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 * $FreeBSD$ 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 */ 74 struct mb_args { 75 int flags; /* Flags for mbuf being allocated */ 76 short type; /* Type of mbuf being allocated */ 77 }; 78 #endif /* _KERNEL */ 79 80 #if defined(__LP64__) 81 #define M_HDR_PAD 6 82 #else 83 #define M_HDR_PAD 2 84 #endif 85 86 /* 87 * Header present at the beginning of every mbuf. 88 */ 89 struct m_hdr { 90 struct mbuf *mh_next; /* next buffer in chain */ 91 struct mbuf *mh_nextpkt; /* next chain in queue/record */ 92 caddr_t mh_data; /* location of data */ 93 int mh_len; /* amount of data in this mbuf */ 94 int mh_flags; /* flags; see below */ 95 short mh_type; /* type of data in this mbuf */ 96 uint8_t pad[M_HDR_PAD];/* word align */ 97 }; 98 99 /* 100 * Packet tag structure (see below for details). 101 */ 102 struct m_tag { 103 SLIST_ENTRY(m_tag) m_tag_link; /* List of packet tags */ 104 u_int16_t m_tag_id; /* Tag ID */ 105 u_int16_t m_tag_len; /* Length of data */ 106 u_int32_t m_tag_cookie; /* ABI/Module ID */ 107 void (*m_tag_free)(struct m_tag *); 108 }; 109 110 /* 111 * Record/packet header in first mbuf of chain; valid only if M_PKTHDR is set. 112 */ 113 struct pkthdr { 114 struct ifnet *rcvif; /* rcv interface */ 115 /* variables for ip and tcp reassembly */ 116 void *header; /* pointer to packet header */ 117 int len; /* total packet length */ 118 uint32_t flowid; /* packet's 4-tuple system 119 * flow identifier 120 */ 121 /* variables for hardware checksum */ 122 int csum_flags; /* flags regarding checksum */ 123 int csum_data; /* data field used by csum routines */ 124 u_int16_t tso_segsz; /* TSO segment size */ 125 u_int16_t ether_vtag; /* Ethernet 802.1p+q vlan tag */ 126 SLIST_HEAD(packet_tags, m_tag) tags; /* list of packet tags */ 127 }; 128 129 /* 130 * Description of external storage mapped into mbuf; valid only if M_EXT is 131 * set. 132 */ 133 struct m_ext { 134 caddr_t ext_buf; /* start of buffer */ 135 void (*ext_free) /* free routine if not the usual */ 136 (void *, void *); 137 void *ext_arg1; /* optional argument pointer */ 138 void *ext_arg2; /* optional argument pointer */ 139 u_int ext_size; /* size of buffer, for ext_free */ 140 volatile u_int *ref_cnt; /* pointer to ref count info */ 141 int ext_type; /* type of external storage */ 142 }; 143 144 /* 145 * The core of the mbuf object along with some shortcut defines for practical 146 * purposes. 147 */ 148 struct mbuf { 149 struct m_hdr m_hdr; 150 union { 151 struct { 152 struct pkthdr MH_pkthdr; /* M_PKTHDR set */ 153 union { 154 struct m_ext MH_ext; /* M_EXT set */ 155 char MH_databuf[MHLEN]; 156 } MH_dat; 157 } MH; 158 char M_databuf[MLEN]; /* !M_PKTHDR, !M_EXT */ 159 } M_dat; 160 }; 161 #define m_next m_hdr.mh_next 162 #define m_len m_hdr.mh_len 163 #define m_data m_hdr.mh_data 164 #define m_type m_hdr.mh_type 165 #define m_flags m_hdr.mh_flags 166 #define m_nextpkt m_hdr.mh_nextpkt 167 #define m_act m_nextpkt 168 #define m_pkthdr M_dat.MH.MH_pkthdr 169 #define m_ext M_dat.MH.MH_dat.MH_ext 170 #define m_pktdat M_dat.MH.MH_dat.MH_databuf 171 #define m_dat M_dat.M_databuf 172 173 /* 174 * mbuf flags. 175 */ 176 #define M_EXT 0x00000001 /* has associated external storage */ 177 #define M_PKTHDR 0x00000002 /* start of record */ 178 #define M_EOR 0x00000004 /* end of record */ 179 #define M_RDONLY 0x00000008 /* associated data is marked read-only */ 180 #define M_PROTO1 0x00000010 /* protocol-specific */ 181 #define M_PROTO2 0x00000020 /* protocol-specific */ 182 #define M_PROTO3 0x00000040 /* protocol-specific */ 183 #define M_PROTO4 0x00000080 /* protocol-specific */ 184 #define M_PROTO5 0x00000100 /* protocol-specific */ 185 #define M_BCAST 0x00000200 /* send/received as link-level broadcast */ 186 #define M_MCAST 0x00000400 /* send/received as link-level multicast */ 187 #define M_FRAG 0x00000800 /* packet is a fragment of a larger packet */ 188 #define M_FIRSTFRAG 0x00001000 /* packet is first fragment */ 189 #define M_LASTFRAG 0x00002000 /* packet is last fragment */ 190 #define M_SKIP_FIREWALL 0x00004000 /* skip firewall processing */ 191 #define M_FREELIST 0x00008000 /* mbuf is on the free list */ 192 #define M_VLANTAG 0x00010000 /* ether_vtag is valid */ 193 #define M_PROMISC 0x00020000 /* packet was not for us */ 194 #define M_NOFREE 0x00040000 /* do not free mbuf, embedded in cluster */ 195 #define M_PROTO6 0x00080000 /* protocol-specific */ 196 #define M_PROTO7 0x00100000 /* protocol-specific */ 197 #define M_PROTO8 0x00200000 /* protocol-specific */ 198 /* 199 * For RELENG_{6,7} steal these flags for limited multiple routing table 200 * support. In RELENG_8 and beyond, use just one flag and a tag. 201 */ 202 #define M_FIB 0xF0000000 /* steal some bits to store fib number. */ 203 204 #define M_NOTIFICATION M_PROTO5 /* SCTP notification */ 205 206 /* 207 * Flags to purge when crossing layers. 208 */ 209 #define M_PROTOFLAGS \ 210 (M_PROTO1|M_PROTO2|M_PROTO3|M_PROTO4|M_PROTO5|M_PROTO6|M_PROTO7|M_PROTO8) 211 212 /* 213 * Flags preserved when copying m_pkthdr. 214 */ 215 #define M_COPYFLAGS \ 216 (M_PKTHDR|M_EOR|M_RDONLY|M_PROTOFLAGS|M_SKIP_FIREWALL|M_BCAST|M_MCAST|\ 217 M_FRAG|M_FIRSTFRAG|M_LASTFRAG|M_VLANTAG|M_PROMISC|M_FIB) 218 219 /* 220 * External buffer types: identify ext_buf type. 221 */ 222 #define EXT_CLUSTER 1 /* mbuf cluster */ 223 #define EXT_SFBUF 2 /* sendfile(2)'s sf_bufs */ 224 #define EXT_JUMBOP 3 /* jumbo cluster 4096 bytes */ 225 #define EXT_JUMBO9 4 /* jumbo cluster 9216 bytes */ 226 #define EXT_JUMBO16 5 /* jumbo cluster 16184 bytes */ 227 #define EXT_PACKET 6 /* mbuf+cluster from packet zone */ 228 #define EXT_MBUF 7 /* external mbuf reference (M_IOVEC) */ 229 #define EXT_NET_DRV 100 /* custom ext_buf provided by net driver(s) */ 230 #define EXT_MOD_TYPE 200 /* custom module's ext_buf type */ 231 #define EXT_DISPOSABLE 300 /* can throw this buffer away w/page flipping */ 232 #define EXT_EXTREF 400 /* has externally maintained ref_cnt ptr */ 233 234 /* 235 * Flags indicating hw checksum support and sw checksum requirements. This 236 * field can be directly tested against if_data.ifi_hwassist. 237 */ 238 #define CSUM_IP 0x0001 /* will csum IP */ 239 #define CSUM_TCP 0x0002 /* will csum TCP */ 240 #define CSUM_UDP 0x0004 /* will csum UDP */ 241 #define CSUM_IP_FRAGS 0x0008 /* will csum IP fragments */ 242 #define CSUM_FRAGMENT 0x0010 /* will do IP fragmentation */ 243 #define CSUM_TSO 0x0020 /* will do TSO */ 244 245 #define CSUM_IP_CHECKED 0x0100 /* did csum IP */ 246 #define CSUM_IP_VALID 0x0200 /* ... the csum is valid */ 247 #define CSUM_DATA_VALID 0x0400 /* csum_data field is valid */ 248 #define CSUM_PSEUDO_HDR 0x0800 /* csum_data has pseudo hdr */ 249 250 #define CSUM_DELAY_DATA (CSUM_TCP | CSUM_UDP) 251 #define CSUM_DELAY_IP (CSUM_IP) /* XXX add ipv6 here too? */ 252 253 /* 254 * mbuf types. 255 */ 256 #define MT_NOTMBUF 0 /* USED INTERNALLY ONLY! Object is not mbuf */ 257 #define MT_DATA 1 /* dynamic (data) allocation */ 258 #define MT_HEADER MT_DATA /* packet header, use M_PKTHDR instead */ 259 #define MT_SONAME 8 /* socket name */ 260 #define MT_CONTROL 14 /* extra-data protocol message */ 261 #define MT_OOBDATA 15 /* expedited data */ 262 #define MT_NTYPES 16 /* number of mbuf types for mbtypes[] */ 263 264 #define MT_NOINIT 255 /* Not a type but a flag to allocate 265 a non-initialized mbuf */ 266 267 #define MB_NOTAGS 0x1UL /* no tags attached to mbuf */ 268 269 /* 270 * General mbuf allocator statistics structure. 271 * 272 * Many of these statistics are no longer used; we instead track many 273 * allocator statistics through UMA's built in statistics mechanism. 