mbuf.h revision 168638
1/*- 2 * Copyright (c) 1982, 1986, 1988, 1993 3 * The Regents of the University of California. 4 * All rights reserved. 5 * 6 * Redistribution and use in source and binary forms, with or without 7 * modification, are permitted provided that the following conditions 8 * are met: 9 * 1. Redistributions of source code must retain the above copyright 10 * notice, this list of conditions and the following disclaimer. 11 * 2. Redistributions in binary form must reproduce the above copyright 12 * notice, this list of conditions and the following disclaimer in the 13 * documentation and/or other materials provided with the distribution. 14 * 3. Neither the name of the University nor the names of its contributors 15 * may be used to endorse or promote products derived from this software 16 * without specific prior written permission. 17 * 18 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 19 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 20 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 21 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 22 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 23 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 24 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 25 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 26 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 27 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 28 * SUCH DAMAGE. 29 * 30 * @(#)mbuf.h 8.5 (Berkeley) 2/19/95 31 * $FreeBSD: head/sys/sys/mbuf.h 168638 2007-04-11 23:13:12Z andre $ 32 */ 33 34#ifndef _SYS_MBUF_H_ 35#define _SYS_MBUF_H_ 36 37/* XXX: These includes suck. Sorry! */ 38#include <sys/queue.h> 39#ifdef _KERNEL 40#include <sys/systm.h> 41#include <vm/uma.h> 42#ifdef WITNESS 43#include <sys/lock.h> 44#endif 45#endif 46 47/* 48 * Mbufs are of a single size, MSIZE (sys/param.h), which includes overhead. 49 * An mbuf may add a single "mbuf cluster" of size MCLBYTES (also in 50 * sys/param.h), which has no additional overhead and is used instead of the 51 * internal data area; this is done when at least MINCLSIZE of data must be 52 * stored. Additionally, it is possible to allocate a separate buffer 53 * externally and attach it to the mbuf in a way similar to that of mbuf 54 * clusters. 55 */ 56#define MLEN (MSIZE - sizeof(struct m_hdr)) /* normal data len */ 57#define MHLEN (MLEN - sizeof(struct pkthdr)) /* data len w/pkthdr */ 58#define MINCLSIZE (MHLEN + 1) /* smallest amount to put in cluster */ 59#define M_MAXCOMPRESS (MHLEN / 2) /* max amount to copy for compression */ 60 61#ifdef _KERNEL 62/*- 63 * Macros for type conversion: 64 * mtod(m, t) -- Convert mbuf pointer to data pointer of correct type. 65 * dtom(x) -- Convert data pointer within mbuf to mbuf pointer (XXX). 66 */ 67#define mtod(m, t) ((t)((m)->m_data)) 68#define dtom(x) ((struct mbuf *)((intptr_t)(x) & ~(MSIZE-1))) 69 70/* 71 * Argument structure passed to UMA routines during mbuf and packet 72 * allocations. 73 */ 74struct mb_args { 75 int flags; /* Flags for mbuf being allocated */ 76 short type; /* Type of mbuf being allocated */ 77}; 78#endif /* _KERNEL */ 79 80/* 81 * Header present at the beginning of every mbuf. 82 */ 83struct m_hdr { 84 struct mbuf *mh_next; /* next buffer in chain */ 85 struct mbuf *mh_nextpkt; /* next chain in queue/record */ 86 caddr_t mh_data; /* location of data */ 87 int mh_len; /* amount of data in this mbuf */ 88 int mh_flags; /* flags; see below */ 89 short mh_type; /* type of data in this mbuf */ 90}; 91 92/* 93 * Packet tag structure (see below for details). 