1 /*- 2 * SPDX-License-Identifier: BSD-3-Clause 3 * 4 * Copyright (c) 1982, 1986, 1988, 1993 5 * The Regents of the University of California. 6 * All rights reserved. 7 * 8 * Redistribution and use in source and binary forms, with or without 9 * modification, are permitted provided that the following conditions 10 * are met: 11 * 1. Redistributions of source code must retain the above copyright 12 * notice, this list of conditions and the following disclaimer. 13 * 2. Redistributions in binary form must reproduce the above copyright 14 * notice, this list of conditions and the following disclaimer in the 15 * documentation and/or other materials provided with the distribution. 16 * 3. Neither the name of the University nor the names of its contributors 17 * may be used to endorse or promote products derived from this software 18 * without specific prior written permission. 19 * 20 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 21 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 22 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 23 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 24 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 25 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 26 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 27 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 28 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 29 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 30 * SUCH DAMAGE. 31 * 32 * @(#)mbuf.h 8.5 (Berkeley) 2/19/95 33 * $FreeBSD$ 34 */ 35 36 #ifndef _SYS_MBUF_H_ 37 #define _SYS_MBUF_H_ 38 39 /* XXX: These includes suck. Sorry! */ 40 #include <sys/queue.h> 41 #ifdef _KERNEL 42 #include <sys/systm.h> 43 #include <sys/refcount.h> 44 #include <vm/uma.h> 45 #ifdef WITNESS 46 #include <sys/lock.h> 47 #endif 48 #endif 49 50 #ifdef _KERNEL 51 #include <sys/sdt.h> 52 53 #define MBUF_PROBE1(probe, arg0) \ 54 SDT_PROBE1(sdt, , , probe, arg0) 55 #define MBUF_PROBE2(probe, arg0, arg1) \ 56 SDT_PROBE2(sdt, , , probe, arg0, arg1) 57 #define MBUF_PROBE3(probe, arg0, arg1, arg2) \ 58 SDT_PROBE3(sdt, , , probe, arg0, arg1, arg2) 59 #define MBUF_PROBE4(probe, arg0, arg1, arg2, arg3) \ 60 SDT_PROBE4(sdt, , , probe, arg0, arg1, arg2, arg3) 61 #define MBUF_PROBE5(probe, arg0, arg1, arg2, arg3, arg4) \ 62 SDT_PROBE5(sdt, , , probe, arg0, arg1, arg2, arg3, arg4) 63 64 SDT_PROBE_DECLARE(sdt, , , m__init); 65 SDT_PROBE_DECLARE(sdt, , , m__gethdr); 66 SDT_PROBE_DECLARE(sdt, , , m__get); 67 SDT_PROBE_DECLARE(sdt, , , m__getcl); 68 SDT_PROBE_DECLARE(sdt, , , m__clget); 69 SDT_PROBE_DECLARE(sdt, , , m__cljget); 70 SDT_PROBE_DECLARE(sdt, , , m__cljset); 71 SDT_PROBE_DECLARE(sdt, , , m__free); 72 SDT_PROBE_DECLARE(sdt, , , m__freem); 73 74 #endif /* _KERNEL */ 75 76 /* 77 * Mbufs are of a single size, MSIZE (sys/param.h), which includes overhead. 78 * An mbuf may add a single "mbuf cluster" of size MCLBYTES (also in 79 * sys/param.h), which has no additional overhead and is used instead of the 80 * internal data area; this is done when at least MINCLSIZE of data must be 81 * stored. Additionally, it is possible to allocate a separate buffer 82 * externally and attach it to the mbuf in a way similar to that of mbuf 83 * clusters. 84 * 85 * NB: These calculation do not take actual compiler-induced alignment and 86 * padding inside the complete struct mbuf into account. Appropriate 87 * attention is required when changing members of struct mbuf. 88 * 89 * MLEN is data length in a normal mbuf. 90 * MHLEN is data length in an mbuf with pktheader. 91 * MINCLSIZE is a smallest amount of data that should be put into cluster. 92 * 93 * Compile-time assertions in uipc_mbuf.c test these values to ensure that 94 * they are sensible. 95 */ 96 struct mbuf; 97 #define MHSIZE offsetof(struct mbuf, m_dat) 98 #define MPKTHSIZE offsetof(struct mbuf, m_pktdat) 99 #define MLEN ((int)(MSIZE - MHSIZE)) 100 #define MHLEN ((int)(MSIZE - MPKTHSIZE)) 101 #define MINCLSIZE (MHLEN + 1) 102 #define M_NODOM 255 103 104 #ifdef _KERNEL 105 /*- 106 * Macro for type conversion: convert mbuf pointer to data pointer of correct 107 * type: 108 * 109 * mtod(m, t) -- Convert mbuf pointer to data pointer of correct type. 110 * mtodo(m, o) -- Same as above but with offset 'o' into data. 111 */ 112 #define mtod(m, t) ((t)((m)->m_data)) 113 #define mtodo(m, o) ((void *)(((m)->m_data) + (o))) 114 115 /* 116 * Argument structure passed to UMA routines during mbuf and packet 117 * allocations. 118 */ 119 struct mb_args { 120 int flags; /* Flags for mbuf being allocated */ 121 short type; /* Type of mbuf being allocated */ 122 }; 123 #endif /* _KERNEL */ 124 125 /* 126 * Packet tag structure (see below for details). 127 */ 128 struct m_tag { 129 SLIST_ENTRY(m_tag) m_tag_link; /* List of packet tags */ 130 u_int16_t m_tag_id; /* Tag ID */ 131 u_int16_t m_tag_len; /* Length of data */ 132 u_int32_t m_tag_cookie; /* ABI/Module ID */ 133 void (*m_tag_free)(struct m_tag *); 134 }; 135 136 /* 137 * Static network interface owned tag. 138 * Allocated through ifp->if_snd_tag_alloc(). 139 */ 140 struct m_snd_tag { 141 struct ifnet *ifp; /* network interface tag belongs to */ 142 volatile u_int refcount; 143 }; 144 145 /* 146 * Record/packet header in first mbuf of chain; valid only if M_PKTHDR is set. 147 * Size ILP32: 48 148 * LP64: 56 149 * Compile-time assertions in uipc_mbuf.c test these values to ensure that 150 * they are correct. 151 */ 152 struct pkthdr { 153 union { 154 struct m_snd_tag *snd_tag; /* send tag, if any */ 155 struct ifnet *rcvif; /* rcv interface */ 156 }; 157 SLIST_HEAD(packet_tags, m_tag) tags; /* list of packet tags */ 158 int32_t len; /* total packet length */ 159 160 /* Layer crossing persistent information. */ 161 uint32_t flowid; /* packet's 4-tuple system */ 162 uint32_t csum_flags; /* checksum and offload features */ 163 uint16_t fibnum; /* this packet should use this fib */ 164 uint8_t numa_domain; /* NUMA domain of recvd pkt */ 165 uint8_t rsstype; /* hash type */ 166 union { 167 uint64_t rcv_tstmp; /* timestamp in ns */ 168 struct { 169 uint8_t l2hlen; /* layer 2 hdr len */ 170 uint8_t l3hlen; /* layer 3 hdr len */ 171 uint8_t l4hlen; /* layer 4 hdr len */ 172 uint8_t l5hlen; /* layer 5 hdr len */ 173 uint32_t spare; 174 }; 175 }; 176 union { 177 uint8_t eight[8]; 178 uint16_t sixteen[4]; 179 uint32_t thirtytwo[2]; 180 uint64_t sixtyfour[1]; 181 uintptr_t unintptr[1]; 182 void *ptr; 183 } PH_per; 184 185 /* Layer specific non-persistent local storage for reassembly, etc. */ 186 union { 187 uint8_t eight[8]; 188 uint16_t sixteen[4]; 189 uint32_t thirtytwo[2]; 190 uint64_t sixtyfour[1]; 191 uintptr_t unintptr[1]; 192 void *ptr; 193 } PH_loc; 194 }; 195 #define ether_vtag PH_per.sixteen[0] 196 #define PH_vt PH_per 197 #define vt_nrecs sixteen[0] /* mld and v6-ND */ 198 #define tso_segsz PH_per.sixteen[1] /* inbound after LRO */ 199 #define lro_nsegs tso_segsz /* inbound after LRO */ 200 #define csum_data PH_per.thirtytwo[1] /* inbound from hardware up */ 201 #define lro_len PH_loc.sixteen[0] /* inbound during LRO (no reassembly) */ 202 #define lro_csum PH_loc.sixteen[1] /* inbound during LRO (no reassembly) */ 203 /* Note PH_loc is used during IP reassembly (all 8 bytes as a ptr) */ 204 205 /* 206 * Description of external storage mapped into mbuf; valid only if M_EXT is 207 * set. 208 * Size ILP32: 28 209 * LP64: 48 210 * Compile-time assertions in uipc_mbuf.c test these values to ensure that 211 * they are correct. 