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 #define CSUM_SCTP 0x0040 /* will csum SCTP */ 245 246 #define CSUM_IP_CHECKED 0x0100 /* did csum IP */ 247 #define CSUM_IP_VALID 0x0200 /* ... the csum is valid */ 248 #define CSUM_DATA_VALID 0x0400 /* csum_data field is valid */ 249 #define CSUM_PSEUDO_HDR 0x0800 /* csum_data has pseudo hdr */ 250 #define CSUM_SCTP_VALID 0x1000 /* SCTP checksum is valid */ 251 252 #define CSUM_DELAY_DATA (CSUM_TCP | CSUM_UDP) 253 #define CSUM_DELAY_IP (CSUM_IP) /* XXX add ipv6 here too? */ 254 255 /* 256 * mbuf types. 257 */ 258 #define MT_NOTMBUF 0 /* USED INTERNALLY ONLY! Object is not mbuf */ 259 #define MT_DATA 1 /* dynamic (data) allocation */ 260 #define MT_HEADER MT_DATA /* packet header, use M_PKTHDR instead */ 261 #define MT_SONAME 8 /* socket name */ 262 #define MT_CONTROL 14 /* extra-data protocol message */ 263 #define MT_OOBDATA 15 /* expedited data */ 264 #define MT_NTYPES 16 /* number of mbuf types for mbtypes[] */ 265 266 #define MT_NOINIT 255 /* Not a type but a flag to allocate 267 a non-initialized mbuf */ 268 269 #define MB_NOTAGS 0x1UL /* no tags attached to mbuf */ 270 271 /* 272 * General mbuf allocator statistics structure. 273 * 274 * Many of these statistics are no longer used; we instead track many 275 * allocator statistics through UMA's built in statistics mechanism. 276 */ 277 struct mbstat { 278 u_long m_mbufs; /* XXX */ 279 u_long m_mclusts; /* XXX */ 280 281 u_long m_drain; /* times drained protocols for space */ 282 u_long m_mcfail; /* XXX: times m_copym failed */ 283 u_long m_mpfail; /* XXX: times m_pullup failed */ 284 u_long m_msize; /* length of an mbuf */ 285 u_long m_mclbytes; /* length of an mbuf cluster */ 286 u_long m_minclsize; /* min length of data to allocate a cluster */ 287 u_long m_mlen; /* length of data in an mbuf */ 288 u_long m_mhlen; /* length of data in a header mbuf */ 289 290 /* Number of mbtypes (gives # elems in mbtypes[] array) */ 291 short m_numtypes; 292 293 /* XXX: Sendfile stats should eventually move to their own struct */ 294 u_long sf_iocnt; /* times sendfile had to do disk I/O */ 295 u_long sf_allocfail; /* times sfbuf allocation failed */ 296 u_long sf_allocwait; /* times sfbuf allocation had to wait */ 297 }; 298 299 /* 300 * Flags specifying how an allocation should be made. 301 * 302 * The flag to use is as follows: 303 * - M_DONTWAIT or M_NOWAIT from an interrupt handler to not block allocation. 304 * - M_WAIT or M_WAITOK from wherever it is safe to block. 305 * 306 * M_DONTWAIT/M_NOWAIT means that we will not block the thread explicitly and 307 * if we cannot allocate immediately we may return NULL, whereas 308 * M_WAIT/M_WAITOK means that if we cannot allocate resources we 309 * will block until they are available, and thus never return NULL. 310 * 311 * XXX Eventually just phase this out to use M_WAITOK/M_NOWAIT. 312 */ 313 #define MBTOM(how) (how) 314 #define M_DONTWAIT M_NOWAIT 315 #define M_TRYWAIT M_WAITOK 316 #define M_WAIT M_WAITOK 317 318 /* 319 * String names of mbuf-related UMA(9) and malloc(9) types. Exposed to 320 * !