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 */
  34
  35#ifndef _SYS_MBUF_H_
  36#define	_SYS_MBUF_H_
  37
  38/* XXX: These includes suck. Sorry! */
  39#include <sys/queue.h>
  40#ifdef _KERNEL
  41#include <sys/systm.h>
  42#include <sys/refcount.h>
  43#include <vm/uma.h>
  44
  45#include <sys/sdt.h>
  46
  47#define	MBUF_PROBE1(probe, arg0)					\
  48	SDT_PROBE1(sdt, , , probe, arg0)
  49#define	MBUF_PROBE2(probe, arg0, arg1)					\
  50	SDT_PROBE2(sdt, , , probe, arg0, arg1)
  51#define	MBUF_PROBE3(probe, arg0, arg1, arg2)				\
  52	SDT_PROBE3(sdt, , , probe, arg0, arg1, arg2)
  53#define	MBUF_PROBE4(probe, arg0, arg1, arg2, arg3)			\
  54	SDT_PROBE4(sdt, , , probe, arg0, arg1, arg2, arg3)
  55#define	MBUF_PROBE5(probe, arg0, arg1, arg2, arg3, arg4)		\
  56	SDT_PROBE5(sdt, , , probe, arg0, arg1, arg2, arg3, arg4)
  57
  58SDT_PROBE_DECLARE(sdt, , , m__init);
  59SDT_PROBE_DECLARE(sdt, , , m__gethdr_raw);
  60SDT_PROBE_DECLARE(sdt, , , m__gethdr);
  61SDT_PROBE_DECLARE(sdt, , , m__get_raw);
  62SDT_PROBE_DECLARE(sdt, , , m__get);
  63SDT_PROBE_DECLARE(sdt, , , m__getcl);
  64SDT_PROBE_DECLARE(sdt, , , m__getjcl);
  65SDT_PROBE_DECLARE(sdt, , , m__clget);
  66SDT_PROBE_DECLARE(sdt, , , m__cljget);
  67SDT_PROBE_DECLARE(sdt, , , m__cljset);
  68SDT_PROBE_DECLARE(sdt, , , m__free);
  69SDT_PROBE_DECLARE(sdt, , , m__freem);
  70SDT_PROBE_DECLARE(sdt, , , m__freemp);
  71
  72#endif /* _KERNEL */
  73
  74/*
  75 * Mbufs are of a single size, MSIZE (sys/param.h), which includes overhead.
  76 * An mbuf may add a single "mbuf cluster" of size MCLBYTES (also in
  77 * sys/param.h), which has no additional overhead and is used instead of the
  78 * internal data area; this is done when at least MINCLSIZE of data must be
  79 * stored.  Additionally, it is possible to allocate a separate buffer
  80 * externally and attach it to the mbuf in a way similar to that of mbuf
  81 * clusters.
  82 *
  83 * NB: These calculation do not take actual compiler-induced alignment and
  84 * padding inside the complete struct mbuf into account.  Appropriate
  85 * attention is required when changing members of struct mbuf.
  86 *
  87 * MLEN is data length in a normal mbuf.
  88 * MHLEN is data length in an mbuf with pktheader.
  89 * MINCLSIZE is a smallest amount of data that should be put into cluster.
  90 *
  91 * Compile-time assertions in uipc_mbuf.c test these values to ensure that
  92 * they are sensible.
  93 */
  94struct mbuf;
  95#define	MHSIZE		offsetof(struct mbuf, m_dat)
  96#define	MPKTHSIZE	offsetof(struct mbuf, m_pktdat)
  97#define	MLEN		((int)(MSIZE - MHSIZE))
  98#define	MHLEN		((int)(MSIZE - MPKTHSIZE))
  99#define	MINCLSIZE	(MHLEN + 1)
 100#define	M_NODOM		255
 101
 102#ifdef _KERNEL
 103/*-
 104 * Macro for type conversion: convert mbuf pointer to data pointer of correct
 105 * type:
 106 *
 107 * mtod(m, t)	-- Convert mbuf pointer to data pointer of correct type.
 108 * mtodo(m, o) -- Same as above but with offset 'o' into data.
 109 */
 110#define	mtod(m, t)	((t)((m)->m_data))
 111#define	mtodo(m, o)	((void *)(((m)->m_data) + (o)))
 112
 113/*
 114 * Argument structure passed to UMA routines during mbuf and packet
 115 * allocations.
 116 */
 117struct mb_args {
 118	int	flags;	/* Flags for mbuf being allocated */
 119	short	type;	/* Type of mbuf being allocated */
 120};
 121#endif /* _KERNEL */
 122
 123/*
 124 * Packet tag structure (see below for details).
 125 */
 126struct m_tag {
 127	SLIST_ENTRY(m_tag)	m_tag_link;	/* List of packet tags */
 128	u_int16_t		m_tag_id;	/* Tag ID */
 129	u_int16_t		m_tag_len;	/* Length of data */
 130	u_int32_t		m_tag_cookie;	/* ABI/Module ID */
 131	void			(*m_tag_free)(struct m_tag *);
 132};
 133
 134/*
 135 * Static network interface owned tag.
 136 * Allocated through ifp->if_snd_tag_alloc().
 137 */
 138struct if_snd_tag_sw;
 139
 140struct m_snd_tag {
 141	struct ifnet *ifp;		/* network interface tag belongs to */
 142	const struct if_snd_tag_sw *sw;
 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: 56
 149 *	 LP64: 64
 150 * Compile-time assertions in uipc_mbuf.c test these values to ensure that
 151 * they are correct.
 152 */
 153struct pkthdr {
 154	union {
 155		struct m_snd_tag *snd_tag;	/* send tag, if any */
 156		struct ifnet	*rcvif;		/* rcv interface */
 157		struct {
 158			uint16_t rcvidx;	/* rcv interface index ... */
 159			uint16_t rcvgen;	/* ... and generation count */
 160		};
 161	};
 162	union {
 163		struct ifnet	*leaf_rcvif;	/* leaf rcv interface */
 164		struct {
 165			uint16_t leaf_rcvidx;	/* leaf rcv interface index ... */
 166			uint16_t leaf_rcvgen;	/* ... and generation count */
 167		};
 168	};
 169	SLIST_HEAD(packet_tags, m_tag) tags; /* list of packet tags */
 170	int32_t		 len;		/* total packet length */
 171
 172	/* Layer crossing persistent information. */
 173	uint32_t	 flowid;	/* packet's 4-tuple system */
 174	uint32_t	 csum_flags;	/* checksum and offload features */
 175	uint16_t	 fibnum;	/* this packet should use this fib */
 176	uint8_t		 numa_domain;	/* NUMA domain of recvd pkt */
 177	uint8_t		 rsstype;	/* hash type */
 178#if !defined(__LP64__)
 179	uint32_t	 pad;		/* pad for 64bit alignment */
 180#endif
 181	union {
 182		uint64_t	rcv_tstmp;	/* timestamp in ns */
 183		struct {
 184			uint8_t		 l2hlen;	/* layer 2 hdr len */
 185			uint8_t		 l3hlen;	/* layer 3 hdr len */
 186			uint8_t		 l4hlen;	/* layer 4 hdr len */
 187			uint8_t		 l5hlen;	/* layer 5 hdr len */
 188			uint8_t		 inner_l2hlen;
 189			uint8_t		 inner_l3hlen;
 190			uint8_t		 inner_l4hlen;
 191			uint8_t		 inner_l5hlen;
 192		};
 193	};
 194	union {
 195		uint8_t  eight[8];
 196		uint16_t sixteen[4];
 197		uint32_t thirtytwo[2];
 198		uint64_t sixtyfour[1];
 199		uintptr_t unintptr[1];
 200		void	*ptr;
 201	} PH_per;
 202
 203	/* Layer specific non-persistent local storage for reassembly, etc. */
 204	union {
 205		union {
 206			uint8_t  eight[8];
 207			uint16_t sixteen[4];
 208			uint32_t thirtytwo[2];
 209			uint64_t sixtyfour[1];
 210			uintptr_t unintptr[1];
 211			void 	*ptr;
 212		} PH_loc;
 213		/* Upon allocation: total packet memory consumption. */
 214		u_int	memlen;
 215	};
 216};
 217#define	ether_vtag	PH_per.sixteen[0]
 218#define tcp_tun_port	PH_per.sixteen[0] /* outbound */
 219#define	vt_nrecs	PH_per.sixteen[0]	  /* mld and v6-ND */
 220#define	tso_segsz	PH_per.sixteen[1] /* inbound after LRO */
 221#define	lro_nsegs	tso_segsz	  /* inbound after LRO */
 222#define	csum_data	PH_per.thirtytwo[1] /* inbound from hardware up */
 223#define	lro_tcp_d_len	PH_loc.sixteen[0] /* inbound during LRO (no reassembly) */
 224#define	lro_tcp_d_csum	PH_loc.sixteen[1] /* inbound during LRO (no reassembly) */
 225#define	lro_tcp_h_off	PH_loc.sixteen[2] /* inbound during LRO (no reassembly) */
 226#define	lro_etype	PH_loc.sixteen[3] /* inbound during LRO (no reassembly) */
 227/* Note PH_loc is used during IP reassembly (all 8 bytes as a ptr) */
 228
 229/*
 230 * TLS records for TLS 1.0-1.2 can have the following header lengths:
 231 * - 5 (AES-CBC with implicit IV)
 232 * - 21 (AES-CBC with explicit IV)
 233 * - 13 (AES-GCM with 8 byte explicit IV)
 234 */
 235#define	MBUF_PEXT_HDR_LEN	23
 236
 237/*
 238 * TLS records for TLS 1.0-1.2 can have the following maximum trailer
 239 * lengths:
 240 * - 16 (AES-GCM)
 241 * - 36 (AES-CBC with SHA1 and up to 16 bytes of padding)
 242 * - 48 (AES-CBC with SHA2-256 and up to 16 bytes of padding)
 243 * - 64 (AES-CBC with SHA2-384 and up to 16 bytes of padding)
 244 */
 245#define	MBUF_PEXT_TRAIL_LEN	64
 246
 247#if defined(__LP64__)
 248#define MBUF_PEXT_MAX_PGS (40 / sizeof(vm_paddr_t))
 249#else
 250#define MBUF_PEXT_MAX_PGS (64 / sizeof(vm_paddr_t))
 251#endif
 252
 253#define	MBUF_PEXT_MAX_BYTES						\
 254    (MBUF_PEXT_MAX_PGS * PAGE_SIZE + MBUF_PEXT_HDR_LEN + MBUF_PEXT_TRAIL_LEN)
 255
 256struct ktls_session;
 257struct socket;
 258
 259/*
 260 * Description of external storage mapped into mbuf; valid only if M_EXT is
 261 * set.
