xref: /freebsd/sys/sys/mbuf.h (revision 90ec6a30353aa7caaf995ea50e2e23aa5a099600)
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 };
145 
146 /*
147  * Record/packet header in first mbuf of chain; valid only if M_PKTHDR is set.
148  * Size ILP32: 48
149  *	 LP64: 56
150  * Compile-time assertions in uipc_mbuf.c test these values to ensure that
151  * they are correct.
152  */
153 struct pkthdr {
154 	union {
155 		struct m_snd_tag *snd_tag;	/* send tag, if any */
156 		struct ifnet	*rcvif;		/* rcv interface */
157 	};
158 	SLIST_HEAD(packet_tags, m_tag) tags; /* list of packet tags */
159 	int32_t		 len;		/* total packet length */
160 
161 	/* Layer crossing persistent information. */
162 	uint32_t	 flowid;	/* packet's 4-tuple system */
163 	uint32_t	 csum_flags;	/* checksum and offload features */
164 	uint16_t	 fibnum;	/* this packet should use this fib */
165 	uint8_t		 numa_domain;	/* NUMA domain of recvd pkt */
166 	uint8_t		 rsstype;	/* hash type */
167 	union {
168 		uint64_t	rcv_tstmp;	/* timestamp in ns */
169 		struct {
170 			uint8_t		 l2hlen;	/* layer 2 hdr len */
171 			uint8_t		 l3hlen;	/* layer 3 hdr len */
172 			uint8_t		 l4hlen;	/* layer 4 hdr len */
173 			uint8_t		 l5hlen;	/* layer 5 hdr len */
174 			uint8_t		 inner_l2hlen;
175 			uint8_t		 inner_l3hlen;
176 			uint8_t		 inner_l4hlen;
177 			uint8_t		 inner_l5hlen;
178 		};
179 	};
180 	union {
181 		uint8_t  eight[8];
182 		uint16_t sixteen[4];
183 		uint32_t thirtytwo[2];
184 		uint64_t sixtyfour[1];
185 		uintptr_t unintptr[1];
186 		void	*ptr;
187 	} PH_per;
188 
189 	/* Layer specific non-persistent local storage for reassembly, etc. */
190 	union {
191 		uint8_t  eight[8];
192 		uint16_t sixteen[4];
193 		uint32_t thirtytwo[2];
194 		uint64_t sixtyfour[1];
195 		uintptr_t unintptr[1];
196 		void 	*ptr;
197 	} PH_loc;
198 };
199 #define	ether_vtag	PH_per.sixteen[0]
200 #define	PH_vt		PH_per
201 #define	vt_nrecs	sixteen[0]	  /* mld and v6-ND */
202 #define	tso_segsz	PH_per.sixteen[1] /* inbound after LRO */
203 #define	lro_nsegs	tso_segsz	  /* inbound after LRO */
204 #define	csum_data	PH_per.thirtytwo[1] /* inbound from hardware up */
205 #define lro_len		PH_loc.sixteen[0] /* inbound during LRO (no reassembly) */
206 #define lro_csum	PH_loc.sixteen[1] /* inbound during LRO (no reassembly) */
207 /* Note PH_loc is used during IP reassembly (all 8 bytes as a ptr) */
208 
209 /*
210  * TLS records for TLS 1.0-1.2 can have the following header lengths:
211  * - 5 (AES-CBC with implicit IV)
212  * - 21 (AES-CBC with explicit IV)
213  * - 13 (AES-GCM with 8 byte explicit IV)
214  */
215 #define	MBUF_PEXT_HDR_LEN	23
216 
217 /*
218  * TLS records for TLS 1.0-1.2 can have the following maximum trailer
219  * lengths:
220  * - 16 (AES-GCM)
221  * - 36 (AES-CBC with SHA1 and up to 16 bytes of padding)
222  * - 48 (AES-CBC with SHA2-256 and up to 16 bytes of padding)
223  * - 64 (AES-CBC with SHA2-384 and up to 16 bytes of padding)
224  */
225 #define	MBUF_PEXT_TRAIL_LEN	64
226 
227 #if defined(__LP64__)
228 #define MBUF_PEXT_MAX_PGS (40 / sizeof(vm_paddr_t))
229 #else
230 #define MBUF_PEXT_MAX_PGS (72 / sizeof(vm_paddr_t))
231 #endif
232 
233 #define	MBUF_PEXT_MAX_BYTES						\
234     (MBUF_PEXT_MAX_PGS * PAGE_SIZE + MBUF_PEXT_HDR_LEN + MBUF_PEXT_TRAIL_LEN)
235 
236 struct ktls_session;
237 struct socket;
238 
239 /*
240  * Description of external storage mapped into mbuf; valid only if M_EXT is
241  * set.
242  * Size ILP32: 28
243  *	 LP64: 48
244  * Compile-time assertions in uipc_mbuf.c test these values to ensure that
245  * they are correct.
246  */
247 typedef	void m_ext_free_t(struct mbuf *);
248 struct m_ext {
249 	union {
250 		/*
251 		 * If EXT_FLAG_EMBREF is set, then we use refcount in the
252 		 * mbuf, the 'ext_count' member.  Otherwise, we have a
253 		 * shadow copy and we use pointer 'ext_cnt'.  The original
254 		 * mbuf is responsible to carry the pointer to free routine
255 		 * and its arguments.  They aren't copied into shadows in
256 		 * mb_dupcl() to avoid dereferencing next cachelines.
257 		 */
258 		volatile u_int	 ext_count;
259 		volatile u_int	*ext_cnt;
260 	};
261 	uint32_t	 ext_size;	/* size of buffer, for ext_free */
262 	uint32_t	 ext_type:8,	/* type of external storage */
263 			 ext_flags:24;	/* external storage mbuf flags */
264 	union {
265 		struct {
266 			/*
267 			 * Regular M_EXT mbuf:
268 			 * o ext_buf always points to the external buffer.
269 			 * o ext_free (below) and two optional arguments
270 			 *   ext_arg1 and ext_arg2 store the free context for
271 			 *   the external storage.  They are set only in the
272 			 *   refcount carrying mbuf, the one with
273 			 *   EXT_FLAG_EMBREF flag, with exclusion for
274 			 *   EXT_EXTREF type, where the free context is copied
275 			 *   into all mbufs that use same external storage.
276 			 */
277 			char 	*ext_buf;	/* start of buffer */
278 #define	m_ext_copylen	offsetof(struct m_ext, ext_arg2)
279 			void	*ext_arg2;
280 		};
281 		struct {
282 			/*
283 			 * Multi-page M_EXTPG mbuf:
284 			 * o extpg_pa - page vector.
285 			 * o extpg_trail and extpg_hdr - TLS trailer and
286 			 *   header.
287 			 * Uses ext_free and may also use ext_arg1.
288 			 */
289 			vm_paddr_t	extpg_pa[MBUF_PEXT_MAX_PGS];
290 			char		extpg_trail[MBUF_PEXT_TRAIL_LEN];
291 			char		extpg_hdr[MBUF_PEXT_HDR_LEN];
292 			/* Pretend these 3 fields are part of mbuf itself. */
293 #define	m_epg_pa	m_ext.extpg_pa
294 #define	m_epg_trail	m_ext.extpg_trail
295 #define	m_epg_hdr	m_ext.extpg_hdr
296 #define	m_epg_ext_copylen	offsetof(struct m_ext, ext_free)
297 		};
298 	};
299 	/*
300 	 * Free method and optional argument pointer, both
301 	 * used by M_EXT and M_EXTPG.
302 	 */
303 	m_ext_free_t	*ext_free;
304 	void		*ext_arg1;
305 };
306 
307 /*
308  * The core of the mbuf object along with some shortcut defines for practical
309  * purposes.
310  */
311 struct mbuf {
312 	/*
313 	 * Header present at the beginning of every mbuf.
314 	 * Size ILP32: 24
315 	 *      LP64: 32
316 	 * Compile-time assertions in uipc_mbuf.c test these values to ensure
317 	 * that they are correct.
318 	 */
319 	union {	/* next buffer in chain */
320 		struct mbuf		*m_next;
321 		SLIST_ENTRY(mbuf)	m_slist;
322 		STAILQ_ENTRY(mbuf)	m_stailq;
323 	};
324 	union {	/* next chain in queue/record */
325 		struct mbuf		*m_nextpkt;
326 		SLIST_ENTRY(mbuf)	m_slistpkt;
327 		STAILQ_ENTRY(mbuf)	m_stailqpkt;
328 	};
329 	caddr_t		 m_data;	/* location of data */
330 	int32_t		 m_len;		/* amount of data in this mbuf */
331 	uint32_t	 m_type:8,	/* type of data in this mbuf */
332 			 m_flags:24;	/* flags; see below */
333 #if !defined(__LP64__)
334 	uint32_t	 m_pad;		/* pad for 64bit alignment */
335 #endif
336 
337 	/*
338 	 * A set of optional headers (packet header, external storage header)
339 	 * and internal data storage.  Historically, these arrays were sized
340 	 * to MHLEN (space left after a packet header) and MLEN (space left
341 	 * after only a regular mbuf header); they are now variable size in
342 	 * order to support future work on variable-size mbufs.
