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 * Macro for type conversion: convert mbuf pointer to data pointer of correct 64 * type: 65 * 66 * mtod(m, t) -- Convert mbuf pointer to data pointer of correct type. 67 */ 68 #define mtod(m, t) ((t)((m)->m_data)) 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 /* deprecated: flowid is valid */ 203 #define M_HASHTYPEBITS 0x0F000000 /* mask of bits holding flowid hash type */ 204 205 /* 206 * For RELENG_{6,7} steal these flags for limited multiple routing table 207 * support. In RELENG_8 and beyond, use just one flag and a tag. 208 */ 209 #define M_FIB 0xF0000000 /* steal some bits to store fib number. */ 210 211 #define M_NOTIFICATION M_PROTO5 /* SCTP notification */ 212 213 /* 214 * Flags to purge when crossing layers. 215 */ 216 #define M_PROTOFLAGS \ 217 (M_PROTO1|M_PROTO2|M_PROTO3|M_PROTO4|M_PROTO5|M_PROTO6|M_PROTO7|M_PROTO8) 218 219 /* 220 * Network interface cards are able to hash protocol fields (such as IPv4 221 * addresses and TCP port numbers) classify packets into flows. These flows 222 * can then be used to maintain ordering while delivering packets to the OS 223 * via parallel input queues, as well as to provide a stateless affinity 224 * model. NIC drivers can pass up the hash via m->m_pkthdr.flowid, and set 225 * m_flag fields to indicate how the hash should be interpreted by the 226 * network stack. 227 * 228 * Most NICs support RSS, which provides ordering and explicit affinity, and 229 * use the hash m_flag bits to indicate what header fields were covered by 230 * the hash. M_HASHTYPE_OPAQUE can be set by non-RSS cards or configurations 231 * that provide an opaque flow identifier, allowing for ordering and 232 * distribution without explicit affinity. 233 */ 234 #define M_HASHTYPE_SHIFT 24 235 #define M_HASHTYPE_NONE 0x0 236 #define M_HASHTYPE_RSS_IPV4 0x1 /* IPv4 2-tuple */ 237 #define M_HASHTYPE_RSS_TCP_IPV4 0x2 /* TCPv4 4-tuple */ 238 #define M_HASHTYPE_RSS_IPV6 0x3 /* IPv6 2-tuple */ 239 #define M_HASHTYPE_RSS_TCP_IPV6 0x4 /* TCPv6 4-tuple */ 240 #define M_HASHTYPE_RSS_IPV6_EX 0x5 /* IPv6 2-tuple + ext hdrs */ 241 #define M_HASHTYPE_RSS_TCP_IPV6_EX 0x6 /* TCPv6 4-tiple + ext hdrs */ 242 #define M_HASHTYPE_OPAQUE 0xf /* ordering, not affinity */ 243 244 #define M_HASHTYPE_CLEAR(m) (m)->m_flags &= ~(M_HASHTYPEBITS) 245 #define M_HASHTYPE_GET(m) (((m)->m_flags & M_HASHTYPEBITS) >> \ 246 M_HASHTYPE_SHIFT) 247 #define M_HASHTYPE_SET(m, v) do { \ 248 (m)->m_flags &= ~M_HASHTYPEBITS; \ 249 (m)->m_flags |= ((v) << M_HASHTYPE_SHIFT); \ 250 } while (0) 251 #define M_HASHTYPE_TEST(m, v) (M_HASHTYPE_GET(m) == (v)) 252 253 /* 254 * Flags preserved when copying m_pkthdr. 255 */ 256 #define M_COPYFLAGS \ 257 (M_PKTHDR|M_EOR|M_RDONLY|M_PROTOFLAGS|M_SKIP_FIREWALL|M_BCAST|M_MCAST|\ 258 M_FRAG|M_FIRSTFRAG|M_LASTFRAG|M_VLANTAG|M_PROMISC|M_FIB|M_HASHTYPEBITS) 259 260 /* 261 * External buffer types: identify ext_buf type. 262 */ 263 #define EXT_CLUSTER 1 /* mbuf cluster */ 264 #define EXT_SFBUF 2 /* sendfile(2)'s sf_bufs */ 265 #define EXT_JUMBOP 3 /* jumbo cluster 4096 bytes */ 266 #define EXT_JUMBO9 4 /* jumbo cluster 9216 bytes */ 267 #define EXT_JUMBO16 5 /* jumbo cluster 16184 bytes */ 268 #define EXT_PACKET 6 /* mbuf+cluster from packet zone */ 269 #define EXT_MBUF 7 /* external mbuf reference (M_IOVEC) */ 270 #define EXT_NET_DRV 100 /* custom ext_buf provided by net driver(s) */ 271 #define EXT_MOD_TYPE 200 /* custom module's ext_buf type */ 272 #define EXT_DISPOSABLE 300 /* can throw this buffer away w/page flipping */ 273 #define EXT_EXTREF 400 /* has externally maintained ref_cnt ptr */ 274 275 /* 276 * Flags indicating hw checksum support and sw checksum requirements. This 277 * field can be directly tested against if_data.ifi_hwassist. 