 262 * Size ILP32: 28
 263 *	 LP64: 48
 264 * Compile-time assertions in uipc_mbuf.c test these values to ensure that
 265 * they are correct.
 266 */
 267typedef	void m_ext_free_t(struct mbuf *);
 268struct m_ext {
 269	union {
 270		/*
 271		 * If EXT_FLAG_EMBREF is set, then we use refcount in the
 272		 * mbuf, the 'ext_count' member.  Otherwise, we have a
 273		 * shadow copy and we use pointer 'ext_cnt'.  The original
 274		 * mbuf is responsible to carry the pointer to free routine
 275		 * and its arguments.  They aren't copied into shadows in
 276		 * mb_dupcl() to avoid dereferencing next cachelines.
 277		 */
 278		volatile u_int	 ext_count;
 279		volatile u_int	*ext_cnt;
 280	};
 281	uint32_t	 ext_size;	/* size of buffer, for ext_free */
 282	uint32_t	 ext_type:8,	/* type of external storage */
 283			 ext_flags:24;	/* external storage mbuf flags */
 284	union {
 285		struct {
 286			/*
 287			 * Regular M_EXT mbuf:
 288			 * o ext_buf always points to the external buffer.
 289			 * o ext_free (below) and two optional arguments
 290			 *   ext_arg1 and ext_arg2 store the free context for
 291			 *   the external storage.  They are set only in the
 292			 *   refcount carrying mbuf, the one with
 293			 *   EXT_FLAG_EMBREF flag, with exclusion for
 294			 *   EXT_EXTREF type, where the free context is copied
 295			 *   into all mbufs that use same external storage.
 296			 */
 297			char 	*ext_buf;	/* start of buffer */
 298#define	m_ext_copylen	offsetof(struct m_ext, ext_arg2)
 299			void	*ext_arg2;
 300		};
 301		struct {
 302			/*
 303			 * Multi-page M_EXTPG mbuf:
 304			 * o extpg_pa - page vector.
 305			 * o extpg_trail and extpg_hdr - TLS trailer and
 306			 *   header.
 307			 * Uses ext_free and may also use ext_arg1.
 308			 */
 309			vm_paddr_t	extpg_pa[MBUF_PEXT_MAX_PGS];
 310			char		extpg_trail[MBUF_PEXT_TRAIL_LEN];
 311			char		extpg_hdr[MBUF_PEXT_HDR_LEN];
 312			/* Pretend these 3 fields are part of mbuf itself. */
 313#define	m_epg_pa	m_ext.extpg_pa
 314#define	m_epg_trail	m_ext.extpg_trail
 315#define	m_epg_hdr	m_ext.extpg_hdr
 316#define	m_epg_ext_copylen	offsetof(struct m_ext, ext_free)
 317		};
 318	};
 319	/*
 320	 * Free method and optional argument pointer, both
 321	 * used by M_EXT and M_EXTPG.
 322	 */
 323	m_ext_free_t	*ext_free;
 324	void		*ext_arg1;
 325};
 326
 327/*
 328 * The core of the mbuf object along with some shortcut defines for practical
 329 * purposes.
 330 */
 331struct mbuf {
 332	/*
 333	 * Header present at the beginning of every mbuf.
 334	 * Size ILP32: 24
 335	 *      LP64: 32
 336	 * Compile-time assertions in uipc_mbuf.c test these values to ensure
 337	 * that they are correct.
 338	 */
 339	union {	/* next buffer in chain */
 340		struct mbuf		*m_next;
 341		SLIST_ENTRY(mbuf)	m_slist;
 342		STAILQ_ENTRY(mbuf)	m_stailq;
 343	};
 344	union {	/* next chain in queue/record */
 345		struct mbuf		*m_nextpkt;
 346		SLIST_ENTRY(mbuf)	m_slistpkt;
 347		STAILQ_ENTRY(mbuf)	m_stailqpkt;
 348	};
 349	caddr_t		 m_data;	/* location of data */
 350	int32_t		 m_len;		/* amount of data in this mbuf */
 351	uint32_t	 m_type:8,	/* type of data in this mbuf */
 352			 m_flags:24;	/* flags; see below */
 353#if !defined(__LP64__)
 354	uint32_t	 m_pad;		/* pad for 64bit alignment */
 355#endif
 356
 357	/*
 358	 * A set of optional headers (packet header, external storage header)
 359	 * and internal data storage.  Historically, these arrays were sized
 360	 * to MHLEN (space left after a packet header) and MLEN (space left
 361	 * after only a regular mbuf header); they are now variable size in
 362	 * order to support future work on variable-size mbufs.
 363	 */
 364	union {
 365		struct {
 366			union {
 367				/* M_PKTHDR set. */
 368				struct pkthdr	m_pkthdr;
 369
 370				/* M_EXTPG set.
 371				 * Multi-page M_EXTPG mbuf has its meta data
 372				 * split between the below anonymous structure
 373				 * and m_ext.  It carries vector of pages,
 374				 * optional header and trailer char vectors
 375				 * and pointers to socket/TLS data.
 376				 */
 377#define	m_epg_startcopy		m_epg_npgs
 378#define	m_epg_endcopy		m_epg_stailq
 379				struct {
 380					/* Overall count of pages and count of
 381					 * pages with I/O pending. */
 382					uint8_t	m_epg_npgs;
 383					uint8_t	m_epg_nrdy;
 384					/* TLS header and trailer lengths.
 385					 * The data itself resides in m_ext. */
 386					uint8_t	m_epg_hdrlen;
 387					uint8_t	m_epg_trllen;
 388					/* Offset into 1st page and length of
 389					 * data in the last page. */
 390					uint16_t m_epg_1st_off;
 391					uint16_t m_epg_last_len;
 392					uint8_t	m_epg_flags;
 393#define	EPG_FLAG_ANON	0x1	/* Data can be encrypted in place. */
 394#define	EPG_FLAG_2FREE	0x2	/* Scheduled for free. */
 395					uint8_t	m_epg_record_type;
 396					uint8_t	__spare[2];
 397					int	m_epg_enc_cnt;
 398					struct ktls_session *m_epg_tls;
 399					struct socket	*m_epg_so;
 400					uint64_t	m_epg_seqno;
 401					STAILQ_ENTRY(mbuf) m_epg_stailq;
 402				};
 403			};
 404			union {
 405				/* M_EXT or M_EXTPG set. */
 406				struct m_ext	m_ext;
 407				/* M_PKTHDR set, neither M_EXT nor M_EXTPG. */
 408				char		m_pktdat[0];
 409			};
 410		};
 411		char	m_dat[0];			/* !M_PKTHDR, !M_EXT */
 412	};
 413};
 414
 415#ifdef _KERNEL
 416static inline int
 417m_epg_pagelen(const struct mbuf *m, int pidx, int pgoff)
 418{
 419
 420	KASSERT(pgoff == 0 || pidx == 0,
 421	    ("page %d with non-zero offset %d in %p", pidx, pgoff, m));
 422
 423	if (pidx == m->m_epg_npgs - 1) {
 424		return (m->m_epg_last_len);
 425	} else {
 426		return (PAGE_SIZE - pgoff);
 427	}
 428}
 429
 430#ifdef INVARIANTS
 431#define	MCHECK(ex, msg)	KASSERT((ex),				\
 432	    ("Multi page mbuf %p with " #msg " at %s:%d",	\
 433	    m, __FILE__, __LINE__))
 434/*
 435 * NB: This expects a non-empty buffer (npgs > 0 and
 436 * last_pg_len > 0).
 437 */
 438#define	MBUF_EXT_PGS_ASSERT_SANITY(m)	do {				\
 439	MCHECK(m->m_epg_npgs > 0, "no valid pages");		\
 440	MCHECK(m->m_epg_npgs <= nitems(m->m_epg_pa),		\
 441	    "too many pages");						\
 442	MCHECK(m->m_epg_nrdy <= m->m_epg_npgs,			\
 443	    "too many ready pages");					\
 444	MCHECK(m->m_epg_1st_off < PAGE_SIZE,			\
 445		"too large page offset");				\
 446	MCHECK(m->m_epg_last_len > 0, "zero last page length");	\
 447	MCHECK(m->m_epg_last_len <= PAGE_SIZE,			\
 448	    "too large last page length");				\
 449	if (m->m_epg_npgs == 1)					\
 450		MCHECK(m->m_epg_1st_off +			\
 451		    m->m_epg_last_len <=	 PAGE_SIZE,		\
 452		    "single page too large");				\
 453	MCHECK(m->m_epg_hdrlen <= sizeof(m->m_epg_hdr),		\
 454	    "too large header length");					\
 455	MCHECK(m->m_epg_trllen <= sizeof(m->m_epg_trail),	\
 456	    "too large header length");					\
 457} while (0)
 458#else
 459#define	MBUF_EXT_PGS_ASSERT_SANITY(m)	do {} while (0)
 460#endif
 461#endif
 462
 463/*
 464 * mbuf flags of global significance and layer crossing.
 465 * Those of only protocol/layer specific significance are to be mapped
 466 * to M_PROTO[1-11] and cleared at layer handoff boundaries.
 467 * NB: Limited to the lower 24 bits.