212 */ 213 typedef void m_ext_free_t(struct mbuf *); 214 struct m_ext { 215 union { 216 /* 217 * If EXT_FLAG_EMBREF is set, then we use refcount in the 218 * mbuf, the 'ext_count' member. Otherwise, we have a 219 * shadow copy and we use pointer 'ext_cnt'. The original 220 * mbuf is responsible to carry the pointer to free routine 221 * and its arguments. They aren't copied into shadows in 222 * mb_dupcl() to avoid dereferencing next cachelines. 223 */ 224 volatile u_int ext_count; 225 volatile u_int *ext_cnt; 226 }; 227 union { 228 /* 229 * If ext_type == EXT_PGS, 'ext_pgs' points to a 230 * structure describing the buffer. Otherwise, 231 * 'ext_buf' points to the start of the buffer. 232 */ 233 struct mbuf_ext_pgs *ext_pgs; 234 char *ext_buf; 235 }; 236 uint32_t ext_size; /* size of buffer, for ext_free */ 237 uint32_t ext_type:8, /* type of external storage */ 238 ext_flags:24; /* external storage mbuf flags */ 239 /* 240 * Fields below store the free context for the external storage. 241 * They are valid only in the refcount carrying mbuf, the one with 242 * EXT_FLAG_EMBREF flag, with exclusion for EXT_EXTREF type, where 243 * the free context is copied into all mbufs that use same external 244 * storage. 245 */ 246 #define m_ext_copylen offsetof(struct m_ext, ext_free) 247 m_ext_free_t *ext_free; /* free routine if not the usual */ 248 void *ext_arg1; /* optional argument pointer */ 249 void *ext_arg2; /* optional argument pointer */ 250 }; 251 252 /* 253 * The core of the mbuf object along with some shortcut defines for practical 254 * purposes. 255 */ 256 struct mbuf { 257 /* 258 * Header present at the beginning of every mbuf. 259 * Size ILP32: 24 260 * LP64: 32 261 * Compile-time assertions in uipc_mbuf.c test these values to ensure 262 * that they are correct. 263 */ 264 union { /* next buffer in chain */ 265 struct mbuf *m_next; 266 SLIST_ENTRY(mbuf) m_slist; 267 STAILQ_ENTRY(mbuf) m_stailq; 268 }; 269 union { /* next chain in queue/record */ 270 struct mbuf *m_nextpkt; 271 SLIST_ENTRY(mbuf) m_slistpkt; 272 STAILQ_ENTRY(mbuf) m_stailqpkt; 273 }; 274 caddr_t m_data; /* location of data */ 275 int32_t m_len; /* amount of data in this mbuf */ 276 uint32_t m_type:8, /* type of data in this mbuf */ 277 m_flags:24; /* flags; see below */ 278 #if !defined(__LP64__) 279 uint32_t m_pad; /* pad for 64bit alignment */ 280 #endif 281 282 /* 283 * A set of optional headers (packet header, external storage header) 284 * and internal data storage. Historically, these arrays were sized 285 * to MHLEN (space left after a packet header) and MLEN (space left 286 * after only a regular mbuf header); they are now variable size in 287 * order to support future work on variable-size mbufs. 288 */ 289 union { 290 struct { 291 struct pkthdr m_pkthdr; /* M_PKTHDR set */ 292 union { 293 struct m_ext m_ext; /* M_EXT set */ 294 char m_pktdat[0]; 295 }; 296 }; 297 char m_dat[0]; /* !M_PKTHDR, !M_EXT */ 298 }; 299 }; 300 301 struct ktls_session; 302 struct socket; 303 304 /* 305 * TLS records for TLS 1.0-1.2 can have the following header lengths: 306 * - 5 (AES-CBC with implicit IV) 307 * - 21 (AES-CBC with explicit IV) 308 * - 13 (AES-GCM with 8 byte explicit IV) 309 */ 310 #define MBUF_PEXT_HDR_LEN 23 311 312 /* 313 * TLS records for TLS 1.0-1.2 can have the following maximum trailer 314 * lengths: 315 * - 16 (AES-GCM) 316 * - 36 (AES-CBC with SHA1 and up to 16 bytes of padding) 317 * - 48 (AES-CBC with SHA2-256 and up to 16 bytes of padding) 318 * - 64 (AES-CBC with SHA2-384 and up to 16 bytes of padding) 319 */ 320 #define MBUF_PEXT_TRAIL_LEN 64 321 322 #ifdef __LP64__ 323 #define MBUF_PEXT_MAX_PGS (152 / sizeof(vm_paddr_t)) 324 #else 325 #define MBUF_PEXT_MAX_PGS (156 / sizeof(vm_paddr_t)) 326 #endif 327 328 #define MBUF_PEXT_MAX_BYTES \ 329 (MBUF_PEXT_MAX_PGS * PAGE_SIZE + MBUF_PEXT_HDR_LEN + MBUF_PEXT_TRAIL_LEN) 330 331 #define MBUF_PEXT_FLAG_ANON 1 /* Data can be encrypted in place. */ 332 333 /* 334 * This struct is 256 bytes in size and is arranged so that the most 335 * common case (accessing the first 4 pages of a 16KB TLS record) will 336 * fit in a single 64 byte cacheline. 337 */ 338 struct mbuf_ext_pgs { 339 uint8_t npgs; /* Number of attached pages */ 340 uint8_t nrdy; /* Pages with I/O pending */ 341 uint8_t hdr_len; /* TLS header length */ 342 uint8_t trail_len; /* TLS trailer length */ 343 uint16_t first_pg_off; /* Offset into 1st page */ 344 uint16_t last_pg_len; /* Length of last page */ 345 vm_paddr_t pa[MBUF_PEXT_MAX_PGS]; /* phys addrs of pages */ 346 char hdr[MBUF_PEXT_HDR_LEN]; /* TLS header */ 347 uint8_t flags; /* Flags */ 348 struct ktls_session *tls; /* TLS session */ 349 #if defined(__i386__) || \ 350 (defined(__powerpc__) && !defined(__powerpc64__) && defined(BOOKE)) 351 /* 352 * i386 and Book-E PowerPC have 64-bit vm_paddr_t, so there is 353 * a 4 byte remainder from the space allocated for pa[]. 354 */ 355 uint32_t pad; 356 #endif 357 union { 358 char trail[MBUF_PEXT_TRAIL_LEN]; /* TLS trailer */ 359 struct { 360 uint8_t record_type; /* Must be first */ 361 struct socket *so; 362 struct mbuf *mbuf; 363 uint64_t seqno; 364 STAILQ_ENTRY(mbuf_ext_pgs) stailq; 365 int enc_cnt; 366 }; 367 }; 368 }; 369 370 #ifdef _KERNEL 371 static inline int 372 mbuf_ext_pg_len(struct mbuf_ext_pgs *ext_pgs, int pidx, int pgoff) 373 { 374 KASSERT(pgoff == 0 || pidx == 0, 375 ("page %d with non-zero offset %d in %p", pidx, pgoff, ext_pgs)); 376 if (pidx == ext_pgs->npgs - 1) { 377 return (ext_pgs->last_pg_len); 378 } else { 379 return (PAGE_SIZE - pgoff); 380 } 381 } 382 383 #ifdef INVARIANT_SUPPORT 384 void mb_ext_pgs_check(struct mbuf_ext_pgs *ext_pgs); 385 #endif 386 #ifdef INVARIANTS 387 #define MBUF_EXT_PGS_ASSERT_SANITY(ext_pgs) mb_ext_pgs_check((ext_pgs)) 388 #else 389 #define MBUF_EXT_PGS_ASSERT_SANITY(ext_pgs) 390 #endif 391 #endif 392 393 /* 394 * mbuf flags of global significance and layer crossing. 