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 union { 126 u_int16_t vt_vtag; /* Ethernet 802.1p+q vlan tag */ 127 u_int16_t vt_nrecs; /* # of IGMPv3 records in this chain */ 128 } PH_vt; 129 SLIST_HEAD(packet_tags, m_tag) tags; /* list of packet tags */ 130 }; 131 #define ether_vtag PH_vt.vt_vtag 132 133 /* 134 * Description of external storage mapped into mbuf; valid only if M_EXT is 135 * set. 136 */ 137 struct m_ext { 138 caddr_t ext_buf; /* start of buffer */ 139 void (*ext_free) /* free routine if not the usual */ 140 (void *, void *); 141 void *ext_arg1; /* optional argument pointer */ 142 void *ext_arg2; /* optional argument pointer */ 143 u_int ext_size; /* size of buffer, for ext_free */ 144 volatile u_int *ref_cnt; /* pointer to ref count info */ 145 int ext_type; /* type of external storage */ 146 }; 147 148 /* 149 * The core of the mbuf object along with some shortcut defines for practical 150 * purposes. 151 */ 152 struct mbuf { 153 struct m_hdr m_hdr; 154 union { 155 struct { 156 struct pkthdr MH_pkthdr; /* M_PKTHDR set */ 157 union { 158 struct m_ext MH_ext; /* M_EXT set */ 159 char MH_databuf[MHLEN]; 160 } MH_dat; 161 } MH; 162 char M_databuf[MLEN]; /* !M_PKTHDR, !M_EXT */ 163 } M_dat; 164 }; 165 #define m_next m_hdr.mh_next 166 #define m_len m_hdr.mh_len 167 #define m_data m_hdr.mh_data 168 #define m_type m_hdr.mh_type 169 #define m_flags m_hdr.mh_flags 170 #define m_nextpkt m_hdr.mh_nextpkt 171 #define m_act m_nextpkt 172 #define m_pkthdr M_dat.MH.MH_pkthdr 173 #define m_ext M_dat.MH.MH_dat.MH_ext 174 #define m_pktdat M_dat.MH.MH_dat.MH_databuf 175 #define m_dat M_dat.M_databuf 176 177 /* 178 * mbuf flags. 179 */ 180 #define M_EXT 0x00000001 /* has associated external storage */ 181 #define M_PKTHDR 0x00000002 /* start of record */ 182 #define M_EOR 0x00000004 /* end of record */ 183 #define M_RDONLY 0x00000008 /* associated data is marked read-only */ 184 #define M_PROTO1 0x00000010 /* protocol-specific */ 185 #define M_PROTO2 0x00000020 /* protocol-specific */ 186 #define M_PROTO3 0x00000040 /* protocol-specific */ 187 #define M_PROTO4 0x00000080 /* protocol-specific */ 188 #define M_PROTO5 0x00000100 /* protocol-specific */ 189 #define M_BCAST 0x00000200 /* send/received as link-level broadcast */ 190 #define M_MCAST 0x00000400 /* send/received as link-level multicast */ 191 #define M_FRAG 0x00000800 /* packet is a fragment of a larger packet */ 192 #define M_FIRSTFRAG 0x00001000 /* packet is first fragment */ 193 #define M_LASTFRAG 0x00002000 /* packet is last fragment */ 194 #define M_SKIP_FIREWALL 0x00004000 /* skip firewall processing */ 195 #define M_FREELIST 0x00008000 /* mbuf is on the free list */ 196 #define M_VLANTAG 0x00010000 /* ether_vtag is valid */ 197 #define M_PROMISC 0x00020000 /* packet was not for us */ 198 #define M_NOFREE 0x00040000 /* do not free mbuf, embedded in cluster */ 199 #define M_PROTO6 0x00080000 /* protocol-specific */ 200 #define M_PROTO7 0x00100000 /* protocol-specific */ 201 #define M_PROTO8 0x00200000 /* protocol-specific */ 202 #define M_FLOWID 0x00400000 /* flowid is valid */ 203 /* 204 * For RELENG_{6,7} steal these flags for limited multiple routing table 205 * support. In RELENG_8 and beyond, use just one flag and a tag. 206 */ 207 #define M_FIB 0xF0000000 /* steal some bits to store fib number. */ 208 209 #define M_NOTIFICATION M_PROTO5 /* SCTP notification */ 210 211 /* 212 * Flags to purge when crossing layers. 213 */ 214 #define M_PROTOFLAGS \ 215 (M_PROTO1|M_PROTO2|M_PROTO3|M_PROTO4|M_PROTO5|M_PROTO6|M_PROTO7|M_PROTO8) 216 217 /* 218 * Flags preserved when copying m_pkthdr. 