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 * Record/packet header in first mbuf of chain; valid only if M_PKTHDR is set. 134 * Size ILP32: 48 135 * LP64: 56 136 * Compile-time assertions in uipc_mbuf.c test these values to ensure that 137 * they are correct. 138 */ 139 struct pkthdr { 140 struct ifnet *rcvif; /* rcv interface */ 141 SLIST_HEAD(packet_tags, m_tag) tags; /* list of packet tags */ 142 int32_t len; /* total packet length */ 143 144 /* Layer crossing persistent information. */ 145 uint32_t flowid; /* packet's 4-tuple system */ 146 uint64_t csum_flags; /* checksum and offload features */ 147 uint16_t fibnum; /* this packet should use this fib */ 148 uint8_t cosqos; /* class/quality of service */ 149 uint8_t rsstype; /* hash type */ 150 uint8_t l2hlen; /* layer 2 header length */ 151 uint8_t l3hlen; /* layer 3 header length */ 152 uint8_t l4hlen; /* layer 4 header length */ 153 uint8_t l5hlen; /* layer 5 header length */ 154 union { 155 uint8_t eight[8]; 156 uint16_t sixteen[4]; 157 uint32_t thirtytwo[2]; 158 uint64_t sixtyfour[1]; 159 uintptr_t unintptr[1]; 160 void *ptr; 161 } PH_per; 162 163 /* Layer specific non-persistent local storage for reassembly, etc. */ 164 union { 165 uint8_t eight[8]; 166 uint16_t sixteen[4]; 167 uint32_t thirtytwo[2]; 168 uint64_t sixtyfour[1]; 169 uintptr_t unintptr[1]; 170 void *ptr; 171 } PH_loc; 172 }; 173 #define ether_vtag PH_per.sixteen[0] 174 #define PH_vt PH_per 175 #define vt_nrecs sixteen[0] 176 #define tso_segsz PH_per.sixteen[1] 177 #define csum_phsum PH_per.sixteen[2] 178 #define csum_data PH_per.thirtytwo[1] 179 180 /* 181 * Description of external storage mapped into mbuf; valid only if M_EXT is 182 * set. 183 * Size ILP32: 28 184 * LP64: 48 185 * Compile-time assertions in uipc_mbuf.c test these values to ensure that 186 * they are correct. 187 */ 188 struct m_ext { 189 union { 190 volatile u_int ext_count; /* value of ref count info */ 191 volatile u_int *ext_cnt; /* pointer to ref count info */ 192 }; 193 caddr_t ext_buf; /* start of buffer */ 194 uint32_t ext_size; /* size of buffer, for ext_free */ 195 uint32_t ext_type:8, /* type of external storage */ 196 ext_flags:24; /* external storage mbuf flags */ 197 void (*ext_free) /* free routine if not the usual */ 198 (struct mbuf *, void *, void *); 199 void *ext_arg1; /* optional argument pointer */ 200 void *ext_arg2; /* optional argument pointer */ 201 }; 202 203 /* 204 * The core of the mbuf object along with some shortcut defines for practical 205 * purposes. 206 */ 207 struct mbuf { 208 /* 209 * Header present at the beginning of every mbuf. 210 * Size ILP32: 24 211 * LP64: 32 212 * Compile-time assertions in uipc_mbuf.c test these values to ensure 213 * that they are correct. 