278 */ 279 #define CSUM_IP 0x0001 /* will csum IP */ 280 #define CSUM_TCP 0x0002 /* will csum TCP */ 281 #define CSUM_UDP 0x0004 /* will csum UDP */ 282 #define CSUM_FRAGMENT 0x0010 /* will do IP fragmentation */ 283 #define CSUM_TSO 0x0020 /* will do TSO */ 284 #define CSUM_SCTP 0x0040 /* will csum SCTP */ 285 #define CSUM_SCTP_IPV6 0x0080 /* will csum IPv6/SCTP */ 286 287 #define CSUM_IP_CHECKED 0x0100 /* did csum IP */ 288 #define CSUM_IP_VALID 0x0200 /* ... the csum is valid */ 289 #define CSUM_DATA_VALID 0x0400 /* csum_data field is valid */ 290 #define CSUM_PSEUDO_HDR 0x0800 /* csum_data has pseudo hdr */ 291 #define CSUM_SCTP_VALID 0x1000 /* SCTP checksum is valid */ 292 #define CSUM_UDP_IPV6 0x2000 /* will csum IPv6/UDP */ 293 #define CSUM_TCP_IPV6 0x4000 /* will csum IPv6/TCP */ 294 /* CSUM_TSO_IPV6 0x8000 will do IPv6/TSO */ 295 296 /* CSUM_FRAGMENT_IPV6 0x10000 will do IPv6 fragementation */ 297 298 #define CSUM_DELAY_DATA_IPV6 (CSUM_TCP_IPV6 | CSUM_UDP_IPV6) 299 #define CSUM_DATA_VALID_IPV6 CSUM_DATA_VALID 300 301 #define CSUM_DELAY_DATA (CSUM_TCP | CSUM_UDP) 302 #define CSUM_DELAY_IP (CSUM_IP) /* Only v4, no v6 IP hdr csum */ 303 304 /* 305 * mbuf types. 306 */ 307 #define MT_NOTMBUF 0 /* USED INTERNALLY ONLY! Object is not mbuf */ 308 #define MT_DATA 1 /* dynamic (data) allocation */ 309 #define MT_HEADER MT_DATA /* packet header, use M_PKTHDR instead */ 310 #define MT_SONAME 8 /* socket name */ 311 #define MT_CONTROL 14 /* extra-data protocol message */ 312 #define MT_OOBDATA 15 /* expedited data */ 313 #define MT_NTYPES 16 /* number of mbuf types for mbtypes[] */ 314 315 #define MT_NOINIT 255 /* Not a type but a flag to allocate 316 a non-initialized mbuf */ 317 318 #define MB_NOTAGS 0x1UL /* no tags attached to mbuf */ 319 320 /* 321 * General mbuf allocator statistics structure. 322 * 323 * Many of these statistics are no longer used; we instead track many 324 * allocator statistics through UMA's built in statistics mechanism. 325 */ 326 struct mbstat { 327 u_long m_mbufs; /* XXX */ 328 u_long m_mclusts; /* XXX */ 329 330 u_long m_drain; /* times drained protocols for space */ 331 u_long m_mcfail; /* XXX: times m_copym failed */ 332 u_long m_mpfail; /* XXX: times m_pullup failed */ 333 u_long m_msize; /* length of an mbuf */ 334 u_long m_mclbytes; /* length of an mbuf cluster */ 335 u_long m_minclsize; /* min length of data to allocate a cluster */ 336 u_long m_mlen; /* length of data in an mbuf */ 337 u_long m_mhlen; /* length of data in a header mbuf */ 338 339 /* Number of mbtypes (gives # elems in mbtypes[] array) */ 340 short m_numtypes; 341 342 /* XXX: Sendfile stats should eventually move to their own struct */ 343 u_long sf_iocnt; /* times sendfile had to do disk I/O */ 344 u_long sf_allocfail; /* times sfbuf allocation failed */ 345 u_long sf_allocwait; /* times sfbuf allocation had to wait */ 346 }; 347 348 /* 349 * Compatibility with historic mbuf allocator. 350 */ 351 #define MBTOM(how) (how) 352 #define M_DONTWAIT M_NOWAIT 353 #define M_TRYWAIT M_WAITOK 354 #define M_WAIT M_WAITOK 355 356 /* 357 * String names of mbuf-related UMA(9) and malloc(9) types. Exposed to 358 * !_KERNEL so that monitoring tools can look up the zones with 359 * libmemstat(3). 360 */ 361 #define MBUF_MEM_NAME "mbuf" 362 #define MBUF_CLUSTER_MEM_NAME "mbuf_cluster" 363 #define MBUF_PACKET_MEM_NAME "mbuf_packet" 364 #define MBUF_JUMBOP_MEM_NAME "mbuf_jumbo_page" 365 #define MBUF_JUMBO9_MEM_NAME "mbuf_jumbo_9k" 366 #define MBUF_JUMBO16_MEM_NAME "mbuf_jumbo_16k" 367 #define MBUF_TAG_MEM_NAME "mbuf_tag" 368 #define MBUF_EXTREFCNT_MEM_NAME "mbuf_ext_refcnt" 369 370 #ifdef _KERNEL 371 372 #ifdef WITNESS 373 #define MBUF_CHECKSLEEP(how) do { \ 374 if (how == M_WAITOK) \ 375 WITNESS_WARN(WARN_GIANTOK | WARN_SLEEPOK, NULL, \ 376 "Sleeping in \"%s\"", __func__); \ 377 } while (0) 378 #else 379 #define MBUF_CHECKSLEEP(how) 380 #endif 381 382 /* 383 * Network buffer allocation API 384 * 385 * The rest of it is defined in kern/kern_mbuf.c 386 */ 387 extern uma_zone_t zone_mbuf; 388 extern uma_zone_t zone_clust; 389 extern uma_zone_t zone_pack; 390 extern uma_zone_t zone_jumbop; 391 extern uma_zone_t zone_jumbo9; 392 extern uma_zone_t zone_jumbo16; 393 extern uma_zone_t zone_ext_refcnt; 394 395 static __inline struct mbuf *m_getcl(int how, short type, int flags); 396 static __inline struct mbuf *m_get(int how, short type); 397 static __inline struct mbuf *m_get2(int how, short type, int flags, 398 u_int size); 399 static __inline struct mbuf *m_gethdr(int how, short type); 400 static __inline struct mbuf *m_getjcl(int how, short type, int flags, 401 int size); 402 static __inline struct mbuf *m_getclr(int how, short type); /* XXX */ 403 static __inline int m_init(struct mbuf *m, uma_zone_t zone, 404 int size, int how, short type, int flags); 405 static __inline struct mbuf *m_free(struct mbuf *m); 406 static __inline void m_clget(struct mbuf *m, int how); 407 static __inline void *m_cljget(struct mbuf *m, int how, int size); 408 static __inline void m_chtype(struct mbuf *m, short new_type); 409 void mb_free_ext(struct mbuf *); 410 static __inline struct mbuf *m_last(struct mbuf *m); 411 int m_pkthdr_init(struct mbuf *m, int how); 412 413 static __inline int 414 m_gettype(int size) 415 { 416 int type; 417 418 switch (size) { 419 case MSIZE: 420 type = EXT_MBUF; 421 break; 422 case MCLBYTES: 423 type = EXT_CLUSTER; 424 break; 425 #if MJUMPAGESIZE != MCLBYTES 426 case MJUMPAGESIZE: 427 type = EXT_JUMBOP; 428 break; 429 #endif 430 case MJUM9BYTES: 431 type = EXT_JUMBO9; 432 break; 433 case MJUM16BYTES: 434 type = EXT_JUMBO16; 435 break; 436 default: 437 panic("%s: invalid cluster size", __func__); 438 } 439 440 return (type); 441 } 442 443 static __inline uma_zone_t 444 m_getzone(int size) 445 { 446 uma_zone_t zone; 447 448 switch (size) { 449 case MCLBYTES: 450 zone = zone_clust; 451 break; 452 #if MJUMPAGESIZE != MCLBYTES 453 case MJUMPAGESIZE: 454 zone = zone_jumbop; 455 break; 456 #endif 457 case MJUM9BYTES: 458 zone = zone_jumbo9; 459 break; 460 case MJUM16BYTES: 461 zone = zone_jumbo16; 462 break; 463 default: 464 panic("%s: invalid cluster size", __func__); 465 } 466 467 return (zone); 468 } 469 470 /* 471 * Initialize an mbuf with linear storage. 472 * 473 * Inline because the consumer text overhead will be roughly the same to 474 * initialize or call a function with this many parameters and M_PKTHDR 475 * should go away with constant propagation for !MGETHDR. 476 */ 477 static __inline int 478 m_init(struct mbuf *m, uma_zone_t zone, int size, int how, short type, 479 int flags) 480 { 481 int error; 482 483 m->m_next = NULL; 484 m->m_nextpkt = NULL; 485 m->m_data = m->m_dat; 486 m->m_len = 0; 487 m->m_flags = flags; 488 m->m_type = type; 489 if (flags & M_PKTHDR) { 490 if ((error = m_pkthdr_init(m, how)) != 0) 491 return (error); 492 } 493 494 return (0); 495 } 496 497 static __inline struct mbuf * 498 m_get(int how, short type) 499 { 500 struct mb_args args; 501 502 args.flags = 0; 503 args.type = type; 504 return ((struct mbuf *)(uma_zalloc_arg(zone_mbuf, &args, how))); 505 } 506 507 /* 508 * XXX This should be deprecated, very little use. 509 */ 510 static __inline struct mbuf * 511 m_getclr(int how, short type) 512 { 513 struct mbuf *m; 514 struct mb_args args; 515 516 args.flags = 0; 517 args.type = type; 518 m = uma_zalloc_arg(zone_mbuf, &args, how); 519 if (m != NULL) 520 bzero(m->m_data, MLEN); 521 return (m); 522 } 523 524 static __inline struct mbuf * 525 m_gethdr(int how, short type) 526 { 527 struct mb_args args; 528 529 args.flags = M_PKTHDR; 530 args.type = type; 531 return ((struct mbuf *)(uma_zalloc_arg(zone_mbuf, &args, how))); 532 } 533 534 static __inline struct mbuf * 535 m_getcl(int how, short type, int flags) 536 { 537 struct mb_args args; 538 539 args.flags = flags; 540 args.type = type; 541 return ((struct mbuf *)(uma_zalloc_arg(zone_pack, &args, how))); 542 } 543 544 /* 545 * m_get2() allocates minimum mbuf that would fit "size" argument. 