CSUM_L5_VALID | \
696     CSUM_COALESCED)
697 
698 /*
699  * CSUM flag description for use with printf(9) %b identifier.
700  */
701 #define	CSUM_BITS \
702     "\20\1CSUM_IP\2CSUM_IP_UDP\3CSUM_IP_TCP\4CSUM_IP_SCTP\5CSUM_IP_TSO" \
703     "\6CSUM_IP_ISCSI\7CSUM_INNER_IP6_UDP\10CSUM_INNER_IP6_TCP" \
704     "\11CSUM_INNER_IP6_TSO\12CSUM_IP6_UDP\13CSUM_IP6_TCP\14CSUM_IP6_SCTP" \
705     "\15CSUM_IP6_TSO\16CSUM_IP6_ISCSI\17CSUM_INNER_IP\20CSUM_INNER_IP_UDP" \
706     "\21CSUM_INNER_IP_TCP\22CSUM_INNER_IP_TSO\23CSUM_ENCAP_VXLAN" \
707     "\24CSUM_ENCAP_RSVD1\25CSUM_INNER_L3_CALC\26CSUM_INNER_L3_VALID" \
708     "\27CSUM_INNER_L4_CALC\30CSUM_INNER_L4_VALID\31CSUM_L3_CALC" \
709     "\32CSUM_L3_VALID\33CSUM_L4_CALC\34CSUM_L4_VALID\35CSUM_L5_CALC" \
710     "\36CSUM_L5_VALID\37CSUM_COALESCED\40CSUM_SND_TAG"
711 
712 /* CSUM flags compatibility mappings. */
713 #define	CSUM_IP_CHECKED		CSUM_L3_CALC
714 #define	CSUM_IP_VALID		CSUM_L3_VALID
715 #define	CSUM_DATA_VALID		CSUM_L4_VALID
716 #define	CSUM_PSEUDO_HDR		CSUM_L4_CALC
717 #define	CSUM_SCTP_VALID		CSUM_L4_VALID
718 #define	CSUM_DELAY_DATA		(CSUM_TCP|CSUM_UDP)
719 #define	CSUM_DELAY_IP		CSUM_IP		/* Only v4, no v6 IP hdr csum */
720 #define	CSUM_DELAY_DATA_IPV6	(CSUM_TCP_IPV6|CSUM_UDP_IPV6)
721 #define	CSUM_DATA_VALID_IPV6	CSUM_DATA_VALID
722 #define	CSUM_TCP		CSUM_IP_TCP
723 #define	CSUM_UDP		CSUM_IP_UDP
724 #define	CSUM_SCTP		CSUM_IP_SCTP
725 #define	CSUM_TSO		(CSUM_IP_TSO|CSUM_IP6_TSO)
726 #define	CSUM_INNER_TSO		(CSUM_INNER_IP_TSO|CSUM_INNER_IP6_TSO)
727 #define	CSUM_UDP_IPV6		CSUM_IP6_UDP
728 #define	CSUM_TCP_IPV6		CSUM_IP6_TCP
729 #define	CSUM_SCTP_IPV6		CSUM_IP6_SCTP
730 #define	CSUM_TLS_MASK		(CSUM_L5_CALC|CSUM_L5_VALID)
731 #define	CSUM_TLS_DECRYPTED	CSUM_L5_CALC
732 
733 /*
734  * mbuf types describing the content of the mbuf (including external storage).
735  */
736 #define	MT_NOTMBUF	0	/* USED INTERNALLY ONLY! Object is not mbuf */
737 #define	MT_DATA		1	/* dynamic (data) allocation */
738 #define	MT_HEADER	MT_DATA	/* packet header, use M_PKTHDR instead */
739 
740 #define	MT_VENDOR1	4	/* for vendor-internal use */
741 #define	MT_VENDOR2	5	/* for vendor-internal use */
742 #define	MT_VENDOR3	6	/* for vendor-internal use */
743 #define	MT_VENDOR4	7	/* for vendor-internal use */
744 
745 #define	MT_SONAME	8	/* socket name */
746 
747 #define	MT_EXP1		9	/* for experimental use */
748 #define	MT_EXP2		10	/* for experimental use */
749 #define	MT_EXP3		11	/* for experimental use */
750 #define	MT_EXP4		12	/* for experimental use */
751 
752 #define	MT_CONTROL	14	/* extra-data protocol message */
753 #define	MT_EXTCONTROL	15	/* control message with externalized contents */
754 #define	MT_OOBDATA	16	/* expedited data  */
755 
756 #define	MT_NOINIT	255	/* Not a type but a flag to allocate
757 				   a non-initialized mbuf */
758 
759 /*
760  * String names of mbuf-related UMA(9) and malloc(9) types.  Exposed to
761  * !_KERNEL so that monitoring tools can look up the zones with
762  * libmemstat(3).
