mbuf.h revision 331722
1103026Ssobomax/*- 2103026Ssobomax * Copyright (c) 1982, 1986, 1988, 1993 3103026Ssobomax * The Regents of the University of California. 4103026Ssobomax * All rights reserved. 5103026Ssobomax * 6103026Ssobomax * Redistribution and use in source and binary forms, with or without 7103026Ssobomax * modification, are permitted provided that the following conditions 8103026Ssobomax * are met: 9103026Ssobomax * 1. Redistributions of source code must retain the above copyright 10103026Ssobomax * notice, this list of conditions and the following disclaimer. 11103026Ssobomax * 2. Redistributions in binary form must reproduce the above copyright 12103026Ssobomax * notice, this list of conditions and the following disclaimer in the 13103026Ssobomax * documentation and/or other materials provided with the distribution. 14103026Ssobomax * 3. Neither the name of the University nor the names of its contributors 15103026Ssobomax * may be used to endorse or promote products derived from this software 16103026Ssobomax * without specific prior written permission. 17103026Ssobomax * 18103026Ssobomax * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 19103026Ssobomax * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 20103026Ssobomax * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 21103026Ssobomax * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 22103026Ssobomax * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 23103026Ssobomax * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 24103026Ssobomax * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 25103026Ssobomax * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 26107726Sru * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 27107726Sru * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 28103026Ssobomax * SUCH DAMAGE. 29103026Ssobomax * 30107726Sru * @(#)mbuf.h 8.5 (Berkeley) 2/19/95 31107726Sru * $FreeBSD: stable/11/sys/sys/mbuf.h 331722 2018-03-29 02:50:57Z eadler $ 32179894Sthompsa */ 33103026Ssobomax 34103026Ssobomax#ifndef _SYS_MBUF_H_ 35103026Ssobomax#define _SYS_MBUF_H_ 36103026Ssobomax 37103026Ssobomax/* XXX: These includes suck. Sorry! */ 38103026Ssobomax#include <sys/queue.h> 39163500Sdanger#ifdef _KERNEL 40163500Sdanger#include <sys/systm.h> 41163500Sdanger#include <vm/uma.h> 42163500Sdanger#ifdef WITNESS 43163500Sdanger#include <sys/lock.h> 44107726Sru#endif 45163500Sdanger#endif 46163500Sdanger 47163500Sdanger#ifdef _KERNEL 48163500Sdanger#include <sys/sdt.h> 49163500Sdanger 50163500Sdanger#define MBUF_PROBE1(probe, arg0) \ 51163500Sdanger SDT_PROBE1(sdt, , , probe, arg0) 52163500Sdanger#define MBUF_PROBE2(probe, arg0, arg1) \ 53163500Sdanger SDT_PROBE2(sdt, , , probe, arg0, arg1) 54103026Ssobomax#define MBUF_PROBE3(probe, arg0, arg1, arg2) \ 55103026Ssobomax SDT_PROBE3(sdt, , , probe, arg0, arg1, arg2) 56107726Sru#define MBUF_PROBE4(probe, arg0, arg1, arg2, arg3) \ 57103026Ssobomax SDT_PROBE4(sdt, , , probe, arg0, arg1, arg2, arg3) 58107726Sru#define MBUF_PROBE5(probe, arg0, arg1, arg2, arg3, arg4) \ 59107726Sru SDT_PROBE5(sdt, , , probe, arg0, arg1, arg2, arg3, arg4) 60103026Ssobomax 61103026SsobomaxSDT_PROBE_DECLARE(sdt, , , m__init); 62103026SsobomaxSDT_PROBE_DECLARE(sdt, , , m__gethdr); 63103026SsobomaxSDT_PROBE_DECLARE(sdt, , , m__get); 64103026SsobomaxSDT_PROBE_DECLARE(sdt, , , m__getcl); 65103026SsobomaxSDT_PROBE_DECLARE(sdt, , , m__clget); 66103026SsobomaxSDT_PROBE_DECLARE(sdt, , , m__cljget); 67103026SsobomaxSDT_PROBE_DECLARE(sdt, , , m__cljset); 68103026SsobomaxSDT_PROBE_DECLARE(sdt, , , m__free); 69103026SsobomaxSDT_PROBE_DECLARE(sdt, , , m__freem); 70103026Ssobomax 71103026Ssobomax#endif /* _KERNEL */ 72103026Ssobomax 73103026Ssobomax/* 74107726Sru * Mbufs are of a single size, MSIZE (sys/param.h), which includes overhead. 75107726Sru * An mbuf may add a single "mbuf cluster" of size MCLBYTES (also in 76103026Ssobomax * sys/param.h), which has no additional overhead and is used instead of the 77107726Sru * internal data area; this is done when at least MINCLSIZE of data must be 78107726Sru * stored. Additionally, it is possible to allocate a separate buffer 79103026Ssobomax * externally and attach it to the mbuf in a way similar to that of mbuf 80131530Sru * clusters. 81107726Sru * 82107726Sru * NB: These calculation do not take actual compiler-induced alignment and 83107726Sru * padding inside the complete struct mbuf into account. Appropriate 84103026Ssobomax * attention is required when changing members of struct mbuf. 85130582Sru * 86125024Ssobomax * MLEN is data length in a normal mbuf. 87125024Ssobomax * MHLEN is data length in an mbuf with pktheader. 88125024Ssobomax * MINCLSIZE is a smallest amount of data that should be put into cluster. 89125024Ssobomax * 90125024Ssobomax * Compile-time assertions in uipc_mbuf.c test these values to ensure that 91125024Ssobomax * they are sensible. 92125024Ssobomax */ 93125024Ssobomaxstruct mbuf; 94125024Ssobomax#define MHSIZE offsetof(struct mbuf, m_dat) 95107726Sru#define MPKTHSIZE offsetof(struct mbuf, m_pktdat) 96103026Ssobomax#define MLEN ((int)(MSIZE - MHSIZE)) 97107726Sru#define MHLEN ((int)(MSIZE - MPKTHSIZE)) 98107726Sru#define MINCLSIZE (MHLEN + 1) 99103026Ssobomax 100107726Sru#ifdef _KERNEL 101107726Sru/*- 102103026Ssobomax * Macro for type conversion: convert mbuf pointer to data pointer of correct 103107726Sru * type: 104103026Ssobomax * 105103026Ssobomax * mtod(m, t) -- Convert mbuf pointer to data pointer of correct type. 106103026Ssobomax * mtodo(m, o) -- Same as above but with offset 'o' into data. 107107726Sru */ 108103026Ssobomax#define mtod(m, t) ((t)((m)->m_data)) 109103026Ssobomax#define mtodo(m, o) ((void *)(((m)->m_data) + (o))) 110103026Ssobomax 111107726Sru/* 112107726Sru * Argument structure passed to UMA routines during mbuf and packet 113107726Sru * allocations. 114107726Sru */ 115107726Srustruct mb_args { 116107726Sru int flags; /* Flags for mbuf being allocated */ 117107726Sru short type; /* Type of mbuf being allocated */ 118107726Sru}; 119103026Ssobomax#endif /* _KERNEL */ 120107726Sru 121107726Sru/* 122107726Sru * Packet tag structure (see below for details). 