214 */ 215 union { /* next buffer in chain */ 216 struct mbuf *m_next; 217 SLIST_ENTRY(mbuf) m_slist; 218 STAILQ_ENTRY(mbuf) m_stailq; 219 }; 220 union { /* next chain in queue/record */ 221 struct mbuf *m_nextpkt; 222 SLIST_ENTRY(mbuf) m_slistpkt; 223 STAILQ_ENTRY(mbuf) m_stailqpkt; 224 }; 225 caddr_t m_data; /* location of data */ 226 int32_t m_len; /* amount of data in this mbuf */ 227 uint32_t m_type:8, /* type of data in this mbuf */ 228 m_flags:24; /* flags; see below */ 229 #if !defined(__LP64__) 230 uint32_t m_pad; /* pad for 64bit alignment */ 231 #endif 232 233 /* 234 * A set of optional headers (packet header, external storage header) 235 * and internal data storage. Historically, these arrays were sized 236 * to MHLEN (space left after a packet header) and MLEN (space left 237 * after only a regular mbuf header); they are now variable size in 238 * order to support future work on variable-size mbufs. 239 */ 240 union { 241 struct { 242 struct pkthdr m_pkthdr; /* M_PKTHDR set */ 243 union { 244 struct m_ext m_ext; /* M_EXT set */ 245 char m_pktdat[0]; 246 }; 247 }; 248 char m_dat[0]; /* !M_PKTHDR, !M_EXT */ 249 }; 250 }; 251 252 /* 253 * mbuf flags of global significance and layer crossing. 254 * Those of only protocol/layer specific significance are to be mapped 255 * to M_PROTO[1-12] and cleared at layer handoff boundaries. 256 * NB: Limited to the lower 24 bits. 257 */ 258 #define M_EXT 0x00000001 /* has associated external storage */ 259 #define M_PKTHDR 0x00000002 /* start of record */ 260 #define M_EOR 0x00000004 /* end of record */ 261 #define M_RDONLY 0x00000008 /* associated data is marked read-only */ 262 #define M_BCAST 0x00000010 /* send/received as link-level broadcast */ 263 #define M_MCAST 0x00000020 /* send/received as link-level multicast */ 264 #define M_PROMISC 0x00000040 /* packet was not for us */ 265 #define M_VLANTAG 0x00000080 /* ether_vtag is valid */ 266 #define M_UNUSED_8 0x00000100 /* --available-- */ 267 #define M_NOFREE 0x00000200 /* do not free mbuf, embedded in cluster */ 268 269 #define M_PROTO1 0x00001000 /* protocol-specific */ 270 #define M_PROTO2 0x00002000 /* protocol-specific */ 271 #define M_PROTO3 0x00004000 /* protocol-specific */ 272 #define M_PROTO4 0x00008000 /* protocol-specific */ 273 #define M_PROTO5 0x00010000 /* protocol-specific */ 274 #define M_PROTO6 0x00020000 /* protocol-specific */ 275 #define M_PROTO7 0x00040000 /* protocol-specific */ 276 #define M_PROTO8 0x00080000 /* protocol-specific */ 277 #define M_PROTO9 0x00100000 /* protocol-specific */ 278 #define M_PROTO10 0x00200000 /* protocol-specific */ 279 #define M_PROTO11 0x00400000 /* protocol-specific */ 280 #define M_PROTO12 0x00800000 /* protocol-specific */ 281 282 /* 283 * Flags to purge when crossing layers. 284 */ 285 #define M_PROTOFLAGS \ 286 (M_PROTO1|M_PROTO2|M_PROTO3|M_PROTO4|M_PROTO5|M_PROTO6|M_PROTO7|M_PROTO8|\ 287 M_PROTO9|M_PROTO10|M_PROTO11|M_PROTO12) 288 289 /* 290 * Flags preserved when copying m_pkthdr. 291 */ 292 #define M_COPYFLAGS \ 293 (M_PKTHDR|M_EOR|M_RDONLY|M_BCAST|M_MCAST|M_PROMISC|M_VLANTAG| \ 294 M_PROTOFLAGS) 295 296 /* 297 * Mbuf flag description for use with printf(9) %b identifier. 