1 /*- 2 * Copyright (c) 1982, 1986, 1988, 1993 3 * The Regents of the University of California. 4 * All rights reserved. 5 * 6 * Redistribution and use in source and binary forms, with or without 7 * modification, are permitted provided that the following conditions 8 * are met: 9 * 1. Redistributions of source code must retain the above copyright 10 * notice, this list of conditions and the following disclaimer. 11 * 2. Redistributions in binary form must reproduce the above copyright 12 * notice, this list of conditions and the following disclaimer in the 13 * documentation and/or other materials provided with the distribution. 14 * 3. Neither the name of the University nor the names of its contributors 15 * may be used to endorse or promote products derived from this software 16 * without specific prior written permission. 17 * 18 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 19 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 20 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 21 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 22 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 23 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 24 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 25 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 26 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 27 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 28 * SUCH DAMAGE. 29 * 30 * @(#)mbuf.h 8.5 (Berkeley) 2/19/95 31 * $FreeBSD$ 32 */ 33 34 #ifndef _SYS_MBUF_H_ 35 #define _SYS_MBUF_H_ 36 37 /* XXX: These includes suck. Sorry! */ 38 #include <sys/queue.h> 39 #ifdef _KERNEL 40 #include <sys/systm.h> 41 #include <vm/uma.h> 42 #ifdef WITNESS 43 #include <sys/lock.h> 44 #endif 45 #endif 46 47 #ifdef _KERNEL 48 #include <sys/sdt.h> 49 50 #define MBUF_PROBE1(probe, arg0) \ 51 SDT_PROBE1(sdt, , , probe, arg0) 52 #define MBUF_PROBE2(probe, arg0, arg1) \ 53 SDT_PROBE2(sdt, , , probe, arg0, arg1) 54 #define MBUF_PROBE3(probe, arg0, arg1, arg2) \ 55 SDT_PROBE3(sdt, , , probe, arg0, arg1, arg2) 56 #define MBUF_PROBE4(probe, arg0, arg1, arg2, arg3) \ 57 SDT_PROBE4(sdt, , , probe, arg0, arg1, arg2, arg3) 58 #define MBUF_PROBE5(probe, arg0, arg1, arg2, arg3, arg4) \ 59 SDT_PROBE5(sdt, , , probe, arg0, arg1, arg2, arg3, arg4) 60 61 SDT_PROBE_DECLARE(sdt, , , m__init); 62 SDT_PROBE_DECLARE(sdt, , , m__gethdr); 63 SDT_PROBE_DECLARE(sdt, , , m__get); 64 SDT_PROBE_DECLARE(sdt, , , m__getcl); 65 SDT_PROBE_DECLARE(sdt, , , m__clget); 66 SDT_PROBE_DECLARE(sdt, , , m__cljget); 67 SDT_PROBE_DECLARE(sdt, , , m__cljset); 68 SDT_PROBE_DECLARE(sdt, , , m__free); 69 SDT_PROBE_DECLARE(sdt, , , m__freem); 70 71 #endif /* _KERNEL */ 72 73 /* 74 * Mbufs are of a single size, MSIZE (sys/param.h), which includes overhead. 75 * An mbuf may add a single "mbuf cluster" of size MCLBYTES (also in 76 * sys/param.h), which has no additional overhead and is used instead of the 77 * internal data area; this is done when at least MINCLSIZE of data must be 78 * stored. Additionally, it is possible to allocate a separate buffer 79 * externally and attach it to the mbuf in a way similar to that of mbuf 80 * clusters. 81 * 82 * NB: These calculation do not take actual compiler-induced alignment and 83 * padding inside the complete struct mbuf into account. Appropriate 84 * attention is required when changing members of struct mbuf. 85 * 86 * MLEN is data length in a normal mbuf. 87 * MHLEN is data length in an mbuf with pktheader. 88 * MINCLSIZE is a smallest amount of data that should be put into cluster. 89 * 90 * Compile-time assertions in uipc_mbuf.c test these values to ensure that 91 * they are sensible. 92 */ 93 struct mbuf; 94 #define MHSIZE offsetof(struct mbuf, m_dat) 95 #define MPKTHSIZE offsetof(struct mbuf, m_pktdat) 96 #define MLEN ((int)(MSIZE - MHSIZE)) 97 #define MHLEN ((int)(MSIZE - MPKTHSIZE)) 98 #define MINCLSIZE (MHLEN + 1) 99 100 #ifdef _KERNEL 101 /*- 102 * Macro for type conversion: convert mbuf pointer to data pointer of correct 103 * type: 104 * 105 * mtod(m, t) -- Convert mbuf pointer to data pointer of correct type. 106 * mtodo(m, o) -- Same as above but with offset 'o' into data. 107 */ 108 #define mtod(m, t) ((t)((m)->m_data)) 109 #define mtodo(m, o) ((void *)(((m)->m_data) + (o))) 110 111 /* 112 * Argument structure passed to UMA routines during mbuf and packet 113 * allocations. 114 */ 115 struct mb_args { 116 int flags; /* Flags for mbuf being allocated */ 117 short type; /* Type of mbuf being allocated */ 118 }; 119 #endif /* _KERNEL */ 120 121 /* 122 * Packet tag structure (see below for details). 123 */ 124 struct m_tag { 125 SLIST_ENTRY(m_tag) m_tag_link; /* List of packet tags */ 126 u_int16_t m_tag_id; /* Tag ID */ 127 u_int16_t m_tag_len; /* Length of data */ 128 u_int32_t m_tag_cookie; /* ABI/Module ID */ 129 void (*m_tag_free)(struct m_tag *); 130 }; 131 132 /* 133 * Static network interface owned tag. 134 * Allocated through ifp->if_snd_tag_alloc(). 135 */ 136 struct m_snd_tag { 137 struct ifnet *ifp; /* network interface tag belongs to */ 138 }; 139 140 /* 141 * Record/packet header in first mbuf of chain; valid only if M_PKTHDR is set. 142 * Size ILP32: 48 143 * LP64: 56 144 * Compile-time assertions in uipc_mbuf.c test these values to ensure that 145 * they are correct. 