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 #define M_PROMISC 0x20000 /* packet was not for us */ 186 187 /* 188 * External buffer types: identify ext_buf type. 189 */ 190 #define EXT_CLUSTER 1 /* mbuf cluster */ 191 #define EXT_SFBUF 2 /* sendfile(2)'s sf_bufs */ 192 #define EXT_JUMBOP 3 /* jumbo cluster 4096 bytes */ 193 #define EXT_JUMBO9 4 /* jumbo cluster 9216 bytes */ 194 #define EXT_JUMBO16 5 /* jumbo cluster 16184 bytes */ 195 #define EXT_PACKET 6 /* mbuf+cluster from packet zone */ 196 #define EXT_NET_DRV 100 /* custom ext_buf provided by net driver(s) */ 197 #define EXT_MOD_TYPE 200 /* custom module's ext_buf type */ 198 #define EXT_DISPOSABLE 300 /* can throw this buffer away w/page flipping */ 199 #define EXT_EXTREF 400 /* has externally maintained ref_cnt ptr */ 200 201 /* 202 * Flags copied when copying m_pkthdr. 203 */ 204 #define M_COPYFLAGS (M_PKTHDR|M_EOR|M_RDONLY|M_PROTO1|M_PROTO1|M_PROTO2|\ 205 M_PROTO3|M_PROTO4|M_PROTO5|M_SKIP_FIREWALL|\ 206 M_BCAST|M_MCAST|M_FRAG|M_FIRSTFRAG|M_LASTFRAG|\ 207 M_VLANTAG|M_PROMISC) 208 209 /* 210 * Flags to purge when crossing layers. 211 */ 212 #define M_PROTOFLAGS (M_PROTO1|M_PROTO2|M_PROTO3|M_PROTO4|M_PROTO5) 213 214 /* 215 * Flags indicating hw checksum support and sw checksum requirements. 216 * This field can be directly tested against if_data.ifi_hwassist. 217 */ 218 #define CSUM_IP 0x0001 /* will csum IP */ 219 #define CSUM_TCP 0x0002 /* will csum TCP */ 220 #define CSUM_UDP 0x0004 /* will csum UDP */ 221 #define CSUM_IP_FRAGS 0x0008 /* will csum IP fragments */ 222 #define CSUM_FRAGMENT 0x0010 /* will do IP fragmentation */ 223 #define CSUM_TSO 0x0020 /* will do TSO */ 224 225 #define CSUM_IP_CHECKED 0x0100 /* did csum IP */ 226 #define CSUM_IP_VALID 0x0200 /* ... the csum is valid */ 227 #define CSUM_DATA_VALID 0x0400 /* csum_data field is valid */ 228 #define CSUM_PSEUDO_HDR 0x0800 /* csum_data has pseudo hdr */ 229 230 #define CSUM_DELAY_DATA (CSUM_TCP | CSUM_UDP) 231 #define CSUM_DELAY_IP (CSUM_IP) /* XXX add ipv6 here too? */ 232 233 /* 234 * mbuf types. 235 */ 236 #define MT_NOTMBUF 0 /* USED INTERNALLY ONLY! Object is not mbuf */ 237 #define MT_DATA 1 /* dynamic (data) allocation */ 238 #define MT_HEADER MT_DATA /* packet header, use M_PKTHDR instead */ 239 #define MT_SONAME 8 /* socket name */ 240 #define MT_CONTROL 14 /* extra-data protocol message */ 241 #define MT_OOBDATA 15 /* expedited data */ 242 #define MT_NTYPES 16 /* number of mbuf types for mbtypes[] */ 243 244 #define MT_NOINIT 255 /* Not a type but a flag to allocate 245 a non-initialized mbuf */ 246 247 /* 248 * General mbuf allocator statistics structure. 249 * 250 * Many of these statistics are no longer used; we instead track many 251 * allocator statistics through UMA's built in statistics mechanism. 