546 * 547 * XXX: This is rather large, should be real function maybe. 548 */ 549 static __inline struct mbuf * 550 m_get2(int how, short type, int flags, u_int size) 551 { 552 struct mb_args args; 553 struct mbuf *m, *n; 554 uma_zone_t zone; 555 556 args.flags = flags; 557 args.type = type; 558 559 if (size <= MHLEN || (size <= MLEN && (flags & M_PKTHDR) == 0)) 560 return ((struct mbuf *)(uma_zalloc_arg(zone_mbuf, &args, how))); 561 if (size <= MCLBYTES) 562 return ((struct mbuf *)(uma_zalloc_arg(zone_pack, &args, how))); 563 564 if (size > MJUM16BYTES) 565 return (NULL); 566 567 m = uma_zalloc_arg(zone_mbuf, &args, how); 568 if (m == NULL) 569 return (NULL); 570 571 #if MJUMPAGESIZE != MCLBYTES 572 if (size <= MJUMPAGESIZE) 573 zone = zone_jumbop; 574 else 575 #endif 576 if (size <= MJUM9BYTES) 577 zone = zone_jumbo9; 578 else 579 zone = zone_jumbo16; 580 581 n = uma_zalloc_arg(zone, m, how); 582 if (n == NULL) { 583 uma_zfree(zone_mbuf, m); 584 return (NULL); 585 } 586 587 return (m); 588 } 589 590 /* 591 * m_getjcl() returns an mbuf with a cluster of the specified size attached. 592 * For size it takes MCLBYTES, MJUMPAGESIZE, MJUM9BYTES, MJUM16BYTES. 593 * 594 * XXX: This is rather large, should be real function maybe. 595 */ 596 static __inline struct mbuf * 597 m_getjcl(int how, short type, int flags, int size) 598 { 599 struct mb_args args; 600 struct mbuf *m, *n; 601 uma_zone_t zone; 602 603 if (size == MCLBYTES) 604 return m_getcl(how, type, flags); 605 606 args.flags = flags; 607 args.type = type; 608 609 m = uma_zalloc_arg(zone_mbuf, &args, how); 610 if (m == NULL) 611 return (NULL); 612 613 zone = m_getzone(size); 614 n = uma_zalloc_arg(zone, m, how); 615 if (n == NULL) { 616 uma_zfree(zone_mbuf, m); 617 return (NULL); 618 } 619 return (m); 620 } 621 622 static __inline void 623 m_free_fast(struct mbuf *m) 624 { 625 #ifdef INVARIANTS 626 if (m->m_flags & M_PKTHDR) 627 KASSERT(SLIST_EMPTY(&m->m_pkthdr.tags), ("doing fast free of mbuf with tags")); 628 #endif 629 630 uma_zfree_arg(zone_mbuf, m, (void *)MB_NOTAGS); 631 } 632 633 static __inline struct mbuf * 634 m_free(struct mbuf *m) 635 { 636 struct mbuf *n = m->m_next; 637 638 if (m->m_flags & M_EXT) 639 mb_free_ext(m); 640 else if ((m->m_flags & M_NOFREE) == 0) 641 uma_zfree(zone_mbuf, m); 642 return (n); 643 } 644 645 static __inline void 646 m_clget(struct mbuf *m, int how) 647 { 648 649 if (m->m_flags & M_EXT) 650 printf("%s: %p mbuf already has cluster\n", __func__, m); 651 m->m_ext.ext_buf = (char *)NULL; 652 uma_zalloc_arg(zone_clust, m, how); 653 /* 654 * On a cluster allocation failure, drain the packet zone and retry, 655 * we might be able to loosen a few clusters up on the drain. 656 */ 657 if ((how & M_NOWAIT) && (m->m_ext.ext_buf == NULL)) { 658 zone_drain(zone_pack); 659 uma_zalloc_arg(zone_clust, m, how); 660 } 661 } 662 663 /* 664 * m_cljget() is different from m_clget() as it can allocate clusters without 665 * attaching them to an mbuf. In that case the return value is the pointer 666 * to the cluster of the requested size. If an mbuf was specified, it gets 667 * the cluster attached to it and the return value can be safely ignored. 668 * For size it takes MCLBYTES, MJUMPAGESIZE, MJUM9BYTES, MJUM16BYTES. 