219 */ 220 #define M_COPYFLAGS \ 221 (M_PKTHDR|M_EOR|M_RDONLY|M_PROTOFLAGS|M_SKIP_FIREWALL|M_BCAST|M_MCAST|\ 222 M_FRAG|M_FIRSTFRAG|M_LASTFRAG|M_VLANTAG|M_PROMISC|M_FIB) 223 224 /* 225 * External buffer types: identify ext_buf type. 226 */ 227 #define EXT_CLUSTER 1 /* mbuf cluster */ 228 #define EXT_SFBUF 2 /* sendfile(2)'s sf_bufs */ 229 #define EXT_JUMBOP 3 /* jumbo cluster 4096 bytes */ 230 #define EXT_JUMBO9 4 /* jumbo cluster 9216 bytes */ 231 #define EXT_JUMBO16 5 /* jumbo cluster 16184 bytes */ 232 #define EXT_PACKET 6 /* mbuf+cluster from packet zone */ 233 #define EXT_MBUF 7 /* external mbuf reference (M_IOVEC) */ 234 #define EXT_NET_DRV 100 /* custom ext_buf provided by net driver(s) */ 235 #define EXT_MOD_TYPE 200 /* custom module's ext_buf type */ 236 #define EXT_DISPOSABLE 300 /* can throw this buffer away w/page flipping */ 237 #define EXT_EXTREF 400 /* has externally maintained ref_cnt ptr */ 238 239 /* 240 * Flags indicating hw checksum support and sw checksum requirements. This 241 * field can be directly tested against if_data.ifi_hwassist. 242 */ 243 #define CSUM_IP 0x0001 /* will csum IP */ 244 #define CSUM_TCP 0x0002 /* will csum TCP */ 245 #define CSUM_UDP 0x0004 /* will csum UDP */ 246 #define CSUM_IP_FRAGS 0x0008 /* will csum IP fragments */ 247 #define CSUM_FRAGMENT 0x0010 /* will do IP fragmentation */ 248 #define CSUM_TSO 0x0020 /* will do TSO */ 249 #define CSUM_SCTP 0x0040 /* will csum SCTP */ 250 251 #define CSUM_IP_CHECKED 0x0100 /* did csum IP */ 252 #define CSUM_IP_VALID 0x0200 /* ... the csum is valid */ 253 #define CSUM_DATA_VALID 0x0400 /* csum_data field is valid */ 254 #define CSUM_PSEUDO_HDR 0x0800 /* csum_data has pseudo hdr */ 255 #define CSUM_SCTP_VALID 0x1000 /* SCTP checksum is valid */ 256 257 #define CSUM_DELAY_DATA (CSUM_TCP | CSUM_UDP) 258 #define CSUM_DELAY_IP (CSUM_IP) /* XXX add ipv6 here too? */ 259 260 /* 261 * mbuf types. 262 */ 263 #define MT_NOTMBUF 0 /* USED INTERNALLY ONLY! Object is not mbuf */ 264 #define MT_DATA 1 /* dynamic (data) allocation */ 265 #define MT_HEADER MT_DATA /* packet header, use M_PKTHDR instead */ 266 #define MT_SONAME 8 /* socket name */ 267 #define MT_CONTROL 14 /* extra-data protocol message */ 268 #define MT_OOBDATA 15 /* expedited data */ 269 #define MT_NTYPES 16 /* number of mbuf types for mbtypes[] */ 270 271 #define MT_NOINIT 255 /* Not a type but a flag to allocate 272 a non-initialized mbuf */ 273 274 #define MB_NOTAGS 0x1UL /* no tags attached to mbuf */ 275 276 /* 277 * General mbuf allocator statistics structure. 278 * 279 * Many of these statistics are no longer used; we instead track many 280 * allocator statistics through UMA's built in statistics mechanism. 281 */ 282 struct mbstat { 283 u_long m_mbufs; /* XXX */ 284 u_long m_mclusts; /* XXX */ 285 286 u_long m_drain; /* times drained protocols for space */ 287 u_long m_mcfail; /* XXX: times m_copym failed */ 288 u_long m_mpfail; /* XXX: times m_pullup failed */ 289 u_long m_msize; /* length of an mbuf */ 290 u_long m_mclbytes; /* length of an mbuf cluster */ 291 u_long m_minclsize; /* min length of data to allocate a cluster */ 292 u_long m_mlen; /* length of data in an mbuf */ 293 u_long m_mhlen; /* length of data in a header mbuf */ 294 295 /* Number of mbtypes (gives # elems in mbtypes[] array) */ 296 short m_numtypes; 297 298 /* XXX: Sendfile stats should eventually move to their own struct */ 299 u_long sf_iocnt; /* times sendfile had to do disk I/O */ 300 u_long sf_allocfail; /* times sfbuf allocation failed */ 301 u_long sf_allocwait; /* times sfbuf allocation had to wait */ 302 }; 303 304 /* 305 * Flags specifying how an allocation should be made. 306 * 307 * The flag to use is as follows: 308 * - M_DONTWAIT or M_NOWAIT from an interrupt handler to not block allocation. 