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_raw); 66 SDT_PROBE_DECLARE(sdt, , , m__gethdr); 67 SDT_PROBE_DECLARE(sdt, , , m__get_raw); 68 SDT_PROBE_DECLARE(sdt, , , m__get); 69 SDT_PROBE_DECLARE(sdt, , , m__getcl); 70 SDT_PROBE_DECLARE(sdt, , , m__getjcl); 71 SDT_PROBE_DECLARE(sdt, , , m__clget); 72 SDT_PROBE_DECLARE(sdt, , , m__cljget); 73 SDT_PROBE_DECLARE(sdt, , , m__cljset); 74 SDT_PROBE_DECLARE(sdt, , , m__free); 75 SDT_PROBE_DECLARE(sdt, , , m__freem); 76 77 #endif /* _KERNEL */ 78 79 /* 80 * Mbufs are of a single size, MSIZE (sys/param.h), which includes overhead. 81 * An mbuf may add a single "mbuf cluster" of size MCLBYTES (also in 82 * sys/param.h), which has no additional overhead and is used instead of the 83 * internal data area; this is done when at least MINCLSIZE of data must be 84 * stored. Additionally, it is possible to allocate a separate buffer 85 * externally and attach it to the mbuf in a way similar to that of mbuf 86 * clusters. 87 * 88 * NB: These calculation do not take actual compiler-induced alignment and 89 * padding inside the complete struct mbuf into account. Appropriate 90 * attention is required when changing members of struct mbuf. 91 * 92 * MLEN is data length in a normal mbuf. 93 * MHLEN is data length in an mbuf with pktheader. 94 * MINCLSIZE is a smallest amount of data that should be put into cluster. 95 * 96 * Compile-time assertions in uipc_mbuf.c test these values to ensure that 97 * they are sensible. 98 */ 99 struct mbuf; 100 #define MHSIZE offsetof(struct mbuf, m_dat) 101 #define MPKTHSIZE offsetof(struct mbuf, m_pktdat) 102 #define MLEN ((int)(MSIZE - MHSIZE)) 103 #define MHLEN ((int)(MSIZE - MPKTHSIZE)) 104 #define MINCLSIZE (MHLEN + 1) 105 #define M_NODOM 255 106 107 #ifdef _KERNEL 108 /*- 109 * Macro for type conversion: convert mbuf pointer to data pointer of correct 110 * type: 111 * 112 * mtod(m, t) -- Convert mbuf pointer to data pointer of correct type. 113 * mtodo(m, o) -- Same as above but with offset 'o' into data. 114 */ 115 #define mtod(m, t) ((t)((m)->m_data)) 116 #define mtodo(m, o) ((void *)(((m)->m_data) + (o))) 117 118 /* 119 * Argument structure passed to UMA routines during mbuf and packet 120 * allocations. 121 */ 122 struct mb_args { 123 int flags; /* Flags for mbuf being allocated */ 124 short type; /* Type of mbuf being allocated */ 125 }; 126 #endif /* _KERNEL */ 127 128 /* 129 * Packet tag structure (see below for details). 130 */ 131 struct m_tag { 132 SLIST_ENTRY(m_tag) m_tag_link; /* List of packet tags */ 133 u_int16_t m_tag_id; /* Tag ID */ 134 u_int16_t m_tag_len; /* Length of data */ 135 u_int32_t m_tag_cookie; /* ABI/Module ID */ 136 void (*m_tag_free)(struct m_tag *); 137 }; 138 139 /* 140 * Static network interface owned tag. 141 * Allocated through ifp->if_snd_tag_alloc(). 142 */ 143 struct if_snd_tag_sw; 144 145 struct m_snd_tag { 146 struct ifnet *ifp; /* network interface tag belongs to */ 147 const struct if_snd_tag_sw *sw; 148 volatile u_int refcount; 149 }; 150 151 /* 152 * Record/packet header in first mbuf of chain; valid only if M_PKTHDR is set. 153 * Size ILP32: 48 154 * LP64: 56 155 * Compile-time assertions in uipc_mbuf.c test these values to ensure that 156 * they are correct. 157 */ 158 struct pkthdr { 159 union { 160 struct m_snd_tag *snd_tag; /* send tag, if any */ 161 struct ifnet *rcvif; /* rcv interface */ 162 struct { 163 uint16_t rcvidx; /* rcv interface index ... */ 164 uint16_t rcvgen; /* ... and generation count */ 165 }; 166 }; 167 SLIST_HEAD(packet_tags, m_tag) tags; /* list of packet tags */ 168 int32_t len; /* total packet length */ 169 170 /* Layer crossing persistent information. */ 171 uint32_t flowid; /* packet's 4-tuple system */ 172 uint32_t csum_flags; /* checksum and offload features */ 173 uint16_t fibnum; /* this packet should use this fib */ 174 uint8_t numa_domain; /* NUMA domain of recvd pkt */ 175 uint8_t rsstype; /* hash type */ 176 union { 177 uint64_t rcv_tstmp; /* timestamp in ns */ 178 struct { 179 uint8_t l2hlen; /* layer 2 hdr len */ 180 uint8_t l3hlen; /* layer 3 hdr len */ 181 uint8_t l4hlen; /* layer 4 hdr len */ 182 uint8_t l5hlen; /* layer 5 hdr len */ 183 uint8_t inner_l2hlen; 184 uint8_t inner_l3hlen; 185 uint8_t inner_l4hlen; 186 uint8_t inner_l5hlen; 187 }; 188 }; 189 union { 190 uint8_t eight[8]; 191 uint16_t sixteen[4]; 192 uint32_t thirtytwo[2]; 193 uint64_t sixtyfour[1]; 194 uintptr_t unintptr[1]; 195 void *ptr; 196 } PH_per; 197 198 /* Layer specific non-persistent local storage for reassembly, etc. */ 199 union { 200 uint8_t eight[8]; 201 uint16_t sixteen[4]; 202 uint32_t thirtytwo[2]; 203 uint64_t sixtyfour[1]; 204 uintptr_t unintptr[1]; 205 void *ptr; 206 } PH_loc; 207 }; 208 #define ether_vtag PH_per.sixteen[0] 209 #define tcp_tun_port PH_per.sixteen[0] /* outbound */ 210 #define vt_nrecs PH_per.sixteen[0] /* mld and v6-ND */ 211 #define tso_segsz PH_per.sixteen[1] /* inbound after LRO */ 212 #define lro_nsegs tso_segsz /* inbound after LRO */ 213 #define csum_data PH_per.thirtytwo[1] /* inbound from hardware up */ 214 #define lro_tcp_d_len PH_loc.sixteen[0] /* inbound during LRO (no reassembly) */ 215 #define lro_tcp_d_csum PH_loc.sixteen[1] /* inbound during LRO (no reassembly) */ 216 #define lro_tcp_h_off PH_loc.sixteen[2] /* inbound during LRO (no reassembly) */ 217 #define lro_etype PH_loc.sixteen[3] /* inbound during LRO (no reassembly) */ 218 /* Note PH_loc is used during IP reassembly (all 8 bytes as a ptr) */ 219 220 /* 221 * TLS records for TLS 1.0-1.2 can have the following header lengths: 222 * - 5 (AES-CBC with implicit IV) 223 * - 21 (AES-CBC with explicit IV) 224 * - 13 (AES-GCM with 8 byte explicit IV) 225 */ 226 #define MBUF_PEXT_HDR_LEN 23 227 228 /* 229 * TLS records for TLS 1.0-1.2 can have the following maximum trailer 230 * lengths: 231 * - 16 (AES-GCM) 232 * - 36 (AES-CBC with SHA1 and up to 16 bytes of padding) 233 * - 48 (AES-CBC with SHA2-256 and up to 16 bytes of padding) 234 * - 64 (AES-CBC with SHA2-384 and up to 16 bytes of padding) 235 */ 236 #define MBUF_PEXT_TRAIL_LEN 64 237 238 #if defined(__LP64__) 239 #define MBUF_PEXT_MAX_PGS (40 / sizeof(vm_paddr_t)) 240 #else 241 #define MBUF_PEXT_MAX_PGS (72 / sizeof(vm_paddr_t)) 242 #endif 243 244 #define MBUF_PEXT_MAX_BYTES \ 245 (MBUF_PEXT_MAX_PGS * PAGE_SIZE + MBUF_PEXT_HDR_LEN + MBUF_PEXT_TRAIL_LEN) 246 247 struct ktls_session; 248 struct socket; 249 250 /* 251 * Description of external storage mapped into mbuf; valid only if M_EXT is 252 * set. 253 * Size ILP32: 28 254 * LP64: 48 255 * Compile-time assertions in uipc_mbuf.c test these values to ensure that 256 * they are correct. 257 */ 258 typedef void m_ext_free_t(struct mbuf *); 259 struct m_ext { 260 union { 261 /* 262 * If EXT_FLAG_EMBREF is set, then we use refcount in the 263 * mbuf, the 'ext_count' member. Otherwise, we have a 264 * shadow copy and we use pointer 'ext_cnt'. The original 265 * mbuf is responsible to carry the pointer to free routine 266 * and its arguments. They aren't copied into shadows in 267 * mb_dupcl() to avoid dereferencing next cachelines. 