252 */ 253 struct mbstat { 254 u_long m_mbufs; /* XXX */ 255 u_long m_mclusts; /* XXX */ 256 257 u_long m_drain; /* times drained protocols for space */ 258 u_long m_mcfail; /* XXX: times m_copym failed */ 259 u_long m_mpfail; /* XXX: times m_pullup failed */ 260 u_long m_msize; /* length of an mbuf */ 261 u_long m_mclbytes; /* length of an mbuf cluster */ 262 u_long m_minclsize; /* min length of data to allocate a cluster */ 263 u_long m_mlen; /* length of data in an mbuf */ 264 u_long m_mhlen; /* length of data in a header mbuf */ 265 266 /* Number of mbtypes (gives # elems in mbtypes[] array: */ 267 short m_numtypes; 268 269 /* XXX: Sendfile stats should eventually move to their own struct */ 270 u_long sf_iocnt; /* times sendfile had to do disk I/O */ 271 u_long sf_allocfail; /* times sfbuf allocation failed */ 272 u_long sf_allocwait; /* times sfbuf allocation had to wait */ 273 }; 274 275 /* 276 * Flags specifying how an allocation should be made. 277 * 278 * The flag to use is as follows: 279 * - M_DONTWAIT or M_NOWAIT from an interrupt handler to not block allocation. 280 * - M_WAIT or M_WAITOK or M_TRYWAIT from wherever it is safe to block. 281 * 282 * M_DONTWAIT/M_NOWAIT means that we will not block the thread explicitly 283 * and if we cannot allocate immediately we may return NULL, 284 * whereas M_WAIT/M_WAITOK/M_TRYWAIT means that if we cannot allocate 285 * resources we will block until they are available, and thus never 286 * return NULL. 287 * 288 * XXX Eventually just phase this out to use M_WAITOK/M_NOWAIT. 289 */ 290 #define MBTOM(how) (how) 291 #define M_DONTWAIT M_NOWAIT 292 #define M_TRYWAIT M_WAITOK 293 #define M_WAIT M_WAITOK 294 295 /* 296 * String names of mbuf-related UMA(9) and malloc(9) types. Exposed to 297 * !_KERNEL so that monitoring tools can look up the zones with 298 * libmemstat(3). 299 */ 300 #define MBUF_MEM_NAME "mbuf" 301 #define MBUF_CLUSTER_MEM_NAME "mbuf_cluster" 302 #define MBUF_PACKET_MEM_NAME "mbuf_packet" 303 #define MBUF_JUMBOP_MEM_NAME "mbuf_jumbo_pagesize" 304 #define MBUF_JUMBO9_MEM_NAME "mbuf_jumbo_9k" 305 #define MBUF_JUMBO16_MEM_NAME "mbuf_jumbo_16k" 306 #define MBUF_TAG_MEM_NAME "mbuf_tag" 307 #define MBUF_EXTREFCNT_MEM_NAME "mbuf_ext_refcnt" 308 309 #ifdef _KERNEL 310 311 #ifdef WITNESS 312 #define MBUF_CHECKSLEEP(how) do { \ 313 if (how == M_WAITOK) \ 314 WITNESS_WARN(WARN_GIANTOK | WARN_SLEEPOK, NULL, \ 315 "Sleeping in \"%s\"", __func__); \ 316 } while (0) 317 #else 318 #define MBUF_CHECKSLEEP(how) 319 #endif 320 321 /* 322 * Network buffer allocation API 323 * 324 * The rest of it is defined in kern/kern_mbuf.c 325 */ 326 327 extern uma_zone_t zone_mbuf; 328 extern uma_zone_t zone_clust; 329 extern uma_zone_t zone_pack; 330 extern uma_zone_t zone_jumbop; 331 extern uma_zone_t zone_jumbo9; 332 extern uma_zone_t zone_jumbo16; 333 extern uma_zone_t zone_ext_refcnt; 334 335 static __inline struct mbuf *m_get(int how, short type); 336 static __inline struct mbuf *m_gethdr(int how, short type); 337 static __inline struct mbuf *m_getcl(int how, short type, int flags); 338 static __inline struct mbuf *m_getjcl(int how, short type, int flags, int size); 339 static __inline struct mbuf *m_getclr(int how, short type); /* XXX */ 340 static __inline struct mbuf *m_free(struct mbuf *m); 341 static __inline void m_clget(struct mbuf *m, int how); 342 static __inline void *m_cljget(struct mbuf *m, int how, int size); 343 static __inline void m_chtype(struct mbuf *m, short new_type); 344 void mb_free_ext(struct mbuf *); 345 346 static __inline 347 struct mbuf * 348 m_get(int how, short type) 349 { 350 struct mb_args args; 351 352 args.flags = 0; 353 args.type = type; 354 return (struct mbuf *)(uma_zalloc_arg(zone_mbuf, &args, how)); 355 } 356 357 /* XXX This should be depracated, very little use */ 358 static __inline 359 struct mbuf * 360 m_getclr(int how, short type) 361 { 362 struct mbuf *m; 363 struct mb_args args; 364 365 args.flags = 0; 366 args.type = type; 367 m = uma_zalloc_arg(zone_mbuf, &args, how); 368 if (m != NULL) 369 bzero(m->m_data, MLEN); 370 return m; 371 } 372 373 static __inline 374 struct mbuf * 375 m_gethdr(int how, short type) 376 { 377 struct mb_args args; 378 379 args.flags = M_PKTHDR; 380 args.type = type; 381 return (struct mbuf *)(uma_zalloc_arg(zone_mbuf, &args, how)); 382 } 383 384 static __inline 385 struct mbuf * 386 m_getcl(int how, short type, int flags) 387 { 388 struct mb_args args; 389 390 args.flags = flags; 391 args.type = type; 392 return (struct mbuf *)(uma_zalloc_arg(zone_pack, &args, how)); 393 } 394 395 /* 396 * m_getjcl() returns an mbuf with a cluster of the specified size attached. 397 * For size it takes MCLBYTES, MJUMPAGESIZE, MJUM9BYTES, MJUM16BYTES. 398 */ 399 static __inline /* XXX: This is rather large, should be real function maybe. */ 400 struct mbuf * 401 m_getjcl(int how, short type, int flags, int size) 402 { 403 struct mb_args args; 404 struct mbuf *m, *n; 405 uma_zone_t zone; 406 407 args.flags = flags; 408 args.type = type; 409 410 m = uma_zalloc_arg(zone_mbuf, &args, how); 411 if (m == NULL) 412 return NULL; 413 414 switch (size) { 415 case MCLBYTES: 416 zone = zone_clust; 417 break; 418 #if MJUMPAGESIZE != MCLBYTES 419 case MJUMPAGESIZE: 420 zone = zone_jumbop; 421 break; 422 #endif 423 case MJUM9BYTES: 424 zone = zone_jumbo9; 425 break; 426 case MJUM16BYTES: 427 zone = zone_jumbo16; 428 break; 429 default: 430 panic("%s: m_getjcl: invalid cluster type", __func__); 431 } 432 n = uma_zalloc_arg(zone, m, how); 433 if (n == NULL) { 434 uma_zfree(zone_mbuf, m); 435 return NULL; 436 } 437 return m; 438 } 439 440 static __inline 441 struct mbuf * 442 m_free(struct mbuf *m) 443 { 444 struct mbuf *n = m->m_next; 445 446 if (m->m_flags & M_EXT) 447 mb_free_ext(m); 448 else 449 uma_zfree(zone_mbuf, m); 450 return n; 451 } 452 453 static __inline 454 void 455 m_clget(struct mbuf *m, int how) 456 { 457 if (m->m_flags & M_EXT) 458 printf("%s: %p mbuf already has cluster\n", __func__, m); 459 m->m_ext.ext_buf = (char *)NULL; 460 uma_zalloc_arg(zone_clust, m, how); 461 /* 462 * On a cluster allocation failure, drain the packet zone and retry, 463 * we might be able to loosen a few clusters up on the drain. 