309 * - M_WAIT or M_WAITOK from wherever it is safe to block. 310 * 311 * M_DONTWAIT/M_NOWAIT means that we will not block the thread explicitly and 312 * if we cannot allocate immediately we may return NULL, whereas 313 * M_WAIT/M_WAITOK means that if we cannot allocate resources we 314 * will block until they are available, and thus never return NULL. 315 * 316 * XXX Eventually just phase this out to use M_WAITOK/M_NOWAIT. 317 */ 318 #define MBTOM(how) (how) 319 #define M_DONTWAIT M_NOWAIT 320 #define M_TRYWAIT M_WAITOK 321 #define M_WAIT M_WAITOK 322 323 /* 324 * String names of mbuf-related UMA(9) and malloc(9) types. Exposed to 325 * !_KERNEL so that monitoring tools can look up the zones with 326 * libmemstat(3). 327 */ 328 #define MBUF_MEM_NAME "mbuf" 329 #define MBUF_CLUSTER_MEM_NAME "mbuf_cluster" 330 #define MBUF_PACKET_MEM_NAME "mbuf_packet" 331 #define MBUF_JUMBOP_MEM_NAME "mbuf_jumbo_page" 332 #define MBUF_JUMBO9_MEM_NAME "mbuf_jumbo_9k" 333 #define MBUF_JUMBO16_MEM_NAME "mbuf_jumbo_16k" 334 #define MBUF_TAG_MEM_NAME "mbuf_tag" 335 #define MBUF_EXTREFCNT_MEM_NAME "mbuf_ext_refcnt" 336 337 #ifdef _KERNEL 338 339 #ifdef WITNESS 340 #define MBUF_CHECKSLEEP(how) do { \ 341 if (how == M_WAITOK) \ 342 WITNESS_WARN(WARN_GIANTOK | WARN_SLEEPOK, NULL, \ 343 "Sleeping in \"%s\"", __func__); \ 344 } while (0) 345 #else 346 #define MBUF_CHECKSLEEP(how) 347 #endif 348 349 /* 350 * Network buffer allocation API 351 * 352 * The rest of it is defined in kern/kern_mbuf.c 353 */ 354 355 extern uma_zone_t zone_mbuf; 356 extern uma_zone_t zone_clust; 357 extern uma_zone_t zone_pack; 358 extern uma_zone_t zone_jumbop; 359 extern uma_zone_t zone_jumbo9; 360 extern uma_zone_t zone_jumbo16; 361 extern uma_zone_t zone_ext_refcnt; 362 363 static __inline struct mbuf *m_getcl(int how, short type, int flags); 364 static __inline struct mbuf *m_get(int how, short type); 365 static __inline struct mbuf *m_gethdr(int how, short type); 366 static __inline struct mbuf *m_getjcl(int how, short type, int flags, 367 int size); 368 static __inline struct mbuf *m_getclr(int how, short type); /* XXX */ 369 static __inline struct mbuf *m_free(struct mbuf *m); 370 static __inline void m_clget(struct mbuf *m, int how); 371 static __inline void *m_cljget(struct mbuf *m, int how, int size); 372 static __inline void m_chtype(struct mbuf *m, short new_type); 373 void mb_free_ext(struct mbuf *); 374 static __inline struct mbuf *m_last(struct mbuf *m); 375 376 static __inline int 377 m_gettype(int size) 378 { 379 int type; 380 381 switch (size) { 382 case MSIZE: 383 type = EXT_MBUF; 384 break; 385 case MCLBYTES: 386 type = EXT_CLUSTER; 387 break; 388 #if MJUMPAGESIZE != MCLBYTES 389 case MJUMPAGESIZE: 390 type = EXT_JUMBOP; 391 break; 392 #endif 393 case MJUM9BYTES: 394 type = EXT_JUMBO9; 395 break; 396 case MJUM16BYTES: 397 type = EXT_JUMBO16; 398 break; 399 default: 400 panic("%s: m_getjcl: invalid cluster size", __func__); 401 } 402 403 return (type); 404 } 405 406 static __inline uma_zone_t 407 m_getzone(int size) 408 { 409 uma_zone_t zone; 410 411 switch (size) { 412 case MSIZE: 413 zone = zone_mbuf; 414 break; 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 } 432 433 return (zone); 434 } 435 436 static __inline struct mbuf * 437 m_get(int how, short type) 438 { 439 struct mb_args args; 440 441 args.flags = 0; 442 args.type = type; 443 return ((struct mbuf *)(uma_zalloc_arg(zone_mbuf, &args, how))); 444 } 445 446 /* 447 * XXX This should be deprecated, very little use. 448 */ 449 static __inline struct mbuf * 