146 */ 147 struct pkthdr { 148 union { 149 struct m_snd_tag *snd_tag; /* send tag, if any */ 150 struct ifnet *rcvif; /* rcv interface */ 151 }; 152 SLIST_HEAD(packet_tags, m_tag) tags; /* list of packet tags */ 153 int32_t len; /* total packet length */ 154 155 /* Layer crossing persistent information. */ 156 uint32_t flowid; /* packet's 4-tuple system */ 157 uint64_t csum_flags; /* checksum and offload features */ 158 uint16_t fibnum; /* this packet should use this fib */ 159 uint8_t cosqos; /* class/quality of service */ 160 uint8_t rsstype; /* hash type */ 161 uint8_t l2hlen; /* layer 2 header length */ 162 uint8_t l3hlen; /* layer 3 header length */ 163 uint8_t l4hlen; /* layer 4 header length */ 164 uint8_t l5hlen; /* layer 5 header length */ 165 union { 166 uint8_t eight[8]; 167 uint16_t sixteen[4]; 168 uint32_t thirtytwo[2]; 169 uint64_t sixtyfour[1]; 170 uintptr_t unintptr[1]; 171 void *ptr; 172 } PH_per; 173 174 /* Layer specific non-persistent local storage for reassembly, etc. */ 175 union { 176 uint8_t eight[8]; 177 uint16_t sixteen[4]; 178 uint32_t thirtytwo[2]; 179 uint64_t sixtyfour[1]; 180 uintptr_t unintptr[1]; 181 void *ptr; 182 } PH_loc; 183 }; 184 #define ether_vtag PH_per.sixteen[0] 185 #define PH_vt PH_per 186 #define vt_nrecs sixteen[0] 187 #define tso_segsz PH_per.sixteen[1] 188 #define lro_nsegs tso_segsz 189 #define csum_phsum PH_per.sixteen[2] 190 #define csum_data PH_per.thirtytwo[1] 191 192 /* 193 * Description of external storage mapped into mbuf; valid only if M_EXT is 194 * set. 195 * Size ILP32: 28 196 * LP64: 48 197 * Compile-time assertions in uipc_mbuf.c test these values to ensure that 198 * they are correct. 199 */ 200 struct m_ext { 201 union { 202 volatile u_int ext_count; /* value of ref count info */ 203 volatile u_int *ext_cnt; /* pointer to ref count info */ 204 }; 205 caddr_t ext_buf; /* start of buffer */ 206 uint32_t ext_size; /* size of buffer, for ext_free */ 207 uint32_t ext_type:8, /* type of external storage */ 208 ext_flags:24; /* external storage mbuf flags */ 209 void (*ext_free) /* free routine if not the usual */ 210 (struct mbuf *, void *, void *); 211 void *ext_arg1; /* optional argument pointer */ 212 void *ext_arg2; /* optional argument pointer */ 213 }; 214 215 /* 216 * The core of the mbuf object along with some shortcut defines for practical 217 * purposes. 218 */ 219 struct mbuf { 220 /* 221 * Header present at the beginning of every mbuf. 222 * Size ILP32: 24 223 * LP64: 32 224 * Compile-time assertions in uipc_mbuf.c test these values to ensure 225 * that they are correct. 226 */ 227 union { /* next buffer in chain */ 228 struct mbuf *m_next; 229 SLIST_ENTRY(mbuf) m_slist; 230 STAILQ_ENTRY(mbuf) m_stailq; 231 }; 232 union { /* next chain in queue/record */ 233 struct mbuf *m_nextpkt; 234 SLIST_ENTRY(mbuf) m_slistpkt; 235 STAILQ_ENTRY(mbuf) m_stailqpkt; 236 }; 237 caddr_t m_data; /* location of data */ 238 int32_t m_len; /* amount of data in this mbuf */ 239 uint32_t m_type:8, /* type of data in this mbuf */ 240 m_flags:24; /* flags; see below */ 241 #if !defined(__LP64__) 242 uint32_t m_pad; /* pad for 64bit alignment */ 243 #endif 244 245 /* 246 * A set of optional headers (packet header, external storage header) 247 * and internal data storage. Historically, these arrays were sized 248 * to MHLEN (space left after a packet header) and MLEN (space left 249 * after only a regular mbuf header); they are now variable size in 250 * order to support future work on variable-size mbufs. 251 */ 252 union { 253 struct { 254 struct pkthdr m_pkthdr; /* M_PKTHDR set */ 255 union { 256 struct m_ext m_ext; /* M_EXT set */ 257 char m_pktdat[0]; 258 }; 259 }; 260 char m_dat[0]; /* !M_PKTHDR, !M_EXT */ 261 }; 262 }; 263 264 /* 265 * mbuf flags of global significance and layer crossing. 266 * Those of only protocol/layer specific significance are to be mapped 267 * to M_PROTO[1-12] and cleared at layer handoff boundaries. 268 * NB: Limited to the lower 24 bits. 269 */ 270 #define M_EXT 0x00000001 /* has associated external storage */ 271 #define M_PKTHDR 0x00000002 /* start of record */ 272 #define M_EOR 0x00000004 /* end of record */ 273 #define M_RDONLY 0x00000008 /* associated data is marked read-only */ 274 #define M_BCAST 0x00000010 /* send/received as link-level broadcast */ 275 #define M_MCAST 0x00000020 /* send/received as link-level multicast */ 276 #define M_PROMISC 0x00000040 /* packet was not for us */ 277 #define M_VLANTAG 0x00000080 /* ether_vtag is valid */ 278 #define M_UNUSED_8 0x00000100 /* --available-- */ 279 #define M_NOFREE 0x00000200 /* do not free mbuf, embedded in cluster */ 280 281 #define M_PROTO1 0x00001000 /* protocol-specific */ 282 #define M_PROTO2 0x00002000 /* protocol-specific */ 283 #define M_PROTO3 0x00004000 /* protocol-specific */ 284 #define M_PROTO4 0x00008000 /* protocol-specific */ 285 #define M_PROTO5 0x00010000 /* protocol-specific */ 286 #define M_PROTO6 0x00020000 /* protocol-specific */ 287 #define M_PROTO7 0x00040000 /* protocol-specific */ 288 #define M_PROTO8 0x00080000 /* protocol-specific */ 289 #define M_PROTO9 0x00100000 /* protocol-specific */ 290 #define M_PROTO10 0x00200000 /* protocol-specific */ 291 #define M_PROTO11 0x00400000 /* protocol-specific */ 292 #define M_PROTO12 0x00800000 /* protocol-specific */ 293 294 #define MB_DTOR_SKIP 0x1 /* don't pollute the cache by touching a freed mbuf */ 295 296 /* 297 * Flags to purge when crossing layers. 298 */ 299 #define M_PROTOFLAGS \ 300 (M_PROTO1|M_PROTO2|M_PROTO3|M_PROTO4|M_PROTO5|M_PROTO6|M_PROTO7|M_PROTO8|\ 301 M_PROTO9|M_PROTO10|M_PROTO11|M_PROTO12) 302 303 /* 304 * Flags preserved when copying m_pkthdr. 305 */ 306 #define M_COPYFLAGS \ 307 (M_PKTHDR|M_EOR|M_RDONLY|M_BCAST|M_MCAST|M_PROMISC|M_VLANTAG| \ 308 M_PROTOFLAGS) 309 310 /* 311 * Mbuf flag description for use with printf(9) %b identifier. 