763  */
764 #define	MBUF_MEM_NAME		"mbuf"
765 #define	MBUF_CLUSTER_MEM_NAME	"mbuf_cluster"
766 #define	MBUF_PACKET_MEM_NAME	"mbuf_packet"
767 #define	MBUF_JUMBOP_MEM_NAME	"mbuf_jumbo_page"
768 #define	MBUF_JUMBO9_MEM_NAME	"mbuf_jumbo_9k"
769 #define	MBUF_JUMBO16_MEM_NAME	"mbuf_jumbo_16k"
770 #define	MBUF_TAG_MEM_NAME	"mbuf_tag"
771 #define	MBUF_EXTREFCNT_MEM_NAME	"mbuf_ext_refcnt"
772 #define	MBUF_EXTPGS_MEM_NAME	"mbuf_extpgs"
773 
774 #ifdef _KERNEL
775 union if_snd_tag_alloc_params;
776 
777 #ifdef WITNESS
778 #define	MBUF_CHECKSLEEP(how) do {					\
779 	if (how == M_WAITOK)						\
780 		WITNESS_WARN(WARN_GIANTOK | WARN_SLEEPOK, NULL,		\
781 		    "Sleeping in \"%s\"", __func__);			\
782 } while (0)
783 #else
784 #define	MBUF_CHECKSLEEP(how)
785 #endif
786 
787 /*
788  * Network buffer allocation API
789  *
790  * The rest of it is defined in kern/kern_mbuf.c
791  */
792 extern uma_zone_t	zone_mbuf;
793 extern uma_zone_t	zone_clust;
794 extern uma_zone_t	zone_pack;
795 extern uma_zone_t	zone_jumbop;
796 extern uma_zone_t	zone_jumbo9;
797 extern uma_zone_t	zone_jumbo16;
798 extern uma_zone_t	zone_extpgs;
799 
800 void		 mb_dupcl(struct mbuf *, struct mbuf *);
801 void		 mb_free_ext(struct mbuf *);
802 void		 mb_free_extpg(struct mbuf *);
803 void		 mb_free_mext_pgs(struct mbuf *);
804 struct mbuf	*mb_alloc_ext_pgs(int, m_ext_free_t);
805 struct mbuf	*mb_alloc_ext_plus_pages(int, int);
806 struct mbuf	*mb_mapped_to_unmapped(struct mbuf *, int, int, int,
807 		    struct mbuf **);
808 int		 mb_unmapped_compress(struct mbuf *m);
809 struct mbuf 	*mb_unmapped_to_ext(struct mbuf *m);
810 void		 mb_free_notready(struct mbuf *m, int count);
811 void		 m_adj(struct mbuf *, int);
812 void		 m_adj_decap(struct mbuf *, int);
813 int		 m_apply(struct mbuf *, int, int,
814 		    int (*)(void *, void *, u_int), void *);
815 int		 m_append(struct mbuf *, int, c_caddr_t);
816 void		 m_cat(struct mbuf *, struct mbuf *);
817 void		 m_catpkt(struct mbuf *, struct mbuf *);
818 int		 m_clget(struct mbuf *m, int how);
819 void 		*m_cljget(struct mbuf *m, int how, int size);
820 struct mbuf	*m_collapse(struct mbuf *, int, int);
821 void		 m_copyback(struct mbuf *, int, int, c_caddr_t);
822 void		 m_copydata(const struct mbuf *, int, int, caddr_t);
823 struct mbuf	*m_copym(struct mbuf *, int, int, int);
824 struct mbuf	*m_copypacket(struct mbuf *, int);
825 void		 m_copy_pkthdr(struct mbuf *, struct mbuf *);
826 struct mbuf	*m_copyup(struct mbuf *, int, int);
827 struct mbuf	*m_defrag(struct mbuf *, int);
828 void		 m_demote_pkthdr(struct mbuf *);
829 void		 m_demote(struct mbuf *, int, int);
830 struct mbuf	*m_devget(char *, int, int, struct ifnet *,
831 		    void (*)(char *, caddr_t, u_int));
832 void		 m_dispose_extcontrolm(struct mbuf *m);
833 struct mbuf	*m_dup(const struct mbuf *, int);
834 int		 m_dup_pkthdr(struct mbuf *, const struct mbuf *, int);
835 void		 m_extadd(struct mbuf *, char *, u_int, m_ext_free_t,
836 		    void *, void *, int, int);
837 u_int		 m_fixhdr(struct mbuf *);
838 struct mbuf	*m_fragment(struct mbuf *, int, int);
839 void		 m_freem(struct mbuf *);
840 void		 m_free_raw(struct mbuf *);
841 struct mbuf	*m_get2(int, int, short, int);
842 struct mbuf	*m_get3(int, int, short, int);
843 struct mbuf	*m_getjcl(int, short, int, int);
844 struct mbuf	*m_getm2(struct mbuf *, int, int, short, int);
845 struct mbuf	*m_getptr(struct mbuf *, int, int *);
846 u_int		 m_length(struct mbuf *, struct mbuf **);
847 int		 m_mbuftouio(struct uio *, const struct mbuf *, int);
848 void		 m_move_pkthdr(struct mbuf *, struct mbuf *);
849 int		 m_pkthdr_init(struct mbuf *, int);
850 struct mbuf	*m_prepend(struct mbuf *, int, int);
851 void		 m_print(const struct mbuf *, int);
852 struct mbuf	*m_pulldown(struct mbuf *, int, int, int *);
853 struct mbuf	*m_pullup(struct mbuf *, int);
854 int		 m_sanity(struct mbuf *, int);
855 struct mbuf	*m_split(struct mbuf *, int, int);
856 struct mbuf	*m_uiotombuf(struct uio *, int, int, int, int);
857 int		 m_unmapped_uiomove(const struct mbuf *, int, struct uio *,
858 		    int);
859 struct mbuf	*m_unshare(struct mbuf *, int);
860 int		 m_snd_tag_alloc(struct ifnet *,
861 		    union if_snd_tag_alloc_params *, struct m_snd_tag **);
862 void		 m_snd_tag_init(struct m_snd_tag *, struct ifnet *,
863 		    const struct if_snd_tag_sw *);
864 void		 m_snd_tag_destroy(struct m_snd_tag *);
865 
866 static __inline int
867 m_gettype(int size)
868 {
869 	int type;
870 
871 	switch (size) {
872 	case MSIZE:
873 		type = EXT_MBUF;
874 		break;
875 	case MCLBYTES:
876 		type = EXT_CLUSTER;
877 		break;
878 #if MJUMPAGESIZE != MCLBYTES
879 	case MJUMPAGESIZE:
880 		type = EXT_JUMBOP;
881 		break;
882 #endif
883 	case MJUM9BYTES:
884 		type = EXT_JUMBO9;
885 		break;
886 	case MJUM16BYTES:
887 		type = EXT_JUMBO16;
888 		break;
889 	default:
890 		panic("%s: invalid cluster size %d", __func__, size);
891 	}
892 
893 	return (type);
894 }
895 
896 /*
897  * Associated an external reference counted buffer with an mbuf.