298 */ 299 #define M_FLAG_BITS \ 300 "\20\1M_EXT\2M_PKTHDR\3M_EOR\4M_RDONLY\5M_BCAST\6M_MCAST" \ 301 "\7M_PROMISC\10M_VLANTAG" 302 #define M_FLAG_PROTOBITS \ 303 "\15M_PROTO1\16M_PROTO2\17M_PROTO3\20M_PROTO4\21M_PROTO5" \ 304 "\22M_PROTO6\23M_PROTO7\24M_PROTO8\25M_PROTO9\26M_PROTO10" \ 305 "\27M_PROTO11\30M_PROTO12" 306 #define M_FLAG_PRINTF (M_FLAG_BITS M_FLAG_PROTOBITS) 307 308 /* 309 * Network interface cards are able to hash protocol fields (such as IPv4 310 * addresses and TCP port numbers) classify packets into flows. These flows 311 * can then be used to maintain ordering while delivering packets to the OS 312 * via parallel input queues, as well as to provide a stateless affinity 313 * model. NIC drivers can pass up the hash via m->m_pkthdr.flowid, and set 314 * m_flag fields to indicate how the hash should be interpreted by the 315 * network stack. 316 * 317 * Most NICs support RSS, which provides ordering and explicit affinity, and 318 * use the hash m_flag bits to indicate what header fields were covered by 319 * the hash. M_HASHTYPE_OPAQUE can be set by non-RSS cards or configurations 320 * that provide an opaque flow identifier, allowing for ordering and 321 * distribution without explicit affinity. 322 */ 323 /* Microsoft RSS standard hash types */ 324 #define M_HASHTYPE_NONE 0 325 #define M_HASHTYPE_RSS_IPV4 1 /* IPv4 2-tuple */ 326 #define M_HASHTYPE_RSS_TCP_IPV4 2 /* TCPv4 4-tuple */ 327 #define M_HASHTYPE_RSS_IPV6 3 /* IPv6 2-tuple */ 328 #define M_HASHTYPE_RSS_TCP_IPV6 4 /* TCPv6 4-tuple */ 329 #define M_HASHTYPE_RSS_IPV6_EX 5 /* IPv6 2-tuple + ext hdrs */ 330 #define M_HASHTYPE_RSS_TCP_IPV6_EX 6 /* TCPv6 4-tiple + ext hdrs */ 331 /* Non-standard RSS hash types */ 332 #define M_HASHTYPE_RSS_UDP_IPV4 7 /* IPv4 UDP 4-tuple */ 333 #define M_HASHTYPE_RSS_UDP_IPV4_EX 8 /* IPv4 UDP 4-tuple + ext hdrs */ 334 #define M_HASHTYPE_RSS_UDP_IPV6 9 /* IPv6 UDP 4-tuple */ 335 #define M_HASHTYPE_RSS_UDP_IPV6_EX 10 /* IPv6 UDP 4-tuple + ext hdrs */ 336 337 #define M_HASHTYPE_OPAQUE 255 /* ordering, not affinity */ 338 339 #define M_HASHTYPE_CLEAR(m) ((m)->m_pkthdr.rsstype = 0) 340 #define M_HASHTYPE_GET(m) ((m)->m_pkthdr.rsstype) 341 #define M_HASHTYPE_SET(m, v) ((m)->m_pkthdr.rsstype = (v)) 342 #define M_HASHTYPE_TEST(m, v) (M_HASHTYPE_GET(m) == (v)) 343 344 /* 345 * COS/QOS class and quality of service tags. 346 * It uses DSCP code points as base. 347 */ 348 #define QOS_DSCP_CS0 0x00 349 #define QOS_DSCP_DEF QOS_DSCP_CS0 350 #define QOS_DSCP_CS1 0x20 351 #define QOS_DSCP_AF11 0x28 352 #define QOS_DSCP_AF12 0x30 353 #define QOS_DSCP_AF13 0x38 354 #define QOS_DSCP_CS2 0x40 355 #define QOS_DSCP_AF21 0x48 356 #define QOS_DSCP_AF22 0x50 357 #define QOS_DSCP_AF23 0x58 358 #define QOS_DSCP_CS3 0x60 359 #define QOS_DSCP_AF31 0x68 360 #define QOS_DSCP_AF32 0x70 361 #define QOS_DSCP_AF33 0x78 362 #define QOS_DSCP_CS4 0x80 363 #define QOS_DSCP_AF41 0x88 364 #define QOS_DSCP_AF42 0x90 365 #define QOS_DSCP_AF43 0x98 366 #define QOS_DSCP_CS5 0xa0 367 #define QOS_DSCP_EF 0xb8 368 #define QOS_DSCP_CS6 0xc0 369 #define QOS_DSCP_CS7 0xe0 370 371 /* 372 * External mbuf storage buffer types. 