312 */ 313 #define M_FLAG_BITS \ 314 "\20\1M_EXT\2M_PKTHDR\3M_EOR\4M_RDONLY\5M_BCAST\6M_MCAST" \ 315 "\7M_PROMISC\10M_VLANTAG" 316 #define M_FLAG_PROTOBITS \ 317 "\15M_PROTO1\16M_PROTO2\17M_PROTO3\20M_PROTO4\21M_PROTO5" \ 318 "\22M_PROTO6\23M_PROTO7\24M_PROTO8\25M_PROTO9\26M_PROTO10" \ 319 "\27M_PROTO11\30M_PROTO12" 320 #define M_FLAG_PRINTF (M_FLAG_BITS M_FLAG_PROTOBITS) 321 322 /* 323 * Network interface cards are able to hash protocol fields (such as IPv4 324 * addresses and TCP port numbers) classify packets into flows. These flows 325 * can then be used to maintain ordering while delivering packets to the OS 326 * via parallel input queues, as well as to provide a stateless affinity 327 * model. NIC drivers can pass up the hash via m->m_pkthdr.flowid, and set 328 * m_flag fields to indicate how the hash should be interpreted by the 329 * network stack. 330 * 331 * Most NICs support RSS, which provides ordering and explicit affinity, and 332 * use the hash m_flag bits to indicate what header fields were covered by 333 * the hash. M_HASHTYPE_OPAQUE and M_HASHTYPE_OPAQUE_HASH can be set by non- 334 * RSS cards or configurations that provide an opaque flow identifier, allowing 335 * for ordering and distribution without explicit affinity. Additionally, 336 * M_HASHTYPE_OPAQUE_HASH indicates that the flow identifier has hash 337 * properties. 338 */ 339 #define M_HASHTYPE_HASHPROP 0x80 /* has hash properties */ 340 #define M_HASHTYPE_HASH(t) (M_HASHTYPE_HASHPROP | (t)) 341 /* Microsoft RSS standard hash types */ 342 #define M_HASHTYPE_NONE 0 343 #define M_HASHTYPE_RSS_IPV4 M_HASHTYPE_HASH(1) /* IPv4 2-tuple */ 344 #define M_HASHTYPE_RSS_TCP_IPV4 M_HASHTYPE_HASH(2) /* TCPv4 4-tuple */ 345 #define M_HASHTYPE_RSS_IPV6 M_HASHTYPE_HASH(3) /* IPv6 2-tuple */ 346 #define M_HASHTYPE_RSS_TCP_IPV6 M_HASHTYPE_HASH(4) /* TCPv6 4-tuple */ 347 #define M_HASHTYPE_RSS_IPV6_EX M_HASHTYPE_HASH(5) /* IPv6 2-tuple + 348 * ext hdrs */ 349 #define M_HASHTYPE_RSS_TCP_IPV6_EX M_HASHTYPE_HASH(6) /* TCPv6 4-tuple + 350 * ext hdrs */ 351 /* Non-standard RSS hash types */ 352 #define M_HASHTYPE_RSS_UDP_IPV4 M_HASHTYPE_HASH(7) /* IPv4 UDP 4-tuple*/ 353 #define M_HASHTYPE_RSS_UDP_IPV4_EX M_HASHTYPE_HASH(8) /* IPv4 UDP 4-tuple + 354 * ext hdrs */ 355 #define M_HASHTYPE_RSS_UDP_IPV6 M_HASHTYPE_HASH(9) /* IPv6 UDP 4-tuple*/ 356 #define M_HASHTYPE_RSS_UDP_IPV6_EX M_HASHTYPE_HASH(10)/* IPv6 UDP 4-tuple + 357 * ext hdrs */ 358 359 #define M_HASHTYPE_OPAQUE 63 /* ordering, not affinity */ 360 #define M_HASHTYPE_OPAQUE_HASH M_HASHTYPE_HASH(M_HASHTYPE_OPAQUE) 361 /* ordering+hash, not affinity*/ 362 363 #define M_HASHTYPE_CLEAR(m) ((m)->m_pkthdr.rsstype = 0) 364 #define M_HASHTYPE_GET(m) ((m)->m_pkthdr.rsstype) 365 #define M_HASHTYPE_SET(m, v) ((m)->m_pkthdr.rsstype = (v)) 366 #define M_HASHTYPE_TEST(m, v) (M_HASHTYPE_GET(m) == (v)) 367 #define M_HASHTYPE_ISHASH(m) (M_HASHTYPE_GET(m) & M_HASHTYPE_HASHPROP) 368 369 /* 370 * COS/QOS class and quality of service tags. 371 * It uses DSCP code points as base. 372 */ 373 #define QOS_DSCP_CS0 0x00 374 #define QOS_DSCP_DEF QOS_DSCP_CS0 375 #define QOS_DSCP_CS1 0x20 376 #define QOS_DSCP_AF11 0x28 377 #define QOS_DSCP_AF12 0x30 378 #define QOS_DSCP_AF13 0x38 379 #define QOS_DSCP_CS2 0x40 380 #define QOS_DSCP_AF21 0x48 381 #define QOS_DSCP_AF22 0x50 382 #define QOS_DSCP_AF23 0x58 383 #define QOS_DSCP_CS3 0x60 384 #define QOS_DSCP_AF31 0x68 385 #define QOS_DSCP_AF32 0x70 386 #define QOS_DSCP_AF33 0x78 387 #define QOS_DSCP_CS4 0x80 388 #define QOS_DSCP_AF41 0x88 389 #define QOS_DSCP_AF42 0x90 390 #define QOS_DSCP_AF43 0x98 391 #define QOS_DSCP_CS5 0xa0 392 #define QOS_DSCP_EF 0xb8 393 #define QOS_DSCP_CS6 0xc0 394 #define QOS_DSCP_CS7 0xe0 395 396 /* 397 * External mbuf storage buffer types. 398 */ 399 #define EXT_CLUSTER 1 /* mbuf cluster */ 400 #define EXT_SFBUF 2 /* sendfile(2)'s sf_buf */ 401 #define EXT_JUMBOP 3 /* jumbo cluster page sized */ 402 #define EXT_JUMBO9 4 /* jumbo cluster 9216 bytes */ 403 #define EXT_JUMBO16 5 /* jumbo cluster 16184 bytes */ 404 #define EXT_PACKET 6 /* mbuf+cluster from packet zone */ 405 #define EXT_MBUF 7 /* external mbuf reference */ 406 #define EXT_SFBUF_NOCACHE 8 /* sendfile(2)'s sf_buf not to be cached */ 407 408 #define EXT_VENDOR1 224 /* for vendor-internal use */ 409 #define EXT_VENDOR2 225 /* for vendor-internal use */ 410 #define EXT_VENDOR3 226 /* for vendor-internal use */ 411 #define EXT_VENDOR4 227 /* for vendor-internal use */ 412 413 #define EXT_EXP1 244 /* for experimental use */ 414 #define EXT_EXP2 245 /* for experimental use */ 415 #define EXT_EXP3 246 /* for experimental use */ 416 #define EXT_EXP4 247 /* for experimental use */ 417 418 #define EXT_NET_DRV 252 /* custom ext_buf provided by net driver(s) */ 419 #define EXT_MOD_TYPE 253 /* custom module's ext_buf type */ 420 #define EXT_DISPOSABLE 254 /* can throw this buffer away w/page flipping */ 421 #define EXT_EXTREF 255 /* has externally maintained ext_cnt ptr */ 422 423 /* 424 * Flags for external mbuf buffer types. 425 * NB: limited to the lower 24 bits. 426 */ 427 #define EXT_FLAG_EMBREF 0x000001 /* embedded ext_count */ 428 #define EXT_FLAG_EXTREF 0x000002 /* external ext_cnt, notyet */ 429 430 #define EXT_FLAG_NOFREE 0x000010 /* don't free mbuf to pool, notyet */ 431 432 #define EXT_FLAG_VENDOR1 0x010000 /* for vendor-internal use */ 433 #define EXT_FLAG_VENDOR2 0x020000 /* for vendor-internal use */ 434 #define EXT_FLAG_VENDOR3 0x040000 /* for vendor-internal use */ 435 #define EXT_FLAG_VENDOR4 0x080000 /* for vendor-internal use */ 436 437 #define EXT_FLAG_EXP1 0x100000 /* for experimental use */ 438 #define EXT_FLAG_EXP2 0x200000 /* for experimental use */ 439 #define EXT_FLAG_EXP3 0x400000 /* for experimental use */ 440 #define EXT_FLAG_EXP4 0x800000 /* for experimental use */ 441 442 /* 443 * EXT flag description for use with printf(9) %b identifier. 