94 */ 95struct m_tag { 96 SLIST_ENTRY(m_tag) m_tag_link; /* List of packet tags */ 97 u_int16_t m_tag_id; /* Tag ID */ 98 u_int16_t m_tag_len; /* Length of data */ 99 u_int32_t m_tag_cookie; /* ABI/Module ID */ 100 void (*m_tag_free)(struct m_tag *); 101}; 102 103/* 104 * Record/packet header in first mbuf of chain; valid only if M_PKTHDR is set. 105 */ 106struct pkthdr { 107 struct ifnet *rcvif; /* rcv interface */ 108 int len; /* total packet length */ 109 /* variables for ip and tcp reassembly */ 110 void *header; /* pointer to packet header */ 111 /* variables for hardware checksum */ 112 int csum_flags; /* flags regarding checksum */ 113 int csum_data; /* data field used by csum routines */ 114 u_int16_t tso_segsz; /* TSO segment size */ 115 u_int16_t ether_vtag; /* Ethernet 802.1p+q vlan tag */ 116 SLIST_HEAD(packet_tags, m_tag) tags; /* list of packet tags */ 117}; 118 119/* 120 * Description of external storage mapped into mbuf; valid only if M_EXT is 121 * set. 122 */ 123struct m_ext { 124 caddr_t ext_buf; /* start of buffer */ 125 void (*ext_free) /* free routine if not the usual */ 126 (void *, void *); 127 void *ext_args; /* optional argument pointer */ 128 u_int ext_size; /* size of buffer, for ext_free */ 129 volatile u_int *ref_cnt; /* pointer to ref count info */ 130 int ext_type; /* type of external storage */ 131}; 132 133/* 134 * The core of the mbuf object along with some shortcut defines for practical 135 * purposes. 136 */ 137struct mbuf { 138 struct m_hdr m_hdr; 139 union { 140 struct { 141 struct pkthdr MH_pkthdr; /* M_PKTHDR set */ 142 union { 143 struct m_ext MH_ext; /* M_EXT set */ 144 char MH_databuf[MHLEN]; 145 } MH_dat; 146 } MH; 147 char M_databuf[MLEN]; /* !M_PKTHDR, !M_EXT */ 148 } M_dat; 149}; 150#define m_next m_hdr.mh_next 151#define m_len m_hdr.mh_len 152#define m_data m_hdr.mh_data 153#define m_type m_hdr.mh_type 154#define m_flags m_hdr.mh_flags 155#define m_nextpkt m_hdr.mh_nextpkt 156#define m_act m_nextpkt 157#define m_pkthdr M_dat.MH.MH_pkthdr 158#define m_ext M_dat.MH.MH_dat.MH_ext 159#define m_pktdat M_dat.MH.MH_dat.MH_databuf 160#define m_dat M_dat.M_databuf 161 162/* 163 * mbuf flags. 164 */ 165#define M_EXT 0x0001 /* has associated external storage */ 166#define M_PKTHDR 0x0002 /* start of record */ 167#define M_EOR 0x0004 /* end of record */ 168#define M_RDONLY 0x0008 /* associated data is marked read-only */ 169#define M_PROTO1 0x0010 /* protocol-specific */ 170#define M_PROTO2 0x0020 /* protocol-specific */ 171#define M_PROTO3 0x0040 /* protocol-specific */ 172#define M_PROTO4 0x0080 /* protocol-specific */ 173#define M_PROTO5 0x0100 /* protocol-specific */ 174#define M_NOTIFICATION 0x2000 /* SCTP notification */ 175#define M_SKIP_FIREWALL 0x4000 /* skip firewall processing */ 176#define M_FREELIST 0x8000 /* mbuf is on the free list */ 177 178/* 179 * mbuf pkthdr flags (also stored in m_flags). 180 */ 181#define M_BCAST 0x0200 /* send/received as link-level broadcast */ 182#define M_MCAST 0x0400 /* send/received as link-level multicast */ 183#define M_FRAG 0x0800 /* packet is a fragment of a larger packet */ 184#define M_FIRSTFRAG 0x1000 /* packet is first fragment */ 185#define M_LASTFRAG 0x2000 /* packet is last fragment */ 186#define M_VLANTAG 0x10000 /* ether_vtag is valid */ 187#define M_PROMISC 0x20000 /* packet was not for us */ 188 189/* 190 * External buffer types: identify ext_buf type. 191 */ 192#define EXT_CLUSTER 1 /* mbuf cluster */ 193#define EXT_SFBUF 2 /* sendfile(2)'s sf_bufs */ 194#define EXT_JUMBOP 3 /* jumbo cluster 4096 bytes */ 195#define EXT_JUMBO9 4 /* jumbo cluster 9216 bytes */ 196#define EXT_JUMBO16 5 /* jumbo cluster 16184 bytes */ 197#define EXT_PACKET 6 /* mbuf+cluster from packet zone */ 198#define EXT_MBUF 7 /* external mbuf reference (M_IOVEC) */ 199#define EXT_NET_DRV 100 /* custom ext_buf provided by net driver(s) */ 200#define EXT_MOD_TYPE 200 /* custom module's ext_buf type */ 201#define EXT_DISPOSABLE 300 /* can throw this buffer away w/page flipping */ 202#define EXT_EXTREF 400 /* has externally maintained ref_cnt ptr */ 203 204/* 205 * Flags copied when copying m_pkthdr. 206 */ 207#define M_COPYFLAGS (M_PKTHDR|M_EOR|M_RDONLY|M_PROTO1|M_PROTO1|M_PROTO2|\ 208 M_PROTO3|M_PROTO4|M_PROTO5|M_SKIP_FIREWALL|\ 209 M_BCAST|M_MCAST|M_FRAG|M_FIRSTFRAG|M_LASTFRAG|\ 210 M_VLANTAG|M_PROMISC) 211 212/* 213 * Flags to purge when crossing layers. 