373 */ 374 #define EXT_CLUSTER 1 /* mbuf cluster */ 375 #define EXT_SFBUF 2 /* sendfile(2)'s sf_buf */ 376 #define EXT_JUMBOP 3 /* jumbo cluster page sized */ 377 #define EXT_JUMBO9 4 /* jumbo cluster 9216 bytes */ 378 #define EXT_JUMBO16 5 /* jumbo cluster 16184 bytes */ 379 #define EXT_PACKET 6 /* mbuf+cluster from packet zone */ 380 #define EXT_MBUF 7 /* external mbuf reference (M_IOVEC) */ 381 #define EXT_SFBUF_NOCACHE 8 /* sendfile(2)'s sf_buf not to be cached */ 382 383 #define EXT_VENDOR1 224 /* for vendor-internal use */ 384 #define EXT_VENDOR2 225 /* for vendor-internal use */ 385 #define EXT_VENDOR3 226 /* for vendor-internal use */ 386 #define EXT_VENDOR4 227 /* for vendor-internal use */ 387 388 #define EXT_EXP1 244 /* for experimental use */ 389 #define EXT_EXP2 245 /* for experimental use */ 390 #define EXT_EXP3 246 /* for experimental use */ 391 #define EXT_EXP4 247 /* for experimental use */ 392 393 #define EXT_NET_DRV 252 /* custom ext_buf provided by net driver(s) */ 394 #define EXT_MOD_TYPE 253 /* custom module's ext_buf type */ 395 #define EXT_DISPOSABLE 254 /* can throw this buffer away w/page flipping */ 396 #define EXT_EXTREF 255 /* has externally maintained ext_cnt ptr */ 397 398 /* 399 * Flags for external mbuf buffer types. 400 * NB: limited to the lower 24 bits. 401 */ 402 #define EXT_FLAG_EMBREF 0x000001 /* embedded ext_count */ 403 #define EXT_FLAG_EXTREF 0x000002 /* external ext_cnt, notyet */ 404 #define EXT_FLAG_NOFREE 0x000010 /* don't free mbuf to pool, notyet */ 405 406 #define EXT_FLAG_VENDOR1 0x010000 /* for vendor-internal use */ 407 #define EXT_FLAG_VENDOR2 0x020000 /* for vendor-internal use */ 408 #define EXT_FLAG_VENDOR3 0x040000 /* for vendor-internal use */ 409 #define EXT_FLAG_VENDOR4 0x080000 /* for vendor-internal use */ 410 411 #define EXT_FLAG_EXP1 0x100000 /* for experimental use */ 412 #define EXT_FLAG_EXP2 0x200000 /* for experimental use */ 413 #define EXT_FLAG_EXP3 0x400000 /* for experimental use */ 414 #define EXT_FLAG_EXP4 0x800000 /* for experimental use */ 415 416 /* 417 * EXT flag description for use with printf(9) %b identifier. 418 */ 419 #define EXT_FLAG_BITS \ 420 "\20\1EXT_FLAG_EMBREF\2EXT_FLAG_EXTREF\5EXT_FLAG_NOFREE" \ 421 "\21EXT_FLAG_VENDOR1\22EXT_FLAG_VENDOR2\23EXT_FLAG_VENDOR3" \ 422 "\24EXT_FLAG_VENDOR4\25EXT_FLAG_EXP1\26EXT_FLAG_EXP2\27EXT_FLAG_EXP3" \ 423 "\30EXT_FLAG_EXP4" 424 425 /* 426 * External reference/free functions. 427 */ 428 void sf_ext_free(void *, void *); 429 void sf_ext_free_nocache(void *, void *); 430 431 /* 432 * Flags indicating checksum, segmentation and other offload work to be 433 * done, or already done, by hardware or lower layers. It is split into 434 * separate inbound and outbound flags. 435 * 436 * Outbound flags that are set by upper protocol layers requesting lower 437 * layers, or ideally the hardware, to perform these offloading tasks. 438 * For outbound packets this field and its flags can be directly tested 439 * against ifnet if_hwassist. 