123107726Sru */ 124107726Srustruct m_tag { 125107726Sru SLIST_ENTRY(m_tag) m_tag_link; /* List of packet tags */ 126107726Sru u_int16_t m_tag_id; /* Tag ID */ 127103026Ssobomax u_int16_t m_tag_len; /* Length of data */ 128107726Sru u_int32_t m_tag_cookie; /* ABI/Module ID */ 129107726Sru void (*m_tag_free)(struct m_tag *); 130107726Sru}; 131107726Sru 132107726Sru/* 133107726Sru * Record/packet header in first mbuf of chain; valid only if M_PKTHDR is set. 134107726Sru * Size ILP32: 48 135107726Sru * LP64: 56 136107726Sru * Compile-time assertions in uipc_mbuf.c test these values to ensure that 137107726Sru * they are correct. 138107726Sru */ 139107726Srustruct pkthdr { 140107726Sru struct ifnet *rcvif; /* rcv interface */ 141107726Sru SLIST_HEAD(packet_tags, m_tag) tags; /* list of packet tags */ 142107726Sru int32_t len; /* total packet length */ 143103026Ssobomax 144107726Sru /* Layer crossing persistent information. */ 145107726Sru uint32_t flowid; /* packet's 4-tuple system */ 146107726Sru uint64_t csum_flags; /* checksum and offload features */ 147107726Sru uint16_t fibnum; /* this packet should use this fib */ 148107726Sru uint8_t cosqos; /* class/quality of service */ 149107726Sru uint8_t rsstype; /* hash type */ 150103026Ssobomax uint8_t l2hlen; /* layer 2 header length */ 151103026Ssobomax uint8_t l3hlen; /* layer 3 header length */ 152103026Ssobomax uint8_t l4hlen; /* layer 4 header length */ 153103026Ssobomax uint8_t l5hlen; /* layer 5 header length */ 154103026Ssobomax union { 155107726Sru uint8_t eight[8]; 156107726Sru uint16_t sixteen[4]; 157107726Sru uint32_t thirtytwo[2]; 158103026Ssobomax uint64_t sixtyfour[1]; 159103026Ssobomax uintptr_t unintptr[1]; 160103026Ssobomax void *ptr; 161107726Sru } PH_per; 162103026Ssobomax 163179894Sthompsa /* Layer specific non-persistent local storage for reassembly, etc. */ 164179894Sthompsa union { 165179894Sthompsa uint8_t eight[8]; 166179894Sthompsa uint16_t sixteen[4]; 167179894Sthompsa uint32_t thirtytwo[2]; 168179894Sthompsa uint64_t sixtyfour[1]; 169103026Ssobomax uintptr_t unintptr[1]; 170103026Ssobomax void *ptr; 171103026Ssobomax } PH_loc; 172103026Ssobomax}; 173103026Ssobomax#define ether_vtag PH_per.sixteen[0] 174131530Sru#define PH_vt PH_per 175103026Ssobomax#define vt_nrecs sixteen[0] 176103026Ssobomax#define tso_segsz PH_per.sixteen[1] 177103026Ssobomax#define csum_phsum PH_per.sixteen[2] 178103026Ssobomax#define csum_data PH_per.thirtytwo[1] 179107726Sru 180103026Ssobomax/* 181103026Ssobomax * Description of external storage mapped into mbuf; valid only if M_EXT is 182107726Sru * set. 183103026Ssobomax * Size ILP32: 28 184107726Sru * LP64: 48 185103026Ssobomax * Compile-time assertions in uipc_mbuf.c test these values to ensure that 186107726Sru * they are correct. 187107726Sru */ 188107726Srustruct m_ext { 189107726Sru union { 190107726Sru volatile u_int ext_count; /* value of ref count info */ 191107726Sru volatile u_int *ext_cnt; /* pointer to ref count info */ 192107726Sru }; 193103026Ssobomax caddr_t ext_buf; /* start of buffer */ 194107726Sru uint32_t ext_size; /* size of buffer, for ext_free */ 195103026Ssobomax uint32_t ext_type:8, /* type of external storage */ 196107726Sru ext_flags:24; /* external storage mbuf flags */ 197107726Sru void (*ext_free) /* free routine if not the usual */ 198107726Sru (struct mbuf *, void *, void *); 199107726Sru void *ext_arg1; /* optional argument pointer */ 200107726Sru void *ext_arg2; /* optional argument pointer */ 201107726Sru}; 202107726Sru 203107726Sru/* 204103026Ssobomax * The core of the mbuf object along with some shortcut defines for practical 205107726Sru * purposes. 206103026Ssobomax */ 207107726Srustruct mbuf { 208103026Ssobomax /* 209107726Sru * Header present at the beginning of every mbuf. 210107726Sru * Size ILP32: 24 211107726Sru * LP64: 32 212107726Sru * Compile-time assertions in uipc_mbuf.c test these values to ensure 213107726Sru * that they are correct. 214156868Sbrd */ 215103026Ssobomax union { /* next buffer in chain */ 216103026Ssobomax struct mbuf *m_next; 217103026Ssobomax SLIST_ENTRY(mbuf) m_slist; 218103026Ssobomax STAILQ_ENTRY(mbuf) m_stailq; 219103026Ssobomax }; 220131530Sru union { /* next chain in queue/record */ 221103026Ssobomax struct mbuf *m_nextpkt; 222107726Sru SLIST_ENTRY(mbuf) m_slistpkt; 223107726Sru STAILQ_ENTRY(mbuf) m_stailqpkt; 224107726Sru }; 225107726Sru caddr_t m_data; /* location of data */ 226107726Sru int32_t m_len; /* amount of data in this mbuf */ 227107726Sru uint32_t m_type:8, /* type of data in this mbuf */ 228103026Ssobomax m_flags:24; /* flags; see below */ 229103026Ssobomax#if !defined(__LP64__) 230103026Ssobomax uint32_t m_pad; /* pad for 64bit alignment */ 231103026Ssobomax#endif 232103026Ssobomax 233103026Ssobomax /* 234107726Sru * A set of optional headers (packet header, external storage header) 235103026Ssobomax * and internal data storage. Historically, these arrays were sized 236107726Sru * to MHLEN (space left after a packet header) and MLEN (space left 237103026Ssobomax * after only a regular mbuf header); they are now variable size in 238103026Ssobomax * order to support future work on variable-size mbufs. 239103026Ssobomax */ 240103026Ssobomax union { 241103026Ssobomax struct { 242107726Sru struct pkthdr m_pkthdr; /* M_PKTHDR set */ 243107726Sru union { 244107726Sru struct m_ext m_ext; /* M_EXT set */ 245107726Sru char m_pktdat[0]; 246107726Sru }; 247103026Ssobomax }; 248103026Ssobomax char m_dat[0]; /* !M_PKTHDR, !M_EXT */ 249103026Ssobomax }; 250107726Sru}; 251107726Sru 252107726Sru/* 253107726Sru * mbuf flags of global significance and layer crossing. 254107726Sru * Those of only protocol/layer specific significance are to be mapped 255103026Ssobomax * to M_PROTO[1-12] and cleared at layer handoff boundaries. 