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] 198 #define tso_segsz PH_per.sixteen[1] 199 #define lro_nsegs tso_segsz 200 #define csum_phsum PH_per.sixteen[2] 201 #define csum_data PH_per.thirtytwo[1] 202 #define lro_len PH_per.sixteen[0] /* inbound during LRO */ 203 #define lro_csum PH_per.sixteen[1] /* inbound during LRO */ 204 #define pace_thoff PH_loc.sixteen[0] 205 #define pace_tlen PH_loc.sixteen[1] 206 #define pace_drphdrlen PH_loc.sixteen[2] 207 #define pace_tos PH_loc.eight[6] 208 #define pace_lock PH_loc.eight[7] 209 210 /* 211 * Description of external storage mapped into mbuf; valid only if M_EXT is 212 * set. 213 * Size ILP32: 28 214 * LP64: 48 215 * Compile-time assertions in uipc_mbuf.c test these values to ensure that 216 * they are correct. 217 */ 218 typedef void m_ext_free_t(struct mbuf *); 219 struct m_ext { 220 union { 221 /* 222 * If EXT_FLAG_EMBREF is set, then we use refcount in the 223 * mbuf, the 'ext_count' member. Otherwise, we have a 224 * shadow copy and we use pointer 'ext_cnt'. The original 225 * mbuf is responsible to carry the pointer to free routine 226 * and its arguments. They aren't copied into shadows in 227 * mb_dupcl() to avoid dereferencing next cachelines. 228 */ 229 volatile u_int ext_count; 230 volatile u_int *ext_cnt; 231 }; 232 union { 233 /* 234 * If ext_type == EXT_PGS, 'ext_pgs' points to a 235 * structure describing the buffer. Otherwise, 236 * 'ext_buf' points to the start of the buffer. 237 */ 238 struct mbuf_ext_pgs *ext_pgs; 239 char *ext_buf; 240 }; 241 uint32_t ext_size; /* size of buffer, for ext_free */ 242 uint32_t ext_type:8, /* type of external storage */ 243 ext_flags:24; /* external storage mbuf flags */ 244 /* 245 * Fields below store the free context for the external storage. 246 * They are valid only in the refcount carrying mbuf, the one with 247 * EXT_FLAG_EMBREF flag, with exclusion for EXT_EXTREF type, where 248 * the free context is copied into all mbufs that use same external 249 * storage. 250 */ 251 #define m_ext_copylen offsetof(struct m_ext, ext_free) 252 m_ext_free_t *ext_free; /* free routine if not the usual */ 253 void *ext_arg1; /* optional argument pointer */ 254 void *ext_arg2; /* optional argument pointer */ 255 }; 256 257 /* 258 * The core of the mbuf object along with some shortcut defines for practical 259 * purposes. 260 */ 261 struct mbuf { 262 /* 263 * Header present at the beginning of every mbuf. 264 * Size ILP32: 24 265 * LP64: 32 266 * Compile-time assertions in uipc_mbuf.c test these values to ensure 267 * that they are correct. 268 */ 269 union { /* next buffer in chain */ 270 struct mbuf *m_next; 271 SLIST_ENTRY(mbuf) m_slist; 272 STAILQ_ENTRY(mbuf) m_stailq; 273 }; 274 union { /* next chain in queue/record */ 275 struct mbuf *m_nextpkt; 276 SLIST_ENTRY(mbuf) m_slistpkt; 277 STAILQ_ENTRY(mbuf) m_stailqpkt; 278 }; 279 caddr_t m_data; /* location of data */ 280 int32_t m_len; /* amount of data in this mbuf */ 281 uint32_t m_type:8, /* type of data in this mbuf */ 282 m_flags:24; /* flags; see below */ 283 #if !defined(__LP64__) 284 uint32_t m_pad; /* pad for 64bit alignment */ 285 #endif 286 287 /* 288 * A set of optional headers (packet header, external storage header) 289 * and internal data storage. Historically, these arrays were sized 290 * to MHLEN (space left after a packet header) and MLEN (space left 291 * after only a regular mbuf header); they are now variable size in 292 * order to support future work on variable-size mbufs. 293 */ 294 union { 295 struct { 296 struct pkthdr m_pkthdr; /* M_PKTHDR set */ 297 union { 298 struct m_ext m_ext; /* M_EXT set */ 299 char m_pktdat[0]; 300 }; 301 }; 302 char m_dat[0]; /* !M_PKTHDR, !M_EXT */ 303 }; 304 }; 305 306 struct ktls_session; 307 struct socket; 308 309 /* 310 * TLS records for TLS 1.0-1.2 can have the following header lengths: 311 * - 5 (AES-CBC with implicit IV) 312 * - 21 (AES-CBC with explicit IV) 313 * - 13 (AES-GCM with 8 byte explicit IV) 314 */ 315 #define MBUF_PEXT_HDR_LEN 23 316 317 /* 318 * TLS records for TLS 1.0-1.2 can have the following maximum trailer 319 * lengths: 320 * - 16 (AES-GCM) 321 * - 36 (AES-CBC with SHA1 and up to 16 bytes of padding) 322 * - 48 (AES-CBC with SHA2-256 and up to 16 bytes of padding) 323 * - 64 (AES-CBC with SHA2-384 and up to 16 bytes of padding) 324 */ 325 #define MBUF_PEXT_TRAIL_LEN 64 326 327 #ifdef __LP64__ 328 #define MBUF_PEXT_MAX_PGS (152 / sizeof(vm_paddr_t)) 329 #else 330 #define MBUF_PEXT_MAX_PGS (156 / sizeof(vm_paddr_t)) 331 #endif 332 333 #define MBUF_PEXT_MAX_BYTES \ 334 (MBUF_PEXT_MAX_PGS * PAGE_SIZE + MBUF_PEXT_HDR_LEN + MBUF_PEXT_TRAIL_LEN) 335 336 #define MBUF_PEXT_FLAG_ANON 1 /* Data can be encrypted in place. */ 337 338 /* 339 * This struct is 256 bytes in size and is arranged so that the most 340 * common case (accessing the first 4 pages of a 16KB TLS record) will 341 * fit in a single 64 byte cacheline. 342 */ 343 struct mbuf_ext_pgs { 344 uint8_t npgs; /* Number of attached pages */ 345 uint8_t nrdy; /* Pages with I/O pending */ 346 uint8_t hdr_len; /* TLS header length */ 347 uint8_t trail_len; /* TLS trailer length */ 348 uint16_t first_pg_off; /* Offset into 1st page */ 349 uint16_t last_pg_len; /* Length of last page */ 350 vm_paddr_t pa[MBUF_PEXT_MAX_PGS]; /* phys addrs of pages */ 351 char hdr[MBUF_PEXT_HDR_LEN]; /* TLS header */ 352 uint8_t flags; /* Flags */ 353 struct ktls_session *tls; /* TLS session */ 354 #if defined(__i386__) || \ 355 (defined(__powerpc__) && !defined(__powerpc64__) && defined(BOOKE)) 356 /* 357 * i386 and Book-E PowerPC have 64-bit vm_paddr_t, so there is 358 * a 4 byte remainder from the space allocated for pa[]. 