xref: /freebsd/sys/sys/mbuf.h (revision bbd421cdf6d8c6102e6fd3979c5bec21ace3c2e3)
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);
66 SDT_PROBE_DECLARE(sdt, , , m__get);
67 SDT_PROBE_DECLARE(sdt, , , m__getcl);
68 SDT_PROBE_DECLARE(sdt, , , m__getjcl);
69 SDT_PROBE_DECLARE(sdt, , , m__clget);
70 SDT_PROBE_DECLARE(sdt, , , m__cljget);
71 SDT_PROBE_DECLARE(sdt, , , m__cljset);
72 SDT_PROBE_DECLARE(sdt, , , m__free);
73 SDT_PROBE_DECLARE(sdt, , , m__freem);
74 
75 #endif /* _KERNEL */
76 
77 /*
78  * Mbufs are of a single size, MSIZE (sys/param.h), which includes overhead.
79  * An mbuf may add a single "mbuf cluster" of size MCLBYTES (also in
80  * sys/param.h), which has no additional overhead and is used instead of the
81  * internal data area; this is done when at least MINCLSIZE of data must be
82  * stored.  Additionally, it is possible to allocate a separate buffer
83  * externally and attach it to the mbuf in a way similar to that of mbuf
84  * clusters.
85  *
86  * NB: These calculation do not take actual compiler-induced alignment and
87  * padding inside the complete struct mbuf into account.  Appropriate
88  * attention is required when changing members of struct mbuf.
89  *
90  * MLEN is data length in a normal mbuf.
91  * MHLEN is data length in an mbuf with pktheader.
92  * MINCLSIZE is a smallest amount of data that should be put into cluster.
93  *
94  * Compile-time assertions in uipc_mbuf.c test these values to ensure that
95  * they are sensible.
96  */
97 struct mbuf;
98 #define	MHSIZE		offsetof(struct mbuf, m_dat)
99 #define	MPKTHSIZE	offsetof(struct mbuf, m_pktdat)
100 #define	MLEN		((int)(MSIZE - MHSIZE))
101 #define	MHLEN		((int)(MSIZE - MPKTHSIZE))
102 #define	MINCLSIZE	(MHLEN + 1)
103 #define	M_NODOM		255
104 
105 #ifdef _KERNEL
106 /*-
107  * Macro for type conversion: convert mbuf pointer to data pointer of correct
108  * type:
109  *
110  * mtod(m, t)	-- Convert mbuf pointer to data pointer of correct type.
111  * mtodo(m, o) -- Same as above but with offset 'o' into data.
112  */
113 #define	mtod(m, t)	((t)((m)->m_data))
114 #define	mtodo(m, o)	((void *)(((m)->m_data) + (o)))
115 
116 /*
117  * Argument structure passed to UMA routines during mbuf and packet
118  * allocations.
119  */
120 struct mb_args {
121 	int	flags;	/* Flags for mbuf being allocated */
122 	short	type;	/* Type of mbuf being allocated */
123 };
124 #endif /* _KERNEL */
125 
126 /*
127  * Packet tag structure (see below for details).
128  */
129 struct m_tag {
130 	SLIST_ENTRY(m_tag)	m_tag_link;	/* List of packet tags */
131 	u_int16_t		m_tag_id;	/* Tag ID */
132 	u_int16_t		m_tag_len;	/* Length of data */
133 	u_int32_t		m_tag_cookie;	/* ABI/Module ID */
134 	void			(*m_tag_free)(struct m_tag *);
135 };
136 
137 /*
138  * Static network interface owned tag.
139  * Allocated through ifp->if_snd_tag_alloc().
140  */
141 struct m_snd_tag {
142 	struct ifnet *ifp;		/* network interface tag belongs to */
143 	volatile u_int refcount;
144 	u_int	type;			/* One of IF_SND_TAG_TYPE_*. */
145 };
146 
147 /*
148  * Record/packet header in first mbuf of chain; valid only if M_PKTHDR is set.
149  * Size ILP32: 48
150  *	 LP64: 56
151  * Compile-time assertions in uipc_mbuf.c test these values to ensure that
152  * they are correct.
153  */
154 struct pkthdr {
155 	union {
156 		struct m_snd_tag *snd_tag;	/* send tag, if any */
157 		struct ifnet	*rcvif;		/* rcv interface */
158 	};
159 	SLIST_HEAD(packet_tags, m_tag) tags; /* list of packet tags */
160 	int32_t		 len;		/* total packet length */
161 
162 	/* Layer crossing persistent information. */
163 	uint32_t	 flowid;	/* packet's 4-tuple system */
164 	uint32_t	 csum_flags;	/* checksum and offload features */
165 	uint16_t	 fibnum;	/* this packet should use this fib */
166 	uint8_t		 numa_domain;	/* NUMA domain of recvd pkt */
167 	uint8_t		 rsstype;	/* hash type */
168 	union {
169 		uint64_t	rcv_tstmp;	/* timestamp in ns */
170 		struct {
171 			uint8_t		 l2hlen;	/* layer 2 hdr len */
172 			uint8_t		 l3hlen;	/* layer 3 hdr len */
173 			uint8_t		 l4hlen;	/* layer 4 hdr len */
174 			uint8_t		 l5hlen;	/* layer 5 hdr len */
175 			uint8_t		 inner_l2hlen;
176 			uint8_t		 inner_l3hlen;
177 			uint8_t		 inner_l4hlen;
178 			uint8_t		 inner_l5hlen;
179 		};
180 	};
181 	union {
182 		uint8_t  eight[8];
183 		uint16_t sixteen[4];
184 		uint32_t thirtytwo[2];
185 		uint64_t sixtyfour[1];
186 		uintptr_t unintptr[1];
187 		void	*ptr;
188 	} PH_per;
189 
190 	/* Layer specific non-persistent local storage for reassembly, etc. */
191 	union {
192 		uint8_t  eight[8];
193 		uint16_t sixteen[4];
194 		uint32_t thirtytwo[2];
195 		uint64_t sixtyfour[1];
196 		uintptr_t unintptr[1];
197 		void 	*ptr;
198 	} PH_loc;
199 };
200 #define	ether_vtag	PH_per.sixteen[0]
201 #define tcp_tun_port	PH_per.sixteen[0] /* outbound */
202 #define	PH_vt		PH_per
203 #define	vt_nrecs	sixteen[0]	  /* mld and v6-ND */
204 #define	tso_segsz	PH_per.sixteen[1] /* inbound after LRO */
205 #define	lro_nsegs	tso_segsz	  /* inbound after LRO */
206 #define	csum_data	PH_per.thirtytwo[1] /* inbound from hardware up */
207 #define lro_len		PH_loc.sixteen[0] /* inbound during LRO (no reassembly) */
208 #define lro_csum	PH_loc.sixteen[1] /* inbound during LRO (no reassembly) */
209 #define lro_etype	PH_loc.sixteen[2] /* inbound during LRO (no reassembly) */
210 /* Note PH_loc is used during IP reassembly (all 8 bytes as a ptr) */
211 
212 /*
213  * TLS records for TLS 1.0-1.2 can have the following header lengths:
214  * - 5 (AES-CBC with implicit IV)
215  * - 21 (AES-CBC with explicit IV)
216  * - 13 (AES-GCM with 8 byte explicit IV)
217  */
218 #define	MBUF_PEXT_HDR_LEN	23
219 
220 /*
221  * TLS records for TLS 1.0-1.2 can have the following maximum trailer
222  * lengths:
223  * - 16 (AES-GCM)
224  * - 36 (AES-CBC with SHA1 and up to 16 bytes of padding)
225  * - 48 (AES-CBC with SHA2-256 and up to 16 bytes of padding)
226  * - 64 (AES-CBC with SHA2-384 and up to 16 bytes of padding)
227  */
228 #define	MBUF_PEXT_TRAIL_LEN	64
229 
230 #if defined(__LP64__)
231 #define MBUF_PEXT_MAX_PGS (40 / sizeof(vm_paddr_t))
232 #else
233 #define MBUF_PEXT_MAX_PGS (72 / sizeof(vm_paddr_t))
234 #endif
235 
236 #define	MBUF_PEXT_MAX_BYTES						\
237     (MBUF_PEXT_MAX_PGS * PAGE_SIZE + MBUF_PEXT_HDR_LEN + MBUF_PEXT_TRAIL_LEN)
238 
239 struct ktls_session;
240 struct socket;
241 
242 /*
243  * Description of external storage mapped into mbuf; valid only if M_EXT is
244  * set.
245  * Size ILP32: 28
246  *	 LP64: 48
247  * Compile-time assertions in uipc_mbuf.c test these values to ensure that
248  * they are correct.
249  */
250 typedef	void m_ext_free_t(struct mbuf *);
251 struct m_ext {
252 	union {
253 		/*
254 		 * If EXT_FLAG_EMBREF is set, then we use refcount in the
255 		 * mbuf, the 'ext_count' member.  Otherwise, we have a
256 		 * shadow copy and we use pointer 'ext_cnt'.  The original
257 		 * mbuf is responsible to carry the pointer to free routine
258 		 * and its arguments.  They aren't copied into shadows in
259 		 * mb_dupcl() to avoid dereferencing next cachelines.
260 		 */
261 		volatile u_int	 ext_count;
262 		volatile u_int	*ext_cnt;
263 	};
264 	uint32_t	 ext_size;	/* size of buffer, for ext_free */
265 	uint32_t	 ext_type:8,	/* type of external storage */
266 			 ext_flags:24;	/* external storage mbuf flags */
267 	union {
268 		struct {
269 			/*
270 			 * Regular M_EXT mbuf:
271 			 * o ext_buf always points to the external buffer.