268 */ 269 volatile u_int ext_count; 270 volatile u_int *ext_cnt; 271 }; 272 uint32_t ext_size; /* size of buffer, for ext_free */ 273 uint32_t ext_type:8, /* type of external storage */ 274 ext_flags:24; /* external storage mbuf flags */ 275 union { 276 struct { 277 /* 278 * Regular M_EXT mbuf: 279 * o ext_buf always points to the external buffer. 280 * o ext_free (below) and two optional arguments 281 * ext_arg1 and ext_arg2 store the free context for 282 * the external storage. They are set only in the 283 * refcount carrying mbuf, the one with 284 * EXT_FLAG_EMBREF flag, with exclusion for 285 * EXT_EXTREF type, where the free context is copied 286 * into all mbufs that use same external storage. 287 */ 288 char *ext_buf; /* start of buffer */ 289 #define m_ext_copylen offsetof(struct m_ext, ext_arg2) 290 void *ext_arg2; 291 }; 292 struct { 293 /* 294 * Multi-page M_EXTPG mbuf: 295 * o extpg_pa - page vector. 296 * o extpg_trail and extpg_hdr - TLS trailer and 297 * header. 298 * Uses ext_free and may also use ext_arg1. 299 */ 300 vm_paddr_t extpg_pa[MBUF_PEXT_MAX_PGS]; 301 char extpg_trail[MBUF_PEXT_TRAIL_LEN]; 302 char extpg_hdr[MBUF_PEXT_HDR_LEN]; 303 /* Pretend these 3 fields are part of mbuf itself. */ 304 #define m_epg_pa m_ext.extpg_pa 305 #define m_epg_trail m_ext.extpg_trail 306 #define m_epg_hdr m_ext.extpg_hdr 307 #define m_epg_ext_copylen offsetof(struct m_ext, ext_free) 308 }; 309 }; 310 /* 311 * Free method and optional argument pointer, both 312 * used by M_EXT and M_EXTPG. 313 */ 314 m_ext_free_t *ext_free; 315 void *ext_arg1; 316 }; 317 318 /* 319 * The core of the mbuf object along with some shortcut defines for practical 320 * purposes. 321 */ 322 struct mbuf { 323 /* 324 * Header present at the beginning of every mbuf. 325 * Size ILP32: 24 326 * LP64: 32 327 * Compile-time assertions in uipc_mbuf.c test these values to ensure 328 * that they are correct. 329 */ 330 union { /* next buffer in chain */ 331 struct mbuf *m_next; 332 SLIST_ENTRY(mbuf) m_slist; 333 STAILQ_ENTRY(mbuf) m_stailq; 334 }; 335 union { /* next chain in queue/record */ 336 struct mbuf *m_nextpkt; 337 SLIST_ENTRY(mbuf) m_slistpkt; 338 STAILQ_ENTRY(mbuf) m_stailqpkt; 339 }; 340 caddr_t m_data; /* location of data */ 341 int32_t m_len; /* amount of data in this mbuf */ 342 uint32_t m_type:8, /* type of data in this mbuf */ 343 m_flags:24; /* flags; see below */ 344 #if !defined(__LP64__) 345 uint32_t m_pad; /* pad for 64bit alignment */ 346 #endif 347 348 /* 349 * A set of optional headers (packet header, external storage header) 350 * and internal data storage. Historically, these arrays were sized 351 * to MHLEN (space left after a packet header) and MLEN (space left 352 * after only a regular mbuf header); they are now variable size in 353 * order to support future work on variable-size mbufs. 354 */ 355 union { 356 struct { 357 union { 358 /* M_PKTHDR set. */ 359 struct pkthdr m_pkthdr; 360 361 /* M_EXTPG set. 362 * Multi-page M_EXTPG mbuf has its meta data 363 * split between the below anonymous structure 364 * and m_ext. It carries vector of pages, 365 * optional header and trailer char vectors 366 * and pointers to socket/TLS data. 367 */ 368 #define m_epg_startcopy m_epg_npgs 369 #define m_epg_endcopy m_epg_stailq 370 struct { 371 /* Overall count of pages and count of 372 * pages with I/O pending. */ 373 uint8_t m_epg_npgs; 374 uint8_t m_epg_nrdy; 375 /* TLS header and trailer lengths. 376 * The data itself resides in m_ext. */ 377 uint8_t m_epg_hdrlen; 378 uint8_t m_epg_trllen; 379 /* Offset into 1st page and length of 380 * data in the last page. */ 381 uint16_t m_epg_1st_off; 382 uint16_t m_epg_last_len; 383 uint8_t m_epg_flags; 384 #define EPG_FLAG_ANON 0x1 /* Data can be encrypted in place. */ 385 #define EPG_FLAG_2FREE 0x2 /* Scheduled for free. */ 386 uint8_t m_epg_record_type; 387 uint8_t __spare[2]; 388 int m_epg_enc_cnt; 389 struct ktls_session *m_epg_tls; 390 struct socket *m_epg_so; 391 uint64_t m_epg_seqno; 392 STAILQ_ENTRY(mbuf) m_epg_stailq; 393 }; 394 }; 395 union { 396 /* M_EXT or M_EXTPG set. */ 397 struct m_ext m_ext; 398 /* M_PKTHDR set, neither M_EXT nor M_EXTPG. */ 399 char m_pktdat[0]; 400 }; 401 }; 402 char m_dat[0]; /* !M_PKTHDR, !M_EXT */ 403 }; 404 }; 405 406 #ifdef _KERNEL 407 static inline int 408 m_epg_pagelen(const struct mbuf *m, int pidx, int pgoff) 409 { 410 411 KASSERT(pgoff == 0 || pidx == 0, 412 ("page %d with non-zero offset %d in %p", pidx, pgoff, m)); 413 414 if (pidx == m->m_epg_npgs - 1) { 415 return (m->m_epg_last_len); 416 } else { 417 return (PAGE_SIZE - pgoff); 418 } 419 } 420 421 #ifdef INVARIANTS 422 #define MCHECK(ex, msg) KASSERT((ex), \ 423 ("Multi page mbuf %p with " #msg " at %s:%d", \ 424 m, __FILE__, __LINE__)) 425 /* 426 * NB: This expects a non-empty buffer (npgs > 0 and 427 * last_pg_len > 0). 428 */ 429 #define MBUF_EXT_PGS_ASSERT_SANITY(m) do { \ 430 MCHECK(m->m_epg_npgs > 0, "no valid pages"); \ 431 MCHECK(m->m_epg_npgs <= nitems(m->m_epg_pa), \ 432 "too many pages"); \ 433 MCHECK(m->m_epg_nrdy <= m->m_epg_npgs, \ 434 "too many ready pages"); \ 435 MCHECK(m->m_epg_1st_off < PAGE_SIZE, \ 436 "too large page offset"); \ 437 MCHECK(m->m_epg_last_len > 0, "zero last page length"); \ 438 MCHECK(m->m_epg_last_len <= PAGE_SIZE, \ 439 "too large last page length"); \ 440 if (m->m_epg_npgs == 1) \ 441 MCHECK(m->m_epg_1st_off + \ 442 m->m_epg_last_len <= PAGE_SIZE, \ 443 "single page too large"); \ 444 MCHECK(m->m_epg_hdrlen <= sizeof(m->m_epg_hdr), \ 445 "too large header length"); \ 446 MCHECK(m->m_epg_trllen <= sizeof(m->m_epg_trail), \ 447 "too large header length"); \ 448 } while (0) 449 #else 450 #define MBUF_EXT_PGS_ASSERT_SANITY(m) do {} while (0) 451 #endif 452 #endif 453 454 /* 455 * mbuf flags of global significance and layer crossing. 456 * Those of only protocol/layer specific significance are to be mapped 457 * to M_PROTO[1-11] and cleared at layer handoff boundaries. 458 * NB: Limited to the lower 24 bits. 459 */ 460 #define M_EXT 0x00000001 /* has associated external storage */ 461 #define M_PKTHDR 0x00000002 /* start of record */ 462 #define M_EOR 0x00000004 /* end of record */ 463 #define M_RDONLY 0x00000008 /* associated data is marked read-only */ 464 #define M_BCAST 0x00000010 /* send/received as link-level broadcast */ 465 #define M_MCAST 0x00000020 /* send/received as link-level multicast */ 466 #define M_PROMISC 0x00000040 /* packet was not for us */ 467 #define M_VLANTAG 0x00000080 /* ether_vtag is valid */ 468 #define M_EXTPG 0x00000100 /* has array of unmapped pages and TLS */ 469 #define M_NOFREE 0x00000200 /* do not free mbuf, embedded in cluster */ 470 #define M_TSTMP 0x00000400 /* rcv_tstmp field is valid */ 471 #define M_TSTMP_HPREC 0x00000800 /* rcv_tstmp is high-prec, typically 472 hw-stamped on port (useful for IEEE 1588 473 and 802.1AS) */ 474 #define M_TSTMP_LRO 0x00001000 /* Time LRO pushed in pkt is valid in (PH_loc) */ 475 476 #define M_PROTO1 0x00002000 /* protocol-specific */ 477 #define M_PROTO2 0x00004000 /* protocol-specific */ 478 #define M_PROTO3 0x00008000 /* protocol-specific */ 479 #define M_PROTO4 0x00010000 /* protocol-specific */ 480 #define M_PROTO5 0x00020000 /* protocol-specific */ 481 #define M_PROTO6 0x00040000 /* protocol-specific */ 482 #define M_PROTO7 0x00080000 /* protocol-specific */ 483 #define M_PROTO8 0x00100000 /* protocol-specific */ 484 #define M_PROTO9 0x00200000 /* protocol-specific */ 485 #define M_PROTO10 0x00400000 /* protocol-specific */ 486 #define M_PROTO11 0x00800000 /* protocol-specific */ 487 488 /* 489 * Flags to purge when crossing layers. 