359 */ 360 uint32_t pad; 361 #endif 362 union { 363 char trail[MBUF_PEXT_TRAIL_LEN]; /* TLS trailer */ 364 struct { 365 uint8_t record_type; /* Must be first */ 366 struct socket *so; 367 struct mbuf *mbuf; 368 uint64_t seqno; 369 STAILQ_ENTRY(mbuf_ext_pgs) stailq; 370 int enc_cnt; 371 }; 372 }; 373 }; 374 375 #ifdef _KERNEL 376 static inline int 377 mbuf_ext_pg_len(struct mbuf_ext_pgs *ext_pgs, int pidx, int pgoff) 378 { 379 KASSERT(pgoff == 0 || pidx == 0, 380 ("page %d with non-zero offset %d in %p", pidx, pgoff, ext_pgs)); 381 if (pidx == ext_pgs->npgs - 1) { 382 return (ext_pgs->last_pg_len); 383 } else { 384 return (PAGE_SIZE - pgoff); 385 } 386 } 387 388 #ifdef INVARIANT_SUPPORT 389 void mb_ext_pgs_check(struct mbuf_ext_pgs *ext_pgs); 390 #endif 391 #ifdef INVARIANTS 392 #define MBUF_EXT_PGS_ASSERT_SANITY(ext_pgs) mb_ext_pgs_check((ext_pgs)) 393 #else 394 #define MBUF_EXT_PGS_ASSERT_SANITY(ext_pgs) 395 #endif 396 #endif 397 398 /* 399 * mbuf flags of global significance and layer crossing. 400 * Those of only protocol/layer specific significance are to be mapped 401 * to M_PROTO[1-11] and cleared at layer handoff boundaries. 402 * NB: Limited to the lower 24 bits. 403 */ 404 #define M_EXT 0x00000001 /* has associated external storage */ 405 #define M_PKTHDR 0x00000002 /* start of record */ 406 #define M_EOR 0x00000004 /* end of record */ 407 #define M_RDONLY 0x00000008 /* associated data is marked read-only */ 408 #define M_BCAST 0x00000010 /* send/received as link-level broadcast */ 409 #define M_MCAST 0x00000020 /* send/received as link-level multicast */ 410 #define M_PROMISC 0x00000040 /* packet was not for us */ 411 #define M_VLANTAG 0x00000080 /* ether_vtag is valid */ 412 #define M_NOMAP 0x00000100 /* mbuf data is unmapped */ 413 #define M_NOFREE 0x00000200 /* do not free mbuf, embedded in cluster */ 414 #define M_TSTMP 0x00000400 /* rcv_tstmp field is valid */ 415 #define M_TSTMP_HPREC 0x00000800 /* rcv_tstmp is high-prec, typically 416 hw-stamped on port (useful for IEEE 1588 417 and 802.1AS) */ 418 #define M_TSTMP_LRO 0x00001000 /* Time LRO pushed in pkt is valid in (PH_loc) */ 419 420 #define M_PROTO1 0x00002000 /* protocol-specific */ 421 #define M_PROTO2 0x00004000 /* protocol-specific */ 422 #define M_PROTO3 0x00008000 /* protocol-specific */ 423 #define M_PROTO4 0x00010000 /* protocol-specific */ 424 #define M_PROTO5 0x00020000 /* protocol-specific */ 425 #define M_PROTO6 0x00040000 /* protocol-specific */ 426 #define M_PROTO7 0x00080000 /* protocol-specific */ 427 #define M_PROTO8 0x00100000 /* protocol-specific */ 428 #define M_PROTO9 0x00200000 /* protocol-specific */ 429 #define M_PROTO10 0x00400000 /* protocol-specific */ 430 #define M_PROTO11 0x00800000 /* protocol-specific */ 431 432 #define MB_DTOR_SKIP 0x1 /* don't pollute the cache by touching a freed mbuf */ 433 434 /* 435 * Flags to purge when crossing layers. 436 */ 437 #define M_PROTOFLAGS \ 438 (M_PROTO1|M_PROTO2|M_PROTO3|M_PROTO4|M_PROTO5|M_PROTO6|M_PROTO7|M_PROTO8|\ 439 M_PROTO9|M_PROTO10|M_PROTO11) 440 441 /* 442 * Flags preserved when copying m_pkthdr. 443 */ 444 #define M_COPYFLAGS \ 445 (M_PKTHDR|M_EOR|M_RDONLY|M_BCAST|M_MCAST|M_PROMISC|M_VLANTAG|M_TSTMP| \ 446 M_TSTMP_HPREC|M_TSTMP_LRO|M_PROTOFLAGS) 447 448 /* 449 * Mbuf flag description for use with printf(9) %b identifier. 450 */ 451 #define M_FLAG_BITS \ 452 "\20\1M_EXT\2M_PKTHDR\3M_EOR\4M_RDONLY\5M_BCAST\6M_MCAST" \ 453 "\7M_PROMISC\10M_VLANTAG\11M_NOMAP\12M_NOFREE\13M_TSTMP\14M_TSTMP_HPREC\15M_TSTMP_LRO" 454 #define M_FLAG_PROTOBITS \ 455 "\16M_PROTO1\17M_PROTO2\20M_PROTO3\21M_PROTO4" \ 456 "\22M_PROTO5\23M_PROTO6\24M_PROTO7\25M_PROTO8\26M_PROTO9" \ 457 "\27M_PROTO10\28M_PROTO11" 458 #define M_FLAG_PRINTF (M_FLAG_BITS M_FLAG_PROTOBITS) 459 460 /* 461 * Network interface cards are able to hash protocol fields (such as IPv4 462 * addresses and TCP port numbers) classify packets into flows. These flows 463 * can then be used to maintain ordering while delivering packets to the OS 464 * via parallel input queues, as well as to provide a stateless affinity 465 * model. NIC drivers can pass up the hash via m->m_pkthdr.flowid, and set 466 * m_flag fields to indicate how the hash should be interpreted by the 467 * network stack. 468 * 469 * Most NICs support RSS, which provides ordering and explicit affinity, and 470 * use the hash m_flag bits to indicate what header fields were covered by 471 * the hash. M_HASHTYPE_OPAQUE and M_HASHTYPE_OPAQUE_HASH can be set by non- 472 * RSS cards or configurations that provide an opaque flow identifier, allowing 473 * for ordering and distribution without explicit affinity. Additionally, 474 * M_HASHTYPE_OPAQUE_HASH indicates that the flow identifier has hash 475 * properties. 476 * 477 * The meaning of the IPV6_EX suffix: 478 * "o Home address from the home address option in the IPv6 destination 479 * options header. If the extension header is not present, use the Source 480 * IPv6 Address. 481 * o IPv6 address that is contained in the Routing-Header-Type-2 from the 482 * associated extension header. If the extension header is not present, 483 * use the Destination IPv6 Address." 