256103026Ssobomax * NB: Limited to the lower 24 bits. 257107726Sru */ 258107726Sru#define M_EXT 0x00000001 /* has associated external storage */ 259107726Sru#define M_PKTHDR 0x00000002 /* start of record */ 260103026Ssobomax#define M_EOR 0x00000004 /* end of record */ 261103026Ssobomax#define M_RDONLY 0x00000008 /* associated data is marked read-only */ 262103026Ssobomax#define M_BCAST 0x00000010 /* send/received as link-level broadcast */ 263103026Ssobomax#define M_MCAST 0x00000020 /* send/received as link-level multicast */ 264107726Sru#define M_PROMISC 0x00000040 /* packet was not for us */ 265107726Sru#define M_VLANTAG 0x00000080 /* ether_vtag is valid */ 266179894Sthompsa#define M_UNUSED_8 0x00000100 /* --available-- */ 267103026Ssobomax#define M_NOFREE 0x00000200 /* do not free mbuf, embedded in cluster */ 268107726Sru 269107726Sru#define M_PROTO1 0x00001000 /* protocol-specific */ 270103026Ssobomax#define M_PROTO2 0x00002000 /* protocol-specific */ 271103026Ssobomax#define M_PROTO3 0x00004000 /* protocol-specific */ 272103026Ssobomax#define M_PROTO4 0x00008000 /* protocol-specific */ 273103026Ssobomax#define M_PROTO5 0x00010000 /* protocol-specific */ 274103026Ssobomax#define M_PROTO6 0x00020000 /* protocol-specific */ 275103026Ssobomax#define M_PROTO7 0x00040000 /* protocol-specific */ 276103026Ssobomax#define M_PROTO8 0x00080000 /* protocol-specific */ 277103026Ssobomax#define M_PROTO9 0x00100000 /* protocol-specific */ 278103026Ssobomax#define M_PROTO10 0x00200000 /* protocol-specific */ 279107726Sru#define M_PROTO11 0x00400000 /* protocol-specific */ 280103026Ssobomax#define M_PROTO12 0x00800000 /* protocol-specific */ 281103026Ssobomax 282103026Ssobomax#define MB_DTOR_SKIP 0x1 /* don't pollute the cache by touching a freed mbuf */ 283107726Sru 284103026Ssobomax/* 285103026Ssobomax * Flags to purge when crossing layers. 286103026Ssobomax */ 287103026Ssobomax#define M_PROTOFLAGS \ 288107726Sru (M_PROTO1|M_PROTO2|M_PROTO3|M_PROTO4|M_PROTO5|M_PROTO6|M_PROTO7|M_PROTO8|\ 289107726Sru M_PROTO9|M_PROTO10|M_PROTO11|M_PROTO12) 290107726Sru 291107726Sru/* 292103026Ssobomax * Flags preserved when copying m_pkthdr. 293103026Ssobomax */ 294107726Sru#define M_COPYFLAGS \ 295163500Sdanger (M_PKTHDR|M_EOR|M_RDONLY|M_BCAST|M_MCAST|M_PROMISC|M_VLANTAG| \ 296107726Sru M_PROTOFLAGS) 297107726Sru 298103026Ssobomax/* 299107726Sru * Mbuf flag description for use with printf(9) %b identifier. 300103026Ssobomax */ 301103026Ssobomax#define M_FLAG_BITS \ 302103026Ssobomax "\20\1M_EXT\2M_PKTHDR\3M_EOR\4M_RDONLY\5M_BCAST\6M_MCAST" \ 303103026Ssobomax "\7M_PROMISC\10M_VLANTAG" 304107726Sru#define M_FLAG_PROTOBITS \ 305103026Ssobomax "\15M_PROTO1\16M_PROTO2\17M_PROTO3\20M_PROTO4\21M_PROTO5" \ 306103026Ssobomax "\22M_PROTO6\23M_PROTO7\24M_PROTO8\25M_PROTO9\26M_PROTO10" \ 307103026Ssobomax "\27M_PROTO11\30M_PROTO12" 308103026Ssobomax#define M_FLAG_PRINTF (M_FLAG_BITS M_FLAG_PROTOBITS) 309103026Ssobomax 310103026Ssobomax/* 311103026Ssobomax * Network interface cards are able to hash protocol fields (such as IPv4 312103026Ssobomax * addresses and TCP port numbers) classify packets into flows. These flows 313103026Ssobomax * can then be used to maintain ordering while delivering packets to the OS 314103026Ssobomax * via parallel input queues, as well as to provide a stateless affinity 315107726Sru * model. NIC drivers can pass up the hash via m->m_pkthdr.flowid, and set 316107726Sru * m_flag fields to indicate how the hash should be interpreted by the 317107726Sru * network stack. 318107726Sru * 319107726Sru * Most NICs support RSS, which provides ordering and explicit affinity, and 320103026Ssobomax * use the hash m_flag bits to indicate what header fields were covered by 321107726Sru * the hash. M_HASHTYPE_OPAQUE and M_HASHTYPE_OPAQUE_HASH can be set by non- 322107726Sru * RSS cards or configurations that provide an opaque flow identifier, allowing 323103026Ssobomax * for ordering and distribution without explicit affinity. Additionally, 324107726Sru * M_HASHTYPE_OPAQUE_HASH indicates that the flow identifier has hash 325107726Sru * properties. 326107726Sru */ 327103026Ssobomax#define M_HASHTYPE_HASHPROP 0x80 /* has hash properties */ 328103026Ssobomax#define M_HASHTYPE_HASH(t) (M_HASHTYPE_HASHPROP | (t)) 329103026Ssobomax/* Microsoft RSS standard hash types */ 330103026Ssobomax#define M_HASHTYPE_NONE 0 331103026Ssobomax#define M_HASHTYPE_RSS_IPV4 M_HASHTYPE_HASH(1) /* IPv4 2-tuple */ 332107726Sru#define M_HASHTYPE_RSS_TCP_IPV4 M_HASHTYPE_HASH(2) /* TCPv4 4-tuple */ 333103026Ssobomax#define M_HASHTYPE_RSS_IPV6 M_HASHTYPE_HASH(3) /* IPv6 2-tuple */ 334103026Ssobomax#define M_HASHTYPE_RSS_TCP_IPV6 M_HASHTYPE_HASH(4) /* TCPv6 4-tuple */ 335179894Sthompsa#define M_HASHTYPE_RSS_IPV6_EX M_HASHTYPE_HASH(5) /* IPv6 2-tuple + 336210933Sjoel * ext hdrs */ 337179894Sthompsa#define M_HASHTYPE_RSS_TCP_IPV6_EX M_HASHTYPE_HASH(6) /* TCPv6 4-tuple + 338179894Sthompsa * ext hdrs */ 339179894Sthompsa/* Non-standard RSS hash types */ 340179894Sthompsa#define M_HASHTYPE_RSS_UDP_IPV4 M_HASHTYPE_HASH(7) /* IPv4 UDP 4-tuple*/ 341#define M_HASHTYPE_RSS_UDP_IPV4_EX M_HASHTYPE_HASH(8) /* IPv4 UDP 4-tuple + 342 * ext hdrs */ 343#define M_HASHTYPE_RSS_UDP_IPV6 M_HASHTYPE_HASH(9) /* IPv6 UDP 4-tuple*/ 344#define M_HASHTYPE_RSS_UDP_IPV6_EX M_HASHTYPE_HASH(10)/* IPv6 UDP 4-tuple + 345 * ext hdrs */ 346 347#define M_HASHTYPE_OPAQUE 63 /* ordering, not affinity */ 348#define M_HASHTYPE_OPAQUE_HASH M_HASHTYPE_HASH(M_HASHTYPE_OPAQUE) 349 /* ordering+hash, not affinity*/ 350 351#define M_HASHTYPE_CLEAR(m) ((m)->m_pkthdr.rsstype = 0) 352#define M_HASHTYPE_GET(m) ((m)->m_pkthdr.rsstype) 353#define M_HASHTYPE_SET(m, v) ((m)->m_pkthdr.rsstype = (v)) 354#define M_HASHTYPE_TEST(m, v) (M_HASHTYPE_GET(m) == (v)) 355#define M_HASHTYPE_ISHASH(m) (M_HASHTYPE_GET(m) & M_HASHTYPE_HASHPROP) 356 357/* 358 * COS/QOS class and quality of service tags. 359 * It uses DSCP code points as base. 360 */ 361#define QOS_DSCP_CS0 0x00 362#define QOS_DSCP_DEF QOS_DSCP_CS0 363#define QOS_DSCP_CS1 0x20 364#define QOS_DSCP_AF11 0x28 365#define QOS_DSCP_AF12 0x30 366#define QOS_DSCP_AF13 0x38 367#define QOS_DSCP_CS2 0x40 368#define QOS_DSCP_AF21 0x48 369#define QOS_DSCP_AF22 0x50 370#define QOS_DSCP_AF23 0x58 371#define QOS_DSCP_CS3 0x60 372#define QOS_DSCP_AF31 0x68 373#define QOS_DSCP_AF32 0x70 374#define QOS_DSCP_AF33 0x78 375#define QOS_DSCP_CS4 0x80 376#define QOS_DSCP_AF41 0x88 377#define QOS_DSCP_AF42 0x90 378#define QOS_DSCP_AF43 0x98 379#define QOS_DSCP_CS5 0xa0 380#define QOS_DSCP_EF 0xb8 381#define QOS_DSCP_CS6 0xc0 382#define QOS_DSCP_CS7 0xe0 383 384/* 385 * External mbuf storage buffer types. 