490 */ 491 #define M_PROTOFLAGS \ 492 (M_PROTO1|M_PROTO2|M_PROTO3|M_PROTO4|M_PROTO5|M_PROTO6|M_PROTO7|M_PROTO8|\ 493 M_PROTO9|M_PROTO10|M_PROTO11) 494 495 /* 496 * Flags preserved when copying m_pkthdr. 497 */ 498 #define M_COPYFLAGS \ 499 (M_PKTHDR|M_EOR|M_RDONLY|M_BCAST|M_MCAST|M_PROMISC|M_VLANTAG|M_TSTMP| \ 500 M_TSTMP_HPREC|M_TSTMP_LRO|M_PROTOFLAGS) 501 502 /* 503 * Flags preserved during demote. 504 */ 505 #define M_DEMOTEFLAGS \ 506 (M_EXT | M_RDONLY | M_NOFREE | M_EXTPG) 507 508 /* 509 * Mbuf flag description for use with printf(9) %b identifier. 510 */ 511 #define M_FLAG_BITS \ 512 "\20\1M_EXT\2M_PKTHDR\3M_EOR\4M_RDONLY\5M_BCAST\6M_MCAST" \ 513 "\7M_PROMISC\10M_VLANTAG\11M_EXTPG\12M_NOFREE\13M_TSTMP\14M_TSTMP_HPREC\15M_TSTMP_LRO" 514 #define M_FLAG_PROTOBITS \ 515 "\16M_PROTO1\17M_PROTO2\20M_PROTO3\21M_PROTO4" \ 516 "\22M_PROTO5\23M_PROTO6\24M_PROTO7\25M_PROTO8\26M_PROTO9" \ 517 "\27M_PROTO10\28M_PROTO11" 518 #define M_FLAG_PRINTF (M_FLAG_BITS M_FLAG_PROTOBITS) 519 520 /* 521 * Network interface cards are able to hash protocol fields (such as IPv4 522 * addresses and TCP port numbers) classify packets into flows. These flows 523 * can then be used to maintain ordering while delivering packets to the OS 524 * via parallel input queues, as well as to provide a stateless affinity 525 * model. NIC drivers can pass up the hash via m->m_pkthdr.flowid, and set 526 * m_flag fields to indicate how the hash should be interpreted by the 527 * network stack. 528 * 529 * Most NICs support RSS, which provides ordering and explicit affinity, and 530 * use the hash m_flag bits to indicate what header fields were covered by 531 * the hash. M_HASHTYPE_OPAQUE and M_HASHTYPE_OPAQUE_HASH can be set by non- 532 * RSS cards or configurations that provide an opaque flow identifier, allowing 533 * for ordering and distribution without explicit affinity. Additionally, 534 * M_HASHTYPE_OPAQUE_HASH indicates that the flow identifier has hash 535 * properties. 536 * 537 * The meaning of the IPV6_EX suffix: 538 * "o Home address from the home address option in the IPv6 destination 539 * options header. If the extension header is not present, use the Source 540 * IPv6 Address. 541 * o IPv6 address that is contained in the Routing-Header-Type-2 from the 542 * associated extension header. If the extension header is not present, 543 * use the Destination IPv6 Address." 544 * Quoted from: 545 * https://docs.microsoft.com/en-us/windows-hardware/drivers/network/rss-hashing-types#ndishashipv6ex 546 */ 547 #define M_HASHTYPE_HASHPROP 0x80 /* has hash properties */ 548 #define M_HASHTYPE_INNER 0x40 /* calculated from inner headers */ 549 #define M_HASHTYPE_HASH(t) (M_HASHTYPE_HASHPROP | (t)) 550 /* Microsoft RSS standard hash types */ 551 #define M_HASHTYPE_NONE 0 552 #define M_HASHTYPE_RSS_IPV4 M_HASHTYPE_HASH(1) /* IPv4 2-tuple */ 553 #define M_HASHTYPE_RSS_TCP_IPV4 M_HASHTYPE_HASH(2) /* TCPv4 4-tuple */ 554 #define M_HASHTYPE_RSS_IPV6 M_HASHTYPE_HASH(3) /* IPv6 2-tuple */ 555 #define M_HASHTYPE_RSS_TCP_IPV6 M_HASHTYPE_HASH(4) /* TCPv6 4-tuple */ 556 #define M_HASHTYPE_RSS_IPV6_EX M_HASHTYPE_HASH(5) /* IPv6 2-tuple + 557 * ext hdrs */ 558 #define M_HASHTYPE_RSS_TCP_IPV6_EX M_HASHTYPE_HASH(6) /* TCPv6 4-tuple + 559 * ext hdrs */ 560 #define M_HASHTYPE_RSS_UDP_IPV4 M_HASHTYPE_HASH(7) /* IPv4 UDP 4-tuple*/ 561 #define M_HASHTYPE_RSS_UDP_IPV6 M_HASHTYPE_HASH(9) /* IPv6 UDP 4-tuple*/ 562 #define M_HASHTYPE_RSS_UDP_IPV6_EX M_HASHTYPE_HASH(10)/* IPv6 UDP 4-tuple + 563 * ext hdrs */ 564 565 #define M_HASHTYPE_OPAQUE 0x3f /* ordering, not affinity */ 566 #define M_HASHTYPE_OPAQUE_HASH M_HASHTYPE_HASH(M_HASHTYPE_OPAQUE) 567 /* ordering+hash, not affinity*/ 568 569 #define M_HASHTYPE_CLEAR(m) ((m)->m_pkthdr.rsstype = 0) 570 #define M_HASHTYPE_GET(m) ((m)->m_pkthdr.rsstype & ~M_HASHTYPE_INNER) 571 #define M_HASHTYPE_SET(m, v) ((m)->m_pkthdr.rsstype = (v)) 572 #define M_HASHTYPE_TEST(m, v) (M_HASHTYPE_GET(m) == (v)) 573 #define M_HASHTYPE_ISHASH(m) \ 574 (((m)->m_pkthdr.rsstype & M_HASHTYPE_HASHPROP) != 0) 575 #define M_HASHTYPE_SETINNER(m) do { \ 576 (m)->m_pkthdr.rsstype |= M_HASHTYPE_INNER; \ 577 } while (0) 578 579 /* 580 * External mbuf storage buffer types. 581 */ 582 #define EXT_CLUSTER 1 /* mbuf cluster */ 583 #define EXT_SFBUF 2 /* sendfile(2)'s sf_buf */ 584 #define EXT_JUMBOP 3 /* jumbo cluster page sized */ 585 #define EXT_JUMBO9 4 /* jumbo cluster 9216 bytes */ 586 #define EXT_JUMBO16 5 /* jumbo cluster 16184 bytes */ 587 #define EXT_PACKET 6 /* mbuf+cluster from packet zone */ 588 #define EXT_MBUF 7 /* external mbuf reference */ 589 #define EXT_RXRING 8 /* data in NIC receive ring */ 590 591 #define EXT_VENDOR1 224 /* for vendor-internal use */ 592 #define EXT_VENDOR2 225 /* for vendor-internal use */ 593 #define EXT_VENDOR3 226 /* for vendor-internal use */ 594 #define EXT_VENDOR4 227 /* for vendor-internal use */ 595 596 #define EXT_EXP1 244 /* for experimental use */ 597 #define EXT_EXP2 245 /* for experimental use */ 598 #define EXT_EXP3 246 /* for experimental use */ 599 #define EXT_EXP4 247 /* for experimental use */ 600 601 #define EXT_NET_DRV 252 /* custom ext_buf provided by net driver(s) */ 602 #define EXT_MOD_TYPE 253 /* custom module's ext_buf type */ 603 #define EXT_DISPOSABLE 254 /* can throw this buffer away w/page flipping */ 604 #define EXT_EXTREF 255 /* has externally maintained ext_cnt ptr */ 605 606 /* 607 * Flags for external mbuf buffer types. 608 * NB: limited to the lower 24 bits. 609 */ 610 #define EXT_FLAG_EMBREF 0x000001 /* embedded ext_count */ 611 #define EXT_FLAG_EXTREF 0x000002 /* external ext_cnt, notyet */ 612 613 #define EXT_FLAG_NOFREE 0x000010 /* don't free mbuf to pool, notyet */ 614 615 #define EXT_FLAG_VENDOR1 0x010000 /* These flags are vendor */ 616 #define EXT_FLAG_VENDOR2 0x020000 /* or submodule specific, */ 617 #define EXT_FLAG_VENDOR3 0x040000 /* not used by mbuf code. */ 618 #define EXT_FLAG_VENDOR4 0x080000 /* Set/read by submodule. */ 619 620 #define EXT_FLAG_EXP1 0x100000 /* for experimental use */ 621 #define EXT_FLAG_EXP2 0x200000 /* for experimental use */ 622 #define EXT_FLAG_EXP3 0x400000 /* for experimental use */ 623 #define EXT_FLAG_EXP4 0x800000 /* for experimental use */ 624 625 /* 626 * EXT flag description for use with printf(9) %b identifier. 