464 */ 465 if ((how & M_NOWAIT) && (m->m_ext.ext_buf == NULL)) { 466 zone_drain(zone_pack); 467 uma_zalloc_arg(zone_clust, m, how); 468 } 469 } 470 471 /* 472 * m_cljget() is different from m_clget() as it can allocate clusters 473 * without attaching them to an mbuf. In that case the return value 474 * is the pointer to the cluster of the requested size. If an mbuf was 475 * specified, it gets the cluster attached to it and the return value 476 * can be safely ignored. 477 * For size it takes MCLBYTES, MJUMPAGESIZE, MJUM9BYTES, MJUM16BYTES. 478 */ 479 static __inline 480 void * 481 m_cljget(struct mbuf *m, int how, int size) 482 { 483 uma_zone_t zone; 484 485 if (m && m->m_flags & M_EXT) 486 printf("%s: %p mbuf already has cluster\n", __func__, m); 487 if (m != NULL) 488 m->m_ext.ext_buf = NULL; 489 490 switch (size) { 491 case MCLBYTES: 492 zone = zone_clust; 493 break; 494 #if MJUMPAGESIZE != MCLBYTES 495 case MJUMPAGESIZE: 496 zone = zone_jumbop; 497 break; 498 #endif 499 case MJUM9BYTES: 500 zone = zone_jumbo9; 501 break; 502 case MJUM16BYTES: 503 zone = zone_jumbo16; 504 break; 505 default: 506 panic("%s: m_getjcl: invalid cluster type", __func__); 507 } 508 509 return (uma_zalloc_arg(zone, m, how)); 510 } 511 512 static __inline 513 void 514 m_chtype(struct mbuf *m, short new_type) 515 { 516 m->m_type = new_type; 517 } 518 519 /* 520 * mbuf, cluster, and external object allocation macros 521 * (for compatibility purposes). 522 */ 523 /* NB: M_COPY_PKTHDR is deprecated. Use M_MOVE_PKTHDR or m_dup_pktdr. */ 524 #define M_MOVE_PKTHDR(to, from) m_move_pkthdr((to), (from)) 525 #define MGET(m, how, type) ((m) = m_get((how), (type))) 526 #define MGETHDR(m, how, type) ((m) = m_gethdr((how), (type))) 527 #define MCLGET(m, how) m_clget((m), (how)) 528 #define MEXTADD(m, buf, size, free, args, flags, type) \ 529 m_extadd((m), (caddr_t)(buf), (size), (free), (args), (flags), (type)) 530 #define m_getm(m, len, how, type) \ 531 m_getm2((m), (len), (how), (type), M_PKTHDR) 532 533 /* 534 * Evaluate TRUE if it's safe to write to the mbuf m's data region (this 535 * can be both the local data payload, or an external buffer area, 536 * depending on whether M_EXT is set). 537 */ 538 #define M_WRITABLE(m) (!((m)->m_flags & M_RDONLY) && \ 539 (!(((m)->m_flags & M_EXT)) || \ 540 (*((m)->m_ext.ref_cnt) == 1)) ) \ 541 542 /* Check if the supplied mbuf has a packet header, or else panic. */ 543 #define M_ASSERTPKTHDR(m) \ 544 KASSERT(m != NULL && m->m_flags & M_PKTHDR, \ 545 ("%s: no mbuf packet header!", __func__)) 546 547 /* Ensure that the supplied mbuf is a valid, non-free mbuf. */ 548 /* XXX: Broken at the moment. Need some UMA magic to make it work again. */ 549 #define M_ASSERTVALID(m) \ 550 KASSERT((((struct mbuf *)m)->m_flags & 0) == 0, \ 551 ("%s: attempted use of a free mbuf!", __func__)) 552 553 /* 554 * Set the m_data pointer of a newly-allocated mbuf (m_get/MGET) to place 555 * an object of the specified size at the end of the mbuf, longword aligned. 