440 */ 441 #define CSUM_IP 0x00000001 /* IP header checksum offload */ 442 #define CSUM_IP_UDP 0x00000002 /* UDP checksum offload */ 443 #define CSUM_IP_TCP 0x00000004 /* TCP checksum offload */ 444 #define CSUM_IP_SCTP 0x00000008 /* SCTP checksum offload */ 445 #define CSUM_IP_TSO 0x00000010 /* TCP segmentation offload */ 446 #define CSUM_IP_ISCSI 0x00000020 /* iSCSI checksum offload */ 447 448 #define CSUM_IP6_UDP 0x00000200 /* UDP checksum offload */ 449 #define CSUM_IP6_TCP 0x00000400 /* TCP checksum offload */ 450 #define CSUM_IP6_SCTP 0x00000800 /* SCTP checksum offload */ 451 #define CSUM_IP6_TSO 0x00001000 /* TCP segmentation offload */ 452 #define CSUM_IP6_ISCSI 0x00002000 /* iSCSI checksum offload */ 453 454 /* Inbound checksum support where the checksum was verified by hardware. */ 455 #define CSUM_L3_CALC 0x01000000 /* calculated layer 3 csum */ 456 #define CSUM_L3_VALID 0x02000000 /* checksum is correct */ 457 #define CSUM_L4_CALC 0x04000000 /* calculated layer 4 csum */ 458 #define CSUM_L4_VALID 0x08000000 /* checksum is correct */ 459 #define CSUM_L5_CALC 0x10000000 /* calculated layer 5 csum */ 460 #define CSUM_L5_VALID 0x20000000 /* checksum is correct */ 461 #define CSUM_COALESED 0x40000000 /* contains merged segments */ 462 463 /* 464 * CSUM flag description for use with printf(9) %b identifier. 465 */ 466 #define CSUM_BITS \ 467 "\20\1CSUM_IP\2CSUM_IP_UDP\3CSUM_IP_TCP\4CSUM_IP_SCTP\5CSUM_IP_TSO" \ 468 "\6CSUM_IP_ISCSI" \ 469 "\12CSUM_IP6_UDP\13CSUM_IP6_TCP\14CSUM_IP6_SCTP\15CSUM_IP6_TSO" \ 470 "\16CSUM_IP6_ISCSI" \ 471 "\31CSUM_L3_CALC\32CSUM_L3_VALID\33CSUM_L4_CALC\34CSUM_L4_VALID" \ 472 "\35CSUM_L5_CALC\36CSUM_L5_VALID\37CSUM_COALESED" 473 474 /* CSUM flags compatibility mappings. */ 475 #define CSUM_IP_CHECKED CSUM_L3_CALC 476 #define CSUM_IP_VALID CSUM_L3_VALID 477 #define CSUM_DATA_VALID CSUM_L4_VALID 478 #define CSUM_PSEUDO_HDR CSUM_L4_CALC 479 #define CSUM_SCTP_VALID CSUM_L4_VALID 480 #define CSUM_DELAY_DATA (CSUM_TCP|CSUM_UDP) 481 #define CSUM_DELAY_IP CSUM_IP /* Only v4, no v6 IP hdr csum */ 482 #define CSUM_DELAY_DATA_IPV6 (CSUM_TCP_IPV6|CSUM_UDP_IPV6) 483 #define CSUM_DATA_VALID_IPV6 CSUM_DATA_VALID 484 #define CSUM_TCP CSUM_IP_TCP 485 #define CSUM_UDP CSUM_IP_UDP 486 #define CSUM_SCTP CSUM_IP_SCTP 487 #define CSUM_TSO (CSUM_IP_TSO|CSUM_IP6_TSO) 488 #define CSUM_UDP_IPV6 CSUM_IP6_UDP 489 #define CSUM_TCP_IPV6 CSUM_IP6_TCP 490 #define CSUM_SCTP_IPV6 CSUM_IP6_SCTP 491 492 /* 493 * mbuf types describing the content of the mbuf (including external storage). 494 */ 495 #define MT_NOTMBUF 0 /* USED INTERNALLY ONLY! Object is not mbuf */ 496 #define MT_DATA 1 /* dynamic (data) allocation */ 497 #define MT_HEADER MT_DATA /* packet header, use M_PKTHDR instead */ 498 499 #define MT_VENDOR1 4 /* for vendor-internal use */ 500 #define MT_VENDOR2 5 /* for vendor-internal use */ 501 #define MT_VENDOR3 6 /* for vendor-internal use */ 502 #define MT_VENDOR4 7 /* for vendor-internal use */ 503 504 #define MT_SONAME 8 /* socket name */ 505 506 #define MT_EXP1 9 /* for experimental use */ 507 #define MT_EXP2 10 /* for experimental use */ 508 #define MT_EXP3 11 /* for experimental use */ 509 #define MT_EXP4 12 /* for experimental use */ 510 511 #define MT_CONTROL 14 /* extra-data protocol message */ 512 #define MT_OOBDATA 15 /* expedited data */ 513 #define MT_NTYPES 16 /* number of mbuf types for mbtypes[] */ 514 515 #define MT_NOINIT 255 /* Not a type but a flag to allocate 516 a non-initialized mbuf */ 517 518 /* 519 * String names of mbuf-related UMA(9) and malloc(9) types. Exposed to 520 * !