343 	 */
344 	union {
345 		struct {
346 			union {
347 				/* M_PKTHDR set. */
348 				struct pkthdr	m_pkthdr;
349 
350 				/* M_EXTPG set.
351 				 * Multi-page M_EXTPG mbuf has its meta data
352 				 * split between the below anonymous structure
353 				 * and m_ext.  It carries vector of pages,
354 				 * optional header and trailer char vectors
355 				 * and pointers to socket/TLS data.
356 				 */
357 #define	m_epg_startcopy		m_epg_npgs
358 #define	m_epg_endcopy		m_epg_stailq
359 				struct {
360 					/* Overall count of pages and count of
361 					 * pages with I/O pending. */
362 					uint8_t	m_epg_npgs;
363 					uint8_t	m_epg_nrdy;
364 					/* TLS header and trailer lengths.
365 					 * The data itself resides in m_ext. */
366 					uint8_t	m_epg_hdrlen;
367 					uint8_t	m_epg_trllen;
368 					/* Offset into 1st page and length of
369 					 * data in the last page. */
370 					uint16_t m_epg_1st_off;
371 					uint16_t m_epg_last_len;
372 					uint8_t	m_epg_flags;
373 #define	EPG_FLAG_ANON	0x1	/* Data can be encrypted in place. */
374 #define	EPG_FLAG_2FREE	0x2	/* Scheduled for free. */
375 					uint8_t	m_epg_record_type;
376 					uint8_t	__spare[2];
377 					int	m_epg_enc_cnt;
378 					struct ktls_session *m_epg_tls;
379 					struct socket	*m_epg_so;
380 					uint64_t	m_epg_seqno;
381 					STAILQ_ENTRY(mbuf) m_epg_stailq;
382 				};
383 			};
384 			union {
385 				/* M_EXT or M_EXTPG set. */
386 				struct m_ext	m_ext;
387 				/* M_PKTHDR set, neither M_EXT nor M_EXTPG. */
388 				char		m_pktdat[0];
389 			};
390 		};
391 		char	m_dat[0];			/* !M_PKTHDR, !M_EXT */
392 	};
393 };
394 
395 #ifdef _KERNEL
396 static inline int
397 m_epg_pagelen(const struct mbuf *m, int pidx, int pgoff)
398 {
399 
400 	KASSERT(pgoff == 0 || pidx == 0,
401 	    ("page %d with non-zero offset %d in %p", pidx, pgoff, m));
402 
403 	if (pidx == m->m_epg_npgs - 1) {
404 		return (m->m_epg_last_len);
405 	} else {
406 		return (PAGE_SIZE - pgoff);
407 	}
408 }
409 
410 #ifdef INVARIANTS
411 #define	MCHECK(ex, msg)	KASSERT((ex),				\
412 	    ("Multi page mbuf %p with " #msg " at %s:%d",	\
413 	    m, __FILE__, __LINE__))
414 /*
415  * NB: This expects a non-empty buffer (npgs > 0 and
416  * last_pg_len > 0).
417  */
418 #define	MBUF_EXT_PGS_ASSERT_SANITY(m)	do {				\
419 	MCHECK(m->m_epg_npgs > 0, "no valid pages");		\
420 	MCHECK(m->m_epg_npgs <= nitems(m->m_epg_pa),		\
421 	    "too many pages");						\
422 	MCHECK(m->m_epg_nrdy <= m->m_epg_npgs,			\
423 	    "too many ready pages");					\
424 	MCHECK(m->m_epg_1st_off < PAGE_SIZE,			\
425 		"too large page offset");				\
426 	MCHECK(m->m_epg_last_len > 0, "zero last page length");	\
427 	MCHECK(m->m_epg_last_len <= PAGE_SIZE,			\
428 	    "too large last page length");				\
429 	if (m->m_epg_npgs == 1)					\
430 		MCHECK(m->m_epg_1st_off +			\
431 		    m->m_epg_last_len <=	 PAGE_SIZE,		\
432 		    "single page too large");				\
433 	MCHECK(m->m_epg_hdrlen <= sizeof(m->m_epg_hdr),		\
434 	    "too large header length");					\
435 	MCHECK(m->m_epg_trllen <= sizeof(m->m_epg_trail),	\
436 	    "too large header length");					\
437 } while (0)
438 #else
439 #define	MBUF_EXT_PGS_ASSERT_SANITY(m)	do {} while (0);
440 #endif
441 #endif
442 
443 /*
444  * mbuf flags of global significance and layer crossing.
445  * Those of only protocol/layer specific significance are to be mapped
446  * to M_PROTO[1-11] and cleared at layer handoff boundaries.
447  * NB: Limited to the lower 24 bits.
448  */
449 #define	M_EXT		0x00000001 /* has associated external storage */
450 #define	M_PKTHDR	0x00000002 /* start of record */
451 #define	M_EOR		0x00000004 /* end of record */
452 #define	M_RDONLY	0x00000008 /* associated data is marked read-only */
453 #define	M_BCAST		0x00000010 /* send/received as link-level broadcast */
454 #define	M_MCAST		0x00000020 /* send/received as link-level multicast */
455 #define	M_PROMISC	0x00000040 /* packet was not for us */
456 #define	M_VLANTAG	0x00000080 /* ether_vtag is valid */
457 #define	M_EXTPG		0x00000100 /* has array of unmapped pages and TLS */
458 #define	M_NOFREE	0x00000200 /* do not free mbuf, embedded in cluster */
459 #define	M_TSTMP		0x00000400 /* rcv_tstmp field is valid */
460 #define	M_TSTMP_HPREC	0x00000800 /* rcv_tstmp is high-prec, typically
461 				      hw-stamped on port (useful for IEEE 1588
462 				      and 802.1AS) */
463 #define M_TSTMP_LRO	0x00001000 /* Time LRO pushed in pkt is valid in (PH_loc) */
464 
465 #define	M_PROTO1	0x00002000 /* protocol-specific */
466 #define	M_PROTO2	0x00004000 /* protocol-specific */
467 #define	M_PROTO3	0x00008000 /* protocol-specific */
468 #define	M_PROTO4	0x00010000 /* protocol-specific */
469 #define	M_PROTO5	0x00020000 /* protocol-specific */
470 #define	M_PROTO6	0x00040000 /* protocol-specific */
471 #define	M_PROTO7	0x00080000 /* protocol-specific */
472 #define	M_PROTO8	0x00100000 /* protocol-specific */
473 #define	M_PROTO9	0x00200000 /* protocol-specific */
474 #define	M_PROTO10	0x00400000 /* protocol-specific */
475 #define	M_PROTO11	0x00800000 /* protocol-specific */
476 
477 #define MB_DTOR_SKIP	0x1	/* don't pollute the cache by touching a freed mbuf */
478 
479 /*
480  * Flags to purge when crossing layers.
481  */
482 #define	M_PROTOFLAGS \
483     (M_PROTO1|M_PROTO2|M_PROTO3|M_PROTO4|M_PROTO5|M_PROTO6|M_PROTO7|M_PROTO8|\
484      M_PROTO9|M_PROTO10|M_PROTO11)
485 
486 /*
487  * Flags preserved when copying m_pkthdr.
488  */
489 #define M_COPYFLAGS \
490     (M_PKTHDR|M_EOR|M_RDONLY|M_BCAST|M_MCAST|M_PROMISC|M_VLANTAG|M_TSTMP| \
491      M_TSTMP_HPREC|M_TSTMP_LRO|M_PROTOFLAGS)
492 
493 /*
494  * Mbuf flag description for use with printf(9) %b identifier.
495  */
496 #define	M_FLAG_BITS \
497     "\20\1M_EXT\2M_PKTHDR\3M_EOR\4M_RDONLY\5M_BCAST\6M_MCAST" \
498     "\7M_PROMISC\10M_VLANTAG\11M_EXTPG\12M_NOFREE\13M_TSTMP\14M_TSTMP_HPREC\15M_TSTMP_LRO"
499 #define	M_FLAG_PROTOBITS \
500     "\16M_PROTO1\17M_PROTO2\20M_PROTO3\21M_PROTO4" \
501     "\22M_PROTO5\23M_PROTO6\24M_PROTO7\25M_PROTO8\26M_PROTO9" \
502     "\27M_PROTO10\28M_PROTO11"
503 #define	M_FLAG_PRINTF (M_FLAG_BITS M_FLAG_PROTOBITS)
504 
505 /*
506  * Network interface cards are able to hash protocol fields (such as IPv4
507  * addresses and TCP port numbers) classify packets into flows.  These flows
508  * can then be used to maintain ordering while delivering packets to the OS
509  * via parallel input queues, as well as to provide a stateless affinity
510  * model.  NIC drivers can pass up the hash via m->m_pkthdr.flowid, and set
511  * m_flag fields to indicate how the hash should be interpreted by the
512  * network stack.