450 m_getclr(int how, short type) 451 { 452 struct mbuf *m; 453 struct mb_args args; 454 455 args.flags = 0; 456 args.type = type; 457 m = uma_zalloc_arg(zone_mbuf, &args, how); 458 if (m != NULL) 459 bzero(m->m_data, MLEN); 460 return (m); 461 } 462 463 static __inline struct mbuf * 464 m_gethdr(int how, short type) 465 { 466 struct mb_args args; 467 468 args.flags = M_PKTHDR; 469 args.type = type; 470 return ((struct mbuf *)(uma_zalloc_arg(zone_mbuf, &args, how))); 471 } 472 473 static __inline struct mbuf * 474 m_getcl(int how, short type, int flags) 475 { 476 struct mb_args args; 477 478 args.flags = flags; 479 args.type = type; 480 return ((struct mbuf *)(uma_zalloc_arg(zone_pack, &args, how))); 481 } 482 483 /* 484 * m_getjcl() returns an mbuf with a cluster of the specified size attached. 485 * For size it takes MCLBYTES, MJUMPAGESIZE, MJUM9BYTES, MJUM16BYTES. 486 * 487 * XXX: This is rather large, should be real function maybe. 488 */ 489 static __inline struct mbuf * 490 m_getjcl(int how, short type, int flags, int size) 491 { 492 struct mb_args args; 493 struct mbuf *m, *n; 494 uma_zone_t zone; 495 496 args.flags = flags; 497 args.type = type; 498 499 m = uma_zalloc_arg(zone_mbuf, &args, how); 500 if (m == NULL) 501 return (NULL); 502 503 zone = m_getzone(size); 504 n = uma_zalloc_arg(zone, m, how); 505 if (n == NULL) { 506 uma_zfree(zone_mbuf, m); 507 return (NULL); 508 } 509 return (m); 510 } 511 512 static __inline void 513 m_free_fast(struct mbuf *m) 514 { 515 #ifdef INVARIANTS 516 if (m->m_flags & M_PKTHDR) 517 KASSERT(SLIST_EMPTY(&m->m_pkthdr.tags), ("doing fast free of mbuf with tags")); 518 #endif 519 520 uma_zfree_arg(zone_mbuf, m, (void *)MB_NOTAGS); 521 } 522 523 static __inline struct mbuf * 524 m_free(struct mbuf *m) 525 { 526 struct mbuf *n = m->m_next; 527 528 if (m->m_flags & M_EXT) 529 mb_free_ext(m); 530 else if ((m->m_flags & M_NOFREE) == 0) 531 uma_zfree(zone_mbuf, m); 532 return (n); 533 } 534 535 static __inline void 536 m_clget(struct mbuf *m, int how) 537 { 538 539 if (m->m_flags & M_EXT) 540 printf("%s: %p mbuf already has cluster\n", __func__, m); 541 m->m_ext.ext_buf = (char *)NULL; 542 uma_zalloc_arg(zone_clust, m, how); 543 /* 544 * On a cluster allocation failure, drain the packet zone and retry, 545 * we might be able to loosen a few clusters up on the drain. 546 */ 547 if ((how & M_NOWAIT) && (m->m_ext.ext_buf == NULL)) { 548 zone_drain(zone_pack); 549 uma_zalloc_arg(zone_clust, m, how); 550 } 551 } 552 553 /* 554 * m_cljget() is different from m_clget() as it can allocate clusters without 555 * attaching them to an mbuf. In that case the return value is the pointer 556 * to the cluster of the requested size. If an mbuf was specified, it gets 557 * the cluster attached to it and the return value can be safely ignored. 558 * For size it takes MCLBYTES, MJUMPAGESIZE, MJUM9BYTES, MJUM16BYTES. 559 */ 560 static __inline void * 561 m_cljget(struct mbuf *m, int how, int size) 562 { 563 uma_zone_t zone; 564 565 if (m && m->m_flags & M_EXT) 566 printf("%s: %p mbuf already has cluster\n", __func__, m); 567 if (m != NULL) 568 m->m_ext.ext_buf = NULL; 569 570 zone = m_getzone(size); 571 return (uma_zalloc_arg(zone, m, how)); 572 } 573 574 static __inline void 575 m_cljset(struct mbuf *m, void *cl, int type) 576 { 577 uma_zone_t zone; 578 int size; 579 580 switch (type) { 581 case EXT_CLUSTER: 582 size = MCLBYTES; 583 zone = zone_clust; 584 break; 585 #if MJUMPAGESIZE != MCLBYTES 586 case EXT_JUMBOP: 587 size = MJUMPAGESIZE; 588 zone = zone_jumbop; 589 break; 590 #endif 591 case EXT_JUMBO9: 592 size = MJUM9BYTES; 593 zone = zone_jumbo9; 594 break; 595 case