274 */ 275 struct mbstat { 276 u_long m_mbufs; /* XXX */ 277 u_long m_mclusts; /* XXX */ 278 279 u_long m_drain; /* times drained protocols for space */ 280 u_long m_mcfail; /* XXX: times m_copym failed */ 281 u_long m_mpfail; /* XXX: times m_pullup failed */ 282 u_long m_msize; /* length of an mbuf */ 283 u_long m_mclbytes; /* length of an mbuf cluster */ 284 u_long m_minclsize; /* min length of data to allocate a cluster */ 285 u_long m_mlen; /* length of data in an mbuf */ 286 u_long m_mhlen; /* length of data in a header mbuf */ 287 288 /* Number of mbtypes (gives # elems in mbtypes[] array) */ 289 short m_numtypes; 290 291 /* XXX: Sendfile stats should eventually move to their own struct */ 292 u_long sf_iocnt; /* times sendfile had to do disk I/O */ 293 u_long sf_allocfail; /* times sfbuf allocation failed */ 294 u_long sf_allocwait; /* times sfbuf allocation had to wait */ 295 }; 296 297 /* 298 * Flags specifying how an allocation should be made. 299 * 300 * The flag to use is as follows: 301 * - M_DONTWAIT or M_NOWAIT from an interrupt handler to not block allocation. 302 * - M_WAIT or M_WAITOK from wherever it is safe to block. 303 * 304 * M_DONTWAIT/M_NOWAIT means that we will not block the thread explicitly and 305 * if we cannot allocate immediately we may return NULL, whereas 306 * M_WAIT/M_WAITOK means that if we cannot allocate resources we 307 * will block until they are available, and thus never return NULL. 308 * 309 * XXX Eventually just phase this out to use M_WAITOK/M_NOWAIT. 310 */ 311 #define MBTOM(how) (how) 312 #define M_DONTWAIT M_NOWAIT 313 #define M_TRYWAIT M_WAITOK 314 #define M_WAIT M_WAITOK 315 316 /* 317 * String names of mbuf-related UMA(9) and malloc(9) types. Exposed to 318 * !_KERNEL so that monitoring tools can look up the zones with 319 * libmemstat(3). 320 */ 321 #define MBUF_MEM_NAME "mbuf" 322 #define MBUF_CLUSTER_MEM_NAME "mbuf_cluster" 323 #define MBUF_PACKET_MEM_NAME "mbuf_packet" 324 #define MBUF_JUMBOP_MEM_NAME "mbuf_jumbo_page" 325 #define MBUF_JUMBO9_MEM_NAME "mbuf_jumbo_9k" 326 #define MBUF_JUMBO16_MEM_NAME "mbuf_jumbo_16k" 327 #define MBUF_TAG_MEM_NAME "mbuf_tag" 328 #define MBUF_EXTREFCNT_MEM_NAME "mbuf_ext_refcnt" 329 330 #ifdef _KERNEL 331 332 #ifdef WITNESS 333 #define MBUF_CHECKSLEEP(how) do { \ 334 if (how == M_WAITOK) \ 335 WITNESS_WARN(WARN_GIANTOK | WARN_SLEEPOK, NULL, \ 336 "Sleeping in \"%s\"", __func__); \ 337 } while (0) 338 #else 339 #define MBUF_CHECKSLEEP(how) 340 #endif 341 342 /* 343 * Network buffer allocation API 344 * 345 * The rest of it is defined in kern/kern_mbuf.c 346 */ 347 348 extern uma_zone_t zone_mbuf; 349 extern uma_zone_t zone_clust; 350 extern uma_zone_t zone_pack; 351 extern uma_zone_t zone_jumbop; 352 extern uma_zone_t zone_jumbo9; 353 extern uma_zone_t zone_jumbo16; 354 extern uma_zone_t zone_ext_refcnt; 355 356 static __inline struct mbuf *m_getcl(int how, short type, int flags); 357 static __inline struct mbuf *m_get(int how, short type); 358 static __inline struct mbuf *m_gethdr(int how, short type); 359 static __inline struct mbuf *m_getjcl(int how, short type, int flags, 360 int size); 361 static __inline struct mbuf *m_getclr(int how, short type); /* XXX */ 362 static __inline struct mbuf *m_free(struct mbuf *m); 363 static __inline void m_clget(struct mbuf *m, int how); 364 static __inline void *m_cljget(struct mbuf *m, int how, int size); 365 static __inline void m_chtype(struct mbuf *m, short new_type); 366 void mb_free_ext(struct mbuf *); 367 static __inline struct mbuf *m_last(struct mbuf *m); 368 369 static __inline int 370 m_gettype(int size) 371 { 372 int type; 373 374 switch (size) { 375 case MSIZE: 376 type = EXT_MBUF; 377 break; 378 case MCLBYTES: 379 type = EXT_CLUSTER; 380 break; 381 #if MJUMPAGESIZE != MCLBYTES 382 case MJUMPAGESIZE: 383 type = EXT_JUMBOP; 384 break; 385 #endif 386 case MJUM9BYTES: 387 type = EXT_JUMBO9; 388 