669 */ 670 static __inline void * 671 m_cljget(struct mbuf *m, int how, int size) 672 { 673 uma_zone_t zone; 674 675 if (m && m->m_flags & M_EXT) 676 printf("%s: %p mbuf already has cluster\n", __func__, m); 677 if (m != NULL) 678 m->m_ext.ext_buf = NULL; 679 680 zone = m_getzone(size); 681 return (uma_zalloc_arg(zone, m, how)); 682 } 683 684 static __inline void 685 m_cljset(struct mbuf *m, void *cl, int type) 686 { 687 uma_zone_t zone; 688 int size; 689 690 switch (type) { 691 case EXT_CLUSTER: 692 size = MCLBYTES; 693 zone = zone_clust; 694 break; 695 #if MJUMPAGESIZE != MCLBYTES 696 case EXT_JUMBOP: 697 size = MJUMPAGESIZE; 698 zone = zone_jumbop; 699 break; 700 #endif 701 case EXT_JUMBO9: 702 size = MJUM9BYTES; 703 zone = zone_jumbo9; 704 break; 705 case EXT_JUMBO16: 706 size = MJUM16BYTES; 707 zone = zone_jumbo16; 708 break; 709 default: 710 panic("%s: unknown cluster type", __func__); 711 break; 712 } 713 714 m->m_data = m->m_ext.ext_buf = cl; 715 m->m_ext.ext_free = m->m_ext.ext_arg1 = m->m_ext.ext_arg2 = NULL; 716 m->m_ext.ext_size = size; 717 m->m_ext.ext_type = type; 718 m->m_ext.ref_cnt = uma_find_refcnt(zone, cl); 719 m->m_flags |= M_EXT; 720 721 } 722 723 static __inline void 724 m_chtype(struct mbuf *m, short new_type) 725 { 726 727 m->m_type = new_type; 728 } 729 730 static __inline struct mbuf * 731 m_last(struct mbuf *m) 732 { 733 734 while (m->m_next) 735 m = m->m_next; 736 return (m); 737 } 738 739 /* 740 * mbuf, cluster, and external object allocation macros (for compatibility 741 * purposes). 742 */ 743 #define M_MOVE_PKTHDR(to, from) m_move_pkthdr((to), (from)) 744 #define MGET(m, how, type) ((m) = m_get((how), (type))) 745 #define MGETHDR(m, how, type) ((m) = m_gethdr((how), (type))) 746 #define MCLGET(m, how) m_clget((m), (how)) 747 #define MEXTADD(m, buf, size, free, arg1, arg2, flags, type) \ 748 m_extadd((m), (caddr_t)(buf), (size), (free),(arg1),(arg2),(flags), (type)) 749 #define m_getm(m, len, how, type) \ 750 m_getm2((m), (len), (how), (type), M_PKTHDR) 751 752 /* 753 * Evaluate TRUE if it's safe to write to the mbuf m's data region (this can 754 * be both the local data payload, or an external buffer area, depending on 755 * whether M_EXT is set). 756 */ 757 #define M_WRITABLE(m) (!((m)->m_flags & M_RDONLY) && \ 758 (!(((m)->m_flags & M_EXT)) || \ 759 (*((m)->m_ext.ref_cnt) == 1)) ) \ 760 761 /* Check if the supplied mbuf has a packet header, or else panic. */ 762 #define M_ASSERTPKTHDR(m) \ 763 KASSERT((m) != NULL && (m)->m_flags & M_PKTHDR, \ 764 ("%s: no mbuf packet header!", __func__)) 765 766 /* 767 * Ensure that the supplied mbuf is a valid, non-free mbuf. 768 * 769 * XXX: Broken at the moment. Need some UMA magic to make it work again. 770 */ 771 #define M_ASSERTVALID(m) \ 772 KASSERT((((struct mbuf *)m)->m_flags & 0) == 0, \ 773 ("%s: attempted use of a free mbuf!", __func__)) 774 775 /* 776 * Set the m_data pointer of a newly-allocated mbuf (m_get/MGET) to place an 777 * object of the specified size at the end of the mbuf, longword aligned. 778 */ 779 #define M_ALIGN(m, len) do { \ 780 KASSERT(!((m)->m_flags & (M_PKTHDR|M_EXT)), \ 781 ("%s: M_ALIGN not normal mbuf", __func__)); \ 782 KASSERT((m)->m_data == (m)->m_dat, \ 783 ("%s: M_ALIGN not a virgin mbuf", __func__)); \ 784 (m)->m_data += (MLEN - (len)) & ~(sizeof(long) - 1); \ 785 } while (0) 786 787 /* 788 * As above, for mbufs allocated with m_gethdr/MGETHDR or initialized by 789 * M_DUP/MOVE_PKTHDR. 790 */ 791 #define MH_ALIGN(m, len) do { \ 792 KASSERT((m)->m_flags & M_PKTHDR && !((m)->m_flags & M_EXT), \ 793 ("%s: MH_ALIGN not PKTHDR mbuf", __func__)); \ 794 KASSERT((m)->m_data == (m)->m_pktdat, \ 795 ("%s: MH_ALIGN not a virgin mbuf", __func__)); \ 796 (m)->m_data += (MHLEN - (len)) & ~(sizeof(long) - 1); \ 797 } while (0) 798 799 /* 800 * Compute the amount of space available before the current start of data in 801 * an mbuf. 802 * 803 * The M_WRITABLE() is a temporary, conservative safety measure: the burden 804 * of checking writability of the mbuf data area rests solely with the caller. 