272 			 * o ext_free (below) and two optional arguments
273 			 *   ext_arg1 and ext_arg2 store the free context for
274 			 *   the external storage.  They are set only in the
275 			 *   refcount carrying mbuf, the one with
276 			 *   EXT_FLAG_EMBREF flag, with exclusion for
277 			 *   EXT_EXTREF type, where the free context is copied
278 			 *   into all mbufs that use same external storage.
279 			 */
280 			char 	*ext_buf;	/* start of buffer */
281 #define	m_ext_copylen	offsetof(struct m_ext, ext_arg2)
282 			void	*ext_arg2;
283 		};
284 		struct {
285 			/*
286 			 * Multi-page M_EXTPG mbuf:
287 			 * o extpg_pa - page vector.
288 			 * o extpg_trail and extpg_hdr - TLS trailer and
289 			 *   header.
290 			 * Uses ext_free and may also use ext_arg1.
291 			 */
292 			vm_paddr_t	extpg_pa[MBUF_PEXT_MAX_PGS];
293 			char		extpg_trail[MBUF_PEXT_TRAIL_LEN];
294 			char		extpg_hdr[MBUF_PEXT_HDR_LEN];
295 			/* Pretend these 3 fields are part of mbuf itself. */
296 #define	m_epg_pa	m_ext.extpg_pa
297 #define	m_epg_trail	m_ext.extpg_trail
298 #define	m_epg_hdr	m_ext.extpg_hdr
299 #define	m_epg_ext_copylen	offsetof(struct m_ext, ext_free)
300 		};
301 	};
302 	/*
303 	 * Free method and optional argument pointer, both
304 	 * used by M_EXT and M_EXTPG.
305 	 */
306 	m_ext_free_t	*ext_free;
307 	void		*ext_arg1;
308 };
309 
310 /*
311  * The core of the mbuf object along with some shortcut defines for practical
312  * purposes.
313  */
314 struct mbuf {
315 	/*
316 	 * Header present at the beginning of every mbuf.
317 	 * Size ILP32: 24
318 	 *      LP64: 32
319 	 * Compile-time assertions in uipc_mbuf.c test these values to ensure
320 	 * that they are correct.
321 	 */
322 	union {	/* next buffer in chain */
323 		struct mbuf		*m_next;
324 		SLIST_ENTRY(mbuf)	m_slist;
325 		STAILQ_ENTRY(mbuf)	m_stailq;
326 	};
327 	union {	/* next chain in queue/record */
328 		struct mbuf		*m_nextpkt;
329 		SLIST_ENTRY(mbuf)	m_slistpkt;
330 		STAILQ_ENTRY(mbuf)	m_stailqpkt;
331 	};
332 	caddr_t		 m_data;	/* location of data */
333 	int32_t		 m_len;		/* amount of data in this mbuf */
334 	uint32_t	 m_type:8,	/* type of data in this mbuf */
335 			 m_flags:24;	/* flags; see below */
336 #if !defined(__LP64__)
337 	uint32_t	 m_pad;		/* pad for 64bit alignment */
338 #endif
339 
340 	/*
341 	 * A set of optional headers (packet header, external storage header)
342 	 * and internal data storage.  Historically, these arrays were sized
343 	 * to MHLEN (space left after a packet header) and MLEN (space left
344 	 * after only a regular mbuf header); they are now variable size in
345 	 * order to support future work on variable-size mbufs.
346 	 */
347 	union {
348 		struct {
349 			union {
350 				/* M_PKTHDR set. */
351 				struct pkthdr	m_pkthdr;
352 
353 				/* M_EXTPG set.
354 				 * Multi-page M_EXTPG mbuf has its meta data
355 				 * split between the below anonymous structure
356 				 * and m_ext.  It carries vector of pages,
357 				 * optional header and trailer char vectors
358 				 * and pointers to socket/TLS data.
359 				 */
360 #define	m_epg_startcopy		m_epg_npgs
361 #define	m_epg_endcopy		m_epg_stailq
362 				struct {
363 					/* Overall count of pages and count of
364 					 * pages with I/O pending. */
365 					uint8_t	m_epg_npgs;
366 					uint8_t	m_epg_nrdy;
367 					/* TLS header and trailer lengths.
368 					 * The data itself resides in m_ext. */
369 					uint8_t	m_epg_hdrlen;
370 					uint8_t	m_epg_trllen;
371 					/* Offset into 1st page and length of
372 					 * data in the last page. */
373 					uint16_t m_epg_1st_off;
374 					uint16_t m_epg_last_len;
375 					uint8_t	m_epg_flags;
376 #define	EPG_FLAG_ANON	0x1	/* Data can be encrypted in place. */
377 #define	EPG_FLAG_2FREE	0x2	/* Scheduled for free. */
378 					uint8_t	m_epg_record_type;
379 					uint8_t	__spare[2];
380 					int	m_epg_enc_cnt;
381 					struct ktls_session *m_epg_tls;
382 					struct socket	*m_epg_so;
383 					uint64_t	m_epg_seqno;
384 					STAILQ_ENTRY(mbuf) m_epg_stailq;
385 				};
386 			};
387 			union {
388 				/* M_EXT or M_EXTPG set. */
389 				struct m_ext	m_ext;
390 				/* M_PKTHDR set, neither M_EXT nor M_EXTPG. */
391 				char		m_pktdat[0];
392 			};
393 		};
394 		char	m_dat[0];			/* !M_PKTHDR, !M_EXT */
395 	};
396 };
397 
398 #ifdef _KERNEL
399 static inline int
400 m_epg_pagelen(const struct mbuf *m, int pidx, int pgoff)
401 {
402 
403 	KASSERT(pgoff == 0 || pidx == 0,
404 	    ("page %d with non-zero offset %d in %p", pidx, pgoff, m));
405 
406 	if (pidx == m->m_epg_npgs - 1) {
407 		return (m->m_epg_last_len);
408 	} else {
409 		return (PAGE_SIZE - pgoff);
410 	}
411 }
412 
413 #ifdef INVARIANTS
414 #define	MCHECK(ex, msg)	KASSERT((ex),				\
415 	    ("Multi page mbuf %p with " #msg " at %s:%d",	\
416 	    m, __FILE__, __LINE__))
417 /*
418  * NB: This expects a non-empty buffer (npgs > 0 and
419  * last_pg_len > 0).
420  */
421 #define	MBUF_EXT_PGS_ASSERT_SANITY(m)	do {				\
422 	MCHECK(m->m_epg_npgs > 0, "no valid pages");		\
423 	MCHECK(m->m_epg_npgs <= nitems(m->m_epg_pa),		\
424 	    "too many pages");						\
425 	MCHECK(m->m_epg_nrdy <= m->m_epg_npgs,			\
426 	    "too many ready pages");					\
427 	MCHECK(m->m_epg_1st_off < PAGE_SIZE,			\
428 		"too large page offset");				\
429 	MCHECK(m->m_epg_last_len > 0, "zero last page length");	\
430 	MCHECK(m->m_epg_last_len <= PAGE_SIZE,			\
431 	    "too large last page length");				\
432 	if (m->m_epg_npgs == 1)					\
433 		MCHECK(m->m_epg_1st_off +			\
434 		    m->m_epg_last_len <=	 PAGE_SIZE,		\
435 		    "single page too large");				\
436 	MCHECK(m->m_epg_hdrlen <= sizeof(m->m_epg_hdr),		\
437 	    "too large header length");					\
438 	MCHECK(m->m_epg_trllen <= sizeof(m->m_epg_trail),	\
439 	    "too large header length");					\
440 } while (0)
441 #else
442 #define	MBUF_EXT_PGS_ASSERT_SANITY(m)	do {} while (0);
443 #endif
444 #endif
445 
446 /*
447  * mbuf flags of global significance and layer crossing.
448  * Those of only protocol/layer specific significance are to be mapped
449  * to M_PROTO[1-11] and cleared at layer handoff boundaries.
450  * NB: Limited to the lower 24 bits.
451  */
452 #define	M_EXT		0x00000001 /* has associated external storage */
453 #define	M_PKTHDR	0x00000002 /* start of record */
454 #define	M_EOR		0x00000004 /* end of record */
455 #define	M_RDONLY	0x00000008 /* associated data is marked read-only */
456 #define	M_BCAST		0x00000010 /* send/received as link-level broadcast */
457 #define	M_MCAST		0x00000020 /* send/received as link-level multicast */
458 #define	M_PROMISC	0x00000040 /* packet was not for us */
459 #define	M_VLANTAG	0x00000080 /* ether_vtag is valid */
460 #define	M_EXTPG		0x00000100 /* has array of unmapped pages and TLS */
461 #define	M_NOFREE	0x00000200 /* do not free mbuf, embedded in cluster */
462 #define	M_TSTMP		0x00000400 /* rcv_tstmp field is valid */
463 #define	M_TSTMP_HPREC	0x00000800 /* rcv_tstmp is high-prec, typically
464 				      hw-stamped on port (useful for IEEE 1588
465 				      and 802.1AS) */
466 #define M_TSTMP_LRO	0x00001000 /* Time LRO pushed in pkt is valid in (PH_loc) */
467 
468 #define	M_PROTO1	0x00002000 /* protocol-specific */
469 #define	M_PROTO2	0x00004000 /* protocol-specific */
470 #define	M_PROTO3	0x00008000 /* protocol-specific */
471 #define	M_PROTO4	0x00010000 /* protocol-specific */
472 #define	M_PROTO5	0x00020000 /* protocol-specific */
473 #define	M_PROTO6	0x00040000 /* protocol-specific */
474 #define	M_PROTO7	0x00080000 /* protocol-specific */
475 #define	M_PROTO8	0x00100000 /* protocol-specific */
476 #define	M_PROTO9	0x00200000 /* protocol-specific */
477 #define	M_PROTO10	0x00400000 /* protocol-specific */
478 #define	M_PROTO11	0x00800000 /* protocol-specific */
479 
480 #define MB_DTOR_SKIP	0x1	/* don't pollute the cache by touching a freed mbuf */
481 
482 /*
483  * Flags to purge when crossing layers.
