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