386 */ 387#define EXT_CLUSTER 1 /* mbuf cluster */ 388#define EXT_SFBUF 2 /* sendfile(2)'s sf_buf */ 389#define EXT_JUMBOP 3 /* jumbo cluster page sized */ 390#define EXT_JUMBO9 4 /* jumbo cluster 9216 bytes */ 391#define EXT_JUMBO16 5 /* jumbo cluster 16184 bytes */ 392#define EXT_PACKET 6 /* mbuf+cluster from packet zone */ 393#define EXT_MBUF 7 /* external mbuf reference (M_IOVEC) */ 394#define EXT_SFBUF_NOCACHE 8 /* sendfile(2)'s sf_buf not to be cached */ 395 396#define EXT_VENDOR1 224 /* for vendor-internal use */ 397#define EXT_VENDOR2 225 /* for vendor-internal use */ 398#define EXT_VENDOR3 226 /* for vendor-internal use */ 399#define EXT_VENDOR4 227 /* for vendor-internal use */ 400 401#define EXT_EXP1 244 /* for experimental use */ 402#define EXT_EXP2 245 /* for experimental use */ 403#define EXT_EXP3 246 /* for experimental use */ 404#define EXT_EXP4 247 /* for experimental use */ 405 406#define EXT_NET_DRV 252 /* custom ext_buf provided by net driver(s) */ 407#define EXT_MOD_TYPE 253 /* custom module's ext_buf type */ 408#define EXT_DISPOSABLE 254 /* can throw this buffer away w/page flipping */ 409#define EXT_EXTREF 255 /* has externally maintained ext_cnt ptr */ 410 411/* 412 * Flags for external mbuf buffer types. 413 * NB: limited to the lower 24 bits. 414 */ 415#define EXT_FLAG_EMBREF 0x000001 /* embedded ext_count */ 416#define EXT_FLAG_EXTREF 0x000002 /* external ext_cnt, notyet */ 417 418#define EXT_FLAG_NOFREE 0x000010 /* don't free mbuf to pool, notyet */ 419 420#define EXT_FLAG_VENDOR1 0x010000 /* for vendor-internal use */ 421#define EXT_FLAG_VENDOR2 0x020000 /* for vendor-internal use */ 422#define EXT_FLAG_VENDOR3 0x040000 /* for vendor-internal use */ 423#define EXT_FLAG_VENDOR4 0x080000 /* for vendor-internal use */ 424 425#define EXT_FLAG_EXP1 0x100000 /* for experimental use */ 426#define EXT_FLAG_EXP2 0x200000 /* for experimental use */ 427#define EXT_FLAG_EXP3 0x400000 /* for experimental use */ 428#define EXT_FLAG_EXP4 0x800000 /* for experimental use */ 429 430/* 431 * EXT flag description for use with printf(9) %b identifier. 432 */ 433#define EXT_FLAG_BITS \ 434 "\20\1EXT_FLAG_EMBREF\2EXT_FLAG_EXTREF\5EXT_FLAG_NOFREE" \ 435 "\21EXT_FLAG_VENDOR1\22EXT_FLAG_VENDOR2\23EXT_FLAG_VENDOR3" \ 436 "\24EXT_FLAG_VENDOR4\25EXT_FLAG_EXP1\26EXT_FLAG_EXP2\27EXT_FLAG_EXP3" \ 437 "\30EXT_FLAG_EXP4" 438 439/* 440 * External reference/free functions. 441 */ 442void sf_ext_free(void *, void *); 443void sf_ext_free_nocache(void *, void *); 444 445/* 446 * Flags indicating checksum, segmentation and other offload work to be 447 * done, or already done, by hardware or lower layers. It is split into 448 * separate inbound and outbound flags. 449 * 450 * Outbound flags that are set by upper protocol layers requesting lower 451 * layers, or ideally the hardware, to perform these offloading tasks. 452 * For outbound packets this field and its flags can be directly tested 453 * against ifnet if_hwassist. 454 */ 455#define CSUM_IP 0x00000001 /* IP header checksum offload */ 456#define CSUM_IP_UDP 0x00000002 /* UDP checksum offload */ 457#define CSUM_IP_TCP 0x00000004 /* TCP checksum offload */ 458#define CSUM_IP_SCTP 0x00000008 /* SCTP checksum offload */ 459#define CSUM_IP_TSO 0x00000010 /* TCP segmentation offload */ 460#define CSUM_IP_ISCSI 0x00000020 /* iSCSI checksum offload */ 461 462#define CSUM_IP6_UDP 0x00000200 /* UDP checksum offload */ 463#define CSUM_IP6_TCP 0x00000400 /* TCP checksum offload */ 464#define CSUM_IP6_SCTP 0x00000800 /* SCTP checksum offload */ 465#define CSUM_IP6_TSO 0x00001000 /* TCP segmentation offload */ 466#define CSUM_IP6_ISCSI 0x00002000 /* iSCSI checksum offload */ 467 468/* Inbound checksum support where the checksum was verified by hardware. */ 469#define CSUM_L3_CALC 0x01000000 /* calculated layer 3 csum */ 470#define CSUM_L3_VALID 0x02000000 /* checksum is correct */ 471#define CSUM_L4_CALC 0x04000000 /* calculated layer 4 csum */ 472#define CSUM_L4_VALID 0x08000000 /* checksum is correct */ 473#define CSUM_L5_CALC 0x10000000 /* calculated layer 5 csum */ 474#define CSUM_L5_VALID 0x20000000 /* checksum is correct */ 475#define CSUM_COALESCED 0x40000000 /* contains merged segments */ 476 477/* 478 * CSUM flag description for use with printf(9) %b identifier. 479 */ 480#define CSUM_BITS \ 481 "\20\1CSUM_IP\2CSUM_IP_UDP\3CSUM_IP_TCP\4CSUM_IP_SCTP\5CSUM_IP_TSO" \ 482 "\6CSUM_IP_ISCSI" \ 483 "\12CSUM_IP6_UDP\13CSUM_IP6_TCP\14CSUM_IP6_SCTP\15CSUM_IP6_TSO" \ 484 "\16CSUM_IP6_ISCSI" \ 485 "\31CSUM_L3_CALC\32CSUM_L3_VALID\33CSUM_L4_CALC\34CSUM_L4_VALID" \ 486 "\35CSUM_L5_CALC\36CSUM_L5_VALID\37CSUM_COALESCED" 487 488/* CSUM flags compatibility mappings. */ 489#define CSUM_IP_CHECKED CSUM_L3_CALC 490#define CSUM_IP_VALID CSUM_L3_VALID 491#define CSUM_DATA_VALID CSUM_L4_VALID 492#define CSUM_PSEUDO_HDR CSUM_L4_CALC 493#define CSUM_SCTP_VALID CSUM_L4_VALID 494#define CSUM_DELAY_DATA (CSUM_TCP|CSUM_UDP) 495#define CSUM_DELAY_IP CSUM_IP /* Only v4, no v6 IP hdr csum */ 496#define CSUM_DELAY_DATA_IPV6 (CSUM_TCP_IPV6|CSUM_UDP_IPV6) 497#define CSUM_DATA_VALID_IPV6 CSUM_DATA_VALID 498#define CSUM_TCP CSUM_IP_TCP 499#define CSUM_UDP CSUM_IP_UDP 500#define CSUM_SCTP CSUM_IP_SCTP 501#define CSUM_TSO (CSUM_IP_TSO|CSUM_IP6_TSO) 502#define CSUM_UDP_IPV6 CSUM_IP6_UDP 503#define CSUM_TCP_IPV6 CSUM_IP6_TCP 504#define CSUM_SCTP_IPV6 CSUM_IP6_SCTP 505 506/* 507 * mbuf types describing the content of the mbuf (including external storage). 