484  */
485 #define	M_PROTOFLAGS \
486     (M_PROTO1|M_PROTO2|M_PROTO3|M_PROTO4|M_PROTO5|M_PROTO6|M_PROTO7|M_PROTO8|\
487      M_PROTO9|M_PROTO10|M_PROTO11)
488 
489 /*
490  * Flags preserved when copying m_pkthdr.
491  */
492 #define M_COPYFLAGS \
493     (M_PKTHDR|M_EOR|M_RDONLY|M_BCAST|M_MCAST|M_PROMISC|M_VLANTAG|M_TSTMP| \
494      M_TSTMP_HPREC|M_TSTMP_LRO|M_PROTOFLAGS)
495 
496 /*
497  * Mbuf flag description for use with printf(9) %b identifier.
498  */
499 #define	M_FLAG_BITS \
500     "\20\1M_EXT\2M_PKTHDR\3M_EOR\4M_RDONLY\5M_BCAST\6M_MCAST" \
501     "\7M_PROMISC\10M_VLANTAG\11M_EXTPG\12M_NOFREE\13M_TSTMP\14M_TSTMP_HPREC\15M_TSTMP_LRO"
502 #define	M_FLAG_PROTOBITS \
503     "\16M_PROTO1\17M_PROTO2\20M_PROTO3\21M_PROTO4" \
504     "\22M_PROTO5\23M_PROTO6\24M_PROTO7\25M_PROTO8\26M_PROTO9" \
505     "\27M_PROTO10\28M_PROTO11"
506 #define	M_FLAG_PRINTF (M_FLAG_BITS M_FLAG_PROTOBITS)
507 
508 /*
509  * Network interface cards are able to hash protocol fields (such as IPv4
510  * addresses and TCP port numbers) classify packets into flows.  These flows
511  * can then be used to maintain ordering while delivering packets to the OS
512  * via parallel input queues, as well as to provide a stateless affinity
513  * model.  NIC drivers can pass up the hash via m->m_pkthdr.flowid, and set
514  * m_flag fields to indicate how the hash should be interpreted by the
515  * network stack.
516  *
517  * Most NICs support RSS, which provides ordering and explicit affinity, and
518  * use the hash m_flag bits to indicate what header fields were covered by
519  * the hash.  M_HASHTYPE_OPAQUE and M_HASHTYPE_OPAQUE_HASH can be set by non-
520  * RSS cards or configurations that provide an opaque flow identifier, allowing
521  * for ordering and distribution without explicit affinity.  Additionally,
522  * M_HASHTYPE_OPAQUE_HASH indicates that the flow identifier has hash
523  * properties.
524  *
525  * The meaning of the IPV6_EX suffix:
526  * "o  Home address from the home address option in the IPv6 destination
527  *     options header.  If the extension header is not present, use the Source
528  *     IPv6 Address.
529  *  o  IPv6 address that is contained in the Routing-Header-Type-2 from the
530  *     associated extension header.  If the extension header is not present,
531  *     use the Destination IPv6 Address."
532  * Quoted from:
533  * https://docs.microsoft.com/en-us/windows-hardware/drivers/network/rss-hashing-types#ndishashipv6ex
534  */
535 #define	M_HASHTYPE_HASHPROP		0x80	/* has hash properties */
536 #define	M_HASHTYPE_INNER		0x40	/* calculated from inner headers */
537 #define	M_HASHTYPE_HASH(t)		(M_HASHTYPE_HASHPROP | (t))
538 /* Microsoft RSS standard hash types */
539 #define	M_HASHTYPE_NONE			0
540 #define	M_HASHTYPE_RSS_IPV4		M_HASHTYPE_HASH(1) /* IPv4 2-tuple */
541 #define	M_HASHTYPE_RSS_TCP_IPV4		M_HASHTYPE_HASH(2) /* TCPv4 4-tuple */
542 #define	M_HASHTYPE_RSS_IPV6		M_HASHTYPE_HASH(3) /* IPv6 2-tuple */
543 #define	M_HASHTYPE_RSS_TCP_IPV6		M_HASHTYPE_HASH(4) /* TCPv6 4-tuple */
544 #define	M_HASHTYPE_RSS_IPV6_EX		M_HASHTYPE_HASH(5) /* IPv6 2-tuple +
545 							    * ext hdrs */
546 #define	M_HASHTYPE_RSS_TCP_IPV6_EX	M_HASHTYPE_HASH(6) /* TCPv6 4-tuple +
547 							    * ext hdrs */
548 #define	M_HASHTYPE_RSS_UDP_IPV4		M_HASHTYPE_HASH(7) /* IPv4 UDP 4-tuple*/
549 #define	M_HASHTYPE_RSS_UDP_IPV6		M_HASHTYPE_HASH(9) /* IPv6 UDP 4-tuple*/
550 #define	M_HASHTYPE_RSS_UDP_IPV6_EX	M_HASHTYPE_HASH(10)/* IPv6 UDP 4-tuple +
551 							    * ext hdrs */
552 
553 #define	M_HASHTYPE_OPAQUE		0x3f	/* ordering, not affinity */
554 #define	M_HASHTYPE_OPAQUE_HASH		M_HASHTYPE_HASH(M_HASHTYPE_OPAQUE)
555 						/* ordering+hash, not affinity*/
556 
557 #define	M_HASHTYPE_CLEAR(m)	((m)->m_pkthdr.rsstype = 0)
558 #define	M_HASHTYPE_GET(m)	((m)->m_pkthdr.rsstype & ~M_HASHTYPE_INNER)
559 #define	M_HASHTYPE_SET(m, v)	((m)->m_pkthdr.rsstype = (v))
560 #define	M_HASHTYPE_TEST(m, v)	(M_HASHTYPE_GET(m) == (v))
561 #define	M_HASHTYPE_ISHASH(m)	\
562     (((m)->m_pkthdr.rsstype & M_HASHTYPE_HASHPROP) != 0)
563 #define	M_HASHTYPE_SETINNER(m)	do {			\
564 	(m)->m_pkthdr.rsstype |= M_HASHTYPE_INNER;	\
565     } while (0)
566 
567 /*
568  * External mbuf storage buffer types.
569  */
570 #define	EXT_CLUSTER	1	/* mbuf cluster */
571 #define	EXT_SFBUF	2	/* sendfile(2)'s sf_buf */
572 #define	EXT_JUMBOP	3	/* jumbo cluster page sized */
573 #define	EXT_JUMBO9	4	/* jumbo cluster 9216 bytes */
574 #define	EXT_JUMBO16	5	/* jumbo cluster 16184 bytes */
575 #define	EXT_PACKET	6	/* mbuf+cluster from packet zone */
576 #define	EXT_MBUF	7	/* external mbuf reference */
577 #define	EXT_RXRING	8	/* data in NIC receive ring */
578 
579 #define	EXT_VENDOR1	224	/* for vendor-internal use */
580 #define	EXT_VENDOR2	225	/* for vendor-internal use */
581 #define	EXT_VENDOR3	226	/* for vendor-internal use */
582 #define	EXT_VENDOR4	227	/* for vendor-internal use */
583 
584 #define	EXT_EXP1	244	/* for experimental use */
585 #define	EXT_EXP2	245	/* for experimental use */
586 #define	EXT_EXP3	246	/* for experimental use */
587 #define	EXT_EXP4	247	/* for experimental use */
588 
589 #define	EXT_NET_DRV	252	/* custom ext_buf provided by net driver(s) */
590 #define	EXT_MOD_TYPE	253	/* custom module's ext_buf type */
591 #define	EXT_DISPOSABLE	254	/* can throw this buffer away w/page flipping */
592 #define	EXT_EXTREF	255	/* has externally maintained ext_cnt ptr */
593 
594 /*
595  * Flags for external mbuf buffer types.
596  * NB: limited to the lower 24 bits.
597  */
598 #define	EXT_FLAG_EMBREF		0x000001	/* embedded ext_count */
599 #define	EXT_FLAG_EXTREF		0x000002	/* external ext_cnt, notyet */
600 
601 #define	EXT_FLAG_NOFREE		0x000010	/* don't free mbuf to pool, notyet */
602 
603 #define	EXT_FLAG_VENDOR1	0x010000	/* These flags are vendor */
604 #define	EXT_FLAG_VENDOR2	0x020000	/* or submodule specific, */
605 #define	EXT_FLAG_VENDOR3	0x040000	/* not used by mbuf code. */
606 #define	EXT_FLAG_VENDOR4	0x080000	/* Set/read by submodule. */
607 
608 #define	EXT_FLAG_EXP1		0x100000	/* for experimental use */
609 #define	EXT_FLAG_EXP2		0x200000	/* for experimental use */
610 #define	EXT_FLAG_EXP3		0x400000	/* for experimental use */
611 #define	EXT_FLAG_EXP4		0x800000	/* for experimental use */
612 
613 /*
614  * EXT flag description for use with printf(9) %b identifier.
