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 <vm/uma.h> 44 #ifdef WITNESS 45 #include <sys/lock.h> 46 #endif 47 #endif 48 49 #ifdef _KERNEL 50 #include <sys/sdt.h> 51 52 #define MBUF_PROBE1(probe, arg0) \ 53 SDT_PROBE1(sdt, , , probe, arg0) 54 #define MBUF_PROBE2(probe, arg0, arg1) \ 55 SDT_PROBE2(sdt, , , probe, arg0, arg1) 56 #define MBUF_PROBE3(probe, arg0, arg1, arg2) \ 57 SDT_PROBE3(sdt, , , probe, arg0, arg1, arg2) 58 #define MBUF_PROBE4(probe, arg0, arg1, arg2, arg3) \ 59 SDT_PROBE4(sdt, , , probe, arg0, arg1, arg2, arg3) 60 #define MBUF_PROBE5(probe, arg0, arg1, arg2, arg3, arg4) \ 61 SDT_PROBE5(sdt, , , probe, arg0, arg1, arg2, arg3, arg4) 62 63 SDT_PROBE_DECLARE(sdt, , , m__init); 64 SDT_PROBE_DECLARE(sdt, , , m__gethdr); 65 SDT_PROBE_DECLARE(sdt, , , m__get); 66 SDT_PROBE_DECLARE(sdt, , , m__getcl); 67 SDT_PROBE_DECLARE(sdt, , , m__clget); 68 SDT_PROBE_DECLARE(sdt, , , m__cljget); 69 SDT_PROBE_DECLARE(sdt, , , m__cljset); 70 SDT_PROBE_DECLARE(sdt, , , m__free); 71 SDT_PROBE_DECLARE(sdt, , , m__freem); 72 73 #endif /* _KERNEL */ 74 75 /* 76 * Mbufs are of a single size, MSIZE (sys/param.h), which includes overhead. 77 * An mbuf may add a single "mbuf cluster" of size MCLBYTES (also in 78 * sys/param.h), which has no additional overhead and is used instead of the 79 * internal data area; this is done when at least MINCLSIZE of data must be 80 * stored. Additionally, it is possible to allocate a separate buffer 81 * externally and attach it to the mbuf in a way similar to that of mbuf 82 * clusters. 83 * 84 * NB: These calculation do not take actual compiler-induced alignment and 85 * padding inside the complete struct mbuf into account. Appropriate 86 * attention is required when changing members of struct mbuf. 87 * 88 * MLEN is data length in a normal mbuf. 89 * MHLEN is data length in an mbuf with pktheader. 90 * MINCLSIZE is a smallest amount of data that should be put into cluster. 91 * 92 * Compile-time assertions in uipc_mbuf.c test these values to ensure that 93 * they are sensible. 94 */ 95 struct mbuf; 96 #define MHSIZE offsetof(struct mbuf, m_dat) 97 #define MPKTHSIZE offsetof(struct mbuf, m_pktdat) 98 #define MLEN ((int)(MSIZE - MHSIZE)) 99 #define MHLEN ((int)(MSIZE - MPKTHSIZE)) 100 #define MINCLSIZE (MHLEN + 1) 101 102 #ifdef _KERNEL 103 /*- 104 * Macro for type conversion: convert mbuf pointer to data pointer of correct 105 * type: 106 * 107 * mtod(m, t) -- Convert mbuf pointer to data pointer of correct type. 108 * mtodo(m, o) -- Same as above but with offset 'o' into data. 109 */ 110 #define mtod(m, t) ((t)((m)->m_data)) 111 #define mtodo(m, o) ((void *)(((m)->m_data) + (o))) 112 113 /* 114 * Argument structure passed to UMA routines during mbuf and packet 115 * allocations. 116 */ 117 struct mb_args { 118 int flags; /* Flags for mbuf being allocated */ 119 short type; /* Type of mbuf being allocated */ 120 }; 121 #endif /* _KERNEL */ 122 123 /* 124 * Packet tag structure (see below for details). 125 */ 126 struct m_tag { 127 SLIST_ENTRY(m_tag) m_tag_link; /* List of packet tags */ 128 u_int16_t m_tag_id; /* Tag ID */ 129 u_int16_t m_tag_len; /* Length of data */ 130 u_int32_t m_tag_cookie; /* ABI/Module ID */ 131 void (*m_tag_free)(struct m_tag *); 132 }; 133 134 /* 135 * Static network interface owned tag. 136 * Allocated through ifp->if_snd_tag_alloc(). 137 */ 138 struct m_snd_tag { 139 struct ifnet *ifp; /* network interface tag belongs to */ 140 }; 141 142 /* 143 * Record/packet header in first mbuf of chain; valid only if M_PKTHDR is set. 144 * Size ILP32: 48 145 * LP64: 56 146 * Compile-time assertions in uipc_mbuf.c test these values to ensure that 147 * they are correct. 148 */ 149 struct pkthdr { 150 union { 151 struct m_snd_tag *snd_tag; /* send tag, if any */ 152 struct ifnet *rcvif; /* rcv interface */ 153 }; 154 SLIST_HEAD(packet_tags, m_tag) tags; /* list of packet tags */ 155 int32_t len; /* total packet length */ 156 157 /* Layer crossing persistent information. */ 158 uint32_t flowid; /* packet's 4-tuple system */ 159 uint32_t csum_flags; /* checksum and offload features */ 160 uint16_t fibnum; /* this packet should use this fib */ 161 uint8_t cosqos; /* class/quality of service */ 162 uint8_t rsstype; /* hash type */ 163 union { 164 uint64_t rcv_tstmp; /* timestamp in ns */ 165 struct { 166 uint8_t l2hlen; /* layer 2 hdr len */ 167 uint8_t l3hlen; /* layer 3 hdr len */ 168 uint8_t l4hlen; /* layer 4 hdr len */ 169 uint8_t l5hlen; /* layer 5 hdr len */ 170 uint32_t spare; 171 }; 172 }; 173 union { 174 uint8_t eight[8]; 175 uint16_t sixteen[4]; 176 uint32_t thirtytwo[2]; 177 uint64_t sixtyfour[1]; 178 uintptr_t unintptr[1]; 179 void *ptr; 180 } PH_per; 181 182 /* Layer specific non-persistent local storage for reassembly, etc. */ 183 union { 184 uint8_t eight[8]; 185 uint16_t sixteen[4]; 186 uint32_t thirtytwo[2]; 187 uint64_t sixtyfour[1]; 188 uintptr_t unintptr[1]; 189 void *ptr; 190 } PH_loc; 191 }; 192 #define ether_vtag PH_per.sixteen[0] 193 #define PH_vt PH_per 194 #define vt_nrecs sixteen[0] 195 #define tso_segsz PH_per.sixteen[1] 196 #define lro_nsegs tso_segsz 197 #define csum_phsum PH_per.sixteen[2] 198 #define csum_data PH_per.thirtytwo[1] 199 200 /* 201 * Description of external storage mapped into mbuf; valid only if M_EXT is 202 * set. 203 * Size ILP32: 28 204 * LP64: 48 205 * Compile-time assertions in uipc_mbuf.c test these values to ensure that 206 * they are correct. 207 */ 208 typedef void m_ext_free_t(struct mbuf *); 209 struct m_ext { 210 union { 211 /* 212 * If EXT_FLAG_EMBREF is set, then we use refcount in the 213 * mbuf, the 'ext_count' member. Otherwise, we have a 214 * shadow copy and we use pointer 'ext_cnt'. The original 215 * mbuf is responsible to carry the pointer to free routine 216 * and its arguments. They aren't copied into shadows in 217 * mb_dupcl() to avoid dereferencing next cachelines. 