627 */ 628 #define EXT_FLAG_BITS \ 629 "\20\1EXT_FLAG_EMBREF\2EXT_FLAG_EXTREF\5EXT_FLAG_NOFREE" \ 630 "\21EXT_FLAG_VENDOR1\22EXT_FLAG_VENDOR2\23EXT_FLAG_VENDOR3" \ 631 "\24EXT_FLAG_VENDOR4\25EXT_FLAG_EXP1\26EXT_FLAG_EXP2\27EXT_FLAG_EXP3" \ 632 "\30EXT_FLAG_EXP4" 633 634 /* 635 * Flags indicating checksum, segmentation and other offload work to be 636 * done, or already done, by hardware or lower layers. It is split into 637 * separate inbound and outbound flags. 638 * 639 * Outbound flags that are set by upper protocol layers requesting lower 640 * layers, or ideally the hardware, to perform these offloading tasks. 641 * For outbound packets this field and its flags can be directly tested 642 * against ifnet if_hwassist. Note that the outbound and the inbound flags do 643 * not collide right now but they could be allowed to (as long as the flags are 644 * scrubbed appropriately when the direction of an mbuf changes). CSUM_BITS 645 * would also have to split into CSUM_BITS_TX and CSUM_BITS_RX. 646 * 647 * CSUM_INNER_<x> is the same as CSUM_<x> but it applies to the inner frame. 648 * The CSUM_ENCAP_<x> bits identify the outer encapsulation. 649 */ 650 #define CSUM_IP 0x00000001 /* IP header checksum offload */ 651 #define CSUM_IP_UDP 0x00000002 /* UDP checksum offload */ 652 #define CSUM_IP_TCP 0x00000004 /* TCP checksum offload */ 653 #define CSUM_IP_SCTP 0x00000008 /* SCTP checksum offload */ 654 #define CSUM_IP_TSO 0x00000010 /* TCP segmentation offload */ 655 #define CSUM_IP_ISCSI 0x00000020 /* iSCSI checksum offload */ 656 657 #define CSUM_INNER_IP6_UDP 0x00000040 658 #define CSUM_INNER_IP6_TCP 0x00000080 659 #define CSUM_INNER_IP6_TSO 0x00000100 660 #define CSUM_IP6_UDP 0x00000200 /* UDP checksum offload */ 661 #define CSUM_IP6_TCP 0x00000400 /* TCP checksum offload */ 662 #define CSUM_IP6_SCTP 0x00000800 /* SCTP checksum offload */ 663 #define CSUM_IP6_TSO 0x00001000 /* TCP segmentation offload */ 664 #define CSUM_IP6_ISCSI 0x00002000 /* iSCSI checksum offload */ 665 666 #define CSUM_INNER_IP 0x00004000 667 #define CSUM_INNER_IP_UDP 0x00008000 668 #define CSUM_INNER_IP_TCP 0x00010000 669 #define CSUM_INNER_IP_TSO 0x00020000 670 671 #define CSUM_ENCAP_VXLAN 0x00040000 /* VXLAN outer encapsulation */ 672 #define CSUM_ENCAP_RSVD1 0x00080000 673 674 /* Inbound checksum support where the checksum was verified by hardware. */ 675 #define CSUM_INNER_L3_CALC 0x00100000 676 #define CSUM_INNER_L3_VALID 0x00200000 677 #define CSUM_INNER_L4_CALC 0x00400000 678 #define CSUM_INNER_L4_VALID 0x00800000 679 #define CSUM_L3_CALC 0x01000000 /* calculated layer 3 csum */ 680 #define CSUM_L3_VALID 0x02000000 /* checksum is correct */ 681 #define CSUM_L4_CALC 0x04000000 /* calculated layer 4 csum */ 682 #define CSUM_L4_VALID 0x08000000 /* checksum is correct */ 683 #define CSUM_L5_CALC 0x10000000 /* calculated layer 5 csum */ 684 #define CSUM_L5_VALID 0x20000000 /* checksum is correct */ 685 #define CSUM_COALESCED 0x40000000 /* contains merged segments */ 686 687 #define CSUM_SND_TAG 0x80000000 /* Packet header has send tag */ 688 689 #define CSUM_FLAGS_TX (CSUM_IP | CSUM_IP_UDP | CSUM_IP_TCP | CSUM_IP_SCTP | \ 690 CSUM_IP_TSO | CSUM_IP_ISCSI | CSUM_INNER_IP6_UDP | CSUM_INNER_IP6_TCP | \ 691 CSUM_INNER_IP6_TSO | CSUM_IP6_UDP | CSUM_IP6_TCP | CSUM_IP6_SCTP | \ 692 CSUM_IP6_TSO | CSUM_IP6_ISCSI | CSUM_INNER_IP | CSUM_INNER_IP_UDP | \ 693 CSUM_INNER_IP_TCP | CSUM_INNER_IP_TSO | CSUM_ENCAP_VXLAN | \ 694 CSUM_ENCAP_RSVD1 | CSUM_SND_TAG) 695 696 #define CSUM_FLAGS_RX (CSUM_INNER_L3_CALC | CSUM_INNER_L3_VALID | \ 697 CSUM_INNER_L4_CALC | CSUM_INNER_L4_VALID | CSUM_L3_CALC | CSUM_L3_VALID | \ 698 CSUM_L4_CALC | CSUM_L4_VALID | CSUM_L5_CALC | CSUM_L5_VALID | \ 699 CSUM_COALESCED) 700 701 /* 702 * CSUM flag description for use with printf(9) %b identifier. 703 */ 704 #define CSUM_BITS \ 705 "\20\1CSUM_IP\2CSUM_IP_UDP\3CSUM_IP_TCP\4CSUM_IP_SCTP\5CSUM_IP_TSO" \ 706 "\6CSUM_IP_ISCSI\7CSUM_INNER_IP6_UDP\10CSUM_INNER_IP6_TCP" \ 707 "\11CSUM_INNER_IP6_TSO\12CSUM_IP6_UDP\13CSUM_IP6_TCP\14CSUM_IP6_SCTP" \ 708 "\15CSUM_IP6_TSO\16CSUM_IP6_ISCSI\17CSUM_INNER_IP\20CSUM_INNER_IP_UDP" \ 709 "\21CSUM_INNER_IP_TCP\22CSUM_INNER_IP_TSO\23CSUM_ENCAP_VXLAN" \ 710 "\24CSUM_ENCAP_RSVD1\25CSUM_INNER_L3_CALC\26CSUM_INNER_L3_VALID" \ 711 "\27CSUM_INNER_L4_CALC\30CSUM_INNER_L4_VALID\31CSUM_L3_CALC" \ 712 "\32CSUM_L3_VALID\33CSUM_L4_CALC\34CSUM_L4_VALID\35CSUM_L5_CALC" \ 713 "\36CSUM_L5_VALID\37CSUM_COALESCED\40CSUM_SND_TAG" 714 715 /* CSUM flags compatibility mappings. */ 716 #define CSUM_IP_CHECKED CSUM_L3_CALC 717 #define CSUM_IP_VALID CSUM_L3_VALID 718 #define CSUM_DATA_VALID CSUM_L4_VALID 719 #define CSUM_PSEUDO_HDR CSUM_L4_CALC 720 #define CSUM_SCTP_VALID CSUM_L4_VALID 721 #define CSUM_DELAY_DATA (CSUM_TCP|CSUM_UDP) 722 #define CSUM_DELAY_IP CSUM_IP /* Only v4, no v6 IP hdr csum */ 723 #define CSUM_DELAY_DATA_IPV6 (CSUM_TCP_IPV6|CSUM_UDP_IPV6) 724 #define CSUM_DATA_VALID_IPV6 CSUM_DATA_VALID 725 #define CSUM_TCP CSUM_IP_TCP 726 #define CSUM_UDP CSUM_IP_UDP 727 #define CSUM_SCTP CSUM_IP_SCTP 728 #define CSUM_TSO (CSUM_IP_TSO|CSUM_IP6_TSO) 729 #define CSUM_INNER_TSO (CSUM_INNER_IP_TSO|CSUM_INNER_IP6_TSO) 730 #define CSUM_UDP_IPV6 CSUM_IP6_UDP 731 #define CSUM_TCP_IPV6 CSUM_IP6_TCP 732 #define CSUM_SCTP_IPV6 CSUM_IP6_SCTP 733 #define CSUM_TLS_MASK (CSUM_L5_CALC|CSUM_L5_VALID) 734 #define CSUM_TLS_DECRYPTED CSUM_L5_CALC 735 736 /* 737 * mbuf types describing the content of the mbuf (including external storage). 738 */ 739 #define MT_NOTMBUF 0 /* USED INTERNALLY ONLY! Object is not mbuf */ 740 #define MT_DATA 1 /* dynamic (data) allocation */ 741 #define MT_HEADER MT_DATA /* packet header, use M_PKTHDR instead */ 742 743 #define MT_VENDOR1 4 /* for vendor-internal use */ 744 #define MT_VENDOR2 5 /* for vendor-internal use */ 745 #define MT_VENDOR3 6 /* for vendor-internal use */ 746 #define MT_VENDOR4 7 /* for vendor-internal use */ 747 748 #define MT_SONAME 8 /* socket name */ 749 750 #define MT_EXP1 9 /* for experimental use */ 751 #define MT_EXP2 10 /* for experimental use */ 752 #define MT_EXP3 11 /* for experimental use */ 753 #define MT_EXP4 12 /* for experimental use */ 754 755 #define MT_CONTROL 14 /* extra-data protocol message */ 756 #define MT_EXTCONTROL 15 /* control message with externalized contents */ 757 #define MT_OOBDATA 16 /* expedited data */ 758 759 #define MT_NOINIT 255 /* Not a type but a flag to allocate 760 a non-initialized mbuf */ 761 762 /* 763 * String names of mbuf-related UMA(9) and malloc(9) types. Exposed to 764 * !_KERNEL so that monitoring tools can look up the zones with 765 * libmemstat(3). 