 468 */
 469#define	M_EXT		0x00000001 /* has associated external storage */
 470#define	M_PKTHDR	0x00000002 /* start of record */
 471#define	M_EOR		0x00000004 /* end of record */
 472#define	M_RDONLY	0x00000008 /* associated data is marked read-only */
 473#define	M_BCAST		0x00000010 /* send/received as link-level broadcast */
 474#define	M_MCAST		0x00000020 /* send/received as link-level multicast */
 475#define	M_PROMISC	0x00000040 /* packet was not for us */
 476#define	M_VLANTAG	0x00000080 /* ether_vtag is valid */
 477#define	M_EXTPG		0x00000100 /* has array of unmapped pages and TLS */
 478#define	M_NOFREE	0x00000200 /* do not free mbuf, embedded in cluster */
 479#define	M_TSTMP		0x00000400 /* rcv_tstmp field is valid */
 480#define	M_TSTMP_HPREC	0x00000800 /* rcv_tstmp is high-prec, typically
 481				      hw-stamped on port (useful for IEEE 1588
 482				      and 802.1AS) */
 483#define M_TSTMP_LRO	0x00001000 /* Time LRO pushed in pkt is valid in (PH_loc) */
 484
 485#define	M_PROTO1	0x00002000 /* protocol-specific */
 486#define	M_PROTO2	0x00004000 /* protocol-specific */
 487#define	M_PROTO3	0x00008000 /* protocol-specific */
 488#define	M_PROTO4	0x00010000 /* protocol-specific */
 489#define	M_PROTO5	0x00020000 /* protocol-specific */
 490#define	M_PROTO6	0x00040000 /* protocol-specific */
 491#define	M_PROTO7	0x00080000 /* protocol-specific */
 492#define	M_PROTO8	0x00100000 /* protocol-specific */
 493#define	M_PROTO9	0x00200000 /* protocol-specific */
 494#define	M_PROTO10	0x00400000 /* protocol-specific */
 495#define	M_PROTO11	0x00800000 /* protocol-specific */
 496
 497/*
 498 * Flags to purge when crossing layers.
 499 */
 500#define	M_PROTOFLAGS \
 501    (M_PROTO1|M_PROTO2|M_PROTO3|M_PROTO4|M_PROTO5|M_PROTO6|M_PROTO7|M_PROTO8|\
 502     M_PROTO9|M_PROTO10|M_PROTO11)
 503
 504/*
 505 * Flags preserved when copying m_pkthdr.
 506 */
 507#define M_COPYFLAGS \
 508    (M_PKTHDR|M_EOR|M_RDONLY|M_BCAST|M_MCAST|M_PROMISC|M_VLANTAG|M_TSTMP| \
 509     M_TSTMP_HPREC|M_TSTMP_LRO|M_PROTOFLAGS)
 510
 511/*
 512 * Flags preserved during demote.
 513 */
 514#define	M_DEMOTEFLAGS \
 515    (M_EXT | M_RDONLY | M_NOFREE | M_EXTPG)
 516
 517/*
 518 * Mbuf flag description for use with printf(9) %b identifier.
 519 */
 520#define	M_FLAG_BITS \
 521    "\20\1M_EXT\2M_PKTHDR\3M_EOR\4M_RDONLY\5M_BCAST\6M_MCAST" \
 522    "\7M_PROMISC\10M_VLANTAG\11M_EXTPG\12M_NOFREE\13M_TSTMP\14M_TSTMP_HPREC\15M_TSTMP_LRO"
 523#define	M_FLAG_PROTOBITS \
 524    "\16M_PROTO1\17M_PROTO2\20M_PROTO3\21M_PROTO4" \
 525    "\22M_PROTO5\23M_PROTO6\24M_PROTO7\25M_PROTO8\26M_PROTO9" \
 526    "\27M_PROTO10\28M_PROTO11"
 527#define	M_FLAG_PRINTF (M_FLAG_BITS M_FLAG_PROTOBITS)
 528
 529/*
 530 * Network interface cards are able to hash protocol fields (such as IPv4
 531 * addresses and TCP port numbers) classify packets into flows.  These flows
 532 * can then be used to maintain ordering while delivering packets to the OS
 533 * via parallel input queues, as well as to provide a stateless affinity
 534 * model.  NIC drivers can pass up the hash via m->m_pkthdr.flowid, and set
 535 * m_flag fields to indicate how the hash should be interpreted by the
 536 * network stack.
 537 *
 538 * Most NICs support RSS, which provides ordering and explicit affinity, and
 539 * use the hash m_flag bits to indicate what header fields were covered by
 540 * the hash.  M_HASHTYPE_OPAQUE and M_HASHTYPE_OPAQUE_HASH can be set by non-
 541 * RSS cards or configurations that provide an opaque flow identifier, allowing
 542 * for ordering and distribution without explicit affinity.  Additionally,
 543 * M_HASHTYPE_OPAQUE_HASH indicates that the flow identifier has hash
 544 * properties.
 545 *
 546 * The meaning of the IPV6_EX suffix:
 547 * "o  Home address from the home address option in the IPv6 destination
 548 *     options header.  If the extension header is not present, use the Source
 549 *     IPv6 Address.
 550 *  o  IPv6 address that is contained in the Routing-Header-Type-2 from the
 551 *     associated extension header.  If the extension header is not present,
 552 *     use the Destination IPv6 Address."
 553 * Quoted from:
 554 * https://docs.microsoft.com/en-us/windows-hardware/drivers/network/rss-hashing-types#ndishashipv6ex
 555 */
 556#define	M_HASHTYPE_HASHPROP		0x80	/* has hash properties */
 557#define	M_HASHTYPE_INNER		0x40	/* calculated from inner headers */
 558#define	M_HASHTYPE_HASH(t)		(M_HASHTYPE_HASHPROP | (t))
 559/* Microsoft RSS standard hash types */
 560#define	M_HASHTYPE_NONE			0
 561#define	M_HASHTYPE_RSS_IPV4		M_HASHTYPE_HASH(1) /* IPv4 2-tuple */
 562#define	M_HASHTYPE_RSS_TCP_IPV4		M_HASHTYPE_HASH(2) /* TCPv4 4-tuple */
 563#define	M_HASHTYPE_RSS_IPV6		M_HASHTYPE_HASH(3) /* IPv6 2-tuple */
 564#define	M_HASHTYPE_RSS_TCP_IPV6		M_HASHTYPE_HASH(4) /* TCPv6 4-tuple */
 565#define	M_HASHTYPE_RSS_IPV6_EX		M_HASHTYPE_HASH(5) /* IPv6 2-tuple +
 566							    * ext hdrs */
 567#define	M_HASHTYPE_RSS_TCP_IPV6_EX	M_HASHTYPE_HASH(6) /* TCPv6 4-tuple +
 568							    * ext hdrs */
 569#define	M_HASHTYPE_RSS_UDP_IPV4		M_HASHTYPE_HASH(7) /* IPv4 UDP 4-tuple*/
 570#define	M_HASHTYPE_RSS_UDP_IPV6		M_HASHTYPE_HASH(9) /* IPv6 UDP 4-tuple*/
 571#define	M_HASHTYPE_RSS_UDP_IPV6_EX	M_HASHTYPE_HASH(10)/* IPv6 UDP 4-tuple +
 572							    * ext hdrs */
 573
 574#define	M_HASHTYPE_OPAQUE		0x3f	/* ordering, not affinity */
 575#define	M_HASHTYPE_OPAQUE_HASH		M_HASHTYPE_HASH(M_HASHTYPE_OPAQUE)
 576						/* ordering+hash, not affinity*/
 577
 578#define	M_HASHTYPE_CLEAR(m)	((m)->m_pkthdr.rsstype = 0)
 579#define	M_HASHTYPE_GET(m)	((m)->m_pkthdr.rsstype & ~M_HASHTYPE_INNER)
 580#define	M_HASHTYPE_SET(m, v)	((m)->m_pkthdr.rsstype = (v))
 581#define	M_HASHTYPE_TEST(m, v)	(M_HASHTYPE_GET(m) == (v))
 582#define	M_HASHTYPE_ISHASH(m)	\
 583    (((m)->m_pkthdr.rsstype & M_HASHTYPE_HASHPROP) != 0)
 584#define	M_HASHTYPE_SETINNER(m)	do {			\
 585	(m)->m_pkthdr.rsstype |= M_HASHTYPE_INNER;	\
 586    } while (0)
 587
 588/*
 589 * External mbuf storage buffer types.
 590 */
 591#define	EXT_CLUSTER	1	/* mbuf cluster */
 592#define	EXT_SFBUF	2	/* sendfile(2)'s sf_buf */
 593#define	EXT_JUMBOP	3	/* jumbo cluster page sized */
 594#define	EXT_JUMBO9	4	/* jumbo cluster 9216 bytes */
 595#define	EXT_JUMBO16	5	/* jumbo cluster 16184 bytes */
 596#define	EXT_PACKET	6	/* mbuf+cluster from packet zone */
 597#define	EXT_MBUF	7	/* external mbuf reference */
 598#define	EXT_RXRING	8	/* data in NIC receive ring */
 599
 600#define	EXT_VENDOR1	224	/* for vendor-internal use */
 601#define	EXT_VENDOR2	225	/* for vendor-internal use */
 602#define	EXT_VENDOR3	226	/* for vendor-internal use */
 603#define	EXT_VENDOR4	227	/* for vendor-internal use */
 604
 605#define	EXT_EXP1	244	/* for experimental use */
 606#define	EXT_EXP2	245	/* for experimental use */
 607#define	EXT_EXP3	246	/* for experimental use */
 608#define	EXT_EXP4	247	/* for experimental use */
 609
 610#define	EXT_NET_DRV	252	/* custom ext_buf provided by net driver(s) */
 611#define	EXT_MOD_TYPE	253	/* custom module's ext_buf type */
 612#define	EXT_DISPOSABLE	254	/* can throw this buffer away w/page flipping */
 613#define	EXT_EXTREF	255	/* has externally maintained ext_cnt ptr */
 614
 615/*
 616 * Flags for external mbuf buffer types.
 617 * NB: limited to the lower 24 bits.