513  *
514  * Most NICs support RSS, which provides ordering and explicit affinity, and
515  * use the hash m_flag bits to indicate what header fields were covered by
516  * the hash.  M_HASHTYPE_OPAQUE and M_HASHTYPE_OPAQUE_HASH can be set by non-
517  * RSS cards or configurations that provide an opaque flow identifier, allowing
518  * for ordering and distribution without explicit affinity.  Additionally,
519  * M_HASHTYPE_OPAQUE_HASH indicates that the flow identifier has hash
520  * properties.
521  *
522  * The meaning of the IPV6_EX suffix:
523  * "o  Home address from the home address option in the IPv6 destination
524  *     options header.  If the extension header is not present, use the Source
525  *     IPv6 Address.
526  *  o  IPv6 address that is contained in the Routing-Header-Type-2 from the
527  *     associated extension header.  If the extension header is not present,
528  *     use the Destination IPv6 Address."
529  * Quoted from:
530  * https://docs.microsoft.com/en-us/windows-hardware/drivers/network/rss-hashing-types#ndishashipv6ex
531  */
532 #define	M_HASHTYPE_HASHPROP		0x80	/* has hash properties */
533 #define	M_HASHTYPE_HASH(t)		(M_HASHTYPE_HASHPROP | (t))
534 /* Microsoft RSS standard hash types */
535 #define	M_HASHTYPE_NONE			0
536 #define	M_HASHTYPE_RSS_IPV4		M_HASHTYPE_HASH(1) /* IPv4 2-tuple */
537 #define	M_HASHTYPE_RSS_TCP_IPV4		M_HASHTYPE_HASH(2) /* TCPv4 4-tuple */
538 #define	M_HASHTYPE_RSS_IPV6		M_HASHTYPE_HASH(3) /* IPv6 2-tuple */
539 #define	M_HASHTYPE_RSS_TCP_IPV6		M_HASHTYPE_HASH(4) /* TCPv6 4-tuple */
540 #define	M_HASHTYPE_RSS_IPV6_EX		M_HASHTYPE_HASH(5) /* IPv6 2-tuple +
541 							    * ext hdrs */
542 #define	M_HASHTYPE_RSS_TCP_IPV6_EX	M_HASHTYPE_HASH(6) /* TCPv6 4-tuple +
543 							    * ext hdrs */
544 #define	M_HASHTYPE_RSS_UDP_IPV4		M_HASHTYPE_HASH(7) /* IPv4 UDP 4-tuple*/
545 #define	M_HASHTYPE_RSS_UDP_IPV6		M_HASHTYPE_HASH(9) /* IPv6 UDP 4-tuple*/
546 #define	M_HASHTYPE_RSS_UDP_IPV6_EX	M_HASHTYPE_HASH(10)/* IPv6 UDP 4-tuple +
547 							    * ext hdrs */
548 
549 #define	M_HASHTYPE_OPAQUE		63	/* ordering, not affinity */
550 #define	M_HASHTYPE_OPAQUE_HASH		M_HASHTYPE_HASH(M_HASHTYPE_OPAQUE)
551 						/* ordering+hash, not affinity*/
552 
553 #define	M_HASHTYPE_CLEAR(m)	((m)->m_pkthdr.rsstype = 0)
554 #define	M_HASHTYPE_GET(m)	((m)->m_pkthdr.rsstype)
555 #define	M_HASHTYPE_SET(m, v)	((m)->m_pkthdr.rsstype = (v))
556 #define	M_HASHTYPE_TEST(m, v)	(M_HASHTYPE_GET(m) == (v))
557 #define	M_HASHTYPE_ISHASH(m)	(M_HASHTYPE_GET(m) & M_HASHTYPE_HASHPROP)
558 
559 /*
560  * External mbuf storage buffer types.
561  */
562 #define	EXT_CLUSTER	1	/* mbuf cluster */
563 #define	EXT_SFBUF	2	/* sendfile(2)'s sf_buf */
564 #define	EXT_JUMBOP	3	/* jumbo cluster page sized */
565 #define	EXT_JUMBO9	4	/* jumbo cluster 9216 bytes */
566 #define	EXT_JUMBO16	5	/* jumbo cluster 16184 bytes */
567 #define	EXT_PACKET	6	/* mbuf+cluster from packet zone */
568 #define	EXT_MBUF	7	/* external mbuf reference */
569 #define	EXT_RXRING	8	/* data in NIC receive ring */
570 
571 #define	EXT_VENDOR1	224	/* for vendor-internal use */
572 #define	EXT_VENDOR2	225	/* for vendor-internal use */
573 #define	EXT_VENDOR3	226	/* for vendor-internal use */
574 #define	EXT_VENDOR4	227	/* for vendor-internal use */
575 
576 #define	EXT_EXP1	244	/* for experimental use */
577 #define	EXT_EXP2	245	/* for experimental use */
578 #define	EXT_EXP3	246	/* for experimental use */
579 #define	EXT_EXP4	247	/* for experimental use */
580 
581 #define	EXT_NET_DRV	252	/* custom ext_buf provided by net driver(s) */
582 #define	EXT_MOD_TYPE	253	/* custom module's ext_buf type */
583 #define	EXT_DISPOSABLE	254	/* can throw this buffer away w/page flipping */
584 #define	EXT_EXTREF	255	/* has externally maintained ext_cnt ptr */
585 
586 /*
587  * Flags for external mbuf buffer types.
588  * NB: limited to the lower 24 bits.
589  */
590 #define	EXT_FLAG_EMBREF		0x000001	/* embedded ext_count */
591 #define	EXT_FLAG_EXTREF		0x000002	/* external ext_cnt, notyet */
592 
593 #define	EXT_FLAG_NOFREE		0x000010	/* don't free mbuf to pool, notyet */
594 
595 #define	EXT_FLAG_VENDOR1	0x010000	/* These flags are vendor */
596 #define	EXT_FLAG_VENDOR2	0x020000	/* or submodule specific, */
597 #define	EXT_FLAG_VENDOR3	0x040000	/* not used by mbuf code. */
598 #define	EXT_FLAG_VENDOR4	0x080000	/* Set/read by submodule. */
599 
600 #define	EXT_FLAG_EXP1		0x100000	/* for experimental use */
601 #define	EXT_FLAG_EXP2		0x200000	/* for experimental use */
602 #define	EXT_FLAG_EXP3		0x400000	/* for experimental use */
603 #define	EXT_FLAG_EXP4		0x800000	/* for experimental use */
604 
605 /*
606  * EXT flag description for use with printf(9) %b identifier.
607  */
608 #define	EXT_FLAG_BITS \
609     "\20\1EXT_FLAG_EMBREF\2EXT_FLAG_EXTREF\5EXT_FLAG_NOFREE" \
610     "\21EXT_FLAG_VENDOR1\22EXT_FLAG_VENDOR2\23EXT_FLAG_VENDOR3" \
611     "\24EXT_FLAG_VENDOR4\25EXT_FLAG_EXP1\26EXT_FLAG_EXP2\27EXT_FLAG_EXP3" \
612     "\30EXT_FLAG_EXP4"
613 
614 /*
615  * Flags indicating checksum, segmentation and other offload work to be
616  * done, or already done, by hardware or lower layers.  It is split into
617  * separate inbound and outbound flags.
618  *
619  * Outbound flags that are set by upper protocol layers requesting lower
620  * layers, or ideally the hardware, to perform these offloading tasks.
621  * For outbound packets this field and its flags can be directly tested
622  * against ifnet if_hwassist.  Note that the outbound and the inbound flags do
623  * not collide right now but they could be allowed to (as long as the flags are
624  * scrubbed appropriately when the direction of an mbuf changes).  CSUM_BITS
625  * would also have to split into CSUM_BITS_TX and CSUM_BITS_RX.
626  *
627  * CSUM_INNER_<x> is the same as CSUM_<x> but it applies to the inner frame.
628  * The CSUM_ENCAP_<x> bits identify the outer encapsulation.