898  */
899 static __inline void
900 m_extaddref(struct mbuf *m, char *buf, u_int size, u_int *ref_cnt,
901     m_ext_free_t freef, void *arg1, void *arg2)
902 {
903 
904 	KASSERT(ref_cnt != NULL, ("%s: ref_cnt not provided", __func__));
905 
906 	atomic_add_int(ref_cnt, 1);
907 	m->m_flags |= M_EXT;
908 	m->m_ext.ext_buf = buf;
909 	m->m_ext.ext_cnt = ref_cnt;
910 	m->m_data = m->m_ext.ext_buf;
911 	m->m_ext.ext_size = size;
912 	m->m_ext.ext_free = freef;
913 	m->m_ext.ext_arg1 = arg1;
914 	m->m_ext.ext_arg2 = arg2;
915 	m->m_ext.ext_type = EXT_EXTREF;
916 	m->m_ext.ext_flags = 0;
917 }
918 
919 static __inline uma_zone_t
920 m_getzone(int size)
921 {
922 	uma_zone_t zone;
923 
924 	switch (size) {
925 	case MCLBYTES:
926 		zone = zone_clust;
927 		break;
928 #if MJUMPAGESIZE != MCLBYTES
929 	case MJUMPAGESIZE:
930 		zone = zone_jumbop;
931 		break;
932 #endif
933 	case MJUM9BYTES:
934 		zone = zone_jumbo9;
935 		break;
936 	case MJUM16BYTES:
937 		zone = zone_jumbo16;
938 		break;
939 	default:
940 		panic("%s: invalid cluster size %d", __func__, size);
941 	}
942 
943 	return (zone);
944 }
945 
946 /*
947  * Initialize an mbuf with linear storage.
948  *
949  * Inline because the consumer text overhead will be roughly the same to
950  * initialize or call a function with this many parameters and M_PKTHDR
951  * should go away with constant propagation for !MGETHDR.
952  */
953 static __inline int
954 m_init(struct mbuf *m, int how, short type, int flags)
955 {
956 	int error;
957 
958 	m->m_next = NULL;
959 	m->m_nextpkt = NULL;
960 	m->m_data = m->m_dat;
961 	m->m_len = 0;
962 	m->m_flags = flags;
963 	m->m_type = type;
964 	if (flags & M_PKTHDR)
965 		error = m_pkthdr_init(m, how);
966 	else
967 		error = 0;
968 
969 	MBUF_PROBE5(m__init, m, how, type, flags, error);
970 	return (error);
971 }
972 
973 static __inline struct mbuf *
974 m_get_raw(int how, short type)
975 {
976 	struct mbuf *m;
977 	struct mb_args args;
978 
979 	args.flags = 0;
980 	args.type = type | MT_NOINIT;
981 	m = uma_zalloc_arg(zone_mbuf, &args, how);
982 	MBUF_PROBE3(m__get_raw, how, type, m);
983 	return (m);
984 }
985 
986 static __inline struct mbuf *
987 m_get(int how, short type)
988 {
989 	struct mbuf *m;
990 	struct mb_args args;
991 
992 	args.flags = 0;
993 	args.type = type;
994 	m = uma_zalloc_arg(zone_mbuf, &args, how);
995 	MBUF_PROBE3(m__get, how, type, m);
996 	return (m);
997 }
998 
999 static __inline struct mbuf *
1000 m_gethdr_raw(int how, short type)
1001 {
1002 	struct mbuf *m;
1003 	struct mb_args args;
1004 
1005 	args.flags = M_PKTHDR;
1006 	args.type = type | MT_NOINIT;
1007 	m = uma_zalloc_arg(zone_mbuf, &args, how);
1008 	MBUF_PROBE3(m__gethdr_raw, how, type, m);
1009 	return (m);
1010 }
1011 
1012 static __inline struct mbuf *
1013 m_gethdr(int how, short type)
1014 {
1015 	struct mbuf *m;
1016 	struct mb_args args;
1017 
1018 	args.flags = M_PKTHDR;
1019 	args.type = type;
1020 	m = uma_zalloc_arg(zone_mbuf, &args, how);
1021 	MBUF_PROBE3(m__gethdr, how, type, m);
1022 	return (m);
1023 }
1024 
1025 static __inline struct mbuf *
1026 m_getcl(int how, short type, int flags)
1027 {
1028 	struct mbuf *m;
1029 	struct mb_args args;
1030 
1031 	args.flags = flags;
1032 	args.type = type;
1033 	m = uma_zalloc_arg(zone_pack, &args, how);
1034 	MBUF_PROBE4(m__getcl, how, type, flags, m);
1035 	return (m);
1036 }
1037 
1038 /*
1039  * XXX: m_cljset() is a dangerous API.  One must attach only a new,
1040  * unreferenced cluster to an mbuf(9).  It is not possible to assert
1041  * that, so care can be taken only by users of the API.