218 */ 219 volatile u_int ext_count; 220 volatile u_int *ext_cnt; 221 }; 222 char *ext_buf; /* start of buffer */ 223 uint32_t ext_size; /* size of buffer, for ext_free */ 224 uint32_t ext_type:8, /* type of external storage */ 225 ext_flags:24; /* external storage mbuf flags */ 226 /* 227 * Fields below store the free context for the external storage. 228 * They are valid only in the refcount carrying mbuf, the one with 229 * EXT_FLAG_EMBREF flag, with exclusion for EXT_EXTREF type, where 230 * the free context is copied into all mbufs that use same external 231 * storage. 232 */ 233 #define m_ext_copylen offsetof(struct m_ext, ext_free) 234 m_ext_free_t *ext_free; /* free routine if not the usual */ 235 void *ext_arg1; /* optional argument pointer */ 236 void *ext_arg2; /* optional argument pointer */ 237 }; 238 239 /* 240 * The core of the mbuf object along with some shortcut defines for practical 241 * purposes. 242 */ 243 struct mbuf { 244 /* 245 * Header present at the beginning of every mbuf. 246 * Size ILP32: 24 247 * LP64: 32 248 * Compile-time assertions in uipc_mbuf.c test these values to ensure 249 * that they are correct. 250 */ 251 union { /* next buffer in chain */ 252 struct mbuf *m_next; 253 SLIST_ENTRY(mbuf) m_slist; 254 STAILQ_ENTRY(mbuf) m_stailq; 255 }; 256 union { /* next chain in queue/record */ 257 struct mbuf *m_nextpkt; 258 SLIST_ENTRY(mbuf) m_slistpkt; 259 STAILQ_ENTRY(mbuf) m_stailqpkt; 260 }; 261 caddr_t m_data; /* location of data */ 262 int32_t m_len; /* amount of data in this mbuf */ 263 uint32_t m_type:8, /* type of data in this mbuf */ 264 m_flags:24; /* flags; see below */ 265 #if !defined(__LP64__) 266 uint32_t m_pad; /* pad for 64bit alignment */ 267 #endif 268 269 /* 270 * A set of optional headers (packet header, external storage header) 271 * and internal data storage. Historically, these arrays were sized 272 * to MHLEN (space left after a packet header) and MLEN (space left 273 * after only a regular mbuf header); they are now variable size in 274 * order to support future work on variable-size mbufs. 275 */ 276 union { 277 struct { 278 struct pkthdr m_pkthdr; /* M_PKTHDR set */ 279 union { 280 struct m_ext m_ext; /* M_EXT set */ 281 char m_pktdat[0]; 282 }; 283 }; 284 char m_dat[0]; /* !M_PKTHDR, !M_EXT */ 285 }; 286 }; 287 288 /* 289 * mbuf flags of global significance and layer crossing. 290 * Those of only protocol/layer specific significance are to be mapped 291 * to M_PROTO[1-12] and cleared at layer handoff boundaries. 292 * NB: Limited to the lower 24 bits. 293 */ 294 #define M_EXT 0x00000001 /* has associated external storage */ 295 #define M_PKTHDR 0x00000002 /* start of record */ 296 #define M_EOR 0x00000004 /* end of record */ 297 #define M_RDONLY 0x00000008 /* associated data is marked read-only */ 298 #define M_BCAST 0x00000010 /* send/received as link-level broadcast */ 299 #define M_MCAST 0x00000020 /* send/received as link-level multicast */ 300 #define M_PROMISC 0x00000040 /* packet was not for us */ 301 #define M_VLANTAG 0x00000080 /* ether_vtag is valid */ 302 #define M_UNUSED_8 0x00000100 /* --available-- */ 303 #define M_NOFREE 0x00000200 /* do not free mbuf, embedded in cluster */ 304 #define M_TSTMP 0x00000400 /* rcv_tstmp field is valid */ 305 #define M_TSTMP_HPREC 0x00000800 /* rcv_tstmp is high-prec, typically 306 hw-stamped on port (useful for IEEE 1588 307 and 802.1AS) */ 308 309 #define M_PROTO1 0x00001000 /* protocol-specific */ 310 #define M_PROTO2 0x00002000 /* protocol-specific */ 311 #define M_PROTO3 0x00004000 /* protocol-specific */ 312 #define M_PROTO4 0x00008000 /* protocol-specific */ 313 #define M_PROTO5 0x00010000 /* protocol-specific */ 314 #define M_PROTO6 0x00020000 /* protocol-specific */ 315 #define M_PROTO7 0x00040000 /* protocol-specific */ 316 #define M_PROTO8 0x00080000 /* protocol-specific */ 317 #define M_PROTO9 0x00100000 /* protocol-specific */ 318 #define M_PROTO10 0x00200000 /* protocol-specific */ 319 #define M_PROTO11 0x00400000 /* protocol-specific */ 320 #define M_PROTO12 0x00800000 /* protocol-specific */ 321 322 #define MB_DTOR_SKIP 0x1 /* don't pollute the cache by touching a freed mbuf */ 323 324 /* 325 * Flags to purge when crossing layers. 326 */ 327 #define M_PROTOFLAGS \ 328 (M_PROTO1|M_PROTO2|M_PROTO3|M_PROTO4|M_PROTO5|M_PROTO6|M_PROTO7|M_PROTO8|\ 329 M_PROTO9|M_PROTO10|M_PROTO11|M_PROTO12) 330 331 /* 332 * Flags preserved when copying m_pkthdr. 333 */ 334 #define M_COPYFLAGS \ 335 (M_PKTHDR|M_EOR|M_RDONLY|M_BCAST|M_MCAST|M_PROMISC|M_VLANTAG|M_TSTMP| \ 336 M_TSTMP_HPREC|M_PROTOFLAGS) 337 338 /* 339 * Mbuf flag description for use with printf(9) %b identifier. 340 */ 341 #define M_FLAG_BITS \ 342 "\20\1M_EXT\2M_PKTHDR\3M_EOR\4M_RDONLY\5M_BCAST\6M_MCAST" \ 343 "\7M_PROMISC\10M_VLANTAG\13M_TSTMP\14M_TSTMP_HPREC" 344 #define M_FLAG_PROTOBITS \ 345 "\15M_PROTO1\16M_PROTO2\17M_PROTO3\20M_PROTO4\21M_PROTO5" \ 346 "\22M_PROTO6\23M_PROTO7\24M_PROTO8\25M_PROTO9\26M_PROTO10" \ 347 "\27M_PROTO11\30M_PROTO12" 348 #define M_FLAG_PRINTF (M_FLAG_BITS M_FLAG_PROTOBITS) 349 350 /* 351 * Network interface cards are able to hash protocol fields (such as IPv4 352 * addresses and TCP port numbers) classify packets into flows. These flows 353 * can then be used to maintain ordering while delivering packets to the OS 354 * via parallel input queues, as well as to provide a stateless affinity 355 * model. NIC drivers can pass up the hash via m->m_pkthdr.flowid, and set 356 * m_flag fields to indicate how the hash should be interpreted by the 357 * network stack. 