805 */ 806 #define M_LEADINGSPACE(m) \ 807 ((m)->m_flags & M_EXT ? \ 808 (M_WRITABLE(m) ? (m)->m_data - (m)->m_ext.ext_buf : 0): \ 809 (m)->m_flags & M_PKTHDR ? (m)->m_data - (m)->m_pktdat : \ 810 (m)->m_data - (m)->m_dat) 811 812 /* 813 * Compute the amount of space available after the end of data in an mbuf. 814 * 815 * The M_WRITABLE() is a temporary, conservative safety measure: the burden 816 * of checking writability of the mbuf data area rests solely with the caller. 817 */ 818 #define M_TRAILINGSPACE(m) \ 819 ((m)->m_flags & M_EXT ? \ 820 (M_WRITABLE(m) ? (m)->m_ext.ext_buf + (m)->m_ext.ext_size \ 821 - ((m)->m_data + (m)->m_len) : 0) : \ 822 &(m)->m_dat[MLEN] - ((m)->m_data + (m)->m_len)) 823 824 /* 825 * Arrange to prepend space of size plen to mbuf m. If a new mbuf must be 826 * allocated, how specifies whether to wait. If the allocation fails, the 827 * original mbuf chain is freed and m is set to NULL. 828 */ 829 #define M_PREPEND(m, plen, how) do { \ 830 struct mbuf **_mmp = &(m); \ 831 struct mbuf *_mm = *_mmp; \ 832 int _mplen = (plen); \ 833 int __mhow = (how); \ 834 \ 835 MBUF_CHECKSLEEP(how); \ 836 if (M_LEADINGSPACE(_mm) >= _mplen) { \ 837 _mm->m_data -= _mplen; \ 838 _mm->m_len += _mplen; \ 839 } else \ 840 _mm = m_prepend(_mm, _mplen, __mhow); \ 841 if (_mm != NULL && _mm->m_flags & M_PKTHDR) \ 842 _mm->m_pkthdr.len += _mplen; \ 843 *_mmp = _mm; \ 844 } while (0) 845 846 /* 847 * Change mbuf to new type. This is a relatively expensive operation and 848 * should be avoided. 849 */ 850 #define MCHTYPE(m, t) m_chtype((m), (t)) 851 852 /* Length to m_copy to copy all. */ 853 #define M_COPYALL 1000000000 854 855 /* Compatibility with 4.3. */ 856 #define m_copy(m, o, l) m_copym((m), (o), (l), M_NOWAIT) 857 858 extern int max_datalen; /* MHLEN - max_hdr */ 859 extern int max_hdr; /* Largest link + protocol header */ 860 extern int max_linkhdr; /* Largest link-level header */ 861 extern int max_protohdr; /* Largest protocol header */ 862 extern struct mbstat mbstat; /* General mbuf stats/infos */ 863 extern int nmbclusters; /* Maximum number of clusters */ 864 865 struct uio; 866 867 void m_adj(struct mbuf *, int); 868 void m_align(struct mbuf *, int); 869 int m_apply(struct mbuf *, int, int, 870 int (*)(void *, void *, u_int), void *); 871 int m_append(struct mbuf *, int, c_caddr_t); 872 void m_cat(struct mbuf *, struct mbuf *); 873 void m_extadd(struct mbuf *, caddr_t, u_int, 874 void (*)(void *, void *), void *, void *, int, int); 875 struct mbuf *m_collapse(struct mbuf *, int, int); 876 void m_copyback(struct mbuf *, int, int, c_caddr_t); 877 void m_copydata(const struct mbuf *, int, int, caddr_t); 878 struct mbuf *m_copym(struct mbuf *, int, int, int); 879 struct mbuf *m_copymdata(struct mbuf *, struct mbuf *, 880 int, int, int, int); 881 struct mbuf *m_copypacket(struct mbuf *, int); 882 void m_copy_pkthdr(struct mbuf *, struct mbuf *); 883 struct mbuf *m_copyup(struct mbuf *n, int len, int dstoff); 884 struct mbuf *m_defrag(struct mbuf *, int); 885 void m_demote(struct mbuf *, int); 886 struct mbuf *m_devget(char *, int, int, struct ifnet *, 887 void (*)(char *, caddr_t, u_int)); 888 struct mbuf *m_dup(struct mbuf *, int); 889 int m_dup_pkthdr(struct mbuf *, struct mbuf *, int); 890 u_int m_fixhdr(struct mbuf *); 891 struct mbuf *m_fragment(struct mbuf *, int, int); 892 void m_freem(struct mbuf *); 893 struct mbuf *m_getm2(struct mbuf *, int, int, short, int); 894 struct mbuf *m_getptr(struct mbuf *, int, int *); 895 u_int m_length(struct mbuf *, struct mbuf **); 896 int m_mbuftouio(struct uio *, struct mbuf *, int); 897 void m_move_pkthdr(struct mbuf *, struct mbuf *); 898 struct mbuf *m_prepend(struct mbuf *, int, int); 899 void m_print(const struct mbuf *, int); 900 struct mbuf *m_pulldown(struct mbuf *, int, int, int *); 901 struct mbuf *m_pullup(struct mbuf *, int); 902 int m_sanity(struct mbuf *, int); 903 struct mbuf *m_split(struct mbuf *, int, int); 