484 * Quoted from: 485 * https://docs.microsoft.com/en-us/windows-hardware/drivers/network/rss-hashing-types#ndishashipv6ex 486 */ 487 #define M_HASHTYPE_HASHPROP 0x80 /* has hash properties */ 488 #define M_HASHTYPE_HASH(t) (M_HASHTYPE_HASHPROP | (t)) 489 /* Microsoft RSS standard hash types */ 490 #define M_HASHTYPE_NONE 0 491 #define M_HASHTYPE_RSS_IPV4 M_HASHTYPE_HASH(1) /* IPv4 2-tuple */ 492 #define M_HASHTYPE_RSS_TCP_IPV4 M_HASHTYPE_HASH(2) /* TCPv4 4-tuple */ 493 #define M_HASHTYPE_RSS_IPV6 M_HASHTYPE_HASH(3) /* IPv6 2-tuple */ 494 #define M_HASHTYPE_RSS_TCP_IPV6 M_HASHTYPE_HASH(4) /* TCPv6 4-tuple */ 495 #define M_HASHTYPE_RSS_IPV6_EX M_HASHTYPE_HASH(5) /* IPv6 2-tuple + 496 * ext hdrs */ 497 #define M_HASHTYPE_RSS_TCP_IPV6_EX M_HASHTYPE_HASH(6) /* TCPv6 4-tuple + 498 * ext hdrs */ 499 #define M_HASHTYPE_RSS_UDP_IPV4 M_HASHTYPE_HASH(7) /* IPv4 UDP 4-tuple*/ 500 #define M_HASHTYPE_RSS_UDP_IPV6 M_HASHTYPE_HASH(9) /* IPv6 UDP 4-tuple*/ 501 #define M_HASHTYPE_RSS_UDP_IPV6_EX M_HASHTYPE_HASH(10)/* IPv6 UDP 4-tuple + 502 * ext hdrs */ 503 504 #define M_HASHTYPE_OPAQUE 63 /* ordering, not affinity */ 505 #define M_HASHTYPE_OPAQUE_HASH M_HASHTYPE_HASH(M_HASHTYPE_OPAQUE) 506 /* ordering+hash, not affinity*/ 507 508 #define M_HASHTYPE_CLEAR(m) ((m)->m_pkthdr.rsstype = 0) 509 #define M_HASHTYPE_GET(m) ((m)->m_pkthdr.rsstype) 510 #define M_HASHTYPE_SET(m, v) ((m)->m_pkthdr.rsstype = (v)) 511 #define M_HASHTYPE_TEST(m, v) (M_HASHTYPE_GET(m) == (v)) 512 #define M_HASHTYPE_ISHASH(m) (M_HASHTYPE_GET(m) & M_HASHTYPE_HASHPROP) 513 514 /* 515 * External mbuf storage buffer types. 516 */ 517 #define EXT_CLUSTER 1 /* mbuf cluster */ 518 #define EXT_SFBUF 2 /* sendfile(2)'s sf_buf */ 519 #define EXT_JUMBOP 3 /* jumbo cluster page sized */ 520 #define EXT_JUMBO9 4 /* jumbo cluster 9216 bytes */ 521 #define EXT_JUMBO16 5 /* jumbo cluster 16184 bytes */ 522 #define EXT_PACKET 6 /* mbuf+cluster from packet zone */ 523 #define EXT_MBUF 7 /* external mbuf reference */ 524 #define EXT_RXRING 8 /* data in NIC receive ring */ 525 #define EXT_PGS 9 /* array of unmapped pages */ 526 527 #define EXT_VENDOR1 224 /* for vendor-internal use */ 528 #define EXT_VENDOR2 225 /* for vendor-internal use */ 529 #define EXT_VENDOR3 226 /* for vendor-internal use */ 530 #define EXT_VENDOR4 227 /* for vendor-internal use */ 531 532 #define EXT_EXP1 244 /* for experimental use */ 533 #define EXT_EXP2 245 /* for experimental use */ 534 #define EXT_EXP3 246 /* for experimental use */ 535 #define EXT_EXP4 247 /* for experimental use */ 536 537 #define EXT_NET_DRV 252 /* custom ext_buf provided by net driver(s) */ 538 #define EXT_MOD_TYPE 253 /* custom module's ext_buf type */ 539 #define EXT_DISPOSABLE 254 /* can throw this buffer away w/page flipping */ 540 #define EXT_EXTREF 255 /* has externally maintained ext_cnt ptr */ 541 542 /* 543 * Flags for external mbuf buffer types. 544 * NB: limited to the lower 24 bits. 545 */ 546 #define EXT_FLAG_EMBREF 0x000001 /* embedded ext_count */ 547 #define EXT_FLAG_EXTREF 0x000002 /* external ext_cnt, notyet */ 548 549 #define EXT_FLAG_NOFREE 0x000010 /* don't free mbuf to pool, notyet */ 550 551 #define EXT_FLAG_VENDOR1 0x010000 /* These flags are vendor */ 552 #define EXT_FLAG_VENDOR2 0x020000 /* or submodule specific, */ 553 #define EXT_FLAG_VENDOR3 0x040000 /* not used by mbuf code. */ 554 #define EXT_FLAG_VENDOR4 0x080000 /* Set/read by submodule. */ 555 556 #define EXT_FLAG_EXP1 0x100000 /* for experimental use */ 557 #define EXT_FLAG_EXP2 0x200000 /* for experimental use */ 558 #define EXT_FLAG_EXP3 0x400000 /* for experimental use */ 559 #define EXT_FLAG_EXP4 0x800000 /* for experimental use */ 560 561 /* 562 * EXT flag description for use with printf(9) %b identifier. 563 */ 564 #define EXT_FLAG_BITS \ 565 "\20\1EXT_FLAG_EMBREF\2EXT_FLAG_EXTREF\5EXT_FLAG_NOFREE" \ 566 "\21EXT_FLAG_VENDOR1\22EXT_FLAG_VENDOR2\23EXT_FLAG_VENDOR3" \ 567 "\24EXT_FLAG_VENDOR4\25EXT_FLAG_EXP1\26EXT_FLAG_EXP2\27EXT_FLAG_EXP3" \ 568 "\30EXT_FLAG_EXP4" 569 570 #define MBUF_EXT_PGS_ASSERT(m) \ 571 KASSERT((((m)->m_flags & M_EXT) != 0) && \ 572 ((m)->m_ext.ext_type == EXT_PGS), \ 573 ("%s: m %p !M_EXT or !EXT_PGS", __func__, m)) 574 575 /* 576 * Flags indicating checksum, segmentation and other offload work to be 577 * done, or already done, by hardware or lower layers. It is split into 578 * separate inbound and outbound flags. 579 * 580 * Outbound flags that are set by upper protocol layers requesting lower 581 * layers, or ideally the hardware, to perform these offloading tasks. 582 * For outbound packets this field and its flags can be directly tested 583 * against ifnet if_hwassist. 584 */ 585 #define CSUM_IP 0x00000001 /* IP header checksum offload */ 586 #define CSUM_IP_UDP 0x00000002 /* UDP checksum offload */ 587 #define CSUM_IP_TCP 0x00000004 /* TCP checksum offload */ 588 #define CSUM_IP_SCTP 0x00000008 /* SCTP checksum offload */ 589 #define CSUM_IP_TSO 0x00000010 /* TCP segmentation offload */ 590 #define CSUM_IP_ISCSI 0x00000020 /* iSCSI checksum offload */ 591 592 #define CSUM_IP6_UDP 0x00000200 /* UDP checksum offload */ 593 #define CSUM_IP6_TCP 0x00000400 /* TCP checksum offload */ 594 #define CSUM_IP6_SCTP 0x00000800 /* SCTP checksum offload */ 595 #define CSUM_IP6_TSO 0x00001000 /* TCP segmentation offload */ 596 #define CSUM_IP6_ISCSI 0x00002000 /* iSCSI checksum offload */ 597 598 /* Inbound checksum support where the checksum was verified by hardware. */ 599 #define CSUM_L3_CALC 0x01000000 /* calculated layer 3 csum */ 600 #define CSUM_L3_VALID 0x02000000 /* checksum is correct */ 601 #define CSUM_L4_CALC 0x04000000 /* calculated layer 4 csum */ 602 #define CSUM_L4_VALID 0x08000000 /* checksum is correct */ 603 #define CSUM_L5_CALC 0x10000000 /* calculated layer 5 csum */ 604 #define CSUM_L5_VALID 0x20000000 /* checksum is correct */ 605 #define CSUM_COALESCED 0x40000000 /* contains merged segments */ 606 607 #define CSUM_SND_TAG 0x80000000 /* Packet header has send tag */ 608 609 /* 610 * CSUM flag description for use with printf(9) %b identifier. 