395 * Those of only protocol/layer specific significance are to be mapped 396 * to M_PROTO[1-11] and cleared at layer handoff boundaries. 397 * NB: Limited to the lower 24 bits. 398 */ 399 #define M_EXT 0x00000001 /* has associated external storage */ 400 #define M_PKTHDR 0x00000002 /* start of record */ 401 #define M_EOR 0x00000004 /* end of record */ 402 #define M_RDONLY 0x00000008 /* associated data is marked read-only */ 403 #define M_BCAST 0x00000010 /* send/received as link-level broadcast */ 404 #define M_MCAST 0x00000020 /* send/received as link-level multicast */ 405 #define M_PROMISC 0x00000040 /* packet was not for us */ 406 #define M_VLANTAG 0x00000080 /* ether_vtag is valid */ 407 #define M_NOMAP 0x00000100 /* mbuf data is unmapped */ 408 #define M_NOFREE 0x00000200 /* do not free mbuf, embedded in cluster */ 409 #define M_TSTMP 0x00000400 /* rcv_tstmp field is valid */ 410 #define M_TSTMP_HPREC 0x00000800 /* rcv_tstmp is high-prec, typically 411 hw-stamped on port (useful for IEEE 1588 412 and 802.1AS) */ 413 #define M_TSTMP_LRO 0x00001000 /* Time LRO pushed in pkt is valid in (PH_loc) */ 414 415 #define M_PROTO1 0x00002000 /* protocol-specific */ 416 #define M_PROTO2 0x00004000 /* protocol-specific */ 417 #define M_PROTO3 0x00008000 /* protocol-specific */ 418 #define M_PROTO4 0x00010000 /* protocol-specific */ 419 #define M_PROTO5 0x00020000 /* protocol-specific */ 420 #define M_PROTO6 0x00040000 /* protocol-specific */ 421 #define M_PROTO7 0x00080000 /* protocol-specific */ 422 #define M_PROTO8 0x00100000 /* protocol-specific */ 423 #define M_PROTO9 0x00200000 /* protocol-specific */ 424 #define M_PROTO10 0x00400000 /* protocol-specific */ 425 #define M_PROTO11 0x00800000 /* protocol-specific */ 426 427 #define MB_DTOR_SKIP 0x1 /* don't pollute the cache by touching a freed mbuf */ 428 429 /* 430 * Flags to purge when crossing layers. 431 */ 432 #define M_PROTOFLAGS \ 433 (M_PROTO1|M_PROTO2|M_PROTO3|M_PROTO4|M_PROTO5|M_PROTO6|M_PROTO7|M_PROTO8|\ 434 M_PROTO9|M_PROTO10|M_PROTO11) 435 436 /* 437 * Flags preserved when copying m_pkthdr. 438 */ 439 #define M_COPYFLAGS \ 440 (M_PKTHDR|M_EOR|M_RDONLY|M_BCAST|M_MCAST|M_PROMISC|M_VLANTAG|M_TSTMP| \ 441 M_TSTMP_HPREC|M_TSTMP_LRO|M_PROTOFLAGS) 442 443 /* 444 * Mbuf flag description for use with printf(9) %b identifier. 445 */ 446 #define M_FLAG_BITS \ 447 "\20\1M_EXT\2M_PKTHDR\3M_EOR\4M_RDONLY\5M_BCAST\6M_MCAST" \ 448 "\7M_PROMISC\10M_VLANTAG\11M_NOMAP\12M_NOFREE\13M_TSTMP\14M_TSTMP_HPREC\15M_TSTMP_LRO" 449 #define M_FLAG_PROTOBITS \ 450 "\16M_PROTO1\17M_PROTO2\20M_PROTO3\21M_PROTO4" \ 451 "\22M_PROTO5\23M_PROTO6\24M_PROTO7\25M_PROTO8\26M_PROTO9" \ 452 "\27M_PROTO10\28M_PROTO11" 453 #define M_FLAG_PRINTF (M_FLAG_BITS M_FLAG_PROTOBITS) 454 455 /* 456 * Network interface cards are able to hash protocol fields (such as IPv4 457 * addresses and TCP port numbers) classify packets into flows. These flows 458 * can then be used to maintain ordering while delivering packets to the OS 459 * via parallel input queues, as well as to provide a stateless affinity 460 * model. NIC drivers can pass up the hash via m->m_pkthdr.flowid, and set 461 * m_flag fields to indicate how the hash should be interpreted by the 462 * network stack. 463 * 464 * Most NICs support RSS, which provides ordering and explicit affinity, and 465 * use the hash m_flag bits to indicate what header fields were covered by 466 * the hash. M_HASHTYPE_OPAQUE and M_HASHTYPE_OPAQUE_HASH can be set by non- 467 * RSS cards or configurations that provide an opaque flow identifier, allowing 468 * for ordering and distribution without explicit affinity. Additionally, 469 * M_HASHTYPE_OPAQUE_HASH indicates that the flow identifier has hash 470 * properties. 471 * 472 * The meaning of the IPV6_EX suffix: 473 * "o Home address from the home address option in the IPv6 destination 474 * options header. If the extension header is not present, use the Source 475 * IPv6 Address. 476 * o IPv6 address that is contained in the Routing-Header-Type-2 from the 477 * associated extension header. If the extension header is not present, 478 * use the Destination IPv6 Address." 479 * Quoted from: 480 * https://docs.microsoft.com/en-us/windows-hardware/drivers/network/rss-hashing-types#ndishashipv6ex 481 */ 482 #define M_HASHTYPE_HASHPROP 0x80 /* has hash properties */ 483 #define M_HASHTYPE_HASH(t) (M_HASHTYPE_HASHPROP | (t)) 484 /* Microsoft RSS standard hash types */ 485 #define M_HASHTYPE_NONE 0 486 #define M_HASHTYPE_RSS_IPV4 M_HASHTYPE_HASH(1) /* IPv4 2-tuple */ 487 #define M_HASHTYPE_RSS_TCP_IPV4 M_HASHTYPE_HASH(2) /* TCPv4 4-tuple */ 488 #define M_HASHTYPE_RSS_IPV6 M_HASHTYPE_HASH(3) /* IPv6 2-tuple */ 489 #define M_HASHTYPE_RSS_TCP_IPV6 M_HASHTYPE_HASH(4) /* TCPv6 4-tuple */ 490 #define M_HASHTYPE_RSS_IPV6_EX M_HASHTYPE_HASH(5) /* IPv6 2-tuple + 491 * ext hdrs */ 492 #define M_HASHTYPE_RSS_TCP_IPV6_EX M_HASHTYPE_HASH(6) /* TCPv6 4-tuple + 493 * ext hdrs */ 494 #define M_HASHTYPE_RSS_UDP_IPV4 M_HASHTYPE_HASH(7) /* IPv4 UDP 4-tuple*/ 495 #define M_HASHTYPE_RSS_UDP_IPV6 M_HASHTYPE_HASH(9) /* IPv6 UDP 4-tuple*/ 496 #define M_HASHTYPE_RSS_UDP_IPV6_EX M_HASHTYPE_HASH(10)/* IPv6 UDP 4-tuple + 497 * ext hdrs */ 498 499 #define M_HASHTYPE_OPAQUE 63 /* ordering, not affinity */ 500 #define M_HASHTYPE_OPAQUE_HASH M_HASHTYPE_HASH(M_HASHTYPE_OPAQUE) 501 /* ordering+hash, not affinity*/ 502 503 #define M_HASHTYPE_CLEAR(m) ((m)->m_pkthdr.rsstype = 0) 504 #define M_HASHTYPE_GET(m) ((m)->m_pkthdr.rsstype) 505 #define M_HASHTYPE_SET(m, v) ((m)->m_pkthdr.rsstype = (v)) 506 #define M_HASHTYPE_TEST(m, v) (M_HASHTYPE_GET(m) == (v)) 507 #define M_HASHTYPE_ISHASH(m) (M_HASHTYPE_GET(m) & M_HASHTYPE_HASHPROP) 508 509 /* 510 * External mbuf storage buffer types. 