 618 */
 619#define	EXT_FLAG_EMBREF		0x000001	/* embedded ext_count */
 620#define	EXT_FLAG_EXTREF		0x000002	/* external ext_cnt, notyet */
 621
 622#define	EXT_FLAG_NOFREE		0x000010	/* don't free mbuf to pool, notyet */
 623
 624#define	EXT_FLAG_VENDOR1	0x010000	/* These flags are vendor */
 625#define	EXT_FLAG_VENDOR2	0x020000	/* or submodule specific, */
 626#define	EXT_FLAG_VENDOR3	0x040000	/* not used by mbuf code. */
 627#define	EXT_FLAG_VENDOR4	0x080000	/* Set/read by submodule. */
 628
 629#define	EXT_FLAG_EXP1		0x100000	/* for experimental use */
 630#define	EXT_FLAG_EXP2		0x200000	/* for experimental use */
 631#define	EXT_FLAG_EXP3		0x400000	/* for experimental use */
 632#define	EXT_FLAG_EXP4		0x800000	/* for experimental use */
 633
 634/*
 635 * EXT flag description for use with printf(9) %b identifier.
 636 */
 637#define	EXT_FLAG_BITS \
 638    "\20\1EXT_FLAG_EMBREF\2EXT_FLAG_EXTREF\5EXT_FLAG_NOFREE" \
 639    "\21EXT_FLAG_VENDOR1\22EXT_FLAG_VENDOR2\23EXT_FLAG_VENDOR3" \
 640    "\24EXT_FLAG_VENDOR4\25EXT_FLAG_EXP1\26EXT_FLAG_EXP2\27EXT_FLAG_EXP3" \
 641    "\30EXT_FLAG_EXP4"
 642
 643/*
 644 * Flags indicating checksum, segmentation and other offload work to be
 645 * done, or already done, by hardware or lower layers.  It is split into
 646 * separate inbound and outbound flags.
 647 *
 648 * Outbound flags that are set by upper protocol layers requesting lower
 649 * layers, or ideally the hardware, to perform these offloading tasks.
 650 * For outbound packets this field and its flags can be directly tested
 651 * against ifnet if_hwassist.  Note that the outbound and the inbound flags do
 652 * not collide right now but they could be allowed to (as long as the flags are
 653 * scrubbed appropriately when the direction of an mbuf changes).  CSUM_BITS
 654 * would also have to split into CSUM_BITS_TX and CSUM_BITS_RX.
 655 *
 656 * CSUM_INNER_<x> is the same as CSUM_<x> but it applies to the inner frame.
 657 * The CSUM_ENCAP_<x> bits identify the outer encapsulation.
 658 */
 659#define	CSUM_IP			0x00000001	/* IP header checksum offload */
 660#define	CSUM_IP_UDP		0x00000002	/* UDP checksum offload */
 661#define	CSUM_IP_TCP		0x00000004	/* TCP checksum offload */
 662#define	CSUM_IP_SCTP		0x00000008	/* SCTP checksum offload */
 663#define	CSUM_IP_TSO		0x00000010	/* TCP segmentation offload */
 664#define	CSUM_IP_ISCSI		0x00000020	/* iSCSI checksum offload */
 665
 666#define	CSUM_INNER_IP6_UDP	0x00000040
 667#define	CSUM_INNER_IP6_TCP	0x00000080
 668#define	CSUM_INNER_IP6_TSO	0x00000100
 669#define	CSUM_IP6_UDP		0x00000200	/* UDP checksum offload */
 670#define	CSUM_IP6_TCP		0x00000400	/* TCP checksum offload */
 671#define	CSUM_IP6_SCTP		0x00000800	/* SCTP checksum offload */
 672#define	CSUM_IP6_TSO		0x00001000	/* TCP segmentation offload */
 673#define	CSUM_IP6_ISCSI		0x00002000	/* iSCSI checksum offload */
 674
 675#define	CSUM_INNER_IP		0x00004000
 676#define	CSUM_INNER_IP_UDP	0x00008000
 677#define	CSUM_INNER_IP_TCP	0x00010000
 678#define	CSUM_INNER_IP_TSO	0x00020000
 679
 680#define	CSUM_ENCAP_VXLAN	0x00040000	/* VXLAN outer encapsulation */
 681#define	CSUM_ENCAP_RSVD1	0x00080000
 682
 683/* Inbound checksum support where the checksum was verified by hardware. */
 684#define	CSUM_INNER_L3_CALC	0x00100000
 685#define	CSUM_INNER_L3_VALID	0x00200000
 686#define	CSUM_INNER_L4_CALC	0x00400000
 687#define	CSUM_INNER_L4_VALID	0x00800000
 688#define	CSUM_L3_CALC		0x01000000	/* calculated layer 3 csum */
 689#define	CSUM_L3_VALID		0x02000000	/* checksum is correct */
 690#define	CSUM_L4_CALC		0x04000000	/* calculated layer 4 csum */
 691#define	CSUM_L4_VALID		0x08000000	/* checksum is correct */
 692#define	CSUM_L5_CALC		0x10000000	/* calculated layer 5 csum */
 693#define	CSUM_L5_VALID		0x20000000	/* checksum is correct */
 694#define	CSUM_COALESCED		0x40000000	/* contains merged segments */
 695
 696#define	CSUM_SND_TAG		0x80000000	/* Packet header has send tag */
 697
 698#define CSUM_FLAGS_TX (CSUM_IP | CSUM_IP_UDP | CSUM_IP_TCP | CSUM_IP_SCTP | \
 699    CSUM_IP_TSO | CSUM_IP_ISCSI | CSUM_INNER_IP6_UDP | CSUM_INNER_IP6_TCP | \
 700    CSUM_INNER_IP6_TSO | CSUM_IP6_UDP | CSUM_IP6_TCP | CSUM_IP6_SCTP | \
 701    CSUM_IP6_TSO | CSUM_IP6_ISCSI | CSUM_INNER_IP | CSUM_INNER_IP_UDP | \
 702    CSUM_INNER_IP_TCP | CSUM_INNER_IP_TSO | CSUM_ENCAP_VXLAN | \
 703    CSUM_ENCAP_RSVD1 | CSUM_SND_TAG)
 704
 705#define CSUM_FLAGS_RX (CSUM_INNER_L3_CALC | CSUM_INNER_L3_VALID | \
 706    CSUM_INNER_L4_CALC | CSUM_INNER_L4_VALID | CSUM_L3_CALC | CSUM_L3_VALID | \
 707    CSUM_L4_CALC | CSUM_L4_VALID | CSUM_L5_CALC | CSUM_L5_VALID | \
 708    CSUM_COALESCED)
 709
 710/*
 711 * CSUM flag description for use with printf(9) %b identifier.
 712 */
 713#define	CSUM_BITS \
 714    "\20\1CSUM_IP\2CSUM_IP_UDP\3CSUM_IP_TCP\4CSUM_IP_SCTP\5CSUM_IP_TSO" \
 715    "\6CSUM_IP_ISCSI\7CSUM_INNER_IP6_UDP\10CSUM_INNER_IP6_TCP" \
 716    "\11CSUM_INNER_IP6_TSO\12CSUM_IP6_UDP\13CSUM_IP6_TCP\14CSUM_IP6_SCTP" \
 717    "\15CSUM_IP6_TSO\16CSUM_IP6_ISCSI\17CSUM_INNER_IP\20CSUM_INNER_IP_UDP" \
 718    "\21CSUM_INNER_IP_TCP\22CSUM_INNER_IP_TSO\23CSUM_ENCAP_VXLAN" \
 719    "\24CSUM_ENCAP_RSVD1\25CSUM_INNER_L3_CALC\26CSUM_INNER_L3_VALID" \
 720    "\27CSUM_INNER_L4_CALC\30CSUM_INNER_L4_VALID\31CSUM_L3_CALC" \
 721    "\32CSUM_L3_VALID\33CSUM_L4_CALC\34CSUM_L4_VALID\35CSUM_L5_CALC" \
 722    "\36CSUM_L5_VALID\37CSUM_COALESCED\40CSUM_SND_TAG"
 723
 724/* CSUM flags compatibility mappings. */
 725#define	CSUM_IP_CHECKED		CSUM_L3_CALC
 726#define	CSUM_IP_VALID		CSUM_L3_VALID
 727#define	CSUM_DATA_VALID		CSUM_L4_VALID
 728#define	CSUM_PSEUDO_HDR		CSUM_L4_CALC
 729#define	CSUM_SCTP_VALID		CSUM_L4_VALID
 730#define	CSUM_DELAY_DATA		(CSUM_TCP|CSUM_UDP)
 731#define	CSUM_DELAY_IP		CSUM_IP		/* Only v4, no v6 IP hdr csum */
 732#define	CSUM_DELAY_DATA_IPV6	(CSUM_TCP_IPV6|CSUM_UDP_IPV6)
 733#define	CSUM_DATA_VALID_IPV6	CSUM_DATA_VALID
 734#define	CSUM_TCP		CSUM_IP_TCP
 735#define	CSUM_UDP		CSUM_IP_UDP
 736#define	CSUM_SCTP		CSUM_IP_SCTP
 737#define	CSUM_TSO		(CSUM_IP_TSO|CSUM_IP6_TSO)
 738#define	CSUM_INNER_TSO		(CSUM_INNER_IP_TSO|CSUM_INNER_IP6_TSO)
 739#define	CSUM_UDP_IPV6		CSUM_IP6_UDP
 740#define	CSUM_TCP_IPV6		CSUM_IP6_TCP
 741#define	CSUM_SCTP_IPV6		CSUM_IP6_SCTP
 742#define	CSUM_TLS_MASK		(CSUM_L5_CALC|CSUM_L5_VALID)
 743#define	CSUM_TLS_DECRYPTED	CSUM_L5_CALC
 744
 745/*
 746 * mbuf types describing the content of the mbuf (including external storage).
 747 */
 748#define	MT_NOTMBUF	0	/* USED INTERNALLY ONLY! Object is not mbuf */
 749#define	MT_DATA		1	/* dynamic (data) allocation */
 750#define	MT_HEADER	MT_DATA	/* packet header, use M_PKTHDR instead */
 751
 752#define	MT_VENDOR1	4	/* for vendor-internal use */
 753#define	MT_VENDOR2	5	/* for vendor-internal use */
 754#define	MT_VENDOR3	6	/* for vendor-internal use */
 755#define	MT_VENDOR4	7	/* for vendor-internal use */
 756
 757#define	MT_SONAME	8	/* socket name */
 758
 759#define	MT_EXP1		9	/* for experimental use */
 760#define	MT_EXP2		10	/* for experimental use */
 761#define	MT_EXP3		11	/* for experimental use */
 762#define	MT_EXP4		12	/* for experimental use */
 763
 764#define	MT_CONTROL	14	/* extra-data protocol message */
 765#define	MT_EXTCONTROL	15	/* control message with externalized contents */
 766#define	MT_OOBDATA	16	/* expedited data  */
 767
 768#define	MT_NOINIT	255	/* Not a type but a flag to allocate
 769				   a non-initialized mbuf */
 770
 771/*
 772 * String names of mbuf-related UMA(9) and malloc(9) types.  Exposed to
 773 * !_KERNEL so that monitoring tools can look up the zones with
 774 * libmemstat(3).