615  */
616 #define	EXT_FLAG_BITS \
617     "\20\1EXT_FLAG_EMBREF\2EXT_FLAG_EXTREF\5EXT_FLAG_NOFREE" \
618     "\21EXT_FLAG_VENDOR1\22EXT_FLAG_VENDOR2\23EXT_FLAG_VENDOR3" \
619     "\24EXT_FLAG_VENDOR4\25EXT_FLAG_EXP1\26EXT_FLAG_EXP2\27EXT_FLAG_EXP3" \
620     "\30EXT_FLAG_EXP4"
621 
622 /*
623  * Flags indicating checksum, segmentation and other offload work to be
624  * done, or already done, by hardware or lower layers.  It is split into
625  * separate inbound and outbound flags.
626  *
627  * Outbound flags that are set by upper protocol layers requesting lower
628  * layers, or ideally the hardware, to perform these offloading tasks.
629  * For outbound packets this field and its flags can be directly tested
630  * against ifnet if_hwassist.  Note that the outbound and the inbound flags do
631  * not collide right now but they could be allowed to (as long as the flags are
632  * scrubbed appropriately when the direction of an mbuf changes).  CSUM_BITS
633  * would also have to split into CSUM_BITS_TX and CSUM_BITS_RX.
634  *
635  * CSUM_INNER_<x> is the same as CSUM_<x> but it applies to the inner frame.
636  * The CSUM_ENCAP_<x> bits identify the outer encapsulation.
637  */
638 #define	CSUM_IP			0x00000001	/* IP header checksum offload */
639 #define	CSUM_IP_UDP		0x00000002	/* UDP checksum offload */
640 #define	CSUM_IP_TCP		0x00000004	/* TCP checksum offload */
641 #define	CSUM_IP_SCTP		0x00000008	/* SCTP checksum offload */
642 #define	CSUM_IP_TSO		0x00000010	/* TCP segmentation offload */
643 #define	CSUM_IP_ISCSI		0x00000020	/* iSCSI checksum offload */
644 
645 #define	CSUM_INNER_IP6_UDP	0x00000040
646 #define	CSUM_INNER_IP6_TCP	0x00000080
647 #define	CSUM_INNER_IP6_TSO	0x00000100
648 #define	CSUM_IP6_UDP		0x00000200	/* UDP checksum offload */
649 #define	CSUM_IP6_TCP		0x00000400	/* TCP checksum offload */
650 #define	CSUM_IP6_SCTP		0x00000800	/* SCTP checksum offload */
651 #define	CSUM_IP6_TSO		0x00001000	/* TCP segmentation offload */
652 #define	CSUM_IP6_ISCSI		0x00002000	/* iSCSI checksum offload */
653 
654 #define	CSUM_INNER_IP		0x00004000
655 #define	CSUM_INNER_IP_UDP	0x00008000
656 #define	CSUM_INNER_IP_TCP	0x00010000
657 #define	CSUM_INNER_IP_TSO	0x00020000
658 
659 #define	CSUM_ENCAP_VXLAN	0x00040000	/* VXLAN outer encapsulation */
660 #define	CSUM_ENCAP_RSVD1	0x00080000
661 
662 /* Inbound checksum support where the checksum was verified by hardware. */
663 #define	CSUM_INNER_L3_CALC	0x00100000
664 #define	CSUM_INNER_L3_VALID	0x00200000
665 #define	CSUM_INNER_L4_CALC	0x00400000
666 #define	CSUM_INNER_L4_VALID	0x00800000
667 #define	CSUM_L3_CALC		0x01000000	/* calculated layer 3 csum */
668 #define	CSUM_L3_VALID		0x02000000	/* checksum is correct */
669 #define	CSUM_L4_CALC		0x04000000	/* calculated layer 4 csum */
670 #define	CSUM_L4_VALID		0x08000000	/* checksum is correct */
671 #define	CSUM_L5_CALC		0x10000000	/* calculated layer 5 csum */
672 #define	CSUM_L5_VALID		0x20000000	/* checksum is correct */
673 #define	CSUM_COALESCED		0x40000000	/* contains merged segments */
674 
675 #define	CSUM_SND_TAG		0x80000000	/* Packet header has send tag */
676 
677 #define CSUM_FLAGS_TX (CSUM_IP | CSUM_IP_UDP | CSUM_IP_TCP | CSUM_IP_SCTP | \
678     CSUM_IP_TSO | CSUM_IP_ISCSI | CSUM_INNER_IP6_UDP | CSUM_INNER_IP6_TCP | \
679     CSUM_INNER_IP6_TSO | CSUM_IP6_UDP | CSUM_IP6_TCP | CSUM_IP6_SCTP | \
680     CSUM_IP6_TSO | CSUM_IP6_ISCSI | CSUM_INNER_IP | CSUM_INNER_IP_UDP | \
681     CSUM_INNER_IP_TCP | CSUM_INNER_IP_TSO | CSUM_ENCAP_VXLAN | \
682     CSUM_ENCAP_RSVD1 | CSUM_SND_TAG)
683 
684 #define CSUM_FLAGS_RX (CSUM_INNER_L3_CALC | CSUM_INNER_L3_VALID | \
685     CSUM_INNER_L4_CALC | CSUM_INNER_L4_VALID | CSUM_L3_CALC | CSUM_L3_VALID | \
686     CSUM_L4_CALC | CSUM_L4_VALID | CSUM_L5_CALC | CSUM_L5_VALID | \
687     CSUM_COALESCED)
688 
689 /*
690  * CSUM flag description for use with printf(9) %b identifier.
691  */
692 #define	CSUM_BITS \
693     "\20\1CSUM_IP\2CSUM_IP_UDP\3CSUM_IP_TCP\4CSUM_IP_SCTP\5CSUM_IP_TSO" \
694     "\6CSUM_IP_ISCSI\7CSUM_INNER_IP6_UDP\10CSUM_INNER_IP6_TCP" \
695     "\11CSUM_INNER_IP6_TSO\12CSUM_IP6_UDP\13CSUM_IP6_TCP\14CSUM_IP6_SCTP" \
696     "\15CSUM_IP6_TSO\16CSUM_IP6_ISCSI\17CSUM_INNER_IP\20CSUM_INNER_IP_UDP" \
697     "\21CSUM_INNER_IP_TCP\22CSUM_INNER_IP_TSO\23CSUM_ENCAP_VXLAN" \
698     "\24CSUM_ENCAP_RSVD1\25CSUM_INNER_L3_CALC\26CSUM_INNER_L3_VALID" \
699     "\27CSUM_INNER_L4_CALC\30CSUM_INNER_L4_VALID\31CSUM_L3_CALC" \
700     "\32CSUM_L3_VALID\33CSUM_L4_CALC\34CSUM_L4_VALID\35CSUM_L5_CALC" \
701     "\36CSUM_L5_VALID\37CSUM_COALESCED\40CSUM_SND_TAG"
702 
703 /* CSUM flags compatibility mappings. */
704 #define	CSUM_IP_CHECKED		CSUM_L3_CALC
705 #define	CSUM_IP_VALID		CSUM_L3_VALID
706 #define	CSUM_DATA_VALID		CSUM_L4_VALID
707 #define	CSUM_PSEUDO_HDR		CSUM_L4_CALC
708 #define	CSUM_SCTP_VALID		CSUM_L4_VALID
709 #define	CSUM_DELAY_DATA		(CSUM_TCP|CSUM_UDP)
710 #define	CSUM_DELAY_IP		CSUM_IP		/* Only v4, no v6 IP hdr csum */
711 #define	CSUM_DELAY_DATA_IPV6	(CSUM_TCP_IPV6|CSUM_UDP_IPV6)
712 #define	CSUM_DATA_VALID_IPV6	CSUM_DATA_VALID
713 #define	CSUM_TCP		CSUM_IP_TCP
714 #define	CSUM_UDP		CSUM_IP_UDP
715 #define	CSUM_SCTP		CSUM_IP_SCTP
716 #define	CSUM_TSO		(CSUM_IP_TSO|CSUM_IP6_TSO)
717 #define	CSUM_INNER_TSO		(CSUM_INNER_IP_TSO|CSUM_INNER_IP6_TSO)
718 #define	CSUM_UDP_IPV6		CSUM_IP6_UDP
719 #define	CSUM_TCP_IPV6		CSUM_IP6_TCP
720 #define	CSUM_SCTP_IPV6		CSUM_IP6_SCTP
721 
722 /*
723  * mbuf types describing the content of the mbuf (including external storage).
724  */
725 #define	MT_NOTMBUF	0	/* USED INTERNALLY ONLY! Object is not mbuf */
726 #define	MT_DATA		1	/* dynamic (data) allocation */
727 #define	MT_HEADER	MT_DATA	/* packet header, use M_PKTHDR instead */
728 
729 #define	MT_VENDOR1	4	/* for vendor-internal use */
730 #define	MT_VENDOR2	5	/* for vendor-internal use */
731 #define	MT_VENDOR3	6	/* for vendor-internal use */
732 #define	MT_VENDOR4	7	/* for vendor-internal use */
733 
734 #define	MT_SONAME	8	/* socket name */
735 
736 #define	MT_EXP1		9	/* for experimental use */
737 #define	MT_EXP2		10	/* for experimental use */
738 #define	MT_EXP3		11	/* for experimental use */
739 #define	MT_EXP4		12	/* for experimental use */
740 
741 #define	MT_CONTROL	14	/* extra-data protocol message */
742 #define	MT_EXTCONTROL	15	/* control message with externalized contents */
743 #define	MT_OOBDATA	16	/* expedited data  */
744 
745 #define	MT_NOINIT	255	/* Not a type but a flag to allocate
746 				   a non-initialized mbuf */
747 
748 /*
749  * String names of mbuf-related UMA(9) and malloc(9) types.  Exposed to
750  * !_KERNEL so that monitoring tools can look up the zones with
751  * libmemstat(3).