1042  */
1043 static __inline void
1044 m_cljset(struct mbuf *m, void *cl, int type)
1045 {
1046 	int size;
1047 
1048 	switch (type) {
1049 	case EXT_CLUSTER:
1050 		size = MCLBYTES;
1051 		break;
1052 #if MJUMPAGESIZE != MCLBYTES
1053 	case EXT_JUMBOP:
1054 		size = MJUMPAGESIZE;
1055 		break;
1056 #endif
1057 	case EXT_JUMBO9:
1058 		size = MJUM9BYTES;
1059 		break;
1060 	case EXT_JUMBO16:
1061 		size = MJUM16BYTES;
1062 		break;
1063 	default:
1064 		panic("%s: unknown cluster type %d", __func__, type);
1065 		break;
1066 	}
1067 
1068 	m->m_data = m->m_ext.ext_buf = cl;
1069 	m->m_ext.ext_free = m->m_ext.ext_arg1 = m->m_ext.ext_arg2 = NULL;
1070 	m->m_ext.ext_size = size;
1071 	m->m_ext.ext_type = type;
1072 	m->m_ext.ext_flags = EXT_FLAG_EMBREF;
1073 	m->m_ext.ext_count = 1;
1074 	m->m_flags |= M_EXT;
1075 	MBUF_PROBE3(m__cljset, m, cl, type);
1076 }
1077 
1078 static __inline void
1079 m_chtype(struct mbuf *m, short new_type)
1080 {
1081 
1082 	m->m_type = new_type;
1083 }
1084 
1085 static __inline void
1086 m_clrprotoflags(struct mbuf *m)
1087 {
1088 
1089 	while (m) {
1090 		m->m_flags &= ~M_PROTOFLAGS;
1091 		m = m->m_next;
1092 	}
1093 }
1094 
1095 static __inline struct mbuf *
1096 m_last(struct mbuf *m)
1097 {
1098 
1099 	while (m->m_next)
1100 		m = m->m_next;
1101 	return (m);
1102 }
1103 
1104 static inline u_int
1105 m_extrefcnt(struct mbuf *m)
1106 {
1107 
1108 	KASSERT(m->m_flags & M_EXT, ("%s: M_EXT missing", __func__));
1109 
1110 	return ((m->m_ext.ext_flags & EXT_FLAG_EMBREF) ? m->m_ext.ext_count :
1111 	    *m->m_ext.ext_cnt);
1112 }
1113 
1114 /*
1115  * mbuf, cluster, and external object allocation macros (for compatibility
1116  * purposes).
1117  */
1118 #define	M_MOVE_PKTHDR(to, from)	m_move_pkthdr((to), (from))
1119 #define	MGET(m, how, type)	((m) = m_get((how), (type)))
1120 #define	MGETHDR(m, how, type)	((m) = m_gethdr((how), (type)))
1121 #define	MCLGET(m, how)		m_clget((m), (how))
1122 #define	MEXTADD(m, buf, size, free, arg1, arg2, flags, type)		\
1123     m_extadd((m), (char *)(buf), (size), (free), (arg1), (arg2),	\
1124     (flags), (type))
1125 #define	m_getm(m, len, how, type)					\
1126     m_getm2((m), (len), (how), (type), M_PKTHDR)
1127 
1128 /*
1129  * Evaluate TRUE if it's safe to write to the mbuf m's data region (this can
1130  * be both the local data payload, or an external buffer area, depending on
1131  * whether M_EXT is set).
1132  */
1133 #define	M_WRITABLE(m)	(((m)->m_flags & (M_RDONLY | M_EXTPG)) == 0 &&	\
1134 			 (!(((m)->m_flags & M_EXT)) ||			\
1135 			 (m_extrefcnt(m) == 1)))
1136 
1137 /* Check if the supplied mbuf has a packet header, or else panic. */
1138 #define	M_ASSERTPKTHDR(m)						\
1139 	KASSERT((m) != NULL && (m)->m_flags & M_PKTHDR,			\
1140 	    ("%s: no mbuf packet header!", __func__))
1141 
1142 /* Check if mbuf is multipage. */
1143 #define M_ASSERTEXTPG(m)						\
1144 	KASSERT(((m)->m_flags & (M_EXTPG|M_PKTHDR)) == M_EXTPG,		\
1145 	    ("%s: m %p is not multipage!", __func__, m))
1146 
1147 /*
1148  * Ensure that the supplied mbuf is a valid, non-free mbuf.
1149  *
1150  * XXX: Broken at the moment.  Need some UMA magic to make it work again.
1151  */
1152 #define	M_ASSERTVALID(m)						\
1153 	KASSERT((((struct mbuf *)m)->m_flags & 0) == 0,			\
1154 	    ("%s: attempted use of a free mbuf!", __func__))
1155 
1156 /* Check whether any mbuf in the chain is unmapped. */
1157 #ifdef INVARIANTS
1158 #define	M_ASSERTMAPPED(m) do {						\
1159 	for (struct mbuf *__m = (m); __m != NULL; __m = __m->m_next)	\
1160 		KASSERT((__m->m_flags & M_EXTPG) == 0,			\
1161 		    ("%s: chain %p contains an unmapped mbuf", __func__, (m)));\
1162 } while (0)
1163 #else
1164 #define	M_ASSERTMAPPED(m)
1165 #endif
1166 
1167 /*
1168  * Return the address of the start of the buffer associated with an mbuf,
1169  * handling external storage, packet-header mbufs, and regular data mbufs.
1170  */
1171 #define	M_START(m)							\
1172 	(((m)->m_flags & M_EXTPG) ? NULL :				\
1173 	 ((m)->m_flags & M_EXT) ? (m)->m_ext.ext_buf :			\
1174 	 ((m)->m_flags & M_PKTHDR) ? &(m)->m_pktdat[0] :		\
1175 	 &(m)->m_dat[0])
1176 
1177 /*
1178  * Return the size of the buffer associated with an mbuf, handling external
1179  * storage, packet-header mbufs, and regular data mbufs.