_KERNEL so that monitoring tools can look up the zones with 321 * libmemstat(3). 322 */ 323 #define MBUF_MEM_NAME "mbuf" 324 #define MBUF_CLUSTER_MEM_NAME "mbuf_cluster" 325 #define MBUF_PACKET_MEM_NAME "mbuf_packet" 326 #define MBUF_JUMBOP_MEM_NAME "mbuf_jumbo_page" 327 #define MBUF_JUMBO9_MEM_NAME "mbuf_jumbo_9k" 328 #define MBUF_JUMBO16_MEM_NAME "mbuf_jumbo_16k" 329 #define MBUF_TAG_MEM_NAME "mbuf_tag" 330 #define MBUF_EXTREFCNT_MEM_NAME "mbuf_ext_refcnt" 331 332 #ifdef _KERNEL 333 334 #ifdef WITNESS 335 #define MBUF_CHECKSLEEP(how) do { \ 336 if (how == M_WAITOK) \ 337 WITNESS_WARN(WARN_GIANTOK | WARN_SLEEPOK, NULL, \ 338 "Sleeping in \"%s\"", __func__); \ 339 } while (0) 340 #else 341 #define MBUF_CHECKSLEEP(how) 342 #endif 343 344 /* 345 * Network buffer allocation API 346 * 347 * The rest of it is defined in kern/kern_mbuf.c 348 */ 349 350 extern uma_zone_t zone_mbuf; 351 extern uma_zone_t zone_clust; 352 extern uma_zone_t zone_pack; 353 extern uma_zone_t zone_jumbop; 354 extern uma_zone_t zone_jumbo9; 355 extern uma_zone_t zone_jumbo16; 356 extern uma_zone_t zone_ext_refcnt; 357 358 static __inline struct mbuf *m_getcl(int how, short type, int flags); 359 static __inline struct mbuf *m_get(int how, short type); 360 static __inline struct mbuf *m_gethdr(int how, short type); 361 static __inline struct mbuf *m_getjcl(int how, short type, int flags, 362 int size); 363 static __inline struct mbuf *m_getclr(int how, short type); /* XXX */ 364 static __inline struct mbuf *m_free(struct mbuf *m); 365 static __inline void m_clget(struct mbuf *m, int how); 366 static __inline void *m_cljget(struct mbuf *m, int how, int size); 367 static __inline void m_chtype(struct mbuf *m, short new_type); 368 void mb_free_ext(struct mbuf *); 369 static __inline struct mbuf *m_last(struct mbuf *m); 370 371 static __inline int 372 m_gettype(int size) 373 { 374 int type; 375 376 switch (size) { 377 case MSIZE: 378 type = EXT_MBUF; 379 break; 380 case MCLBYTES: 381 type = EXT_CLUSTER; 382 break; 383 #if MJUMPAGESIZE != MCLBYTES 384 case MJUMPAGESIZE: 385 type = EXT_JUMBOP; 386 break; 387 #endif 388 case MJUM9BYTES: 389 type = EXT_JUMBO9; 390 break; 391 case MJUM16BYTES: 392 type = EXT_JUMBO16; 393 break; 394 default: 395 panic("%s: m_getjcl: invalid cluster size", __func__); 396 } 397 398 return (type); 399 } 400 401 static __inline uma_zone_t 402 m_getzone(int size) 403 { 404 uma_zone_t zone; 405 406 switch (size) { 407 case MSIZE: 408 zone = zone_mbuf; 409 break; 410 case MCLBYTES: 411 zone = zone_clust; 412 break; 413 #if MJUMPAGESIZE != MCLBYTES 414 case MJUMPAGESIZE: 415 zone = zone_jumbop; 416 break; 417 #endif 418 case MJUM9BYTES: 419 zone = zone_jumbo9; 420 break; 421 case MJUM16BYTES: 422 zone = zone_jumbo16; 423 break; 424 default: 425 panic("%s: m_getjcl: invalid cluster type", __func__); 426 } 427 428 return (zone); 429 } 430 431 static __inline struct mbuf * 432 m_get(int how, short type) 433 { 434 struct mb_args args; 435 436 args.flags = 0; 437 args.type = type; 438 return ((struct mbuf *)(uma_zalloc_arg(zone_mbuf, &args, how))); 439 } 440 441 /* 442 * XXX This should be deprecated, very little use. 443 */ 444 static __inline struct mbuf * 445 m_getclr(int how, short type) 446 { 447 struct mbuf *m; 448 struct mb_args args; 449 450 args.flags = 0; 451 args.type = type; 452 m = uma_zalloc_arg(zone_mbuf, &args, how); 453 if (m != NULL) 454 bzero(m->m_data, MLEN); 455 return (m); 456 } 457 458 static __inline struct mbuf * 459 m_gethdr(int how, short type) 460 { 461 struct mb_args args; 462 463 args.flags = M_PKTHDR; 464 args.type = type; 465 return ((struct mbuf *)(uma_zalloc_arg(zone_mbuf, &args, how))); 466 } 467 468 static __inline struct mbuf * 469 m_getcl(int how, short type, int flags) 470 { 471 struct mb_args args; 472 473 args.flags = flags; 474 args.type = type; 475 return ((struct mbuf *)(uma_zalloc_arg(zone_pack, &args, how))); 476 } 477 478 /* 479 * m_getjcl() returns an mbuf with a cluster of the specified size attached. 480 * For size it takes MCLBYTES, MJUMPAGESIZE, MJUM9BYTES, MJUM16BYTES. 481 * 482 * XXX: This is rather large, should be real function maybe. 483 */ 484 static __inline struct mbuf * 485 m_getjcl(int how, short type, int flags, int size) 486 { 487 struct mb_args args; 488 struct mbuf *m, *n; 489 uma_zone_t zone; 490 491 args.flags = flags; 492 args.type = type; 493 494 m = uma_zalloc_arg(zone_mbuf, &args, how); 495 if (m == NULL) 496 return (NULL); 497 498 zone = m_getzone(size); 499 n = uma_zalloc_arg(zone, m, how); 500 if (n == NULL) { 501 uma_zfree(zone_mbuf, m); 502 return (NULL); 503 } 504 return (m); 505 } 506 507 static __inline void 508 m_free_fast(struct mbuf *m) 509 { 510 #ifdef INVARIANTS 511 if (m->m_flags & M_PKTHDR) 512 KASSERT(SLIST_EMPTY(&m->m_pkthdr.tags), ("doing fast free of mbuf with tags")); 513 #endif 514 515 uma_zfree_arg(zone_mbuf, m, (void *)MB_NOTAGS); 516 } 517 518 static __inline struct mbuf * 519 m_free(struct mbuf *m) 520 { 521 struct mbuf *n = m->m_next; 522 523 if (m->m_flags & M_EXT) 524 mb_free_ext(m); 525 else if ((m->m_flags & M_NOFREE) == 0) 526 uma_zfree(zone_mbuf, m); 527 return (n); 528 } 529 530 static __inline void 531 m_clget(struct mbuf *m, int how) 532 { 533 534 if (m->m_flags & M_EXT) 535 printf("%s: %p mbuf already has cluster\n", __func__, m); 536 m->m_ext.ext_buf = (char *)NULL; 537 uma_zalloc_arg(zone_clust, m, how); 538 /* 539 * On a cluster allocation failure, drain the packet zone and retry, 540 * we might be able to loosen a few clusters up on the drain. 541 */ 542 if ((how & M_NOWAIT) && (m->m_ext.ext_buf == NULL)) { 543 zone_drain(zone_pack); 544 uma_zalloc_arg(zone_clust, m, how); 545 } 546 } 547 548 /* 549 * m_cljget() is different from m_clget() as it can allocate clusters without 550 * attaching them to an mbuf. In that case the return value is the pointer 551 * to the cluster of the requested size. If an mbuf was specified, it gets 552 * the cluster attached to it and the return value can be safely ignored. 553 * For size it takes MCLBYTES, MJUMPAGESIZE, MJUM9BYTES, MJUM16BYTES. 