508 */ 509#define MT_NOTMBUF 0 /* USED INTERNALLY ONLY! Object is not mbuf */ 510#define MT_DATA 1 /* dynamic (data) allocation */ 511#define MT_HEADER MT_DATA /* packet header, use M_PKTHDR instead */ 512 513#define MT_VENDOR1 4 /* for vendor-internal use */ 514#define MT_VENDOR2 5 /* for vendor-internal use */ 515#define MT_VENDOR3 6 /* for vendor-internal use */ 516#define MT_VENDOR4 7 /* for vendor-internal use */ 517 518#define MT_SONAME 8 /* socket name */ 519 520#define MT_EXP1 9 /* for experimental use */ 521#define MT_EXP2 10 /* for experimental use */ 522#define MT_EXP3 11 /* for experimental use */ 523#define MT_EXP4 12 /* for experimental use */ 524 525#define MT_CONTROL 14 /* extra-data protocol message */ 526#define MT_OOBDATA 15 /* expedited data */ 527#define MT_NTYPES 16 /* number of mbuf types for mbtypes[] */ 528 529#define MT_NOINIT 255 /* Not a type but a flag to allocate 530 a non-initialized mbuf */ 531 532/* 533 * String names of mbuf-related UMA(9) and malloc(9) types. Exposed to 534 * !_KERNEL so that monitoring tools can look up the zones with 535 * libmemstat(3). 536 */ 537#define MBUF_MEM_NAME "mbuf" 538#define MBUF_CLUSTER_MEM_NAME "mbuf_cluster" 539#define MBUF_PACKET_MEM_NAME "mbuf_packet" 540#define MBUF_JUMBOP_MEM_NAME "mbuf_jumbo_page" 541#define MBUF_JUMBO9_MEM_NAME "mbuf_jumbo_9k" 542#define MBUF_JUMBO16_MEM_NAME "mbuf_jumbo_16k" 543#define MBUF_TAG_MEM_NAME "mbuf_tag" 544#define MBUF_EXTREFCNT_MEM_NAME "mbuf_ext_refcnt" 545 546#ifdef _KERNEL 547 548#ifdef WITNESS 549#define MBUF_CHECKSLEEP(how) do { \ 550 if (how == M_WAITOK) \ 551 WITNESS_WARN(WARN_GIANTOK | WARN_SLEEPOK, NULL, \ 552 "Sleeping in \"%s\"", __func__); \ 553} while (0) 554#else 555#define MBUF_CHECKSLEEP(how) 556#endif 557 558/* 559 * Network buffer allocation API 560 * 561 * The rest of it is defined in kern/kern_mbuf.c 562 */ 563extern uma_zone_t zone_mbuf; 564extern uma_zone_t zone_clust; 565extern uma_zone_t zone_pack; 566extern uma_zone_t zone_jumbop; 567extern uma_zone_t zone_jumbo9; 568extern uma_zone_t zone_jumbo16; 569 570void mb_dupcl(struct mbuf *, struct mbuf *); 571void mb_free_ext(struct mbuf *); 572void m_adj(struct mbuf *, int); 573int m_apply(struct mbuf *, int, int, 574 int (*)(void *, void *, u_int), void *); 575int m_append(struct mbuf *, int, c_caddr_t); 576void m_cat(struct mbuf *, struct mbuf *); 577void m_catpkt(struct mbuf *, struct mbuf *); 578int m_clget(struct mbuf *m, int how); 579void *m_cljget(struct mbuf *m, int how, int size); 580struct mbuf *m_collapse(struct mbuf *, int, int); 581void m_copyback(struct mbuf *, int, int, c_caddr_t); 582void m_copydata(const struct mbuf *, int, int, caddr_t); 583struct mbuf *m_copym(struct mbuf *, int, int, int); 584struct mbuf *m_copypacket(struct mbuf *, int); 585void m_copy_pkthdr(struct mbuf *, struct mbuf *); 586struct mbuf *m_copyup(struct mbuf *, int, int); 587struct mbuf *m_defrag(struct mbuf *, int); 588void m_demote_pkthdr(struct mbuf *); 589void m_demote(struct mbuf *, int, int); 590struct mbuf *m_devget(char *, int, int, struct ifnet *, 591 void (*)(char *, caddr_t, u_int)); 592struct mbuf *m_dup(const struct mbuf *, int); 593int m_dup_pkthdr(struct mbuf *, const struct mbuf *, int); 594void m_extadd(struct mbuf *, caddr_t, u_int, 595 void (*)(struct mbuf *, void *, void *), void *, void *, 596 int, int); 597u_int m_fixhdr(struct mbuf *); 598struct mbuf *m_fragment(struct mbuf *, int, int); 599void m_freem(struct mbuf *); 600struct mbuf *m_get2(int, int, short, int); 601struct mbuf *m_getjcl(int, short, int, int); 602struct mbuf *m_getm2(struct mbuf *, int, int, short, int); 603struct mbuf *m_getptr(struct mbuf *, int, int *); 604u_int m_length(struct mbuf *, struct mbuf **); 605int m_mbuftouio(struct uio *, struct mbuf *, int); 606void m_move_pkthdr(struct mbuf *, struct mbuf *); 607int m_pkthdr_init(struct mbuf *, int); 608struct mbuf *m_prepend(struct mbuf *, int, int); 609void m_print(const struct mbuf *, int); 610struct mbuf *m_pulldown(struct mbuf *, int, int, int *); 611struct mbuf *m_pullup(struct mbuf *, int); 612int m_sanity(struct mbuf *, int); 613struct mbuf *m_split(struct mbuf *, int, int); 614struct mbuf *m_uiotombuf(struct uio *, int, int, int, int); 615struct mbuf *m_unshare(struct mbuf *, int); 616 617static __inline int 618m_gettype(int size) 619{ 620 int type; 621 622 switch (size) { 623 case MSIZE: 624 type = EXT_MBUF; 625 break; 626 case MCLBYTES: 627 type = EXT_CLUSTER; 628 break; 629#if MJUMPAGESIZE != MCLBYTES 630 case MJUMPAGESIZE: 631 type = EXT_JUMBOP; 632 break; 633#endif 634 case MJUM9BYTES: 635 type = EXT_JUMBO9; 636 break; 637 case MJUM16BYTES: 638 type = EXT_JUMBO16; 639 break; 640 default: 641 panic("%s: invalid cluster size %d", __func__, size); 642 } 643 644 return (type); 645} 646 647/* 648 * Associated an external reference counted buffer with an mbuf. 649 */ 650static __inline void 651m_extaddref(struct mbuf *m, caddr_t buf, u_int size, u_int *ref_cnt, 652 void (*freef)(struct mbuf *, void *, void *), void *arg1, void *arg2) 653{ 654 655 KASSERT(ref_cnt != NULL, ("%s: ref_cnt not provided", __func__)); 656 657 atomic_add_int(ref_cnt, 1); 658 m->m_flags |= M_EXT; 659 m->m_ext.ext_buf = buf; 660 m->m_ext.ext_cnt = ref_cnt; 661 m->m_data = m->m_ext.ext_buf; 662 m->m_ext.ext_size = size; 663 m->m_ext.ext_free = freef; 664 m->m_ext.ext_arg1 = arg1; 665 m->m_ext.ext_arg2 = arg2; 666 m->m_ext.ext_type = EXT_EXTREF; 667 m->m_ext.ext_flags = 0; 668} 669 670static __inline uma_zone_t 671m_getzone(int size) 672{ 673 uma_zone_t zone; 674 675 switch (size) { 676 case MCLBYTES: 677 zone = zone_clust; 678 break; 679#if MJUMPAGESIZE != MCLBYTES 680 case MJUMPAGESIZE: 681 zone = zone_jumbop; 682 break; 683#endif 684 case MJUM9BYTES: 685 zone = zone_jumbo9; 686 break; 687 case MJUM16BYTES: 688 zone = zone_jumbo16; 689 break; 690 default: 691 panic("%s: invalid cluster size %d", __func__, size); 692 } 693 694 return (zone); 695} 696 697/* 698 * Initialize an mbuf with linear storage. 699 * 700 * Inline because the consumer text overhead will be roughly the same to 701 * initialize or call a function with this many parameters and M_PKTHDR 702 * should go away with constant propagation for !MGETHDR. 