1180  */
1181 #define	M_SIZE(m)							\
1182 	(((m)->m_flags & M_EXT) ? (m)->m_ext.ext_size :			\
1183 	 ((m)->m_flags & M_PKTHDR) ? MHLEN :				\
1184 	 MLEN)
1185 
1186 /*
1187  * Set the m_data pointer of a newly allocated mbuf to place an object of the
1188  * specified size at the end of the mbuf, longword aligned.
1189  *
1190  * NB: Historically, we had M_ALIGN(), MH_ALIGN(), and MEXT_ALIGN() as
1191  * separate macros, each asserting that it was called at the proper moment.
1192  * This required callers to themselves test the storage type and call the
1193  * right one.  Rather than require callers to be aware of those layout
1194  * decisions, we centralize here.
1195  */
1196 static __inline void
1197 m_align(struct mbuf *m, int len)
1198 {
1199 #ifdef INVARIANTS
1200 	const char *msg = "%s: not a virgin mbuf";
1201 #endif
1202 	int adjust;
1203 
1204 	KASSERT(m->m_data == M_START(m), (msg, __func__));
1205 
1206 	adjust = M_SIZE(m) - len;
1207 	m->m_data += adjust &~ (sizeof(long)-1);
1208 }
1209 
1210 #define	M_ALIGN(m, len)		m_align(m, len)
1211 #define	MH_ALIGN(m, len)	m_align(m, len)
1212 #define	MEXT_ALIGN(m, len)	m_align(m, len)
1213 
1214 /*
1215  * Compute the amount of space available before the current start of data in
1216  * an mbuf.
1217  *
1218  * The M_WRITABLE() is a temporary, conservative safety measure: the burden
1219  * of checking writability of the mbuf data area rests solely with the caller.
1220  *
1221  * NB: In previous versions, M_LEADINGSPACE() would only check M_WRITABLE()
1222  * for mbufs with external storage.  We now allow mbuf-embedded data to be
1223  * read-only as well.
1224  */
1225 #define	M_LEADINGSPACE(m)						\
1226 	(M_WRITABLE(m) ? ((m)->m_data - M_START(m)) : 0)
1227 
1228 /*
1229  * Compute the amount of space available after the end of data in an mbuf.
1230  *
1231  * The M_WRITABLE() is a temporary, conservative safety measure: the burden
1232  * of checking writability of the mbuf data area rests solely with the caller.
1233  *
1234  * NB: In previous versions, M_TRAILINGSPACE() would only check M_WRITABLE()
1235  * for mbufs with external storage.  We now allow mbuf-embedded data to be
1236  * read-only as well.
1237  */
1238 #define	M_TRAILINGSPACE(m)						\
1239 	(M_WRITABLE(m) ?						\
1240 	    ((M_START(m) + M_SIZE(m)) - ((m)->m_data + (m)->m_len)) : 0)
1241 
1242 /*
1243  * Arrange to prepend space of size plen to mbuf m.  If a new mbuf must be
1244  * allocated, how specifies whether to wait.  If the allocation fails, the
1245  * original mbuf chain is freed and m is set to NULL.
1246  */
1247 #define	M_PREPEND(m, plen, how) do {					\
1248 	struct mbuf **_mmp = &(m);					\
1249 	struct mbuf *_mm = *_mmp;					\
1250 	int _mplen = (plen);						\
1251 	int __mhow = (how);						\
1252 									\
1253 	MBUF_CHECKSLEEP(how);						\
1254 	if (M_LEADINGSPACE(_mm) >= _mplen) {				\
1255 		_mm->m_data -= _mplen;					\
1256 		_mm->m_len += _mplen;					\
1257 	} else								\
1258 		_mm = m_prepend(_mm, _mplen, __mhow);			\
1259 	if (_mm != NULL && _mm->m_flags & M_PKTHDR)			\
1260 		_mm->m_pkthdr.len += _mplen;				\
1261 	*_mmp = _mm;							\
1262 } while (0)
1263 
1264 /*
1265  * Change mbuf to new type.  This is a relatively expensive operation and
1266  * should be avoided.
1267  */
1268 #define	MCHTYPE(m, t)	m_chtype((m), (t))
1269 
1270 /* Return the rcvif of a packet header. */
1271 static __inline struct ifnet *
1272 m_rcvif(struct mbuf *m)
1273 {
1274 
1275 	M_ASSERTPKTHDR(m);
1276 	if (m->m_pkthdr.csum_flags & CSUM_SND_TAG)
1277 		return (NULL);
1278 	return (m->m_pkthdr.rcvif);
1279 }
1280 
1281 /* Length to m_copy to copy all. */
1282 #define	M_COPYALL	1000000000
1283 
1284 extern int		max_datalen;	/* MHLEN - max_hdr */
1285 extern int		max_hdr;	/* Largest link + protocol header */
1286 extern int		max_linkhdr;	/* Largest link-level header */
1287 extern int		max_protohdr;	/* Largest protocol header */
1288 extern int		nmbclusters;	/* Maximum number of clusters */
1289 extern bool		mb_use_ext_pgs;	/* Use ext_pgs for sendfile */
1290 
1291 /*-
1292  * Network packets may have annotations attached by affixing a list of
1293  * "packet tags" to the pkthdr structure.  Packet tags are dynamically
1294  * allocated semi-opaque data structures that have a fixed header
1295  * (struct m_tag) that specifies the size of the memory block and a
1296  * <cookie,type> pair that identifies it.  The cookie is a 32-bit unique
1297  * unsigned value used to identify a module or ABI.  By convention this value
1298  * is chosen as the date+time that the module is created, expressed as the
1299  * number of seconds since the epoch (e.g., using date -u +'%s').  The type
1300  * value is an ABI/module-specific value that identifies a particular
1301  * annotation and is private to the module.  For compatibility with systems
1302  * like OpenBSD that define packet tags w/o an ABI/module cookie, the value
1303  * PACKET_ABI_COMPAT is used to implement m_tag_get and m_tag_find
1304  * compatibility shim functions and several tag types are defined below.