break; 389 case MJUM16BYTES: 390 type = EXT_JUMBO16; 391 break; 392 default: 393 panic("%s: m_getjcl: invalid cluster size", __func__); 394 } 395 396 return (type); 397 } 398 399 static __inline uma_zone_t 400 m_getzone(int size) 401 { 402 uma_zone_t zone; 403 404 switch (size) { 405 case MSIZE: 406 zone = zone_mbuf; 407 break; 408 case MCLBYTES: 409 zone = zone_clust; 410 break; 411 #if MJUMPAGESIZE != MCLBYTES 412 case MJUMPAGESIZE: 413 zone = zone_jumbop; 414 break; 415 #endif 416 case MJUM9BYTES: 417 zone = zone_jumbo9; 418 break; 419 case MJUM16BYTES: 420 zone = zone_jumbo16; 421 break; 422 default: 423 panic("%s: m_getjcl: invalid cluster type", __func__); 424 } 425 426 return (zone); 427 } 428 429 static __inline struct mbuf * 430 m_get(int how, short type) 431 { 432 struct mb_args args; 433 434 args.flags = 0; 435 args.type = type; 436 return ((struct mbuf *)(uma_zalloc_arg(zone_mbuf, &args, how))); 437 } 438 439 /* 440 * XXX This should be deprecated, very little use. 441 */ 442 static __inline struct mbuf * 443 m_getclr(int how, short type) 444 { 445 struct mbuf *m; 446 struct mb_args args; 447 448 args.flags = 0; 449 args.type = type; 450 m = uma_zalloc_arg(zone_mbuf, &args, how); 451 if (m != NULL) 452 bzero(m->m_data, MLEN); 453 return (m); 454 } 455 456 static __inline struct mbuf * 457 m_gethdr(int how, short type) 458 { 459 struct mb_args args; 460 461 args.flags = M_PKTHDR; 462 args.type = type; 463 return ((struct mbuf *)(uma_zalloc_arg(zone_mbuf, &args, how))); 464 } 465 466 static __inline struct mbuf * 467 m_getcl(int how, short type, int flags) 468 { 469 struct mb_args args; 470 471 args.flags = flags; 472 args.type = type; 473 return ((struct mbuf *)(uma_zalloc_arg(zone_pack, &args, how))); 474 } 475 476 /* 477 * m_getjcl() returns an mbuf with a cluster of the specified size attached. 478 * For size it takes MCLBYTES, MJUMPAGESIZE, MJUM9BYTES, MJUM16BYTES. 479 * 480 * XXX: This is rather large, should be real function maybe. 481 */ 482 static __inline struct mbuf * 483 m_getjcl(int how, short type, int flags, int size) 484 { 485 struct mb_args args; 486 struct mbuf *m, *n; 487 uma_zone_t zone; 488 489 args.flags = flags; 490 args.type = type; 491 492 m = uma_zalloc_arg(zone_mbuf, &args, how); 493 if (m == NULL) 494 return (NULL); 495 496 zone = m_getzone(size); 497 n = uma_zalloc_arg(zone, m, how); 498 if (n == NULL) { 499 uma_zfree(zone_mbuf, m); 500 return (NULL); 501 } 502 return (m); 503 } 504 505 static __inline void 506 m_free_fast(struct mbuf *m) 507 { 508 #ifdef INVARIANTS 509 if (m->m_flags & M_PKTHDR) 510 KASSERT(SLIST_EMPTY(&m->m_pkthdr.tags), ("doing fast free of mbuf with tags")); 511 #endif 512 513 uma_zfree_arg(zone_mbuf, m, (void *)MB_NOTAGS); 514 } 515 516 static __inline struct mbuf * 517 m_free(struct mbuf *m) 518 { 519 struct mbuf *n = m->m_next; 520 521 if (m->m_flags & M_EXT) 522 mb_free_ext(m); 523 else if ((m->m_flags & M_NOFREE) == 0) 524 uma_zfree(zone_mbuf, m); 525 return (n); 526 } 527 528 static __inline void 529 m_clget(struct mbuf *m, int how) 530 { 531 532 if (m->m_flags & M_EXT) 533 printf("%s: %p mbuf already has cluster\n", __func__, m); 534 m->m_ext.ext_buf = (char *)NULL; 535 uma_zalloc_arg(zone_clust, m, how); 536 /* 537 * On a cluster allocation failure, drain the packet zone and retry, 538 * we might be able to loosen a few clusters up on the drain. 539 */ 540 if ((how & M_NOWAIT) && (m->m_ext.ext_buf == NULL)) { 541 zone_drain(zone_pack); 542 uma_zalloc_arg(zone_clust, m, how); 543 } 544 } 545 546 /* 547 * m_cljget() is different from m_clget() as it can allocate clusters without 548 * attaching them to an mbuf. In that case the return value is the pointer 549 * to the cluster of the requested size. If an mbuf was specified, it gets 550 * the cluster attached to it and the return value can be safely ignored. 