511 */ 512 #define EXT_CLUSTER 1 /* mbuf cluster */ 513 #define EXT_SFBUF 2 /* sendfile(2)'s sf_buf */ 514 #define EXT_JUMBOP 3 /* jumbo cluster page sized */ 515 #define EXT_JUMBO9 4 /* jumbo cluster 9216 bytes */ 516 #define EXT_JUMBO16 5 /* jumbo cluster 16184 bytes */ 517 #define EXT_PACKET 6 /* mbuf+cluster from packet zone */ 518 #define EXT_MBUF 7 /* external mbuf reference */ 519 #define EXT_RXRING 8 /* data in NIC receive ring */ 520 #define EXT_PGS 9 /* array of unmapped pages */ 521 522 #define EXT_VENDOR1 224 /* for vendor-internal use */ 523 #define EXT_VENDOR2 225 /* for vendor-internal use */ 524 #define EXT_VENDOR3 226 /* for vendor-internal use */ 525 #define EXT_VENDOR4 227 /* for vendor-internal use */ 526 527 #define EXT_EXP1 244 /* for experimental use */ 528 #define EXT_EXP2 245 /* for experimental use */ 529 #define EXT_EXP3 246 /* for experimental use */ 530 #define EXT_EXP4 247 /* for experimental use */ 531 532 #define EXT_NET_DRV 252 /* custom ext_buf provided by net driver(s) */ 533 #define EXT_MOD_TYPE 253 /* custom module's ext_buf type */ 534 #define EXT_DISPOSABLE 254 /* can throw this buffer away w/page flipping */ 535 #define EXT_EXTREF 255 /* has externally maintained ext_cnt ptr */ 536 537 /* 538 * Flags for external mbuf buffer types. 539 * NB: limited to the lower 24 bits. 540 */ 541 #define EXT_FLAG_EMBREF 0x000001 /* embedded ext_count */ 542 #define EXT_FLAG_EXTREF 0x000002 /* external ext_cnt, notyet */ 543 544 #define EXT_FLAG_NOFREE 0x000010 /* don't free mbuf to pool, notyet */ 545 546 #define EXT_FLAG_VENDOR1 0x010000 /* These flags are vendor */ 547 #define EXT_FLAG_VENDOR2 0x020000 /* or submodule specific, */ 548 #define EXT_FLAG_VENDOR3 0x040000 /* not used by mbuf code. */ 549 #define EXT_FLAG_VENDOR4 0x080000 /* Set/read by submodule. */ 550 551 #define EXT_FLAG_EXP1 0x100000 /* for experimental use */ 552 #define EXT_FLAG_EXP2 0x200000 /* for experimental use */ 553 #define EXT_FLAG_EXP3 0x400000 /* for experimental use */ 554 #define EXT_FLAG_EXP4 0x800000 /* for experimental use */ 555 556 /* 557 * EXT flag description for use with printf(9) %b identifier. 558 */ 559 #define EXT_FLAG_BITS \ 560 "\20\1EXT_FLAG_EMBREF\2EXT_FLAG_EXTREF\5EXT_FLAG_NOFREE" \ 561 "\21EXT_FLAG_VENDOR1\22EXT_FLAG_VENDOR2\23EXT_FLAG_VENDOR3" \ 562 "\24EXT_FLAG_VENDOR4\25EXT_FLAG_EXP1\26EXT_FLAG_EXP2\27EXT_FLAG_EXP3" \ 563 "\30EXT_FLAG_EXP4" 564 565 #define MBUF_EXT_PGS_ASSERT(m) \ 566 KASSERT((((m)->m_flags & M_EXT) != 0) && \ 567 ((m)->m_ext.ext_type == EXT_PGS), \ 568 ("%s: m %p !M_EXT or !EXT_PGS", __func__, m)) 569 570 /* 571 * Flags indicating checksum, segmentation and other offload work to be 572 * done, or already done, by hardware or lower layers. It is split into 573 * separate inbound and outbound flags. 574 * 575 * Outbound flags that are set by upper protocol layers requesting lower 576 * layers, or ideally the hardware, to perform these offloading tasks. 577 * For outbound packets this field and its flags can be directly tested 578 * against ifnet if_hwassist. 579 */ 580 #define CSUM_IP 0x00000001 /* IP header checksum offload */ 581 #define CSUM_IP_UDP 0x00000002 /* UDP checksum offload */ 582 #define CSUM_IP_TCP 0x00000004 /* TCP checksum offload */ 583 #define CSUM_IP_SCTP 0x00000008 /* SCTP checksum offload */ 584 #define CSUM_IP_TSO 0x00000010 /* TCP segmentation offload */ 585 #define CSUM_IP_ISCSI 0x00000020 /* iSCSI checksum offload */ 586 587 #define CSUM_IP6_UDP 0x00000200 /* UDP checksum offload */ 588 #define CSUM_IP6_TCP 0x00000400 /* TCP checksum offload */ 589 #define CSUM_IP6_SCTP 0x00000800 /* SCTP checksum offload */ 590 #define CSUM_IP6_TSO 0x00001000 /* TCP segmentation offload */ 591 #define CSUM_IP6_ISCSI 0x00002000 /* iSCSI checksum offload */ 592 593 /* Inbound checksum support where the checksum was verified by hardware. */ 594 #define CSUM_L3_CALC 0x01000000 /* calculated layer 3 csum */ 595 #define CSUM_L3_VALID 0x02000000 /* checksum is correct */ 596 #define CSUM_L4_CALC 0x04000000 /* calculated layer 4 csum */ 597 #define CSUM_L4_VALID 0x08000000 /* checksum is correct */ 598 #define CSUM_L5_CALC 0x10000000 /* calculated layer 5 csum */ 599 #define CSUM_L5_VALID 0x20000000 /* checksum is correct */ 600 #define CSUM_COALESCED 0x40000000 /* contains merged segments */ 601 602 #define CSUM_SND_TAG 0x80000000 /* Packet header has send tag */ 603 604 /* 605 * CSUM flag description for use with printf(9) %b identifier. 606 */ 607 #define CSUM_BITS \ 608 "\20\1CSUM_IP\2CSUM_IP_UDP\3CSUM_IP_TCP\4CSUM_IP_SCTP\5CSUM_IP_TSO" \ 609 "\6CSUM_IP_ISCSI" \ 610 "\12CSUM_IP6_UDP\13CSUM_IP6_TCP\14CSUM_IP6_SCTP\15CSUM_IP6_TSO" \ 611 "\16CSUM_IP6_ISCSI" \ 612 "\31CSUM_L3_CALC\32CSUM_L3_VALID\33CSUM_L4_CALC\34CSUM_L4_VALID" \ 613 "\35CSUM_L5_CALC\36CSUM_L5_VALID\37CSUM_COALESCED\40CSUM_SND_TAG" 614 615 /* CSUM flags compatibility mappings. */ 616 #define CSUM_IP_CHECKED CSUM_L3_CALC 617 #define CSUM_IP_VALID CSUM_L3_VALID 618 #define CSUM_DATA_VALID CSUM_L4_VALID 619 #define CSUM_PSEUDO_HDR CSUM_L4_CALC 620 #define CSUM_SCTP_VALID CSUM_L4_VALID 621 #define CSUM_DELAY_DATA (CSUM_TCP|CSUM_UDP) 622 #define CSUM_DELAY_IP CSUM_IP /* Only v4, no v6 IP hdr csum */ 623 #define CSUM_DELAY_DATA_IPV6 (CSUM_TCP_IPV6|CSUM_UDP_IPV6) 624 #define CSUM_DATA_VALID_IPV6 CSUM_DATA_VALID 625 #define CSUM_TCP CSUM_IP_TCP 626 #define CSUM_UDP CSUM_IP_UDP 627 #define CSUM_SCTP CSUM_IP_SCTP 628 #define CSUM_TSO (CSUM_IP_TSO|CSUM_IP6_TSO) 629 #define CSUM_UDP_IPV6 CSUM_IP6_UDP 630 #define CSUM_TCP_IPV6 CSUM_IP6_TCP 631 #define CSUM_SCTP_IPV6 CSUM_IP6_SCTP 632 633 /* 634 * mbuf types describing the content of the mbuf (including external storage). 635 */ 636 #define MT_NOTMBUF 0 /* USED INTERNALLY ONLY! Object is not mbuf */ 637 #define MT_DATA 1 /* dynamic (data) allocation */ 638 #define MT_HEADER MT_DATA /* packet header, use M_PKTHDR instead */ 639 640 #define MT_VENDOR1 4 /* for vendor-internal use */ 641 #define MT_VENDOR2 5 /* for vendor-internal use */ 642 #define MT_VENDOR3 6 /* for vendor-internal use */ 643 #define MT_VENDOR4 7 /* for vendor-internal use */ 644 645 #define MT_SONAME 8 /* socket name */ 646 647 #define MT_EXP1 9 /* for experimental use */ 648 #define MT_EXP2 10 /* for experimental use */ 649 #define MT_EXP3 11 /* for experimental use */ 650 #define MT_EXP4 12 /* for experimental use */ 651 652 #define MT_CONTROL 14 /* extra-data protocol message */ 653 #define MT_EXTCONTROL 15 /* control message with externalized contents */ 654 #define MT_OOBDATA 16 /* expedited data */ 655 656 #define MT_NOINIT 255 /* Not a type but a flag to allocate 657 a non-initialized mbuf */ 658 659 /* 660 * String names of mbuf-related UMA(9) and malloc(9) types. Exposed to 661 * !_KERNEL so that monitoring tools can look up the zones with 662 * libmemstat(3). 