444 */ 445 #define EXT_FLAG_BITS \ 446 "\20\1EXT_FLAG_EMBREF\2EXT_FLAG_EXTREF\5EXT_FLAG_NOFREE" \ 447 "\21EXT_FLAG_VENDOR1\22EXT_FLAG_VENDOR2\23EXT_FLAG_VENDOR3" \ 448 "\24EXT_FLAG_VENDOR4\25EXT_FLAG_EXP1\26EXT_FLAG_EXP2\27EXT_FLAG_EXP3" \ 449 "\30EXT_FLAG_EXP4" 450 451 /* 452 * External reference/free functions. 453 */ 454 void sf_ext_free(void *, void *); 455 void sf_ext_free_nocache(void *, void *); 456 457 /* 458 * Flags indicating checksum, segmentation and other offload work to be 459 * done, or already done, by hardware or lower layers. It is split into 460 * separate inbound and outbound flags. 461 * 462 * Outbound flags that are set by upper protocol layers requesting lower 463 * layers, or ideally the hardware, to perform these offloading tasks. 464 * For outbound packets this field and its flags can be directly tested 465 * against ifnet if_hwassist. 466 */ 467 #define CSUM_IP 0x00000001 /* IP header checksum offload */ 468 #define CSUM_IP_UDP 0x00000002 /* UDP checksum offload */ 469 #define CSUM_IP_TCP 0x00000004 /* TCP checksum offload */ 470 #define CSUM_IP_SCTP 0x00000008 /* SCTP checksum offload */ 471 #define CSUM_IP_TSO 0x00000010 /* TCP segmentation offload */ 472 #define CSUM_IP_ISCSI 0x00000020 /* iSCSI checksum offload */ 473 474 #define CSUM_IP6_UDP 0x00000200 /* UDP checksum offload */ 475 #define CSUM_IP6_TCP 0x00000400 /* TCP checksum offload */ 476 #define CSUM_IP6_SCTP 0x00000800 /* SCTP checksum offload */ 477 #define CSUM_IP6_TSO 0x00001000 /* TCP segmentation offload */ 478 #define CSUM_IP6_ISCSI 0x00002000 /* iSCSI checksum offload */ 479 480 /* Inbound checksum support where the checksum was verified by hardware. */ 481 #define CSUM_L3_CALC 0x01000000 /* calculated layer 3 csum */ 482 #define CSUM_L3_VALID 0x02000000 /* checksum is correct */ 483 #define CSUM_L4_CALC 0x04000000 /* calculated layer 4 csum */ 484 #define CSUM_L4_VALID 0x08000000 /* checksum is correct */ 485 #define CSUM_L5_CALC 0x10000000 /* calculated layer 5 csum */ 486 #define CSUM_L5_VALID 0x20000000 /* checksum is correct */ 487 #define CSUM_COALESCED 0x40000000 /* contains merged segments */ 488 489 /* 490 * CSUM flag description for use with printf(9) %b identifier. 491 */ 492 #define CSUM_BITS \ 493 "\20\1CSUM_IP\2CSUM_IP_UDP\3CSUM_IP_TCP\4CSUM_IP_SCTP\5CSUM_IP_TSO" \ 494 "\6CSUM_IP_ISCSI" \ 495 "\12CSUM_IP6_UDP\13CSUM_IP6_TCP\14CSUM_IP6_SCTP\15CSUM_IP6_TSO" \ 496 "\16CSUM_IP6_ISCSI" \ 497 "\31CSUM_L3_CALC\32CSUM_L3_VALID\33CSUM_L4_CALC\34CSUM_L4_VALID" \ 498 "\35CSUM_L5_CALC\36CSUM_L5_VALID\37CSUM_COALESCED" 499 500 /* CSUM flags compatibility mappings. */ 501 #define CSUM_IP_CHECKED CSUM_L3_CALC 502 #define CSUM_IP_VALID CSUM_L3_VALID 503 #define CSUM_DATA_VALID CSUM_L4_VALID 504 #define CSUM_PSEUDO_HDR CSUM_L4_CALC 505 #define CSUM_SCTP_VALID CSUM_L4_VALID 506 #define CSUM_DELAY_DATA (CSUM_TCP|CSUM_UDP) 507 #define CSUM_DELAY_IP CSUM_IP /* Only v4, no v6 IP hdr csum */ 508 #define CSUM_DELAY_DATA_IPV6 (CSUM_TCP_IPV6|CSUM_UDP_IPV6) 509 #define CSUM_DATA_VALID_IPV6 CSUM_DATA_VALID 510 #define CSUM_TCP CSUM_IP_TCP 511 #define CSUM_UDP CSUM_IP_UDP 512 #define CSUM_SCTP CSUM_IP_SCTP 513 #define CSUM_TSO (CSUM_IP_TSO|CSUM_IP6_TSO) 514 #define CSUM_UDP_IPV6 CSUM_IP6_UDP 515 #define CSUM_TCP_IPV6 CSUM_IP6_TCP 516 #define CSUM_SCTP_IPV6 CSUM_IP6_SCTP 517 518 /* 519 * mbuf types describing the content of the mbuf (including external storage). 520 */ 521 #define MT_NOTMBUF 0 /* USED INTERNALLY ONLY! Object is not mbuf */ 522 #define MT_DATA 1 /* dynamic (data) allocation */ 523 #define MT_HEADER MT_DATA /* packet header, use M_PKTHDR instead */ 524 525 #define MT_VENDOR1 4 /* for vendor-internal use */ 526 #define MT_VENDOR2 5 /* for vendor-internal use */ 527 #define MT_VENDOR3 6 /* for vendor-internal use */ 528 #define MT_VENDOR4 7 /* for vendor-internal use */ 529 530 #define MT_SONAME 8 /* socket name */ 531 532 #define MT_EXP1 9 /* for experimental use */ 533 #define MT_EXP2 10 /* for experimental use */ 534 #define MT_EXP3 11 /* for experimental use */ 535 #define MT_EXP4 12 /* for experimental use */ 536 537 #define MT_CONTROL 14 /* extra-data protocol message */ 538 #define MT_OOBDATA 15 /* expedited data */ 539 #define MT_NTYPES 16 /* number of mbuf types for mbtypes[] */ 540 541 #define MT_NOINIT 255 /* Not a type but a flag to allocate 542 a non-initialized mbuf */ 543 544 /* 545 * String names of mbuf-related UMA(9) and malloc(9) types. Exposed to 546 * !_KERNEL so that monitoring tools can look up the zones with 547 * libmemstat(3). 548 */ 549 #define MBUF_MEM_NAME "mbuf" 550 #define MBUF_CLUSTER_MEM_NAME "mbuf_cluster" 551 #define MBUF_PACKET_MEM_NAME "mbuf_packet" 552 #define MBUF_JUMBOP_MEM_NAME "mbuf_jumbo_page" 553 #define MBUF_JUMBO9_MEM_NAME "mbuf_jumbo_9k" 554 #define MBUF_JUMBO16_MEM_NAME "mbuf_jumbo_16k" 555 #define MBUF_TAG_MEM_NAME "mbuf_tag" 556 #define MBUF_EXTREFCNT_MEM_NAME "mbuf_ext_refcnt" 557 558 #ifdef _KERNEL 559 560 #ifdef WITNESS 561 #define MBUF_CHECKSLEEP(how) do { \ 562 if (how == M_WAITOK) \ 563 WITNESS_WARN(WARN_GIANTOK | WARN_SLEEPOK, NULL, \ 564 "Sleeping in \"%s\"", __func__); \ 565 } while (0) 566 #else 567 #define MBUF_CHECKSLEEP(how) 568 #endif 569 570 /* 571 * Network buffer allocation API 572 * 573 * The rest of it is defined in kern/kern_mbuf.c 574 */ 575 extern uma_zone_t zone_mbuf; 576 extern uma_zone_t zone_clust; 577 extern uma_zone_t zone_pack; 578 extern uma_zone_t zone_jumbop; 579 extern uma_zone_t zone_jumbo9; 580 extern uma_zone_t zone_jumbo16; 581 582 void mb_dupcl(struct mbuf *, struct mbuf *); 583 void mb_free_ext(struct mbuf *); 584 void m_adj(struct mbuf *, int); 585 int m_apply(struct mbuf *, int, int, 586 int (*)(void *, void *, u_int), void *); 587 int m_append(struct mbuf *, int, c_caddr_t); 588 void m_cat(struct mbuf *, struct mbuf *); 589 void m_catpkt(struct