556 */ 557 #define M_ALIGN(m, len) do { \ 558 KASSERT(!((m)->m_flags & (M_PKTHDR|M_EXT)), \ 559 ("%s: M_ALIGN not normal mbuf", __func__)); \ 560 KASSERT((m)->m_data == (m)->m_dat, \ 561 ("%s: M_ALIGN not a virgin mbuf", __func__)); \ 562 (m)->m_data += (MLEN - (len)) & ~(sizeof(long) - 1); \ 563 } while (0) 564 565 /* 566 * As above, for mbufs allocated with m_gethdr/MGETHDR 567 * or initialized by M_COPY_PKTHDR. 568 */ 569 #define MH_ALIGN(m, len) do { \ 570 KASSERT((m)->m_flags & M_PKTHDR && !((m)->m_flags & M_EXT), \ 571 ("%s: MH_ALIGN not PKTHDR mbuf", __func__)); \ 572 KASSERT((m)->m_data == (m)->m_pktdat, \ 573 ("%s: MH_ALIGN not a virgin mbuf", __func__)); \ 574 (m)->m_data += (MHLEN - (len)) & ~(sizeof(long) - 1); \ 575 } while (0) 576 577 /* 578 * Compute the amount of space available 579 * before the current start of data in an mbuf. 580 * 581 * The M_WRITABLE() is a temporary, conservative safety measure: the burden 582 * of checking writability of the mbuf data area rests solely with the caller. 583 */ 584 #define M_LEADINGSPACE(m) \ 585 ((m)->m_flags & M_EXT ? \ 586 (M_WRITABLE(m) ? (m)->m_data - (m)->m_ext.ext_buf : 0): \ 587 (m)->m_flags & M_PKTHDR ? (m)->m_data - (m)->m_pktdat : \ 588 (m)->m_data - (m)->m_dat) 589 590 /* 591 * Compute the amount of space available 592 * after the end of data in an mbuf. 593 * 594 * The M_WRITABLE() is a temporary, conservative safety measure: the burden 595 * of checking writability of the mbuf data area rests solely with the caller. 596 */ 597 #define M_TRAILINGSPACE(m) \ 598 ((m)->m_flags & M_EXT ? \ 599 (M_WRITABLE(m) ? (m)->m_ext.ext_buf + (m)->m_ext.ext_size \ 600 - ((m)->m_data + (m)->m_len) : 0) : \ 601 &(m)->m_dat[MLEN] - ((m)->m_data + (m)->m_len)) 602 603 /* 604 * Arrange to prepend space of size plen to mbuf m. 605 * If a new mbuf must be allocated, how specifies whether to wait. 606 * If the allocation fails, the original mbuf chain is freed and m is 607 * set to NULL. 608 */ 609 #define M_PREPEND(m, plen, how) do { \ 610 struct mbuf **_mmp = &(m); \ 611 struct mbuf *_mm = *_mmp; \ 612 int _mplen = (plen); \ 613 int __mhow = (how); \ 614 \ 615 MBUF_CHECKSLEEP(how); \ 616 if (M_LEADINGSPACE(_mm) >= _mplen) { \ 617 _mm->m_data -= _mplen; \ 618 _mm->m_len += _mplen; \ 619 } else \ 620 _mm = m_prepend(_mm, _mplen, __mhow); \ 621 if (_mm != NULL && _mm->m_flags & M_PKTHDR) \ 622 _mm->m_pkthdr.len += _mplen; \ 623 *_mmp = _mm; \ 624 } while (0) 625 626 /* 627 * Change mbuf to new type. 628 * This is a relatively expensive operation and should be avoided. 