663 */ 664 #define MBUF_MEM_NAME "mbuf" 665 #define MBUF_CLUSTER_MEM_NAME "mbuf_cluster" 666 #define MBUF_PACKET_MEM_NAME "mbuf_packet" 667 #define MBUF_JUMBOP_MEM_NAME "mbuf_jumbo_page" 668 #define MBUF_JUMBO9_MEM_NAME "mbuf_jumbo_9k" 669 #define MBUF_JUMBO16_MEM_NAME "mbuf_jumbo_16k" 670 #define MBUF_TAG_MEM_NAME "mbuf_tag" 671 #define MBUF_EXTREFCNT_MEM_NAME "mbuf_ext_refcnt" 672 #define MBUF_EXTPGS_MEM_NAME "mbuf_extpgs" 673 674 #ifdef _KERNEL 675 676 #ifdef WITNESS 677 #define MBUF_CHECKSLEEP(how) do { \ 678 if (how == M_WAITOK) \ 679 WITNESS_WARN(WARN_GIANTOK | WARN_SLEEPOK, NULL, \ 680 "Sleeping in \"%s\"", __func__); \ 681 } while (0) 682 #else 683 #define MBUF_CHECKSLEEP(how) 684 #endif 685 686 /* 687 * Network buffer allocation API 688 * 689 * The rest of it is defined in kern/kern_mbuf.c 690 */ 691 extern uma_zone_t zone_mbuf; 692 extern uma_zone_t zone_clust; 693 extern uma_zone_t zone_pack; 694 extern uma_zone_t zone_jumbop; 695 extern uma_zone_t zone_jumbo9; 696 extern uma_zone_t zone_jumbo16; 697 extern uma_zone_t zone_extpgs; 698 699 void mb_dupcl(struct mbuf *, struct mbuf *); 700 void mb_free_ext(struct mbuf *); 701 void mb_free_mext_pgs(struct mbuf *); 702 struct mbuf *mb_alloc_ext_pgs(int, bool, m_ext_free_t); 703 int mb_unmapped_compress(struct mbuf *m); 704 struct mbuf *mb_unmapped_to_ext(struct mbuf *m); 705 void mb_free_notready(struct mbuf *m, int count); 706 void m_adj(struct mbuf *, int); 707 int m_apply(struct mbuf *, int, int, 708 int (*)(void *, void *, u_int), void *); 709 int m_append(struct mbuf *, int, c_caddr_t); 710 void m_cat(struct mbuf *, struct mbuf *); 711 void m_catpkt(struct mbuf *, struct mbuf *); 712 int m_clget(struct mbuf *m, int how); 713 void *m_cljget(struct mbuf *m, int how, int size); 714 struct mbuf *m_collapse(struct mbuf *, int, int); 715 void m_copyback(struct mbuf *, int, int, c_caddr_t); 716 void m_copydata(const struct mbuf *, int, int, caddr_t); 717 struct mbuf *m_copym(struct mbuf *, int, int, int); 718 struct mbuf *m_copypacket(struct mbuf *, int); 719 void m_copy_pkthdr(struct mbuf *, struct mbuf *); 720 struct mbuf *m_copyup(struct mbuf *, int, int); 721 struct mbuf *m_defrag(struct mbuf *, int); 722 void m_demote_pkthdr(struct mbuf *); 723 void m_demote(struct mbuf *, int, int); 724 struct mbuf *m_devget(char *, int, int, struct ifnet *, 725 void (*)(char *, caddr_t, u_int)); 726 void m_dispose_extcontrolm(struct mbuf *m); 727 struct mbuf *m_dup(const struct mbuf *, int); 728 int m_dup_pkthdr(struct mbuf *, const struct mbuf *, int); 729 void m_extadd(struct mbuf *, char *, u_int, m_ext_free_t, 730 void *, void *, int, int); 731 u_int m_fixhdr(struct mbuf *); 732 struct mbuf *m_fragment(struct mbuf *, int, int); 733 void m_freem(struct mbuf *); 734 struct mbuf *m_get2(int, int, short, int); 735 struct mbuf *m_getjcl(int, short, int, int); 736 struct mbuf *m_getm2(struct mbuf *, int, int, short, int); 737 struct mbuf *m_getptr(struct mbuf *, int, int *); 738 u_int m_length(struct mbuf *, struct mbuf **); 739 int m_mbuftouio(struct uio *, const struct mbuf *, int); 740 int m_unmappedtouio(const struct mbuf *, int, struct uio *, int); 741 void m_move_pkthdr(struct mbuf *, struct mbuf *); 742 int m_pkthdr_init(struct mbuf *, int); 743 struct mbuf *m_prepend(struct mbuf *, int, int); 744 void m_print(const struct mbuf *, int); 745 struct mbuf *m_pulldown(struct mbuf *, int, int, int *); 746 struct mbuf *m_pullup(struct mbuf *, int); 747 int m_sanity(struct mbuf *, int); 748 struct mbuf *m_split(struct mbuf *, int, int); 749 struct mbuf *m_uiotombuf(struct uio *, int, int, int, int); 750 struct mbuf *m_unshare(struct mbuf *, int); 751 void m_snd_tag_init(struct m_snd_tag *, struct ifnet *); 752 void m_snd_tag_destroy(struct m_snd_tag *); 753 754 static __inline int 755 m_gettype(int size) 756 { 757 int type; 758 759 switch (size) { 760 case MSIZE: 761 type = EXT_MBUF; 762 break; 763 case MCLBYTES: 764 type = EXT_CLUSTER; 765 break; 766 #if MJUMPAGESIZE != MCLBYTES 767 case MJUMPAGESIZE: 768 type = EXT_JUMBOP; 769 break; 770 #endif 771 case MJUM9BYTES: 772 type = EXT_JUMBO9; 773 break; 774 case MJUM16BYTES: 775 type = EXT_JUMBO16; 776 break; 777 default: 778 panic("%s: invalid cluster size %d", __func__, size); 779 } 780 781 return (type); 782 } 783 784 /* 785 * Associated an external reference counted buffer with an mbuf. 786 */ 787 static __inline void 788 m_extaddref(struct mbuf *m, char *buf, u_int size, u_int *ref_cnt, 789 m_ext_free_t freef, void *arg1, void *arg2) 790 { 791 792 KASSERT(ref_cnt != NULL, ("%s: ref_cnt not provided", __func__)); 793 794 atomic_add_int(ref_cnt, 1); 795 m->m_flags |= M_EXT; 796 m->m_ext.ext_buf = buf; 797 m->m_ext.ext_cnt = ref_cnt; 798 m->m_data = m->m_ext.ext_buf; 799 m->m_ext.ext_size = size; 800 m->m_ext.ext_free = freef; 801 m->m_ext.ext_arg1 = arg1; 802 m->m_ext.ext_arg2 = arg2; 803 m->m_ext.ext_type = EXT_EXTREF; 804 m->m_ext.ext_flags = 0; 805 } 806 807 static __inline uma_zone_t 808 m_getzone(int size) 809 { 810 uma_zone_t zone; 811 812 switch (size) { 813 case MCLBYTES: 814 zone = zone_clust; 815 break; 816 #if MJUMPAGESIZE != MCLBYTES 817 case MJUMPAGESIZE: 818 zone = zone_jumbop; 819 break; 820 #endif 821 case MJUM9BYTES: 822 zone = zone_jumbo9; 823 break; 824 case MJUM16BYTES: 825 zone = zone_jumbo16; 826 break; 827 default: 828 panic("%s: invalid cluster size %d", __func__, size); 829 } 830 831 return (zone); 832 } 833 834 /* 835 * Initialize an mbuf with linear storage. 836 * 837 * Inline because the consumer text overhead will be roughly the same to 838 * initialize or call a function with this many parameters and M_PKTHDR 839 * should go away with constant propagation for !MGETHDR. 840 */ 841 static __inline int 842 m_init(struct mbuf *m, int how, short type, int flags) 843 { 844 int error; 845 846 m->m_next = NULL; 847 m->m_nextpkt = NULL; 848 m->m_data = m->m_dat; 849 m->m_len = 0; 850 m->m_flags = flags; 851 m->m_type = type; 852 if (flags & M_PKTHDR) 853 error = m_pkthdr_init(m, how); 854 else 855 error = 0; 856 857 MBUF_PROBE5(m__init, m, how, type, flags, error); 858 return (error); 859 } 860 861 static __inline struct mbuf * 862 m_get(int how, short type) 863 { 864 struct mbuf *m; 865 struct mb_args args; 866 867 args.flags = 0; 868 args.type = type; 869 m = uma_zalloc_arg(zone_mbuf, &args, how); 870 MBUF_PROBE3(m__get, how, type, m); 871 return (m); 872 } 873 874 static __inline struct mbuf * 875 m_gethdr(int how, short type) 876 { 877 struct mbuf *m; 878 struct mb_args args; 879 880 args.flags = M_PKTHDR; 881 args.type = type; 882 m = uma_zalloc_arg(zone_mbuf, &args, how); 883 MBUF_PROBE3(m__gethdr, how, type, m); 884 return (m); 885 } 886 887 static __inline struct mbuf * 888 m_getcl(int how, short type, int flags) 889 { 890 struct mbuf *m; 891 struct mb_args args; 892 893 args.flags = flags; 894 args.type = type; 895 m = uma_zalloc_arg(zone_pack, &args, how); 