554 */ 555 static __inline void * 556 m_cljget(struct mbuf *m, int how, int size) 557 { 558 uma_zone_t zone; 559 560 if (m && m->m_flags & M_EXT) 561 printf("%s: %p mbuf already has cluster\n", __func__, m); 562 if (m != NULL) 563 m->m_ext.ext_buf = NULL; 564 565 zone = m_getzone(size); 566 return (uma_zalloc_arg(zone, m, how)); 567 } 568 569 static __inline void 570 m_cljset(struct mbuf *m, void *cl, int type) 571 { 572 uma_zone_t zone; 573 int size; 574 575 switch (type) { 576 case EXT_CLUSTER: 577 size = MCLBYTES; 578 zone = zone_clust; 579 break; 580 #if MJUMPAGESIZE != MCLBYTES 581 case EXT_JUMBOP: 582 size = MJUMPAGESIZE; 583 zone = zone_jumbop; 584 break; 585 #endif 586 case EXT_JUMBO9: 587 size = MJUM9BYTES; 588 zone = zone_jumbo9; 589 break; 590 case EXT_JUMBO16: 591 size = MJUM16BYTES; 592 zone = zone_jumbo16; 593 break; 594 default: 595 panic("unknown cluster type"); 596 break; 597 } 598 599 m->m_data = m->m_ext.ext_buf = cl; 600 m->m_ext.ext_free = m->m_ext.ext_arg1 = m->m_ext.ext_arg2 = NULL; 601 m->m_ext.ext_size = size; 602 m->m_ext.ext_type = type; 603 m->m_ext.ref_cnt = uma_find_refcnt(zone, cl); 604 m->m_flags |= M_EXT; 605 606 } 607 608 static __inline void 609 m_chtype(struct mbuf *m, short new_type) 610 { 611 612 m->m_type = new_type; 613 } 614 615 static __inline struct mbuf * 616 m_last(struct mbuf *m) 617 { 618 619 while (m->m_next) 620 m = m->m_next; 621 return (m); 622 } 623 624 /* 625 * mbuf, cluster, and external object allocation macros (for compatibility 626 * purposes). 627 */ 628 #define M_MOVE_PKTHDR(to, from) m_move_pkthdr((to), (from)) 629 #define MGET(m, how, type) ((m) = m_get((how), (type))) 630 #define MGETHDR(m, how, type) ((m) = m_gethdr((how), (type))) 631 #define MCLGET(m, how) m_clget((m), (how)) 632 #define MEXTADD(m, buf, size, free, arg1, arg2, flags, type) \ 633 m_extadd((m), (caddr_t)(buf), (size), (free),(arg1),(arg2),(flags), (type)) 634 #define m_getm(m, len, how, type) \ 635 m_getm2((m), (len), (how), (type), M_PKTHDR) 636 637 /* 638 * Evaluate TRUE if it's safe to write to the mbuf m's data region (this can 639 * be both the local data payload, or an external buffer area, depending on 640 * whether M_EXT is set). 641 */ 642 #define M_WRITABLE(m) (!((m)->m_flags & M_RDONLY) && \ 643 (!(((m)->m_flags & M_EXT)) || \ 644 (*((m)->m_ext.ref_cnt) == 1)) ) \ 645 646 /* Check if the supplied mbuf has a packet header, or else panic. */ 647 #define M_ASSERTPKTHDR(m) \ 648 KASSERT((m) != NULL && (m)->m_flags & M_PKTHDR, \ 649 ("%s: no mbuf packet header!", __func__)) 650 651 /* 652 * Ensure that the supplied mbuf is a valid, non-free mbuf. 653 * 654 * XXX: Broken at the moment. Need some UMA magic to make it work again. 655 */ 656 #define M_ASSERTVALID(m) \ 657 KASSERT((((struct mbuf *)m)->m_flags & 0) == 0, \ 658 ("%s: attempted use of a free mbuf!", __func__)) 659 660 /* 661 * Set the m_data pointer of a newly-allocated mbuf (m_get/MGET) to place an 662 * object of the specified size at the end of the mbuf, longword aligned. 663 */ 664 #define M_ALIGN(m, len) do { \ 665 KASSERT(!