629 */ 630 #define MCHTYPE(m, t) m_chtype((m), (t)) 631 632 /* Length to m_copy to copy all. */ 633 #define M_COPYALL 1000000000 634 635 /* Compatibility with 4.3. */ 636 #define m_copy(m, o, l) m_copym((m), (o), (l), M_DONTWAIT) 637 638 extern int max_datalen; /* MHLEN - max_hdr */ 639 extern int max_hdr; /* Largest link + protocol header */ 640 extern int max_linkhdr; /* Largest link-level header */ 641 extern int max_protohdr; /* Largest protocol header */ 642 extern struct mbstat mbstat; /* General mbuf stats/infos */ 643 extern int nmbclusters; /* Maximum number of clusters */ 644 645 struct uio; 646 647 void m_adj(struct mbuf *, int); 648 void m_align(struct mbuf *, int); 649 int m_apply(struct mbuf *, int, int, 650 int (*)(void *, void *, u_int), void *); 651 int m_append(struct mbuf *, int, c_caddr_t); 652 void m_cat(struct mbuf *, struct mbuf *); 653 void m_extadd(struct mbuf *, caddr_t, u_int, 654 void (*)(void *, void *), void *, int, int); 655 void m_copyback(struct mbuf *, int, int, c_caddr_t); 656 void m_copydata(const struct mbuf *, int, int, caddr_t); 657 struct mbuf *m_copym(struct mbuf *, int, int, int); 658 struct mbuf *m_copymdata(struct mbuf *, struct mbuf *, 659 int, int, int, int); 660 struct mbuf *m_copypacket(struct mbuf *, int); 661 void m_copy_pkthdr(struct mbuf *, struct mbuf *); 662 struct mbuf *m_copyup(struct mbuf *n, int len, int dstoff); 663 struct mbuf *m_defrag(struct mbuf *, int); 664 void m_demote(struct mbuf *, int); 665 struct mbuf *m_devget(char *, int, int, struct ifnet *, 666 void (*)(char *, caddr_t, u_int)); 667 struct mbuf *m_dup(struct mbuf *, int); 668 int m_dup_pkthdr(struct mbuf *, struct mbuf *, int); 669 u_int m_fixhdr(struct mbuf *); 670 struct mbuf *m_fragment(struct mbuf *, int, int); 671 void m_freem(struct mbuf *); 672 struct mbuf *m_getm2(struct mbuf *, int, int, short, int); 673 struct mbuf *m_getptr(struct mbuf *, int, int *); 674 u_int m_length(struct mbuf *, struct mbuf **); 675 void m_move_pkthdr(struct mbuf *, struct mbuf *); 676 struct mbuf *m_prepend(struct mbuf *, int, int); 677 void m_print(const struct mbuf *, int); 678 struct mbuf *m_pulldown(struct mbuf *, int, int, int *); 679 struct mbuf *m_pullup(struct mbuf *, int); 680 int m_sanity(struct mbuf *, int); 681 struct mbuf *m_split(struct mbuf *, int, int); 682 struct mbuf *m_uiotombuf(struct uio *, int, int, int, int); 683 struct mbuf *m_unshare(struct mbuf *, int how); 684 685 /*- 686 * Network packets may have annotations attached by affixing a list 687 * of "packet tags" to the pkthdr structure. Packet tags are 688 * dynamically allocated semi-opaque data structures that have 689 * a fixed header (struct m_tag) that specifies the size of the 690 * memory block and a <cookie,type> pair that identifies it. 691 * The cookie is a 32-bit unique unsigned value used to identify 692 * a module or ABI. By convention this value is chosen as the 693 * date+time that the module is created, expressed as the number of 694 * seconds since the epoch (e.g., using date -u +'%s'). The type value 695 * is an ABI/module-specific value that identifies a particular annotation 696 * and is private to the module. For compatibility with systems 697 * like OpenBSD that define packet tags w/o an ABI/module cookie, 698 * the value PACKET_ABI_COMPAT is used to implement m_tag_get and 699 * m_tag_find compatibility shim functions and several tag types are 700 * defined below. Users that do not require compatibility should use 701 * a private cookie value so that packet tag-related definitions 702 * can be maintained privately. 