1305  * Users that do not require compatibility should use a private cookie value
1306  * so that packet tag-related definitions can be maintained privately.
1307  *
1308  * Note that the packet tag returned by m_tag_alloc has the default memory
1309  * alignment implemented by malloc.  To reference private data one can use a
1310  * construct like:
1311  *
1312  *	struct m_tag *mtag = m_tag_alloc(...);
1313  *	struct foo *p = (struct foo *)(mtag+1);
1314  *
1315  * if the alignment of struct m_tag is sufficient for referencing members of
1316  * struct foo.  Otherwise it is necessary to embed struct m_tag within the
1317  * private data structure to insure proper alignment; e.g.,
1318  *
1319  *	struct foo {
1320  *		struct m_tag	tag;
1321  *		...
1322  *	};
1323  *	struct foo *p = (struct foo *) m_tag_alloc(...);
1324  *	struct m_tag *mtag = &p->tag;
1325  */
1326 
1327 /*
1328  * Persistent tags stay with an mbuf until the mbuf is reclaimed.  Otherwise
1329  * tags are expected to ``vanish'' when they pass through a network
1330  * interface.  For most interfaces this happens normally as the tags are
1331  * reclaimed when the mbuf is free'd.  However in some special cases
1332  * reclaiming must be done manually.  An example is packets that pass through
1333  * the loopback interface.  Also, one must be careful to do this when
1334  * ``turning around'' packets (e.g., icmp_reflect).
1335  *
1336  * To mark a tag persistent bit-or this flag in when defining the tag id.
1337  * The tag will then be treated as described above.
1338  */
1339 #define	MTAG_PERSISTENT				0x800
1340 
1341 #define	PACKET_TAG_NONE				0  /* Nadda */
1342 
1343 /* Packet tags for use with PACKET_ABI_COMPAT. */
1344 #define	PACKET_TAG_IPSEC_IN_DONE		1  /* IPsec applied, in */
1345 #define	PACKET_TAG_IPSEC_OUT_DONE		2  /* IPsec applied, out */
1346 #define	PACKET_TAG_IPSEC_IN_CRYPTO_DONE		3  /* NIC IPsec crypto done */
1347 #define	PACKET_TAG_IPSEC_OUT_CRYPTO_NEEDED	4  /* NIC IPsec crypto req'ed */
1348 #define	PACKET_TAG_IPSEC_IN_COULD_DO_CRYPTO	5  /* NIC notifies IPsec */
1349 #define	PACKET_TAG_IPSEC_PENDING_TDB		6  /* Reminder to do IPsec */
1350 #define	PACKET_TAG_BRIDGE			7  /* Bridge processing done */
1351 #define	PACKET_TAG_GIF				8  /* GIF processing done */
1352 #define	PACKET_TAG_GRE				9  /* GRE processing done */
1353 #define	PACKET_TAG_IN_PACKET_CHECKSUM		10 /* NIC checksumming done */
1354 #define	PACKET_TAG_ENCAP			11 /* Encap.  processing */
1355 #define	PACKET_TAG_IPSEC_SOCKET			12 /* IPSEC socket ref */
1356 #define	PACKET_TAG_IPSEC_HISTORY		13 /* IPSEC history */
1357 #define	PACKET_TAG_IPV6_INPUT			14 /* IPV6 input processing */
1358 #define	PACKET_TAG_DUMMYNET			15 /* dummynet info */
1359 #define	PACKET_TAG_DIVERT			17 /* divert info */
1360 #define	PACKET_TAG_IPFORWARD			18 /* ipforward info */
1361 #define	PACKET_TAG_MACLABEL	(19 | MTAG_PERSISTENT) /* MAC label */
1362 #define	PACKET_TAG_PF				21 /* PF/ALTQ information */
1363 #define	PACKET_TAG_RTSOCKFAM			25 /* rtsock sa family */
1364 #define	PACKET_TAG_IPOPTIONS			27 /* Saved IP options */
1365 #define	PACKET_TAG_CARP				28 /* CARP info */
1366 #define	PACKET_TAG_IPSEC_NAT_T_PORTS		29 /* two uint16_t */
1367 #define	PACKET_TAG_ND_OUTGOING			30 /* ND outgoing */
1368 
1369 /* Specific cookies and tags. */
1370 
1371 /* Packet tag routines. */
1372 struct m_tag	*m_tag_alloc(u_int32_t, int, int, int);
1373 void		 m_tag_delete(struct mbuf *, struct m_tag *);
1374 void		 m_tag_delete_chain(struct mbuf *, struct m_tag *);
1375 void		 m_tag_free_default(struct m_tag *);
1376 struct m_tag	*m_tag_locate(struct mbuf *, u_int32_t, int, struct m_tag *);
1377 struct m_tag	*m_tag_copy(struct m_tag *, int);
1378 int		 m_tag_copy_chain(struct mbuf *, const struct mbuf *, int);
1379 void		 m_tag_delete_nonpersistent(struct mbuf *);
1380 
1381 /*
1382  * Initialize the list of tags associated with an mbuf.