629  */
630 #define	CSUM_IP			0x00000001	/* IP header checksum offload */
631 #define	CSUM_IP_UDP		0x00000002	/* UDP checksum offload */
632 #define	CSUM_IP_TCP		0x00000004	/* TCP checksum offload */
633 #define	CSUM_IP_SCTP		0x00000008	/* SCTP checksum offload */
634 #define	CSUM_IP_TSO		0x00000010	/* TCP segmentation offload */
635 #define	CSUM_IP_ISCSI		0x00000020	/* iSCSI checksum offload */
636 
637 #define	CSUM_INNER_IP6_UDP	0x00000040
638 #define	CSUM_INNER_IP6_TCP	0x00000080
639 #define	CSUM_INNER_IP6_TSO	0x00000100
640 #define	CSUM_IP6_UDP		0x00000200	/* UDP checksum offload */
641 #define	CSUM_IP6_TCP		0x00000400	/* TCP checksum offload */
642 #define	CSUM_IP6_SCTP		0x00000800	/* SCTP checksum offload */
643 #define	CSUM_IP6_TSO		0x00001000	/* TCP segmentation offload */
644 #define	CSUM_IP6_ISCSI		0x00002000	/* iSCSI checksum offload */
645 
646 #define	CSUM_INNER_IP		0x00004000
647 #define	CSUM_INNER_IP_UDP	0x00008000
648 #define	CSUM_INNER_IP_TCP	0x00010000
649 #define	CSUM_INNER_IP_TSO	0x00020000
650 
651 #define	CSUM_ENCAP_VXLAN	0x00040000	/* VXLAN outer encapsulation */
652 #define	CSUM_ENCAP_RSVD1	0x00080000
653 
654 /* Inbound checksum support where the checksum was verified by hardware. */
655 #define	CSUM_INNER_L3_CALC	0x00100000
656 #define	CSUM_INNER_L3_VALID	0x00200000
657 #define	CSUM_INNER_L4_CALC	0x00400000
658 #define	CSUM_INNER_L4_VALID	0x00800000
659 #define	CSUM_L3_CALC		0x01000000	/* calculated layer 3 csum */
660 #define	CSUM_L3_VALID		0x02000000	/* checksum is correct */
661 #define	CSUM_L4_CALC		0x04000000	/* calculated layer 4 csum */
662 #define	CSUM_L4_VALID		0x08000000	/* checksum is correct */
663 #define	CSUM_L5_CALC		0x10000000	/* calculated layer 5 csum */
664 #define	CSUM_L5_VALID		0x20000000	/* checksum is correct */
665 #define	CSUM_COALESCED		0x40000000	/* contains merged segments */
666 
667 #define	CSUM_SND_TAG		0x80000000	/* Packet header has send tag */
668 
669 #define CSUM_FLAGS_TX (CSUM_IP | CSUM_IP_UDP | CSUM_IP_TCP | CSUM_IP_SCTP | \
670     CSUM_IP_TSO | CSUM_IP_ISCSI | CSUM_INNER_IP6_UDP | CSUM_INNER_IP6_TCP | \
671     CSUM_INNER_IP6_TSO | CSUM_IP6_UDP | CSUM_IP6_TCP | CSUM_IP6_SCTP | \
672     CSUM_IP6_TSO | CSUM_IP6_ISCSI | CSUM_INNER_IP | CSUM_INNER_IP_UDP | \
673     CSUM_INNER_IP_TCP | CSUM_INNER_IP_TSO | CSUM_ENCAP_VXLAN | \
674     CSUM_ENCAP_RSVD1 | CSUM_SND_TAG)
675 
676 #define CSUM_FLAGS_RX (CSUM_INNER_L3_CALC | CSUM_INNER_L3_VALID | \
677     CSUM_INNER_L4_CALC | CSUM_INNER_L4_VALID | CSUM_L3_CALC | CSUM_L3_VALID | \
678     CSUM_L4_CALC | CSUM_L4_VALID | CSUM_L5_CALC | CSUM_L5_VALID | \
679     CSUM_COALESCED)
680 
681 /*
682  * CSUM flag description for use with printf(9) %b identifier.
683  */
684 #define	CSUM_BITS \
685     "\20\1CSUM_IP\2CSUM_IP_UDP\3CSUM_IP_TCP\4CSUM_IP_SCTP\5CSUM_IP_TSO" \
686     "\6CSUM_IP_ISCSI\7CSUM_INNER_IP6_UDP\10CSUM_INNER_IP6_TCP" \
687     "\11CSUM_INNER_IP6_TSO\12CSUM_IP6_UDP\13CSUM_IP6_TCP\14CSUM_IP6_SCTP" \
688     "\15CSUM_IP6_TSO\16CSUM_IP6_ISCSI\17CSUM_INNER_IP\20CSUM_INNER_IP_UDP" \
689     "\21CSUM_INNER_IP_TCP\22CSUM_INNER_IP_TSO\23CSUM_ENCAP_VXLAN" \
690     "\24CSUM_ENCAP_RSVD1\25CSUM_INNER_L3_CALC\26CSUM_INNER_L3_VALID" \
691     "\27CSUM_INNER_L4_CALC\30CSUM_INNER_L4_VALID\31CSUM_L3_CALC" \
692     "\32CSUM_L3_VALID\33CSUM_L4_CALC\34CSUM_L4_VALID\35CSUM_L5_CALC" \
693     "\36CSUM_L5_VALID\37CSUM_COALESCED\40CSUM_SND_TAG"
694 
695 /* CSUM flags compatibility mappings. */
696 #define	CSUM_IP_CHECKED		CSUM_L3_CALC
697 #define	CSUM_IP_VALID		CSUM_L3_VALID
698 #define	CSUM_DATA_VALID		CSUM_L4_VALID
699 #define	CSUM_PSEUDO_HDR		CSUM_L4_CALC
700 #define	CSUM_SCTP_VALID		CSUM_L4_VALID
701 #define	CSUM_DELAY_DATA		(CSUM_TCP|CSUM_UDP)
702 #define	CSUM_DELAY_IP		CSUM_IP		/* Only v4, no v6 IP hdr csum */
703 #define	CSUM_DELAY_DATA_IPV6	(CSUM_TCP_IPV6|CSUM_UDP_IPV6)
704 #define	CSUM_DATA_VALID_IPV6	CSUM_DATA_VALID
705 #define	CSUM_TCP		CSUM_IP_TCP
706 #define	CSUM_UDP		CSUM_IP_UDP
707 #define	CSUM_SCTP		CSUM_IP_SCTP
708 #define	CSUM_TSO		(CSUM_IP_TSO|CSUM_IP6_TSO)
709 #define	CSUM_INNER_TSO		(CSUM_INNER_IP_TSO|CSUM_INNER_IP6_TSO)
710 #define	CSUM_UDP_IPV6		CSUM_IP6_UDP
711 #define	CSUM_TCP_IPV6		CSUM_IP6_TCP
712 #define	CSUM_SCTP_IPV6		CSUM_IP6_SCTP
713 
714 /*
715  * mbuf types describing the content of the mbuf (including external storage).
716  */
717 #define	MT_NOTMBUF	0	/* USED INTERNALLY ONLY! Object is not mbuf */
718 #define	MT_DATA		1	/* dynamic (data) allocation */
719 #define	MT_HEADER	MT_DATA	/* packet header, use M_PKTHDR instead */
720 
721 #define	MT_VENDOR1	4	/* for vendor-internal use */
722 #define	MT_VENDOR2	5	/* for vendor-internal use */
723 #define	MT_VENDOR3	6	/* for vendor-internal use */
724 #define	MT_VENDOR4	7	/* for vendor-internal use */
725 
726 #define	MT_SONAME	8	/* socket name */
727 
728 #define	MT_EXP1		9	/* for experimental use */
729 #define	MT_EXP2		10	/* for experimental use */
730 #define	MT_EXP3		11	/* for experimental use */
731 #define	MT_EXP4		12	/* for experimental use */
732 
733 #define	MT_CONTROL	14	/* extra-data protocol message */
734 #define	MT_EXTCONTROL	15	/* control message with externalized contents */
735 #define	MT_OOBDATA	16	/* expedited data  */
736 
737 #define	MT_NOINIT	255	/* Not a type but a flag to allocate
738 				   a non-initialized mbuf */
739 
740 /*
741  * String names of mbuf-related UMA(9) and malloc(9) types.  Exposed to
742  * !_KERNEL so that monitoring tools can look up the zones with
743  * libmemstat(3).