611 */ 612 #define CSUM_BITS \ 613 "\20\1CSUM_IP\2CSUM_IP_UDP\3CSUM_IP_TCP\4CSUM_IP_SCTP\5CSUM_IP_TSO" \ 614 "\6CSUM_IP_ISCSI" \ 615 "\12CSUM_IP6_UDP\13CSUM_IP6_TCP\14CSUM_IP6_SCTP\15CSUM_IP6_TSO" \ 616 "\16CSUM_IP6_ISCSI" \ 617 "\31CSUM_L3_CALC\32CSUM_L3_VALID\33CSUM_L4_CALC\34CSUM_L4_VALID" \ 618 "\35CSUM_L5_CALC\36CSUM_L5_VALID\37CSUM_COALESCED\40CSUM_SND_TAG" 619 620 /* CSUM flags compatibility mappings. */ 621 #define CSUM_IP_CHECKED CSUM_L3_CALC 622 #define CSUM_IP_VALID CSUM_L3_VALID 623 #define CSUM_DATA_VALID CSUM_L4_VALID 624 #define CSUM_PSEUDO_HDR CSUM_L4_CALC 625 #define CSUM_SCTP_VALID CSUM_L4_VALID 626 #define CSUM_DELAY_DATA (CSUM_TCP|CSUM_UDP) 627 #define CSUM_DELAY_IP CSUM_IP /* Only v4, no v6 IP hdr csum */ 628 #define CSUM_DELAY_DATA_IPV6 (CSUM_TCP_IPV6|CSUM_UDP_IPV6) 629 #define CSUM_DATA_VALID_IPV6 CSUM_DATA_VALID 630 #define CSUM_TCP CSUM_IP_TCP 631 #define CSUM_UDP CSUM_IP_UDP 632 #define CSUM_SCTP CSUM_IP_SCTP 633 #define CSUM_TSO (CSUM_IP_TSO|CSUM_IP6_TSO) 634 #define CSUM_UDP_IPV6 CSUM_IP6_UDP 635 #define CSUM_TCP_IPV6 CSUM_IP6_TCP 636 #define CSUM_SCTP_IPV6 CSUM_IP6_SCTP 637 638 /* 639 * mbuf types describing the content of the mbuf (including external storage). 640 */ 641 #define MT_NOTMBUF 0 /* USED INTERNALLY ONLY! Object is not mbuf */ 642 #define MT_DATA 1 /* dynamic (data) allocation */ 643 #define MT_HEADER MT_DATA /* packet header, use M_PKTHDR instead */ 644 645 #define MT_VENDOR1 4 /* for vendor-internal use */ 646 #define MT_VENDOR2 5 /* for vendor-internal use */ 647 #define MT_VENDOR3 6 /* for vendor-internal use */ 648 #define MT_VENDOR4 7 /* for vendor-internal use */ 649 650 #define MT_SONAME 8 /* socket name */ 651 652 #define MT_EXP1 9 /* for experimental use */ 653 #define MT_EXP2 10 /* for experimental use */ 654 #define MT_EXP3 11 /* for experimental use */ 655 #define MT_EXP4 12 /* for experimental use */ 656 657 #define MT_CONTROL 14 /* extra-data protocol message */ 658 #define MT_EXTCONTROL 15 /* control message with externalized contents */ 659 #define MT_OOBDATA 16 /* expedited data */ 660 661 #define MT_NOINIT 255 /* Not a type but a flag to allocate 662 a non-initialized mbuf */ 663 664 /* 665 * String names of mbuf-related UMA(9) and malloc(9) types. Exposed to 666 * !_KERNEL so that monitoring tools can look up the zones with 667 * libmemstat(3). 668 */ 669 #define MBUF_MEM_NAME "mbuf" 670 #define MBUF_CLUSTER_MEM_NAME "mbuf_cluster" 671 #define MBUF_PACKET_MEM_NAME "mbuf_packet" 672 #define MBUF_JUMBOP_MEM_NAME "mbuf_jumbo_page" 673 #define MBUF_JUMBO9_MEM_NAME "mbuf_jumbo_9k" 674 #define MBUF_JUMBO16_MEM_NAME "mbuf_jumbo_16k" 675 #define MBUF_TAG_MEM_NAME "mbuf_tag" 676 #define MBUF_EXTREFCNT_MEM_NAME "mbuf_ext_refcnt" 677 #define MBUF_EXTPGS_MEM_NAME "mbuf_extpgs" 678 679 #ifdef _KERNEL 680 681 #ifdef WITNESS 682 #define MBUF_CHECKSLEEP(how) do { \ 683 if (how == M_WAITOK) \ 684 WITNESS_WARN(WARN_GIANTOK | WARN_SLEEPOK, NULL, \ 685 "Sleeping in \"%s\"", __func__); \ 686 } while (0) 687 #else 688 #define MBUF_CHECKSLEEP(how) 689 #endif 690 691 /* 692 * Network buffer allocation API 693 * 694 * The rest of it is defined in kern/kern_mbuf.c 695 */ 696 extern uma_zone_t zone_mbuf; 697 extern uma_zone_t zone_clust; 698 extern uma_zone_t zone_pack; 699 extern uma_zone_t zone_jumbop; 700 extern uma_zone_t zone_jumbo9; 701 extern uma_zone_t zone_jumbo16; 702 extern uma_zone_t zone_extpgs; 703 704 void mb_dupcl(struct mbuf *, struct mbuf *); 705 void mb_free_ext(struct mbuf *); 706 void mb_free_mext_pgs(struct mbuf *); 707 struct mbuf *mb_alloc_ext_pgs(int, bool, m_ext_free_t); 708 int mb_unmapped_compress(struct mbuf *m); 709 struct mbuf *mb_unmapped_to_ext(struct mbuf *m); 710 void mb_free_notready(struct mbuf *m, int count); 711 void m_adj(struct mbuf *, int); 712 int m_apply(struct mbuf *, int, int, 713 int (*)(void *, void *, u_int), void *); 714 int m_append(struct mbuf *, int, c_caddr_t); 715 void m_cat(struct mbuf *, struct mbuf *); 716 void m_catpkt(struct mbuf *, struct mbuf *); 717 int m_clget(struct mbuf *m, int how); 718 void *m_cljget(struct mbuf *m, int how, int size); 719 struct mbuf *m_collapse(struct mbuf *, int, int); 720 void m_copyback(struct mbuf *, int, int, c_caddr_t); 721 void m_copydata(const struct mbuf *, int, int, caddr_t); 722 struct mbuf *m_copym(struct mbuf *, int, int, int); 723 struct mbuf *m_copypacket(struct mbuf *, int); 724 void m_copy_pkthdr(struct mbuf *, struct mbuf *); 725 struct mbuf *m_copyup(struct mbuf *, int, int); 726 struct mbuf *m_defrag(struct mbuf *, int); 727 void m_demote_pkthdr(struct mbuf *); 728 void m_demote(struct mbuf *, int, int); 729 struct mbuf *m_devget(char *, int, int, struct ifnet *, 730 void (*)(char *, caddr_t, u_int)); 731 void m_dispose_extcontrolm(struct mbuf *m); 732 struct mbuf *m_dup(const struct mbuf *, int); 733 int m_dup_pkthdr(struct mbuf *, const struct mbuf *, int); 734 void m_extadd(struct mbuf *, char *, u_int, m_ext_free_t, 735 void *, void *, int, int); 736 u_int m_fixhdr(struct mbuf *); 737 struct mbuf *m_fragment(struct mbuf *, int, int); 738 void m_freem(struct mbuf *); 739 struct mbuf *m_get2(int, int, short, int); 740 struct mbuf *m_getjcl(int, short, int, int); 741 struct mbuf *m_getm2(struct mbuf *, int, int, short, int); 742 struct mbuf *m_getptr(struct mbuf *, int, int *); 743 u_int m_length(struct mbuf *, struct mbuf **); 744 int m_mbuftouio(struct uio *, const struct mbuf *, int); 745 int m_unmappedtouio(const struct mbuf *, int, struct uio *, int); 746 void m_move_pkthdr(struct mbuf *, struct mbuf *); 747 int m_pkthdr_init(struct mbuf *, int); 748 struct mbuf *m_prepend(struct mbuf *, int, int); 749 void m_print(const struct mbuf *, int); 750 struct mbuf *m_pulldown(struct mbuf *, int, int, int *); 751 struct mbuf *m_pullup(struct mbuf *, int); 752 int m_sanity(struct mbuf *, int); 753 struct mbuf *m_split(struct mbuf *, int, int); 754 struct mbuf *m_uiotombuf(struct uio *, int, int, int, int); 755 struct mbuf *m_unshare(struct mbuf *, int); 756 