 775 */
 776#define	MBUF_MEM_NAME		"mbuf"
 777#define	MBUF_CLUSTER_MEM_NAME	"mbuf_cluster"
 778#define	MBUF_PACKET_MEM_NAME	"mbuf_packet"
 779#define	MBUF_JUMBOP_MEM_NAME	"mbuf_jumbo_page"
 780#define	MBUF_JUMBO9_MEM_NAME	"mbuf_jumbo_9k"
 781#define	MBUF_JUMBO16_MEM_NAME	"mbuf_jumbo_16k"
 782#define	MBUF_TAG_MEM_NAME	"mbuf_tag"
 783#define	MBUF_EXTREFCNT_MEM_NAME	"mbuf_ext_refcnt"
 784#define	MBUF_EXTPGS_MEM_NAME	"mbuf_extpgs"
 785
 786#ifdef _KERNEL
 787union if_snd_tag_alloc_params;
 788
 789#define	MBUF_CHECKSLEEP(how) do {					\
 790	if (how == M_WAITOK)						\
 791		WITNESS_WARN(WARN_GIANTOK | WARN_SLEEPOK, NULL,		\
 792		    "Sleeping in \"%s\"", __func__);			\
 793} while (0)
 794
 795/*
 796 * Network buffer allocation API
 797 *
 798 * The rest of it is defined in kern/kern_mbuf.c
 799 */
 800extern uma_zone_t	zone_mbuf;
 801extern uma_zone_t	zone_clust;
 802extern uma_zone_t	zone_pack;
 803extern uma_zone_t	zone_jumbop;
 804extern uma_zone_t	zone_jumbo9;
 805extern uma_zone_t	zone_jumbo16;
 806extern uma_zone_t	zone_extpgs;
 807
 808void		 mb_dupcl(struct mbuf *, struct mbuf *);
 809void		 mb_free_ext(struct mbuf *);
 810void		 mb_free_extpg(struct mbuf *);
 811void		 mb_free_mext_pgs(struct mbuf *);
 812struct mbuf	*mb_alloc_ext_pgs(int, m_ext_free_t);
 813struct mbuf	*mb_alloc_ext_plus_pages(int, int);
 814struct mbuf	*mb_mapped_to_unmapped(struct mbuf *, int, int, int,
 815		    struct mbuf **);
 816int		 mb_unmapped_compress(struct mbuf *m);
 817struct mbuf 	*mb_unmapped_to_ext(struct mbuf *m);
 818void		 mb_free_notready(struct mbuf *m, int count);
 819void		 m_adj(struct mbuf *, int);
 820void		 m_adj_decap(struct mbuf *, int);
 821int		 m_apply(struct mbuf *, int, int,
 822		    int (*)(void *, void *, u_int), void *);
 823int		 m_append(struct mbuf *, int, c_caddr_t);
 824void		 m_cat(struct mbuf *, struct mbuf *);
 825void		 m_catpkt(struct mbuf *, struct mbuf *);
 826int		 m_clget(struct mbuf *m, int how);
 827void 		*m_cljget(struct mbuf *m, int how, int size);
 828struct mbuf	*m_collapse(struct mbuf *, int, int);
 829void		 m_copyback(struct mbuf *, int, int, c_caddr_t);
 830void		 m_copydata(const struct mbuf *, int, int, caddr_t);
 831struct mbuf	*m_copym(struct mbuf *, int, int, int);
 832struct mbuf	*m_copypacket(struct mbuf *, int);
 833void		 m_copy_pkthdr(struct mbuf *, struct mbuf *);
 834struct mbuf	*m_copyup(struct mbuf *, int, int);
 835struct mbuf	*m_defrag(struct mbuf *, int);
 836void		 m_demote_pkthdr(struct mbuf *);
 837void		 m_demote(struct mbuf *, int, int);
 838struct mbuf	*m_devget(char *, int, int, struct ifnet *,
 839		    void (*)(char *, caddr_t, u_int));
 840void		 m_dispose_extcontrolm(struct mbuf *m);
 841struct mbuf	*m_dup(const struct mbuf *, int);
 842int		 m_dup_pkthdr(struct mbuf *, const struct mbuf *, int);
 843void		 m_extadd(struct mbuf *, char *, u_int, m_ext_free_t,
 844		    void *, void *, int, int);
 845u_int		 m_fixhdr(struct mbuf *);
 846struct mbuf	*m_fragment(struct mbuf *, int, int);
 847void		 m_freem(struct mbuf *);
 848void		 m_freemp(struct mbuf *);
 849void		 m_free_raw(struct mbuf *);
 850struct mbuf	*m_get2(int, int, short, int);
 851struct mbuf	*m_get3(int, int, short, int);
 852struct mbuf	*m_getjcl(int, short, int, int);
 853struct mbuf	*m_getm2(struct mbuf *, int, int, short, int);
 854struct mbuf	*m_getptr(struct mbuf *, int, int *);
 855u_int		 m_length(struct mbuf *, struct mbuf **);
 856int		 m_mbuftouio(struct uio *, const struct mbuf *, int);
 857void		 m_move_pkthdr(struct mbuf *, struct mbuf *);
 858int		 m_pkthdr_init(struct mbuf *, int);
 859struct mbuf	*m_prepend(struct mbuf *, int, int);
 860void		 m_print(const struct mbuf *, int);
 861struct mbuf	*m_pulldown(struct mbuf *, int, int, int *);
 862struct mbuf	*m_pullup(struct mbuf *, int);
 863int		 m_sanity(struct mbuf *, int);
 864struct mbuf	*m_split(struct mbuf *, int, int);
 865struct mbuf	*m_uiotombuf(struct uio *, int, int, int, int);
 866int		 m_unmapped_uiomove(const struct mbuf *, int, struct uio *,
 867		    int);
 868struct mbuf	*m_unshare(struct mbuf *, int);
 869int		 m_snd_tag_alloc(struct ifnet *,
 870		    union if_snd_tag_alloc_params *, struct m_snd_tag **);
 871void		 m_snd_tag_init(struct m_snd_tag *, struct ifnet *,
 872		    const struct if_snd_tag_sw *);
 873void		 m_snd_tag_destroy(struct m_snd_tag *);
 874void		 m_rcvif_serialize(struct mbuf *);
 875struct ifnet	*m_rcvif_restore(struct mbuf *);
 876
 877static __inline int
 878m_gettype(int size)
 879{
 880	int type;
 881
 882	switch (size) {
 883	case MSIZE:
 884		type = EXT_MBUF;
 885		break;
 886	case MCLBYTES:
 887		type = EXT_CLUSTER;
 888		break;
 889	case MJUMPAGESIZE:
 890		type = EXT_JUMBOP;
 891		break;
 892	case MJUM9BYTES:
 893		type = EXT_JUMBO9;
 894		break;
 895	case MJUM16BYTES:
 896		type = EXT_JUMBO16;
 897		break;
 898	default:
 899		panic("%s: invalid cluster size %d", __func__, size);
 900	}
 901
 902	return (type);
 903}
 904
 905/*
 906 * Associated an external reference counted buffer with an mbuf.
 907 */
 908static __inline void
 909m_extaddref(struct mbuf *m, char *buf, u_int size, u_int *ref_cnt,
 910    m_ext_free_t freef, void *arg1, void *arg2)
 911{
 912
 913	KASSERT(ref_cnt != NULL, ("%s: ref_cnt not provided", __func__));
 914
 915	atomic_add_int(ref_cnt, 1);
 916	m->m_flags |= M_EXT;
 917	m->m_ext.ext_buf = buf;
 918	m->m_ext.ext_cnt = ref_cnt;
 919	m->m_data = m->m_ext.ext_buf;
 920	m->m_ext.ext_size = size;
 921	m->m_ext.ext_free = freef;
 922	m->m_ext.ext_arg1 = arg1;
 923	m->m_ext.ext_arg2 = arg2;
 924	m->m_ext.ext_type = EXT_EXTREF;
 925	m->m_ext.ext_flags = 0;
 926}
 927
 928static __inline uma_zone_t
 929m_getzone(int size)
 930{
 931	uma_zone_t zone;
 932
 933	switch (size) {
 934	case MCLBYTES:
 935		zone = zone_clust;
 936		break;
 937	case MJUMPAGESIZE:
 938		zone = zone_jumbop;
 939		break;
 940	case MJUM9BYTES:
 941		zone = zone_jumbo9;
 942		break;
 943	case MJUM16BYTES:
 944		zone = zone_jumbo16;
 945		break;
 946	default:
 947		panic("%s: invalid cluster size %d", __func__, size);
 948	}
 949
 950	return (zone);
 951}
 952
 953/*
 954 * Initialize an mbuf with linear storage.
 955 *
 956 * Inline because the consumer text overhead will be roughly the same to
 957 * initialize or call a function with this many parameters and M_PKTHDR
 958 * should go away with constant propagation for !MGETHDR.