744  */
745 #define	MBUF_MEM_NAME		"mbuf"
746 #define	MBUF_CLUSTER_MEM_NAME	"mbuf_cluster"
747 #define	MBUF_PACKET_MEM_NAME	"mbuf_packet"
748 #define	MBUF_JUMBOP_MEM_NAME	"mbuf_jumbo_page"
749 #define	MBUF_JUMBO9_MEM_NAME	"mbuf_jumbo_9k"
750 #define	MBUF_JUMBO16_MEM_NAME	"mbuf_jumbo_16k"
751 #define	MBUF_TAG_MEM_NAME	"mbuf_tag"
752 #define	MBUF_EXTREFCNT_MEM_NAME	"mbuf_ext_refcnt"
753 #define	MBUF_EXTPGS_MEM_NAME	"mbuf_extpgs"
754 
755 #ifdef _KERNEL
756 
757 #ifdef WITNESS
758 #define	MBUF_CHECKSLEEP(how) do {					\
759 	if (how == M_WAITOK)						\
760 		WITNESS_WARN(WARN_GIANTOK | WARN_SLEEPOK, NULL,		\
761 		    "Sleeping in \"%s\"", __func__);			\
762 } while (0)
763 #else
764 #define	MBUF_CHECKSLEEP(how)
765 #endif
766 
767 /*
768  * Network buffer allocation API
769  *
770  * The rest of it is defined in kern/kern_mbuf.c
771  */
772 extern uma_zone_t	zone_mbuf;
773 extern uma_zone_t	zone_clust;
774 extern uma_zone_t	zone_pack;
775 extern uma_zone_t	zone_jumbop;
776 extern uma_zone_t	zone_jumbo9;
777 extern uma_zone_t	zone_jumbo16;
778 extern uma_zone_t	zone_extpgs;
779 
780 void		 mb_dupcl(struct mbuf *, struct mbuf *);
781 void		 mb_free_ext(struct mbuf *);
782 void		 mb_free_extpg(struct mbuf *);
783 void		 mb_free_mext_pgs(struct mbuf *);
784 struct mbuf	*mb_alloc_ext_pgs(int, m_ext_free_t);
785 struct mbuf	*mb_alloc_ext_plus_pages(int, int);
786 struct mbuf	*mb_mapped_to_unmapped(struct mbuf *, int, int, int,
787 		    struct mbuf **);
788 int		 mb_unmapped_compress(struct mbuf *m);
789 struct mbuf 	*mb_unmapped_to_ext(struct mbuf *m);
790 void		 mb_free_notready(struct mbuf *m, int count);
791 void		 m_adj(struct mbuf *, int);
792 int		 m_apply(struct mbuf *, int, int,
793 		    int (*)(void *, void *, u_int), void *);
794 int		 m_append(struct mbuf *, int, c_caddr_t);
795 void		 m_cat(struct mbuf *, struct mbuf *);
796 void		 m_catpkt(struct mbuf *, struct mbuf *);
797 int		 m_clget(struct mbuf *m, int how);
798 void 		*m_cljget(struct mbuf *m, int how, int size);
799 struct mbuf	*m_collapse(struct mbuf *, int, int);
800 void		 m_copyback(struct mbuf *, int, int, c_caddr_t);
801 void		 m_copydata(const struct mbuf *, int, int, caddr_t);
802 struct mbuf	*m_copym(struct mbuf *, int, int, int);
803 struct mbuf	*m_copypacket(struct mbuf *, int);
804 void		 m_copy_pkthdr(struct mbuf *, struct mbuf *);
805 struct mbuf	*m_copyup(struct mbuf *, int, int);
806 struct mbuf	*m_defrag(struct mbuf *, int);
807 void		 m_demote_pkthdr(struct mbuf *);
808 void		 m_demote(struct mbuf *, int, int);
809 struct mbuf	*m_devget(char *, int, int, struct ifnet *,
810 		    void (*)(char *, caddr_t, u_int));
811 void		 m_dispose_extcontrolm(struct mbuf *m);
812 struct mbuf	*m_dup(const struct mbuf *, int);
813 int		 m_dup_pkthdr(struct mbuf *, const struct mbuf *, int);
814 void		 m_extadd(struct mbuf *, char *, u_int, m_ext_free_t,
815 		    void *, void *, int, int);
816 u_int		 m_fixhdr(struct mbuf *);
817 struct mbuf	*m_fragment(struct mbuf *, int, int);
818 void		 m_freem(struct mbuf *);
819 struct mbuf	*m_get2(int, int, short, int);
820 struct mbuf	*m_getjcl(int, short, int, int);
821 struct mbuf	*m_getm2(struct mbuf *, int, int, short, int);
822 struct mbuf	*m_getptr(struct mbuf *, int, int *);
823 u_int		 m_length(struct mbuf *, struct mbuf **);
824 int		 m_mbuftouio(struct uio *, const struct mbuf *, int);
825 int		 m_unmappedtouio(const struct mbuf *, int, struct uio *, int);
826 void		 m_move_pkthdr(struct mbuf *, struct mbuf *);
827 int		 m_pkthdr_init(struct mbuf *, int);
828 struct mbuf	*m_prepend(struct mbuf *, int, int);
829 void		 m_print(const struct mbuf *, int);
830 struct mbuf	*m_pulldown(struct mbuf *, int, int, int *);
831 struct mbuf	*m_pullup(struct mbuf *, int);
832 int		 m_sanity(struct mbuf *, int);
833 struct mbuf	*m_split(struct mbuf *, int, int);
834 struct mbuf	*m_uiotombuf(struct uio *, int, int, int, int);
835 struct mbuf	*m_unshare(struct mbuf *, int);
836 void		 m_snd_tag_init(struct m_snd_tag *, struct ifnet *);
837 void		 m_snd_tag_destroy(struct m_snd_tag *);
838 
839 static __inline int
840 m_gettype(int size)
841 {
842 	int type;
843 
844 	switch (size) {
845 	case MSIZE:
846 		type = EXT_MBUF;
847 		break;
848 	case MCLBYTES:
849 		type = EXT_CLUSTER;
850 		break;
851 #if MJUMPAGESIZE != MCLBYTES
852 	case MJUMPAGESIZE:
853 		type = EXT_JUMBOP;
854 		break;
855 #endif
856 	case MJUM9BYTES:
857 		type = EXT_JUMBO9;
858 		break;
859 	case MJUM16BYTES:
860 		type = EXT_JUMBO16;
861 		break;
862 	default:
863 		panic("%s: invalid cluster size %d", __func__, size);
864 	}
865 
866 	return (type);
867 }
868 
869 /*
870  * Associated an external reference counted buffer with an mbuf.
871  */
872 static __inline void
873 m_extaddref(struct mbuf *m, char *buf, u_int size, u_int *ref_cnt,
874     m_ext_free_t freef, void *arg1, void *arg2)
875 {
876 
877 	KASSERT(ref_cnt != NULL, ("%s: ref_cnt not provided", __func__));
878 
879 	atomic_add_int(ref_cnt, 1);
880 	m->m_flags |= M_EXT;
881 	m->m_ext.ext_buf = buf;
882 	m->m_ext.ext_cnt = ref_cnt;
883 	m->m_data = m->m_ext.ext_buf;
884 	m->m_ext.ext_size = size;
885 	m->m_ext.ext_free = freef;
886 	m->m_ext.ext_arg1 = arg1;
887 	m->m_ext.ext_arg2 = arg2;
888 	m->m_ext.ext_type = EXT_EXTREF;
889 	m->m_ext.ext_flags = 0;
890 }
891 
892 static __inline uma_zone_t
893 m_getzone(int size)
894 {
895 	uma_zone_t zone;
896 
897 	switch (size) {
898 	case MCLBYTES:
899 		zone = zone_clust;
900 		break;
901 #if MJUMPAGESIZE != MCLBYTES
902 	case MJUMPAGESIZE:
903 		zone = zone_jumbop;
904 		break;
905 #endif
906 	case MJUM9BYTES:
907 		zone = zone_jumbo9;
908 		break;
909 	case MJUM16BYTES:
910 		zone = zone_jumbo16;
911 		break;
912 	default:
913 		panic("%s: invalid cluster size %d", __func__, size);
914 	}
915 
916 	return (zone);
917 }
918 
919 /*
920  * Initialize an mbuf with linear storage.
921  *
922  * Inline because the consumer text overhead will be roughly the same to
923  * initialize or call a function with this many parameters and M_PKTHDR
924  * should go away with constant propagation for !MGETHDR.
925  */
926 static __inline int
927 m_init(struct mbuf *m, int how, short type, int flags)
928 {
929 	int error;
930 
931 	m->m_next = NULL;
932 	m->m_nextpkt = NULL;
933 	m->m_data = m->m_dat;
934 	m->m_len = 0;
935 	m->m_flags = flags;
936 	m->m_type = type;
937 	if (flags & M_PKTHDR)
938 		error = m_pkthdr_init(m, how);
939 	else
940 		error = 0;
941 
942 	MBUF_PROBE5(m__init, m, how, type, flags, error);
943 	return (error);
944 }
945 
946 static __inline struct mbuf *
947 m_get(int how, short type)
948 {
949 	struct mbuf *m;
950 	struct mb_args args;
951 
952 	args.flags = 0;
953 	args.type = type;
954 	m = uma_zalloc_arg(zone_mbuf, &args, how);
955 	MBUF_PROBE3(m__get, how, type, m);
956 	return (m);
957 }
958 
959 static __inline struct mbuf *
960 m_gethdr(int how, short type)
961 {
962 	struct mbuf *m;
963 	struct mb_args args;
964 
965 	args.flags = M_PKTHDR;
966 	args.type = type;
967 	m = uma_zalloc_arg(zone_mbuf, &args, how);
968 	MBUF_PROBE3(m__gethdr, how, type, m);
969 	return (m);
970 }
971 
972 static __inline struct mbuf *
973 m_getcl(int how, short type, int flags)
974 {
975 	struct mbuf *m;
976 	struct mb_args args;
977 
978 	args.flags = flags;
979 	args.type = type;
980 	m = uma_zalloc_arg(zone_pack, &args, how);
981 	MBUF_PROBE4(m__getcl, how, type, flags, m);
982 	return (m);
983 }
984 
985 /*
986  * XXX: m_cljset() is a dangerous API.  One must attach only a new,
987  * unreferenced cluster to an mbuf(9).  It is not possible to assert
988  * that, so care can be taken only by users of the API.