((m)->m_flags & (M_PKTHDR|M_EXT)), \ 666 ("%s: M_ALIGN not normal mbuf", __func__)); \ 667 KASSERT((m)->m_data == (m)->m_dat, \ 668 ("%s: M_ALIGN not a virgin mbuf", __func__)); \ 669 (m)->m_data += (MLEN - (len)) & ~(sizeof(long) - 1); \ 670 } while (0) 671 672 /* 673 * As above, for mbufs allocated with m_gethdr/MGETHDR or initialized by 674 * M_DUP/MOVE_PKTHDR. 675 */ 676 #define MH_ALIGN(m, len) do { \ 677 KASSERT((m)->m_flags & M_PKTHDR && !((m)->m_flags & M_EXT), \ 678 ("%s: MH_ALIGN not PKTHDR mbuf", __func__)); \ 679 KASSERT((m)->m_data == (m)->m_pktdat, \ 680 ("%s: MH_ALIGN not a virgin mbuf", __func__)); \ 681 (m)->m_data += (MHLEN - (len)) & ~(sizeof(long) - 1); \ 682 } while (0) 683 684 /* 685 * Compute the amount of space available before the current start of data in 686 * an mbuf. 687 * 688 * The M_WRITABLE() is a temporary, conservative safety measure: the burden 689 * of checking writability of the mbuf data area rests solely with the caller. 690 */ 691 #define M_LEADINGSPACE(m) \ 692 ((m)->m_flags & M_EXT ? \ 693 (M_WRITABLE(m) ? (m)->m_data - (m)->m_ext.ext_buf : 0): \ 694 (m)->m_flags & M_PKTHDR ? (m)->m_data - (m)->m_pktdat : \ 695 (m)->m_data - (m)->m_dat) 696 697 /* 698 * Compute the amount of space available after the end of data in an mbuf. 699 * 700 * The M_WRITABLE() is a temporary, conservative safety measure: the burden 701 * of checking writability of the mbuf data area rests solely with the caller. 702 */ 703 #define M_TRAILINGSPACE(m) \ 704 ((m)->m_flags & M_EXT ? \ 705 (M_WRITABLE(m) ? (m)->m_ext.ext_buf + (m)->m_ext.ext_size \ 706 - ((m)->m_data + (m)->m_len) : 0) : \ 707 &(m)->m_dat[MLEN] - ((m)->m_data + (m)->m_len)) 708 709 /* 710 * Arrange to prepend space of size plen to mbuf m. If a new mbuf must be 711 * allocated, how specifies whether to wait. If the allocation fails, the 712 * original mbuf chain is freed and m is set to NULL. 713 */ 714 #define M_PREPEND(m, plen, how) do { \ 715 struct mbuf **_mmp = &(m); \ 716 struct mbuf *_mm = *_mmp; \ 717 int _mplen = (plen); \ 718 int __mhow = (how); \ 719 \ 720 MBUF_CHECKSLEEP(how); \ 721 if (M_LEADINGSPACE(_mm) >= _mplen) { \ 722 _mm->m_data -= _mplen; \ 723 _mm->m_len += _mplen; \ 724 } else \ 725 _mm = m_prepend(_mm, _mplen, __mhow); \ 726 if (_mm != NULL && _mm->m_flags & M_PKTHDR) \ 727 _mm->m_pkthdr.len += _mplen; \ 728 *_mmp = _mm; \ 729 } while (0) 730 731 /* 732 * Change mbuf to new type. This is a relatively expensive operation and 733 * should be avoided. 734 */ 735 #define MCHTYPE(m, t) m_chtype((m), (t)) 736 737 /* Length to m_copy to copy all. */ 738 #define M_COPYALL 1000000000 739 740 /* Compatibility with 4.3. */ 741 #define m_copy(m, o, l) m_copym((m), (o), (l), M_DONTWAIT) 742 743 extern int max_datalen; /* MHLEN - max_hdr */ 744 extern int max_hdr; /* Largest link + protocol header */ 745 extern int max_linkhdr; /* Largest link-level header */ 746 extern int max_protohdr; /* Largest protocol header */ 747 extern struct mbstat mbstat; /* General mbuf stats/infos */ 748 extern int nmbclusters; /* Maximum number of clusters */ 749 750 struct