551 * For size it takes MCLBYTES, MJUMPAGESIZE, MJUM9BYTES, MJUM16BYTES. 552 */ 553 static __inline void * 554 m_cljget(struct mbuf *m, int how, int size) 555 { 556 uma_zone_t zone; 557 558 if (m && m->m_flags & M_EXT) 559 printf("%s: %p mbuf already has cluster\n", __func__, m); 560 if (m != NULL) 561 m->m_ext.ext_buf = NULL; 562 563 zone = m_getzone(size); 564 return (uma_zalloc_arg(zone, m, how)); 565 } 566 567 static __inline void 568 m_cljset(struct mbuf *m, void *cl, int type) 569 { 570 uma_zone_t zone; 571 int size; 572 573 switch (type) { 574 case EXT_CLUSTER: 575 size = MCLBYTES; 576 zone = zone_clust; 577 break; 578 #if MJUMPAGESIZE != MCLBYTES 579 case EXT_JUMBOP: 580 size = MJUMPAGESIZE; 581 zone = zone_jumbop; 582 break; 583 #endif 584 case EXT_JUMBO9: 585 size = MJUM9BYTES; 586 zone = zone_jumbo9; 587 break; 588 case EXT_JUMBO16: 589 size = MJUM16BYTES; 590 zone = zone_jumbo16; 591 break; 592 default: 593 panic("unknown cluster type"); 594 break; 595 } 596 597 m->m_data = m->m_ext.ext_buf = cl; 598 m->m_ext.ext_free = m->m_ext.ext_arg1 = m->m_ext.ext_arg2 = NULL; 599 m->m_ext.ext_size = size; 600 m->m_ext.ext_type = type; 601 m->m_ext.ref_cnt = uma_find_refcnt(zone, cl); 602 m->m_flags |= M_EXT; 603 604 } 605 606 static __inline void 607 m_chtype(struct mbuf *m, short new_type) 608 { 609 610 m->m_type = new_type; 611 } 612 613 static __inline struct mbuf * 614 m_last(struct mbuf *m) 615 { 616 617 while (m->m_next) 618 m = m->m_next; 619 return (m); 620 } 621 622 /* 623 * mbuf, cluster, and external object allocation macros (for compatibility 624 * purposes). 625 */ 626 #define M_MOVE_PKTHDR(to, from) m_move_pkthdr((to), (from)) 627 #define MGET(m, how, type) ((m) = m_get((how), (type))) 628 #define MGETHDR(m, how, type) ((m) = m_gethdr((how), (type))) 629 #define MCLGET(m, how) m_clget((m), (how)) 630 #define MEXTADD(m, buf, size, free, arg1, arg2, flags, type) \ 631 m_extadd((m), (caddr_t)(buf), (size), (free),(arg1),(arg2),(flags), (type)) 632 #define m_getm(m, len, how, type) \ 633 m_getm2((m), (len), (how), (type), M_PKTHDR) 634 635 /* 636 * Evaluate TRUE if it's safe to write to the mbuf m's data region (this can 637 * be both the local data payload, or an external buffer area, depending on 638 * whether M_EXT is set). 639 */ 640 #define M_WRITABLE(m) (!((m)->m_flags & M_RDONLY) && \ 641 (!(((m)->m_flags & M_EXT)) || \ 642 (*((m)->m_ext.ref_cnt) == 1)) ) \ 643 644 /* Check if the supplied mbuf has a packet header, or else panic. */ 645 #define M_ASSERTPKTHDR(m) \ 646 KASSERT((m) != NULL && (m)->m_flags & M_PKTHDR, \ 647 ("%s: no mbuf packet header!", __func__)) 648 649 /* 650 * Ensure that the supplied mbuf is a valid, non-free mbuf. 651 * 652 * XXX: Broken at the moment. Need some UMA magic to make it work again. 653 */ 654 #define M_ASSERTVALID(m) \ 655 KASSERT((((struct mbuf *)m)->m_flags & 0) == 0, \ 656 ("%s: attempted use of a free mbuf!", __func__)) 657 658 /* 659 * Set the m_data pointer of a newly-allocated mbuf (m_get/MGET) to place an 660 * object of the specified size at the end of the mbuf, longword aligned. 661 */ 662 #define M_ALIGN(m, len) do { \ 663 KASSERT(!((m)->m_flags & (M_PKTHDR|M_EXT)), \ 664 ("%s: M_ALIGN not normal mbuf", __func__)); \ 665 KASSERT((m)->m_data == (m)->m_dat, \ 666 ("%s: M_ALIGN not a virgin mbuf", __func__)); \ 667 (m)->m_data += (MLEN - (len)) & ~(sizeof(long) - 1); \ 668 } while (0) 669 670 /* 671 * As above, for mbufs allocated with m_gethdr/MGETHDR or initialized by 672 * M_DUP/MOVE_PKTHDR. 673 */ 674 #define MH_ALIGN(m, len) do { \ 675 KASSERT((m)->m_flags & M_PKTHDR && !((m)->m_flags & M_EXT), \ 676 ("%s: MH_ALIGN not PKTHDR mbuf", __func__)); \ 677 KASSERT((m)->m_data == (m)->m_pktdat, \ 678 ("%s: MH_ALIGN not a virgin mbuf", __func__)); \ 679 (m)->m_data += (MHLEN - (len)) & ~(sizeof(long) - 1); \ 680 } while (0) 681 682 /* 683 * Compute the amount of space available before the current start of data in 684 * an mbuf. 685 * 686 * The M_WRITABLE() is a temporary, conservative safety measure: the burden 687 * of checking writability of the mbuf data area rests solely with the caller. 