214 */ 215#define M_PROTOFLAGS (M_PROTO1|M_PROTO2|M_PROTO3|M_PROTO4|M_PROTO5) 216 217/* 218 * Flags indicating hw checksum support and sw checksum requirements. This 219 * field can be directly tested against if_data.ifi_hwassist. 220 */ 221#define CSUM_IP 0x0001 /* will csum IP */ 222#define CSUM_TCP 0x0002 /* will csum TCP */ 223#define CSUM_UDP 0x0004 /* will csum UDP */ 224#define CSUM_IP_FRAGS 0x0008 /* will csum IP fragments */ 225#define CSUM_FRAGMENT 0x0010 /* will do IP fragmentation */ 226#define CSUM_TSO 0x0020 /* will do TSO */ 227 228#define CSUM_IP_CHECKED 0x0100 /* did csum IP */ 229#define CSUM_IP_VALID 0x0200 /* ... the csum is valid */ 230#define CSUM_DATA_VALID 0x0400 /* csum_data field is valid */ 231#define CSUM_PSEUDO_HDR 0x0800 /* csum_data has pseudo hdr */ 232 233#define CSUM_DELAY_DATA (CSUM_TCP | CSUM_UDP) 234#define CSUM_DELAY_IP (CSUM_IP) /* XXX add ipv6 here too? */ 235 236/* 237 * mbuf types. 238 */ 239#define MT_NOTMBUF 0 /* USED INTERNALLY ONLY! Object is not mbuf */ 240#define MT_DATA 1 /* dynamic (data) allocation */ 241#define MT_HEADER MT_DATA /* packet header, use M_PKTHDR instead */ 242#define MT_SONAME 8 /* socket name */ 243#define MT_CONTROL 14 /* extra-data protocol message */ 244#define MT_OOBDATA 15 /* expedited data */ 245#define MT_NTYPES 16 /* number of mbuf types for mbtypes[] */ 246 247#define MT_NOINIT 255 /* Not a type but a flag to allocate 248 a non-initialized mbuf */ 249 250/* 251 * General mbuf allocator statistics structure. 252 * 253 * Many of these statistics are no longer used; we instead track many 254 * allocator statistics through UMA's built in statistics mechanism. 255 */ 256struct mbstat { 257 u_long m_mbufs; /* XXX */ 258 u_long m_mclusts; /* XXX */ 259 260 u_long m_drain; /* times drained protocols for space */ 261 u_long m_mcfail; /* XXX: times m_copym failed */ 262 u_long m_mpfail; /* XXX: times m_pullup failed */ 263 u_long m_msize; /* length of an mbuf */ 264 u_long m_mclbytes; /* length of an mbuf cluster */ 265 u_long m_minclsize; /* min length of data to allocate a cluster */ 266 u_long m_mlen; /* length of data in an mbuf */ 267 u_long m_mhlen; /* length of data in a header mbuf */ 268 269 /* Number of mbtypes (gives # elems in mbtypes[] array: */ 270 short m_numtypes; 271 272 /* XXX: Sendfile stats should eventually move to their own struct */ 273 u_long sf_iocnt; /* times sendfile had to do disk I/O */ 274 u_long sf_allocfail; /* times sfbuf allocation failed */ 275 u_long sf_allocwait; /* times sfbuf allocation had to wait */ 276}; 277 278/* 279 * Flags specifying how an allocation should be made. 280 * 281 * The flag to use is as follows: 282 * - M_DONTWAIT or M_NOWAIT from an interrupt handler to not block allocation. 283 * - M_WAIT or M_WAITOK or M_TRYWAIT from wherever it is safe to block. 284 * 285 * M_DONTWAIT/M_NOWAIT means that we will not block the thread explicitly and 286 * if we cannot allocate immediately we may return NULL, whereas 287 * M_WAIT/M_WAITOK/M_TRYWAIT means that if we cannot allocate resources we 288 * will block until they are available, and thus never return NULL. 289 * 290 * XXX Eventually just phase this out to use M_WAITOK/M_NOWAIT. 291 */ 292#define MBTOM(how) (how) 293#define M_DONTWAIT M_NOWAIT 294#define M_TRYWAIT M_WAITOK 295#define M_WAIT M_WAITOK 296 297/* 298 * String names of mbuf-related UMA(9) and malloc(9) types. Exposed to 299 * !_KERNEL so that monitoring tools can look up the zones with 300 * libmemstat(3). 