904 struct mbuf *m_uiotombuf(struct uio *, int, int, int, int); 905 struct mbuf *m_unshare(struct mbuf *, int how); 906 907 /*- 908 * Network packets may have annotations attached by affixing a list of 909 * "packet tags" to the pkthdr structure. Packet tags are dynamically 910 * allocated semi-opaque data structures that have a fixed header 911 * (struct m_tag) that specifies the size of the memory block and a 912 * <cookie,type> pair that identifies it. The cookie is a 32-bit unique 913 * unsigned value used to identify a module or ABI. By convention this value 914 * is chosen as the date+time that the module is created, expressed as the 915 * number of seconds since the epoch (e.g., using date -u +'%s'). The type 916 * value is an ABI/module-specific value that identifies a particular 917 * annotation and is private to the module. For compatibility with systems 918 * like OpenBSD that define packet tags w/o an ABI/module cookie, the value 919 * PACKET_ABI_COMPAT is used to implement m_tag_get and m_tag_find 920 * compatibility shim functions and several tag types are defined below. 921 * Users that do not require compatibility should use a private cookie value 922 * so that packet tag-related definitions can be maintained privately. 923 * 924 * Note that the packet tag returned by m_tag_alloc has the default memory 925 * alignment implemented by malloc. To reference private data one can use a 926 * construct like: 927 * 928 * struct m_tag *mtag = m_tag_alloc(...); 929 * struct foo *p = (struct foo *)(mtag+1); 930 * 931 * if the alignment of struct m_tag is sufficient for referencing members of 932 * struct foo. Otherwise it is necessary to embed struct m_tag within the 933 * private data structure to insure proper alignment; e.g., 934 * 935 * struct foo { 936 * struct m_tag tag; 937 * ... 938 * }; 939 * struct foo *p = (struct foo *) m_tag_alloc(...); 940 * struct m_tag *mtag = &p->tag; 941 */ 942 943 /* 944 * Persistent tags stay with an mbuf until the mbuf is reclaimed. Otherwise 945 * tags are expected to ``vanish'' when they pass through a network 946 * interface. For most interfaces this happens normally as the tags are 947 * reclaimed when the mbuf is free'd. However in some special cases 948 * reclaiming must be done manually. An example is packets that pass through 949 * the loopback interface. Also, one must be careful to do this when 950 * ``turning around'' packets (e.g., icmp_reflect). 951 * 952 * To mark a tag persistent bit-or this flag in when defining the tag id. 953 * The tag will then be treated as described above. 954 */ 955 #define MTAG_PERSISTENT 0x800 956 957 #define PACKET_TAG_NONE 0 /* Nadda */ 958 959 /* Packet tags for use with PACKET_ABI_COMPAT. */ 960 #define PACKET_TAG_IPSEC_IN_DONE 1 /* IPsec applied, in */ 961 #define PACKET_TAG_IPSEC_OUT_DONE 2 /* IPsec applied, out */ 962 #define PACKET_TAG_IPSEC_IN_CRYPTO_DONE 3 /* NIC IPsec crypto done */ 963 #define PACKET_TAG_IPSEC_OUT_CRYPTO_NEEDED 4 /* NIC IPsec crypto req'ed */ 964 #define PACKET_TAG_IPSEC_IN_COULD_DO_CRYPTO 5 /* NIC notifies IPsec */ 965 #define PACKET_TAG_IPSEC_PENDING_TDB 6 /* Reminder to do IPsec */ 966 #define PACKET_TAG_BRIDGE 7 /* Bridge processing done */ 967 #define PACKET_TAG_GIF 8 /* GIF processing done */ 968 #define PACKET_TAG_GRE 9 /* GRE processing done */ 969 #define PACKET_TAG_IN_PACKET_CHECKSUM 10 /* NIC checksumming done */ 970 #define PACKET_TAG_ENCAP 11 /* Encap. processing */ 971 #define PACKET_TAG_IPSEC_SOCKET 12 /* IPSEC socket ref */ 972 #define PACKET_TAG_IPSEC_HISTORY 13 /* IPSEC history */ 973 #define PACKET_TAG_IPV6_INPUT 14 /* IPV6 input processing */ 974 #define PACKET_TAG_DUMMYNET 15 /* dummynet info */ 975 #define PACKET_TAG_DIVERT 17 /* divert info */ 976 #define PACKET_TAG_IPFORWARD 18 /* ipforward info */ 977 #define PACKET_TAG_MACLABEL (19 | MTAG_PERSISTENT) /* MAC label */ 978 #define PACKET_TAG_PF (21 | MTAG_PERSISTENT) /* PF/ALTQ information */ 979 #define PACKET_TAG_RTSOCKFAM 25 /* rtsock sa family */ 980 #define PACKET_TAG_IPOPTIONS 27 /* Saved IP options */ 981 #define PACKET_TAG_CARP 28 /* CARP info */ 982 #define PACKET_TAG_IPSEC_NAT_T_PORTS 29 /* two uint16_t */ 983 #define PACKET_TAG_ND_OUTGOING 30 /* ND outgoing */ 984 985 /* Specific cookies and tags. */ 986 987 /* Packet tag routines. */ 988 struct m_tag *m_tag_alloc(u_int32_t, int, int, int); 989 void m_tag_delete(struct mbuf *, struct m_tag *); 990 void m_tag_delete_chain(struct mbuf *, struct m_tag *); 991 void m_tag_free_default(struct m_tag *); 992 struct m_tag *m_tag_locate(struct mbuf *, u_int32_t, int, struct m_tag *); 993 struct m_tag *m_tag_copy(struct m_tag *, int); 994 int m_tag_copy_chain(struct mbuf *, struct mbuf *, int); 995 void m_tag_delete_nonpersistent(struct mbuf *); 996 997 /* 998 * Initialize the list of tags associated with an mbuf. 999 */ 1000 static __inline void 1001 m_tag_init(struct mbuf *m) 1002 { 1003 1004 SLIST_INIT(&m->m_pkthdr.tags); 1005 } 1006 1007 /* 1008 * Set up the contents of a tag. Note that this does not fill in the free 1009 * method; the caller is expected to do that. 1010 * 1011 * XXX probably should be called m_tag_init, but that was already taken. 1012 */ 1013 static __inline void 1014 m_tag_setup(struct m_tag *t, u_int32_t cookie, int type, int len) 1015 { 1016 1017 t->m_tag_id = type; 1018 t->m_tag_len = len; 1019 t->m_tag_cookie = cookie; 1020 } 1021 1022 /* 1023 * Reclaim resources associated with a tag. 1024 */ 1025 static __inline void 1026 m_tag_free(struct m_tag *t) 1027 { 1028 1029 (*t->m_tag_free)(t); 1030 } 1031 1032 /* 1033 * Return the first tag associated with an mbuf. 1034 */ 1035 static __inline struct m_tag * 1036 m_tag_first(struct mbuf *m) 1037 { 1038 1039 return (SLIST_FIRST(&m->m_pkthdr.tags)); 1040 } 1041 1042 /* 1043 * Return the next tag in the list of tags associated with an mbuf. 1044 */ 1045 static __inline struct m_tag * 1046 m_tag_next(struct mbuf *m, struct m_tag *t) 1047 { 1048 1049 return (SLIST_NEXT(t, m_tag_link)); 1050 } 1051 1052 /* 1053 * Prepend a tag to the list of tags associated with an mbuf. 1054 */ 1055 static __inline void 1056 m_tag_prepend(struct mbuf *m, struct m_tag *t) 1057 { 1058 1059 SLIST_INSERT_HEAD(&m->m_pkthdr.tags, t, m_tag_link); 1060 } 1061 1062 /* 1063 * Unlink a tag from the list of tags associated with an mbuf. 1064 */ 1065 static __inline void 1066 m_tag_unlink(struct mbuf *m, struct m_tag *t) 1067 { 1068 1069 SLIST_REMOVE(&m->m_pkthdr.tags, t, m_tag, m_tag_link); 1070 } 1071 1072 /* These are for OpenBSD compatibility. */ 1073 #define MTAG_ABI_COMPAT 0 /* compatibility ABI */ 1074 1075 static __inline struct m_tag * 1076 m_tag_get(int type, int length, int wait) 1077 { 1078 return (m_tag_alloc(MTAG_ABI_COMPAT, type, length, wait)); 1079 } 1080 1081 static __inline struct m_tag * 1082 m_tag_find(struct mbuf *m, int type, struct m_tag *start) 1083 { 1084 return (SLIST_EMPTY(&m->m_pkthdr.tags) ? (struct m_tag *)NULL : 1085 m_tag_locate(m, MTAG_ABI_COMPAT, type, start)); 1086 } 1087 1088 /* XXX temporary FIB methods probably eventually use tags.*/ 1089 #define M_FIBSHIFT 28 1090 #define M_FIBMASK 0x0F 1091 1092 /* get the fib from an mbuf and if it is not set, return the default */ 1093 #define M_GETFIB(_m) \ 1094 ((((_m)->m_flags & M_FIB) >> M_FIBSHIFT) & M_FIBMASK) 1095 1096 #define M_SETFIB(_m, _fib) do { \ 1097 _m->m_flags &= ~M_FIB; \ 1098 _m->m_flags |= (((_fib) << M_FIBSHIFT) & M_FIB); \ 1099 } while (0) 1100 1101 #endif /* _KERNEL */ 1102 1103 #ifdef MBUF_PROFILING 1104 void m_profile(struct mbuf *m); 1105 #define M_PROFILE(m) m_profile(m) 1106 #else 1107 #define M_PROFILE(m) 1108 #endif 1109 1110 1111 #endif /* !_SYS_MBUF_H_ */ 1112