989  */
990 static __inline void
991 m_cljset(struct mbuf *m, void *cl, int type)
992 {
993 	int size;
994 
995 	switch (type) {
996 	case EXT_CLUSTER:
997 		size = MCLBYTES;
998 		break;
999 #if MJUMPAGESIZE != MCLBYTES
1000 	case EXT_JUMBOP:
1001 		size = MJUMPAGESIZE;
1002 		break;
1003 #endif
1004 	case EXT_JUMBO9:
1005 		size = MJUM9BYTES;
1006 		break;
1007 	case EXT_JUMBO16:
1008 		size = MJUM16BYTES;
1009 		break;
1010 	default:
1011 		panic("%s: unknown cluster type %d", __func__, type);
1012 		break;
1013 	}
1014 
1015 	m->m_data = m->m_ext.ext_buf = cl;
1016 	m->m_ext.ext_free = m->m_ext.ext_arg1 = m->m_ext.ext_arg2 = NULL;
1017 	m->m_ext.ext_size = size;
1018 	m->m_ext.ext_type = type;
1019 	m->m_ext.ext_flags = EXT_FLAG_EMBREF;
1020 	m->m_ext.ext_count = 1;
1021 	m->m_flags |= M_EXT;
1022 	MBUF_PROBE3(m__cljset, m, cl, type);
1023 }
1024 
1025 static __inline void
1026 m_chtype(struct mbuf *m, short new_type)
1027 {
1028 
1029 	m->m_type = new_type;
1030 }
1031 
1032 static __inline void
1033 m_clrprotoflags(struct mbuf *m)
1034 {
1035 
1036 	while (m) {
1037 		m->m_flags &= ~M_PROTOFLAGS;
1038 		m = m->m_next;
1039 	}
1040 }
1041 
1042 static __inline struct mbuf *
1043 m_last(struct mbuf *m)
1044 {
1045 
1046 	while (m->m_next)
1047 		m = m->m_next;
1048 	return (m);
1049 }
1050 
1051 static inline u_int
1052 m_extrefcnt(struct mbuf *m)
1053 {
1054 
1055 	KASSERT(m->m_flags & M_EXT, ("%s: M_EXT missing", __func__));
1056 
1057 	return ((m->m_ext.ext_flags & EXT_FLAG_EMBREF) ? m->m_ext.ext_count :
1058 	    *m->m_ext.ext_cnt);
1059 }
1060 
1061 /*
1062  * mbuf, cluster, and external object allocation macros (for compatibility
1063  * purposes).
1064  */
1065 #define	M_MOVE_PKTHDR(to, from)	m_move_pkthdr((to), (from))
1066 #define	MGET(m, how, type)	((m) = m_get((how), (type)))
1067 #define	MGETHDR(m, how, type)	((m) = m_gethdr((how), (type)))
1068 #define	MCLGET(m, how)		m_clget((m), (how))
1069 #define	MEXTADD(m, buf, size, free, arg1, arg2, flags, type)		\
1070     m_extadd((m), (char *)(buf), (size), (free), (arg1), (arg2),	\
1071     (flags), (type))
1072 #define	m_getm(m, len, how, type)					\
1073     m_getm2((m), (len), (how), (type), M_PKTHDR)
1074 
1075 /*
1076  * Evaluate TRUE if it's safe to write to the mbuf m's data region (this can
1077  * be both the local data payload, or an external buffer area, depending on
1078  * whether M_EXT is set).
1079  */
1080 #define	M_WRITABLE(m)	(((m)->m_flags & (M_RDONLY | M_EXTPG)) == 0 &&	\
1081 			 (!(((m)->m_flags & M_EXT)) ||			\
1082 			 (m_extrefcnt(m) == 1)))
1083 
1084 /* Check if the supplied mbuf has a packet header, or else panic. */
1085 #define	M_ASSERTPKTHDR(m)						\
1086 	KASSERT((m) != NULL && (m)->m_flags & M_PKTHDR,			\
1087 	    ("%s: no mbuf packet header!", __func__))
1088 
1089 /* Check if mbuf is multipage. */
1090 #define M_ASSERTEXTPG(m)						\
1091 	KASSERT(((m)->m_flags & (M_EXTPG|M_PKTHDR)) == M_EXTPG,		\
1092 	    ("%s: m %p is not multipage!", __func__, m))
1093 
1094 /*
1095  * Ensure that the supplied mbuf is a valid, non-free mbuf.
1096  *
1097  * XXX: Broken at the moment.  Need some UMA magic to make it work again.
1098  */
1099 #define	M_ASSERTVALID(m)						\
1100 	KASSERT((((struct mbuf *)m)->m_flags & 0) == 0,			\
1101 	    ("%s: attempted use of a free mbuf!", __func__))
1102 
1103 /*
1104  * Return the address of the start of the buffer associated with an mbuf,
1105  * handling external storage, packet-header mbufs, and regular data mbufs.
1106  */
1107 #define	M_START(m)							\
1108 	(((m)->m_flags & M_EXTPG) ? NULL :				\
1109 	 ((m)->m_flags & M_EXT) ? (m)->m_ext.ext_buf :			\
1110 	 ((m)->m_flags & M_PKTHDR) ? &(m)->m_pktdat[0] :		\
1111 	 &(m)->m_dat[0])
1112 
1113 /*
1114  * Return the size of the buffer associated with an mbuf, handling external
1115  * storage, packet-header mbufs, and regular data mbufs.
1116  */
1117 #define	M_SIZE(m)							\
1118 	(((m)->m_flags & M_EXT) ? (m)->m_ext.ext_size :			\
1119 	 ((m)->m_flags & M_PKTHDR) ? MHLEN :				\
1120 	 MLEN)
1121 
1122 /*
1123  * Set the m_data pointer of a newly allocated mbuf to place an object of the
1124  * specified size at the end of the mbuf, longword aligned.
1125  *
1126  * NB: Historically, we had M_ALIGN(), MH_ALIGN(), and MEXT_ALIGN() as
1127  * separate macros, each asserting that it was called at the proper moment.
1128  * This required callers to themselves test the storage type and call the
1129  * right one.  Rather than require callers to be aware of those layout
1130  * decisions, we centralize here.
1131  */
1132 static __inline void
1133 m_align(struct mbuf *m, int len)
1134 {
1135 #ifdef INVARIANTS
1136 	const char *msg = "%s: not a virgin mbuf";
1137 #endif
1138 	int adjust;
1139 
1140 	KASSERT(m->m_data == M_START(m), (msg, __func__));
1141 
1142 	adjust = M_SIZE(m) - len;
1143 	m->m_data += adjust &~ (sizeof(long)-1);
1144 }
1145 
1146 #define	M_ALIGN(m, len)		m_align(m, len)
1147 #define	MH_ALIGN(m, len)	m_align(m, len)
1148 #define	MEXT_ALIGN(m, len)	m_align(m, len)
1149 
1150 /*
1151  * Compute the amount of space available before the current start of data in
1152  * an mbuf.
1153  *
1154  * The M_WRITABLE() is a temporary, conservative safety measure: the burden
1155  * of checking writability of the mbuf data area rests solely with the caller.
1156  *
1157  * NB: In previous versions, M_LEADINGSPACE() would only check M_WRITABLE()
1158  * for mbufs with external storage.  We now allow mbuf-embedded data to be
1159  * read-only as well.
1160  */
1161 #define	M_LEADINGSPACE(m)						\
1162 	(M_WRITABLE(m) ? ((m)->m_data - M_START(m)) : 0)
1163 
1164 /*
1165  * Compute the amount of space available after the end of data in an mbuf.
1166  *
1167  * The M_WRITABLE() is a temporary, conservative safety measure: the burden
1168  * of checking writability of the mbuf data area rests solely with the caller.
1169  *
1170  * NB: In previous versions, M_TRAILINGSPACE() would only check M_WRITABLE()
1171  * for mbufs with external storage.  We now allow mbuf-embedded data to be
1172  * read-only as well.
1173  */
1174 #define	M_TRAILINGSPACE(m)						\
1175 	(M_WRITABLE(m) ?						\
1176 	    ((M_START(m) + M_SIZE(m)) - ((m)->m_data + (m)->m_len)) : 0)
1177 
1178 /*
1179  * Arrange to prepend space of size plen to mbuf m.  If a new mbuf must be
1180  * allocated, how specifies whether to wait.  If the allocation fails, the
1181  * original mbuf chain is freed and m is set to NULL.
1182  */
1183 #define	M_PREPEND(m, plen, how) do {					\
1184 	struct mbuf **_mmp = &(m);					\
1185 	struct mbuf *_mm = *_mmp;					\
1186 	int _mplen = (plen);						\
1187 	int __mhow = (how);						\
1188 									\
1189 	MBUF_CHECKSLEEP(how);						\
1190 	if (M_LEADINGSPACE(_mm) >= _mplen) {				\
1191 		_mm->m_data -= _mplen;					\
1192 		_mm->m_len += _mplen;					\
1193 	} else								\
1194 		_mm = m_prepend(_mm, _mplen, __mhow);			\
1195 	if (_mm != NULL && _mm->m_flags & M_PKTHDR)			\
1196 		_mm->m_pkthdr.len += _mplen;				\
1197 	*_mmp = _mm;							\
1198 } while (0)
1199 
1200 /*
1201  * Change mbuf to new type.  This is a relatively expensive operation and
1202  * should be avoided.