 959 */
 960static __inline int
 961m_init(struct mbuf *m, int how, short type, int flags)
 962{
 963	int error;
 964
 965	m->m_next = NULL;
 966	m->m_nextpkt = NULL;
 967	m->m_data = m->m_dat;
 968	m->m_len = 0;
 969	m->m_flags = flags;
 970	m->m_type = type;
 971	if (flags & M_PKTHDR)
 972		error = m_pkthdr_init(m, how);
 973	else
 974		error = 0;
 975
 976	MBUF_PROBE5(m__init, m, how, type, flags, error);
 977	return (error);
 978}
 979
 980static __inline struct mbuf *
 981m_get_raw(int how, short type)
 982{
 983	struct mbuf *m;
 984	struct mb_args args;
 985
 986	args.flags = 0;
 987	args.type = type | MT_NOINIT;
 988	m = uma_zalloc_arg(zone_mbuf, &args, how);
 989	MBUF_PROBE3(m__get_raw, how, type, m);
 990	return (m);
 991}
 992
 993static __inline struct mbuf *
 994m_get(int how, short type)
 995{
 996	struct mbuf *m;
 997	struct mb_args args;
 998
 999	args.flags = 0;
1000	args.type = type;
1001	m = uma_zalloc_arg(zone_mbuf, &args, how);
1002	MBUF_PROBE3(m__get, how, type, m);
1003	return (m);
1004}
1005
1006static __inline struct mbuf *
1007m_gethdr_raw(int how, short type)
1008{
1009	struct mbuf *m;
1010	struct mb_args args;
1011
1012	args.flags = M_PKTHDR;
1013	args.type = type | MT_NOINIT;
1014	m = uma_zalloc_arg(zone_mbuf, &args, how);
1015	MBUF_PROBE3(m__gethdr_raw, how, type, m);
1016	return (m);
1017}
1018
1019static __inline struct mbuf *
1020m_gethdr(int how, short type)
1021{
1022	struct mbuf *m;
1023	struct mb_args args;
1024
1025	args.flags = M_PKTHDR;
1026	args.type = type;
1027	m = uma_zalloc_arg(zone_mbuf, &args, how);
1028	MBUF_PROBE3(m__gethdr, how, type, m);
1029	return (m);
1030}
1031
1032static __inline struct mbuf *
1033m_getcl(int how, short type, int flags)
1034{
1035	struct mbuf *m;
1036	struct mb_args args;
1037
1038	args.flags = flags;
1039	args.type = type;
1040	m = uma_zalloc_arg(zone_pack, &args, how);
1041	MBUF_PROBE4(m__getcl, how, type, flags, m);
1042	return (m);
1043}
1044
1045/*
1046 * XXX: m_cljset() is a dangerous API.  One must attach only a new,
1047 * unreferenced cluster to an mbuf(9).  It is not possible to assert
1048 * that, so care can be taken only by users of the API.
1049 */
1050static __inline void
1051m_cljset(struct mbuf *m, void *cl, int type)
1052{
1053	int size;
1054
1055	switch (type) {
1056	case EXT_CLUSTER:
1057		size = MCLBYTES;
1058		break;
1059	case EXT_JUMBOP:
1060		size = MJUMPAGESIZE;
1061		break;
1062	case EXT_JUMBO9:
1063		size = MJUM9BYTES;
1064		break;
1065	case EXT_JUMBO16:
1066		size = MJUM16BYTES;
1067		break;
1068	default:
1069		panic("%s: unknown cluster type %d", __func__, type);
1070		break;
1071	}
1072
1073	m->m_data = m->m_ext.ext_buf = cl;
1074	m->m_ext.ext_free = m->m_ext.ext_arg1 = m->m_ext.ext_arg2 = NULL;
1075	m->m_ext.ext_size = size;
1076	m->m_ext.ext_type = type;
1077	m->m_ext.ext_flags = EXT_FLAG_EMBREF;
1078	m->m_ext.ext_count = 1;
1079	m->m_flags |= M_EXT;
1080	MBUF_PROBE3(m__cljset, m, cl, type);
1081}
1082
1083static __inline void
1084m_chtype(struct mbuf *m, short new_type)
1085{
1086
1087	m->m_type = new_type;
1088}
1089
1090static __inline void
1091m_clrprotoflags(struct mbuf *m)
1092{
1093
1094	while (m) {
1095		m->m_flags &= ~M_PROTOFLAGS;
1096		m = m->m_next;
1097	}
1098}
1099
1100static __inline struct mbuf *
1101m_last(struct mbuf *m)
1102{
1103
1104	while (m->m_next)
1105		m = m->m_next;
1106	return (m);
1107}
1108
1109static inline u_int
1110m_extrefcnt(struct mbuf *m)
1111{
1112
1113	KASSERT(m->m_flags & M_EXT, ("%s: M_EXT missing", __func__));
1114
1115	return ((m->m_ext.ext_flags & EXT_FLAG_EMBREF) ? m->m_ext.ext_count :
1116	    *m->m_ext.ext_cnt);
1117}
1118
1119/*
1120 * mbuf, cluster, and external object allocation macros (for compatibility
1121 * purposes).
1122 */
1123#define	M_MOVE_PKTHDR(to, from)	m_move_pkthdr((to), (from))
1124#define	MGET(m, how, type)	((m) = m_get((how), (type)))
1125#define	MGETHDR(m, how, type)	((m) = m_gethdr((how), (type)))
1126#define	MCLGET(m, how)		m_clget((m), (how))
1127#define	MEXTADD(m, buf, size, free, arg1, arg2, flags, type)		\
1128    m_extadd((m), (char *)(buf), (size), (free), (arg1), (arg2),	\
1129    (flags), (type))
1130#define	m_getm(m, len, how, type)					\
1131    m_getm2((m), (len), (how), (type), M_PKTHDR)
1132
1133/*
1134 * Evaluate TRUE if it's safe to write to the mbuf m's data region (this can
1135 * be both the local data payload, or an external buffer area, depending on
1136 * whether M_EXT is set).
1137 */
1138#define	M_WRITABLE(m)	(((m)->m_flags & (M_RDONLY | M_EXTPG)) == 0 &&	\
1139			 (!(((m)->m_flags & M_EXT)) ||			\
1140			 (m_extrefcnt(m) == 1)))
1141
1142/* Check if the supplied mbuf has a packet header, or else panic. */
1143#define	M_ASSERTPKTHDR(m)						\
1144	KASSERT((m) != NULL && (m)->m_flags & M_PKTHDR,			\
1145	    ("%s: no mbuf packet header!", __func__))
1146
1147/* Check if the supplied mbuf has no send tag, or else panic. */
1148#define	M_ASSERT_NO_SND_TAG(m)						\
1149	KASSERT((m) != NULL && (m)->m_flags & M_PKTHDR &&		\
1150	       ((m)->m_pkthdr.csum_flags & CSUM_SND_TAG) == 0,		\
1151	    ("%s: receive mbuf has send tag!", __func__))
1152
1153/* Check if mbuf is multipage. */
1154#define M_ASSERTEXTPG(m)						\
1155	KASSERT(((m)->m_flags & (M_EXTPG|M_PKTHDR)) == M_EXTPG,		\
1156	    ("%s: m %p is not multipage!", __func__, m))
1157
1158/*
1159 * Ensure that the supplied mbuf is a valid, non-free mbuf.
1160 *
1161 * XXX: Broken at the moment.  Need some UMA magic to make it work again.
1162 */
1163#define	M_ASSERTVALID(m)						\
1164	KASSERT((((struct mbuf *)m)->m_flags & 0) == 0,			\
1165	    ("%s: attempted use of a free mbuf!", __func__))
1166
1167/* Check whether any mbuf in the chain is unmapped. */
1168#ifdef INVARIANTS
1169#define	M_ASSERTMAPPED(m) do {						\
1170	for (struct mbuf *__m = (m); __m != NULL; __m = __m->m_next)	\
1171		KASSERT((__m->m_flags & M_EXTPG) == 0,			\
1172		    ("%s: chain %p contains an unmapped mbuf", __func__, (m)));\
1173} while (0)
1174#else
1175#define	M_ASSERTMAPPED(m) do {} while (0)
1176#endif
1177
1178/*
1179 * Return the address of the start of the buffer associated with an mbuf,
1180 * handling external storage, packet-header mbufs, and regular data mbufs.
1181 */
1182#define	M_START(m)							\
1183	(((m)->m_flags & M_EXTPG) ? NULL :				\
1184	 ((m)->m_flags & M_EXT) ? (m)->m_ext.ext_buf :			\
1185	 ((m)->m_flags & M_PKTHDR) ? &(m)->m_pktdat[0] :		\
1186	 &(m)->m_dat[0])
1187
1188/*
1189 * Return the size of the buffer associated with an mbuf, handling external
1190 * storage, packet-header mbufs, and regular data mbufs.
1191 */
1192#define	M_SIZE(m)							\
1193	(((m)->m_flags & M_EXT) ? (m)->m_ext.ext_size :			\
1194	 ((m)->m_flags & M_PKTHDR) ? MHLEN :				\
1195	 MLEN)
1196
1197/*
1198 * Set the m_data pointer of a newly allocated mbuf to place an object of the
1199 * specified size at the end of the mbuf, longword aligned.
1200 *
1201 * NB: Historically, we had M_ALIGN(), MH_ALIGN(), and MEXT_ALIGN() as
1202 * separate macros, each asserting that it was called at the proper moment.
1203 * This required callers to themselves test the storage type and call the
1204 * right one.  Rather than require callers to be aware of those layout
1205 * decisions, we centralize here.
1206 */
1207static __inline void
1208m_align(struct mbuf *m, int len)
1209{
1210#ifdef INVARIANTS
1211	const char *msg = "%s: not a virgin mbuf";
1212#endif
1213	int adjust;
1214
1215	KASSERT(m->m_data == M_START(m), (msg, __func__));
1216
1217	adjust = M_SIZE(m) - len;
1218	m->m_data += adjust &~ (sizeof(long)-1);
1219}
1220
1221#define	M_ALIGN(m, len)		m_align(m, len)
1222#define	MH_ALIGN(m, len)	m_align(m, len)
1223#define	MEXT_ALIGN(m, len)	m_align(m, len)
1224
1225/*
1226 * Compute the amount of space available before the current start of data in
1227 * an mbuf.
1228 *
1229 * The M_WRITABLE() is a temporary, conservative safety measure: the burden
1230 * of checking writability of the mbuf data area rests solely with the caller.
1231 *
1232 * NB: In previous versions, M_LEADINGSPACE() would only check M_WRITABLE()
1233 * for mbufs with external storage.  We now allow mbuf-embedded data to be
1234 * read-only as well.
1235 */
1236#define	M_LEADINGSPACE(m)						\
1237	(M_WRITABLE(m) ? ((m)->m_data - M_START(m)) : 0)
1238
1239/*
1240 * So M_TRAILINGROOM() is for when you want to know how much space
1241 * would be there if it was writable. This can be used to
1242 * detect changes in mbufs by knowing the value at one point
1243 * and then being able to compare it later to the current M_TRAILINGROOM().