688 */ 689 #define M_LEADINGSPACE(m) \ 690 ((m)->m_flags & M_EXT ? \ 691 (M_WRITABLE(m) ? (m)->m_data - (m)->m_ext.ext_buf : 0): \ 692 (m)->m_flags & M_PKTHDR ? (m)->m_data - (m)->m_pktdat : \ 693 (m)->m_data - (m)->m_dat) 694 695 /* 696 * Compute the amount of space available after the end of data in an mbuf. 697 * 698 * The M_WRITABLE() is a temporary, conservative safety measure: the burden 699 * of checking writability of the mbuf data area rests solely with the caller. 700 */ 701 #define M_TRAILINGSPACE(m) \ 702 ((m)->m_flags & M_EXT ? \ 703 (M_WRITABLE(m) ? (m)->m_ext.ext_buf + (m)->m_ext.ext_size \ 704 - ((m)->m_data + (m)->m_len) : 0) : \ 705 &(m)->m_dat[MLEN] - ((m)->m_data + (m)->m_len)) 706 707 /* 708 * Arrange to prepend space of size plen to mbuf m. If a new mbuf must be 709 * allocated, how specifies whether to wait. If the allocation fails, the 710 * original mbuf chain is freed and m is set to NULL. 711 */ 712 #define M_PREPEND(m, plen, how) do { \ 713 struct mbuf **_mmp = &(m); \ 714 struct mbuf *_mm = *_mmp; \ 715 int _mplen = (plen); \ 716 int __mhow = (how); \ 717 \ 718 MBUF_CHECKSLEEP(how); \ 719 if (M_LEADINGSPACE(_mm) >= _mplen) { \ 720 _mm->m_data -= _mplen; \ 721 _mm->m_len += _mplen; \ 722 } else \ 723 _mm = m_prepend(_mm, _mplen, __mhow); \ 724 if (_mm != NULL && _mm->m_flags & M_PKTHDR) \ 725 _mm->m_pkthdr.len += _mplen; \ 726 *_mmp = _mm; \ 727 } while (0) 728 729 /* 730 * Change mbuf to new type. This is a relatively expensive operation and 731 * should be avoided. 732 */ 733 #define MCHTYPE(m, t) m_chtype((m), (t)) 734 735 /* Length to m_copy to copy all. */ 736 #define M_COPYALL 1000000000 737 738 /* Compatibility with 4.3. */ 739 #define m_copy(m, o, l) m_copym((m), (o), (l), M_DONTWAIT) 740 741 extern int max_datalen; /* MHLEN - max_hdr */ 742 extern int max_hdr; /* Largest link + protocol header */ 743 extern int max_linkhdr; /* Largest link-level header */ 744 extern int max_protohdr; /* Largest protocol header */ 745 extern struct mbstat mbstat; /* General mbuf stats/infos */ 746 extern int nmbclusters; /* Maximum number of clusters */ 747 748 struct uio; 749 750 void m_adj(struct mbuf *, int); 751 void m_align(struct mbuf *, int); 752 int m_apply(struct mbuf *, int, int, 753 int (*)(void *, void *, u_int), void *); 754 int m_append(struct mbuf *, int, c_caddr_t); 755 void m_cat(struct mbuf *, struct mbuf *); 756 void m_extadd(struct mbuf *, caddr_t, u_int, 757 void (*)(void *, void *), void *, void *, int, int); 758 struct mbuf *m_collapse(struct mbuf *, int, int); 759 void m_copyback(struct mbuf *, int, int, c_caddr_t); 760 void m_copydata(const struct mbuf *, int, int, caddr_t); 761 struct mbuf *m_copym(struct mbuf *, int, int, int); 762 struct mbuf *m_copymdata(struct mbuf *, struct mbuf *, 763 int, int, int, int); 764 struct mbuf *m_copypacket(struct mbuf *, int); 765 void m_copy_pkthdr(struct mbuf *, struct mbuf *); 766 struct mbuf *m_copyup(struct mbuf *n, int len, int dstoff); 767 struct mbuf *m_defrag(struct mbuf *, int); 768 void m_demote(struct mbuf *, int); 769 struct mbuf *m_devget(char *, int, int, struct ifnet *, 770 void (*)(char *, caddr_t, u_int)); 771 struct mbuf *m_dup(struct mbuf *, int); 772 int m_dup_pkthdr(struct mbuf *, struct mbuf *, int); 773 u_int m_fixhdr(struct mbuf *); 774 struct mbuf *m_fragment(struct mbuf *, int, int); 775 void m_freem(struct mbuf *); 776 struct mbuf *m_getm2(struct mbuf *, int, int, short, int); 777 struct mbuf *m_getptr(struct mbuf *, int, int *); 778 u_int m_length(struct mbuf *, struct mbuf **); 779 void m_move_pkthdr(struct mbuf *, struct mbuf *); 780 struct