703 */ 704static __inline int 705m_init(struct mbuf *m, int how, short type, int flags) 706{ 707 int error; 708 709 m->m_next = NULL; 710 m->m_nextpkt = NULL; 711 m->m_data = m->m_dat; 712 m->m_len = 0; 713 m->m_flags = flags; 714 m->m_type = type; 715 if (flags & M_PKTHDR) 716 error = m_pkthdr_init(m, how); 717 else 718 error = 0; 719 720 MBUF_PROBE5(m__init, m, how, type, flags, error); 721 return (error); 722} 723 724static __inline struct mbuf * 725m_get(int how, short type) 726{ 727 struct mbuf *m; 728 struct mb_args args; 729 730 args.flags = 0; 731 args.type = type; 732 m = uma_zalloc_arg(zone_mbuf, &args, how); 733 MBUF_PROBE3(m__get, how, type, m); 734 return (m); 735} 736 737static __inline struct mbuf * 738m_gethdr(int how, short type) 739{ 740 struct mbuf *m; 741 struct mb_args args; 742 743 args.flags = M_PKTHDR; 744 args.type = type; 745 m = uma_zalloc_arg(zone_mbuf, &args, how); 746 MBUF_PROBE3(m__gethdr, how, type, m); 747 return (m); 748} 749 750static __inline struct mbuf * 751m_getcl(int how, short type, int flags) 752{ 753 struct mbuf *m; 754 struct mb_args args; 755 756 args.flags = flags; 757 args.type = type; 758 m = uma_zalloc_arg(zone_pack, &args, how); 759 MBUF_PROBE4(m__getcl, how, type, flags, m); 760 return (m); 761} 762 763/* 764 * XXX: m_cljset() is a dangerous API. One must attach only a new, 765 * unreferenced cluster to an mbuf(9). It is not possible to assert 766 * that, so care can be taken only by users of the API. 767 */ 768static __inline void 769m_cljset(struct mbuf *m, void *cl, int type) 770{ 771 int size; 772 773 switch (type) { 774 case EXT_CLUSTER: 775 size = MCLBYTES; 776 break; 777#if MJUMPAGESIZE != MCLBYTES 778 case EXT_JUMBOP: 779 size = MJUMPAGESIZE; 780 break; 781#endif 782 case EXT_JUMBO9: 783 size = MJUM9BYTES; 784 break; 785 case EXT_JUMBO16: 786 size = MJUM16BYTES; 787 break; 788 default: 789 panic("%s: unknown cluster type %d", __func__, type); 790 break; 791 } 792 793 m->m_data = m->m_ext.ext_buf = cl; 794 m->m_ext.ext_free = m->m_ext.ext_arg1 = m->m_ext.ext_arg2 = NULL; 795 m->m_ext.ext_size = size; 796 m->m_ext.ext_type = type; 797 m->m_ext.ext_flags = EXT_FLAG_EMBREF; 798 m->m_ext.ext_count = 1; 799 m->m_flags |= M_EXT; 800 MBUF_PROBE3(m__cljset, m, cl, type); 801} 802 803static __inline void 804m_chtype(struct mbuf *m, short new_type) 805{ 806 807 m->m_type = new_type; 808} 809 810static __inline void 811m_clrprotoflags(struct mbuf *m) 812{ 813 814 while (m) { 815 m->m_flags &= ~M_PROTOFLAGS; 816 m = m->m_next; 817 } 818} 819 820static __inline struct mbuf * 821m_last(struct mbuf *m) 822{ 823 824 while (m->m_next) 825 m = m->m_next; 826 return (m); 827} 828 829static inline u_int 830m_extrefcnt(struct mbuf *m) 831{ 832 833 KASSERT(m->m_flags & M_EXT, ("%s: M_EXT missing", __func__)); 834 835 return ((m->m_ext.ext_flags & EXT_FLAG_EMBREF) ? m->m_ext.ext_count : 836 *m->m_ext.ext_cnt); 837} 838 839/* 840 * mbuf, cluster, and external object allocation macros (for compatibility 841 * purposes). 842 */ 843#define M_MOVE_PKTHDR(to, from) m_move_pkthdr((to), (from)) 844#define MGET(m, how, type) ((m) = m_get((how), (type))) 845#define MGETHDR(m, how, type) ((m) = m_gethdr((how), (type))) 846#define MCLGET(m, how) m_clget((m), (how)) 847#define MEXTADD(m, buf, size, free, arg1, arg2, flags, type) \ 848 m_extadd((m), (caddr_t)(buf), (size), (free), (arg1), (arg2), \ 849 (flags), (type)) 850#define m_getm(m, len, how, type) \ 851 m_getm2((m), (len), (how), (type), M_PKTHDR) 852 853/* 854 * Evaluate TRUE if it's safe to write to the mbuf m's data region (this can 855 * be both the local data payload, or an external buffer area, depending on 856 * whether M_EXT is set). 857 */ 858#define M_WRITABLE(m) (!((m)->m_flags & M_RDONLY) && \ 859 (!(((m)->m_flags & M_EXT)) || \ 860 (m_extrefcnt(m) == 1))) 861 862/* Check if the supplied mbuf has a packet header, or else panic. */ 863#define M_ASSERTPKTHDR(m) \ 864 KASSERT((m) != NULL && (m)->m_flags & M_PKTHDR, \ 865 ("%s: no mbuf packet header!", __func__)) 866 867/* 868 * Ensure that the supplied mbuf is a valid, non-free mbuf. 869 * 870 * XXX: Broken at the moment. Need some UMA magic to make it work again. 871 */ 872#define M_ASSERTVALID(m) \ 873 KASSERT((((struct mbuf *)m)->m_flags & 0) == 0, \ 874 ("%s: attempted use of a free mbuf!", __func__)) 875 876/* 877 * Return the address of the start of the buffer associated with an mbuf, 878 * handling external storage, packet-header mbufs, and regular data mbufs. 879 */ 880#define M_START(m) \ 881 (((m)->m_flags & M_EXT) ? (m)->m_ext.ext_buf : \ 882 ((m)->m_flags & M_PKTHDR) ? &(m)->m_pktdat[0] : \ 883 &(m)->m_dat[0]) 884 885/* 886 * Return the size of the buffer associated with an mbuf, handling external 887 * storage, packet-header mbufs, and regular data mbufs. 888 */ 889#define M_SIZE(m) \ 890 (((m)->m_flags & M_EXT) ? (m)->m_ext.ext_size : \ 891 ((m)->m_flags & M_PKTHDR) ? MHLEN : \ 892 MLEN) 893 894/* 895 * Set the m_data pointer of a newly allocated mbuf to place an object of the 896 * specified size at the end of the mbuf, longword aligned. 897 * 898 * NB: Historically, we had M_ALIGN(), MH_ALIGN(), and MEXT_ALIGN() as 899 * separate macros, each asserting that it was called at the proper moment. 900 * This required callers to themselves test the storage type and call the 901 * right one. Rather than require callers to be aware of those layout 902 * decisions, we centralize here. 903 */ 904static __inline void 905m_align(struct mbuf *m, int len) 906{ 907#ifdef INVARIANTS 908 const char *msg = "%s: not a virgin mbuf"; 909#endif 910 int adjust; 911 912 KASSERT(m->m_data == M_START(m), (msg, __func__)); 913 914 adjust = M_SIZE(m) - len; 915 m->m_data += adjust &~ (sizeof(long)-1); 916} 917 918#define M_ALIGN(m, len) m_align(m, len) 919#define MH_ALIGN(m, len) m_align(m, len) 920#define MEXT_ALIGN(m, len) m_align(m, len) 921 922/* 923 * Compute the amount of space available before the current start of data in 924 * an mbuf. 925 * 926 * The M_WRITABLE() is a temporary, conservative safety measure: the burden 927 * of checking writability of the mbuf data area rests solely with the caller. 