896 MBUF_PROBE4(m__getcl, how, type, flags, m); 897 return (m); 898 } 899 900 /* 901 * XXX: m_cljset() is a dangerous API. One must attach only a new, 902 * unreferenced cluster to an mbuf(9). It is not possible to assert 903 * that, so care can be taken only by users of the API. 904 */ 905 static __inline void 906 m_cljset(struct mbuf *m, void *cl, int type) 907 { 908 int size; 909 910 switch (type) { 911 case EXT_CLUSTER: 912 size = MCLBYTES; 913 break; 914 #if MJUMPAGESIZE != MCLBYTES 915 case EXT_JUMBOP: 916 size = MJUMPAGESIZE; 917 break; 918 #endif 919 case EXT_JUMBO9: 920 size = MJUM9BYTES; 921 break; 922 case EXT_JUMBO16: 923 size = MJUM16BYTES; 924 break; 925 default: 926 panic("%s: unknown cluster type %d", __func__, type); 927 break; 928 } 929 930 m->m_data = m->m_ext.ext_buf = cl; 931 m->m_ext.ext_free = m->m_ext.ext_arg1 = m->m_ext.ext_arg2 = NULL; 932 m->m_ext.ext_size = size; 933 m->m_ext.ext_type = type; 934 m->m_ext.ext_flags = EXT_FLAG_EMBREF; 935 m->m_ext.ext_count = 1; 936 m->m_flags |= M_EXT; 937 MBUF_PROBE3(m__cljset, m, cl, type); 938 } 939 940 static __inline void 941 m_chtype(struct mbuf *m, short new_type) 942 { 943 944 m->m_type = new_type; 945 } 946 947 static __inline void 948 m_clrprotoflags(struct mbuf *m) 949 { 950 951 while (m) { 952 m->m_flags &= ~M_PROTOFLAGS; 953 m = m->m_next; 954 } 955 } 956 957 static __inline struct mbuf * 958 m_last(struct mbuf *m) 959 { 960 961 while (m->m_next) 962 m = m->m_next; 963 return (m); 964 } 965 966 static inline u_int 967 m_extrefcnt(struct mbuf *m) 968 { 969 970 KASSERT(m->m_flags & M_EXT, ("%s: M_EXT missing", __func__)); 971 972 return ((m->m_ext.ext_flags & EXT_FLAG_EMBREF) ? m->m_ext.ext_count : 973 *m->m_ext.ext_cnt); 974 } 975 976 /* 977 * mbuf, cluster, and external object allocation macros (for compatibility 978 * purposes). 979 */ 980 #define M_MOVE_PKTHDR(to, from) m_move_pkthdr((to), (from)) 981 #define MGET(m, how, type) ((m) = m_get((how), (type))) 982 #define MGETHDR(m, how, type) ((m) = m_gethdr((how), (type))) 983 #define MCLGET(m, how) m_clget((m), (how)) 984 #define MEXTADD(m, buf, size, free, arg1, arg2, flags, type) \ 985 m_extadd((m), (char *)(buf), (size), (free), (arg1), (arg2), \ 986 (flags), (type)) 987 #define m_getm(m, len, how, type) \ 988 m_getm2((m), (len), (how), (type), M_PKTHDR) 989 990 /* 991 * Evaluate TRUE if it's safe to write to the mbuf m's data region (this can 992 * be both the local data payload, or an external buffer area, depending on 993 * whether M_EXT is set). 994 */ 995 #define M_WRITABLE(m) (((m)->m_flags & (M_RDONLY | M_NOMAP)) == 0 && \ 996 (!(((m)->m_flags & M_EXT)) || \ 997 (m_extrefcnt(m) == 1))) 998 999 /* Check if the supplied mbuf has a packet header, or else panic. */ 1000 #define M_ASSERTPKTHDR(m) \ 1001 KASSERT((m) != NULL && (m)->m_flags & M_PKTHDR, \ 1002 ("%s: no mbuf packet header!", __func__)) 1003 1004 /* 1005 * Ensure that the supplied mbuf is a valid, non-free mbuf. 1006 * 1007 * XXX: Broken at the moment. Need some UMA magic to make it work again. 1008 */ 1009 #define M_ASSERTVALID(m) \ 1010 KASSERT((((struct mbuf *)m)->m_flags & 0) == 0, \ 1011 ("%s: attempted use of a free mbuf!", __func__)) 1012 1013 /* 1014 * Return the address of the start of the buffer associated with an mbuf, 1015 * handling external storage, packet-header mbufs, and regular data mbufs. 1016 */ 1017 #define M_START(m) \ 1018 (((m)->m_flags & M_NOMAP) ? NULL : \ 1019 ((m)->m_flags & M_EXT) ? (m)->m_ext.ext_buf : \ 1020 ((m)->m_flags & M_PKTHDR) ? &(m)->m_pktdat[0] : \ 1021 &(m)->m_dat[0]) 1022 1023 /* 1024 * Return the size of the buffer associated with an mbuf, handling external 1025 * storage, packet-header mbufs, and regular data mbufs. 1026 */ 1027 #define M_SIZE(m) \ 1028 (((m)->m_flags & M_EXT) ? (m)->m_ext.ext_size : \ 1029 ((m)->m_flags & M_PKTHDR) ? MHLEN : \ 1030 MLEN) 1031 1032 /* 1033 * Set the m_data pointer of a newly allocated mbuf to place an object of the 1034 * specified size at the end of the mbuf, longword aligned. 1035 * 1036 * NB: Historically, we had M_ALIGN(), MH_ALIGN(), and MEXT_ALIGN() as 1037 * separate macros, each asserting that it was called at the proper moment. 1038 * This required callers to themselves test the storage type and call the 1039 * right one. Rather than require callers to be aware of those layout 1040 * decisions, we centralize here. 1041 */ 1042 static __inline void 1043 m_align(struct mbuf *m, int len) 1044 { 1045 #ifdef INVARIANTS 1046 const char *msg = "%s: not a virgin mbuf"; 1047 #endif 1048 int adjust; 1049 1050 KASSERT(m->m_data == M_START(m), (msg, __func__)); 1051 1052 adjust = M_SIZE(m) - len; 1053 m->m_data += adjust &~ (sizeof(long)-1); 1054 } 1055 1056 #define M_ALIGN(m, len) m_align(m, len) 1057 #define MH_ALIGN(m, len) m_align(m, len) 1058 #define MEXT_ALIGN(m, len) m_align(m, len) 1059 1060 /* 1061 * Compute the amount of space available before the current start of data in 1062 * an mbuf. 1063 * 1064 * The M_WRITABLE() is a temporary, conservative safety measure: the burden 1065 * of checking writability of the mbuf data area rests solely with the caller. 1066 * 1067 * NB: In previous versions, M_LEADINGSPACE() would only check M_WRITABLE() 1068 * for mbufs with external storage. We now allow mbuf-embedded data to be 1069 * read-only as well. 1070 */ 1071 #define M_LEADINGSPACE(m) \ 1072 (M_WRITABLE(m) ? ((m)->m_data - M_START(m)) : 0) 1073 1074 /* 1075 * Compute the amount of space available after the end of data in an mbuf. 1076 * 1077 * The M_WRITABLE() is a temporary, conservative safety measure: the burden 1078 * of checking writability of the mbuf data area rests solely with the caller. 1079 * 1080 * NB: In previous versions, M_TRAILINGSPACE() would only check M_WRITABLE() 1081 * for mbufs with external storage. We now allow mbuf-embedded data to be 1082 * read-only as well. 1083 */ 1084 #define M_TRAILINGSPACE(m) \ 1085 (M_WRITABLE(m) ? \ 1086 ((M_START(m) + M_SIZE(m)) - ((m)->m_data + (m)->m_len)) : 0) 1087 1088 /* 1089 * Arrange to prepend space of size plen to mbuf m. If a new mbuf must be 1090 * allocated, how specifies whether to wait. If the allocation fails, the 1091 * original mbuf chain is freed and m is set to NULL. 1092 */ 1093 #define M_PREPEND(m, plen, how) do { \ 1094 struct mbuf **_mmp = &(m); \ 1095 struct mbuf *_mm = *_mmp; \ 1096 int _mplen = (plen); \ 1097 int __mhow = (how); \ 1098 \ 1099 MBUF_CHECKSLEEP(how); \ 1100 if (M_LEADINGSPACE(_mm) >= _mplen) { \ 1101 _mm->m_data -= _mplen; \ 1102 _mm->m_len += _mplen; \ 1103 } else \ 1104 _mm = m_prepend(_mm, _mplen, __mhow); \ 1105 if (_mm != NULL && _mm->m_flags & M_PKTHDR) \ 1106 _mm->m_pkthdr.len += _mplen; \ 1107 *_mmp = _mm; \ 1108 } while (0) 1109 1110 /* 1111 * Change mbuf to new type. This is a relatively expensive operation and 1112 * should be avoided. 