752  */
753 #define	MBUF_MEM_NAME		"mbuf"
754 #define	MBUF_CLUSTER_MEM_NAME	"mbuf_cluster"
755 #define	MBUF_PACKET_MEM_NAME	"mbuf_packet"
756 #define	MBUF_JUMBOP_MEM_NAME	"mbuf_jumbo_page"
757 #define	MBUF_JUMBO9_MEM_NAME	"mbuf_jumbo_9k"
758 #define	MBUF_JUMBO16_MEM_NAME	"mbuf_jumbo_16k"
759 #define	MBUF_TAG_MEM_NAME	"mbuf_tag"
760 #define	MBUF_EXTREFCNT_MEM_NAME	"mbuf_ext_refcnt"
761 #define	MBUF_EXTPGS_MEM_NAME	"mbuf_extpgs"
762 
763 #ifdef _KERNEL
764 union if_snd_tag_alloc_params;
765 
766 #ifdef WITNESS
767 #define	MBUF_CHECKSLEEP(how) do {					\
768 	if (how == M_WAITOK)						\
769 		WITNESS_WARN(WARN_GIANTOK | WARN_SLEEPOK, NULL,		\
770 		    "Sleeping in \"%s\"", __func__);			\
771 } while (0)
772 #else
773 #define	MBUF_CHECKSLEEP(how)
774 #endif
775 
776 /*
777  * Network buffer allocation API
778  *
779  * The rest of it is defined in kern/kern_mbuf.c
780  */
781 extern uma_zone_t	zone_mbuf;
782 extern uma_zone_t	zone_clust;
783 extern uma_zone_t	zone_pack;
784 extern uma_zone_t	zone_jumbop;
785 extern uma_zone_t	zone_jumbo9;
786 extern uma_zone_t	zone_jumbo16;
787 extern uma_zone_t	zone_extpgs;
788 
789 void		 mb_dupcl(struct mbuf *, struct mbuf *);
790 void		 mb_free_ext(struct mbuf *);
791 void		 mb_free_extpg(struct mbuf *);
792 void		 mb_free_mext_pgs(struct mbuf *);
793 struct mbuf	*mb_alloc_ext_pgs(int, m_ext_free_t);
794 struct mbuf	*mb_alloc_ext_plus_pages(int, int);
795 struct mbuf	*mb_mapped_to_unmapped(struct mbuf *, int, int, int,
796 		    struct mbuf **);
797 int		 mb_unmapped_compress(struct mbuf *m);
798 struct mbuf 	*mb_unmapped_to_ext(struct mbuf *m);
799 void		 mb_free_notready(struct mbuf *m, int count);
800 void		 m_adj(struct mbuf *, int);
801 void		 m_adj_decap(struct mbuf *, int);
802 int		 m_apply(struct mbuf *, int, int,
803 		    int (*)(void *, void *, u_int), void *);
804 int		 m_append(struct mbuf *, int, c_caddr_t);
805 void		 m_cat(struct mbuf *, struct mbuf *);
806 void		 m_catpkt(struct mbuf *, struct mbuf *);
807 int		 m_clget(struct mbuf *m, int how);
808 void 		*m_cljget(struct mbuf *m, int how, int size);
809 struct mbuf	*m_collapse(struct mbuf *, int, int);
810 void		 m_copyback(struct mbuf *, int, int, c_caddr_t);
811 void		 m_copydata(const struct mbuf *, int, int, caddr_t);
812 struct mbuf	*m_copym(struct mbuf *, int, int, int);
813 struct mbuf	*m_copypacket(struct mbuf *, int);
814 void		 m_copy_pkthdr(struct mbuf *, struct mbuf *);
815 struct mbuf	*m_copyup(struct mbuf *, int, int);
816 struct mbuf	*m_defrag(struct mbuf *, int);
817 void		 m_demote_pkthdr(struct mbuf *);
818 void		 m_demote(struct mbuf *, int, int);
819 struct mbuf	*m_devget(char *, int, int, struct ifnet *,
820 		    void (*)(char *, caddr_t, u_int));
821 void		 m_dispose_extcontrolm(struct mbuf *m);
822 struct mbuf	*m_dup(const struct mbuf *, int);
823 int		 m_dup_pkthdr(struct mbuf *, const struct mbuf *, int);
824 void		 m_extadd(struct mbuf *, char *, u_int, m_ext_free_t,
825 		    void *, void *, int, int);
826 u_int		 m_fixhdr(struct mbuf *);
827 struct mbuf	*m_fragment(struct mbuf *, int, int);
828 void		 m_freem(struct mbuf *);
829 struct mbuf	*m_get2(int, int, short, int);
830 struct mbuf	*m_getjcl(int, short, int, int);
831 struct mbuf	*m_getm2(struct mbuf *, int, int, short, int);
832 struct mbuf	*m_getptr(struct mbuf *, int, int *);
833 u_int		 m_length(struct mbuf *, struct mbuf **);
834 int		 m_mbuftouio(struct uio *, const struct mbuf *, int);
835 int		 m_unmappedtouio(const struct mbuf *, int, struct uio *, int);
836 void		 m_move_pkthdr(struct mbuf *, struct mbuf *);
837 int		 m_pkthdr_init(struct mbuf *, int);
838 struct mbuf	*m_prepend(struct mbuf *, int, int);
839 void		 m_print(const struct mbuf *, int);
840 struct mbuf	*m_pulldown(struct mbuf *, int, int, int *);
841 struct mbuf	*m_pullup(struct mbuf *, int);
842 int		 m_sanity(struct mbuf *, int);
843 struct mbuf	*m_split(struct mbuf *, int, int);
844 struct mbuf	*m_uiotombuf(struct uio *, int, int, int, int);
845 struct mbuf	*m_unshare(struct mbuf *, int);
846 int		 m_snd_tag_alloc(struct ifnet *,
847 		    union if_snd_tag_alloc_params *, struct m_snd_tag **);
848 void		 m_snd_tag_init(struct m_snd_tag *, struct ifnet *, u_int);
849 void		 m_snd_tag_destroy(struct m_snd_tag *);
850 
851 static __inline int
852 m_gettype(int size)
853 {
854 	int type;
855 
856 	switch (size) {
857 	case MSIZE:
858 		type = EXT_MBUF;
859 		break;
860 	case MCLBYTES:
861 		type = EXT_CLUSTER;
862 		break;
863 #if MJUMPAGESIZE != MCLBYTES
864 	case MJUMPAGESIZE:
865 		type = EXT_JUMBOP;
866 		break;
867 #endif
868 	case MJUM9BYTES:
869 		type = EXT_JUMBO9;
870 		break;
871 	case MJUM16BYTES:
872 		type = EXT_JUMBO16;
873 		break;
874 	default:
875 		panic("%s: invalid cluster size %d", __func__, size);
876 	}
877 
878 	return (type);
879 }
880 
881 /*
882  * Associated an external reference counted buffer with an mbuf.
883  */
884 static __inline void
885 m_extaddref(struct mbuf *m, char *buf, u_int size, u_int *ref_cnt,
886     m_ext_free_t freef, void *arg1, void *arg2)
887 {
888 
889 	KASSERT(ref_cnt != NULL, ("%s: ref_cnt not provided", __func__));
890 
891 	atomic_add_int(ref_cnt, 1);
892 	m->m_flags |= M_EXT;
893 	m->m_ext.ext_buf = buf;
894 	m->m_ext.ext_cnt = ref_cnt;
895 	m->m_data = m->m_ext.ext_buf;
896 	m->m_ext.ext_size = size;
897 	m->m_ext.ext_free = freef;
898 	m->m_ext.ext_arg1 = arg1;
899 	m->m_ext.ext_arg2 = arg2;
900 	m->m_ext.ext_type = EXT_EXTREF;
901 	m->m_ext.ext_flags = 0;
902 }
903 
904 static __inline uma_zone_t
905 m_getzone(int size)
906 {
907 	uma_zone_t zone;
908 
909 	switch (size) {
910 	case MCLBYTES:
911 		zone = zone_clust;
912 		break;
913 #if MJUMPAGESIZE != MCLBYTES
914 	case MJUMPAGESIZE:
915 		zone = zone_jumbop;
916 		break;
917 #endif
918 	case MJUM9BYTES:
919 		zone = zone_jumbo9;
920 		break;
921 	case MJUM16BYTES:
922 		zone = zone_jumbo16;
923 		break;
924 	default:
925 		panic("%s: invalid cluster size %d", __func__, size);
926 	}
927 
928 	return (zone);
929 }
930 
931 /*
932  * Initialize an mbuf with linear storage.
933  *
934  * Inline because the consumer text overhead will be roughly the same to
935  * initialize or call a function with this many parameters and M_PKTHDR
936  * should go away with constant propagation for !MGETHDR.