EXT_JUMBO16: 596 size = MJUM16BYTES; 597 zone = zone_jumbo16; 598 break; 599 default: 600 panic("unknown cluster type"); 601 break; 602 } 603 604 m->m_data = m->m_ext.ext_buf = cl; 605 m->m_ext.ext_free = m->m_ext.ext_arg1 = m->m_ext.ext_arg2 = NULL; 606 m->m_ext.ext_size = size; 607 m->m_ext.ext_type = type; 608 m->m_ext.ref_cnt = uma_find_refcnt(zone, cl); 609 m->m_flags |= M_EXT; 610 611 } 612 613 static __inline void 614 m_chtype(struct mbuf *m, short new_type) 615 { 616 617 m->m_type = new_type; 618 } 619 620 static __inline struct mbuf * 621 m_last(struct mbuf *m) 622 { 623 624 while (m->m_next) 625 m = m->m_next; 626 return (m); 627 } 628 629 /* 630 * mbuf, cluster, and external object allocation macros (for compatibility 631 * purposes). 632 */ 633 #define M_MOVE_PKTHDR(to, from) m_move_pkthdr((to), (from)) 634 #define MGET(m, how, type) ((m) = m_get((how), (type))) 635 #define MGETHDR(m, how, type) ((m) = m_gethdr((how), (type))) 636 #define MCLGET(m, how) m_clget((m), (how)) 637 #define MEXTADD(m, buf, size, free, arg1, arg2, flags, type) \ 638 m_extadd((m), (caddr_t)(buf), (size), (free),(arg1),(arg2),(flags), (type)) 639 #define m_getm(m, len, how, type) \ 640 m_getm2((m), (len), (how), (type), M_PKTHDR) 641 642 /* 643 * Evaluate TRUE if it's safe to write to the mbuf m's data region (this can 644 * be both the local data payload, or an external buffer area, depending on 645 * whether M_EXT is set). 646 */ 647 #define M_WRITABLE(m) (!((m)->m_flags & M_RDONLY) && \ 648 (!(((m)->m_flags & M_EXT)) || \ 649 (*((m)->m_ext.ref_cnt) == 1)) ) \ 650 651 /* Check if the supplied mbuf has a packet header, or else panic. */ 652 #define M_ASSERTPKTHDR(m) \ 653 KASSERT((m) != NULL && (m)->m_flags & M_PKTHDR, \ 654 ("%s: no mbuf packet header!", __func__)) 655 656 /* 657 * Ensure that the supplied mbuf is a valid, non-free mbuf. 658 * 659 * XXX: Broken at the moment. Need some UMA magic to make it work again. 660 */ 661 #define M_ASSERTVALID(m) \ 662 KASSERT((((struct mbuf *)m)->m_flags & 0) == 0, \ 663 ("%s: attempted use of a free mbuf!", __func__)) 664 665 /* 666 * Set the m_data pointer of a newly-allocated mbuf (m_get/MGET) to place an 667 * object of the specified size at the end of the mbuf, longword aligned. 668 */ 669 #define M_ALIGN(m, len) do { \ 670 KASSERT(!((m)->m_flags & (M_PKTHDR|M_EXT)), \ 671 ("%s: M_ALIGN not normal mbuf", __func__)); \ 672 KASSERT((m)->m_data == (m)->m_dat, \ 673 ("%s: M_ALIGN not a virgin mbuf", __func__)); \ 674 (m)->m_data += (MLEN - (len)) & ~(sizeof(long) - 1); \ 675 } while (0) 676 677 /* 678 * As above, for mbufs allocated with m_gethdr/MGETHDR or initialized by 679 * M_DUP/MOVE_PKTHDR. 680 */ 681 #define MH_ALIGN(m, len) do { \ 682 KASSERT((m)->m_flags & M_PKTHDR && !((m)->m_flags & M_EXT), \ 683 ("%s: MH_ALIGN not PKTHDR mbuf", __func__)); \ 684 KASSERT((m)->m_data == (m)->m_pktdat, \ 685 ("%s: MH_ALIGN not a virgin mbuf", __func__)); \ 686 (m)->m_data += (MHLEN - (len)) & ~(sizeof(long) - 1); \ 687 } while (0) 688 689 /* 690 * Compute the amount of space available before the current start of data in 691 * an mbuf. 692 * 693 * The M_WRITABLE() is a temporary, conservative safety measure: the burden 694 * of checking writability of the mbuf data area rests solely with the caller. 695 */ 696 #define M_LEADINGSPACE(m) \ 697 ((m)->m_flags & M_EXT ? \ 698 (M_WRITABLE(m) ? (m)->m_data - (m)->m_ext.ext_buf : 0): \ 699 (m)->m_flags & M_PKTHDR ? (m)->m_data - (m)->m_pktdat : \ 700 (m)->m_data - (m)->m_dat) 701 702 /* 703 * Compute the amount of space available after the end of data in an mbuf. 704 * 705 * The M_WRITABLE() is a temporary, conservative safety measure: the burden 706 * of checking writability of the mbuf data area rests solely with the caller. 