703 * 704 * Note that the packet tag returned by m_tag_alloc has the default 705 * memory alignment implemented by malloc. To reference private data 706 * one can use a construct like: 707 * 708 * struct m_tag *mtag = m_tag_alloc(...); 709 * struct foo *p = (struct foo *)(mtag+1); 710 * 711 * if the alignment of struct m_tag is sufficient for referencing members 712 * of struct foo. Otherwise it is necessary to embed struct m_tag within 713 * the private data structure to insure proper alignment; e.g., 714 * 715 * struct foo { 716 * struct m_tag tag; 717 * ... 718 * }; 719 * struct foo *p = (struct foo *) m_tag_alloc(...); 720 * struct m_tag *mtag = &p->tag; 721 */ 722 723 /* 724 * Persistent tags stay with an mbuf until the mbuf is reclaimed. 725 * Otherwise tags are expected to ``vanish'' when they pass through 726 * a network interface. For most interfaces this happens normally 727 * as the tags are reclaimed when the mbuf is free'd. However in 728 * some special cases reclaiming must be done manually. An example 729 * is packets that pass through the loopback interface. Also, one 730 * must be careful to do this when ``turning around'' packets (e.g., 731 * icmp_reflect). 732 * 733 * To mark a tag persistent bit-or this flag in when defining the 734 * tag id. The tag will then be treated as described above. 735 */ 736 #define MTAG_PERSISTENT 0x800 737 738 #define PACKET_TAG_NONE 0 /* Nadda */ 739 740 /* Packet tags for use with PACKET_ABI_COMPAT. */ 741 #define PACKET_TAG_IPSEC_IN_DONE 1 /* IPsec applied, in */ 742 #define PACKET_TAG_IPSEC_OUT_DONE 2 /* IPsec applied, out */ 743 #define PACKET_TAG_IPSEC_IN_CRYPTO_DONE 3 /* NIC IPsec crypto done */ 744 #define PACKET_TAG_IPSEC_OUT_CRYPTO_NEEDED 4 /* NIC IPsec crypto req'ed */ 745 #define PACKET_TAG_IPSEC_IN_COULD_DO_CRYPTO 5 /* NIC notifies IPsec */ 746 #define PACKET_TAG_IPSEC_PENDING_TDB 6 /* Reminder to do IPsec */ 747 #define PACKET_TAG_BRIDGE 7 /* Bridge processing done */ 748 #define PACKET_TAG_GIF 8 /* GIF processing done */ 749 #define PACKET_TAG_GRE 9 /* GRE processing done */ 750 #define PACKET_TAG_IN_PACKET_CHECKSUM 10 /* NIC checksumming done */ 751 #define PACKET_TAG_ENCAP 11 /* Encap. processing */ 752 #define PACKET_TAG_IPSEC_SOCKET 12 /* IPSEC socket ref */ 753 #define PACKET_TAG_IPSEC_HISTORY 13 /* IPSEC history */ 754 #define PACKET_TAG_IPV6_INPUT 14 /* IPV6 input processing */ 755 #define PACKET_TAG_DUMMYNET 15 /* dummynet info */ 756 #define PACKET_TAG_DIVERT 17 /* divert info */ 757 #define PACKET_TAG_IPFORWARD 18 /* ipforward info */ 758 #define PACKET_TAG_MACLABEL (19 | MTAG_PERSISTENT) /* MAC label */ 759 #define PACKET_TAG_PF_ROUTED 21 /* PF routed, avoid loops */ 760 #define PACKET_TAG_PF_FRAGCACHE 22 /* PF fragment cached */ 761 #define PACKET_TAG_PF_QID 23 /* PF ALTQ queue id */ 762 #define PACKET_TAG_PF_TAG 