937  */
938 static __inline int
939 m_init(struct mbuf *m, int how, short type, int flags)
940 {
941 	int error;
942 
943 	m->m_next = NULL;
944 	m->m_nextpkt = NULL;
945 	m->m_data = m->m_dat;
946 	m->m_len = 0;
947 	m->m_flags = flags;
948 	m->m_type = type;
949 	if (flags & M_PKTHDR)
950 		error = m_pkthdr_init(m, how);
951 	else
952 		error = 0;
953 
954 	MBUF_PROBE5(m__init, m, how, type, flags, error);
955 	return (error);
956 }
957 
958 static __inline struct mbuf *
959 m_get(int how, short type)
960 {
961 	struct mbuf *m;
962 	struct mb_args args;
963 
964 	args.flags = 0;
965 	args.type = type;
966 	m = uma_zalloc_arg(zone_mbuf, &args, how);
967 	MBUF_PROBE3(m__get, how, type, m);
968 	return (m);
969 }
970 
971 static __inline struct mbuf *
972 m_gethdr(int how, short type)
973 {
974 	struct mbuf *m;
975 	struct mb_args args;
976 
977 	args.flags = M_PKTHDR;
978 	args.type = type;
979 	m = uma_zalloc_arg(zone_mbuf, &args, how);
980 	MBUF_PROBE3(m__gethdr, how, type, m);
981 	return (m);
982 }
983 
984 static __inline struct mbuf *
985 m_getcl(int how, short type, int flags)
986 {
987 	struct mbuf *m;
988 	struct mb_args args;
989 
990 	args.flags = flags;
991 	args.type = type;
992 	m = uma_zalloc_arg(zone_pack, &args, how);
993 	MBUF_PROBE4(m__getcl, how, type, flags, m);
994 	return (m);
995 }
996 
997 /*
998  * XXX: m_cljset() is a dangerous API.  One must attach only a new,
999  * unreferenced cluster to an mbuf(9).  It is not possible to assert
1000  * that, so care can be taken only by users of the API.
1001  */
1002 static __inline void
1003 m_cljset(struct mbuf *m, void *cl, int type)
1004 {
1005 	int size;
1006 
1007 	switch (type) {
1008 	case EXT_CLUSTER:
1009 		size = MCLBYTES;
1010 		break;
1011 #if MJUMPAGESIZE != MCLBYTES
1012 	case EXT_JUMBOP:
1013 		size = MJUMPAGESIZE;
1014 		break;
1015 #endif
1016 	case EXT_JUMBO9:
1017 		size = MJUM9BYTES;
1018 		break;
1019 	case EXT_JUMBO16:
1020 		size = MJUM16BYTES;
1021 		break;
1022 	default:
1023 		panic("%s: unknown cluster type %d", __func__, type);
1024 		break;
1025 	}
1026 
1027 	m->m_data = m->m_ext.ext_buf = cl;
1028 	m->m_ext.ext_free = m->m_ext.ext_arg1 = m->m_ext.ext_arg2 = NULL;
1029 	m->m_ext.ext_size = size;
1030 	m->m_ext.ext_type = type;
1031 	m->m_ext.ext_flags = EXT_FLAG_EMBREF;
1032 	m->m_ext.ext_count = 1;
1033 	m->m_flags |= M_EXT;
1034 	MBUF_PROBE3(m__cljset, m, cl, type);
1035 }
1036 
1037 static __inline void
1038 m_chtype(struct mbuf *m, short new_type)
1039 {
1040 
1041 	m->m_type = new_type;
1042 }
1043 
1044 static __inline void
1045 m_clrprotoflags(struct mbuf *m)
1046 {
1047 
1048 	while (m) {
1049 		m->m_flags &= ~M_PROTOFLAGS;
1050 		m = m->m_next;
1051 	}
1052 }
1053 
1054 static __inline struct mbuf *
1055 m_last(struct mbuf *m)
1056 {
1057 
1058 	while (m->m_next)
1059 		m = m->m_next;
1060 	return (m);
1061 }
1062 
1063 static inline u_int
1064 m_extrefcnt(struct mbuf *m)
1065 {
1066 
1067 	KASSERT(m->m_flags & M_EXT, ("%s: M_EXT missing", __func__));
1068 
1069 	return ((m->m_ext.ext_flags & EXT_FLAG_EMBREF) ? m->m_ext.ext_count :
1070 	    *m->m_ext.ext_cnt);
1071 }
1072 
1073 /*
1074  * mbuf, cluster, and external object allocation macros (for compatibility
1075  * purposes).
1076  */
1077 #define	M_MOVE_PKTHDR(to, from)	m_move_pkthdr((to), (from))
1078 #define	MGET(m, how, type)	((m) = m_get((how), (type)))
1079 #define	MGETHDR(m, how, type)	((m) = m_gethdr((how), (type)))
1080 #define	MCLGET(m, how)		m_clget((m), (how))
1081 #define	MEXTADD(m, buf, size, free, arg1, arg2, flags, type)		\
1082     m_extadd((m), (char *)(buf), (size), (free), (arg1), (arg2),	\
1083     (flags), (type))
1084 #define	m_getm(m, len, how, type)					\
1085     m_getm2((m), (len), (how), (type), M_PKTHDR)
1086 
1087 /*
1088  * Evaluate TRUE if it's safe to write to the mbuf m's data region (this can
1089  * be both the local data payload, or an external buffer area, depending on
1090  * whether M_EXT is set).
1091  */
1092 #define	M_WRITABLE(m)	(((m)->m_flags & (M_RDONLY | M_EXTPG)) == 0 &&	\
1093 			 (!(((m)->m_flags & M_EXT)) ||			\
1094 			 (m_extrefcnt(m) == 1)))
1095 
1096 /* Check if the supplied mbuf has a packet header, or else panic. */
1097 #define	M_ASSERTPKTHDR(m)						\
1098 	KASSERT((m) != NULL && (m)->m_flags & M_PKTHDR,			\
1099 	    ("%s: no mbuf packet header!", __func__))
1100 
1101 /* Check if mbuf is multipage. */
1102 #define M_ASSERTEXTPG(m)						\
1103 	KASSERT(((m)->m_flags & (M_EXTPG|M_PKTHDR)) == M_EXTPG,		\
1104 	    ("%s: m %p is not multipage!", __func__, m))
1105 
1106 /*
1107  * Ensure that the supplied mbuf is a valid, non-free mbuf.
1108  *
1109  * XXX: Broken at the moment.  Need some UMA magic to make it work again.
1110  */
1111 #define	M_ASSERTVALID(m)						\
1112 	KASSERT((((struct mbuf *)m)->m_flags & 0) == 0,			\
1113 	    ("%s: attempted use of a free mbuf!", __func__))
1114 
1115 /*
1116  * Return the address of the start of the buffer associated with an mbuf,
1117  * handling external storage, packet-header mbufs, and regular data mbufs.
1118  */
1119 #define	M_START(m)							\
1120 	(((m)->m_flags & M_EXTPG) ? NULL :				\
1121 	 ((m)->m_flags & M_EXT) ? (m)->m_ext.ext_buf :			\
1122 	 ((m)->m_flags & M_PKTHDR) ? &(m)->m_pktdat[0] :		\
1123 	 &(m)->m_dat[0])
1124 
1125 /*
1126  * Return the size of the buffer associated with an mbuf, handling external
1127  * storage, packet-header mbufs, and regular data mbufs.
1128  */
1129 #define	M_SIZE(m)							\
1130 	(((m)->m_flags & M_EXT) ? (m)->m_ext.ext_size :			\
1131 	 ((m)->m_flags & M_PKTHDR) ? MHLEN :				\
1132 	 MLEN)
1133 
1134 /*
1135  * Set the m_data pointer of a newly allocated mbuf to place an object of the
1136  * specified size at the end of the mbuf, longword aligned.
1137  *
1138  * NB: Historically, we had M_ALIGN(), MH_ALIGN(), and MEXT_ALIGN() as
1139  * separate macros, each asserting that it was called at the proper moment.
1140  * This required callers to themselves test the storage type and call the
1141  * right one.  Rather than require callers to be aware of those layout
1142  * decisions, we centralize here.
1143  */
1144 static __inline void
1145 m_align(struct mbuf *m, int len)
1146 {
1147 #ifdef INVARIANTS
1148 	const char *msg = "%s: not a virgin mbuf";
1149 #endif
1150 	int adjust;
1151 
1152 	KASSERT(m->m_data == M_START(m), (msg, __func__));
1153 
1154 	adjust = M_SIZE(m) - len;
1155 	m->m_data += adjust &~ (sizeof(long)-1);
1156 }
1157 
1158 #define	M_ALIGN(m, len)		m_align(m, len)
1159 #define	MH_ALIGN(m, len)	m_align(m, len)
1160 #define	MEXT_ALIGN(m, len)	m_align(m, len)
1161 
1162 /*
1163  * Compute the amount of space available before the current start of data in
1164  * an mbuf.
1165  *
1166  * The M_WRITABLE() is a temporary, conservative safety measure: the burden
1167  * of checking writability of the mbuf data area rests solely with the caller.
1168  *
1169  * NB: In previous versions, M_LEADINGSPACE() would only check M_WRITABLE()
1170  * for mbufs with external storage.  We now allow mbuf-embedded data to be
1171  * read-only as well.
1172  */
1173 #define	M_LEADINGSPACE(m)						\
1174 	(M_WRITABLE(m) ? ((m)->m_data - M_START(m)) : 0)
1175 
1176 /*
1177  * Compute the amount of space available after the end of data in an mbuf.
1178  *
1179  * The M_WRITABLE() is a temporary, conservative safety measure: the burden
1180  * of checking writability of the mbuf data area rests solely with the caller.
1181  *
1182  * NB: In previous versions, M_TRAILINGSPACE() would only check M_WRITABLE()
1183  * for mbufs with external storage.  We now allow mbuf-embedded data to be
1184  * read-only as well.
1185  */
1186 #define	M_TRAILINGSPACE(m)						\
1187 	(M_WRITABLE(m) ?						\
1188 	    ((M_START(m) + M_SIZE(m)) - ((m)->m_data + (m)->m_len)) : 0)
1189 
1190 /*
1191  * Arrange to prepend space of size plen to mbuf m.  If a new mbuf must be
1192  * allocated, how specifies whether to wait.  If the allocation fails, the
1193  * original mbuf chain is freed and m is set to NULL.