mbuf *, struct mbuf *); 590 int m_clget(struct mbuf *m, int how); 591 void *m_cljget(struct mbuf *m, int how, int size); 592 struct mbuf *m_collapse(struct mbuf *, int, int); 593 void m_copyback(struct mbuf *, int, int, c_caddr_t); 594 void m_copydata(const struct mbuf *, int, int, caddr_t); 595 struct mbuf *m_copym(struct mbuf *, int, int, int); 596 struct mbuf *m_copypacket(struct mbuf *, int); 597 void m_copy_pkthdr(struct mbuf *, struct mbuf *); 598 struct mbuf *m_copyup(struct mbuf *, int, int); 599 struct mbuf *m_defrag(struct mbuf *, int); 600 void m_demote_pkthdr(struct mbuf *); 601 void m_demote(struct mbuf *, int, int); 602 struct mbuf *m_devget(char *, int, int, struct ifnet *, 603 void (*)(char *, caddr_t, u_int)); 604 struct mbuf *m_dup(const struct mbuf *, int); 605 int m_dup_pkthdr(struct mbuf *, const struct mbuf *, int); 606 void m_extadd(struct mbuf *, caddr_t, u_int, 607 void (*)(struct mbuf *, void *, void *), void *, void *, 608 int, int); 609 u_int m_fixhdr(struct mbuf *); 610 struct mbuf *m_fragment(struct mbuf *, int, int); 611 void m_freem(struct mbuf *); 612 struct mbuf *m_get2(int, int, short, int); 613 struct mbuf *m_getjcl(int, short, int, int); 614 struct mbuf *m_getm2(struct mbuf *, int, int, short, int); 615 struct mbuf *m_getptr(struct mbuf *, int, int *); 616 u_int m_length(struct mbuf *, struct mbuf **); 617 int m_mbuftouio(struct uio *, const struct mbuf *, int); 618 void m_move_pkthdr(struct mbuf *, struct mbuf *); 619 int m_pkthdr_init(struct mbuf *, int); 620 struct mbuf *m_prepend(struct mbuf *, int, int); 621 void m_print(const struct mbuf *, int); 622 struct mbuf *m_pulldown(struct mbuf *, int, int, int *); 623 struct mbuf *m_pullup(struct mbuf *, int); 624 int m_sanity(struct mbuf *, int); 625 struct mbuf *m_split(struct mbuf *, int, int); 626 struct mbuf *m_uiotombuf(struct uio *, int, int, int, int); 627 struct mbuf *m_unshare(struct mbuf *, int); 628 629 static __inline int 630 m_gettype(int size) 631 { 632 int type; 633 634 switch (size) { 635 case MSIZE: 636 type = EXT_MBUF; 637 break; 638 case MCLBYTES: 639 type = EXT_CLUSTER; 640 break; 641 #if MJUMPAGESIZE != MCLBYTES 642 case MJUMPAGESIZE: 643 type = EXT_JUMBOP; 644 break; 645 #endif 646 case MJUM9BYTES: 647 type = EXT_JUMBO9; 648 break; 649 case MJUM16BYTES: 650 type = EXT_JUMBO16; 651 break; 652 default: 653 panic("%s: invalid cluster size %d", __func__, size); 654 } 655 656 return (type); 657 } 658 659 /* 660 * Associated an external reference counted buffer with an mbuf. 661 */ 662 static __inline void 663 m_extaddref(struct mbuf *m, caddr_t buf, u_int size, u_int *ref_cnt, 664 void (*freef)(struct mbuf *, void *, void *), void *arg1, void *arg2) 665 { 666 667 KASSERT(ref_cnt != NULL, ("%s: ref_cnt not provided", __func__)); 668 669 atomic_add_int(ref_cnt, 1); 670 m->m_flags |= M_EXT; 671 m->m_ext.ext_buf = buf; 672 m->m_ext.ext_cnt = ref_cnt; 673 m->m_data = m->m_ext.ext_buf; 674 m->m_ext.ext_size = size; 675 m->m_ext.ext_free = freef; 676 m->m_ext.ext_arg1 = arg1; 677 m->m_ext.ext_arg2 = arg2; 678 m->m_ext.ext_type = EXT_EXTREF; 679 m->m_ext.ext_flags = 0; 680 } 681 682 static __inline uma_zone_t 683 m_getzone(int size) 684 { 685 uma_zone_t zone; 686 687 switch (size) { 688 case MCLBYTES: 689 zone = zone_clust; 690 break; 691 #if MJUMPAGESIZE != MCLBYTES 692 case MJUMPAGESIZE: 693 zone = zone_jumbop; 694 break; 695 #endif 696 case MJUM9BYTES: 697 zone = zone_jumbo9; 698 break; 699 case MJUM16BYTES: 700 zone = zone_jumbo16; 701 break; 702 default: 703 panic("%s: invalid cluster size %d", __func__, size); 704 } 705 706 return (zone); 707 } 708 709 /* 710 * Initialize an mbuf with linear storage. 711 * 712 * Inline because the consumer text overhead will be roughly the same to 713 * initialize or call a function with this many parameters and M_PKTHDR 714 * should go away with constant propagation for !MGETHDR. 715 */ 716 static __inline int 717 m_init(struct mbuf *m, int how, short type, int flags) 718 { 719 int error; 720 721 m->m_next = NULL; 722 m->m_nextpkt = NULL; 723 m->m_data = m->m_dat; 724 m->m_len = 0; 725 m->m_flags = flags; 726 m->m_type = type; 727 if (flags & M_PKTHDR) 728 error = m_pkthdr_init(m, how); 729 else 730 error = 0; 731 732 MBUF_PROBE5(m__init, m, how, type, flags, error); 733 return (error); 734 } 735 736 static __inline struct mbuf * 737 m_get(int how, short type) 738 { 739 struct mbuf *m; 740 struct mb_args args; 741 742 args.flags = 0; 743 args.type = type; 744 m = uma_zalloc_arg(zone_mbuf, &args, how); 745 MBUF_PROBE3(m__get, how, type, m); 746 return (m); 747 } 748 749 static __inline struct mbuf * 750 m_gethdr(int how, short type) 751 { 752 struct mbuf *m; 753 struct mb_args args; 754 755 args.flags = M_PKTHDR; 756 args.type = type; 757 m = uma_zalloc_arg(zone_mbuf, &args, how); 758 MBUF_PROBE3(m__gethdr, how, type, m); 759 return (m); 760 } 761 762 static __inline struct mbuf * 763 m_getcl(int how, short type, int flags) 764 { 765 struct mbuf *m; 766 struct mb_args args; 767 768 args.flags = flags; 769 args.type = type; 770 m = uma_zalloc_arg(zone_pack, &args, how); 771 MBUF_PROBE4(m__getcl, how, type, flags, m); 772 return (m); 773 } 774 775 /* 776 * XXX: m_cljset() is a dangerous API. One must attach only a new, 777 * unreferenced cluster to an mbuf(9). It is not possible to assert 778 * that, so care can be taken only by users of the API. 779 */ 780 static __inline void 781 m_cljset(struct mbuf *m, void *cl, int type) 782 { 783 int size; 784 785 switch (type) { 786 case EXT_CLUSTER: 787 size = MCLBYTES; 788 break; 789 #if MJUMPAGESIZE != MCLBYTES 790 case EXT_JUMBOP: 791 size = MJUMPAGESIZE; 792 break; 793 #endif 794 case EXT_JUMBO9: 795 size = MJUM9BYTES; 796 break; 797 