24 /* PF tagged */ 763 #define PACKET_TAG_RTSOCKFAM 25 /* rtsock sa family */ 764 #define PACKET_TAG_PF_TRANSLATE_LOCALHOST 26 /* PF translate localhost */ 765 #define PACKET_TAG_IPOPTIONS 27 /* Saved IP options */ 766 #define PACKET_TAG_CARP 28 /* CARP info */ 767 768 /* Specific cookies and tags. */ 769 770 /* Packet tag routines. */ 771 struct m_tag *m_tag_alloc(u_int32_t, int, int, int); 772 void m_tag_delete(struct mbuf *, struct m_tag *); 773 void m_tag_delete_chain(struct mbuf *, struct m_tag *); 774 void m_tag_free_default(struct m_tag *); 775 struct m_tag *m_tag_locate(struct mbuf *, u_int32_t, int, struct m_tag *); 776 struct m_tag *m_tag_copy(struct m_tag *, int); 777 int m_tag_copy_chain(struct mbuf *, struct mbuf *, int); 778 void m_tag_delete_nonpersistent(struct mbuf *); 779 780 /* 781 * Initialize the list of tags associated with an mbuf. 782 */ 783 static __inline void 784 m_tag_init(struct mbuf *m) 785 { 786 SLIST_INIT(&m->m_pkthdr.tags); 787 } 788 789 /* 790 * Set up the contents of a tag. Note that this does not 791 * fill in the free method; the caller is expected to do that. 792 * 793 * XXX probably should be called m_tag_init, but that was 794 * already taken. 795 */ 796 static __inline void 797 m_tag_setup(struct m_tag *t, u_int32_t cookie, int type, int len) 798 { 799 t->m_tag_id = type; 800 t->m_tag_len = len; 801 t->m_tag_cookie = cookie; 802 } 803 804 /* 805 * Reclaim resources associated with a tag. 806 */ 807 static __inline void 808 m_tag_free(struct m_tag *t) 809 { 810 (*t->m_tag_free)(t); 811 } 812 813 /* 814 * Return the first tag associated with an mbuf. 815 */ 816 static __inline struct m_tag * 817 m_tag_first(struct mbuf *m) 818 { 819 return (SLIST_FIRST(&m->m_pkthdr.tags)); 820 } 821 822 /* 823 * Return the next tag in the list of tags associated with an mbuf. 824 */ 825 static __inline struct m_tag * 826 m_tag_next(struct mbuf *m, struct m_tag *t) 827 { 828 return (SLIST_NEXT(t, m_tag_link)); 829 } 830 831 /* 832 * Prepend a tag to the list of tags associated with an mbuf. 833 */ 834 static __inline void 835 m_tag_prepend(struct mbuf *m, struct m_tag *t) 836 { 837 SLIST_INSERT_HEAD(&m->m_pkthdr.tags, t, m_tag_link); 838 } 839 840 /* 841 * Unlink a tag from the list of tags associated with an mbuf. 842 */ 843 static __inline void 844 m_tag_unlink(struct mbuf *m, struct m_tag *t) 845 { 846 SLIST_REMOVE(&m->m_pkthdr.tags, t, m_tag, m_tag_link); 847 } 848 849 /* These are for OpenBSD compatibility. */ 850 #define MTAG_ABI_COMPAT 0 /* compatibility ABI */ 851 852 static __inline struct m_tag * 853 m_tag_get(int type, int length, int wait) 854 { 855 return (m_tag_alloc(MTAG_ABI_COMPAT, type, length, wait)); 856 } 857 858 static __inline struct m_tag * 859 m_tag_find(struct mbuf *m, int type, struct m_tag *start) 860 { 861 return (SLIST_EMPTY(&m->m_pkthdr.tags) ? (struct m_tag *)NULL : 862 m_tag_locate(m, MTAG_ABI_COMPAT, type, start)); 863 } 864 865 #endif /* _KERNEL */ 866 867 #endif /* !_SYS_MBUF_H_ */ 868