707 */ 708 #define M_TRAILINGSPACE(m) \ 709 ((m)->m_flags & M_EXT ? \ 710 (M_WRITABLE(m) ? (m)->m_ext.ext_buf + (m)->m_ext.ext_size \ 711 - ((m)->m_data + (m)->m_len) : 0) : \ 712 &(m)->m_dat[MLEN] - ((m)->m_data + (m)->m_len)) 713 714 /* 715 * Arrange to prepend space of size plen to mbuf m. If a new mbuf must be 716 * allocated, how specifies whether to wait. If the allocation fails, the 717 * original mbuf chain is freed and m is set to NULL. 718 */ 719 #define M_PREPEND(m, plen, how) do { \ 720 struct mbuf **_mmp = &(m); \ 721 struct mbuf *_mm = *_mmp; \ 722 int _mplen = (plen); \ 723 int __mhow = (how); \ 724 \ 725 MBUF_CHECKSLEEP(how); \ 726 if (M_LEADINGSPACE(_mm) >= _mplen) { \ 727 _mm->m_data -= _mplen; \ 728 _mm->m_len += _mplen; \ 729 } else \ 730 _mm = m_prepend(_mm, _mplen, __mhow); \ 731 if (_mm != NULL && _mm->m_flags & M_PKTHDR) \ 732 _mm->m_pkthdr.len += _mplen; \ 733 *_mmp = _mm; \ 734 } while (0) 735 736 /* 737 * Change mbuf to new type. This is a relatively expensive operation and 738 * should be avoided. 739 */ 740 #define MCHTYPE(m, t) m_chtype((m), (t)) 741 742 /* Length to m_copy to copy all. */ 743 #define M_COPYALL 1000000000 744 745 /* Compatibility with 4.3. */ 746 #define m_copy(m, o, l) m_copym((m), (o), (l), M_DONTWAIT) 747 748 extern int max_datalen; /* MHLEN - max_hdr */ 749 extern int max_hdr; /* Largest link + protocol header */ 750 extern int max_linkhdr; /* Largest link-level header */ 751 extern int max_protohdr; /* Largest protocol header */ 752 extern struct mbstat mbstat; /* General mbuf stats/infos */ 753 extern int nmbclusters; /* Maximum number of clusters */ 754 755 struct uio; 756 757 void m_adj(struct mbuf *, int); 758 void m_align(struct mbuf *, int); 759 int m_apply(struct mbuf *, int, int, 760 int (*)(void *, void *, u_int), void *); 761 int m_append(struct mbuf *, int, c_caddr_t); 762 void m_cat(struct mbuf *, struct mbuf *); 763 void m_extadd(struct mbuf *, caddr_t, u_int, 764 void (*)(void *, void *), void *, void *, int, int); 765 struct mbuf *m_collapse(struct mbuf *, int, int); 766 void m_copyback(struct mbuf *, int, int, c_caddr_t); 767 void m_copydata(const struct mbuf *, int, int, caddr_t); 768 struct mbuf *m_copym(struct mbuf *, int, int, int); 769 struct mbuf *m_copymdata(struct mbuf *, struct mbuf *, 770 int, int, int, int); 771 struct mbuf *m_copypacket(struct mbuf *, int); 772 void m_copy_pkthdr(struct mbuf *, struct mbuf *); 773 struct mbuf *m_copyup(struct mbuf *n, int len, int dstoff); 774 struct mbuf *m_defrag(struct mbuf *, int); 775 void m_demote(struct mbuf *, int); 776 struct mbuf *m_devget(char *, int, int, struct ifnet *, 777 void (*)(char *, caddr_t, u_int)); 778 struct mbuf *m_dup(struct mbuf *, int); 779 int m_dup_pkthdr(struct mbuf *, struct mbuf *, int); 780 u_int m_fixhdr(struct mbuf *); 781 struct mbuf *m_fragment(struct mbuf *, int, int); 782 void m_freem(struct mbuf *); 783 struct mbuf *m_getm2(struct mbuf *, int, int, short, int); 784 struct mbuf *m_getptr(struct mbuf *, int, int *); 785 u_int m_length(struct mbuf *, struct mbuf **); 786 void m_move_pkthdr(struct mbuf *, struct mbuf *); 787 struct mbuf *m_prepend(struct mbuf *, int, int); 788 void m_print(const struct mbuf *, int); 789 struct mbuf *m_pulldown(struct mbuf *, int, int, int *); 790 struct mbuf *m_pullup(struct mbuf *, int); 791 int m_sanity(struct mbuf *, int); 792 struct mbuf *m_split(struct mbuf *, int, int); 793 struct mbuf *m_uiotombuf(struct uio *, int, int, int, int); 794 struct mbuf *m_unshare(struct mbuf *, int how); 795 796 /*- 