358 * 359 * Most NICs support RSS, which provides ordering and explicit affinity, and 360 * use the hash m_flag bits to indicate what header fields were covered by 361 * the hash. M_HASHTYPE_OPAQUE and M_HASHTYPE_OPAQUE_HASH can be set by non- 362 * RSS cards or configurations that provide an opaque flow identifier, allowing 363 * for ordering and distribution without explicit affinity. Additionally, 364 * M_HASHTYPE_OPAQUE_HASH indicates that the flow identifier has hash 365 * properties. 366 * 367 * The meaning of the IPV6_EX suffix: 368 * "o Home address from the home address option in the IPv6 destination 369 * options header. If the extension header is not present, use the Source 370 * IPv6 Address. 371 * o IPv6 address that is contained in the Routing-Header-Type-2 from the 372 * associated extension header. If the extension header is not present, 373 * use the Destination IPv6 Address." 374 * Quoted from: 375 * https://docs.microsoft.com/en-us/windows-hardware/drivers/network/rss-hashing-types#ndishashipv6ex 376 */ 377 #define M_HASHTYPE_HASHPROP 0x80 /* has hash properties */ 378 #define M_HASHTYPE_HASH(t) (M_HASHTYPE_HASHPROP | (t)) 379 /* Microsoft RSS standard hash types */ 380 #define M_HASHTYPE_NONE 0 381 #define M_HASHTYPE_RSS_IPV4 M_HASHTYPE_HASH(1) /* IPv4 2-tuple */ 382 #define M_HASHTYPE_RSS_TCP_IPV4 M_HASHTYPE_HASH(2) /* TCPv4 4-tuple */ 383 #define M_HASHTYPE_RSS_IPV6 M_HASHTYPE_HASH(3) /* IPv6 2-tuple */ 384 #define M_HASHTYPE_RSS_TCP_IPV6 M_HASHTYPE_HASH(4) /* TCPv6 4-tuple */ 385 #define M_HASHTYPE_RSS_IPV6_EX M_HASHTYPE_HASH(5) /* IPv6 2-tuple + 386 * ext hdrs */ 387 #define M_HASHTYPE_RSS_TCP_IPV6_EX M_HASHTYPE_HASH(6) /* TCPv6 4-tuple + 388 * ext hdrs */ 389 #define M_HASHTYPE_RSS_UDP_IPV4 M_HASHTYPE_HASH(7) /* IPv4 UDP 4-tuple*/ 390 #define M_HASHTYPE_RSS_UDP_IPV6 M_HASHTYPE_HASH(9) /* IPv6 UDP 4-tuple*/ 391 #define M_HASHTYPE_RSS_UDP_IPV6_EX M_HASHTYPE_HASH(10)/* IPv6 UDP 4-tuple + 392 * ext hdrs */ 393 394 #define M_HASHTYPE_OPAQUE 63 /* ordering, not affinity */ 395 #define M_HASHTYPE_OPAQUE_HASH M_HASHTYPE_HASH(M_HASHTYPE_OPAQUE) 396 /* ordering+hash, not affinity*/ 397 398 #define M_HASHTYPE_CLEAR(m) ((m)->m_pkthdr.rsstype = 0) 399 #define M_HASHTYPE_GET(m) ((m)->m_pkthdr.rsstype) 400 #define M_HASHTYPE_SET(m, v) ((m)->m_pkthdr.rsstype = (v)) 401 #define M_HASHTYPE_TEST(m, v) (M_HASHTYPE_GET(m) == (v)) 402 #define M_HASHTYPE_ISHASH(m) (M_HASHTYPE_GET(m) & M_HASHTYPE_HASHPROP) 403 404 /* 405 * COS/QOS class and quality of service tags. 406 * It uses DSCP code points as base. 407 */ 408 #define QOS_DSCP_CS0 0x00 409 #define QOS_DSCP_DEF QOS_DSCP_CS0 410 #define QOS_DSCP_CS1 0x20 411 #define QOS_DSCP_AF11 0x28 412 #define QOS_DSCP_AF12 0x30 413 #define QOS_DSCP_AF13 0x38 414 #define QOS_DSCP_CS2 0x40 415 #define QOS_DSCP_AF21 0x48 416 #define QOS_DSCP_AF22 0x50 417 #define QOS_DSCP_AF23 0x58 418 #define QOS_DSCP_CS3 0x60 419 #define QOS_DSCP_AF31 0x68 420 #define QOS_DSCP_AF32 0x70 421 #define QOS_DSCP_AF33 0x78 422 #define QOS_DSCP_CS4 0x80 423 #define QOS_DSCP_AF41 0x88 424 #define QOS_DSCP_AF42 0x90 425 #define QOS_DSCP_AF43 0x98 426 #define QOS_DSCP_CS5 0xa0 427 #define QOS_DSCP_EF 0xb8 428 #define QOS_DSCP_CS6 0xc0 429 #define QOS_DSCP_CS7 0xe0 430 431 /* 432 * External mbuf storage buffer types. 433 */ 434 #define EXT_CLUSTER 1 /* mbuf cluster */ 435 #define EXT_SFBUF 2 /* sendfile(2)'s sf_buf */ 436 #define EXT_JUMBOP 3 /* jumbo cluster page sized */ 437 #define EXT_JUMBO9 4 /* jumbo cluster 9216 bytes */ 438 #define EXT_JUMBO16 5 /* jumbo cluster 16184 bytes */ 439 #define EXT_PACKET 6 /* mbuf+cluster from packet zone */ 440 #define EXT_MBUF 7 /* external mbuf reference */ 441 442 #define EXT_VENDOR1 224 /* for vendor-internal use */ 443 #define EXT_VENDOR2 225 /* for vendor-internal use */ 444 #define EXT_VENDOR3 226 /* for vendor-internal use */ 445 #define EXT_VENDOR4 227 /* for vendor-internal use */ 446 447 #define EXT_EXP1 244 /* for experimental use */ 448 #define EXT_EXP2 245 /* for experimental use */ 449 #define EXT_EXP3 246 /* for experimental use */ 450 #define EXT_EXP4 247 /* for experimental use */ 451 452 #define EXT_NET_DRV 252 /* custom ext_buf provided by net driver(s) */ 453 #define EXT_MOD_TYPE 253 /* custom module's ext_buf type */ 454 #define EXT_DISPOSABLE 254 /* can throw this buffer away w/page flipping */ 455 #define EXT_EXTREF 255 /* has externally maintained ext_cnt ptr */ 456 457 /* 458 * Flags for external mbuf buffer types. 459 * NB: limited to the lower 24 bits. 460 */ 461 #define EXT_FLAG_EMBREF 0x000001 /* embedded ext_count */ 462 #define EXT_FLAG_EXTREF 0x000002 /* external ext_cnt, notyet */ 463 464 #define EXT_FLAG_NOFREE 0x000010 /* don't free mbuf to pool, notyet */ 465 466 #define EXT_FLAG_VENDOR1 0x010000 /* These flags are vendor */ 467 #define EXT_FLAG_VENDOR2 0x020000 /* or submodule specific, */ 468 #define EXT_FLAG_VENDOR3 0x040000 /* not used by mbuf code. */ 469 #define EXT_FLAG_VENDOR4 0x080000 /* Set/read by submodule. */ 470 471 #define EXT_FLAG_EXP1 0x100000 /* for experimental use */ 472 #define EXT_FLAG_EXP2 0x200000 /* for experimental use */ 473 #define EXT_FLAG_EXP3 0x400000 /* for experimental use */ 474 #define EXT_FLAG_EXP4 0x800000 /* for experimental use */ 475 476 /* 477 * EXT flag description for use with printf(9) %b identifier. 478 */ 479 #define EXT_FLAG_BITS \ 480 "\20\1EXT_FLAG_EMBREF\2EXT_FLAG_EXTREF\5EXT_FLAG_NOFREE" \ 481 "\21EXT_FLAG_VENDOR1\22EXT_FLAG_VENDOR2\23EXT_FLAG_VENDOR3" \ 482 "\24EXT_FLAG_VENDOR4\25EXT_FLAG_EXP1\26EXT_FLAG_EXP2\27EXT_FLAG_EXP3" \ 483 "\30EXT_FLAG_EXP4" 484 485 /* 486 * Flags indicating checksum, segmentation and other offload work to be 487 * done, or already done, by hardware or lower layers. It is split into 488 * separate inbound and outbound flags. 