1113 */ 1114 #define MCHTYPE(m, t) m_chtype((m), (t)) 1115 1116 /* Return the rcvif of a packet header. */ 1117 static __inline struct ifnet * 1118 m_rcvif(struct mbuf *m) 1119 { 1120 1121 M_ASSERTPKTHDR(m); 1122 if (m->m_pkthdr.csum_flags & CSUM_SND_TAG) 1123 return (NULL); 1124 return (m->m_pkthdr.rcvif); 1125 } 1126 1127 /* Length to m_copy to copy all. */ 1128 #define M_COPYALL 1000000000 1129 1130 extern int max_datalen; /* MHLEN - max_hdr */ 1131 extern int max_hdr; /* Largest link + protocol header */ 1132 extern int max_linkhdr; /* Largest link-level header */ 1133 extern int max_protohdr; /* Largest protocol header */ 1134 extern int nmbclusters; /* Maximum number of clusters */ 1135 extern bool mb_use_ext_pgs; /* Use ext_pgs for sendfile */ 1136 1137 /*- 1138 * Network packets may have annotations attached by affixing a list of 1139 * "packet tags" to the pkthdr structure. Packet tags are dynamically 1140 * allocated semi-opaque data structures that have a fixed header 1141 * (struct m_tag) that specifies the size of the memory block and a 1142 * <cookie,type> pair that identifies it. The cookie is a 32-bit unique 1143 * unsigned value used to identify a module or ABI. By convention this value 1144 * is chosen as the date+time that the module is created, expressed as the 1145 * number of seconds since the epoch (e.g., using date -u +'%s'). The type 1146 * value is an ABI/module-specific value that identifies a particular 1147 * annotation and is private to the module. For compatibility with systems 1148 * like OpenBSD that define packet tags w/o an ABI/module cookie, the value 1149 * PACKET_ABI_COMPAT is used to implement m_tag_get and m_tag_find 1150 * compatibility shim functions and several tag types are defined below. 1151 * Users that do not require compatibility should use a private cookie value 1152 * so that packet tag-related definitions can be maintained privately. 1153 * 1154 * Note that the packet tag returned by m_tag_alloc has the default memory 1155 * alignment implemented by malloc. To reference private data one can use a 1156 * construct like: 1157 * 1158 * struct m_tag *mtag = m_tag_alloc(...); 1159 * struct foo *p = (struct foo *)(mtag+1); 1160 * 1161 * if the alignment of struct m_tag is sufficient for referencing members of 1162 * struct foo. Otherwise it is necessary to embed struct m_tag within the 1163 * private data structure to insure proper alignment; e.g., 1164 * 1165 * struct foo { 1166 * struct m_tag tag; 1167 * ... 1168 * }; 1169 * struct foo *p = (struct foo *) m_tag_alloc(...); 1170 * struct m_tag *mtag = &p->tag; 1171 */ 1172 1173 /* 1174 * Persistent tags stay with an mbuf until the mbuf is reclaimed. Otherwise 1175 * tags are expected to ``vanish'' when they pass through a network 1176 * interface. For most interfaces this happens normally as the tags are 1177 * reclaimed when the mbuf is free'd. However in some special cases 1178 * reclaiming must be done manually. An example is packets that pass through 1179 * the loopback interface. Also, one must be careful to do this when 1180 * ``turning around'' packets (e.g., icmp_reflect). 1181 * 1182 * To mark a tag persistent bit-or this flag in when defining the tag id. 1183 * The tag will then be treated as described above. 1184 */ 1185 #define MTAG_PERSISTENT 0x800 1186 1187 #define PACKET_TAG_NONE 0 /* Nadda */ 1188 1189 /* Packet tags for use with PACKET_ABI_COMPAT. */ 1190 #define PACKET_TAG_IPSEC_IN_DONE 1 /* IPsec applied, in */ 1191 #define PACKET_TAG_IPSEC_OUT_DONE 2 /* IPsec applied, out */ 1192 #define PACKET_TAG_IPSEC_IN_CRYPTO_DONE 3 /* NIC IPsec crypto done */ 1193 #define PACKET_TAG_IPSEC_OUT_CRYPTO_NEEDED 4 /* NIC IPsec crypto req'ed */ 1194 #define PACKET_TAG_IPSEC_IN_COULD_DO_CRYPTO 5 /* NIC notifies IPsec */ 1195 #define PACKET_TAG_IPSEC_PENDING_TDB 6 /* Reminder to do IPsec */ 1196 #define PACKET_TAG_BRIDGE 7 /* Bridge processing done */ 1197 #define PACKET_TAG_GIF 8 /* GIF processing done */ 1198 #define PACKET_TAG_GRE 9 /* GRE processing done */ 1199 #define PACKET_TAG_IN_PACKET_CHECKSUM 10 /* NIC checksumming done */ 1200 #define PACKET_TAG_ENCAP 11 /* Encap. processing */ 1201 #define PACKET_TAG_IPSEC_SOCKET 12 /* IPSEC socket ref */ 1202 #define PACKET_TAG_IPSEC_HISTORY 13 /* IPSEC history */ 1203 #define PACKET_TAG_IPV6_INPUT 14 /* IPV6 input processing */ 1204 #define PACKET_TAG_DUMMYNET 15 /* dummynet info */ 1205 #define PACKET_TAG_DIVERT 17 /* divert info */ 1206 #define PACKET_TAG_IPFORWARD 18 /* ipforward info */ 1207 #define PACKET_TAG_MACLABEL (19 | MTAG_PERSISTENT) /* MAC label */ 1208 #define PACKET_TAG_PF (21 | MTAG_PERSISTENT) /* PF/ALTQ information */ 1209 #define PACKET_TAG_RTSOCKFAM 25 /* rtsock sa family */ 1210 #define PACKET_TAG_IPOPTIONS 27 /* Saved IP options */ 1211 #define PACKET_TAG_CARP 28 /* CARP info */ 1212 #define PACKET_TAG_IPSEC_NAT_T_PORTS 29 /* two uint16_t */ 1213 #define PACKET_TAG_ND_OUTGOING 30 /* ND outgoing */ 1214 1215 /* Specific cookies and tags. */ 1216 1217 /* Packet tag routines. */ 1218 struct m_tag *m_tag_alloc(u_int32_t, int, int, int); 1219 void m_tag_delete(struct mbuf *, struct m_tag *); 1220 void m_tag_delete_chain(struct mbuf *, struct m_tag *); 1221 void m_tag_free_default(struct m_tag *); 1222 struct m_tag *m_tag_locate(struct mbuf *, u_int32_t, int, struct m_tag *); 1223 struct m_tag *m_tag_copy(struct m_tag *, int); 1224 int m_tag_copy_chain(struct mbuf *, const struct mbuf *, int); 1225 void m_tag_delete_nonpersistent(struct mbuf *); 1226 1227 /* 1228 * Initialize the list of tags associated with an mbuf. 1229 */ 1230 static __inline void 1231 m_tag_init(struct mbuf *m) 1232 { 1233 1234 SLIST_INIT(&m->m_pkthdr.tags); 1235 } 1236 1237 /* 1238 * Set up the contents of a tag. Note that this does not fill in the free 1239 * method; the caller is expected to do that. 1240 * 1241 * XXX probably should be called m_tag_init, but that was already taken. 