301 */ 302#define MBUF_MEM_NAME "mbuf" 303#define MBUF_CLUSTER_MEM_NAME "mbuf_cluster" 304#define MBUF_PACKET_MEM_NAME "mbuf_packet" 305#define MBUF_JUMBOP_MEM_NAME "mbuf_jumbo_pagesize" 306#define MBUF_JUMBO9_MEM_NAME "mbuf_jumbo_9k" 307#define MBUF_JUMBO16_MEM_NAME "mbuf_jumbo_16k" 308#define MBUF_TAG_MEM_NAME "mbuf_tag" 309#define MBUF_EXTREFCNT_MEM_NAME "mbuf_ext_refcnt" 310 311#ifdef _KERNEL 312 313#ifdef WITNESS 314#define MBUF_CHECKSLEEP(how) do { \ 315 if (how == M_WAITOK) \ 316 WITNESS_WARN(WARN_GIANTOK | WARN_SLEEPOK, NULL, \ 317 "Sleeping in \"%s\"", __func__); \ 318} while (0) 319#else 320#define MBUF_CHECKSLEEP(how) 321#endif 322 323/* 324 * Network buffer allocation API 325 * 326 * The rest of it is defined in kern/kern_mbuf.c 327 */ 328 329extern uma_zone_t zone_mbuf; 330extern uma_zone_t zone_clust; 331extern uma_zone_t zone_pack; 332extern uma_zone_t zone_jumbop; 333extern uma_zone_t zone_jumbo9; 334extern uma_zone_t zone_jumbo16; 335extern uma_zone_t zone_ext_refcnt; 336 337static __inline struct mbuf *m_get(int how, short type); 338static __inline struct mbuf *m_gethdr(int how, short type); 339static __inline struct mbuf *m_getcl(int how, short type, int flags); 340static __inline struct mbuf *m_getjcl(int how, short type, int flags, 341 int size); 342static __inline struct mbuf *m_getclr(int how, short type); /* XXX */ 343static __inline struct mbuf *m_free(struct mbuf *m); 344static __inline void m_clget(struct mbuf *m, int how); 345static __inline void *m_cljget(struct mbuf *m, int how, int size); 346static __inline void m_chtype(struct mbuf *m, short new_type); 347void mb_free_ext(struct mbuf *); 348static __inline struct mbuf *m_last(struct mbuf *m); 349 350static __inline int 351m_gettype(int size) 352{ 353 int type; 354 355 switch (size) { 356 case MSIZE: 357 type = EXT_MBUF; 358 break; 359 case MCLBYTES: 360 type = EXT_CLUSTER; 361 break; 362#if MJUMPAGESIZE != MCLBYTES 363 case MJUMPAGESIZE: 364 type = EXT_JUMBOP; 365 break; 366#endif 367 case MJUM9BYTES: 368 type = EXT_JUMBO9; 369 break; 370 case MJUM16BYTES: 371 type = EXT_JUMBO16; 372 break; 373 default: 374 panic("%s: m_getjcl: invalid cluster size", __func__); 375 } 376 377 return (type); 378} 379 380static __inline uma_zone_t 381m_getzone(int size) 382{ 383 uma_zone_t zone; 384 385 switch (size) { 386 case MSIZE: 387 zone = zone_mbuf; 388 break; 389 case MCLBYTES: 390 zone = zone_clust; 391 break; 392#if MJUMPAGESIZE != MCLBYTES 393 case MJUMPAGESIZE: 394 zone = zone_jumbop; 395 break; 396#endif 397 case MJUM9BYTES: 398 zone = zone_jumbo9; 399 break; 400 case MJUM16BYTES: 401 zone = zone_jumbo16; 402 break; 403 default: 404 panic("%s: m_getjcl: invalid cluster type", __func__); 405 } 406 407 return (zone); 408} 409 410static __inline struct mbuf * 411m_get(int how, short type) 412{ 413 struct mb_args args; 414 415 args.flags = 0; 416 args.type = type; 417 return ((struct mbuf *)(uma_zalloc_arg(zone_mbuf, &args, how))); 418} 419 420/* 421 * XXX This should be deprecated, very little use. 422 */ 423static __inline struct mbuf * 424m_getclr(int how, short type) 425{ 426 struct mbuf *m; 427 struct mb_args args; 428 429 args.flags = 0; 430 args.type = type; 431 m = uma_zalloc_arg(zone_mbuf, &args, how); 432 if (m != NULL) 433 bzero(m->m_data, MLEN); 434 return (m); 435} 436 437static __inline struct mbuf * 438m_gethdr(int how, short type) 439{ 440 struct mb_args args; 441 442 args.flags = M_PKTHDR; 443 args.type = type; 444 return ((struct mbuf *)(uma_zalloc_arg(zone_mbuf, &args, how))); 445} 446 447static __inline struct mbuf * 448m_getcl(int how, short type, int flags) 449{ 450 struct mb_args args; 451 452 args.flags = flags; 453 args.type = type; 454 return ((struct mbuf *)(uma_zalloc_arg(zone_pack, &args, how))); 455} 456 457/* 458 * m_getjcl() returns an mbuf with a cluster of the specified size attached. 459 * For size it takes MCLBYTES, MJUMPAGESIZE, MJUM9BYTES, MJUM16BYTES. 460 * 461 * XXX: This is rather large, should be real function maybe. 