1244 * The TRAILINGSPACE() macro is not suitable for this since an mbuf
1245 * at one point might not be writable and then later it becomes writable
1246 * even though the space at the back of it has not changed.
1247 */
1248#define M_TRAILINGROOM(m) ((M_START(m) + M_SIZE(m)) - ((m)->m_data + (m)->m_len))
1249/*
1250 * Compute the amount of space available after the end of data in an mbuf.
1251 *
1252 * The M_WRITABLE() is a temporary, conservative safety measure: the burden
1253 * of checking writability of the mbuf data area rests solely with the caller.
1254 *
1255 * NB: In previous versions, M_TRAILINGSPACE() would only check M_WRITABLE()
1256 * for mbufs with external storage.  We now allow mbuf-embedded data to be
1257 * read-only as well.
1258 */
1259#define	M_TRAILINGSPACE(m) (M_WRITABLE(m) ? M_TRAILINGROOM(m) : 0)
1260
1261/*
1262 * Arrange to prepend space of size plen to mbuf m.  If a new mbuf must be
1263 * allocated, how specifies whether to wait.  If the allocation fails, the
1264 * original mbuf chain is freed and m is set to NULL.
1265 */
1266#define	M_PREPEND(m, plen, how) do {					\
1267	struct mbuf **_mmp = &(m);					\
1268	struct mbuf *_mm = *_mmp;					\
1269	int _mplen = (plen);						\
1270	int __mhow = (how);						\
1271									\
1272	MBUF_CHECKSLEEP(how);						\
1273	if (M_LEADINGSPACE(_mm) >= _mplen) {				\
1274		_mm->m_data -= _mplen;					\
1275		_mm->m_len += _mplen;					\
1276	} else								\
1277		_mm = m_prepend(_mm, _mplen, __mhow);			\
1278	if (_mm != NULL && _mm->m_flags & M_PKTHDR)			\
1279		_mm->m_pkthdr.len += _mplen;				\
1280	*_mmp = _mm;							\
1281} while (0)
1282
1283/*
1284 * Change mbuf to new type.  This is a relatively expensive operation and
1285 * should be avoided.
1286 */
1287#define	MCHTYPE(m, t)	m_chtype((m), (t))
1288
1289/* Return the rcvif of a packet header. */
1290static __inline struct ifnet *
1291m_rcvif(struct mbuf *m)
1292{
1293
1294	M_ASSERTPKTHDR(m);
1295	if (m->m_pkthdr.csum_flags & CSUM_SND_TAG)
1296		return (NULL);
1297	return (m->m_pkthdr.rcvif);
1298}
1299
1300/* Length to m_copy to copy all. */
1301#define	M_COPYALL	1000000000
1302
1303extern u_int		max_linkhdr;	/* Largest link-level header */
1304extern u_int		max_hdr;	/* Largest link + protocol header */
1305extern u_int		max_protohdr;	/* Largest protocol header */
1306void max_linkhdr_grow(u_int);
1307void max_protohdr_grow(u_int);
1308
1309extern int		nmbclusters;	/* Maximum number of clusters */
1310extern bool		mb_use_ext_pgs;	/* Use ext_pgs for sendfile */
1311
1312/*-
1313 * Network packets may have annotations attached by affixing a list of
1314 * "packet tags" to the pkthdr structure.  Packet tags are dynamically
1315 * allocated semi-opaque data structures that have a fixed header
1316 * (struct m_tag) that specifies the size of the memory block and a
1317 * <cookie,type> pair that identifies it.  The cookie is a 32-bit unique
1318 * unsigned value used to identify a module or ABI.  By convention this value
1319 * is chosen as the date+time that the module is created, expressed as the
1320 * number of seconds since the epoch (e.g., using date -u +'%s').  The type
1321 * value is an ABI/module-specific value that identifies a particular
1322 * annotation and is private to the module.  For compatibility with systems
1323 * like OpenBSD that define packet tags w/o an ABI/module cookie, the value
1324 * PACKET_ABI_COMPAT is used to implement m_tag_get and m_tag_find
1325 * compatibility shim functions and several tag types are defined below.
1326 * Users that do not require compatibility should use a private cookie value
1327 * so that packet tag-related definitions can be maintained privately.
1328 *
1329 * Note that the packet tag returned by m_tag_alloc has the default memory
1330 * alignment implemented by malloc.  To reference private data one can use a
1331 * construct like:
1332 *
1333 *	struct m_tag *mtag = m_tag_alloc(...);
1334 *	struct foo *p = (struct foo *)(mtag+1);
1335 *
1336 * if the alignment of struct m_tag is sufficient for referencing members of
1337 * struct foo.  Otherwise it is necessary to embed struct m_tag within the
1338 * private data structure to insure proper alignment; e.g.,
1339 *
1340 *	struct foo {
1341 *		struct m_tag	tag;
1342 *		...
1343 *	};
1344 *	struct foo *p = (struct foo *) m_tag_alloc(...);
1345 *	struct m_tag *mtag = &p->tag;
1346 */
1347
1348/*
1349 * Persistent tags stay with an mbuf until the mbuf is reclaimed.  Otherwise
1350 * tags are expected to ``vanish'' when they pass through a network
1351 * interface.  For most interfaces this happens normally as the tags are
1352 * reclaimed when the mbuf is free'd.  However in some special cases
1353 * reclaiming must be done manually.  An example is packets that pass through
1354 * the loopback interface.  Also, one must be careful to do this when
1355 * ``turning around'' packets (e.g., icmp_reflect).
1356 *
1357 * To mark a tag persistent bit-or this flag in when defining the tag id.
1358 * The tag will then be treated as described above.
1359 */
1360#define	MTAG_PERSISTENT				0x800
1361
1362#define	PACKET_TAG_NONE				0  /* Nadda */
1363
1364/* Packet tags for use with PACKET_ABI_COMPAT. */
1365#define	PACKET_TAG_IPSEC_IN_DONE		1  /* IPsec applied, in */
1366#define	PACKET_TAG_IPSEC_OUT_DONE		2  /* IPsec applied, out */
1367#define	PACKET_TAG_IPSEC_IN_CRYPTO_DONE		3  /* NIC IPsec crypto done */
1368#define	PACKET_TAG_IPSEC_OUT_CRYPTO_NEEDED	4  /* NIC IPsec crypto req'ed */
1369#define	PACKET_TAG_IPSEC_IN_COULD_DO_CRYPTO	5  /* NIC notifies IPsec */
1370#define	PACKET_TAG_IPSEC_PENDING_TDB		6  /* Reminder to do IPsec */
1371#define	PACKET_TAG_BRIDGE			7  /* Bridge processing done */
1372#define	PACKET_TAG_GIF				8  /* GIF processing done */
1373#define	PACKET_TAG_GRE				9  /* GRE processing done */
1374#define	PACKET_TAG_IN_PACKET_CHECKSUM		10 /* NIC checksumming done */
1375#define	PACKET_TAG_ENCAP			11 /* Encap.  processing */
1376#define	PACKET_TAG_IPSEC_SOCKET			12 /* IPSEC socket ref */
1377#define	PACKET_TAG_IPSEC_HISTORY		13 /* IPSEC history */
1378#define	PACKET_TAG_IPV6_INPUT			14 /* IPV6 input processing */
1379#define	PACKET_TAG_DUMMYNET			15 /* dummynet info */
1380#define	PACKET_TAG_DIVERT			17 /* divert info */
1381#define	PACKET_TAG_IPFORWARD			18 /* ipforward info */
1382#define	PACKET_TAG_MACLABEL	(19 | MTAG_PERSISTENT) /* MAC label */
1383#define	PACKET_TAG_PF				21 /* PF/ALTQ information */
1384/* was	PACKET_TAG_RTSOCKFAM			25    rtsock sa family */
1385#define	PACKET_TAG_IPOPTIONS			27 /* Saved IP options */
1386#define	PACKET_TAG_CARP				28 /* CARP info */
1387#define	PACKET_TAG_IPSEC_NAT_T_PORTS		29 /* two uint16_t */
1388#define	PACKET_TAG_ND_OUTGOING			30 /* ND outgoing */
1389#define	PACKET_TAG_PF_REASSEMBLED		31
1390
1391/* Specific cookies and tags. */
1392
1393/* Packet tag routines. */
1394struct m_tag	*m_tag_alloc(uint32_t, uint16_t, int, int);
1395void		 m_tag_delete(struct mbuf *, struct m_tag *);
1396void		 m_tag_delete_chain(struct mbuf *, struct m_tag *);
1397void		 m_tag_free_default(struct m_tag *);
1398struct m_tag	*m_tag_locate(struct mbuf *, uint32_t, uint16_t,
1399    struct m_tag *);
1400struct m_tag	*m_tag_copy(struct m_tag *, int);
1401int		 m_tag_copy_chain(struct mbuf *, const struct mbuf *, int);
1402void		 m_tag_delete_nonpersistent(struct mbuf *);
1403
1404/*
1405 * Initialize the list of tags associated with an mbuf.
1406 */
1407static __inline void
1408m_tag_init(struct mbuf *m)
1409{
1410
1411	SLIST_INIT(&m->m_pkthdr.tags);
1412}
1413
1414/*
1415 * Set up the contents of a tag.  Note that this does not fill in the free
1416 * method; the caller is expected to do that.
1417 *
1418 * XXX probably should be called m_tag_init, but that was already taken.
1419 */
1420static __inline void
1421m_tag_setup(struct m_tag *t, uint32_t cookie, uint16_t type, int len)
1422{
1423
1424	t->m_tag_id = type;
1425	t->m_tag_len = len;
1426	t->m_tag_cookie = cookie;
1427}
1428
1429/*
1430 * Reclaim resources associated with a tag.
1431 */
1432static __inline void
1433m_tag_free(struct m_tag *t)
1434{
1435
1436	(*t->m_tag_free)(t);
1437}
1438
1439/*
1440 * Return the first tag associated with an mbuf.