1194  */
1195 #define	M_PREPEND(m, plen, how) do {					\
1196 	struct mbuf **_mmp = &(m);					\
1197 	struct mbuf *_mm = *_mmp;					\
1198 	int _mplen = (plen);						\
1199 	int __mhow = (how);						\
1200 									\
1201 	MBUF_CHECKSLEEP(how);						\
1202 	if (M_LEADINGSPACE(_mm) >= _mplen) {				\
1203 		_mm->m_data -= _mplen;					\
1204 		_mm->m_len += _mplen;					\
1205 	} else								\
1206 		_mm = m_prepend(_mm, _mplen, __mhow);			\
1207 	if (_mm != NULL && _mm->m_flags & M_PKTHDR)			\
1208 		_mm->m_pkthdr.len += _mplen;				\
1209 	*_mmp = _mm;							\
1210 } while (0)
1211 
1212 /*
1213  * Change mbuf to new type.  This is a relatively expensive operation and
1214  * should be avoided.
1215  */
1216 #define	MCHTYPE(m, t)	m_chtype((m), (t))
1217 
1218 /* Return the rcvif of a packet header. */
1219 static __inline struct ifnet *
1220 m_rcvif(struct mbuf *m)
1221 {
1222 
1223 	M_ASSERTPKTHDR(m);
1224 	if (m->m_pkthdr.csum_flags & CSUM_SND_TAG)
1225 		return (NULL);
1226 	return (m->m_pkthdr.rcvif);
1227 }
1228 
1229 /* Length to m_copy to copy all. */
1230 #define	M_COPYALL	1000000000
1231 
1232 extern int		max_datalen;	/* MHLEN - max_hdr */
1233 extern int		max_hdr;	/* Largest link + protocol header */
1234 extern int		max_linkhdr;	/* Largest link-level header */
1235 extern int		max_protohdr;	/* Largest protocol header */
1236 extern int		nmbclusters;	/* Maximum number of clusters */
1237 extern bool		mb_use_ext_pgs;	/* Use ext_pgs for sendfile */
1238 
1239 /*-
1240  * Network packets may have annotations attached by affixing a list of
1241  * "packet tags" to the pkthdr structure.  Packet tags are dynamically
1242  * allocated semi-opaque data structures that have a fixed header
1243  * (struct m_tag) that specifies the size of the memory block and a
1244  * <cookie,type> pair that identifies it.  The cookie is a 32-bit unique
1245  * unsigned value used to identify a module or ABI.  By convention this value
1246  * is chosen as the date+time that the module is created, expressed as the
1247  * number of seconds since the epoch (e.g., using date -u +'%s').  The type
1248  * value is an ABI/module-specific value that identifies a particular
1249  * annotation and is private to the module.  For compatibility with systems
1250  * like OpenBSD that define packet tags w/o an ABI/module cookie, the value
1251  * PACKET_ABI_COMPAT is used to implement m_tag_get and m_tag_find
1252  * compatibility shim functions and several tag types are defined below.
1253  * Users that do not require compatibility should use a private cookie value
1254  * so that packet tag-related definitions can be maintained privately.
1255  *
1256  * Note that the packet tag returned by m_tag_alloc has the default memory
1257  * alignment implemented by malloc.  To reference private data one can use a
1258  * construct like:
1259  *
1260  *	struct m_tag *mtag = m_tag_alloc(...);
1261  *	struct foo *p = (struct foo *)(mtag+1);
1262  *
1263  * if the alignment of struct m_tag is sufficient for referencing members of
1264  * struct foo.  Otherwise it is necessary to embed struct m_tag within the
1265  * private data structure to insure proper alignment; e.g.,
1266  *
1267  *	struct foo {
1268  *		struct m_tag	tag;
1269  *		...
1270  *	};
1271  *	struct foo *p = (struct foo *) m_tag_alloc(...);
1272  *	struct m_tag *mtag = &p->tag;
1273  */
1274 
1275 /*
1276  * Persistent tags stay with an mbuf until the mbuf is reclaimed.  Otherwise
1277  * tags are expected to ``vanish'' when they pass through a network
1278  * interface.  For most interfaces this happens normally as the tags are
1279  * reclaimed when the mbuf is free'd.  However in some special cases
1280  * reclaiming must be done manually.  An example is packets that pass through
1281  * the loopback interface.  Also, one must be careful to do this when
1282  * ``turning around'' packets (e.g., icmp_reflect).
1283  *
1284  * To mark a tag persistent bit-or this flag in when defining the tag id.
1285  * The tag will then be treated as described above.
1286  */
1287 #define	MTAG_PERSISTENT				0x800
1288 
1289 #define	PACKET_TAG_NONE				0  /* Nadda */
1290 
1291 /* Packet tags for use with PACKET_ABI_COMPAT. */
1292 #define	PACKET_TAG_IPSEC_IN_DONE		1  /* IPsec applied, in */
1293 #define	PACKET_TAG_IPSEC_OUT_DONE		2  /* IPsec applied, out */
1294 #define	PACKET_TAG_IPSEC_IN_CRYPTO_DONE		3  /* NIC IPsec crypto done */
1295 #define	PACKET_TAG_IPSEC_OUT_CRYPTO_NEEDED	4  /* NIC IPsec crypto req'ed */
1296 #define	PACKET_TAG_IPSEC_IN_COULD_DO_CRYPTO	5  /* NIC notifies IPsec */
1297 #define	PACKET_TAG_IPSEC_PENDING_TDB		6  /* Reminder to do IPsec */
1298 #define	PACKET_TAG_BRIDGE			7  /* Bridge processing done */
1299 #define	PACKET_TAG_GIF				8  /* GIF processing done */
1300 #define	PACKET_TAG_GRE				9  /* GRE processing done */
1301 #define	PACKET_TAG_IN_PACKET_CHECKSUM		10 /* NIC checksumming done */
1302 #define	PACKET_TAG_ENCAP			11 /* Encap.  processing */
1303 #define	PACKET_TAG_IPSEC_SOCKET			12 /* IPSEC socket ref */
1304 #define	PACKET_TAG_IPSEC_HISTORY		13 /* IPSEC history */
1305 #define	PACKET_TAG_IPV6_INPUT			14 /* IPV6 input processing */
1306 #define	PACKET_TAG_DUMMYNET			15 /* dummynet info */
1307 #define	PACKET_TAG_DIVERT			17 /* divert info */
1308 #define	PACKET_TAG_IPFORWARD			18 /* ipforward info */
1309 #define	PACKET_TAG_MACLABEL	(19 | MTAG_PERSISTENT) /* MAC label */
1310 #define	PACKET_TAG_PF		(21 | MTAG_PERSISTENT) /* PF/ALTQ information */
1311 #define	PACKET_TAG_RTSOCKFAM			25 /* rtsock sa family */
1312 #define	PACKET_TAG_IPOPTIONS			27 /* Saved IP options */
1313 #define	PACKET_TAG_CARP				28 /* CARP info */
1314 #define	PACKET_TAG_IPSEC_NAT_T_PORTS		29 /* two uint16_t */
1315 #define	PACKET_TAG_ND_OUTGOING			30 /* ND outgoing */
1316 
1317 /* Specific cookies and tags. */
1318 
1319 /* Packet tag routines. */
1320 struct m_tag	*m_tag_alloc(u_int32_t, int, int, int);
1321 void		 m_tag_delete(struct mbuf *, struct m_tag *);
1322 void		 m_tag_delete_chain(struct mbuf *, struct m_tag *);
1323 void		 m_tag_free_default(struct m_tag *);
1324 struct m_tag	*m_tag_locate(struct mbuf *, u_int32_t, int, struct m_tag *);
1325 struct m_tag	*m_tag_copy(struct m_tag *, int);
1326 int		 m_tag_copy_chain(struct mbuf *, const struct mbuf *, int);
1327 void		 m_tag_delete_nonpersistent(struct mbuf *);
1328 
1329 /*
1330  * Initialize the list of tags associated with an mbuf.
1331  */
1332 static __inline void
1333 m_tag_init(struct mbuf *m)
1334 {
1335 
1336 	SLIST_INIT(&m->m_pkthdr.tags);
1337 }
1338 
1339 /*
1340  * Set up the contents of a tag.  Note that this does not fill in the free
1341  * method; the caller is expected to do that.
1342  *
1343  * XXX probably should be called m_tag_init, but that was already taken.
1344  */
1345 static __inline void
1346 m_tag_setup(struct m_tag *t, u_int32_t cookie, int type, int len)
1347 {
1348 
1349 	t->m_tag_id = type;
1350 	t->m_tag_len = len;
1351 	t->m_tag_cookie = cookie;
1352 }
1353 
1354 /*
1355  * Reclaim resources associated with a tag.
1356  */
1357 static __inline void
1358 m_tag_free(struct m_tag *t)
1359 {
1360 
1361 	(*t->m_tag_free)(t);
1362 }
1363 
1364 /*
1365  * Return the first tag associated with an mbuf.
1366  */
1367 static __inline struct m_tag *
1368 m_tag_first(struct mbuf *m)
1369 {
1370 
1371 	return (SLIST_FIRST(&m->m_pkthdr.tags));
1372 }
1373 
1374 /*
1375  * Return the next tag in the list of tags associated with an mbuf.
1376  */
1377 static __inline struct m_tag *
1378 m_tag_next(struct mbuf *m __unused, struct m_tag *t)
1379 {
1380 
1381 	return (SLIST_NEXT(t, m_tag_link));
1382 }
1383 
1384 /*
1385  * Prepend a tag to the list of tags associated with an mbuf.