void m_snd_tag_init(struct m_snd_tag *, struct ifnet *); 757 void m_snd_tag_destroy(struct m_snd_tag *); 758 759 static __inline int 760 m_gettype(int size) 761 { 762 int type; 763 764 switch (size) { 765 case MSIZE: 766 type = EXT_MBUF; 767 break; 768 case MCLBYTES: 769 type = EXT_CLUSTER; 770 break; 771 #if MJUMPAGESIZE != MCLBYTES 772 case MJUMPAGESIZE: 773 type = EXT_JUMBOP; 774 break; 775 #endif 776 case MJUM9BYTES: 777 type = EXT_JUMBO9; 778 break; 779 case MJUM16BYTES: 780 type = EXT_JUMBO16; 781 break; 782 default: 783 panic("%s: invalid cluster size %d", __func__, size); 784 } 785 786 return (type); 787 } 788 789 /* 790 * Associated an external reference counted buffer with an mbuf. 791 */ 792 static __inline void 793 m_extaddref(struct mbuf *m, char *buf, u_int size, u_int *ref_cnt, 794 m_ext_free_t freef, void *arg1, void *arg2) 795 { 796 797 KASSERT(ref_cnt != NULL, ("%s: ref_cnt not provided", __func__)); 798 799 atomic_add_int(ref_cnt, 1); 800 m->m_flags |= M_EXT; 801 m->m_ext.ext_buf = buf; 802 m->m_ext.ext_cnt = ref_cnt; 803 m->m_data = m->m_ext.ext_buf; 804 m->m_ext.ext_size = size; 805 m->m_ext.ext_free = freef; 806 m->m_ext.ext_arg1 = arg1; 807 m->m_ext.ext_arg2 = arg2; 808 m->m_ext.ext_type = EXT_EXTREF; 809 m->m_ext.ext_flags = 0; 810 } 811 812 static __inline uma_zone_t 813 m_getzone(int size) 814 { 815 uma_zone_t zone; 816 817 switch (size) { 818 case MCLBYTES: 819 zone = zone_clust; 820 break; 821 #if MJUMPAGESIZE != MCLBYTES 822 case MJUMPAGESIZE: 823 zone = zone_jumbop; 824 break; 825 #endif 826 case MJUM9BYTES: 827 zone = zone_jumbo9; 828 break; 829 case MJUM16BYTES: 830 zone = zone_jumbo16; 831 break; 832 default: 833 panic("%s: invalid cluster size %d", __func__, size); 834 } 835 836 return (zone); 837 } 838 839 /* 840 * Initialize an mbuf with linear storage. 841 * 842 * Inline because the consumer text overhead will be roughly the same to 843 * initialize or call a function with this many parameters and M_PKTHDR 844 * should go away with constant propagation for !MGETHDR. 845 */ 846 static __inline int 847 m_init(struct mbuf *m, int how, short type, int flags) 848 { 849 int error; 850 851 m->m_next = NULL; 852 m->m_nextpkt = NULL; 853 m->m_data = m->m_dat; 854 m->m_len = 0; 855 m->m_flags = flags; 856 m->m_type = type; 857 if (flags & M_PKTHDR) 858 error = m_pkthdr_init(m, how); 859 else 860 error = 0; 861 862 MBUF_PROBE5(m__init, m, how, type, flags, error); 863 return (error); 864 } 865 866 static __inline struct mbuf * 867 m_get(int how, short type) 868 { 869 struct mbuf *m; 870 struct mb_args args; 871 872 args.flags = 0; 873 args.type = type; 874 m = uma_zalloc_arg(zone_mbuf, &args, how); 875 MBUF_PROBE3(m__get, how, type, m); 876 return (m); 877 } 878 879 static __inline struct mbuf * 880 m_gethdr(int how, short type) 881 { 882 struct mbuf *m; 883 struct mb_args args; 884 885 args.flags = M_PKTHDR; 886 args.type = type; 887 m = uma_zalloc_arg(zone_mbuf, &args, how); 888 MBUF_PROBE3(m__gethdr, how, type, m); 889 return (m); 890 } 891 892 static __inline struct mbuf * 893 m_getcl(int how, short type, int flags) 894 { 895 struct mbuf *m; 896 struct mb_args args; 897 898 args.flags = flags; 899 args.type = type; 900 m = uma_zalloc_arg(zone_pack, &args, how); 901 MBUF_PROBE4(m__getcl, how, type, flags, m); 902 return (m); 903 } 904 905 /* 906 * XXX: m_cljset() is a dangerous API. One must attach only a new, 907 * unreferenced cluster to an mbuf(9). It is not possible to assert 908 * that, so care can be taken only by users of the API. 909 */ 910 static __inline void 911 m_cljset(struct mbuf *m, void *cl, int type) 912 { 913 int size; 914 915 switch (type) { 916 case EXT_CLUSTER: 917 size = MCLBYTES; 918 break; 919 #if MJUMPAGESIZE != MCLBYTES 920 case EXT_JUMBOP: 921 size = MJUMPAGESIZE; 922 break; 923 #endif 924 case EXT_JUMBO9: 925 size = MJUM9BYTES; 926 break; 927 case EXT_JUMBO16: 928 size = MJUM16BYTES; 929 break; 930 default: 931 panic("%s: unknown cluster type %d", __func__, type); 932 break; 933 } 934 935 m->m_data = m->m_ext.ext_buf = cl; 936 m->m_ext.ext_free = m->m_ext.ext_arg1 = m->m_ext.ext_arg2 = NULL; 937 m->m_ext.ext_size = size; 938 m->m_ext.ext_type = type; 939 m->m_ext.ext_flags = EXT_FLAG_EMBREF; 940 m->m_ext.ext_count = 1; 941 m->m_flags |= M_EXT; 942 MBUF_PROBE3(m__cljset, m, cl, type); 943 } 944 945 static __inline void 946 m_chtype(struct mbuf *m, short new_type) 947 { 948 949 m->m_type = new_type; 950 } 951 952 static __inline void 953 m_clrprotoflags(struct mbuf *m) 954 { 955 956 while (m) { 957 m->m_flags &= ~M_PROTOFLAGS; 958 m = m->m_next; 959 } 960 } 961 962 static __inline struct mbuf * 963 m_last(struct mbuf *m) 964 { 965 966 while (m->m_next) 967 m = m->m_next; 968 return (m); 969 } 970 971 static inline u_int 972 m_extrefcnt(struct mbuf *m) 973 { 974 975 KASSERT(m->m_flags & M_EXT, ("%s: M_EXT missing", __func__)); 976 977 return ((m->m_ext.ext_flags & EXT_FLAG_EMBREF) ? m->m_ext.ext_count : 978 *m->m_ext.ext_cnt); 979 } 980 981 /* 982 * mbuf, cluster, and external object allocation macros (for compatibility 983 * purposes). 984 */ 985 #define M_MOVE_PKTHDR(to, from) m_move_pkthdr((to), (from)) 986 #define MGET(m, how, type) ((m) = m_get((how), (type))) 987 #define MGETHDR(m, how, type) ((m) = m_gethdr((how), (type))) 988 #define MCLGET(m, how) m_clget((m), (how)) 989 #define MEXTADD(m, buf, size, free, arg1, arg2, flags, type) \ 990 m_extadd((m), (char *)(buf), (size), (free), (arg1), (arg2), \ 991 (flags), (type)) 992 #define m_getm(m, len, how, type) \ 993 m_getm2((m), (len), (how), (type), M_PKTHDR) 994 995 /* 996 * Evaluate TRUE if it's safe to write to the mbuf m's data region (this can 997 * be both the local data payload, or an external buffer area, depending on 998 * whether M_EXT is set). 