462 */ 463static __inline struct mbuf * 464m_getjcl(int how, short type, int flags, int size) 465{ 466 struct mb_args args; 467 struct mbuf *m, *n; 468 uma_zone_t zone; 469 470 args.flags = flags; 471 args.type = type; 472 473 m = uma_zalloc_arg(zone_mbuf, &args, how); 474 if (m == NULL) 475 return (NULL); 476 477 zone = m_getzone(size); 478 n = uma_zalloc_arg(zone, m, how); 479 if (n == NULL) { 480 uma_zfree(zone_mbuf, m); 481 return (NULL); 482 } 483 return (m); 484} 485 486static __inline struct mbuf * 487m_free(struct mbuf *m) 488{ 489 struct mbuf *n = m->m_next; 490 491 if (m->m_flags & M_EXT) 492 mb_free_ext(m); 493 else 494 uma_zfree(zone_mbuf, m); 495 return (n); 496} 497 498static __inline void 499m_clget(struct mbuf *m, int how) 500{ 501 502 if (m->m_flags & M_EXT) 503 printf("%s: %p mbuf already has cluster\n", __func__, m); 504 m->m_ext.ext_buf = (char *)NULL; 505 uma_zalloc_arg(zone_clust, m, how); 506 /* 507 * On a cluster allocation failure, drain the packet zone and retry, 508 * we might be able to loosen a few clusters up on the drain. 509 */ 510 if ((how & M_NOWAIT) && (m->m_ext.ext_buf == NULL)) { 511 zone_drain(zone_pack); 512 uma_zalloc_arg(zone_clust, m, how); 513 } 514} 515 516/* 517 * m_cljget() is different from m_clget() as it can allocate clusters without 518 * attaching them to an mbuf. In that case the return value is the pointer 519 * to the cluster of the requested size. If an mbuf was specified, it gets 520 * the cluster attached to it and the return value can be safely ignored. 521 * For size it takes MCLBYTES, MJUMPAGESIZE, MJUM9BYTES, MJUM16BYTES. 522 */ 523static __inline void * 524m_cljget(struct mbuf *m, int how, int size) 525{ 526 uma_zone_t zone; 527 528 if (m && m->m_flags & M_EXT) 529 printf("%s: %p mbuf already has cluster\n", __func__, m); 530 if (m != NULL) 531 m->m_ext.ext_buf = NULL; 532 533 zone = m_getzone(size); 534 return (uma_zalloc_arg(zone, m, how)); 535} 536 537static __inline void 538m_cljset(struct mbuf *m, void *cl, int type) 539{ 540 uma_zone_t zone; 541 int size; 542 543 switch (type) { 544 case EXT_CLUSTER: 545 size = MCLBYTES; 546 zone = zone_clust; 547 break; 548#if MJUMPAGESIZE != MCLBYTES 549 case EXT_JUMBOP: 550 size = MJUMPAGESIZE; 551 zone = zone_jumbop; 552 break; 553#endif 554 case EXT_JUMBO9: 555 size = MJUM9BYTES; 556 zone = zone_jumbo9; 557 break; 558 case EXT_JUMBO16: 559 size = MJUM16BYTES; 560 zone = zone_jumbo16; 561 break; 562 default: 563 panic("unknown cluster type"); 564 break; 565 } 566 567 m->m_data = m->m_ext.ext_buf = cl; 568 m->m_ext.ext_free = m->m_ext.ext_args = NULL; 569 m->m_ext.ext_size = size; 570 m->m_ext.ext_type = type; 571 m->m_ext.ref_cnt = uma_find_refcnt(zone, cl); 572 m->m_flags |= M_EXT; 573 574} 575 576static __inline void 577m_chtype(struct mbuf *m, short new_type) 578{ 579 580 m->m_type = new_type; 581} 582 583static __inline struct mbuf * 584m_last(struct mbuf *m) 585{ 586 587 while (m->m_next) 588 m = m->m_next; 589 return (m); 590} 591 592/* 593 * mbuf, cluster, and external object allocation macros (for compatibility 594 * purposes). 595 */ 596#define M_MOVE_PKTHDR(to, from) m_move_pkthdr((to), (from)) 597#define MGET(m, how, type) ((m) = m_get((how), (type))) 598#define MGETHDR(m, how, type) ((m) = m_gethdr((how), (type))) 599#define MCLGET(m, how) m_clget((m), (how)) 600#define MEXTADD(m, buf, size, free, args, flags, type) \ 601 m_extadd((m), (caddr_t)(buf), (size), (free), (args), (flags), (type)) 602#define m_getm(m, len, how, type) \ 603 m_getm2((m), (len), (how), (type), M_PKTHDR) 604 605/* 606 * Evaluate TRUE if it's safe to write to the mbuf m's data region (this can 607 * be both the local data payload, or an external buffer area, depending on 608 * whether M_EXT is set). 609 */ 610#define M_WRITABLE(m) (!((m)->m_flags & M_RDONLY) && \ 611 (!(((m)->m_flags & M_EXT)) || \ 612 (*((m)->m_ext.ref_cnt) == 1)) ) \ 613 614/* Check if the supplied mbuf has a packet header, or else panic. */ 615#define M_ASSERTPKTHDR(m) \ 616 KASSERT(m != NULL && m->m_flags & M_PKTHDR, \ 617 ("%s: no mbuf packet header!", __func__)) 618 619/* 620 * Ensure that the supplied mbuf is a valid, non-free mbuf. 621 * 622 * XXX: Broken at the moment. Need some UMA magic to make it work again. 623 */ 624#define M_ASSERTVALID(m) \ 625 KASSERT((((struct mbuf *)m)->m_flags & 0) == 0, \ 626 ("%s: attempted use of a free mbuf!", __func__)) 627 628/* 629 * Set the m_data pointer of a newly-allocated mbuf (m_get/MGET) to place an 630 * object of the specified size at the end of the mbuf, longword aligned. 