797 * Network packets may have annotations attached by affixing a list of 798 * "packet tags" to the pkthdr structure. Packet tags are dynamically 799 * allocated semi-opaque data structures that have a fixed header 800 * (struct m_tag) that specifies the size of the memory block and a 801 * <cookie,type> pair that identifies it. The cookie is a 32-bit unique 802 * unsigned value used to identify a module or ABI. By convention this value 803 * is chosen as the date+time that the module is created, expressed as the 804 * number of seconds since the epoch (e.g., using date -u +'%s'). The type 805 * value is an ABI/module-specific value that identifies a particular 806 * annotation and is private to the module. For compatibility with systems 807 * like OpenBSD that define packet tags w/o an ABI/module cookie, the value 808 * PACKET_ABI_COMPAT is used to implement m_tag_get and m_tag_find 809 * compatibility shim functions and several tag types are defined below. 810 * Users that do not require compatibility should use a private cookie value 811 * so that packet tag-related definitions can be maintained privately. 812 * 813 * Note that the packet tag returned by m_tag_alloc has the default memory 814 * alignment implemented by malloc. To reference private data one can use a 815 * construct like: 816 * 817 * struct m_tag *mtag = m_tag_alloc(...); 818 * struct foo *p = (struct foo *)(mtag+1); 819 * 820 * if the alignment of struct m_tag is sufficient for referencing members of 821 * struct foo. Otherwise it is necessary to embed struct m_tag within the 822 * private data structure to insure proper alignment; e.g., 823 * 824 * struct foo { 825 * struct m_tag tag; 826 * ... 827 * }; 828 * struct foo *p = (struct foo *) m_tag_alloc(...); 829 * struct m_tag *mtag = &p->tag; 830 */ 831 832 /* 833 * Persistent tags stay with an mbuf until the mbuf is reclaimed. Otherwise 834 * tags are expected to ``vanish'' when they pass through a network 835 * interface. For most interfaces this happens normally as the tags are 836 * reclaimed when the mbuf is free'd. However in some special cases 837 * reclaiming must be done manually. An example is packets that pass through 838 * the loopback interface. Also, one must be careful to do this when 839 * ``turning around'' packets (e.g., icmp_reflect). 840 * 841 * To mark a tag persistent bit-or this flag in when defining the tag id. 842 * The tag will then be treated as described above. 843 */ 844 #define MTAG_PERSISTENT 0x800 845 846 #define PACKET_TAG_NONE 0 /* Nadda */ 847 848 /* Packet tags for use with PACKET_ABI_COMPAT. */ 849 #define PACKET_TAG_IPSEC_IN_DONE 1 /* IPsec applied, in */ 850 #define PACKET_TAG_IPSEC_OUT_DONE 2 /* IPsec applied, out */ 851 #define PACKET_TAG_IPSEC_IN_CRYPTO_DONE 3 /* NIC IPsec crypto done */ 852 #define PACKET_TAG_IPSEC_OUT_CRYPTO_NEEDED 4 /* NIC IPsec crypto req'ed */ 853 #define PACKET_TAG_IPSEC_IN_COULD_DO_CRYPTO 5 /* NIC notifies IPsec */ 854 #define PACKET_TAG_IPSEC_PENDING_TDB 6 /* Reminder to do IPsec */ 855 #define PACKET_TAG_BRIDGE 7 /* Bridge processing done */ 856 #define PACKET_TAG_GIF 8 /* GIF processing done */ 857 #define PACKET_TAG_GRE 9 /* GRE processing done */ 858 #define PACKET_TAG_IN_PACKET_CHECKSUM 10 /* NIC checksumming done */ 859 #define PACKET_TAG_ENCAP 11 /* Encap. processing */ 860 #define PACKET_TAG_IPSEC_SOCKET 12 /* IPSEC socket ref */ 861 #define PACKET_TAG_IPSEC_HISTORY 13 /* IPSEC history */ 862 #define PACKET_TAG_IPV6_INPUT 14 /* IPV6 input processing */ 863 #define PACKET_TAG_DUMMYNET 15 /* dummynet info */ 864 #define PACKET_TAG_DIVERT 17 /* divert info */ 865 #define PACKET_TAG_IPFORWARD 