999 */ 1000 #define M_WRITABLE(m) (((m)->m_flags & (M_RDONLY | M_NOMAP)) == 0 && \ 1001 (!(((m)->m_flags & M_EXT)) || \ 1002 (m_extrefcnt(m) == 1))) 1003 1004 /* Check if the supplied mbuf has a packet header, or else panic. */ 1005 #define M_ASSERTPKTHDR(m) \ 1006 KASSERT((m) != NULL && (m)->m_flags & M_PKTHDR, \ 1007 ("%s: no mbuf packet header!", __func__)) 1008 1009 /* 1010 * Ensure that the supplied mbuf is a valid, non-free mbuf. 1011 * 1012 * XXX: Broken at the moment. Need some UMA magic to make it work again. 1013 */ 1014 #define M_ASSERTVALID(m) \ 1015 KASSERT((((struct mbuf *)m)->m_flags & 0) == 0, \ 1016 ("%s: attempted use of a free mbuf!", __func__)) 1017 1018 /* 1019 * Return the address of the start of the buffer associated with an mbuf, 1020 * handling external storage, packet-header mbufs, and regular data mbufs. 1021 */ 1022 #define M_START(m) \ 1023 (((m)->m_flags & M_NOMAP) ? NULL : \ 1024 ((m)->m_flags & M_EXT) ? (m)->m_ext.ext_buf : \ 1025 ((m)->m_flags & M_PKTHDR) ? &(m)->m_pktdat[0] : \ 1026 &(m)->m_dat[0]) 1027 1028 /* 1029 * Return the size of the buffer associated with an mbuf, handling external 1030 * storage, packet-header mbufs, and regular data mbufs. 1031 */ 1032 #define M_SIZE(m) \ 1033 (((m)->m_flags & M_EXT) ? (m)->m_ext.ext_size : \ 1034 ((m)->m_flags & M_PKTHDR) ? MHLEN : \ 1035 MLEN) 1036 1037 /* 1038 * Set the m_data pointer of a newly allocated mbuf to place an object of the 1039 * specified size at the end of the mbuf, longword aligned. 1040 * 1041 * NB: Historically, we had M_ALIGN(), MH_ALIGN(), and MEXT_ALIGN() as 1042 * separate macros, each asserting that it was called at the proper moment. 1043 * This required callers to themselves test the storage type and call the 1044 * right one. Rather than require callers to be aware of those layout 1045 * decisions, we centralize here. 1046 */ 1047 static __inline void 1048 m_align(struct mbuf *m, int len) 1049 { 1050 #ifdef INVARIANTS 1051 const char *msg = "%s: not a virgin mbuf"; 1052 #endif 1053 int adjust; 1054 1055 KASSERT(m->m_data == M_START(m), (msg, __func__)); 1056 1057 adjust = M_SIZE(m) - len; 1058 m->m_data += adjust &~ (sizeof(long)-1); 1059 } 1060 1061 #define M_ALIGN(m, len) m_align(m, len) 1062 #define MH_ALIGN(m, len) m_align(m, len) 1063 #define MEXT_ALIGN(m, len) m_align(m, len) 1064 1065 /* 1066 * Compute the amount of space available before the current start of data in 1067 * an mbuf. 1068 * 1069 * The M_WRITABLE() is a temporary, conservative safety measure: the burden 1070 * of checking writability of the mbuf data area rests solely with the caller. 1071 * 1072 * NB: In previous versions, M_LEADINGSPACE() would only check M_WRITABLE() 1073 * for mbufs with external storage. We now allow mbuf-embedded data to be 1074 * read-only as well. 1075 */ 1076 #define M_LEADINGSPACE(m) \ 1077 (M_WRITABLE(m) ? ((m)->m_data - M_START(m)) : 0) 1078 1079 /* 1080 * Compute the amount of space available after the end of data in an mbuf. 1081 * 1082 * The M_WRITABLE() is a temporary, conservative safety measure: the burden 1083 * of checking writability of the mbuf data area rests solely with the caller. 1084 * 1085 * NB: In previous versions, M_TRAILINGSPACE() would only check M_WRITABLE() 1086 * for mbufs with external storage. We now allow mbuf-embedded data to be 1087 * read-only as well. 1088 */ 1089 #define M_TRAILINGSPACE(m) \ 1090 (M_WRITABLE(m) ? \ 1091 ((M_START(m) + M_SIZE(m)) - ((m)->m_data + (m)->m_len)) : 0) 1092 1093 /* 1094 * Arrange to prepend space of size plen to mbuf m. If a new mbuf must be 1095 * allocated, how specifies whether to wait. If the allocation fails, the 1096 * original mbuf chain is freed and m is set to NULL. 1097 */ 1098 #define M_PREPEND(m, plen, how) do { \ 1099 struct mbuf **_mmp = &(m); \ 1100 struct mbuf *_mm = *_mmp; \ 1101 int _mplen = (plen); \ 1102 int __mhow = (how); \ 1103 \ 1104 MBUF_CHECKSLEEP(how); \ 1105 if (M_LEADINGSPACE(_mm) >= _mplen) { \ 1106 _mm->m_data -= _mplen; \ 1107 _mm->m_len += _mplen; \ 1108 } else \ 1109 _mm = m_prepend(_mm, _mplen, __mhow); \ 1110 if (_mm != NULL && _mm->m_flags & M_PKTHDR) \ 1111 _mm->m_pkthdr.len += _mplen; \ 1112 *_mmp = _mm; \ 1113 } while (0) 1114 1115 /* 1116 * Change mbuf to new type. This is a relatively expensive operation and 1117 * should be avoided. 1118 */ 1119 #define MCHTYPE(m, t) m_chtype((m), (t)) 1120 1121 /* Return the rcvif of a packet header. */ 1122 static __inline struct ifnet * 1123 m_rcvif(struct mbuf *m) 1124 { 1125 1126 M_ASSERTPKTHDR(m); 1127 if (m->m_pkthdr.csum_flags & CSUM_SND_TAG) 1128 return (NULL); 1129 return (m->m_pkthdr.rcvif); 1130 } 1131 1132 /* Length to m_copy to copy all. */ 1133 #define M_COPYALL 1000000000 1134 1135 extern int max_datalen; /* MHLEN - max_hdr */ 1136 extern int max_hdr; /* Largest link + protocol header */ 1137 extern int max_linkhdr; /* Largest link-level header */ 1138 extern int max_protohdr; /* Largest protocol header */ 1139 extern int nmbclusters; /* Maximum number of clusters */ 1140 extern bool mb_use_ext_pgs; /* Use ext_pgs for sendfile */ 1141 1142 /*- 1143 * Network packets may have annotations attached by affixing a list of 1144 * "packet tags" to the pkthdr structure. Packet tags are dynamically 1145 * allocated semi-opaque data structures that have a fixed header 1146 * (struct m_tag) that specifies the size of the memory block and a 1147 * <cookie,type> pair that identifies it. The cookie is a 32-bit unique 1148 * unsigned value used to identify a module or ABI. By convention this value 1149 * is chosen as the date+time that the module is created, expressed as the 1150 * number of seconds since the epoch (e.g., using date -u +'%s'). The type 1151 * value is an ABI/module-specific value that identifies a particular 1152 * annotation and is private to the module. For compatibility with systems 1153 * like OpenBSD that define packet tags w/o an ABI/module cookie, the value 1154 * PACKET_ABI_COMPAT is used to implement m_tag_get and m_tag_find 1155 * compatibility shim functions and several tag types are defined below. 