1386  */
1387 static __inline void
1388 m_tag_prepend(struct mbuf *m, struct m_tag *t)
1389 {
1390 
1391 	SLIST_INSERT_HEAD(&m->m_pkthdr.tags, t, m_tag_link);
1392 }
1393 
1394 /*
1395  * Unlink a tag from the list of tags associated with an mbuf.
1396  */
1397 static __inline void
1398 m_tag_unlink(struct mbuf *m, struct m_tag *t)
1399 {
1400 
1401 	SLIST_REMOVE(&m->m_pkthdr.tags, t, m_tag, m_tag_link);
1402 }
1403 
1404 /* These are for OpenBSD compatibility. */
1405 #define	MTAG_ABI_COMPAT		0		/* compatibility ABI */
1406 
1407 static __inline struct m_tag *
1408 m_tag_get(int type, int length, int wait)
1409 {
1410 	return (m_tag_alloc(MTAG_ABI_COMPAT, type, length, wait));
1411 }
1412 
1413 static __inline struct m_tag *
1414 m_tag_find(struct mbuf *m, int type, struct m_tag *start)
1415 {
1416 	return (SLIST_EMPTY(&m->m_pkthdr.tags) ? (struct m_tag *)NULL :
1417 	    m_tag_locate(m, MTAG_ABI_COMPAT, type, start));
1418 }
1419 
1420 static inline struct m_snd_tag *
1421 m_snd_tag_ref(struct m_snd_tag *mst)
1422 {
1423 
1424 	refcount_acquire(&mst->refcount);
1425 	return (mst);
1426 }
1427 
1428 static inline void
1429 m_snd_tag_rele(struct m_snd_tag *mst)
1430 {
1431 
1432 	if (refcount_release(&mst->refcount))
1433 		m_snd_tag_destroy(mst);
1434 }
1435 
1436 static __inline struct mbuf *
1437 m_free(struct mbuf *m)
1438 {
1439 	struct mbuf *n = m->m_next;
1440 
1441 	MBUF_PROBE1(m__free, m);
1442 	if ((m->m_flags & (M_PKTHDR|M_NOFREE)) == (M_PKTHDR|M_NOFREE))
1443 		m_tag_delete_chain(m, NULL);
1444 	if (m->m_flags & M_PKTHDR && m->m_pkthdr.csum_flags & CSUM_SND_TAG)
1445 		m_snd_tag_rele(m->m_pkthdr.snd_tag);
1446 	if (m->m_flags & M_EXTPG)
1447 		mb_free_extpg(m);
1448 	else if (m->m_flags & M_EXT)
1449 		mb_free_ext(m);
1450 	else if ((m->m_flags & M_NOFREE) == 0)
1451 		uma_zfree(zone_mbuf, m);
1452 	return (n);
1453 }
1454 
1455 static __inline int
1456 rt_m_getfib(struct mbuf *m)
1457 {
1458 	KASSERT(m->m_flags & M_PKTHDR , ("Attempt to get FIB from non header mbuf."));
1459 	return (m->m_pkthdr.fibnum);
1460 }
1461 
1462 #define M_GETFIB(_m)   rt_m_getfib(_m)
1463 
1464 #define M_SETFIB(_m, _fib) do {						\
1465         KASSERT((_m)->m_flags & M_PKTHDR, ("Attempt to set FIB on non header mbuf."));	\
1466 	((_m)->m_pkthdr.fibnum) = (_fib);				\
1467 } while (0)
1468 
1469 /* flags passed as first argument for "m_xxx_tcpip_hash()" */
1470 #define	MBUF_HASHFLAG_L2	(1 << 2)
1471 #define	MBUF_HASHFLAG_L3	(1 << 3)
1472 #define	MBUF_HASHFLAG_L4	(1 << 4)
1473 
1474 /* mbuf hashing helper routines */
1475 uint32_t	m_ether_tcpip_hash_init(void);
1476 uint32_t	m_ether_tcpip_hash(const uint32_t, const struct mbuf *, uint32_t);
1477 uint32_t	m_infiniband_tcpip_hash_init(void);
1478 uint32_t	m_infiniband_tcpip_hash(const uint32_t, const struct mbuf *, uint32_t);
1479 
1480 #ifdef MBUF_PROFILING
1481  void m_profile(struct mbuf *m);
1482  #define M_PROFILE(m) m_profile(m)
1483 #else
1484  #define M_PROFILE(m)
1485 #endif
1486 
1487 struct mbufq {
1488 	STAILQ_HEAD(, mbuf)	mq_head;
1489 	int			mq_len;
1490 	int			mq_maxlen;
1491 };
1492 
1493 static inline void
1494 mbufq_init(struct mbufq *mq, int maxlen)
1495 {
1496 
1497 	STAILQ_INIT(&mq->mq_head);
1498 	mq->mq_maxlen = maxlen;
1499 	mq->mq_len = 0;
1500 }
1501 
1502 static inline struct mbuf *
1503 mbufq_flush(struct mbufq *mq)
1504 {
1505 	struct mbuf *m;
1506 
1507 	m = STAILQ_FIRST(&mq->mq_head);
1508 	STAILQ_INIT(&mq->mq_head);
1509 	mq->mq_len = 0;
1510 	return (m);
1511 }
1512 
1513 static inline void
1514 mbufq_drain(struct mbufq *mq)
1515 {
1516 	struct mbuf *m, *n;
1517 
1518 	n = mbufq_flush(mq);
1519 	while ((m = n) != NULL) {
1520 		n = STAILQ_NEXT(m, m_stailqpkt);
1521 		m_freem(m);
1522 	}
1523 }
1524 
1525 static inline struct mbuf *
1526 mbufq_first(const struct mbufq *mq)
1527 {
1528 
1529 	return (STAILQ_FIRST(&mq->mq_head));
1530 }
1531 
1532 static inline struct mbuf *
1533 mbufq_last(const struct mbufq *mq)
1534 {
1535 
1536 	return (STAILQ_LAST(&mq->mq_head, mbuf, m_stailqpkt));
1537 }
1538 
1539 static inline int
1540 mbufq_full(const struct mbufq *mq)
1541 {
1542 
1543 	return (mq->mq_maxlen > 0 && mq->mq_len >= mq->mq_maxlen);
1544 }
1545 
1546 static inline int
1547 mbufq_len(const struct mbufq *mq)
1548 {
1549 
1550 	return (mq->mq_len);
1551 }
1552 
1553 static inline int
1554 mbufq_enqueue(struct mbufq *mq, struct mbuf *m)
1555 {
1556 
1557 	if (mbufq_full(mq))
1558 		return (ENOBUFS);
1559 	STAILQ_INSERT_TAIL(&mq->mq_head, m, m_stailqpkt);
1560 	mq->mq_len++;
1561 	return (0);
1562 }
1563 
1564 static inline struct mbuf *
1565 mbufq_dequeue(struct mbufq *mq)
1566 {
1567 	struct mbuf *m;
1568 
1569 	m = STAILQ_FIRST(&mq->mq_head);
1570 	if (m) {
1571 		STAILQ_REMOVE_HEAD(&mq->mq_head, m_stailqpkt);
1572 		m->m_nextpkt = NULL;
1573 		mq->mq_len--;
1574 	}
1575 	return (m);
1576 }
1577 
1578 static inline void
1579 mbufq_prepend(struct mbufq *mq, struct mbuf *m)
1580 {
1581 
1582 	STAILQ_INSERT_HEAD(&mq->mq_head, m, m_stailqpkt);
1583 	mq->mq_len++;
1584 }
1585 
1586 /*
1587  * Note: this doesn't enforce the maximum list size for dst.
1588  */
1589 static inline void
1590 mbufq_concat(struct mbufq *mq_dst, struct mbufq *mq_src)
1591 {
1592 
1593 	mq_dst->mq_len += mq_src->mq_len;
1594 	STAILQ_CONCAT(&mq_dst->mq_head, &mq_src->mq_head);
1595 	mq_src->mq_len = 0;
1596 }
1597 
1598 #ifdef _SYS_TIMESPEC_H_
1599 static inline void
1600 mbuf_tstmp2timespec(struct mbuf *m, struct timespec *ts)
1601 {
1602 
1603 	KASSERT((m->m_flags & M_PKTHDR) != 0, ("mbuf %p no M_PKTHDR", m));
1604 	KASSERT((m->m_flags & (M_TSTMP|M_TSTMP_LRO)) != 0, ("mbuf %p no M_TSTMP or M_TSTMP_LRO", m));
1605 	ts->tv_sec = m->m_pkthdr.rcv_tstmp / 1000000000;
1606 	ts->tv_nsec = m->m_pkthdr.rcv_tstmp % 1000000000;
1607 }
1608 #endif
1609 
1610 #ifdef DEBUGNET
1611 /* Invoked from the debugnet client code. */
1612 void	debugnet_mbuf_drain(void);
1613 void	debugnet_mbuf_start(void);
1614 void	debugnet_mbuf_finish(void);
1615 void	debugnet_mbuf_reinit(int nmbuf, int nclust, int clsize);
1616 #endif
1617 
1618 static inline bool
1619 mbuf_has_tls_session(struct mbuf *m)
1620 {
1621 
1622 	if (m->m_flags & M_EXTPG) {
1623 		if (m->m_epg_tls != NULL) {
1624 			return (true);
1625 		}
1626 	}
1627 	return (false);
1628 }
1629 
1630 #endif /* _KERNEL */
1631 #endif /* !_SYS_MBUF_H_ */
1632