631 */ 632#define M_ALIGN(m, len) do { \ 633 KASSERT(!((m)->m_flags & (M_PKTHDR|M_EXT)), \ 634 ("%s: M_ALIGN not normal mbuf", __func__)); \ 635 KASSERT((m)->m_data == (m)->m_dat, \ 636 ("%s: M_ALIGN not a virgin mbuf", __func__)); \ 637 (m)->m_data += (MLEN - (len)) & ~(sizeof(long) - 1); \ 638} while (0) 639 640/* 641 * As above, for mbufs allocated with m_gethdr/MGETHDR or initialized by 642 * M_DUP/MOVE_PKTHDR. 643 */ 644#define MH_ALIGN(m, len) do { \ 645 KASSERT((m)->m_flags & M_PKTHDR && !((m)->m_flags & M_EXT), \ 646 ("%s: MH_ALIGN not PKTHDR mbuf", __func__)); \ 647 KASSERT((m)->m_data == (m)->m_pktdat, \ 648 ("%s: MH_ALIGN not a virgin mbuf", __func__)); \ 649 (m)->m_data += (MHLEN - (len)) & ~(sizeof(long) - 1); \ 650} while (0) 651 652/* 653 * Compute the amount of space available before the current start of data in 654 * an mbuf. 655 * 656 * The M_WRITABLE() is a temporary, conservative safety measure: the burden 657 * of checking writability of the mbuf data area rests solely with the caller. 658 */ 659#define M_LEADINGSPACE(m) \ 660 ((m)->m_flags & M_EXT ? \ 661 (M_WRITABLE(m) ? (m)->m_data - (m)->m_ext.ext_buf : 0): \ 662 (m)->m_flags & M_PKTHDR ? (m)->m_data - (m)->m_pktdat : \ 663 (m)->m_data - (m)->m_dat) 664 665/* 666 * Compute the amount of space available after the end of data in an mbuf. 667 * 668 * The M_WRITABLE() is a temporary, conservative safety measure: the burden 669 * of checking writability of the mbuf data area rests solely with the caller. 670 */ 671#define M_TRAILINGSPACE(m) \ 672 ((m)->m_flags & M_EXT ? \ 673 (M_WRITABLE(m) ? (m)->m_ext.ext_buf + (m)->m_ext.ext_size \ 674 - ((m)->m_data + (m)->m_len) : 0) : \ 675 &(m)->m_dat[MLEN] - ((m)->m_data + (m)->m_len)) 676 677/* 678 * Arrange to prepend space of size plen to mbuf m. If a new mbuf must be 679 * allocated, how specifies whether to wait. If the allocation fails, the 680 * original mbuf chain is freed and m is set to NULL. 681 */ 682#define M_PREPEND(m, plen, how) do { \ 683 struct mbuf **_mmp = &(m); \ 684 struct mbuf *_mm = *_mmp; \ 685 int _mplen = (plen); \ 686 int __mhow = (how); \ 687 \ 688 MBUF_CHECKSLEEP(how); \ 689 if (M_LEADINGSPACE(_mm) >= _mplen) { \ 690 _mm->m_data -= _mplen; \ 691 _mm->m_len += _mplen; \ 692 } else \ 693 _mm = m_prepend(_mm, _mplen, __mhow); \ 694 if (_mm != NULL && _mm->m_flags & M_PKTHDR) \ 695 _mm->m_pkthdr.len += _mplen; \ 696 *_mmp = _mm; \ 697} while (0) 698 699/* 700 * Change mbuf to new type. This is a relatively expensive operation and 701 * should be avoided. 702 */ 703#define MCHTYPE(m, t) m_chtype((m), (t)) 704 705/* Length to m_copy to copy all. */ 706#define M_COPYALL 1000000000 707 708/* Compatibility with 4.3. */ 709#define m_copy(m, o, l) m_copym((m), (o), (l), M_DONTWAIT) 710 711extern int max_datalen; /* MHLEN - max_hdr */ 712extern int max_hdr; /* Largest link + protocol header */ 713extern int max_linkhdr; /* Largest link-level header */ 714extern int max_protohdr; /* Largest protocol header */ 715extern struct mbstat mbstat; /* General mbuf stats/infos */ 716extern int nmbclusters; /* Maximum number of clusters */ 717 718struct uio; 719 720void m_adj(struct mbuf *, int); 721void m_align(struct mbuf *, int); 722int m_apply(struct mbuf *, int, int, 723 int (*)(void *, void *, u_int), void *); 724int m_append(struct mbuf *, int, c_caddr_t); 725void m_cat(struct mbuf *, struct mbuf *); 726void m_extadd(struct mbuf *, caddr_t, u_int, 727 void (*)(void *, void *), void *, int, int); 728void m_copyback(struct mbuf *, int, int, c_caddr_t); 729void m_copydata(const struct mbuf *, int, int, caddr_t); 730struct mbuf *m_copym(struct mbuf *, int, int, int); 731struct mbuf *m_copymdata(struct mbuf *, struct mbuf *, 732 int, int, int, int); 733struct mbuf *m_copypacket(struct mbuf *, int); 734void m_copy_pkthdr(struct mbuf *, struct mbuf *); 735struct mbuf *m_copyup(struct mbuf *n, int len, int dstoff); 736struct mbuf *m_defrag(struct mbuf *, int); 737void m_demote(struct mbuf *, int); 738struct mbuf *m_devget(char *, int, int, struct ifnet *, 739 void (*)(char *, caddr_t, u_int)); 740struct mbuf *m_dup(struct mbuf *, int); 741int m_dup_pkthdr(struct mbuf *, struct mbuf *, int); 742u_int m_fixhdr(struct mbuf *); 743struct mbuf *m_fragment(struct mbuf *, int, int); 744void m_freem(struct mbuf *); 745struct mbuf *m_getm2(struct mbuf *, int, int, short, int); 746struct mbuf *m_getptr(struct mbuf *, int, int *); 747u_int m_length(struct mbuf *, struct mbuf **); 748void m_move_pkthdr(struct mbuf *, struct mbuf *); 749struct mbuf *m_prepend(struct mbuf *, int, int); 750void m_print(const struct mbuf *, int); 751struct mbuf *m_pulldown(struct mbuf *, int, int, int *); 752struct mbuf *m_pullup(struct mbuf *, int); 753int m_sanity(struct mbuf *, int); 754struct mbuf *m_split(struct mbuf *, int, int); 755struct mbuf *m_uiotombuf(struct uio *, int, int, int, int); 756struct mbuf *m_unshare(struct mbuf *, int how); 757 758/*- 759 * Network packets may have annotations attached by affixing a list of 760 * "packet tags" to the pkthdr structure. Packet tags are dynamically 761 * allocated semi-opaque data structures that have a fixed header 762 * (struct m_tag) that specifies the size of the memory block and a 763 * <cookie,type> pair that identifies it. The cookie is a 32-bit unique 764 * unsigned value used to identify a module or ABI. By convention this value 765 * is chosen as the date+time that the module is created, expressed as the 766 * number of seconds since the epoch (e.g., using date -u +'%s'). The type 767 * value is an ABI/module-specific value that identifies a particular 768 * annotation and is private to the module. For compatibility with systems 769 * like OpenBSD that define packet tags w/o an ABI/module cookie, the value 770 * PACKET_ABI_COMPAT is used to implement m_tag_get and m_tag_find 771 * compatibility shim functions and several tag types are defined below. 772 * Users that do not require compatibility should use a private cookie value 773 * so that packet tag-related definitions can be maintained privately. 774 * 775 * Note that the packet tag returned by m_tag_alloc has the default memory 776 * alignment implemented by malloc. To reference private data one can use a 777 * construct like: 778 * 779 * struct m_tag *mtag = m_tag_alloc(...); 780 * struct foo *p = (struct foo *)(mtag+1); 781 * 782 * if the alignment of struct m_tag is sufficient for referencing members of 783 * struct foo. Otherwise it is necessary to embed struct m_tag within the 784 * private data structure to insure proper alignment; e.g., 785 * 786 * struct foo { 787 * struct m_tag tag; 788 * ... 789 * }; 790 * struct foo *p = (struct foo *) m_tag_alloc(...); 791 * struct m_tag *mtag = &p->tag; 792 */ 793 794/* 795 * Persistent tags stay with an mbuf until the mbuf is reclaimed. Otherwise 796 * tags are expected to ``vanish'' when they pass through a network 797 * interface. For most interfaces this happens normally as the tags are 798 * reclaimed when the mbuf is free'd. However in some special cases 799 * reclaiming must be done manually. An example is packets that pass through 800 * the loopback interface. Also, one must be careful to do this when 801 * ``turning around'' packets (e.g., icmp_reflect). 802 * 803 * To mark a tag persistent bit-or this flag in when defining the tag id. 804 * The tag will then be treated as described above. 