case EXT_JUMBO16: 798 size = MJUM16BYTES; 799 break; 800 default: 801 panic("%s: unknown cluster type %d", __func__, type); 802 break; 803 } 804 805 m->m_data = m->m_ext.ext_buf = cl; 806 m->m_ext.ext_free = m->m_ext.ext_arg1 = m->m_ext.ext_arg2 = NULL; 807 m->m_ext.ext_size = size; 808 m->m_ext.ext_type = type; 809 m->m_ext.ext_flags = EXT_FLAG_EMBREF; 810 m->m_ext.ext_count = 1; 811 m->m_flags |= M_EXT; 812 MBUF_PROBE3(m__cljset, m, cl, type); 813 } 814 815 static __inline void 816 m_chtype(struct mbuf *m, short new_type) 817 { 818 819 m->m_type = new_type; 820 } 821 822 static __inline void 823 m_clrprotoflags(struct mbuf *m) 824 { 825 826 while (m) { 827 m->m_flags &= ~M_PROTOFLAGS; 828 m = m->m_next; 829 } 830 } 831 832 static __inline struct mbuf * 833 m_last(struct mbuf *m) 834 { 835 836 while (m->m_next) 837 m = m->m_next; 838 return (m); 839 } 840 841 static inline u_int 842 m_extrefcnt(struct mbuf *m) 843 { 844 845 KASSERT(m->m_flags & M_EXT, ("%s: M_EXT missing", __func__)); 846 847 return ((m->m_ext.ext_flags & EXT_FLAG_EMBREF) ? m->m_ext.ext_count : 848 *m->m_ext.ext_cnt); 849 } 850 851 /* 852 * mbuf, cluster, and external object allocation macros (for compatibility 853 * purposes). 854 */ 855 #define M_MOVE_PKTHDR(to, from) m_move_pkthdr((to), (from)) 856 #define MGET(m, how, type) ((m) = m_get((how), (type))) 857 #define MGETHDR(m, how, type) ((m) = m_gethdr((how), (type))) 858 #define MCLGET(m, how) m_clget((m), (how)) 859 #define MEXTADD(m, buf, size, free, arg1, arg2, flags, type) \ 860 m_extadd((m), (caddr_t)(buf), (size), (free), (arg1), (arg2), \ 861 (flags), (type)) 862 #define m_getm(m, len, how, type) \ 863 m_getm2((m), (len), (how), (type), M_PKTHDR) 864 865 /* 866 * Evaluate TRUE if it's safe to write to the mbuf m's data region (this can 867 * be both the local data payload, or an external buffer area, depending on 868 * whether M_EXT is set). 869 */ 870 #define M_WRITABLE(m) (!((m)->m_flags & M_RDONLY) && \ 871 (!(((m)->m_flags & M_EXT)) || \ 872 (m_extrefcnt(m) == 1))) 873 874 /* Check if the supplied mbuf has a packet header, or else panic. */ 875 #define M_ASSERTPKTHDR(m) \ 876 KASSERT((m) != NULL && (m)->m_flags & M_PKTHDR, \ 877 ("%s: no mbuf packet header!", __func__)) 878 879 /* 880 * Ensure that the supplied mbuf is a valid, non-free mbuf. 881 * 882 * XXX: Broken at the moment. Need some UMA magic to make it work again. 883 */ 884 #define M_ASSERTVALID(m) \ 885 KASSERT((((struct mbuf *)m)->m_flags & 0) == 0, \ 886 ("%s: attempted use of a free mbuf!", __func__)) 887 888 /* 889 * Return the address of the start of the buffer associated with an mbuf, 890 * handling external storage, packet-header mbufs, and regular data mbufs. 891 */ 892 #define M_START(m) \ 893 (((m)->m_flags & M_EXT) ? (m)->m_ext.ext_buf : \ 894 ((m)->m_flags & M_PKTHDR) ? &(m)->m_pktdat[0] : \ 895 &(m)->m_dat[0]) 896 897 /* 898 * Return the size of the buffer associated with an mbuf, handling external 899 * storage, packet-header mbufs, and regular data mbufs. 900 */ 901 #define M_SIZE(m) \ 902 (((m)->m_flags & M_EXT) ? (m)->m_ext.ext_size : \ 903 ((m)->m_flags & M_PKTHDR) ? MHLEN : \ 904 MLEN) 905 906 /* 907 * Set the m_data pointer of a newly allocated mbuf to place an object of the 908 * specified size at the end of the mbuf, longword aligned. 909 * 910 * NB: Historically, we had M_ALIGN(), MH_ALIGN(), and MEXT_ALIGN() as 911 * separate macros, each asserting that it was called at the proper moment. 912 * This required callers to themselves test the storage type and call the 913 * right one. Rather than require callers to be aware of those layout 914 * decisions, we centralize here. 915 */ 916 static __inline void 917 m_align(struct mbuf *m, int len) 918 { 919 #ifdef INVARIANTS 920 const char *msg = "%s: not a virgin mbuf"; 921 #endif 922 int adjust; 923 924 KASSERT(m->m_data == M_START(m), (msg, __func__)); 925 926 adjust = M_SIZE(m) - len; 927 m->m_data += adjust &~ (sizeof(long)-1); 928 } 929 930 #define M_ALIGN(m, len) m_align(m, len) 931 #define MH_ALIGN(m, len) m_align(m, len) 932 #define MEXT_ALIGN(m, len) m_align(m, len) 933 934 /* 935 * Compute the amount of space available before the current start of data in 936 * an mbuf. 937 * 938 * The M_WRITABLE() is a temporary, conservative safety measure: the burden 939 * of checking writability of the mbuf data area rests solely with the caller. 940 * 941 * NB: In previous versions, M_LEADINGSPACE() would only check M_WRITABLE() 942 * for mbufs with external storage. We now allow mbuf-embedded data to be 943 * read-only as well. 944 */ 945 #define M_LEADINGSPACE(m) \ 946 (M_WRITABLE(m) ? ((m)->m_data - M_START(m)) : 0) 947 948 /* 949 * Compute the amount of space available after the end of data in an mbuf. 950 * 951 * The M_WRITABLE() is a temporary, conservative safety measure: the burden 952 * of checking writability of the mbuf data area rests solely with the caller. 953 * 954 * NB: In previous versions, M_TRAILINGSPACE() would only check M_WRITABLE() 955 * for mbufs with external storage. We now allow mbuf-embedded data to be 956 * read-only as well. 957 */ 958 #define M_TRAILINGSPACE(m) \ 959 (M_WRITABLE(m) ? \ 960 ((M_START(m) + M_SIZE(m)) - ((m)->m_data + (m)->m_len)) : 0) 961 962 /* 963 * Arrange to prepend space of size plen to mbuf m. If a new mbuf must be 964 * allocated, how specifies whether to wait. If the allocation fails, the 965 * original mbuf chain is freed and m is set to NULL. 966 */ 967 #define M_PREPEND(m, plen, how) do { \ 968 struct mbuf **_mmp = &(m); \ 969 struct mbuf *_mm = *_mmp; \ 970 int _mplen = (plen); \ 971 int __mhow = (how); \ 972 \ 973 MBUF_CHECKSLEEP(how); \ 974 if (M_LEADINGSPACE(_mm) >= _mplen) { \ 975 _mm->m_data -= _mplen; \ 976 _mm->m_len += _mplen; \ 977 } else \ 978 _mm = m_prepend(_mm, _mplen, __mhow); \ 979 if (_mm != NULL && _mm->m_flags & M_PKTHDR) \ 980 _mm->m_pkthdr.len += _mplen; \ 981 *_mmp = _mm; \ 982 } while (0) 983 984 /* 985 * Change mbuf to new type. This is a relatively expensive operation and 986 * should be avoided. 