766 */ 767 #define MBUF_MEM_NAME "mbuf" 768 #define MBUF_CLUSTER_MEM_NAME "mbuf_cluster" 769 #define MBUF_PACKET_MEM_NAME "mbuf_packet" 770 #define MBUF_JUMBOP_MEM_NAME "mbuf_jumbo_page" 771 #define MBUF_JUMBO9_MEM_NAME "mbuf_jumbo_9k" 772 #define MBUF_JUMBO16_MEM_NAME "mbuf_jumbo_16k" 773 #define MBUF_TAG_MEM_NAME "mbuf_tag" 774 #define MBUF_EXTREFCNT_MEM_NAME "mbuf_ext_refcnt" 775 #define MBUF_EXTPGS_MEM_NAME "mbuf_extpgs" 776 777 #ifdef _KERNEL 778 union if_snd_tag_alloc_params; 779 780 #ifdef WITNESS 781 #define MBUF_CHECKSLEEP(how) do { \ 782 if (how == M_WAITOK) \ 783 WITNESS_WARN(WARN_GIANTOK | WARN_SLEEPOK, NULL, \ 784 "Sleeping in \"%s\"", __func__); \ 785 } while (0) 786 #else 787 #define MBUF_CHECKSLEEP(how) do {} while (0) 788 #endif 789 790 /* 791 * Network buffer allocation API 792 * 793 * The rest of it is defined in kern/kern_mbuf.c 794 */ 795 extern uma_zone_t zone_mbuf; 796 extern uma_zone_t zone_clust; 797 extern uma_zone_t zone_pack; 798 extern uma_zone_t zone_jumbop; 799 extern uma_zone_t zone_jumbo9; 800 extern uma_zone_t zone_jumbo16; 801 extern uma_zone_t zone_extpgs; 802 803 void mb_dupcl(struct mbuf *, struct mbuf *); 804 void mb_free_ext(struct mbuf *); 805 void mb_free_extpg(struct mbuf *); 806 void mb_free_mext_pgs(struct mbuf *); 807 struct mbuf *mb_alloc_ext_pgs(int, m_ext_free_t); 808 struct mbuf *mb_alloc_ext_plus_pages(int, int); 809 struct mbuf *mb_mapped_to_unmapped(struct mbuf *, int, int, int, 810 struct mbuf **); 811 int mb_unmapped_compress(struct mbuf *m); 812 struct mbuf *mb_unmapped_to_ext(struct mbuf *m); 813 void mb_free_notready(struct mbuf *m, int count); 814 void m_adj(struct mbuf *, int); 815 void m_adj_decap(struct mbuf *, int); 816 int m_apply(struct mbuf *, int, int, 817 int (*)(void *, void *, u_int), void *); 818 int m_append(struct mbuf *, int, c_caddr_t); 819 void m_cat(struct mbuf *, struct mbuf *); 820 void m_catpkt(struct mbuf *, struct mbuf *); 821 int m_clget(struct mbuf *m, int how); 822 void *m_cljget(struct mbuf *m, int how, int size); 823 struct mbuf *m_collapse(struct mbuf *, int, int); 824 void m_copyback(struct mbuf *, int, int, c_caddr_t); 825 void m_copydata(const struct mbuf *, int, int, caddr_t); 826 struct mbuf *m_copym(struct mbuf *, int, int, int); 827 struct mbuf *m_copypacket(struct mbuf *, int); 828 void m_copy_pkthdr(struct mbuf *, struct mbuf *); 829 struct mbuf *m_copyup(struct mbuf *, int, int); 830 struct mbuf *m_defrag(struct mbuf *, int); 831 void m_demote_pkthdr(struct mbuf *); 832 void m_demote(struct mbuf *, int, int); 833 struct mbuf *m_devget(char *, int, int, struct ifnet *, 834 void (*)(char *, caddr_t, u_int)); 835 void m_dispose_extcontrolm(struct mbuf *m); 836 struct mbuf *m_dup(const struct mbuf *, int); 837 int m_dup_pkthdr(struct mbuf *, const struct mbuf *, int); 838 void m_extadd(struct mbuf *, char *, u_int, m_ext_free_t, 839 void *, void *, int, int); 840 u_int m_fixhdr(struct mbuf *); 841 struct mbuf *m_fragment(struct mbuf *, int, int); 842 void m_freem(struct mbuf *); 843 void m_free_raw(struct mbuf *); 844 struct mbuf *m_get2(int, int, short, int); 845 struct mbuf *m_get3(int, int, short, int); 846 struct mbuf *m_getjcl(int, short, int, int); 847 struct mbuf *m_getm2(struct mbuf *, int, int, short, int); 848 struct mbuf *m_getptr(struct mbuf *, int, int *); 849 u_int m_length(struct mbuf *, struct mbuf **); 850 int m_mbuftouio(struct uio *, const struct mbuf *, int); 851 void m_move_pkthdr(struct mbuf *, struct mbuf *); 852 int m_pkthdr_init(struct mbuf *, int); 853 struct mbuf *m_prepend(struct mbuf *, int, int); 854 void m_print(const struct mbuf *, int); 855 struct mbuf *m_pulldown(struct mbuf *, int, int, int *); 856 struct mbuf *m_pullup(struct mbuf *, int); 857 int m_sanity(struct mbuf *, int); 858 struct mbuf *m_split(struct mbuf *, int, int); 859 struct mbuf *m_uiotombuf(struct uio *, int, int, int, int); 860 int m_unmapped_uiomove(const struct mbuf *, int, struct uio *, 861 int); 862 struct mbuf *m_unshare(struct mbuf *, int); 863 int m_snd_tag_alloc(struct ifnet *, 864 union if_snd_tag_alloc_params *, struct m_snd_tag **); 865 void m_snd_tag_init(struct m_snd_tag *, struct ifnet *, 866 const struct if_snd_tag_sw *); 867 void m_snd_tag_destroy(struct m_snd_tag *); 868 void m_rcvif_serialize(struct mbuf *); 869 struct ifnet *m_rcvif_restore(struct mbuf *); 870 871 static __inline int 872 m_gettype(int size) 873 { 874 int type; 875 876 switch (size) { 877 case MSIZE: 878 type = EXT_MBUF; 879 break; 880 case MCLBYTES: 881 type = EXT_CLUSTER; 882 break; 883 #if MJUMPAGESIZE != MCLBYTES 884 case MJUMPAGESIZE: 885 type = EXT_JUMBOP; 886 break; 887 #endif 888 case MJUM9BYTES: 889 type = EXT_JUMBO9; 890 break; 891 case MJUM16BYTES: 892 type = EXT_JUMBO16; 893 break; 894 default: 895 panic("%s: invalid cluster size %d", __func__, size); 896 } 897 898 return (type); 899 } 900 901 /* 902 * Associated an external reference counted buffer with an mbuf. 903 */ 904 static __inline void 905 m_extaddref(struct mbuf *m, char *buf, u_int size, u_int *ref_cnt, 906 m_ext_free_t freef, void *arg1, void *arg2) 907 { 908 909 KASSERT(ref_cnt != NULL, ("%s: ref_cnt not provided", __func__)); 910 911 atomic_add_int(ref_cnt, 1); 912 m->m_flags |= M_EXT; 913 m->m_ext.ext_buf = buf; 914 m->m_ext.ext_cnt = ref_cnt; 915 m->m_data = m->m_ext.ext_buf; 916 m->m_ext.ext_size = size; 917 m->m_ext.ext_free = freef; 918 m->m_ext.ext_arg1 = arg1; 919 m->m_ext.ext_arg2 = arg2; 920 m->m_ext.ext_type = EXT_EXTREF; 921 m->m_ext.ext_flags = 0; 922 } 923 924 static __inline uma_zone_t 925 m_getzone(int size) 926 { 927 uma_zone_t zone; 928 929 switch (size) { 930 case MCLBYTES: 931 zone = zone_clust; 932 break; 933 #if MJUMPAGESIZE != MCLBYTES 934 case MJUMPAGESIZE: 935 zone = zone_jumbop; 936 break; 937 #endif 938 case MJUM9BYTES: 939 zone = zone_jumbo9; 940 break; 941 case MJUM16BYTES: 942 zone = zone_jumbo16; 943 break; 944 default: 945 panic("%s: invalid cluster size %d", __func__, size); 946 } 947 948 return (zone); 949 } 950 951 /* 952 * Initialize an mbuf with linear storage. 953 * 954 * Inline because the consumer text overhead will be roughly the same to 955 * initialize or call a function with this many parameters and M_PKTHDR 956 * should go away with constant propagation for !MGETHDR. 957 */ 958 static __inline int 959 m_init(struct mbuf *m, int how, short type, int flags) 960 { 961 int error; 962 963 m->m_next = NULL; 964 m->m_nextpkt = NULL; 965 m->m_data = m->m_dat; 966 m->m_len = 0; 967 m->m_flags = flags; 968 m->m_type = type; 969 if (flags & M_PKTHDR) 970 error = m_pkthdr_init(m, how); 971 else 972 error = 0; 973 974 MBUF_PROBE5(m__init, m, how, type, flags, error); 975 return (error); 976 } 977 978 static __inline struct mbuf * 979 m_get_raw(int how, short type) 980 { 981 struct mbuf *m; 982 struct mb_args args; 983 984 args.flags = 0; 985 args.type = type | MT_NOINIT; 986 m = uma_zalloc_arg(zone_mbuf, &args, how); 987 MBUF_PROBE3(m__get_raw, how, type, m); 988 return (m); 989 } 990 991 static __inline struct mbuf * 992 m_get(int how, short type) 993 { 994 struct mbuf *m; 995 struct mb_args args; 996 997 args.flags = 0; 998 args.type = type; 999 m = uma_zalloc_arg(zone_mbuf, &args, how); 1000 MBUF_PROBE3(m__get, how, type, m); 1001 return (m); 1002 } 1003 1004 static __inline struct mbuf * 1005 m_gethdr_raw(int how, short type) 1006 { 1007 struct mbuf *m; 1008 struct mb_args args; 1009 1010 args.flags = M_PKTHDR; 1011 args.type = type | MT_NOINIT; 1012 m = uma_zalloc_arg(zone_mbuf, &args, how); 1013 MBUF_PROBE3(m__gethdr_raw, how, type, m); 1014 return (m); 1015 } 1016 1017 static __inline struct mbuf * 1018 m_gethdr(int how, short type) 1019 { 1020 struct mbuf *m; 1021 struct mb_args args; 1022 1023 args.flags = M_PKTHDR; 1024 args.type = type; 1025 m = uma_zalloc_arg(zone_mbuf, &args, how); 1026 MBUF_PROBE3(m__gethdr, how, type, m); 1027 return (m); 1028 } 1029 1030 static __inline struct mbuf * 1031 m_getcl(int how, short type, int flags) 1032 { 1033 struct mbuf *m; 1034 struct mb_args args; 1035 1036 args.flags = flags; 1037 args.type = type; 1038 m = uma_zalloc_arg(zone_pack, &args, how); 1039 MBUF_PROBE4(m__getcl, how, type, flags, m); 1040 return (m); 1041 } 1042 1043 /* 1044 * XXX: m_cljset() is a dangerous API. One must attach only a new, 1045 * unreferenced cluster to an mbuf(9). It is not possible to assert 1046 * that, so care can be taken only by users of the API. 1047 */ 1048 static __inline void 1049 m_cljset(struct mbuf *m, void *cl, int type) 1050 { 1051 int size; 1052 1053 switch (type) { 1054 case EXT_CLUSTER: 1055 size = MCLBYTES; 1056 break; 1057 #if MJUMPAGESIZE != MCLBYTES 1058 case EXT_JUMBOP: 1059 size = MJUMPAGESIZE; 1060 break; 1061 #endif 1062 case EXT_JUMBO9: 1063 size = MJUM9BYTES; 1064 break; 1065 case EXT_JUMBO16: 1066 size = MJUM16BYTES; 1067 break; 1068 default: 1069 panic("%s: unknown cluster type %d", __func__, type); 1070 break; 1071 } 1072 1073 m->m_data = m->m_ext.ext_buf = cl; 1074 m->m_ext.ext_free = m->m_ext.ext_arg1 = m->m_ext.ext_arg2 = NULL; 1075 m->m_ext.ext_size = size; 1076 m->m_ext.ext_type = type; 1077 m->m_ext.ext_flags = EXT_FLAG_EMBREF; 1078 m->m_ext.ext_count = 1; 1079 m->m_flags |= M_EXT; 1080 MBUF_PROBE3(m__cljset, m, cl, type); 1081 } 1082 1083 static __inline void 1084 m_chtype(struct mbuf *m, short new_type) 1085 { 1086 1087 m->m_type = new_type; 1088 } 1089 1090 static __inline void 1091 m_clrprotoflags(struct mbuf *m) 1092 { 1093 1094 while (m) { 1095 m->m_flags &= ~M_PROTOFLAGS; 1096 m = m->m_next; 1097 } 1098 } 1099 1100 static __inline struct mbuf * 1101 m_last(struct mbuf *m) 1102 { 1103 1104 while (m->m_next) 1105 m = m->m_next; 1106 return (m); 1107 } 1108 1109 static inline u_int 1110 m_extrefcnt(struct mbuf *m) 1111 { 1112 1113 KASSERT(m->m_flags & M_EXT, ("%s: M_EXT missing", __func__)); 1114 1115 return ((m->m_ext.ext_flags & EXT_FLAG_EMBREF) ? m->m_ext.ext_count : 1116 *m->m_ext.ext_cnt); 1117 } 1118 1119 /* 1120 * mbuf, cluster, and external object allocation macros (for compatibility 1121 * purposes). 1122 */ 1123 #define M_MOVE_PKTHDR(to, from) m_move_pkthdr((to), (from)) 1124 #define MGET(m, how, type) ((m) = m_get((how), (type))) 1125 #define MGETHDR(m, how, type) ((m) = m_gethdr((how), (type))) 1126 #define MCLGET(m, how) m_clget((m), (how)) 1127 #define MEXTADD(m, buf, size, free, arg1, arg2, flags, type) \ 1128 m_extadd((m), (char *)(buf), (size), (free), (arg1), (arg2), \ 1129 (flags), (type)) 1130 #define m_getm(m, len, how, type) \ 1131 m_getm2((m), (len), (how), (type), M_PKTHDR) 1132 1133 /* 1134 * Evaluate TRUE if it's safe to write to the mbuf m's data region (this can 1135 * be both the local data payload, or an external buffer area, depending on 1136 * whether M_EXT is set). 1137 */ 1138 #define M_WRITABLE(m) (((m)->m_flags & (M_RDONLY | M_EXTPG)) == 0 && \ 1139 (!(((m)->m_flags & M_EXT)) || \ 1140 (m_extrefcnt(m) == 1))) 1141 1142 /* Check if the supplied mbuf has a packet header, or else panic. */ 1143 #define M_ASSERTPKTHDR(m) \ 1144 KASSERT((m) != NULL && (m)->m_flags & M_PKTHDR, \ 1145 ("%s: no mbuf packet header!", __func__)) 1146 1147 /* Check if the supplied mbuf has no send tag, or else panic. */ 1148 #define M_ASSERT_NO_SND_TAG(m) \ 1149 KASSERT((m) != NULL && (m)->m_flags & M_PKTHDR && \ 1150 ((m)->m_pkthdr.csum_flags & CSUM_SND_TAG) == 0, \ 1151 ("%s: receive mbuf has send tag!", __func__)) 1152 1153 /* Check if mbuf is multipage. */ 1154 #define M_ASSERTEXTPG(m) \ 1155 KASSERT(((m)->m_flags & (M_EXTPG|M_PKTHDR)) == M_EXTPG, \ 1156 ("%s: m %p is not multipage!", __func__, m)) 1157 1158 /* 1159 * Ensure that the supplied mbuf is a valid, non-free mbuf. 1160 * 1161 * XXX: Broken at the moment. Need some UMA magic to make it work again. 1162 */ 1163 #define M_ASSERTVALID(m) \ 1164 KASSERT((((struct mbuf *)m)->m_flags & 0) == 0, \ 1165 ("%s: attempted use of a free mbuf!", __func__)) 1166 1167 /* Check whether any mbuf in the chain is unmapped. */ 1168 #ifdef INVARIANTS 1169 #define M_ASSERTMAPPED(m) do { \ 1170 for (struct mbuf *__m = (m); __m != NULL; __m = __m->m_next) \ 1171 KASSERT((__m->m_flags & M_EXTPG) == 0, \ 1172 ("%s: chain %p contains an unmapped mbuf", __func__, (m)));\ 1173 } while (0) 1174 #else 1175 #define M_ASSERTMAPPED(m) do {} while (0) 1176 #endif 1177 1178 /* 1179 * Return the address of the start of the buffer associated with an mbuf, 1180 * handling external storage, packet-header mbufs, and regular data mbufs. 1181 */ 1182 #define M_START(m) \ 1183 (((m)->m_flags & M_EXTPG) ? NULL : \ 1184 ((m)->m_flags & M_EXT) ? (m)->m_ext.ext_buf : \ 1185 ((m)->m_flags & M_PKTHDR) ? &(m)->m_pktdat[0] : \ 1186 &(m)->m_dat[0]) 1187 1188 /* 1189 * Return the size of the buffer associated with an mbuf, handling external 1190 * storage, packet-header mbufs, and regular data mbufs. 1191 */ 1192 #define M_SIZE(m) \ 1193 (((m)->m_flags & M_EXT) ? (m)->m_ext.ext_size : \ 1194 ((m)->m_flags & M_PKTHDR) ? MHLEN : \ 1195 MLEN) 1196 1197 /* 1198 * Set the m_data pointer of a newly allocated mbuf to place an object of the 1199 * specified size at the end of the mbuf, longword aligned. 1200 * 1201 * NB: Historically, we had M_ALIGN(), MH_ALIGN(), and MEXT_ALIGN() as 1202 * separate macros, each asserting that it was called at the proper moment. 1203 * This required callers to themselves test the storage type and call the 1204 * right one. Rather than require callers to be aware of those layout 1205 * decisions, we centralize here. 1206 */ 1207 static __inline void 1208 m_align(struct mbuf *m, int len) 1209 { 1210 #ifdef INVARIANTS 1211 const char *msg = "%s: not a virgin mbuf"; 1212 #endif 1213 int adjust; 1214 1215 KASSERT(m->m_data == M_START(m), (msg, __func__)); 1216 1217 adjust = M_SIZE(m) - len; 1218 m->m_data += adjust &~ (sizeof(long)-1); 1219 } 1220 1221 #define M_ALIGN(m, len) m_align(m, len) 1222 #define MH_ALIGN(m, len) m_align(m, len) 1223 #define MEXT_ALIGN(m, len) m_align(m, len) 1224 1225 /* 1226 * Compute the amount of space available before the current start of data in 1227 * an mbuf. 1228 * 1229 * The M_WRITABLE() is a temporary, conservative safety measure: the burden 1230 * of checking writability of the mbuf data area rests solely with the caller. 1231 * 1232 * NB: In previous versions, M_LEADINGSPACE() would only check M_WRITABLE() 1233 * for mbufs with external storage. We now allow mbuf-embedded data to be 1234 * read-only as well. 1235 */ 1236 #define M_LEADINGSPACE(m) \ 1237 (M_WRITABLE(m) ? ((m)->m_data - M_START(m)) : 0) 1238 1239 /* 1240 * Compute the amount of space available after the end of data in an mbuf. 1241 * 1242 * The M_WRITABLE() is a temporary, conservative safety measure: the burden 1243 * of checking writability of the mbuf data area rests solely with the caller. 