mbuf *m_prepend(struct mbuf *, int, int); 781 void m_print(const struct mbuf *, int); 782 struct mbuf *m_pulldown(struct mbuf *, int, int, int *); 783 struct mbuf *m_pullup(struct mbuf *, int); 784 int m_sanity(struct mbuf *, int); 785 struct mbuf *m_split(struct mbuf *, int, int); 786 struct mbuf *m_uiotombuf(struct uio *, int, int, int, int); 787 struct mbuf *m_unshare(struct mbuf *, int how); 788 789 /*- 790 * Network packets may have annotations attached by affixing a list of 791 * "packet tags" to the pkthdr structure. Packet tags are dynamically 792 * allocated semi-opaque data structures that have a fixed header 793 * (struct m_tag) that specifies the size of the memory block and a 794 * <cookie,type> pair that identifies it. The cookie is a 32-bit unique 795 * unsigned value used to identify a module or ABI. By convention this value 796 * is chosen as the date+time that the module is created, expressed as the 797 * number of seconds since the epoch (e.g., using date -u +'%s'). The type 798 * value is an ABI/module-specific value that identifies a particular 799 * annotation and is private to the module. For compatibility with systems 800 * like OpenBSD that define packet tags w/o an ABI/module cookie, the value 801 * PACKET_ABI_COMPAT is used to implement m_tag_get and m_tag_find 802 * compatibility shim functions and several tag types are defined below. 803 * Users that do not require compatibility should use a private cookie value 804 * so that packet tag-related definitions can be maintained privately. 805 * 806 * Note that the packet tag returned by m_tag_alloc has the default memory 807 * alignment implemented by malloc. To reference private data one can use a 808 * construct like: 809 * 810 * struct m_tag *mtag = m_tag_alloc(...); 811 * struct foo *p = (struct foo *)(mtag+1); 812 * 813 * if the alignment of struct m_tag is sufficient for referencing members of 814 * struct foo. Otherwise it is necessary to embed struct m_tag within the 815 * private data structure to insure proper alignment; e.g., 816 * 817 * struct foo { 818 * struct m_tag tag; 819 * ... 820 * }; 821 * struct foo *p = (struct foo *) m_tag_alloc(...); 822 * struct m_tag *mtag = &p->tag; 823 */ 824 825 /* 826 * Persistent tags stay with an mbuf until the mbuf is reclaimed. Otherwise 827 * tags are expected to ``vanish'' when they pass through a network 828 * interface. For most interfaces this happens normally as the tags are 829 * reclaimed when the mbuf is free'd. However in some special cases 830 * reclaiming must be done manually. An example is packets that pass through 831 * the loopback interface. Also, one must be careful to do this when 832 * ``turning around'' packets (e.g., icmp_reflect). 833 * 834 * To mark a tag persistent bit-or this flag in when defining the tag id. 835 * The tag will then be treated as described above. 836 */ 837 #define MTAG_PERSISTENT 0x800 838 839 #define PACKET_TAG_NONE 0 /* Nadda */ 840 841 /* Packet tags for use with PACKET_ABI_COMPAT. */ 842 #define PACKET_TAG_IPSEC_IN_DONE 1 /* IPsec applied, in */ 843 #define PACKET_TAG_IPSEC_OUT_DONE 2 /* IPsec applied, out */ 844 #define PACKET_TAG_IPSEC_IN_CRYPTO_DONE 3 /* NIC IPsec crypto done */ 845 #define PACKET_TAG_IPSEC_OUT_CRYPTO_NEEDED 4 /* NIC IPsec crypto req'ed */ 846 #define PACKET_TAG_IPSEC_IN_COULD_DO_CRYPTO 5 /* NIC notifies IPsec */ 847 #define PACKET_TAG_IPSEC_PENDING_TDB 6 /* Reminder to do IPsec */ 848 #define PACKET_TAG_BRIDGE 7 /* Bridge processing done */ 849 #define PACKET_TAG_GIF 8 /* GIF processing done */ 850 #define PACKET_TAG_GRE 9 /* GRE processing done */ 851 #define PACKET_TAG_IN_PACKET_CHECKSUM 10 /* NIC checksumming done */ 852 #define PACKET_TAG_ENCAP 11 /* Encap. processing */ 853 #define PACKET_TAG_IPSEC_SOCKET 12 /* IPSEC socket ref */ 854 #define PACKET_TAG_IPSEC_HISTORY 13 /* IPSEC history */ 