1156 * Users that do not require compatibility should use a private cookie value 1157 * so that packet tag-related definitions can be maintained privately. 1158 * 1159 * Note that the packet tag returned by m_tag_alloc has the default memory 1160 * alignment implemented by malloc. To reference private data one can use a 1161 * construct like: 1162 * 1163 * struct m_tag *mtag = m_tag_alloc(...); 1164 * struct foo *p = (struct foo *)(mtag+1); 1165 * 1166 * if the alignment of struct m_tag is sufficient for referencing members of 1167 * struct foo. Otherwise it is necessary to embed struct m_tag within the 1168 * private data structure to insure proper alignment; e.g., 1169 * 1170 * struct foo { 1171 * struct m_tag tag; 1172 * ... 1173 * }; 1174 * struct foo *p = (struct foo *) m_tag_alloc(...); 1175 * struct m_tag *mtag = &p->tag; 1176 */ 1177 1178 /* 1179 * Persistent tags stay with an mbuf until the mbuf is reclaimed. Otherwise 1180 * tags are expected to ``vanish'' when they pass through a network 1181 * interface. For most interfaces this happens normally as the tags are 1182 * reclaimed when the mbuf is free'd. However in some special cases 1183 * reclaiming must be done manually. An example is packets that pass through 1184 * the loopback interface. Also, one must be careful to do this when 1185 * ``turning around'' packets (e.g., icmp_reflect). 1186 * 1187 * To mark a tag persistent bit-or this flag in when defining the tag id. 1188 * The tag will then be treated as described above. 1189 */ 1190 #define MTAG_PERSISTENT 0x800 1191 1192 #define PACKET_TAG_NONE 0 /* Nadda */ 1193 1194 /* Packet tags for use with PACKET_ABI_COMPAT. */ 1195 #define PACKET_TAG_IPSEC_IN_DONE 1 /* IPsec applied, in */ 1196 #define PACKET_TAG_IPSEC_OUT_DONE 2 /* IPsec applied, out */ 1197 #define PACKET_TAG_IPSEC_IN_CRYPTO_DONE 3 /* NIC IPsec crypto done */ 1198 #define PACKET_TAG_IPSEC_OUT_CRYPTO_NEEDED 4 /* NIC IPsec crypto req'ed */ 1199 #define PACKET_TAG_IPSEC_IN_COULD_DO_CRYPTO 5 /* NIC notifies IPsec */ 1200 #define PACKET_TAG_IPSEC_PENDING_TDB 6 /* Reminder to do IPsec */ 1201 #define PACKET_TAG_BRIDGE 7 /* Bridge processing done */ 1202 #define PACKET_TAG_GIF 8 /* GIF processing done */ 1203 #define PACKET_TAG_GRE 9 /* GRE processing done */ 1204 #define PACKET_TAG_IN_PACKET_CHECKSUM 10 /* NIC checksumming done */ 1205 #define PACKET_TAG_ENCAP 11 /* Encap. processing */ 1206 #define PACKET_TAG_IPSEC_SOCKET 12 /* IPSEC socket ref */ 1207 #define PACKET_TAG_IPSEC_HISTORY 13 /* IPSEC history */ 1208 #define PACKET_TAG_IPV6_INPUT 14 /* IPV6 input processing */ 1209 #define PACKET_TAG_DUMMYNET 15 /* dummynet info */ 1210 #define PACKET_TAG_DIVERT 17 /* divert info */ 1211 #define PACKET_TAG_IPFORWARD 18 /* ipforward info */ 1212 #define PACKET_TAG_MACLABEL (19 | MTAG_PERSISTENT) /* MAC label */ 1213 #define PACKET_TAG_PF (21 | MTAG_PERSISTENT) /* PF/ALTQ information */ 1214 #define PACKET_TAG_RTSOCKFAM 25 /* rtsock sa family */ 1215 #define PACKET_TAG_IPOPTIONS 27 /* Saved IP options */ 1216 #define PACKET_TAG_CARP 28 /* CARP info */ 1217 #define PACKET_TAG_IPSEC_NAT_T_PORTS 29 /* two uint16_t */ 1218 #define PACKET_TAG_ND_OUTGOING 30 /* ND outgoing */ 1219 1220 /* Specific cookies and tags. */ 1221 1222 /* Packet tag routines. */ 1223 struct m_tag *m_tag_alloc(u_int32_t, int, int, int); 1224 void m_tag_delete(struct mbuf *, struct m_tag *); 1225 void m_tag_delete_chain(struct mbuf *, struct m_tag *); 1226 void m_tag_free_default(struct m_tag *); 1227 struct m_tag *m_tag_locate(struct mbuf *, u_int32_t, int, struct m_tag *); 1228 struct m_tag *m_tag_copy(struct m_tag *, int); 1229 int m_tag_copy_chain(struct mbuf *, const struct mbuf *, int); 1230 void m_tag_delete_nonpersistent(struct mbuf *); 1231 1232 /* 1233 * Initialize the list of tags associated with an mbuf. 1234 */ 1235 static __inline void 1236 m_tag_init(struct mbuf *m) 1237 { 1238 1239 SLIST_INIT(&m->m_pkthdr.tags); 1240 } 1241 1242 /* 1243 * Set up the contents of a tag. Note that this does not fill in the free 1244 * method; the caller is expected to do that. 1245 * 1246 * XXX probably should be called m_tag_init, but that was already taken. 1247 */ 1248 static __inline void 1249 m_tag_setup(struct m_tag *t, u_int32_t cookie, int type, int len) 1250 { 1251 1252 t->m_tag_id = type; 1253 t->m_tag_len = len; 1254 t->m_tag_cookie = cookie; 1255 } 1256 1257 /* 1258 * Reclaim resources associated with a tag. 1259 */ 1260 static __inline void 1261 m_tag_free(struct m_tag *t) 1262 { 1263 1264 (*t->m_tag_free)(t); 1265 } 1266 1267 /* 1268 * Return the first tag associated with an mbuf. 1269 */ 1270 static __inline struct m_tag * 1271 m_tag_first(struct mbuf *m) 1272 { 1273 1274 return (SLIST_FIRST(&m->m_pkthdr.tags)); 1275 } 1276 1277 /* 1278 * Return the next tag in the list of tags associated with an mbuf. 1279 */ 1280 static __inline struct m_tag * 1281 m_tag_next(struct mbuf *m __unused, struct m_tag *t) 1282 { 1283 1284 return (SLIST_NEXT(t, m_tag_link)); 1285 } 1286 1287 /* 1288 * Prepend a tag to the list of tags associated with an mbuf. 1289 */ 1290 static __inline void 1291 m_tag_prepend(struct mbuf *m, struct m_tag *t) 1292 { 1293 1294 SLIST_INSERT_HEAD(&m->m_pkthdr.tags, t, m_tag_link); 1295 } 1296 1297 /* 1298 * Unlink a tag from the list of tags associated with an mbuf. 1299 */ 