1242 */ 1243 static __inline void 1244 m_tag_setup(struct m_tag *t, u_int32_t cookie, int type, int len) 1245 { 1246 1247 t->m_tag_id = type; 1248 t->m_tag_len = len; 1249 t->m_tag_cookie = cookie; 1250 } 1251 1252 /* 1253 * Reclaim resources associated with a tag. 1254 */ 1255 static __inline void 1256 m_tag_free(struct m_tag *t) 1257 { 1258 1259 (*t->m_tag_free)(t); 1260 } 1261 1262 /* 1263 * Return the first tag associated with an mbuf. 1264 */ 1265 static __inline struct m_tag * 1266 m_tag_first(struct mbuf *m) 1267 { 1268 1269 return (SLIST_FIRST(&m->m_pkthdr.tags)); 1270 } 1271 1272 /* 1273 * Return the next tag in the list of tags associated with an mbuf. 1274 */ 1275 static __inline struct m_tag * 1276 m_tag_next(struct mbuf *m __unused, struct m_tag *t) 1277 { 1278 1279 return (SLIST_NEXT(t, m_tag_link)); 1280 } 1281 1282 /* 1283 * Prepend a tag to the list of tags associated with an mbuf. 1284 */ 1285 static __inline void 1286 m_tag_prepend(struct mbuf *m, struct m_tag *t) 1287 { 1288 1289 SLIST_INSERT_HEAD(&m->m_pkthdr.tags, t, m_tag_link); 1290 } 1291 1292 /* 1293 * Unlink a tag from the list of tags associated with an mbuf. 1294 */ 1295 static __inline void 1296 m_tag_unlink(struct mbuf *m, struct m_tag *t) 1297 { 1298 1299 SLIST_REMOVE(&m->m_pkthdr.tags, t, m_tag, m_tag_link); 1300 } 1301 1302 /* These are for OpenBSD compatibility. */ 1303 #define MTAG_ABI_COMPAT 0 /* compatibility ABI */ 1304 1305 static __inline struct m_tag * 1306 m_tag_get(int type, int length, int wait) 1307 { 1308 return (m_tag_alloc(MTAG_ABI_COMPAT, type, length, wait)); 1309 } 1310 1311 static __inline struct m_tag * 1312 m_tag_find(struct mbuf *m, int type, struct m_tag *start) 1313 { 1314 return (SLIST_EMPTY(&m->m_pkthdr.tags) ? (struct m_tag *)NULL : 1315 m_tag_locate(m, MTAG_ABI_COMPAT, type, start)); 1316 } 1317 1318 static inline struct m_snd_tag * 1319 m_snd_tag_ref(struct m_snd_tag *mst) 1320 { 1321 1322 refcount_acquire(&mst->refcount); 1323 return (mst); 1324 } 1325 1326 static inline void 1327 m_snd_tag_rele(struct m_snd_tag *mst) 1328 { 1329 1330 if (refcount_release(&mst->refcount)) 1331 m_snd_tag_destroy(mst); 1332 } 1333 1334 static __inline struct mbuf * 1335 m_free(struct mbuf *m) 1336 { 1337 struct mbuf *n = m->m_next; 1338 1339 MBUF_PROBE1(m__free, m); 1340 if ((m->m_flags & (M_PKTHDR|M_NOFREE)) == (M_PKTHDR|M_NOFREE)) 1341 m_tag_delete_chain(m, NULL); 1342 if (m->m_flags & M_PKTHDR && m->m_pkthdr.csum_flags & CSUM_SND_TAG) 1343 m_snd_tag_rele(m->m_pkthdr.snd_tag); 1344 if (m->m_flags & M_EXT) 1345 mb_free_ext(m); 1346 else if ((m->m_flags & M_NOFREE) == 0) 1347 uma_zfree(zone_mbuf, m); 1348 return (n); 1349 } 1350 1351 static __inline int 1352 rt_m_getfib(struct mbuf *m) 1353 { 1354 KASSERT(m->m_flags & M_PKTHDR , ("Attempt to get FIB from non header mbuf.")); 1355 return (m->m_pkthdr.fibnum); 1356 } 1357 1358 #define M_GETFIB(_m) rt_m_getfib(_m) 1359 1360 #define M_SETFIB(_m, _fib) do { \ 1361 KASSERT((_m)->m_flags & M_PKTHDR, ("Attempt to set FIB on non header mbuf.")); \ 1362 ((_m)->m_pkthdr.fibnum) = (_fib); \ 1363 } while (0) 1364 1365 /* flags passed as first argument for "m_ether_tcpip_hash()" */ 1366 #define MBUF_HASHFLAG_L2 (1 << 2) 1367 #define MBUF_HASHFLAG_L3 (1 << 3) 1368 #define MBUF_HASHFLAG_L4 (1 << 4) 1369 1370 /* mbuf hashing helper routines */ 1371 uint32_t m_ether_tcpip_hash_init(void); 1372 uint32_t m_ether_tcpip_hash(const uint32_t, const struct mbuf *, const uint32_t); 1373 1374 #ifdef MBUF_PROFILING 1375 void m_profile(struct mbuf *m); 1376 #define M_PROFILE(m) m_profile(m) 1377 #else 1378 #define M_PROFILE(m) 1379 #endif 1380 1381 struct mbufq { 1382 STAILQ_HEAD(, mbuf) mq_head; 1383 int mq_len; 1384 int mq_maxlen; 1385 }; 1386 1387 static inline void 1388 mbufq_init(struct mbufq *mq, int maxlen) 1389 { 1390 1391 STAILQ_INIT(&mq->mq_head); 1392 mq->mq_maxlen = maxlen; 1393 mq->mq_len = 0; 1394 } 1395 1396 static inline struct mbuf * 1397 mbufq_flush(struct mbufq *mq) 1398 { 1399 struct mbuf *m; 1400 1401 m = STAILQ_FIRST(&mq->mq_head); 1402 STAILQ_INIT(&mq->mq_head); 1403 mq->mq_len = 0; 1404 return (m); 1405 } 1406 1407 static inline void 1408 mbufq_drain(struct mbufq *mq) 1409 { 1410 struct mbuf *m, *n; 1411 1412 n = mbufq_flush(mq); 1413 while ((m = n) != NULL) { 1414 n = STAILQ_NEXT(m, m_stailqpkt); 1415 m_freem(m); 1416 } 1417 } 1418 1419 static inline struct mbuf * 1420 mbufq_first(const struct mbufq *mq) 1421 { 1422 1423 return (STAILQ_FIRST(&mq->mq_head)); 1424 } 1425 1426 static inline struct mbuf * 1427 mbufq_last(const struct mbufq *mq) 1428 { 1429 1430 return (STAILQ_LAST(&mq->mq_head, mbuf, m_stailqpkt)); 1431 } 1432 1433 static inline int 1434 mbufq_full(const struct mbufq *mq) 1435 { 1436 1437 return (mq->mq_maxlen > 0 && mq->mq_len >= mq->mq_maxlen); 1438 } 1439 1440 static inline int 1441 mbufq_len(const struct mbufq *mq) 1442 { 1443 1444 return (mq->mq_len); 1445 } 1446 1447 static inline int 1448 mbufq_enqueue(struct mbufq *mq, struct mbuf *m) 1449 { 1450 1451 if (mbufq_full(mq)) 1452 return (ENOBUFS); 1453 STAILQ_INSERT_TAIL(&mq->mq_head, m, m_stailqpkt); 1454 mq->mq_len++; 1455 return (0); 1456 } 1457 1458 static inline struct mbuf * 1459 mbufq_dequeue(struct mbufq *mq) 1460 { 1461 struct mbuf *m; 1462 1463 m = STAILQ_FIRST(&mq->mq_head); 1464 if (m) { 1465 STAILQ_REMOVE_HEAD(&mq->mq_head, m_stailqpkt); 1466 m->m_nextpkt = NULL; 1467 mq->mq_len--; 1468 } 1469 return (m); 1470 } 1471 1472 static inline void 1473 mbufq_prepend(struct mbufq *mq, struct mbuf *m) 1474 { 1475 1476 STAILQ_INSERT_HEAD(&mq->mq_head, m, m_stailqpkt); 1477 mq->mq_len++; 1478 } 1479 1480 /* 1481 * Note: this doesn't enforce the maximum list size for dst. 1482 */ 1483 static inline void 1484 mbufq_concat(struct mbufq *mq_dst, struct mbufq *mq_src) 1485 { 1486 1487 mq_dst->mq_len += mq_src->mq_len; 1488 STAILQ_CONCAT(&mq_dst->mq_head, &mq_src->mq_head); 1489 mq_src->mq_len = 0; 1490 } 1491 1492 #ifdef _SYS_TIMESPEC_H_ 1493 static inline void 1494 mbuf_tstmp2timespec(struct mbuf *m, struct timespec *ts) 1495 { 1496 1497 KASSERT((m->m_flags & M_PKTHDR) != 0, ("mbuf %p no M_PKTHDR", m)); 1498 KASSERT((m->m_flags & (M_TSTMP|M_TSTMP_LRO)) != 0, ("mbuf %p no M_TSTMP or M_TSTMP_LRO", m)); 1499 ts->tv_sec = m->m_pkthdr.rcv_tstmp / 1000000000; 1500 ts->tv_nsec = m->m_pkthdr.rcv_tstmp % 1000000000; 1501 } 1502 #endif 1503 1504 #ifdef DEBUGNET 1505 /* Invoked from the debugnet client code. */ 1506 void debugnet_mbuf_drain(void); 1507 void debugnet_mbuf_start(void); 1508 void debugnet_mbuf_finish(void); 1509 void debugnet_mbuf_reinit(int nmbuf, int nclust, int clsize); 1510 #endif 1511 1512 static inline bool 1513 mbuf_has_tls_session(struct mbuf *m) 1514 { 1515 1516 if (m->m_flags & M_NOMAP) { 1517 MBUF_EXT_PGS_ASSERT(m); 1518 if (m->m_ext.ext_pgs->tls != NULL) { 1519 return (true); 1520 } 1521 } 1522 return (false); 1523 } 1524 1525 #endif /* _KERNEL */ 1526 #endif /* !_SYS_MBUF_H_ */ 1527