1203  */
1204 #define	MCHTYPE(m, t)	m_chtype((m), (t))
1205 
1206 /* Return the rcvif of a packet header. */
1207 static __inline struct ifnet *
1208 m_rcvif(struct mbuf *m)
1209 {
1210 
1211 	M_ASSERTPKTHDR(m);
1212 	if (m->m_pkthdr.csum_flags & CSUM_SND_TAG)
1213 		return (NULL);
1214 	return (m->m_pkthdr.rcvif);
1215 }
1216 
1217 /* Length to m_copy to copy all. */
1218 #define	M_COPYALL	1000000000
1219 
1220 extern int		max_datalen;	/* MHLEN - max_hdr */
1221 extern int		max_hdr;	/* Largest link + protocol header */
1222 extern int		max_linkhdr;	/* Largest link-level header */
1223 extern int		max_protohdr;	/* Largest protocol header */
1224 extern int		nmbclusters;	/* Maximum number of clusters */
1225 extern bool		mb_use_ext_pgs;	/* Use ext_pgs for sendfile */
1226 
1227 /*-
1228  * Network packets may have annotations attached by affixing a list of
1229  * "packet tags" to the pkthdr structure.  Packet tags are dynamically
1230  * allocated semi-opaque data structures that have a fixed header
1231  * (struct m_tag) that specifies the size of the memory block and a
1232  * <cookie,type> pair that identifies it.  The cookie is a 32-bit unique
1233  * unsigned value used to identify a module or ABI.  By convention this value
1234  * is chosen as the date+time that the module is created, expressed as the
1235  * number of seconds since the epoch (e.g., using date -u +'%s').  The type
1236  * value is an ABI/module-specific value that identifies a particular
1237  * annotation and is private to the module.  For compatibility with systems
1238  * like OpenBSD that define packet tags w/o an ABI/module cookie, the value
1239  * PACKET_ABI_COMPAT is used to implement m_tag_get and m_tag_find
1240  * compatibility shim functions and several tag types are defined below.
1241  * Users that do not require compatibility should use a private cookie value
1242  * so that packet tag-related definitions can be maintained privately.
1243  *
1244  * Note that the packet tag returned by m_tag_alloc has the default memory
1245  * alignment implemented by malloc.  To reference private data one can use a
1246  * construct like:
1247  *
1248  *	struct m_tag *mtag = m_tag_alloc(...);
1249  *	struct foo *p = (struct foo *)(mtag+1);
1250  *
1251  * if the alignment of struct m_tag is sufficient for referencing members of
1252  * struct foo.  Otherwise it is necessary to embed struct m_tag within the
1253  * private data structure to insure proper alignment; e.g.,
1254  *
1255  *	struct foo {
1256  *		struct m_tag	tag;
1257  *		...
1258  *	};
1259  *	struct foo *p = (struct foo *) m_tag_alloc(...);
1260  *	struct m_tag *mtag = &p->tag;
1261  */
1262 
1263 /*
1264  * Persistent tags stay with an mbuf until the mbuf is reclaimed.  Otherwise
1265  * tags are expected to ``vanish'' when they pass through a network
1266  * interface.  For most interfaces this happens normally as the tags are
1267  * reclaimed when the mbuf is free'd.  However in some special cases
1268  * reclaiming must be done manually.  An example is packets that pass through
1269  * the loopback interface.  Also, one must be careful to do this when
1270  * ``turning around'' packets (e.g., icmp_reflect).
1271  *
1272  * To mark a tag persistent bit-or this flag in when defining the tag id.
1273  * The tag will then be treated as described above.
1274  */
1275 #define	MTAG_PERSISTENT				0x800
1276 
1277 #define	PACKET_TAG_NONE				0  /* Nadda */
1278 
1279 /* Packet tags for use with PACKET_ABI_COMPAT. */
1280 #define	PACKET_TAG_IPSEC_IN_DONE		1  /* IPsec applied, in */
1281 #define	PACKET_TAG_IPSEC_OUT_DONE		2  /* IPsec applied, out */
1282 #define	PACKET_TAG_IPSEC_IN_CRYPTO_DONE		3  /* NIC IPsec crypto done */
1283 #define	PACKET_TAG_IPSEC_OUT_CRYPTO_NEEDED	4  /* NIC IPsec crypto req'ed */
1284 #define	PACKET_TAG_IPSEC_IN_COULD_DO_CRYPTO	5  /* NIC notifies IPsec */
1285 #define	PACKET_TAG_IPSEC_PENDING_TDB		6  /* Reminder to do IPsec */
1286 #define	PACKET_TAG_BRIDGE			7  /* Bridge processing done */
1287 #define	PACKET_TAG_GIF				8  /* GIF processing done */
1288 #define	PACKET_TAG_GRE				9  /* GRE processing done */
1289 #define	PACKET_TAG_IN_PACKET_CHECKSUM		10 /* NIC checksumming done */
1290 #define	PACKET_TAG_ENCAP			11 /* Encap.  processing */
1291 #define	PACKET_TAG_IPSEC_SOCKET			12 /* IPSEC socket ref */
1292 #define	PACKET_TAG_IPSEC_HISTORY		13 /* IPSEC history */
1293 #define	PACKET_TAG_IPV6_INPUT			14 /* IPV6 input processing */
1294 #define	PACKET_TAG_DUMMYNET			15 /* dummynet info */
1295 #define	PACKET_TAG_DIVERT			17 /* divert info */
1296 #define	PACKET_TAG_IPFORWARD			18 /* ipforward info */
1297 #define	PACKET_TAG_MACLABEL	(19 | MTAG_PERSISTENT) /* MAC label */
1298 #define	PACKET_TAG_PF		(21 | MTAG_PERSISTENT) /* PF/ALTQ information */
1299 #define	PACKET_TAG_RTSOCKFAM			25 /* rtsock sa family */
1300 #define	PACKET_TAG_IPOPTIONS			27 /* Saved IP options */
1301 #define	PACKET_TAG_CARP				28 /* CARP info */
1302 #define	PACKET_TAG_IPSEC_NAT_T_PORTS		29 /* two uint16_t */
1303 #define	PACKET_TAG_ND_OUTGOING			30 /* ND outgoing */
1304 
1305 /* Specific cookies and tags. */
1306 
1307 /* Packet tag routines. */
1308 struct m_tag	*m_tag_alloc(u_int32_t, int, int, int);
1309 void		 m_tag_delete(struct mbuf *, struct m_tag *);
1310 void		 m_tag_delete_chain(struct mbuf *, struct m_tag *);
1311 void		 m_tag_free_default(struct m_tag *);
1312 struct m_tag	*m_tag_locate(struct mbuf *, u_int32_t, int, struct m_tag *);
1313 struct m_tag	*m_tag_copy(struct m_tag *, int);
1314 int		 m_tag_copy_chain(struct mbuf *, const struct mbuf *, int);
1315 void		 m_tag_delete_nonpersistent(struct mbuf *);
1316 
1317 /*
1318  * Initialize the list of tags associated with an mbuf.
1319  */
1320 static __inline void
1321 m_tag_init(struct mbuf *m)
1322 {
1323 
1324 	SLIST_INIT(&m->m_pkthdr.tags);
1325 }
1326 
1327 /*
1328  * Set up the contents of a tag.  Note that this does not fill in the free
1329  * method; the caller is expected to do that.
1330  *
1331  * XXX probably should be called m_tag_init, but that was already taken.
1332  */
1333 static __inline void
1334 m_tag_setup(struct m_tag *t, u_int32_t cookie, int type, int len)
1335 {
1336 
1337 	t->m_tag_id = type;
1338 	t->m_tag_len = len;
1339 	t->m_tag_cookie = cookie;
1340 }
1341 
1342 /*
1343  * Reclaim resources associated with a tag.
1344  */
1345 static __inline void
1346 m_tag_free(struct m_tag *t)
1347 {
1348 
1349 	(*t->m_tag_free)(t);
1350 }
1351 
1352 /*
1353  * Return the first tag associated with an mbuf.
1354  */
1355 static __inline struct m_tag *
1356 m_tag_first(struct mbuf *m)
1357 {
1358 
1359 	return (SLIST_FIRST(&m->m_pkthdr.tags));
1360 }
1361 
1362 /*
1363  * Return the next tag in the list of tags associated with an mbuf.
1364  */
1365 static __inline struct m_tag *
1366 m_tag_next(struct mbuf *m __unused, struct m_tag *t)
1367 {
1368 
1369 	return (SLIST_NEXT(t, m_tag_link));
1370 }
1371 
1372 /*
1373  * Prepend a tag to the list of tags associated with an mbuf.
1374  */
1375 static __inline void
1376 m_tag_prepend(struct mbuf *m, struct m_tag *t)
1377 {
1378 
1379 	SLIST_INSERT_HEAD(&m->m_pkthdr.tags, t, m_tag_link);
1380 }
1381 
1382 /*
1383  * Unlink a tag from the list of tags associated with an mbuf.