1383  */
1384 static __inline void
1385 m_tag_init(struct mbuf *m)
1386 {
1387 
1388 	SLIST_INIT(&m->m_pkthdr.tags);
1389 }
1390 
1391 /*
1392  * Set up the contents of a tag.  Note that this does not fill in the free
1393  * method; the caller is expected to do that.
1394  *
1395  * XXX probably should be called m_tag_init, but that was already taken.
1396  */
1397 static __inline void
1398 m_tag_setup(struct m_tag *t, u_int32_t cookie, int type, int len)
1399 {
1400 
1401 	t->m_tag_id = type;
1402 	t->m_tag_len = len;
1403 	t->m_tag_cookie = cookie;
1404 }
1405 
1406 /*
1407  * Reclaim resources associated with a tag.
1408  */
1409 static __inline void
1410 m_tag_free(struct m_tag *t)
1411 {
1412 
1413 	(*t->m_tag_free)(t);
1414 }
1415 
1416 /*
1417  * Return the first tag associated with an mbuf.
1418  */
1419 static __inline struct m_tag *
1420 m_tag_first(struct mbuf *m)
1421 {
1422 
1423 	return (SLIST_FIRST(&m->m_pkthdr.tags));
1424 }
1425 
1426 /*
1427  * Return the next tag in the list of tags associated with an mbuf.
1428  */
1429 static __inline struct m_tag *
1430 m_tag_next(struct mbuf *m __unused, struct m_tag *t)
1431 {
1432 
1433 	return (SLIST_NEXT(t, m_tag_link));
1434 }
1435 
1436 /*
1437  * Prepend a tag to the list of tags associated with an mbuf.
1438  */
1439 static __inline void
1440 m_tag_prepend(struct mbuf *m, struct m_tag *t)
1441 {
1442 
1443 	SLIST_INSERT_HEAD(&m->m_pkthdr.tags, t, m_tag_link);
1444 }
1445 
1446 /*
1447  * Unlink a tag from the list of tags associated with an mbuf.
1448  */
1449 static __inline void
1450 m_tag_unlink(struct mbuf *m, struct m_tag *t)
1451 {
1452 
1453 	SLIST_REMOVE(&m->m_pkthdr.tags, t, m_tag, m_tag_link);
1454 }
1455 
1456 /* These are for OpenBSD compatibility. */
1457 #define	MTAG_ABI_COMPAT		0		/* compatibility ABI */
1458 
1459 static __inline struct m_tag *
1460 m_tag_get(int type, int length, int wait)
1461 {
1462 	return (m_tag_alloc(MTAG_ABI_COMPAT, type, length, wait));
1463 }
1464 
1465 static __inline struct m_tag *
1466 m_tag_find(struct mbuf *m, int type, struct m_tag *start)
1467 {
1468 	return (SLIST_EMPTY(&m->m_pkthdr.tags) ? (struct m_tag *)NULL :
1469 	    m_tag_locate(m, MTAG_ABI_COMPAT, type, start));
1470 }
1471 
1472 static inline struct m_snd_tag *
1473 m_snd_tag_ref(struct m_snd_tag *mst)
1474 {
1475 
1476 	refcount_acquire(&mst->refcount);
1477 	return (mst);
1478 }
1479 
1480 static inline void
1481 m_snd_tag_rele(struct m_snd_tag *mst)
1482 {
1483 
1484 	if (refcount_release(&mst->refcount))
1485 		m_snd_tag_destroy(mst);
1486 }
1487 
1488 static __inline struct mbuf *
1489 m_free(struct mbuf *m)
1490 {
1491 	struct mbuf *n = m->m_next;
1492 
1493 	MBUF_PROBE1(m__free, m);
1494 	if ((m->m_flags & (M_PKTHDR|M_NOFREE)) == (M_PKTHDR|M_NOFREE))
1495 		m_tag_delete_chain(m, NULL);
1496 	if (m->m_flags & M_PKTHDR && m->m_pkthdr.csum_flags & CSUM_SND_TAG)
1497 		m_snd_tag_rele(m->m_pkthdr.snd_tag);
1498 	if (m->m_flags & M_EXTPG)
1499 		mb_free_extpg(m);
1500 	else if (m->m_flags & M_EXT)
1501 		mb_free_ext(m);
1502 	else if ((m->m_flags & M_NOFREE) == 0)
1503 		uma_zfree(zone_mbuf, m);
1504 	return (n);
1505 }
1506 
1507 static __inline int
1508 rt_m_getfib(struct mbuf *m)
1509 {
1510 	KASSERT(m->m_flags & M_PKTHDR , ("Attempt to get FIB from non header mbuf."));
1511 	return (m->m_pkthdr.fibnum);
1512 }
1513 
1514 #define M_GETFIB(_m)   rt_m_getfib(_m)
1515 
1516 #define M_SETFIB(_m, _fib) do {						\
1517         KASSERT((_m)->m_flags & M_PKTHDR, ("Attempt to set FIB on non header mbuf."));	\
1518 	((_m)->m_pkthdr.fibnum) = (_fib);				\
1519 } while (0)
1520 
1521 /* flags passed as first argument for "m_xxx_tcpip_hash()" */
1522 #define	MBUF_HASHFLAG_L2	(1 << 2)
1523 #define	MBUF_HASHFLAG_L3	(1 << 3)
1524 #define	MBUF_HASHFLAG_L4	(1 << 4)
1525 
