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