1244 * 1245 * NB: In previous versions, M_TRAILINGSPACE() would only check M_WRITABLE() 1246 * for mbufs with external storage. We now allow mbuf-embedded data to be 1247 * read-only as well. 1248 */ 1249 #define M_TRAILINGSPACE(m) \ 1250 (M_WRITABLE(m) ? \ 1251 ((M_START(m) + M_SIZE(m)) - ((m)->m_data + (m)->m_len)) : 0) 1252 1253 /* 1254 * Arrange to prepend space of size plen to mbuf m. If a new mbuf must be 1255 * allocated, how specifies whether to wait. If the allocation fails, the 1256 * original mbuf chain is freed and m is set to NULL. 1257 */ 1258 #define M_PREPEND(m, plen, how) do { \ 1259 struct mbuf **_mmp = &(m); \ 1260 struct mbuf *_mm = *_mmp; \ 1261 int _mplen = (plen); \ 1262 int __mhow = (how); \ 1263 \ 1264 MBUF_CHECKSLEEP(how); \ 1265 if (M_LEADINGSPACE(_mm) >= _mplen) { \ 1266 _mm->m_data -= _mplen; \ 1267 _mm->m_len += _mplen; \ 1268 } else \ 1269 _mm = m_prepend(_mm, _mplen, __mhow); \ 1270 if (_mm != NULL && _mm->m_flags & M_PKTHDR) \ 1271 _mm->m_pkthdr.len += _mplen; \ 1272 *_mmp = _mm; \ 1273 } while (0) 1274 1275 /* 1276 * Change mbuf to new type. This is a relatively expensive operation and 1277 * should be avoided. 1278 */ 1279 #define MCHTYPE(m, t) m_chtype((m), (t)) 1280 1281 /* Return the rcvif of a packet header. */ 1282 static __inline struct ifnet * 1283 m_rcvif(struct mbuf *m) 1284 { 1285 1286 M_ASSERTPKTHDR(m); 1287 if (m->m_pkthdr.csum_flags & CSUM_SND_TAG) 1288 return (NULL); 1289 return (m->m_pkthdr.rcvif); 1290 } 1291 1292 /* Length to m_copy to copy all. */ 1293 #define M_COPYALL 1000000000 1294 1295 extern int max_datalen; /* MHLEN - max_hdr */ 1296 extern int max_hdr; /* Largest link + protocol header */ 1297 extern int max_linkhdr; /* Largest link-level header */ 1298 extern int max_protohdr; /* Largest protocol header */ 1299 extern int nmbclusters; /* Maximum number of clusters */ 1300 extern bool mb_use_ext_pgs; /* Use ext_pgs for sendfile */ 1301 1302 /*- 1303 * Network packets may have annotations attached by affixing a list of 1304 * "packet tags" to the pkthdr structure. Packet tags are dynamically 1305 * allocated semi-opaque data structures that have a fixed header 1306 * (struct m_tag) that specifies the size of the memory block and a 1307 * <cookie,type> pair that identifies it. The cookie is a 32-bit unique 1308 * unsigned value used to identify a module or ABI. By convention this value 1309 * is chosen as the date+time that the module is created, expressed as the 1310 * number of seconds since the epoch (e.g., using date -u +'%s'). The type 1311 * value is an ABI/module-specific value that identifies a particular 1312 * annotation and is private to the module. For compatibility with systems 1313 * like OpenBSD that define packet tags w/o an ABI/module cookie, the value 1314 * PACKET_ABI_COMPAT is used to implement m_tag_get and m_tag_find 1315 * compatibility shim functions and several tag types are defined below. 1316 * Users that do not require compatibility should use a private cookie value 1317 * so that packet tag-related definitions can be maintained privately. 1318 * 1319 * Note that the packet tag returned by m_tag_alloc has the default memory 1320 * alignment implemented by malloc. To reference private data one can use a 1321 * construct like: 1322 * 1323 * struct m_tag *mtag = m_tag_alloc(...); 1324 * struct foo *p = (struct foo *)(mtag+1); 1325 * 1326 * if the alignment of struct m_tag is sufficient for referencing members of 1327 * struct foo. Otherwise it is necessary to embed struct m_tag within the 1328 * private data structure to insure proper alignment; e.g., 1329 * 1330 * struct foo { 1331 * struct m_tag tag; 1332 * ... 1333 * }; 1334 * struct foo *p = (struct foo *) m_tag_alloc(...); 1335 * struct m_tag *mtag = &p->tag; 1336 */ 1337 1338 /* 1339 * Persistent tags stay with an mbuf until the mbuf is reclaimed. Otherwise 1340 * tags are expected to ``vanish'' when they pass through a network 1341 * interface. For most interfaces this happens normally as the tags are 1342 * reclaimed when the mbuf is free'd. However in some special cases 1343 * reclaiming must be done manually. An example is packets that pass through 1344 * the loopback interface. Also, one must be careful to do this when 1345 * ``turning around'' packets (e.g., icmp_reflect). 1346 * 1347 * To mark a tag persistent bit-or this flag in when defining the tag id. 1348 * The tag will then be treated as described above. 1349 */ 1350 #define MTAG_PERSISTENT 0x800 1351 1352 #define PACKET_TAG_NONE 0 /* Nadda */ 1353 1354 /* Packet tags for use with PACKET_ABI_COMPAT. */ 1355 #define PACKET_TAG_IPSEC_IN_DONE 1 /* IPsec applied, in */ 1356 #define PACKET_TAG_IPSEC_OUT_DONE 2 /* IPsec applied, out */ 1357 #define PACKET_TAG_IPSEC_IN_CRYPTO_DONE 3 /* NIC IPsec crypto done */ 1358 #define PACKET_TAG_IPSEC_OUT_CRYPTO_NEEDED 4 /* NIC IPsec crypto req'ed */ 1359 #define PACKET_TAG_IPSEC_IN_COULD_DO_CRYPTO 5 /* NIC notifies IPsec */ 1360 #define PACKET_TAG_IPSEC_PENDING_TDB 6 /* Reminder to do IPsec */ 1361 #define PACKET_TAG_BRIDGE 7 /* Bridge processing done */ 1362 #define PACKET_TAG_GIF 8 /* GIF processing done */ 1363 #define PACKET_TAG_GRE 9 /* GRE processing done */ 1364 #define PACKET_TAG_IN_PACKET_CHECKSUM 10 /* NIC checksumming done */ 1365 #define PACKET_TAG_ENCAP 11 /* Encap. processing */ 1366 #define PACKET_TAG_IPSEC_SOCKET 12 /* IPSEC socket ref */ 1367 #define PACKET_TAG_IPSEC_HISTORY 13 /* IPSEC history */ 1368 #define PACKET_TAG_IPV6_INPUT 14 /* IPV6 input processing */ 1369 #define PACKET_TAG_DUMMYNET 15 /* dummynet info */ 1370 #define PACKET_TAG_DIVERT 17 /* divert info */ 1371 #define PACKET_TAG_IPFORWARD 18 /* ipforward info */ 1372 #define PACKET_TAG_MACLABEL (19 | MTAG_PERSISTENT) /* MAC label */ 1373 #define PACKET_TAG_PF 21 /* PF/ALTQ information */ 1374 #define PACKET_TAG_RTSOCKFAM 25 /* rtsock sa family */ 1375 #define PACKET_TAG_IPOPTIONS 27 /* Saved IP options */ 1376 #define PACKET_TAG_CARP 28 /* CARP info */ 1377 #define PACKET_TAG_IPSEC_NAT_T_PORTS 29 /* two uint16_t */ 1378 #define PACKET_TAG_ND_OUTGOING 30 /* ND outgoing */ 1379 1380 /* Specific cookies and tags. */ 1381 1382 /* Packet tag routines. */ 1383 struct m_tag *m_tag_alloc(uint32_t, uint16_t, int, int); 1384 void m_tag_delete(struct mbuf *, struct m_tag *); 1385 void m_tag_delete_chain(struct mbuf *, struct m_tag *); 1386 void m_tag_free_default(struct m_tag *); 1387 struct m_tag *m_tag_locate(struct mbuf *, uint32_t, uint16_t, 1388 struct m_tag *); 1389 struct m_tag *m_tag_copy(struct m_tag *, int); 1390 int m_tag_copy_chain(struct mbuf *, const struct mbuf *, int); 1391 void m_tag_delete_nonpersistent(struct mbuf *); 1392 1393 /* 1394 * Initialize the list of tags associated with an mbuf. 1395 */ 1396 static __inline void 1397 m_tag_init(struct mbuf *m) 1398 { 1399 1400 SLIST_INIT(&m->m_pkthdr.tags); 1401 } 1402 1403 /* 1404 * Set up the contents of a tag. Note that this does not fill in the free 1405 * method; the caller is expected to do that. 1406 * 1407 * XXX probably should be called m_tag_init, but that was already taken. 