1526 /* mbuf hashing helper routines */
1527 uint32_t	m_ether_tcpip_hash_init(void);
1528 uint32_t	m_ether_tcpip_hash(const uint32_t, const struct mbuf *, uint32_t);
1529 uint32_t	m_infiniband_tcpip_hash_init(void);
1530 uint32_t	m_infiniband_tcpip_hash(const uint32_t, const struct mbuf *, uint32_t);
1531 
1532 #ifdef MBUF_PROFILING
1533  void m_profile(struct mbuf *m);
1534  #define M_PROFILE(m) m_profile(m)
1535 #else
1536  #define M_PROFILE(m)
1537 #endif
1538 
1539 struct mbufq {
1540 	STAILQ_HEAD(, mbuf)	mq_head;
1541 	int			mq_len;
1542 	int			mq_maxlen;
1543 };
1544 
1545 static inline void
1546 mbufq_init(struct mbufq *mq, int maxlen)
1547 {
1548 
1549 	STAILQ_INIT(&mq->mq_head);
1550 	mq->mq_maxlen = maxlen;
1551 	mq->mq_len = 0;
1552 }
1553 
1554 static inline struct mbuf *
1555 mbufq_flush(struct mbufq *mq)
1556 {
1557 	struct mbuf *m;
1558 
1559 	m = STAILQ_FIRST(&mq->mq_head);
1560 	STAILQ_INIT(&mq->mq_head);
1561 	mq->mq_len = 0;
1562 	return (m);
1563 }
1564 
1565 static inline void
1566 mbufq_drain(struct mbufq *mq)
1567 {
1568 	struct mbuf *m, *n;
1569 
1570 	n = mbufq_flush(mq);
1571 	while ((m = n) != NULL) {
1572 		n = STAILQ_NEXT(m, m_stailqpkt);
1573 		m_freem(m);
1574 	}
1575 }
1576 
1577 static inline struct mbuf *
1578 mbufq_first(const struct mbufq *mq)
1579 {
1580 
1581 	return (STAILQ_FIRST(&mq->mq_head));
1582 }
1583 
1584 static inline struct mbuf *
1585 mbufq_last(const struct mbufq *mq)
1586 {
1587 
1588 	return (STAILQ_LAST(&mq->mq_head, mbuf, m_stailqpkt));
1589 }
1590 
1591 static inline int
1592 mbufq_full(const struct mbufq *mq)
1593 {
1594 
1595 	return (mq->mq_maxlen > 0 && mq->mq_len >= mq->mq_maxlen);
1596 }
1597 
1598 static inline int
1599 mbufq_len(const struct mbufq *mq)
1600 {
1601 
1602 	return (mq->mq_len);
1603 }
1604 
1605 static inline int
1606 mbufq_enqueue(struct mbufq *mq, struct mbuf *m)
1607 {
1608 
1609 	if (mbufq_full(mq))
1610 		return (ENOBUFS);
1611 	STAILQ_INSERT_TAIL(&mq->mq_head, m, m_stailqpkt);
1612 	mq->mq_len++;
1613 	return (0);
1614 }
1615 
1616 static inline struct mbuf *
1617 mbufq_dequeue(struct mbufq *mq)
1618 {
1619 	struct mbuf *m;
1620 
1621 	m = STAILQ_FIRST(&mq->mq_head);
1622 	if (m) {
1623 		STAILQ_REMOVE_HEAD(&mq->mq_head, m_stailqpkt);
1624 		m->m_nextpkt = NULL;
1625 		mq->mq_len--;
1626 	}
1627 	return (m);
1628 }
1629 
1630 static inline void
1631 mbufq_prepend(struct mbufq *mq, struct mbuf *m)
1632 {
1633 
1634 	STAILQ_INSERT_HEAD(&mq->mq_head, m, m_stailqpkt);
1635 	mq->mq_len++;
1636 }
1637 
1638 /*
1639  * Note: this doesn't enforce the maximum list size for dst.
1640  */
1641 static inline void
1642 mbufq_concat(struct mbufq *mq_dst, struct mbufq *mq_src)
1643 {
1644 
1645 	mq_dst->mq_len += mq_src->mq_len;
1646 	STAILQ_CONCAT(&mq_dst->mq_head, &mq_src->mq_head);
1647 	mq_src->mq_len = 0;
1648 }
1649 
1650 #ifdef _SYS_TIMESPEC_H_
1651 static inline void
1652 mbuf_tstmp2timespec(struct mbuf *m, struct timespec *ts)
1653 {
1654 
1655 	KASSERT((m->m_flags & M_PKTHDR) != 0, ("mbuf %p no M_PKTHDR", m));
1656 	KASSERT((m->m_flags & (M_TSTMP|M_TSTMP_LRO)) != 0, ("mbuf %p no M_TSTMP or M_TSTMP_LRO", m));
1657 	ts->tv_sec = m->m_pkthdr.rcv_tstmp / 1000000000;
1658 	ts->tv_nsec = m->m_pkthdr.rcv_tstmp % 1000000000;
1659 }
1660 #endif
1661 
1662 #ifdef DEBUGNET
1663 /* Invoked from the debugnet client code. */
1664 void	debugnet_mbuf_drain(void);
1665 void	debugnet_mbuf_start(void);
1666 void	debugnet_mbuf_finish(void);
1667 void	debugnet_mbuf_reinit(int nmbuf, int nclust, int clsize);
1668 #endif
1669 
1670 static inline bool
1671 mbuf_has_tls_session(struct mbuf *m)
1672 {
1673 
1674 	if (m->m_flags & M_EXTPG) {
1675 		if (m->m_epg_tls != NULL) {
1676 			return (true);
1677 		}
1678 	}
1679 	return (false);
1680 }
1681 
1682 #endif /* _KERNEL */
1683 #endif /* !_SYS_MBUF_H_ */
1684