855 #define PACKET_TAG_IPV6_INPUT 14 /* IPV6 input processing */ 856 #define PACKET_TAG_DUMMYNET 15 /* dummynet info */ 857 #define PACKET_TAG_DIVERT 17 /* divert info */ 858 #define PACKET_TAG_IPFORWARD 18 /* ipforward info */ 859 #define PACKET_TAG_MACLABEL (19 | MTAG_PERSISTENT) /* MAC label */ 860 #define PACKET_TAG_PF 21 /* PF + ALTQ information */ 861 #define PACKET_TAG_RTSOCKFAM 25 /* rtsock sa family */ 862 #define PACKET_TAG_IPOPTIONS 27 /* Saved IP options */ 863 #define PACKET_TAG_CARP 28 /* CARP info */ 864 865 /* Specific cookies and tags. */ 866 867 /* Packet tag routines. */ 868 struct m_tag *m_tag_alloc(u_int32_t, int, int, int); 869 void m_tag_delete(struct mbuf *, struct m_tag *); 870 void m_tag_delete_chain(struct mbuf *, struct m_tag *); 871 void m_tag_free_default(struct m_tag *); 872 struct m_tag *m_tag_locate(struct mbuf *, u_int32_t, int, struct m_tag *); 873 struct m_tag *m_tag_copy(struct m_tag *, int); 874 int m_tag_copy_chain(struct mbuf *, struct mbuf *, int); 875 void m_tag_delete_nonpersistent(struct mbuf *); 876 877 /* 878 * Initialize the list of tags associated with an mbuf. 879 */ 880 static __inline void 881 m_tag_init(struct mbuf *m) 882 { 883 884 SLIST_INIT(&m->m_pkthdr.tags); 885 } 886 887 /* 888 * Set up the contents of a tag. Note that this does not fill in the free 889 * method; the caller is expected to do that. 890 * 891 * XXX probably should be called m_tag_init, but that was already taken. 892 */ 893 static __inline void 894 m_tag_setup(struct m_tag *t, u_int32_t cookie, int type, int len) 895 { 896 897 t->m_tag_id = type; 898 t->m_tag_len = len; 899 t->m_tag_cookie = cookie; 900 } 901 902 /* 903 * Reclaim resources associated with a tag. 904 */ 905 static __inline void 906 m_tag_free(struct m_tag *t) 907 { 908 909 (*t->m_tag_free)(t); 910 } 911 912 /* 913 * Return the first tag associated with an mbuf. 914 */ 915 static __inline struct m_tag * 916 m_tag_first(struct mbuf *m) 917 { 918 919 return (SLIST_FIRST(&m->m_pkthdr.tags)); 920 } 921 922 /* 923 * Return the next tag in the list of tags associated with an mbuf. 924 */ 925 static __inline struct m_tag * 926 m_tag_next(struct mbuf *m, struct m_tag *t) 927 { 928 929 return (SLIST_NEXT(t, m_tag_link)); 930 } 931 932 /* 933 * Prepend a tag to the list of tags associated with an mbuf. 934 */ 935 static __inline void 936 m_tag_prepend(struct mbuf *m, struct m_tag *t) 937 { 938 939 SLIST_INSERT_HEAD(&m->m_pkthdr.tags, t, m_tag_link); 940 } 941 942 /* 943 * Unlink a tag from the list of tags associated with an mbuf. 944 */ 945 static __inline void 946 m_tag_unlink(struct mbuf *m, struct m_tag *t) 947 { 948 949 SLIST_REMOVE(&m->m_pkthdr.tags, t, m_tag, m_tag_link); 950 } 951 952 /* These are for OpenBSD compatibility. */ 953 #define MTAG_ABI_COMPAT 0 /* compatibility ABI */ 954 955 static __inline struct m_tag * 956 m_tag_get(int type, int length, int wait) 957 { 958 return (m_tag_alloc(MTAG_ABI_COMPAT, type, length, wait)); 959 } 960 961 static __inline struct m_tag * 962 m_tag_find(struct mbuf *m, int type, struct m_tag *start) 963 { 964 return (SLIST_EMPTY(&m->m_pkthdr.tags) ? (struct m_tag *)NULL : 965 m_tag_locate(m, MTAG_ABI_COMPAT, type, start)); 966 } 967 968 /* XXX temporary FIB methods probably eventually use tags.*/ 969 #define M_FIBSHIFT 28 970 #define M_FIBMASK 0x0F 971 972 /* get the fib from an mbuf and if it is not set, return the default */ 973 #define M_GETFIB(_m) \ 974 ((((_m)->m_flags & M_FIB) >> M_FIBSHIFT) & M_FIBMASK) 975 976 #define M_SETFIB(_m, _fib) do { \ 977 _m->m_flags &= ~M_FIB; \ 978 _m->m_flags |= (((_fib) << M_FIBSHIFT) & M_FIB); \ 979 } while (0) 980 981 #endif /* _KERNEL */ 982 983 #ifdef MBUF_PROFILING 984 void m_profile(struct mbuf *m); 985 #define M_PROFILE(m) m_profile(m) 986 #else 987 #define M_PROFILE(m) 988 #endif 989 990 991 #endif /* !_SYS_MBUF_H_ */ 992