18 /* ipforward info */ 866 #define PACKET_TAG_MACLABEL (19 | MTAG_PERSISTENT) /* MAC label */ 867 #define PACKET_TAG_PF 21 /* PF + ALTQ information */ 868 #define PACKET_TAG_RTSOCKFAM 25 /* rtsock sa family */ 869 #define PACKET_TAG_IPOPTIONS 27 /* Saved IP options */ 870 #define PACKET_TAG_CARP 28 /* CARP info */ 871 872 /* Specific cookies and tags. */ 873 874 /* Packet tag routines. */ 875 struct m_tag *m_tag_alloc(u_int32_t, int, int, int); 876 void m_tag_delete(struct mbuf *, struct m_tag *); 877 void m_tag_delete_chain(struct mbuf *, struct m_tag *); 878 void m_tag_free_default(struct m_tag *); 879 struct m_tag *m_tag_locate(struct mbuf *, u_int32_t, int, struct m_tag *); 880 struct m_tag *m_tag_copy(struct m_tag *, int); 881 int m_tag_copy_chain(struct mbuf *, struct mbuf *, int); 882 void m_tag_delete_nonpersistent(struct mbuf *); 883 884 /* 885 * Initialize the list of tags associated with an mbuf. 886 */ 887 static __inline void 888 m_tag_init(struct mbuf *m) 889 { 890 891 SLIST_INIT(&m->m_pkthdr.tags); 892 } 893 894 /* 895 * Set up the contents of a tag. Note that this does not fill in the free 896 * method; the caller is expected to do that. 897 * 898 * XXX probably should be called m_tag_init, but that was already taken. 899 */ 900 static __inline void 901 m_tag_setup(struct m_tag *t, u_int32_t cookie, int type, int len) 902 { 903 904 t->m_tag_id = type; 905 t->m_tag_len = len; 906 t->m_tag_cookie = cookie; 907 } 908 909 /* 910 * Reclaim resources associated with a tag. 911 */ 912 static __inline void 913 m_tag_free(struct m_tag *t) 914 { 915 916 (*t->m_tag_free)(t); 917 } 918 919 /* 920 * Return the first tag associated with an mbuf. 921 */ 922 static __inline struct m_tag * 923 m_tag_first(struct mbuf *m) 924 { 925 926 return (SLIST_FIRST(&m->m_pkthdr.tags)); 927 } 928 929 /* 930 * Return the next tag in the list of tags associated with an mbuf. 931 */ 932 static __inline struct m_tag * 933 m_tag_next(struct mbuf *m, struct m_tag *t) 934 { 935 936 return (SLIST_NEXT(t, m_tag_link)); 937 } 938 939 /* 940 * Prepend a tag to the list of tags associated with an mbuf. 941 */ 942 static __inline void 943 m_tag_prepend(struct mbuf *m, struct m_tag *t) 944 { 945 946 SLIST_INSERT_HEAD(&m->m_pkthdr.tags, t, m_tag_link); 947 } 948 949 /* 950 * Unlink a tag from the list of tags associated with an mbuf. 951 */ 952 static __inline void 953 m_tag_unlink(struct mbuf *m, struct m_tag *t) 954 { 955 956 SLIST_REMOVE(&m->m_pkthdr.tags, t, m_tag, m_tag_link); 957 } 958 959 /* These are for OpenBSD compatibility. */ 960 #define MTAG_ABI_COMPAT 0 /* compatibility ABI */ 961 962 static __inline struct m_tag * 963 m_tag_get(int type, int length, int wait) 964 { 965 return (m_tag_alloc(MTAG_ABI_COMPAT, type, length, wait)); 966 } 967 968 static __inline struct m_tag * 969 m_tag_find(struct mbuf *m, int type, struct m_tag *start) 970 { 971 return (SLIST_EMPTY(&m->m_pkthdr.tags) ? (struct m_tag *)NULL : 972 m_tag_locate(m, MTAG_ABI_COMPAT, type, start)); 973 } 974 975 /* XXX temporary FIB methods probably eventually use tags.*/ 976 #define M_FIBSHIFT 28 977 #define M_FIBMASK 0x0F 978 979 /* get the fib from an mbuf and if it is not set, return the default */ 980 #define M_GETFIB(_m) \ 981 ((((_m)->m_flags & M_FIB) >> M_FIBSHIFT) & M_FIBMASK) 982 983 #define M_SETFIB(_m, _fib) do { \ 984 _m->m_flags &= ~M_FIB; \ 985 _m->m_flags |= (((_fib) << M_FIBSHIFT) & M_FIB); \ 986 } while (0) 987 988 #endif /* _KERNEL */ 989 990 #ifdef MBUF_PROFILING 991 void m_profile(struct mbuf *m); 992 #define M_PROFILE(m) m_profile(m) 993 #else 994 #define M_PROFILE(m) 995 #endif 996 997 998 #endif /* !_SYS_MBUF_H_ */ 999