489 * 490 * Outbound flags that are set by upper protocol layers requesting lower 491 * layers, or ideally the hardware, to perform these offloading tasks. 492 * For outbound packets this field and its flags can be directly tested 493 * against ifnet if_hwassist. 494 */ 495 #define CSUM_IP 0x00000001 /* IP header checksum offload */ 496 #define CSUM_IP_UDP 0x00000002 /* UDP checksum offload */ 497 #define CSUM_IP_TCP 0x00000004 /* TCP checksum offload */ 498 #define CSUM_IP_SCTP 0x00000008 /* SCTP checksum offload */ 499 #define CSUM_IP_TSO 0x00000010 /* TCP segmentation offload */ 500 #define CSUM_IP_ISCSI 0x00000020 /* iSCSI checksum offload */ 501 502 #define CSUM_IP6_UDP 0x00000200 /* UDP checksum offload */ 503 #define CSUM_IP6_TCP 0x00000400 /* TCP checksum offload */ 504 #define CSUM_IP6_SCTP 0x00000800 /* SCTP checksum offload */ 505 #define CSUM_IP6_TSO 0x00001000 /* TCP segmentation offload */ 506 #define CSUM_IP6_ISCSI 0x00002000 /* iSCSI checksum offload */ 507 508 /* Inbound checksum support where the checksum was verified by hardware. */ 509 #define CSUM_L3_CALC 0x01000000 /* calculated layer 3 csum */ 510 #define CSUM_L3_VALID 0x02000000 /* checksum is correct */ 511 #define CSUM_L4_CALC 0x04000000 /* calculated layer 4 csum */ 512 #define CSUM_L4_VALID 0x08000000 /* checksum is correct */ 513 #define CSUM_L5_CALC 0x10000000 /* calculated layer 5 csum */ 514 #define CSUM_L5_VALID 0x20000000 /* checksum is correct */ 515 #define CSUM_COALESCED 0x40000000 /* contains merged segments */ 516 517 /* 518 * CSUM flag description for use with printf(9) %b identifier. 519 */ 520 #define CSUM_BITS \ 521 "\20\1CSUM_IP\2CSUM_IP_UDP\3CSUM_IP_TCP\4CSUM_IP_SCTP\5CSUM_IP_TSO" \ 522 "\6CSUM_IP_ISCSI" \ 523 "\12CSUM_IP6_UDP\13CSUM_IP6_TCP\14CSUM_IP6_SCTP\15CSUM_IP6_TSO" \ 524 "\16CSUM_IP6_ISCSI" \ 525 "\31CSUM_L3_CALC\32CSUM_L3_VALID\33CSUM_L4_CALC\34CSUM_L4_VALID" \ 526 "\35CSUM_L5_CALC\36CSUM_L5_VALID\37CSUM_COALESCED" 527 528 /* CSUM flags compatibility mappings. */ 529 #define CSUM_IP_CHECKED CSUM_L3_CALC 530 #define CSUM_IP_VALID CSUM_L3_VALID 531 #define CSUM_DATA_VALID CSUM_L4_VALID 532 #define CSUM_PSEUDO_HDR CSUM_L4_CALC 533 #define CSUM_SCTP_VALID CSUM_L4_VALID 534 #define CSUM_DELAY_DATA (CSUM_TCP|CSUM_UDP) 535 #define CSUM_DELAY_IP CSUM_IP /* Only v4, no v6 IP hdr csum */ 536 #define CSUM_DELAY_DATA_IPV6 (CSUM_TCP_IPV6|CSUM_UDP_IPV6) 537 #define CSUM_DATA_VALID_IPV6 CSUM_DATA_VALID 538 #define CSUM_TCP CSUM_IP_TCP 539 #define CSUM_UDP CSUM_IP_UDP 540 #define CSUM_SCTP CSUM_IP_SCTP 541 #define CSUM_TSO (CSUM_IP_TSO|CSUM_IP6_TSO) 542 #define CSUM_UDP_IPV6 CSUM_IP6_UDP 543 #define CSUM_TCP_IPV6 CSUM_IP6_TCP 544 #define CSUM_SCTP_IPV6 CSUM_IP6_SCTP 545 546 /* 547 * mbuf types describing the content of the mbuf (including external storage). 548 */ 549 #define MT_NOTMBUF 0 /* USED INTERNALLY ONLY! Object is not mbuf */ 550 #define MT_DATA 1 /* dynamic (data) allocation */ 551 #define MT_HEADER MT_DATA /* packet header, use M_PKTHDR instead */ 552 553 #define MT_VENDOR1 4 /* for vendor-internal use */ 554 #define MT_VENDOR2 5 /* for vendor-internal use */ 555 #define MT_VENDOR3 6 /* for vendor-internal use */ 556 #define MT_VENDOR4 7 /* for vendor-internal use */ 557 558 #define MT_SONAME 8 /* socket name */ 559 560 #define MT_EXP1 9 /* for experimental use */ 561 #define MT_EXP2 10 /* for experimental use */ 562 #define MT_EXP3 11 /* for experimental use */ 563 #define MT_EXP4 12 /* for experimental use */ 564 565 #define MT_CONTROL 14 /* extra-data protocol message */ 566 #define MT_OOBDATA 15 /* expedited data */ 567 #define MT_NTYPES 16 /* number of mbuf types for mbtypes[] */ 568 569 #define MT_NOINIT 255 /* Not a type but a flag to allocate 570 a non-initialized mbuf */ 571 572 /* 573 * String names of mbuf-related UMA(9) and malloc(9) types. Exposed to 574 * !_KERNEL so that monitoring tools can look up the zones with 575 * libmemstat(3). 576 */ 577 #define MBUF_MEM_NAME "mbuf" 578 #define MBUF_CLUSTER_MEM_NAME "mbuf_cluster" 579 #define MBUF_PACKET_MEM_NAME "mbuf_packet" 580 #define MBUF_JUMBOP_MEM_NAME "mbuf_jumbo_page" 581 #define MBUF_JUMBO9_MEM_NAME "mbuf_jumbo_9k" 582 #define MBUF_JUMBO16_MEM_NAME "mbuf_jumbo_16k" 583 #define MBUF_TAG_MEM_NAME "mbuf_tag" 584 #define MBUF_EXTREFCNT_MEM_NAME "mbuf_ext_refcnt" 585 586 #ifdef _KERNEL 587 588 #ifdef WITNESS 589 #define MBUF_CHECKSLEEP(how) do { \ 590 if (how == M_WAITOK) \ 591 WITNESS_WARN(WARN_GIANTOK | WARN_SLEEPOK, NULL, \ 592 "Sleeping in \"%s\"", __func__); \ 593 } while (0) 594 #else 595 #define MBUF_CHECKSLEEP(how) 596 #endif 597 598 /* 599 * Network buffer allocation API 600 * 601 * The rest of it is defined in kern/kern_mbuf.c 602 */ 603 extern uma_zone_t zone_mbuf; 604 extern uma_zone_t zone_clust; 605 extern uma_zone_t zone_pack; 606 extern uma_zone_t zone_jumbop; 607 extern uma_zone_t zone_jumbo9; 608 extern uma_zone_t zone_jumbo16; 609 610 void mb_dupcl(struct mbuf *, struct mbuf *); 611 void mb_free_ext(struct mbuf *); 612 void m_adj(struct mbuf *, int); 613 int m_apply(struct mbuf *, int, int, 614 int (*)(void *, void *, u_int), void *); 615 int m_append(struct mbuf *, int, c_caddr_t); 616 void m_cat(struct mbuf *, struct mbuf *); 617 void m_catpkt(struct mbuf *, struct mbuf *); 618 int m_clget(struct mbuf *m, int how); 619 void *m_cljget(struct mbuf *m, int how, int size); 620 struct mbuf *m_collapse(struct mbuf *, int, int); 621 void m_copyback(struct mbuf *, int, int, c_caddr_t); 622 void m_copydata(const struct mbuf *, int, int, caddr_t); 623 struct mbuf *m_copym(struct mbuf *, int, int, int); 624 struct mbuf *m_copypacket(struct mbuf *, int); 625 void m_copy_pkthdr(struct mbuf *, struct mbuf *); 626 struct mbuf *m_copyup(struct mbuf *, int, int); 627 struct mbuf *m_defrag(struct mbuf *, int); 628 void m_demote_pkthdr(struct mbuf *); 629 void m_demote(struct mbuf *, int, int); 630 struct mbuf *m_devget(char *, int, int, struct ifnet *, 631 void (*)(char *, caddr_t, u_int)); 632 struct mbuf *m_dup(const struct mbuf *, int); 633 int m_dup_pkthdr(struct mbuf *, const struct mbuf *, int); 634 void m_extadd(struct mbuf *, char *, u_int, m_ext_free_t, 635 void *, void *, int, int); 636 u_int m_fixhdr(struct mbuf *); 637 struct mbuf *m_fragment(struct mbuf *, int, int); 638 void m_freem(struct mbuf *); 639 struct mbuf *m_get2(int, int, short, int); 640 struct mbuf *m_getjcl(int, short, int, int); 641 struct mbuf *m_getm2(struct mbuf *, int, int, short, int); 642 struct mbuf *m_getptr(struct mbuf *, int, int *); 643 u_int m_length(struct mbuf *, struct mbuf **); 644 int m_mbuftouio(struct uio *, const struct mbuf *, int); 645 void m_move_pkthdr(struct mbuf *, struct mbuf *); 646 int m_pkthdr_init(struct mbuf *, int); 647 struct mbuf *m_prepend(struct mbuf *, int, int); 648 void m_print(const struct mbuf *, int); 649 struct mbuf *m_pulldown(struct mbuf *, int, int, int *); 650 struct mbuf *m_pullup(struct mbuf *, int); 651 int m_sanity(struct mbuf *, int); 652 struct mbuf *m_split(struct mbuf *, int, int); 653 struct mbuf *m_uiotombuf(struct uio *, int, int, int, int); 654 struct mbuf *m_unshare(struct mbuf *, int); 655 656 static __inline int 657 m_gettype(int size) 658 { 659 int type; 660 661 switch (size) { 662 case MSIZE: 663 type = EXT_MBUF; 664 break; 665 case MCLBYTES: 666 type = EXT_CLUSTER; 667 break; 668 #if MJUMPAGESIZE != MCLBYTES 669 case MJUMPAGESIZE: 670 type = EXT_JUMBOP; 671 break; 672 #endif 673 case MJUM9BYTES: 674 type = EXT_JUMBO9; 675 break; 676 case MJUM16BYTES: 677 type = EXT_JUMBO16; 678 break; 679 default: 680 panic("%s: invalid cluster size %d", __func__, size); 681 } 682 683 return (type); 684 } 685 686 /* 687 * Associated an external reference counted buffer with an mbuf. 688 */ 689 static __inline void 690 m_extaddref(struct mbuf *m, char *buf, u_int size, u_int *ref_cnt, 691 m_ext_free_t freef, void *arg1, void *arg2) 692 { 693 694 KASSERT(ref_cnt != NULL, ("%s: ref_cnt not provided", __func__)); 695 696 atomic_add_int(ref_cnt, 1); 697 m->m_flags |= M_EXT; 698 m->m_ext.ext_buf = buf; 699 m->m_ext.ext_cnt = ref_cnt; 700 m->m_data = m->m_ext.ext_buf; 701 m->m_ext.ext_size = size; 702 m->m_ext.ext_free = freef; 703 m->m_ext.ext_arg1 = arg1; 704 m->m_ext.ext_arg2 = arg2; 705 m->m_ext.ext_type = EXT_EXTREF; 706 m->m_ext.ext_flags = 0; 707 } 708 709 static __inline uma_zone_t 710 m_getzone(int size) 711 { 712 uma_zone_t zone; 713 714 switch (size) { 715 case MCLBYTES: 716 zone = zone_clust; 717 break; 718 #if MJUMPAGESIZE != MCLBYTES 719 case MJUMPAGESIZE: 720 zone = zone_jumbop; 721 break; 722 #endif 723 case MJUM9BYTES: 724 zone = zone_jumbo9; 725 break; 726 case MJUM16BYTES: 727 zone = zone_jumbo16; 728 break; 729 default: 730 panic("%s: invalid cluster size %d", __func__, size); 731 } 732 733 return (zone); 734 } 735 736 /* 737 * Initialize an mbuf with linear storage. 738 * 739 * Inline because the consumer text overhead will be roughly the same to 740 * initialize or call a function with this many parameters and M_PKTHDR 741 * should go away with constant propagation for !MGETHDR. 742 */ 743 static __inline int 744 m_init(struct mbuf *m, int how, short type, int flags) 745 { 746 int error; 747 748 m->m_next = NULL; 749 m->m_nextpkt = NULL; 750 m->m_data = m->m_dat; 751 m->m_len = 0; 752 m->m_flags = flags; 753 m->m_type = type; 754 if (flags & M_PKTHDR) 755 error = m_pkthdr_init(m, how); 756 else 757 error = 0; 758 759 MBUF_PROBE5(m__init, m, how, type, flags, error); 760 return (error); 761 } 762 763 static __inline struct mbuf * 764 m_get(int how, short type) 765 { 766 struct mbuf *m; 767 struct mb_args args; 768 769 args.flags = 0; 770 args.type = type; 771 m = uma_zalloc_arg(zone_mbuf, &args, how); 772 MBUF_PROBE3(m__get, how, type, m); 773 return (m); 774 } 775 776 static __inline struct mbuf * 777 m_gethdr(int how, short type) 778 { 779 struct mbuf *m; 780 struct mb_args args; 781 782 args.flags = M_PKTHDR; 783 args.type = type; 784 m = uma_zalloc_arg(zone_mbuf, &args, how); 785 MBUF_PROBE3(m__gethdr, how, type, m); 786 return (m); 787 } 788 789 static __inline struct mbuf * 790 m_getcl(int how, short type, int flags) 791 { 792 struct mbuf *m; 793 struct mb_args args; 794 795 args.flags = flags; 796 args.type = type; 797 m = uma_zalloc_arg(zone_pack, &args, how); 798 MBUF_PROBE4(m__getcl, how, type, flags, m); 799 return (m); 800 } 801 802 /* 803 * XXX: m_cljset() is a dangerous API. One must attach only a new, 804 * unreferenced cluster to an mbuf(9). It is not possible to assert 805 * that, so care can be taken only by users of the API. 806 */ 807 static __inline void 808 m_cljset(struct mbuf *m, void *cl, int type) 809 { 810 int size; 811 812 switch (type) { 813 case EXT_CLUSTER: 814 size = MCLBYTES; 815 break; 816 #if MJUMPAGESIZE != MCLBYTES 817 case EXT_JUMBOP: 818 size = MJUMPAGESIZE; 819 break; 820 #endif 821 case EXT_JUMBO9: 822 size = MJUM9BYTES; 823 break; 824 case EXT_JUMBO16: 825 size = MJUM16BYTES; 826 break; 827 default: 828 panic("%s: unknown cluster type %d", __func__, type); 829 break; 830 } 831 832 m->m_data = m->m_ext.ext_buf = cl; 833 m->m_ext.ext_free = m->m_ext.ext_arg1 = m->m_ext.ext_arg2 = NULL; 834 m->m_ext.ext_size = size; 835 m->m_ext.ext_type = type; 836 m->m_ext.ext_flags = EXT_FLAG_EMBREF; 837 m->m_ext.ext_count = 1; 838 m->m_flags |= M_EXT; 839 MBUF_PROBE3(m__cljset, m, cl, type); 840 } 841 842 static __inline void 843 m_chtype(struct mbuf *m, short new_type) 844 { 845 846 m->m_type = new_type; 847 } 848 849 static __inline void 850 m_clrprotoflags(struct mbuf *m) 851 { 852 853 while (m) { 854 m->m_flags &= ~M_PROTOFLAGS; 855 m = m->m_next; 856 } 857 } 858 859 static __inline struct mbuf * 860 m_last(struct mbuf *m) 861 { 862 863 while (m->m_next) 864 m = m->m_next; 865 return (m); 866 } 867 868 static inline u_int 869 m_extrefcnt(struct mbuf *m) 870 { 871 872 KASSERT(m->m_flags & M_EXT, ("%s: M_EXT missing", __func__)); 873 874 return ((m->m_ext.ext_flags & EXT_FLAG_EMBREF) ? m->m_ext.ext_count : 875 *m->m_ext.ext_cnt); 876 } 877 878 /* 879 * mbuf, cluster, and external object allocation macros (for compatibility 880 * purposes). 