928 * 929 * NB: In previous versions, M_LEADINGSPACE() would only check M_WRITABLE() 930 * for mbufs with external storage. We now allow mbuf-embedded data to be 931 * read-only as well. 932 */ 933#define M_LEADINGSPACE(m) \ 934 (M_WRITABLE(m) ? ((m)->m_data - M_START(m)) : 0) 935 936/* 937 * Compute the amount of space available after the end of data in an mbuf. 938 * 939 * The M_WRITABLE() is a temporary, conservative safety measure: the burden 940 * of checking writability of the mbuf data area rests solely with the caller. 941 * 942 * NB: In previous versions, M_TRAILINGSPACE() would only check M_WRITABLE() 943 * for mbufs with external storage. We now allow mbuf-embedded data to be 944 * read-only as well. 945 */ 946#define M_TRAILINGSPACE(m) \ 947 (M_WRITABLE(m) ? \ 948 ((M_START(m) + M_SIZE(m)) - ((m)->m_data + (m)->m_len)) : 0) 949 950/* 951 * Arrange to prepend space of size plen to mbuf m. If a new mbuf must be 952 * allocated, how specifies whether to wait. If the allocation fails, the 953 * original mbuf chain is freed and m is set to NULL. 954 */ 955#define M_PREPEND(m, plen, how) do { \ 956 struct mbuf **_mmp = &(m); \ 957 struct mbuf *_mm = *_mmp; \ 958 int _mplen = (plen); \ 959 int __mhow = (how); \ 960 \ 961 MBUF_CHECKSLEEP(how); \ 962 if (M_LEADINGSPACE(_mm) >= _mplen) { \ 963 _mm->m_data -= _mplen; \ 964 _mm->m_len += _mplen; \ 965 } else \ 966 _mm = m_prepend(_mm, _mplen, __mhow); \ 967 if (_mm != NULL && _mm->m_flags & M_PKTHDR) \ 968 _mm->m_pkthdr.len += _mplen; \ 969 *_mmp = _mm; \ 970} while (0) 971 972/* 973 * Change mbuf to new type. This is a relatively expensive operation and 974 * should be avoided. 975 */ 976#define MCHTYPE(m, t) m_chtype((m), (t)) 977 978/* Length to m_copy to copy all. */ 979#define M_COPYALL 1000000000 980 981/* Compatibility with 4.3. */ 982#define m_copy(m, o, l) m_copym((m), (o), (l), M_NOWAIT) 983 984extern int max_datalen; /* MHLEN - max_hdr */ 985extern int max_hdr; /* Largest link + protocol header */ 986extern int max_linkhdr; /* Largest link-level header */ 987extern int max_protohdr; /* Largest protocol header */ 988extern int nmbclusters; /* Maximum number of clusters */ 989 990/*- 991 * Network packets may have annotations attached by affixing a list of 992 * "packet tags" to the pkthdr structure. Packet tags are dynamically 993 * allocated semi-opaque data structures that have a fixed header 994 * (struct m_tag) that specifies the size of the memory block and a 995 * <cookie,type> pair that identifies it. The cookie is a 32-bit unique 996 * unsigned value used to identify a module or ABI. By convention this value 997 * is chosen as the date+time that the module is created, expressed as the 998 * number of seconds since the epoch (e.g., using date -u +'%s'). The type 999 * value is an ABI/module-specific value that identifies a particular 1000 * annotation and is private to the module. For compatibility with systems 1001 * like OpenBSD that define packet tags w/o an ABI/module cookie, the value 1002 * PACKET_ABI_COMPAT is used to implement m_tag_get and m_tag_find 1003 * compatibility shim functions and several tag types are defined below. 1004 * Users that do not require compatibility should use a private cookie value 1005 * so that packet tag-related definitions can be maintained privately. 1006 * 1007 * Note that the packet tag returned by m_tag_alloc has the default memory 1008 * alignment implemented by malloc. To reference private data one can use a 1009 * construct like: 1010 * 1011 * struct m_tag *mtag = m_tag_alloc(...); 1012 * struct foo *p = (struct foo *)(mtag+1); 1013 * 1014 * if the alignment of struct m_tag is sufficient for referencing members of 1015 * struct foo. Otherwise it is necessary to embed struct m_tag within the 1016 * private data structure to insure proper alignment; e.g., 1017 * 1018 * struct foo { 1019 * struct m_tag tag; 1020 * ... 1021 * }; 1022 * struct foo *p = (struct foo *) m_tag_alloc(...); 1023 * struct m_tag *mtag = &p->tag; 1024 */ 1025 1026/* 1027 * Persistent tags stay with an mbuf until the mbuf is reclaimed. Otherwise 1028 * tags are expected to ``vanish'' when they pass through a network 1029 * interface. For most interfaces this happens normally as the tags are 1030 * reclaimed when the mbuf is free'd. However in some special cases 1031 * reclaiming must be done manually. An example is packets that pass through 1032 * the loopback interface. Also, one must be careful to do this when 1033 * ``turning around'' packets (e.g., icmp_reflect). 1034 * 1035 * To mark a tag persistent bit-or this flag in when defining the tag id. 1036 * The tag will then be treated as described above. 1037 */ 1038#define MTAG_PERSISTENT 0x800 1039 1040#define PACKET_TAG_NONE 0 /* Nadda */ 1041 1042/* Packet tags for use with PACKET_ABI_COMPAT. */ 1043#define PACKET_TAG_IPSEC_IN_DONE 1 /* IPsec applied, in */ 1044#define PACKET_TAG_IPSEC_OUT_DONE 2 /* IPsec applied, out */ 1045#define PACKET_TAG_IPSEC_IN_CRYPTO_DONE 3 /* NIC IPsec crypto done */ 1046#define PACKET_TAG_IPSEC_OUT_CRYPTO_NEEDED 4 /* NIC IPsec crypto req'ed */ 1047#define PACKET_TAG_IPSEC_IN_COULD_DO_CRYPTO 5 /* NIC notifies IPsec */ 1048#define PACKET_TAG_IPSEC_PENDING_TDB 6 /* Reminder to do IPsec */ 1049#define PACKET_TAG_BRIDGE 7 /* Bridge processing done */ 1050#define PACKET_TAG_GIF 8 /* GIF processing done */ 1051#define PACKET_TAG_GRE 9 /* GRE processing done */ 1052#define PACKET_TAG_IN_PACKET_CHECKSUM 10 /* NIC checksumming done */ 1053#define PACKET_TAG_ENCAP 11 /* Encap. processing */ 1054#define PACKET_TAG_IPSEC_SOCKET 12 /* IPSEC socket ref */ 1055#define PACKET_TAG_IPSEC_HISTORY 13 /* IPSEC history */ 1056#define PACKET_TAG_IPV6_INPUT 14 /* IPV6 input processing */ 1057#define PACKET_TAG_DUMMYNET 15 /* dummynet info */ 1058#define PACKET_TAG_DIVERT 17 /* divert info */ 1059#define PACKET_TAG_IPFORWARD 18 /* ipforward info */ 1060#define PACKET_TAG_MACLABEL (19 | MTAG_PERSISTENT) /* MAC label */ 1061#define PACKET_TAG_PF (21 | MTAG_PERSISTENT) /* PF/ALTQ information */ 1062#define PACKET_TAG_RTSOCKFAM 25 /* rtsock sa family */ 1063#define PACKET_TAG_IPOPTIONS 27 /* Saved IP options */ 1064#define PACKET_TAG_CARP 28 /* CARP info */ 1065#define PACKET_TAG_IPSEC_NAT_T_PORTS 29 /* two uint16_t */ 1066#define