uio; 751 752 void m_adj(struct mbuf *, int); 753 void m_align(struct mbuf *, int); 754 int m_apply(struct mbuf *, int, int, 755 int (*)(void *, void *, u_int), void *); 756 int m_append(struct mbuf *, int, c_caddr_t); 757 void m_cat(struct mbuf *, struct mbuf *); 758 void m_extadd(struct mbuf *, caddr_t, u_int, 759 void (*)(void *, void *), void *, void *, int, int); 760 struct mbuf *m_collapse(struct mbuf *, int, int); 761 void m_copyback(struct mbuf *, int, int, c_caddr_t); 762 void m_copydata(const struct mbuf *, int, int, caddr_t); 763 struct mbuf *m_copym(struct mbuf *, int, int, int); 764 struct mbuf *m_copymdata(struct mbuf *, struct mbuf *, 765 int, int, int, int); 766 struct mbuf *m_copypacket(struct mbuf *, int); 767 void m_copy_pkthdr(struct mbuf *, struct mbuf *); 768 struct mbuf *m_copyup(struct mbuf *n, int len, int dstoff); 769 struct mbuf *m_defrag(struct mbuf *, int); 770 void m_demote(struct mbuf *, int); 771 struct mbuf *m_devget(char *, int, int, struct ifnet *, 772 void (*)(char *, caddr_t, u_int)); 773 struct mbuf *m_dup(struct mbuf *, int); 774 int m_dup_pkthdr(struct mbuf *, struct mbuf *, int); 775 u_int m_fixhdr(struct mbuf *); 776 struct mbuf *m_fragment(struct mbuf *, int, int); 777 void m_freem(struct mbuf *); 778 struct mbuf *m_getm2(struct mbuf *, int, int, short, int); 779 struct mbuf *m_getptr(struct mbuf *, int, int *); 780 u_int m_length(struct mbuf *, struct mbuf **); 781 void m_move_pkthdr(struct mbuf *, struct mbuf *); 782 struct mbuf *m_prepend(struct mbuf *, int, int); 783 void m_print(const struct mbuf *, int); 784 struct mbuf *m_pulldown(struct mbuf *, int, int, int *); 785 struct mbuf *m_pullup(struct mbuf *, int); 786 int m_sanity(struct mbuf *, int); 787 struct mbuf *m_split(struct mbuf *, int, int); 788 struct mbuf *m_uiotombuf(struct uio *, int, int, int, int); 789 struct mbuf *m_unshare(struct mbuf *, int how); 790 791 /*- 792 * Network packets may have annotations attached by affixing a list of 793 * "packet tags" to the pkthdr structure. Packet tags are dynamically 794 * allocated semi-opaque data structures that have a fixed header 795 * (struct m_tag) that specifies the size of the memory block and a 796 * <cookie,type> pair that identifies it. The cookie is a 32-bit unique 797 * unsigned value used to identify a module or ABI. By convention this value 798 * is chosen as the date+time that the module is created, expressed as the 799 * number of seconds since the epoch (e.g., using date -u +'%s'). The type 800 * value is an ABI/module-specific value that identifies a particular 801 * annotation and is private to the module. For compatibility with systems 802 * like OpenBSD that define packet tags w/o an ABI/module cookie, the value 803 * PACKET_ABI_COMPAT is used to implement m_tag_get and m_tag_find 804 * compatibility shim functions and several tag types are defined below. 805 * Users that do not require compatibility should use a private cookie value 806 * so that packet tag-related definitions can be maintained privately. 