881 */ 882 #define M_MOVE_PKTHDR(to, from) m_move_pkthdr((to), (from)) 883 #define MGET(m, how, type) ((m) = m_get((how), (type))) 884 #define MGETHDR(m, how, type) ((m) = m_gethdr((how), (type))) 885 #define MCLGET(m, how) m_clget((m), (how)) 886 #define MEXTADD(m, buf, size, free, arg1, arg2, flags, type) \ 887 m_extadd((m), (char *)(buf), (size), (free), (arg1), (arg2), \ 888 (flags), (type)) 889 #define m_getm(m, len, how, type) \ 890 m_getm2((m), (len), (how), (type), M_PKTHDR) 891 892 /* 893 * Evaluate TRUE if it's safe to write to the mbuf m's data region (this can 894 * be both the local data payload, or an external buffer area, depending on 895 * whether M_EXT is set). 896 */ 897 #define M_WRITABLE(m) (!((m)->m_flags & M_RDONLY) && \ 898 (!(((m)->m_flags & M_EXT)) || \ 899 (m_extrefcnt(m) == 1))) 900 901 /* Check if the supplied mbuf has a packet header, or else panic. */ 902 #define M_ASSERTPKTHDR(m) \ 903 KASSERT((m) != NULL && (m)->m_flags & M_PKTHDR, \ 904 ("%s: no mbuf packet header!", __func__)) 905 906 /* 907 * Ensure that the supplied mbuf is a valid, non-free mbuf. 908 * 909 * XXX: Broken at the moment. Need some UMA magic to make it work again. 910 */ 911 #define M_ASSERTVALID(m) \ 912 KASSERT((((struct mbuf *)m)->m_flags & 0) == 0, \ 913 ("%s: attempted use of a free mbuf!", __func__)) 914 915 /* 916 * Return the address of the start of the buffer associated with an mbuf, 917 * handling external storage, packet-header mbufs, and regular data mbufs. 918 */ 919 #define M_START(m) \ 920 (((m)->m_flags & M_EXT) ? (m)->m_ext.ext_buf : \ 921 ((m)->m_flags & M_PKTHDR) ? &(m)->m_pktdat[0] : \ 922 &(m)->m_dat[0]) 923 924 /* 925 * Return the size of the buffer associated with an mbuf, handling external 926 * storage, packet-header mbufs, and regular data mbufs. 927 */ 928 #define M_SIZE(m) \ 929 (((m)->m_flags & M_EXT) ? (m)->m_ext.ext_size : \ 930 ((m)->m_flags & M_PKTHDR) ? MHLEN : \ 931 MLEN) 932 933 /* 934 * Set the m_data pointer of a newly allocated mbuf to place an object of the 935 * specified size at the end of the mbuf, longword aligned. 936 * 937 * NB: Historically, we had M_ALIGN(), MH_ALIGN(), and MEXT_ALIGN() as 938 * separate macros, each asserting that it was called at the proper moment. 939 * This required callers to themselves test the storage type and call the 940 * right one. Rather than require callers to be aware of those layout 941 * decisions, we centralize here. 942 */ 943 static __inline void 944 m_align(struct mbuf *m, int len) 945 { 946 #ifdef INVARIANTS 947 const char *msg = "%s: not a virgin mbuf"; 948 #endif 949 int adjust; 950 951 KASSERT(m->m_data == M_START(m), (msg, __func__)); 952 953 adjust = M_SIZE(m) - len; 954 m->m_data += adjust &~ (sizeof(long)-1); 955 } 956 957 #define M_ALIGN(m, len) m_align(m, len) 958 #define MH_ALIGN(m, len) m_align(m, len) 959 #define MEXT_ALIGN(m, len) m_align(m, len) 960 961 /* 962 * Compute the amount of space available before the current start of data in 963 * an mbuf. 964 * 965 * The M_WRITABLE() is a temporary, conservative safety measure: the burden 966 * of checking writability of the mbuf data area rests solely with the caller. 967 * 968 * NB: In previous versions, M_LEADINGSPACE() would only check M_WRITABLE() 969 * for mbufs with external storage. We now allow mbuf-embedded data to be 970 * read-only as well. 971 */ 972 #define M_LEADINGSPACE(m) \ 973 (M_WRITABLE(m) ? ((m)->m_data - M_START(m)) : 0) 974 975 /* 976 * Compute the amount of space available after the end of data in an mbuf. 977 * 978 * The M_WRITABLE() is a temporary, conservative safety measure: the burden 979 * of checking writability of the mbuf data area rests solely with the caller. 980 * 981 * NB: In previous versions, M_TRAILINGSPACE() would only check M_WRITABLE() 982 * for mbufs with external storage. We now allow mbuf-embedded data to be 983 * read-only as well. 984 */ 985 #define M_TRAILINGSPACE(m) \ 986 (M_WRITABLE(m) ? \ 987 ((M_START(m) + M_SIZE(m)) - ((m)->m_data + (m)->m_len)) : 0) 988 989 /* 990 * Arrange to prepend space of size plen to mbuf m. If a new mbuf must be 991 * allocated, how specifies whether to wait. If the allocation fails, the 992 * original mbuf chain is freed and m is set to NULL. 993 */ 994 #define M_PREPEND(m, plen, how) do { \ 995 struct mbuf **_mmp = &(m); \ 996 struct mbuf *_mm = *_mmp; \ 997 int _mplen = (plen); \ 998 int __mhow = (how); \ 999 \ 1000 MBUF_CHECKSLEEP(how); \ 1001 if (M_LEADINGSPACE(_mm) >= _mplen) { \ 1002 _mm->m_data -= _mplen; \ 1003 _mm->m_len += _mplen; \ 1004 } else \ 1005 _mm = m_prepend(_mm, _mplen, __mhow); \ 1006 if (_mm != NULL && _mm->m_flags & M_PKTHDR) \ 1007 _mm->m_pkthdr.len += _mplen; \ 1008 *_mmp = _mm; \ 1009 } while (0) 1010 1011 /* 1012 * Change mbuf to new type. This is a relatively expensive operation and 1013 * should be avoided. 