1408 */ 1409 static __inline void 1410 m_tag_setup(struct m_tag *t, uint32_t cookie, uint16_t type, int len) 1411 { 1412 1413 t->m_tag_id = type; 1414 t->m_tag_len = len; 1415 t->m_tag_cookie = cookie; 1416 } 1417 1418 /* 1419 * Reclaim resources associated with a tag. 1420 */ 1421 static __inline void 1422 m_tag_free(struct m_tag *t) 1423 { 1424 1425 (*t->m_tag_free)(t); 1426 } 1427 1428 /* 1429 * Return the first tag associated with an mbuf. 1430 */ 1431 static __inline struct m_tag * 1432 m_tag_first(struct mbuf *m) 1433 { 1434 1435 return (SLIST_FIRST(&m->m_pkthdr.tags)); 1436 } 1437 1438 /* 1439 * Return the next tag in the list of tags associated with an mbuf. 1440 */ 1441 static __inline struct m_tag * 1442 m_tag_next(struct mbuf *m __unused, struct m_tag *t) 1443 { 1444 1445 return (SLIST_NEXT(t, m_tag_link)); 1446 } 1447 1448 /* 1449 * Prepend a tag to the list of tags associated with an mbuf. 1450 */ 1451 static __inline void 1452 m_tag_prepend(struct mbuf *m, struct m_tag *t) 1453 { 1454 1455 SLIST_INSERT_HEAD(&m->m_pkthdr.tags, t, m_tag_link); 1456 } 1457 1458 /* 1459 * Unlink a tag from the list of tags associated with an mbuf. 1460 */ 1461 static __inline void 1462 m_tag_unlink(struct mbuf *m, struct m_tag *t) 1463 { 1464 1465 SLIST_REMOVE(&m->m_pkthdr.tags, t, m_tag, m_tag_link); 1466 } 1467 1468 /* These are for OpenBSD compatibility. */ 1469 #define MTAG_ABI_COMPAT 0 /* compatibility ABI */ 1470 1471 static __inline struct m_tag * 1472 m_tag_get(uint16_t type, int length, int wait) 1473 { 1474 return (m_tag_alloc(MTAG_ABI_COMPAT, type, length, wait)); 1475 } 1476 1477 static __inline struct m_tag * 1478 m_tag_find(struct mbuf *m, uint16_t type, struct m_tag *start) 1479 { 1480 return (SLIST_EMPTY(&m->m_pkthdr.tags) ? (struct m_tag *)NULL : 1481 m_tag_locate(m, MTAG_ABI_COMPAT, type, start)); 1482 } 1483 1484 static inline struct m_snd_tag * 1485 m_snd_tag_ref(struct m_snd_tag *mst) 1486 { 1487 1488 refcount_acquire(&mst->refcount); 1489 return (mst); 1490 } 1491 1492 static inline void 1493 m_snd_tag_rele(struct m_snd_tag *mst) 1494 { 1495 1496 if (refcount_release(&mst->refcount)) 1497 m_snd_tag_destroy(mst); 1498 } 1499 1500 static __inline struct mbuf * 1501 m_free(struct mbuf *m) 1502 { 1503 struct mbuf *n = m->m_next; 1504 1505 MBUF_PROBE1(m__free, m); 1506 if ((m->m_flags & (M_PKTHDR|M_NOFREE)) == (M_PKTHDR|M_NOFREE)) 1507 m_tag_delete_chain(m, NULL); 1508 if (m->m_flags & M_PKTHDR && m->m_pkthdr.csum_flags & CSUM_SND_TAG) 1509 m_snd_tag_rele(m->m_pkthdr.snd_tag); 1510 if (m->m_flags & M_EXTPG) 1511 mb_free_extpg(m); 1512 else if (m->m_flags & M_EXT) 1513 mb_free_ext(m); 1514 else if ((m->m_flags & M_NOFREE) == 0) 1515 uma_zfree(zone_mbuf, m); 1516 return (n); 1517 } 1518 1519 static __inline int 1520 rt_m_getfib(struct mbuf *m) 1521 { 1522 KASSERT(m->m_flags & M_PKTHDR , ("Attempt to get FIB from non header mbuf.")); 1523 return (m->m_pkthdr.fibnum); 1524 } 1525 1526 #define M_GETFIB(_m) rt_m_getfib(_m) 1527 1528 #define M_SETFIB(_m, _fib) do { \ 1529 KASSERT((_m)->m_flags & M_PKTHDR, ("Attempt to set FIB on non header mbuf.")); \ 1530 ((_m)->m_pkthdr.fibnum) = (_fib); \ 1531 } while (0) 1532 1533 /* flags passed as first argument for "m_xxx_tcpip_hash()" */ 1534 #define MBUF_HASHFLAG_L2 (1 << 2) 1535 #define MBUF_HASHFLAG_L3 (1 << 3) 1536 #define MBUF_HASHFLAG_L4 (1 << 4) 1537 1538 /* mbuf hashing helper routines */ 1539 uint32_t m_ether_tcpip_hash_init(void); 1540 uint32_t m_ether_tcpip_hash(const uint32_t, const struct mbuf *, uint32_t); 1541 uint32_t m_infiniband_tcpip_hash_init(void); 1542 uint32_t m_infiniband_tcpip_hash(const uint32_t, const struct mbuf *, uint32_t); 1543 1544 #ifdef MBUF_PROFILING 1545 void m_profile(struct mbuf *m); 1546 #define M_PROFILE(m) m_profile(m) 1547 #else 1548 #define M_PROFILE(m) 1549 #endif 1550 1551 struct mbufq { 1552 STAILQ_HEAD(, mbuf) mq_head; 1553 int mq_len; 1554 int mq_maxlen; 1555 }; 1556 1557 static inline void 1558 mbufq_init(struct mbufq *mq, int maxlen) 1559 { 1560 1561 STAILQ_INIT(&mq->mq_head); 1562 mq->mq_maxlen = maxlen; 1563 mq->mq_len = 0; 1564 } 1565 1566 static inline struct mbuf * 1567 mbufq_flush(struct mbufq *mq) 1568 { 1569 struct mbuf *m; 1570 1571 m = STAILQ_FIRST(&mq->mq_head); 1572 STAILQ_INIT(&mq->mq_head); 1573 mq->mq_len = 0; 1574 return (m); 1575 } 1576 1577 static inline void 1578 mbufq_drain(struct mbufq *mq) 1579 { 1580 struct mbuf *m, *n; 1581 1582 n = mbufq_flush(mq); 1583 while ((m = n) != NULL) { 1584 n = STAILQ_NEXT(m, m_stailqpkt); 1585 m_freem(m); 1586 } 1587 } 1588 1589 static inline struct mbuf * 1590 mbufq_first(const struct mbufq *mq) 1591 { 1592 1593 return (STAILQ_FIRST(&mq->mq_head)); 1594 } 1595 1596 static inline struct mbuf * 1597 mbufq_last(const struct mbufq *mq) 1598 { 1599 1600 return (STAILQ_LAST(&mq->mq_head, mbuf, m_stailqpkt)); 1601 } 1602 1603 static inline int 1604 mbufq_full(const struct mbufq *mq) 1605 { 1606 1607 return (mq->mq_maxlen > 0 && mq->mq_len >= mq->mq_maxlen); 1608 } 1609 1610 static inline int 1611 mbufq_len(const struct mbufq *mq) 1612 { 1613 1614 return (mq->mq_len); 1615 } 1616 1617 static inline int 1618 mbufq_enqueue(struct mbufq *mq, struct mbuf *m) 1619 { 1620 1621 if (mbufq_full(mq)) 1622 return (ENOBUFS); 1623 STAILQ_INSERT_TAIL(&mq->mq_head, m, m_stailqpkt); 1624 mq->mq_len++; 1625 return (0); 1626 } 1627 1628 static inline struct mbuf * 1629 mbufq_dequeue(struct mbufq *mq) 1630 { 1631 struct mbuf *m; 1632 1633 m = STAILQ_FIRST(&mq->mq_head); 1634 if (m) { 1635 STAILQ_REMOVE_HEAD(&mq->mq_head, m_stailqpkt); 1636 m->m_nextpkt = NULL; 1637 mq->mq_len--; 1638 } 1639 return (m); 1640 } 1641 1642 static inline void 1643 mbufq_prepend(struct mbufq *mq, struct mbuf *m) 1644 { 1645 1646 STAILQ_INSERT_HEAD(&mq->mq_head, m, m_stailqpkt); 1647 mq->mq_len++; 1648 } 1649 1650 /* 1651 * Note: this doesn't enforce the maximum list size for dst. 1652 */ 1653 static inline void 1654 mbufq_concat(struct mbufq *mq_dst, struct mbufq *mq_src) 1655 { 1656 1657 mq_dst->mq_len += mq_src->mq_len; 1658 STAILQ_CONCAT(&mq_dst->mq_head, &mq_src->mq_head); 1659 mq_src->mq_len = 0; 1660 } 1661 1662 #ifdef _SYS_TIMESPEC_H_ 1663 static inline void 1664 mbuf_tstmp2timespec(struct mbuf *m, struct timespec *ts) 1665 { 1666 1667 KASSERT((m->m_flags & M_PKTHDR) != 0, ("mbuf %p no M_PKTHDR", m)); 1668 KASSERT((m->m_flags & (M_TSTMP|M_TSTMP_LRO)) != 0, ("mbuf %p no M_TSTMP or M_TSTMP_LRO", m)); 1669 ts->tv_sec = m->m_pkthdr.rcv_tstmp / 1000000000; 1670 ts->tv_nsec = m->m_pkthdr.rcv_tstmp % 1000000000; 1671 } 1672 #endif 1673 1674 #ifdef DEBUGNET 1675 /* Invoked from the debugnet client code. */ 1676 void debugnet_mbuf_drain(void); 1677 void debugnet_mbuf_start(void); 1678 void debugnet_mbuf_finish(void); 1679 void debugnet_mbuf_reinit(int nmbuf, int nclust, int clsize); 1680 #endif 1681 1682 static inline bool 1683 mbuf_has_tls_session(struct mbuf *m) 1684 { 1685 1686 if (m->m_flags & M_EXTPG) { 1687 if (m->m_epg_tls != NULL) { 1688 return (true); 1689 } 1690 } 1691 return (false); 1692 } 1693 1694 #endif /* _KERNEL */ 1695 #endif /* !_SYS_MBUF_H_ */ 1696