807 * 808 * Note that the packet tag returned by m_tag_alloc has the default memory 809 * alignment implemented by malloc. To reference private data one can use a 810 * construct like: 811 * 812 * struct m_tag *mtag = m_tag_alloc(...); 813 * struct foo *p = (struct foo *)(mtag+1); 814 * 815 * if the alignment of struct m_tag is sufficient for referencing members of 816 * struct foo. Otherwise it is necessary to embed struct m_tag within the 817 * private data structure to insure proper alignment; e.g., 818 * 819 * struct foo { 820 * struct m_tag tag; 821 * ... 822 * }; 823 * struct foo *p = (struct foo *) m_tag_alloc(...); 824 * struct m_tag *mtag = &p->tag; 825 */ 826 827 /* 828 * Persistent tags stay with an mbuf until the mbuf is reclaimed. Otherwise 829 * tags are expected to ``vanish'' when they pass through a network 830 * interface. For most interfaces this happens normally as the tags are 831 * reclaimed when the mbuf is free'd. However in some special cases 832 * reclaiming must be done manually. An example is packets that pass through 833 * the loopback interface. Also, one must be careful to do this when 834 * ``turning around'' packets (e.g., icmp_reflect). 835 * 836 * To mark a tag persistent bit-or this flag in when defining the tag id. 837 * The tag will then be treated as described above. 838 */ 839 #define MTAG_PERSISTENT 0x800 840 841 #define PACKET_TAG_NONE 0 /* Nadda */ 842 843 /* Packet tags for use with PACKET_ABI_COMPAT. */ 844 #define PACKET_TAG_IPSEC_IN_DONE 1 /* IPsec applied, in */ 845 #define PACKET_TAG_IPSEC_OUT_DONE 2 /* IPsec applied, out */ 846 #define PACKET_TAG_IPSEC_IN_CRYPTO_DONE 3 /* NIC IPsec crypto done */ 847 #define PACKET_TAG_IPSEC_OUT_CRYPTO_NEEDED 4 /* NIC IPsec crypto req'ed */ 848 #define PACKET_TAG_IPSEC_IN_COULD_DO_CRYPTO 5 /* NIC notifies IPsec */ 849 #define PACKET_TAG_IPSEC_PENDING_TDB 6 /* Reminder to do IPsec */ 850 #define PACKET_TAG_BRIDGE 7 /* Bridge processing done */ 851 #define PACKET_TAG_GIF 8 /* GIF processing done */ 852 #define PACKET_TAG_GRE 9 /* GRE processing done */ 853 #define PACKET_TAG_IN_PACKET_CHECKSUM 10 /* NIC checksumming done */ 854 #define PACKET_TAG_ENCAP 11 /* Encap. processing */ 855 #define PACKET_TAG_IPSEC_SOCKET 12 /* IPSEC socket ref */ 856 #define PACKET_TAG_IPSEC_HISTORY 13 /* IPSEC history */ 857 #define PACKET_TAG_IPV6_INPUT 14 /* IPV6 input processing */ 858 #define PACKET_TAG_DUMMYNET 15 /* dummynet info */ 859 #define PACKET_TAG_DIVERT 17 /* divert info */ 860 #define PACKET_TAG_IPFORWARD 18 /* ipforward info */ 861 #define PACKET_TAG_MACLABEL (19 | MTAG_PERSISTENT) /* MAC label */ 862 #define PACKET_TAG_PF 21 /* PF + ALTQ information */ 863 #define PACKET_TAG_RTSOCKFAM 25 /* rtsock sa family */ 864 #define PACKET_TAG_IPOPTIONS 27 /* Saved IP options */ 865 #define PACKET_TAG_CARP 28 /* CARP info */ 866 867 /* Specific cookies and tags. */ 868 869 /* Packet tag routines. */ 870 struct m_tag *m_tag_alloc(u_int32_t, int, int, int); 871 void m_tag_delete(struct mbuf *, struct m_tag *); 872 void m_tag_delete_chain(struct mbuf *, struct m_tag *); 873 void