987 */ 988 #define MCHTYPE(m, t) m_chtype((m), (t)) 989 990 /* Length to m_copy to copy all. */ 991 #define M_COPYALL 1000000000 992 993 extern int max_datalen; /* MHLEN - max_hdr */ 994 extern int max_hdr; /* Largest link + protocol header */ 995 extern int max_linkhdr; /* Largest link-level header */ 996 extern int max_protohdr; /* Largest protocol header */ 997 extern int nmbclusters; /* Maximum number of clusters */ 998 999 /*- 1000 * Network packets may have annotations attached by affixing a list of 1001 * "packet tags" to the pkthdr structure. Packet tags are dynamically 1002 * allocated semi-opaque data structures that have a fixed header 1003 * (struct m_tag) that specifies the size of the memory block and a 1004 * <cookie,type> pair that identifies it. The cookie is a 32-bit unique 1005 * unsigned value used to identify a module or ABI. By convention this value 1006 * is chosen as the date+time that the module is created, expressed as the 1007 * number of seconds since the epoch (e.g., using date -u +'%s'). The type 1008 * value is an ABI/module-specific value that identifies a particular 1009 * annotation and is private to the module. For compatibility with systems 1010 * like OpenBSD that define packet tags w/o an ABI/module cookie, the value 1011 * PACKET_ABI_COMPAT is used to implement m_tag_get and m_tag_find 1012 * compatibility shim functions and several tag types are defined below. 1013 * Users that do not require compatibility should use a private cookie value 1014 * so that packet tag-related definitions can be maintained privately. 1015 * 1016 * Note that the packet tag returned by m_tag_alloc has the default memory 1017 * alignment implemented by malloc. To reference private data one can use a 1018 * construct like: 1019 * 1020 * struct m_tag *mtag = m_tag_alloc(...); 1021 * struct foo *p = (struct foo *)(mtag+1); 1022 * 1023 * if the alignment of struct m_tag is sufficient for referencing members of 1024 * struct foo. Otherwise it is necessary to embed struct m_tag within the 1025 * private data structure to insure proper alignment; e.g., 1026 * 1027 * struct foo { 1028 * struct m_tag tag; 1029 * ... 1030 * }; 1031 * struct foo *p = (struct foo *) m_tag_alloc(...); 1032 * struct m_tag *mtag = &p->tag; 1033 */ 1034 1035 /* 1036 * Persistent tags stay with an mbuf until the mbuf is reclaimed. Otherwise 1037 * tags are expected to ``vanish'' when they pass through a network 1038 * interface. For most interfaces this happens normally as the tags are 1039 * reclaimed when the mbuf is free'd. However in some special cases 1040 * reclaiming must be done manually. An example is packets that pass through 1041 * the loopback interface. Also, one must be careful to do this when 1042 * ``turning around'' packets (e.g., icmp_reflect). 1043 * 1044 * To mark a tag persistent bit-or this flag in when defining the tag id. 1045 * The tag will then be treated as described above. 1046 */ 1047 #define MTAG_PERSISTENT 0x800 1048 1049 #define PACKET_TAG_NONE 0 /* Nadda */ 1050 1051 /* Packet tags for use with PACKET_ABI_COMPAT. */ 1052 #define PACKET_TAG_IPSEC_IN_DONE 1 /* IPsec applied, in */ 1053 #define PACKET_TAG_IPSEC_OUT_DONE 2 /* IPsec applied, out */ 1054 #define PACKET_TAG_IPSEC_IN_CRYPTO_DONE 3 /* NIC IPsec crypto done */ 1055 #define PACKET_TAG_IPSEC_OUT_CRYPTO_NEEDED 4 /* NIC IPsec crypto req'ed */ 1056 #define PACKET_TAG_IPSEC_IN_COULD_DO_CRYPTO 5 /* NIC notifies IPsec */ 1057 #define PACKET_TAG_IPSEC_PENDING_TDB 6 /* Reminder to do IPsec */ 1058 #define PACKET_TAG_BRIDGE 7 /* Bridge processing done */ 1059 #define PACKET_TAG_GIF 8 /* GIF processing done */ 1060 #define PACKET_TAG_GRE 9 /* GRE processing done */ 1061 #define PACKET_TAG_IN_PACKET_CHECKSUM 10 /* NIC checksumming done */ 1062 #define PACKET_TAG_ENCAP 11 /* Encap. processing */ 1063 #define PACKET_TAG_IPSEC_SOCKET 12 /* IPSEC socket ref */ 1064 #define PACKET_TAG_IPSEC_HISTORY 13 /* IPSEC history */ 1065 #define PACKET_TAG_IPV6_INPUT 14 /* IPV6 input processing */ 1066 #define PACKET_TAG_DUMMYNET 15 /* dummynet info */ 1067 #define PACKET_TAG_DIVERT 17 /* divert info */ 1068 #define PACKET_TAG_IPFORWARD 18 /* ipforward info */ 1069 #define PACKET_TAG_MACLABEL (19 | MTAG_PERSISTENT) /* MAC label */ 1070 #define PACKET_TAG_PF (21 | MTAG_PERSISTENT) /* PF/ALTQ information */ 1071 #define PACKET_TAG_RTSOCKFAM 25 /* rtsock sa family */ 1072 #define PACKET_TAG_IPOPTIONS 27 /* Saved IP options */ 1073 #define PACKET_TAG_CARP 28 /* CARP info */ 1074 #define PACKET_TAG_IPSEC_NAT_T_PORTS 29 /* two uint16_t */ 1075 #define PACKET_TAG_ND_OUTGOING 30 /* ND outgoing */ 1076 1077 /* Specific cookies and tags. */ 1078 1079 /* Packet tag routines. */ 1080 struct m_tag *m_tag_alloc(u_int32_t, int, int, int); 1081 void m_tag_delete(struct mbuf *, struct m_tag *); 1082 void m_tag_delete_chain(struct mbuf *, struct m_tag *); 1083 void m_tag_free_default(struct m_tag *); 1084 struct m_tag *m_tag_locate(struct mbuf *, u_int32_t, int, struct m_tag *); 1085 struct m_tag *m_tag_copy(struct m_tag *, int); 1086 int m_tag_copy_chain(struct mbuf *, const struct mbuf *, int); 1087 void m_tag_delete_nonpersistent(struct mbuf *); 1088 1089 /* 1090 * Initialize the list of tags associated with an mbuf. 1091 */ 1092 static __inline void 1093 m_tag_init(struct mbuf *m) 1094 { 1095 1096 SLIST_INIT(&m->m_pkthdr.tags); 1097 } 1098 1099 /* 1100 * Set up the contents of a tag. Note that this does not fill in the free 1101 * method; the caller is expected to do that. 1102 * 1103 * XXX probably should be called m_tag_init, but that was already taken. 