1384  */
1385 static __inline void
1386 m_tag_unlink(struct mbuf *m, struct m_tag *t)
1387 {
1388 
1389 	SLIST_REMOVE(&m->m_pkthdr.tags, t, m_tag, m_tag_link);
1390 }
1391 
1392 /* These are for OpenBSD compatibility. */
1393 #define	MTAG_ABI_COMPAT		0		/* compatibility ABI */
1394 
1395 static __inline struct m_tag *
1396 m_tag_get(int type, int length, int wait)
1397 {
1398 	return (m_tag_alloc(MTAG_ABI_COMPAT, type, length, wait));
1399 }
1400 
1401 static __inline struct m_tag *
1402 m_tag_find(struct mbuf *m, int type, struct m_tag *start)
1403 {
1404 	return (SLIST_EMPTY(&m->m_pkthdr.tags) ? (struct m_tag *)NULL :
1405 	    m_tag_locate(m, MTAG_ABI_COMPAT, type, start));
1406 }
1407 
1408 static inline struct m_snd_tag *
1409 m_snd_tag_ref(struct m_snd_tag *mst)
1410 {
1411 
1412 	refcount_acquire(&mst->refcount);
1413 	return (mst);
1414 }
1415 
1416 static inline void
1417 m_snd_tag_rele(struct m_snd_tag *mst)
1418 {
1419 
1420 	if (refcount_release(&mst->refcount))
1421 		m_snd_tag_destroy(mst);
1422 }
1423 
1424 static __inline struct mbuf *
1425 m_free(struct mbuf *m)
1426 {
1427 	struct mbuf *n = m->m_next;
1428 
1429 	MBUF_PROBE1(m__free, m);
1430 	if ((m->m_flags & (M_PKTHDR|M_NOFREE)) == (M_PKTHDR|M_NOFREE))
1431 		m_tag_delete_chain(m, NULL);
1432 	if (m->m_flags & M_PKTHDR && m->m_pkthdr.csum_flags & CSUM_SND_TAG)
1433 		m_snd_tag_rele(m->m_pkthdr.snd_tag);
1434 	if (m->m_flags & M_EXTPG)
1435 		mb_free_extpg(m);
1436 	else if (m->m_flags & M_EXT)
1437 		mb_free_ext(m);
1438 	else if ((m->m_flags & M_NOFREE) == 0)
1439 		uma_zfree(zone_mbuf, m);
1440 	return (n);
1441 }
1442 
1443 static __inline int
1444 rt_m_getfib(struct mbuf *m)
1445 {
1446 	KASSERT(m->m_flags & M_PKTHDR , ("Attempt to get FIB from non header mbuf."));
1447 	return (m->m_pkthdr.fibnum);
1448 }
1449 
1450 #define M_GETFIB(_m)   rt_m_getfib(_m)
1451 
1452 #define M_SETFIB(_m, _fib) do {						\
1453         KASSERT((_m)->m_flags & M_PKTHDR, ("Attempt to set FIB on non header mbuf."));	\
1454 	((_m)->m_pkthdr.fibnum) = (_fib);				\
1455 } while (0)
1456 
1457 /* flags passed as first argument for "m_ether_tcpip_hash()" */
1458 #define	MBUF_HASHFLAG_L2	(1 << 2)
1459 #define	MBUF_HASHFLAG_L3	(1 << 3)
1460 #define	MBUF_HASHFLAG_L4	(1 << 4)
1461 
1462 /* mbuf hashing helper routines */
1463 uint32_t	m_ether_tcpip_hash_init(void);
1464 uint32_t	m_ether_tcpip_hash(const uint32_t, const struct mbuf *, const uint32_t);
1465 
1466 #ifdef MBUF_PROFILING
1467  void m_profile(struct mbuf *m);
1468  #define M_PROFILE(m) m_profile(m)
1469 #else
1470  #define M_PROFILE(m)
1471 #endif
1472 
1473 struct mbufq {
1474 	STAILQ_HEAD(, mbuf)	mq_head;
1475 	int			mq_len;
1476 	int			mq_maxlen;
1477 };
1478 
1479 static inline void
1480 mbufq_init(struct mbufq *mq, int maxlen)
1481 {
1482 
1483 	STAILQ_INIT(&mq->mq_head);
1484 	mq->mq_maxlen = maxlen;
1485 	mq->mq_len = 0;
1486 }
1487 
1488 static inline struct mbuf *
1489 mbufq_flush(struct mbufq *mq)
1490 {
1491 	struct mbuf *m;
1492 
1493 	m = STAILQ_FIRST(&mq->mq_head);
1494 	STAILQ_INIT(&mq->mq_head);
1495 	mq->mq_len = 0;
1496 	return (m);
1497 }
1498 
1499 static inline void
1500 mbufq_drain(struct mbufq *mq)
1501 {
1502 	struct mbuf *m, *n;
1503 
1504 	n = mbufq_flush(mq);
1505 	while ((m = n) != NULL) {
1506 		n = STAILQ_NEXT(m, m_stailqpkt);
1507 		m_freem(m);
1508 	}
1509 }
1510 
1511 static inline struct mbuf *
1512 mbufq_first(const struct mbufq *mq)
1513 {
1514 
1515 	return (STAILQ_FIRST(&mq->mq_head));
1516 }
1517 
1518 static inline struct mbuf *
1519 mbufq_last(const struct mbufq *mq)
1520 {
1521 
1522 	return (STAILQ_LAST(&mq->mq_head, mbuf, m_stailqpkt));
1523 }
1524 
1525 static inline int
1526 mbufq_full(const struct mbufq *mq)
1527 {
1528 
1529 	return (mq->mq_maxlen > 0 && mq->mq_len >= mq->mq_maxlen);
1530 }
1531 
1532 static inline int
1533 mbufq_len(const struct mbufq *mq)
1534 {
1535 
1536 	return (mq->mq_len);
1537 }
1538 
1539 static inline int
1540 mbufq_enqueue(struct mbufq *mq, struct mbuf *m)
1541 {
1542 
1543 	if (mbufq_full(mq))
1544 		return (ENOBUFS);
1545 	STAILQ_INSERT_TAIL(&mq->mq_head, m, m_stailqpkt);
1546 	mq->mq_len++;
1547 	return (0);
1548 }
1549 
1550 static inline struct mbuf *
1551 mbufq_dequeue(struct mbufq *mq)
1552 {
1553 	struct mbuf *m;
1554 
1555 	m = STAILQ_FIRST(&mq->mq_head);
1556 	if (m) {
1557 		STAILQ_REMOVE_HEAD(&mq->mq_head, m_stailqpkt);
1558 		m->m_nextpkt = NULL;
1559 		mq->mq_len--;
1560 	}
1561 	return (m);
1562 }
1563 
1564 static inline void
1565 mbufq_prepend(struct mbufq *mq, struct mbuf *m)
1566 {
1567 
1568 	STAILQ_INSERT_HEAD(&mq->mq_head, m, m_stailqpkt);
1569 	mq->mq_len++;
1570 }
1571 
1572 /*
1573  * Note: this doesn't enforce the maximum list size for dst.
1574  */
1575 static inline void
1576 mbufq_concat(struct mbufq *mq_dst, struct mbufq *mq_src)
1577 {
1578 
1579 	mq_dst->mq_len += mq_src->mq_len;
1580 	STAILQ_CONCAT(&mq_dst->mq_head, &mq_src->mq_head);
1581 	mq_src->mq_len = 0;
1582 }
1583 
1584 #ifdef _SYS_TIMESPEC_H_
1585 static inline void
1586 mbuf_tstmp2timespec(struct mbuf *m, struct timespec *ts)
1587 {
1588 
1589 	KASSERT((m->m_flags & M_PKTHDR) != 0, ("mbuf %p no M_PKTHDR", m));
1590 	KASSERT((m->m_flags & (M_TSTMP|M_TSTMP_LRO)) != 0, ("mbuf %p no M_TSTMP or M_TSTMP_LRO", m));
1591 	ts->tv_sec = m->m_pkthdr.rcv_tstmp / 1000000000;
1592 	ts->tv_nsec = m->m_pkthdr.rcv_tstmp % 1000000000;
1593 }
1594 #endif
1595 
1596 #ifdef DEBUGNET
1597 /* Invoked from the debugnet client code. */
1598 void	debugnet_mbuf_drain(void);
1599 void	debugnet_mbuf_start(void);
1600 void	debugnet_mbuf_finish(void);
1601 void	debugnet_mbuf_reinit(int nmbuf, int nclust, int clsize);
1602 #endif
1603 
1604 static inline bool
1605 mbuf_has_tls_session(struct mbuf *m)
1606 {
1607 
1608 	if (m->m_flags & M_EXTPG) {
1609 		if (m->m_epg_tls != NULL) {
1610 			return (true);
1611 		}
1612 	}
1613 	return (false);
1614 }
1615 
1616 #endif /* _KERNEL */
1617 #endif /* !_SYS_MBUF_H_ */
1618