1441 */
1442static __inline struct m_tag *
1443m_tag_first(struct mbuf *m)
1444{
1445
1446	return (SLIST_FIRST(&m->m_pkthdr.tags));
1447}
1448
1449/*
1450 * Return the next tag in the list of tags associated with an mbuf.
1451 */
1452static __inline struct m_tag *
1453m_tag_next(struct mbuf *m __unused, struct m_tag *t)
1454{
1455
1456	return (SLIST_NEXT(t, m_tag_link));
1457}
1458
1459/*
1460 * Prepend a tag to the list of tags associated with an mbuf.
1461 */
1462static __inline void
1463m_tag_prepend(struct mbuf *m, struct m_tag *t)
1464{
1465
1466	SLIST_INSERT_HEAD(&m->m_pkthdr.tags, t, m_tag_link);
1467}
1468
1469/*
1470 * Unlink a tag from the list of tags associated with an mbuf.
1471 */
1472static __inline void
1473m_tag_unlink(struct mbuf *m, struct m_tag *t)
1474{
1475
1476	SLIST_REMOVE(&m->m_pkthdr.tags, t, m_tag, m_tag_link);
1477}
1478
1479/* These are for OpenBSD compatibility. */
1480#define	MTAG_ABI_COMPAT		0		/* compatibility ABI */
1481
1482static __inline struct m_tag *
1483m_tag_get(uint16_t type, int length, int wait)
1484{
1485	return (m_tag_alloc(MTAG_ABI_COMPAT, type, length, wait));
1486}
1487
1488static __inline struct m_tag *
1489m_tag_find(struct mbuf *m, uint16_t type, struct m_tag *start)
1490{
1491	return (SLIST_EMPTY(&m->m_pkthdr.tags) ? (struct m_tag *)NULL :
1492	    m_tag_locate(m, MTAG_ABI_COMPAT, type, start));
1493}
1494
1495static inline struct m_snd_tag *
1496m_snd_tag_ref(struct m_snd_tag *mst)
1497{
1498
1499	refcount_acquire(&mst->refcount);
1500	return (mst);
1501}
1502
1503static inline void
1504m_snd_tag_rele(struct m_snd_tag *mst)
1505{
1506
1507	if (refcount_release(&mst->refcount))
1508		m_snd_tag_destroy(mst);
1509}
1510
1511static __inline struct mbuf *
1512m_free(struct mbuf *m)
1513{
1514	struct mbuf *n = m->m_next;
1515
1516	MBUF_PROBE1(m__free, m);
1517	if ((m->m_flags & (M_PKTHDR|M_NOFREE)) == (M_PKTHDR|M_NOFREE))
1518		m_tag_delete_chain(m, NULL);
1519	if (m->m_flags & M_PKTHDR && m->m_pkthdr.csum_flags & CSUM_SND_TAG)
1520		m_snd_tag_rele(m->m_pkthdr.snd_tag);
1521	if (m->m_flags & M_EXTPG)
1522		mb_free_extpg(m);
1523	else if (m->m_flags & M_EXT)
1524		mb_free_ext(m);
1525	else if ((m->m_flags & M_NOFREE) == 0)
1526		uma_zfree(zone_mbuf, m);
1527	return (n);
1528}
1529
1530static __inline int
1531rt_m_getfib(struct mbuf *m)
1532{
1533	KASSERT(m->m_flags & M_PKTHDR , ("Attempt to get FIB from non header mbuf."));
1534	return (m->m_pkthdr.fibnum);
1535}
1536
1537#define M_GETFIB(_m)   rt_m_getfib(_m)
1538
1539#define M_SETFIB(_m, _fib) do {						\
1540        KASSERT((_m)->m_flags & M_PKTHDR, ("Attempt to set FIB on non header mbuf."));	\
1541	((_m)->m_pkthdr.fibnum) = (_fib);				\
1542} while (0)
1543
1544/* flags passed as first argument for "m_xxx_tcpip_hash()" */
1545#define	MBUF_HASHFLAG_L2	(1 << 2)
1546#define	MBUF_HASHFLAG_L3	(1 << 3)
1547#define	MBUF_HASHFLAG_L4	(1 << 4)
1548
1549/* mbuf hashing helper routines */
1550uint32_t	m_ether_tcpip_hash_init(void);
1551uint32_t	m_ether_tcpip_hash(const uint32_t, const struct mbuf *, uint32_t);
1552uint32_t	m_infiniband_tcpip_hash_init(void);
1553uint32_t	m_infiniband_tcpip_hash(const uint32_t, const struct mbuf *, uint32_t);
1554
1555#ifdef MBUF_PROFILING
1556 void m_profile(struct mbuf *m);
1557 #define M_PROFILE(m) m_profile(m)
1558#else
1559 #define M_PROFILE(m)
1560#endif
1561
1562struct mbufq {
1563	STAILQ_HEAD(, mbuf)	mq_head;
1564	int			mq_len;
1565	int			mq_maxlen;
1566};
1567
1568static inline void
1569mbufq_init(struct mbufq *mq, int maxlen)
1570{
1571
1572	STAILQ_INIT(&mq->mq_head);
1573	mq->mq_maxlen = maxlen;
1574	mq->mq_len = 0;
1575}
1576
1577static inline struct mbuf *
1578mbufq_flush(struct mbufq *mq)
1579{
1580	struct mbuf *m;
1581
1582	m = STAILQ_FIRST(&mq->mq_head);
1583	STAILQ_INIT(&mq->mq_head);
1584	mq->mq_len = 0;
1585	return (m);
1586}
1587
1588static inline void
1589mbufq_drain(struct mbufq *mq)
1590{
1591	struct mbuf *m, *n;
1592
1593	n = mbufq_flush(mq);
1594	while ((m = n) != NULL) {
1595		n = STAILQ_NEXT(m, m_stailqpkt);
1596		m_freem(m);
1597	}
1598}
1599
1600static inline struct mbuf *
1601mbufq_first(const struct mbufq *mq)
1602{
1603
1604	return (STAILQ_FIRST(&mq->mq_head));
1605}
1606
1607static inline struct mbuf *
1608mbufq_last(const struct mbufq *mq)
1609{
1610
1611	return (STAILQ_LAST(&mq->mq_head, mbuf, m_stailqpkt));
1612}
1613
1614static inline bool
1615mbufq_empty(const struct mbufq *mq)
1616{
1617	return (mq->mq_len == 0);
1618}
1619
1620static inline int
1621mbufq_full(const struct mbufq *mq)
1622{
1623
1624	return (mq->mq_maxlen > 0 && mq->mq_len >= mq->mq_maxlen);
1625}
1626
1627static inline int
1628mbufq_len(const struct mbufq *mq)
1629{
1630
1631	return (mq->mq_len);
1632}
1633
1634static inline int
1635mbufq_enqueue(struct mbufq *mq, struct mbuf *m)
1636{
1637
1638	if (mbufq_full(mq))
1639		return (ENOBUFS);
1640	STAILQ_INSERT_TAIL(&mq->mq_head, m, m_stailqpkt);
1641	mq->mq_len++;
1642	return (0);
1643}
1644
1645static inline struct mbuf *
1646mbufq_dequeue(struct mbufq *mq)
1647{
1648	struct mbuf *m;
1649
1650	m = STAILQ_FIRST(&mq->mq_head);
1651	if (m) {
1652		STAILQ_REMOVE_HEAD(&mq->mq_head, m_stailqpkt);
1653		m->m_nextpkt = NULL;
1654		mq->mq_len--;
1655	}
1656	return (m);
1657}
1658
1659static inline void
1660mbufq_prepend(struct mbufq *mq, struct mbuf *m)
1661{
1662
1663	STAILQ_INSERT_HEAD(&mq->mq_head, m, m_stailqpkt);
1664	mq->mq_len++;
1665}
1666
1667/*
1668 * Note: this doesn't enforce the maximum list size for dst.
1669 */
1670static inline void
1671mbufq_concat(struct mbufq *mq_dst, struct mbufq *mq_src)
1672{
1673
1674	mq_dst->mq_len += mq_src->mq_len;
1675	STAILQ_CONCAT(&mq_dst->mq_head, &mq_src->mq_head);
1676	mq_src->mq_len = 0;
1677}
1678
1679#ifdef _SYS_TIMESPEC_H_
1680static inline void
1681mbuf_tstmp2timespec(struct mbuf *m, struct timespec *ts)
1682{
1683
1684	KASSERT((m->m_flags & M_PKTHDR) != 0, ("mbuf %p no M_PKTHDR", m));
1685	KASSERT((m->m_flags & (M_TSTMP|M_TSTMP_LRO)) != 0,
1686	    ("mbuf %p no M_TSTMP or M_TSTMP_LRO", m));
1687	ts->tv_sec = m->m_pkthdr.rcv_tstmp / 1000000000;
1688	ts->tv_nsec = m->m_pkthdr.rcv_tstmp % 1000000000;
1689}
1690#endif
1691
1692static inline void
1693mbuf_tstmp2timeval(struct mbuf *m, struct timeval *tv)
1694{
1695
1696	KASSERT((m->m_flags & M_PKTHDR) != 0, ("mbuf %p no M_PKTHDR", m));
1697	KASSERT((m->m_flags & (M_TSTMP|M_TSTMP_LRO)) != 0,
1698	    ("mbuf %p no M_TSTMP or M_TSTMP_LRO", m));
1699	tv->tv_sec = m->m_pkthdr.rcv_tstmp / 1000000000;
1700	tv->tv_usec = (m->m_pkthdr.rcv_tstmp % 1000000000) / 1000;
1701}
1702
1703#ifdef DEBUGNET
1704/* Invoked from the debugnet client code. */
1705void	debugnet_mbuf_drain(void);
1706void	debugnet_mbuf_start(void);
1707void	debugnet_mbuf_finish(void);
1708void	debugnet_mbuf_reinit(int nmbuf, int nclust, int clsize);
1709#endif
1710
1711static inline bool
1712mbuf_has_tls_session(struct mbuf *m)
1713{
1714
1715	if (m->m_flags & M_EXTPG) {
1716		if (m->m_epg_tls != NULL) {
1717			return (true);
1718		}
1719	}
1720	return (false);
1721}
1722
1723#endif /* _KERNEL */
1724#endif /* !_SYS_MBUF_H_ */