m_tag_free_default(struct m_tag *); 874 struct m_tag *m_tag_locate(struct mbuf *, u_int32_t, int, struct m_tag *); 875 struct m_tag *m_tag_copy(struct m_tag *, int); 876 int m_tag_copy_chain(struct mbuf *, struct mbuf *, int); 877 void m_tag_delete_nonpersistent(struct mbuf *); 878 879 /* 880 * Initialize the list of tags associated with an mbuf. 881 */ 882 static __inline void 883 m_tag_init(struct mbuf *m) 884 { 885 886 SLIST_INIT(&m->m_pkthdr.tags); 887 } 888 889 /* 890 * Set up the contents of a tag. Note that this does not fill in the free 891 * method; the caller is expected to do that. 892 * 893 * XXX probably should be called m_tag_init, but that was already taken. 894 */ 895 static __inline void 896 m_tag_setup(struct m_tag *t, u_int32_t cookie, int type, int len) 897 { 898 899 t->m_tag_id = type; 900 t->m_tag_len = len; 901 t->m_tag_cookie = cookie; 902 } 903 904 /* 905 * Reclaim resources associated with a tag. 906 */ 907 static __inline void 908 m_tag_free(struct m_tag *t) 909 { 910 911 (*t->m_tag_free)(t); 912 } 913 914 /* 915 * Return the first tag associated with an mbuf. 916 */ 917 static __inline struct m_tag * 918 m_tag_first(struct mbuf *m) 919 { 920 921 return (SLIST_FIRST(&m->m_pkthdr.tags)); 922 } 923 924 /* 925 * Return the next tag in the list of tags associated with an mbuf. 926 */ 927 static __inline struct m_tag * 928 m_tag_next(struct mbuf *m, struct m_tag *t) 929 { 930 931 return (SLIST_NEXT(t, m_tag_link)); 932 } 933 934 /* 935 * Prepend a tag to the list of tags associated with an mbuf. 936 */ 937 static __inline void 938 m_tag_prepend(struct mbuf *m, struct m_tag *t) 939 { 940 941 SLIST_INSERT_HEAD(&m->m_pkthdr.tags, t, m_tag_link); 942 } 943 944 /* 945 * Unlink a tag from the list of tags associated with an mbuf. 946 */ 947 static __inline void 948 m_tag_unlink(struct mbuf *m, struct m_tag *t) 949 { 950 951 SLIST_REMOVE(&m->m_pkthdr.tags, t, m_tag, m_tag_link); 952 } 953 954 /* These are for OpenBSD compatibility. */ 955 #define MTAG_ABI_COMPAT 0 /* compatibility ABI */ 956 957 static __inline struct m_tag * 958 m_tag_get(int type, int length, int wait) 959 { 960 return (m_tag_alloc(MTAG_ABI_COMPAT, type, length, wait)); 961 } 962 963 static __inline struct m_tag * 964 m_tag_find(struct mbuf *m, int type, struct m_tag *start) 965 { 966 return (SLIST_EMPTY(&m->m_pkthdr.tags) ? (struct m_tag *)NULL : 967 m_tag_locate(m, MTAG_ABI_COMPAT, type, start)); 968 } 969 970 /* XXX temporary FIB methods probably eventually use tags.*/ 971 #define M_FIBSHIFT 28 972 #define M_FIBMASK 0x0F 973 974 /* get the fib from an mbuf and if it is not set, return the default */ 975 #define M_GETFIB(_m) \ 976 ((((_m)->m_flags & M_FIB) >> M_FIBSHIFT) & M_FIBMASK) 977 978 #define M_SETFIB(_m, _fib) do { \ 979 _m->m_flags &= ~M_FIB; \ 980 _m->m_flags |= (((_fib) << M_FIBSHIFT) & M_FIB); \ 981 } while (0) 982 983 #endif /* _KERNEL */ 984 985 #ifdef MBUF_PROFILING 986 void m_profile(struct mbuf *m); 987 #define M_PROFILE(m) m_profile(m) 988 #else 989 #define M_PROFILE(m) 990 #endif 991 992 993 #endif /* !_SYS_MBUF_H_ */ 994