1300 static __inline void 1301 m_tag_unlink(struct mbuf *m, struct m_tag *t) 1302 { 1303 1304 SLIST_REMOVE(&m->m_pkthdr.tags, t, m_tag, m_tag_link); 1305 } 1306 1307 /* These are for OpenBSD compatibility. */ 1308 #define MTAG_ABI_COMPAT 0 /* compatibility ABI */ 1309 1310 static __inline struct m_tag * 1311 m_tag_get(int type, int length, int wait) 1312 { 1313 return (m_tag_alloc(MTAG_ABI_COMPAT, type, length, wait)); 1314 } 1315 1316 static __inline struct m_tag * 1317 m_tag_find(struct mbuf *m, int type, struct m_tag *start) 1318 { 1319 return (SLIST_EMPTY(&m->m_pkthdr.tags) ? (struct m_tag *)NULL : 1320 m_tag_locate(m, MTAG_ABI_COMPAT, type, start)); 1321 } 1322 1323 static inline struct m_snd_tag * 1324 m_snd_tag_ref(struct m_snd_tag *mst) 1325 { 1326 1327 refcount_acquire(&mst->refcount); 1328 return (mst); 1329 } 1330 1331 static inline void 1332 m_snd_tag_rele(struct m_snd_tag *mst) 1333 { 1334 1335 if (refcount_release(&mst->refcount)) 1336 m_snd_tag_destroy(mst); 1337 } 1338 1339 static __inline struct mbuf * 1340 m_free(struct mbuf *m) 1341 { 1342 struct mbuf *n = m->m_next; 1343 1344 MBUF_PROBE1(m__free, m); 1345 if ((m->m_flags & (M_PKTHDR|M_NOFREE)) == (M_PKTHDR|M_NOFREE)) 1346 m_tag_delete_chain(m, NULL); 1347 if (m->m_flags & M_PKTHDR && m->m_pkthdr.csum_flags & CSUM_SND_TAG) 1348 m_snd_tag_rele(m->m_pkthdr.snd_tag); 1349 if (m->m_flags & M_EXT) 1350 mb_free_ext(m); 1351 else if ((m->m_flags & M_NOFREE) == 0) 1352 uma_zfree(zone_mbuf, m); 1353 return (n); 1354 } 1355 1356 static __inline int 1357 rt_m_getfib(struct mbuf *m) 1358 { 1359 KASSERT(m->m_flags & M_PKTHDR , ("Attempt to get FIB from non header mbuf.")); 1360 return (m->m_pkthdr.fibnum); 1361 } 1362 1363 #define M_GETFIB(_m) rt_m_getfib(_m) 1364 1365 #define M_SETFIB(_m, _fib) do { \ 1366 KASSERT((_m)->m_flags & M_PKTHDR, ("Attempt to set FIB on non header mbuf.")); \ 1367 ((_m)->m_pkthdr.fibnum) = (_fib); \ 1368 } while (0) 1369 1370 /* flags passed as first argument for "m_ether_tcpip_hash()" */ 1371 #define MBUF_HASHFLAG_L2 (1 << 2) 1372 #define MBUF_HASHFLAG_L3 (1 << 3) 1373 #define MBUF_HASHFLAG_L4 (1 << 4) 1374 1375 /* mbuf hashing helper routines */ 1376 uint32_t m_ether_tcpip_hash_init(void); 1377 uint32_t m_ether_tcpip_hash(const uint32_t, const struct mbuf *, const uint32_t); 1378 1379 #ifdef MBUF_PROFILING 1380 void m_profile(struct mbuf *m); 1381 #define M_PROFILE(m) m_profile(m) 1382 #else 1383 #define M_PROFILE(m) 1384 #endif 1385 1386 struct mbufq { 1387 STAILQ_HEAD(, mbuf) mq_head; 1388 int mq_len; 1389 int mq_maxlen; 1390 }; 1391 1392 static inline void 1393 mbufq_init(struct mbufq *mq, int maxlen) 1394 { 1395 1396 STAILQ_INIT(&mq->mq_head); 1397 mq->mq_maxlen = maxlen; 1398 mq->mq_len = 0; 1399 } 1400 1401 static inline struct mbuf * 1402 mbufq_flush(struct mbufq *mq) 1403 { 1404 struct mbuf *m; 1405 1406 m = STAILQ_FIRST(&mq->mq_head); 1407 STAILQ_INIT(&mq->mq_head); 1408 mq->mq_len = 0; 1409 return (m); 1410 } 1411 1412 static inline void 1413 mbufq_drain(struct mbufq *mq) 1414 { 1415 struct mbuf *m, *n; 1416 1417 n = mbufq_flush(mq); 1418 while ((m = n) != NULL) { 1419 n = STAILQ_NEXT(m, m_stailqpkt); 1420 m_freem(m); 1421 } 1422 } 1423 1424 static inline struct mbuf * 1425 mbufq_first(const struct mbufq *mq) 1426 { 1427 1428 return (STAILQ_FIRST(&mq->mq_head)); 1429 } 1430 1431 static inline struct mbuf * 1432 mbufq_last(const struct mbufq *mq) 1433 { 1434 1435 return (STAILQ_LAST(&mq->mq_head, mbuf, m_stailqpkt)); 1436 } 1437 1438 static inline int 1439 mbufq_full(const struct mbufq *mq) 1440 { 1441 1442 return (mq->mq_maxlen > 0 && mq->mq_len >= mq->mq_maxlen); 1443 } 1444 1445 static inline int 1446 mbufq_len(const struct mbufq *mq) 1447 { 1448 1449 return (mq->mq_len); 1450 } 1451 1452 static inline int 1453 mbufq_enqueue(struct mbufq *mq, struct mbuf *m) 1454 { 1455 1456 if (mbufq_full(mq)) 1457 return (ENOBUFS); 1458 STAILQ_INSERT_TAIL(&mq->mq_head, m, m_stailqpkt); 1459 mq->mq_len++; 1460 return (0); 1461 } 1462 1463 static inline struct mbuf * 1464 mbufq_dequeue(struct mbufq *mq) 1465 { 1466 struct mbuf *m; 1467 1468 m = STAILQ_FIRST(&mq->mq_head); 1469 if (m) { 1470 STAILQ_REMOVE_HEAD(&mq->mq_head, m_stailqpkt); 1471 m->m_nextpkt = NULL; 1472 mq->mq_len--; 1473 } 1474 return (m); 1475 } 1476 1477 static inline void 1478 mbufq_prepend(struct mbufq *mq, struct mbuf *m) 1479 { 1480 1481 STAILQ_INSERT_HEAD(&mq->mq_head, m, m_stailqpkt); 1482 mq->mq_len++; 1483 } 1484 1485 /* 1486 * Note: this doesn't enforce the maximum list size for dst. 1487 */ 1488 static inline void 1489 mbufq_concat(struct mbufq *mq_dst, struct mbufq *mq_src) 1490 { 1491 1492 mq_dst->mq_len += mq_src->mq_len; 1493 STAILQ_CONCAT(&mq_dst->mq_head, &mq_src->mq_head); 1494 mq_src->mq_len = 0; 1495 } 1496 1497 #ifdef _SYS_TIMESPEC_H_ 1498 static inline void 1499 mbuf_tstmp2timespec(struct mbuf *m, struct timespec *ts) 1500 { 1501 1502 KASSERT((m->m_flags & M_PKTHDR) != 0, ("mbuf %p no M_PKTHDR", m)); 1503 KASSERT((m->m_flags & (M_TSTMP|M_TSTMP_LRO)) != 0, ("mbuf %p no M_TSTMP or M_TSTMP_LRO", m)); 1504 ts->tv_sec = m->m_pkthdr.rcv_tstmp / 1000000000; 1505 ts->tv_nsec = m->m_pkthdr.rcv_tstmp % 1000000000; 1506 } 1507 #endif 1508 1509 #ifdef NETDUMP 1510 /* Invoked from the netdump client code. */ 1511 void netdump_mbuf_drain(void); 1512 void netdump_mbuf_dump(void); 1513 void netdump_mbuf_reinit(int nmbuf, int nclust, int clsize); 1514 #endif 1515 1516 static inline bool 1517 mbuf_has_tls_session(struct mbuf *m) 1518 { 1519 1520 if (m->m_flags & M_NOMAP) { 1521 MBUF_EXT_PGS_ASSERT(m); 1522 if (m->m_ext.ext_pgs->tls != NULL) { 1523 return (true); 1524 } 1525 } 1526 return (false); 1527 } 1528 1529 #endif /* _KERNEL */ 1530 #endif /* !_SYS_MBUF_H_ */ 1531