805 */ 806#define MTAG_PERSISTENT 0x800 807 808#define PACKET_TAG_NONE 0 /* Nadda */ 809 810/* Packet tags for use with PACKET_ABI_COMPAT. */ 811#define PACKET_TAG_IPSEC_IN_DONE 1 /* IPsec applied, in */ 812#define PACKET_TAG_IPSEC_OUT_DONE 2 /* IPsec applied, out */ 813#define PACKET_TAG_IPSEC_IN_CRYPTO_DONE 3 /* NIC IPsec crypto done */ 814#define PACKET_TAG_IPSEC_OUT_CRYPTO_NEEDED 4 /* NIC IPsec crypto req'ed */ 815#define PACKET_TAG_IPSEC_IN_COULD_DO_CRYPTO 5 /* NIC notifies IPsec */ 816#define PACKET_TAG_IPSEC_PENDING_TDB 6 /* Reminder to do IPsec */ 817#define PACKET_TAG_BRIDGE 7 /* Bridge processing done */ 818#define PACKET_TAG_GIF 8 /* GIF processing done */ 819#define PACKET_TAG_GRE 9 /* GRE processing done */ 820#define PACKET_TAG_IN_PACKET_CHECKSUM 10 /* NIC checksumming done */ 821#define PACKET_TAG_ENCAP 11 /* Encap. processing */ 822#define PACKET_TAG_IPSEC_SOCKET 12 /* IPSEC socket ref */ 823#define PACKET_TAG_IPSEC_HISTORY 13 /* IPSEC history */ 824#define PACKET_TAG_IPV6_INPUT 14 /* IPV6 input processing */ 825#define PACKET_TAG_DUMMYNET 15 /* dummynet info */ 826#define PACKET_TAG_DIVERT 17 /* divert info */ 827#define PACKET_TAG_IPFORWARD 18 /* ipforward info */ 828#define PACKET_TAG_MACLABEL (19 | MTAG_PERSISTENT) /* MAC label */ 829#define PACKET_TAG_PF_ROUTED 21 /* PF routed, avoid loops */ 830#define PACKET_TAG_PF_FRAGCACHE 22 /* PF fragment cached */ 831#define PACKET_TAG_PF_QID 23 /* PF ALTQ queue id */ 832#define PACKET_TAG_PF_TAG 24 /* PF tagged */ 833#define PACKET_TAG_RTSOCKFAM 25 /* rtsock sa family */ 834#define PACKET_TAG_PF_TRANSLATE_LOCALHOST 26 /* PF translate localhost */ 835#define PACKET_TAG_IPOPTIONS 27 /* Saved IP options */ 836#define PACKET_TAG_CARP 28 /* CARP info */ 837 838/* Specific cookies and tags. */ 839 840/* Packet tag routines. */ 841struct m_tag *m_tag_alloc(u_int32_t, int, int, int); 842void m_tag_delete(struct mbuf *, struct m_tag *); 843void m_tag_delete_chain(struct mbuf *, struct m_tag *); 844void m_tag_free_default(struct m_tag *); 845struct m_tag *m_tag_locate(struct mbuf *, u_int32_t, int, struct m_tag *); 846struct m_tag *m_tag_copy(struct m_tag *, int); 847int m_tag_copy_chain(struct mbuf *, struct mbuf *, int); 848void m_tag_delete_nonpersistent(struct mbuf *); 849 850/* 851 * Initialize the list of tags associated with an mbuf. 852 */ 853static __inline void 854m_tag_init(struct mbuf *m) 855{ 856 857 SLIST_INIT(&m->m_pkthdr.tags); 858} 859 860/* 861 * Set up the contents of a tag. Note that this does not fill in the free 862 * method; the caller is expected to do that. 863 * 864 * XXX probably should be called m_tag_init, but that was already taken. 865 */ 866static __inline void 867m_tag_setup(struct m_tag *t, u_int32_t cookie, int type, int len) 868{ 869 870 t->m_tag_id = type; 871 t->m_tag_len = len; 872 t->m_tag_cookie = cookie; 873} 874 875/* 876 * Reclaim resources associated with a tag. 877 */ 878static __inline void 879m_tag_free(struct m_tag *t) 880{ 881 882 (*t->m_tag_free)(t); 883} 884 885/* 886 * Return the first tag associated with an mbuf. 887 */ 888static __inline struct m_tag * 889m_tag_first(struct mbuf *m) 890{ 891 892 return (SLIST_FIRST(&m->m_pkthdr.tags)); 893} 894 895/* 896 * Return the next tag in the list of tags associated with an mbuf. 897 */ 898static __inline struct m_tag * 899m_tag_next(struct mbuf *m, struct m_tag *t) 900{ 901 902 return (SLIST_NEXT(t, m_tag_link)); 903} 904 905/* 906 * Prepend a tag to the list of tags associated with an mbuf. 907 */ 908static __inline void 909m_tag_prepend(struct mbuf *m, struct m_tag *t) 910{ 911 912 SLIST_INSERT_HEAD(&m->m_pkthdr.tags, t, m_tag_link); 913} 914 915/* 916 * Unlink a tag from the list of tags associated with an mbuf. 917 */ 918static __inline void 919m_tag_unlink(struct mbuf *m, struct m_tag *t) 920{ 921 922 SLIST_REMOVE(&m->m_pkthdr.tags, t, m_tag, m_tag_link); 923} 924 925/* These are for OpenBSD compatibility. */ 926#define MTAG_ABI_COMPAT 0 /* compatibility ABI */ 927 928static __inline struct m_tag * 929m_tag_get(int type, int length, int wait) 930{ 931 return (m_tag_alloc(MTAG_ABI_COMPAT, type, length, wait)); 932} 933 934static __inline struct m_tag * 935m_tag_find(struct mbuf *m, int type, struct m_tag *start) 936{ 937 return (SLIST_EMPTY(&m->m_pkthdr.tags) ? (struct m_tag *)NULL : 938 m_tag_locate(m, MTAG_ABI_COMPAT, type, start)); 939} 940 941#endif /* _KERNEL */ 942 943#endif /* !_SYS_MBUF_H_ */ 944