1014 */ 1015 #define MCHTYPE(m, t) m_chtype((m), (t)) 1016 1017 /* Length to m_copy to copy all. */ 1018 #define M_COPYALL 1000000000 1019 1020 extern int max_datalen; /* MHLEN - max_hdr */ 1021 extern int max_hdr; /* Largest link + protocol header */ 1022 extern int max_linkhdr; /* Largest link-level header */ 1023 extern int max_protohdr; /* Largest protocol header */ 1024 extern int nmbclusters; /* Maximum number of clusters */ 1025 1026 /*- 1027 * Network packets may have annotations attached by affixing a list of 1028 * "packet tags" to the pkthdr structure. Packet tags are dynamically 1029 * allocated semi-opaque data structures that have a fixed header 1030 * (struct m_tag) that specifies the size of the memory block and a 1031 * <cookie,type> pair that identifies it. The cookie is a 32-bit unique 1032 * unsigned value used to identify a module or ABI. By convention this value 1033 * is chosen as the date+time that the module is created, expressed as the 1034 * number of seconds since the epoch (e.g., using date -u +'%s'). The type 1035 * value is an ABI/module-specific value that identifies a particular 1036 * annotation and is private to the module. For compatibility with systems 1037 * like OpenBSD that define packet tags w/o an ABI/module cookie, the value 1038 * PACKET_ABI_COMPAT is used to implement m_tag_get and m_tag_find 1039 * compatibility shim functions and several tag types are defined below. 1040 * Users that do not require compatibility should use a private cookie value 1041 * so that packet tag-related definitions can be maintained privately. 1042 * 1043 * Note that the packet tag returned by m_tag_alloc has the default memory 1044 * alignment implemented by malloc. To reference private data one can use a 1045 * construct like: 1046 * 1047 * struct m_tag *mtag = m_tag_alloc(...); 1048 * struct foo *p = (struct foo *)(mtag+1); 1049 * 1050 * if the alignment of struct m_tag is sufficient for referencing members of 1051 * struct foo. Otherwise it is necessary to embed struct m_tag within the 1052 * private data structure to insure proper alignment; e.g., 1053 * 1054 * struct foo { 1055 * struct m_tag tag; 1056 * ... 1057 * }; 1058 * struct foo *p = (struct foo *) m_tag_alloc(...); 1059 * struct m_tag *mtag = &p->tag; 1060 */ 1061 1062 /* 1063 * Persistent tags stay with an mbuf until the mbuf is reclaimed. Otherwise 1064 * tags are expected to ``vanish'' when they pass through a network 1065 * interface. For most interfaces this happens normally as the tags are 1066 * reclaimed when the mbuf is free'd. However in some special cases 1067 * reclaiming must be done manually. An example is packets that pass through 1068 * the loopback interface. Also, one must be careful to do this when 1069 * ``turning around'' packets (e.g., icmp_reflect). 1070 * 1071 * To mark a tag persistent bit-or this flag in when defining the tag id. 1072 * The tag will then be treated as described above. 1073 */ 1074 #define MTAG_PERSISTENT 0x800 1075 1076 #define PACKET_TAG_NONE 0 /* Nadda */ 1077 1078 /* Packet tags for use with PACKET_ABI_COMPAT. */ 1079 #define PACKET_TAG_IPSEC_IN_DONE 1 /* IPsec applied, in */ 1080 #define PACKET_TAG_IPSEC_OUT_DONE 2 /* IPsec applied, out */ 1081 #define PACKET_TAG_IPSEC_IN_CRYPTO_DONE 3 /* NIC IPsec crypto done */ 1082 #define PACKET_TAG_IPSEC_OUT_CRYPTO_NEEDED 4 /* NIC IPsec crypto req'ed */ 1083 #define PACKET_TAG_IPSEC_IN_COULD_DO_CRYPTO 5 /* NIC notifies IPsec */ 1084 #define PACKET_TAG_IPSEC_PENDING_TDB 6 /* Reminder to do IPsec */ 1085 #define PACKET_TAG_BRIDGE 7 /* Bridge processing done */ 1086 #define PACKET_TAG_GIF 8 /* GIF processing done */ 1087 #define PACKET_TAG_GRE 9 /* GRE processing done */ 1088 #define PACKET_TAG_IN_PACKET_CHECKSUM 10 /* NIC checksumming done */ 1089 #define PACKET_TAG_ENCAP 11 /* Encap. processing */ 1090 #define PACKET_TAG_IPSEC_SOCKET 12 /* IPSEC socket ref */ 1091 #define PACKET_TAG_IPSEC_HISTORY 13 /* IPSEC history */ 1092 #define PACKET_TAG_IPV6_INPUT 14 /* IPV6 input processing */ 1093 #define PACKET_TAG_DUMMYNET 15 /* dummynet info */ 1094 #define PACKET_TAG_DIVERT 17 /* divert info */ 1095 #define PACKET_TAG_IPFORWARD 18 /* ipforward info */ 1096 #define PACKET_TAG_MACLABEL (19 | MTAG_PERSISTENT) /* MAC label */ 1097 #define PACKET_TAG_PF (21 | MTAG_PERSISTENT) /* PF/ALTQ information */ 1098 #define PACKET_TAG_RTSOCKFAM 25 /* rtsock sa family */ 1099 #define PACKET_TAG_IPOPTIONS 27 /* Saved IP options */ 1100 #define PACKET_TAG_CARP 28 /* CARP info */ 1101 #define PACKET_TAG_IPSEC_NAT_T_PORTS 29 /* two uint16_t */ 1102 #define PACKET_TAG_ND_OUTGOING 30 /* ND outgoing */ 1103 1104 /* Specific cookies and tags. */ 1105 1106 /* Packet tag routines. */ 1107 struct m_tag *m_tag_alloc(u_int32_t, int, int, int); 1108 void m_tag_delete(struct mbuf *, struct m_tag *); 1109 void m_tag_delete_chain(struct mbuf *, struct m_tag *); 1110 void m_tag_free_default(struct m_tag *); 1111 struct m_tag *m_tag_locate(struct mbuf *, u_int32_t, int, struct m_tag *); 1112 struct m_tag *m_tag_copy(struct m_tag *, int); 1113 int m_tag_copy_chain(struct mbuf *, const struct mbuf *, int); 1114 void m_tag_delete_nonpersistent(struct mbuf *); 1115 1116 /* 1117 * Initialize the list of tags associated with an mbuf. 1118 */ 1119 static __inline void 1120 m_tag_init(struct mbuf *m) 1121 { 1122 1123 SLIST_INIT(&m->m_pkthdr.tags); 1124 } 1125 1126 /* 1127 * Set up the contents of a tag. Note that this does not fill in the free 1128 * method; the caller is expected to do that. 1129 * 1130 * XXX probably should be called m_tag_init, but that was already taken. 