1104 */ 1105 static __inline void 1106 m_tag_setup(struct m_tag *t, u_int32_t cookie, int type, int len) 1107 { 1108 1109 t->m_tag_id = type; 1110 t->m_tag_len = len; 1111 t->m_tag_cookie = cookie; 1112 } 1113 1114 /* 1115 * Reclaim resources associated with a tag. 1116 */ 1117 static __inline void 1118 m_tag_free(struct m_tag *t) 1119 { 1120 1121 (*t->m_tag_free)(t); 1122 } 1123 1124 /* 1125 * Return the first tag associated with an mbuf. 1126 */ 1127 static __inline struct m_tag * 1128 m_tag_first(struct mbuf *m) 1129 { 1130 1131 return (SLIST_FIRST(&m->m_pkthdr.tags)); 1132 } 1133 1134 /* 1135 * Return the next tag in the list of tags associated with an mbuf. 1136 */ 1137 static __inline struct m_tag * 1138 m_tag_next(struct mbuf *m __unused, struct m_tag *t) 1139 { 1140 1141 return (SLIST_NEXT(t, m_tag_link)); 1142 } 1143 1144 /* 1145 * Prepend a tag to the list of tags associated with an mbuf. 1146 */ 1147 static __inline void 1148 m_tag_prepend(struct mbuf *m, struct m_tag *t) 1149 { 1150 1151 SLIST_INSERT_HEAD(&m->m_pkthdr.tags, t, m_tag_link); 1152 } 1153 1154 /* 1155 * Unlink a tag from the list of tags associated with an mbuf. 1156 */ 1157 static __inline void 1158 m_tag_unlink(struct mbuf *m, struct m_tag *t) 1159 { 1160 1161 SLIST_REMOVE(&m->m_pkthdr.tags, t, m_tag, m_tag_link); 1162 } 1163 1164 /* These are for OpenBSD compatibility. */ 1165 #define MTAG_ABI_COMPAT 0 /* compatibility ABI */ 1166 1167 static __inline struct m_tag * 1168 m_tag_get(int type, int length, int wait) 1169 { 1170 return (m_tag_alloc(MTAG_ABI_COMPAT, type, length, wait)); 1171 } 1172 1173 static __inline struct m_tag * 1174 m_tag_find(struct mbuf *m, int type, struct m_tag *start) 1175 { 1176 return (SLIST_EMPTY(&m->m_pkthdr.tags) ? (struct m_tag *)NULL : 1177 m_tag_locate(m, MTAG_ABI_COMPAT, type, start)); 1178 } 1179 1180 static __inline struct mbuf * 1181 m_free(struct mbuf *m) 1182 { 1183 struct mbuf *n = m->m_next; 1184 1185 MBUF_PROBE1(m__free, m); 1186 if ((m->m_flags & (M_PKTHDR|M_NOFREE)) == (M_PKTHDR|M_NOFREE)) 1187 m_tag_delete_chain(m, NULL); 1188 if (m->m_flags & M_EXT) 1189 mb_free_ext(m); 1190 else if ((m->m_flags & M_NOFREE) == 0) 1191 uma_zfree(zone_mbuf, m); 1192 return (n); 1193 } 1194 1195 static __inline int 1196 rt_m_getfib(struct mbuf *m) 1197 { 1198 KASSERT(m->m_flags & M_PKTHDR , ("Attempt to get FIB from non header mbuf.")); 1199 return (m->m_pkthdr.fibnum); 1200 } 1201 1202 #define M_GETFIB(_m) rt_m_getfib(_m) 1203 1204 #define M_SETFIB(_m, _fib) do { \ 1205 KASSERT((_m)->m_flags & M_PKTHDR, ("Attempt to set FIB on non header mbuf.")); \ 1206 ((_m)->m_pkthdr.fibnum) = (_fib); \ 1207 } while (0) 1208 1209 /* flags passed as first argument for "m_ether_tcpip_hash()" */ 1210 #define MBUF_HASHFLAG_L2 (1 << 2) 1211 #define MBUF_HASHFLAG_L3 (1 << 3) 1212 #define MBUF_HASHFLAG_L4 (1 << 4) 1213 1214 /* mbuf hashing helper routines */ 1215 uint32_t m_ether_tcpip_hash_init(void); 1216 uint32_t m_ether_tcpip_hash(const uint32_t, const struct mbuf *, const uint32_t); 1217 1218 #ifdef MBUF_PROFILING 1219 void m_profile(struct mbuf *m); 1220 #define M_PROFILE(m) m_profile(m) 1221 #else 1222 #define M_PROFILE(m) 1223 #endif 1224 1225 struct mbufq { 1226 STAILQ_HEAD(, mbuf) mq_head; 1227 int mq_len; 1228 int mq_maxlen; 1229 }; 1230 1231 static inline void 1232 mbufq_init(struct mbufq *mq, int maxlen) 1233 { 1234 1235 STAILQ_INIT(&mq->mq_head); 1236 mq->mq_maxlen = maxlen; 1237 mq->mq_len = 0; 1238 } 1239 1240 static inline struct mbuf * 1241 mbufq_flush(struct mbufq *mq) 1242 { 1243 struct mbuf *m; 1244 1245 m = STAILQ_FIRST(&mq->mq_head); 1246 STAILQ_INIT(&mq->mq_head); 1247 mq->mq_len = 0; 1248 return (m); 1249 } 1250 1251 static inline void 1252 mbufq_drain(struct mbufq *mq) 1253 { 1254 struct mbuf *m, *n; 1255 1256 n = mbufq_flush(mq); 1257 while ((m = n) != NULL) { 1258 n = STAILQ_NEXT(m, m_stailqpkt); 1259 m_freem(m); 1260 } 1261 } 1262 1263 static inline struct mbuf * 1264 mbufq_first(const struct mbufq *mq) 1265 { 1266 1267 return (STAILQ_FIRST(&mq->mq_head)); 1268 } 1269 1270 static inline struct mbuf * 1271 mbufq_last(const struct mbufq *mq) 1272 { 1273 1274 return (STAILQ_LAST(&mq->mq_head, mbuf, m_stailqpkt)); 1275 } 1276 1277 static inline int 1278 mbufq_full(const struct mbufq *mq) 1279 { 1280 1281 return (mq->mq_len >= mq->mq_maxlen); 1282 } 1283 1284 static inline int 1285 mbufq_len(const struct mbufq *mq) 1286 { 1287 1288 return (mq->mq_len); 1289 } 1290 1291 static inline int 1292 mbufq_enqueue(struct mbufq *mq, struct mbuf *m) 1293 { 1294 1295 if (mbufq_full(mq)) 1296 return (ENOBUFS); 1297 STAILQ_INSERT_TAIL(&mq->mq_head, m, m_stailqpkt); 1298 mq->mq_len++; 1299 return (0); 1300 } 1301 1302 static inline struct mbuf * 1303 mbufq_dequeue(struct mbufq *mq) 1304 { 1305 struct mbuf *m; 1306 1307 m = STAILQ_FIRST(&mq->mq_head); 1308 if (m) { 1309 STAILQ_REMOVE_HEAD(&mq->mq_head, m_stailqpkt); 1310 m->m_nextpkt = NULL; 1311 mq->mq_len--; 1312 } 1313 return (m); 1314 } 1315 1316 static inline void 1317 mbufq_prepend(struct mbufq *mq, struct mbuf *m) 1318 { 1319 1320 STAILQ_INSERT_HEAD(&mq->mq_head, m, m_stailqpkt); 1321 mq->mq_len++; 1322 } 1323 1324 /* 1325 * Note: this doesn't enforce the maximum list size for dst. 1326 */ 1327 static inline void 1328 mbufq_concat(struct mbufq *mq_dst, struct mbufq *mq_src) 1329 { 1330 1331 mq_dst->mq_len += mq_src->mq_len; 1332 STAILQ_CONCAT(&mq_dst->mq_head, &mq_src->mq_head); 1333 mq_src->mq_len = 0; 1334 } 1335 1336 #endif /* _KERNEL */ 1337 #endif /* !_SYS_MBUF_H_ */ 1338