PACKET_TAG_ND_OUTGOING 30 /* ND outgoing */ 1067 1068/* Specific cookies and tags. */ 1069 1070/* Packet tag routines. */ 1071struct m_tag *m_tag_alloc(u_int32_t, int, int, int); 1072void m_tag_delete(struct mbuf *, struct m_tag *); 1073void m_tag_delete_chain(struct mbuf *, struct m_tag *); 1074void m_tag_free_default(struct m_tag *); 1075struct m_tag *m_tag_locate(struct mbuf *, u_int32_t, int, struct m_tag *); 1076struct m_tag *m_tag_copy(struct m_tag *, int); 1077int m_tag_copy_chain(struct mbuf *, const struct mbuf *, int); 1078void m_tag_delete_nonpersistent(struct mbuf *); 1079 1080/* 1081 * Initialize the list of tags associated with an mbuf. 1082 */ 1083static __inline void 1084m_tag_init(struct mbuf *m) 1085{ 1086 1087 SLIST_INIT(&m->m_pkthdr.tags); 1088} 1089 1090/* 1091 * Set up the contents of a tag. Note that this does not fill in the free 1092 * method; the caller is expected to do that. 1093 * 1094 * XXX probably should be called m_tag_init, but that was already taken. 1095 */ 1096static __inline void 1097m_tag_setup(struct m_tag *t, u_int32_t cookie, int type, int len) 1098{ 1099 1100 t->m_tag_id = type; 1101 t->m_tag_len = len; 1102 t->m_tag_cookie = cookie; 1103} 1104 1105/* 1106 * Reclaim resources associated with a tag. 1107 */ 1108static __inline void 1109m_tag_free(struct m_tag *t) 1110{ 1111 1112 (*t->m_tag_free)(t); 1113} 1114 1115/* 1116 * Return the first tag associated with an mbuf. 1117 */ 1118static __inline struct m_tag * 1119m_tag_first(struct mbuf *m) 1120{ 1121 1122 return (SLIST_FIRST(&m->m_pkthdr.tags)); 1123} 1124 1125/* 1126 * Return the next tag in the list of tags associated with an mbuf. 1127 */ 1128static __inline struct m_tag * 1129m_tag_next(struct mbuf *m __unused, struct m_tag *t) 1130{ 1131 1132 return (SLIST_NEXT(t, m_tag_link)); 1133} 1134 1135/* 1136 * Prepend a tag to the list of tags associated with an mbuf. 1137 */ 1138static __inline void 1139m_tag_prepend(struct mbuf *m, struct m_tag *t) 1140{ 1141 1142 SLIST_INSERT_HEAD(&m->m_pkthdr.tags, t, m_tag_link); 1143} 1144 1145/* 1146 * Unlink a tag from the list of tags associated with an mbuf. 1147 */ 1148static __inline void 1149m_tag_unlink(struct mbuf *m, struct m_tag *t) 1150{ 1151 1152 SLIST_REMOVE(&m->m_pkthdr.tags, t, m_tag, m_tag_link); 1153} 1154 1155/* These are for OpenBSD compatibility. */ 1156#define MTAG_ABI_COMPAT 0 /* compatibility ABI */ 1157 1158static __inline struct m_tag * 1159m_tag_get(int type, int length, int wait) 1160{ 1161 return (m_tag_alloc(MTAG_ABI_COMPAT, type, length, wait)); 1162} 1163 1164static __inline struct m_tag * 1165m_tag_find(struct mbuf *m, int type, struct m_tag *start) 1166{ 1167 return (SLIST_EMPTY(&m->m_pkthdr.tags) ? (struct m_tag *)NULL : 1168 m_tag_locate(m, MTAG_ABI_COMPAT, type, start)); 1169} 1170 1171static __inline struct mbuf * 1172m_free(struct mbuf *m) 1173{ 1174 struct mbuf *n = m->m_next; 1175 1176 MBUF_PROBE1(m__free, m); 1177 if ((m->m_flags & (M_PKTHDR|M_NOFREE)) == (M_PKTHDR|M_NOFREE)) 1178 m_tag_delete_chain(m, NULL); 1179 if (m->m_flags & M_EXT) 1180 mb_free_ext(m); 1181 else if ((m->m_flags & M_NOFREE) == 0) 1182 uma_zfree(zone_mbuf, m); 1183 return (n); 1184} 1185 1186static __inline int 1187rt_m_getfib(struct mbuf *m) 1188{ 1189 KASSERT(m->m_flags & M_PKTHDR , ("Attempt to get FIB from non header mbuf.")); 1190 return (m->m_pkthdr.fibnum); 1191} 1192 1193#define M_GETFIB(_m) rt_m_getfib(_m) 1194 1195#define M_SETFIB(_m, _fib) do { \ 1196 KASSERT((_m)->m_flags & M_PKTHDR, ("Attempt to set FIB on non header mbuf.")); \ 1197 ((_m)->m_pkthdr.fibnum) = (_fib); \ 1198} while (0) 1199 1200/* flags passed as first argument for "m_ether_tcpip_hash()" */ 1201#define MBUF_HASHFLAG_L2 (1 << 2) 1202#define MBUF_HASHFLAG_L3 (1 << 3) 1203#define MBUF_HASHFLAG_L4 (1 << 4) 1204 1205/* mbuf hashing helper routines */ 1206uint32_t m_ether_tcpip_hash_init(void); 1207uint32_t m_ether_tcpip_hash(const uint32_t, const struct mbuf *, const uint32_t); 1208 1209#ifdef MBUF_PROFILING 1210 void m_profile(struct mbuf *m); 1211 #define M_PROFILE(m) m_profile(m) 1212#else 1213 #define M_PROFILE(m) 1214#endif 1215 1216struct mbufq { 1217 STAILQ_HEAD(, mbuf) mq_head; 1218 int mq_len; 1219 int mq_maxlen; 1220}; 1221 1222static inline void 1223mbufq_init(struct mbufq *mq, int maxlen) 1224{ 1225 1226 STAILQ_INIT(&mq->mq_head); 1227 mq->mq_maxlen = maxlen; 1228 mq->mq_len = 0; 1229} 1230 1231static inline struct mbuf * 1232mbufq_flush(struct mbufq *mq) 1233{ 1234 struct mbuf *m; 1235 1236 m = STAILQ_FIRST(&mq->mq_head); 1237 STAILQ_INIT(&mq->mq_head); 1238 mq->mq_len = 0; 1239 return (m); 1240} 1241 1242static inline void 1243mbufq_drain(struct mbufq *mq) 1244{ 1245 struct mbuf *m, *n; 1246 1247 n = mbufq_flush(mq); 1248 while ((m = n) != NULL) { 1249 n = STAILQ_NEXT(m, m_stailqpkt); 1250 m_freem(m); 1251 } 1252} 1253 1254static inline struct mbuf * 1255mbufq_first(const struct mbufq *mq) 1256{ 1257 1258 return (STAILQ_FIRST(&mq->mq_head)); 1259} 1260 1261static inline struct mbuf * 1262mbufq_last(const struct mbufq *mq) 1263{ 1264 1265 return (STAILQ_LAST(&mq->mq_head, mbuf, m_stailqpkt)); 1266} 1267 1268static inline int 1269mbufq_full(const struct mbufq *mq) 1270{ 1271 1272 return (mq->mq_len >= mq->mq_maxlen); 1273} 1274 1275static inline int 1276mbufq_len(const struct mbufq *mq) 1277{ 1278 1279 return (mq->mq_len); 1280} 1281 1282static inline int 1283mbufq_enqueue(struct mbufq *mq, struct mbuf *m) 1284{ 1285 1286 if (mbufq_full(mq)) 1287 return (ENOBUFS); 1288 STAILQ_INSERT_TAIL(&mq->mq_head, m, m_stailqpkt); 1289 mq->mq_len++; 1290 return (0); 1291} 1292 1293static inline struct mbuf * 1294mbufq_dequeue(struct mbufq *mq) 1295{ 1296 struct mbuf *m; 1297 1298 m = STAILQ_FIRST(&mq->mq_head); 1299 if (m) { 1300 STAILQ_REMOVE_HEAD(&mq->mq_head, m_stailqpkt); 1301 m->m_nextpkt = NULL; 1302 mq->mq_len--; 1303 } 1304 return (m); 1305} 1306 1307static inline void 1308mbufq_prepend(struct mbufq *mq, struct mbuf *m) 1309{ 1310 1311 STAILQ_INSERT_HEAD(&mq->mq_head, m, m_stailqpkt); 1312 mq->mq_len++; 1313} 1314#endif /* _KERNEL */ 1315#endif /* !_SYS_MBUF_H_ */ 1316