1131 */ 1132 static __inline void 1133 m_tag_setup(struct m_tag *t, u_int32_t cookie, int type, int len) 1134 { 1135 1136 t->m_tag_id = type; 1137 t->m_tag_len = len; 1138 t->m_tag_cookie = cookie; 1139 } 1140 1141 /* 1142 * Reclaim resources associated with a tag. 1143 */ 1144 static __inline void 1145 m_tag_free(struct m_tag *t) 1146 { 1147 1148 (*t->m_tag_free)(t); 1149 } 1150 1151 /* 1152 * Return the first tag associated with an mbuf. 1153 */ 1154 static __inline struct m_tag * 1155 m_tag_first(struct mbuf *m) 1156 { 1157 1158 return (SLIST_FIRST(&m->m_pkthdr.tags)); 1159 } 1160 1161 /* 1162 * Return the next tag in the list of tags associated with an mbuf. 1163 */ 1164 static __inline struct m_tag * 1165 m_tag_next(struct mbuf *m __unused, struct m_tag *t) 1166 { 1167 1168 return (SLIST_NEXT(t, m_tag_link)); 1169 } 1170 1171 /* 1172 * Prepend a tag to the list of tags associated with an mbuf. 1173 */ 1174 static __inline void 1175 m_tag_prepend(struct mbuf *m, struct m_tag *t) 1176 { 1177 1178 SLIST_INSERT_HEAD(&m->m_pkthdr.tags, t, m_tag_link); 1179 } 1180 1181 /* 1182 * Unlink a tag from the list of tags associated with an mbuf. 1183 */ 1184 static __inline void 1185 m_tag_unlink(struct mbuf *m, struct m_tag *t) 1186 { 1187 1188 SLIST_REMOVE(&m->m_pkthdr.tags, t, m_tag, m_tag_link); 1189 } 1190 1191 /* These are for OpenBSD compatibility. */ 1192 #define MTAG_ABI_COMPAT 0 /* compatibility ABI */ 1193 1194 static __inline struct m_tag * 1195 m_tag_get(int type, int length, int wait) 1196 { 1197 return (m_tag_alloc(MTAG_ABI_COMPAT, type, length, wait)); 1198 } 1199 1200 static __inline struct m_tag * 1201 m_tag_find(struct mbuf *m, int type, struct m_tag *start) 1202 { 1203 return (SLIST_EMPTY(&m->m_pkthdr.tags) ? (struct m_tag *)NULL : 1204 m_tag_locate(m, MTAG_ABI_COMPAT, type, start)); 1205 } 1206 1207 static __inline struct mbuf * 1208 m_free(struct mbuf *m) 1209 { 1210 struct mbuf *n = m->m_next; 1211 1212 MBUF_PROBE1(m__free, m); 1213 if ((m->m_flags & (M_PKTHDR|M_NOFREE)) == (M_PKTHDR|M_NOFREE)) 1214 m_tag_delete_chain(m, NULL); 1215 if (m->m_flags & M_EXT) 1216 mb_free_ext(m); 1217 else if ((m->m_flags & M_NOFREE) == 0) 1218 uma_zfree(zone_mbuf, m); 1219 return (n); 1220 } 1221 1222 static __inline int 1223 rt_m_getfib(struct mbuf *m) 1224 { 1225 KASSERT(m->m_flags & M_PKTHDR , ("Attempt to get FIB from non header mbuf.")); 1226 return (m->m_pkthdr.fibnum); 1227 } 1228 1229 #define M_GETFIB(_m) rt_m_getfib(_m) 1230 1231 #define M_SETFIB(_m, _fib) do { \ 1232 KASSERT((_m)->m_flags & M_PKTHDR, ("Attempt to set FIB on non header mbuf.")); \ 1233 ((_m)->m_pkthdr.fibnum) = (_fib); \ 1234 } while (0) 1235 1236 /* flags passed as first argument for "m_ether_tcpip_hash()" */ 1237 #define MBUF_HASHFLAG_L2 (1 << 2) 1238 #define MBUF_HASHFLAG_L3 (1 << 3) 1239 #define MBUF_HASHFLAG_L4 (1 << 4) 1240 1241 /* mbuf hashing helper routines */ 1242 uint32_t m_ether_tcpip_hash_init(void); 1243 uint32_t m_ether_tcpip_hash(const uint32_t, const struct mbuf *, const uint32_t); 1244 1245 #ifdef MBUF_PROFILING 1246 void m_profile(struct mbuf *m); 1247 #define M_PROFILE(m) m_profile(m) 1248 #else 1249 #define M_PROFILE(m) 1250 #endif 1251 1252 struct mbufq { 1253 STAILQ_HEAD(, mbuf) mq_head; 1254 int mq_len; 1255 int mq_maxlen; 1256 }; 1257 1258 static inline void 1259 mbufq_init(struct mbufq *mq, int maxlen) 1260 { 1261 1262 STAILQ_INIT(&mq->mq_head); 1263 mq->mq_maxlen = maxlen; 1264 mq->mq_len = 0; 1265 } 1266 1267 static inline struct mbuf * 1268 mbufq_flush(struct mbufq *mq) 1269 { 1270 struct mbuf *m; 1271 1272 m = STAILQ_FIRST(&mq->mq_head); 1273 STAILQ_INIT(&mq->mq_head); 1274 mq->mq_len = 0; 1275 return (m); 1276 } 1277 1278 static inline void 1279 mbufq_drain(struct mbufq *mq) 1280 { 1281 struct mbuf *m, *n; 1282 1283 n = mbufq_flush(mq); 1284 while ((m = n) != NULL) { 1285 n = STAILQ_NEXT(m, m_stailqpkt); 1286 m_freem(m); 1287 } 1288 } 1289 1290 static inline struct mbuf * 1291 mbufq_first(const struct mbufq *mq) 1292 { 1293 1294 return (STAILQ_FIRST(&mq->mq_head)); 1295 } 1296 1297 static inline struct mbuf * 1298 mbufq_last(const struct mbufq *mq) 1299 { 1300 1301 return (STAILQ_LAST(&mq->mq_head, mbuf, m_stailqpkt)); 1302 } 1303 1304 static inline int 1305 mbufq_full(const struct mbufq *mq) 1306 { 1307 1308 return (mq->mq_len >= mq->mq_maxlen); 1309 } 1310 1311 static inline int 1312 mbufq_len(const struct mbufq *mq) 1313 { 1314 1315 return (mq->mq_len); 1316 } 1317 1318 static inline int 1319 mbufq_enqueue(struct mbufq *mq, struct mbuf *m) 1320 { 1321 1322 if (mbufq_full(mq)) 1323 return (ENOBUFS); 1324 STAILQ_INSERT_TAIL(&mq->mq_head, m, m_stailqpkt); 1325 mq->mq_len++; 1326 return (0); 1327 } 1328 1329 static inline struct mbuf * 1330 mbufq_dequeue(struct mbufq *mq) 1331 { 1332 struct mbuf *m; 1333 1334 m = STAILQ_FIRST(&mq->mq_head); 1335 if (m) { 1336 STAILQ_REMOVE_HEAD(&mq->mq_head, m_stailqpkt); 1337 m->m_nextpkt = NULL; 1338 mq->mq_len--; 1339 } 1340 return (m); 1341 } 1342 1343 static inline void 1344 mbufq_prepend(struct mbufq *mq, struct mbuf *m) 1345 { 1346 1347 STAILQ_INSERT_HEAD(&mq->mq_head, m, m_stailqpkt); 1348 mq->mq_len++; 1349 } 1350 1351 /* 1352 * Note: this doesn't enforce the maximum list size for dst. 1353 */ 1354 static inline void 1355 mbufq_concat(struct mbufq *mq_dst, struct mbufq *mq_src) 1356 { 1357 1358 mq_dst->mq_len += mq_src->mq_len; 1359 STAILQ_CONCAT(&mq_dst->mq_head, &mq_src->mq_head); 1360 mq_src->mq_len = 0; 1361 } 1362 1363 #ifdef _SYS_TIMESPEC_H_ 1364 static inline void 1365 mbuf_tstmp2timespec(struct mbuf *m, struct timespec *ts) 1366 { 1367 1368 KASSERT((m->m_flags & M_PKTHDR) != 0, ("mbuf %p no M_PKTHDR", m)); 1369 KASSERT((m->m_flags & M_TSTMP) != 0, ("mbuf %p no M_TSTMP", m)); 1370 ts->tv_sec = m->m_pkthdr.rcv_tstmp / 1000000000; 1371 ts->tv_nsec = m->m_pkthdr.rcv_tstmp % 1000000000; 1372 } 1373 #endif 1374 1375 #endif /* _KERNEL */ 1376 #endif /* !_SYS_MBUF_H_ */ 1377