_KERNEL so that monitoring tools can look up the zones with 521 * libmemstat(3). 522 */ 523 #define MBUF_MEM_NAME "mbuf" 524 #define MBUF_CLUSTER_MEM_NAME "mbuf_cluster" 525 #define MBUF_PACKET_MEM_NAME "mbuf_packet" 526 #define MBUF_JUMBOP_MEM_NAME "mbuf_jumbo_page" 527 #define MBUF_JUMBO9_MEM_NAME "mbuf_jumbo_9k" 528 #define MBUF_JUMBO16_MEM_NAME "mbuf_jumbo_16k" 529 #define MBUF_TAG_MEM_NAME "mbuf_tag" 530 #define MBUF_EXTREFCNT_MEM_NAME "mbuf_ext_refcnt" 531 532 #ifdef _KERNEL 533 534 #ifdef WITNESS 535 #define MBUF_CHECKSLEEP(how) do { \ 536 if (how == M_WAITOK) \ 537 WITNESS_WARN(WARN_GIANTOK | WARN_SLEEPOK, NULL, \ 538 "Sleeping in \"%s\"", __func__); \ 539 } while (0) 540 #else 541 #define MBUF_CHECKSLEEP(how) 542 #endif 543 544 /* 545 * Network buffer allocation API 546 * 547 * The rest of it is defined in kern/kern_mbuf.c 548 */ 549 extern uma_zone_t zone_mbuf; 550 extern uma_zone_t zone_clust; 551 extern uma_zone_t zone_pack; 552 extern uma_zone_t zone_jumbop; 553 extern uma_zone_t zone_jumbo9; 554 extern uma_zone_t zone_jumbo16; 555 556 void mb_dupcl(struct mbuf *, struct mbuf *); 557 void mb_free_ext(struct mbuf *); 558 void m_adj(struct mbuf *, int); 559 int m_apply(struct mbuf *, int, int, 560 int (*)(void *, void *, u_int), void *); 561 int m_append(struct mbuf *, int, c_caddr_t); 562 void m_cat(struct mbuf *, struct mbuf *); 563 void m_catpkt(struct mbuf *, struct mbuf *); 564 int m_clget(struct mbuf *m, int how); 565 void *m_cljget(struct mbuf *m, int how, int size); 566 struct mbuf *m_collapse(struct mbuf *, int, int); 567 void m_copyback(struct mbuf *, int, int, c_caddr_t); 568 void m_copydata(const struct mbuf *, int, int, caddr_t); 569 struct mbuf *m_copym(struct mbuf *, int, int, int); 570 struct mbuf *m_copypacket(struct mbuf *, int); 571 void m_copy_pkthdr(struct mbuf *, struct mbuf *); 572 struct mbuf *m_copyup(struct mbuf *, int, int); 573 struct mbuf *m_defrag(struct mbuf *, int); 574 void m_demote_pkthdr(struct mbuf *); 575 void m_demote(struct mbuf *, int, int); 576 struct mbuf *m_devget(char *, int, int, struct ifnet *, 577 void (*)(char *, caddr_t, u_int)); 578 struct mbuf *m_dup(const struct mbuf *, int); 579 int m_dup_pkthdr(struct mbuf *, const struct mbuf *, int); 580 void m_extadd(struct mbuf *, caddr_t, u_int, 581 void (*)(struct mbuf *, void *, void *), void *, void *, 582 int, int); 583 u_int m_fixhdr(struct mbuf *); 584 struct mbuf *m_fragment(struct mbuf *, int, int); 585 void m_freem(struct mbuf *); 586 struct mbuf *m_get2(int, int, short, int); 587 struct mbuf *m_getjcl(int, short, int, int); 588 struct mbuf *m_getm2(struct mbuf *, int, int, short, int); 589 struct mbuf *m_getptr(struct mbuf *, int, int *); 590 u_int m_length(struct mbuf *, struct mbuf **); 591 int m_mbuftouio(struct uio *, struct mbuf *, int); 592 void m_move_pkthdr(struct mbuf *, struct mbuf *); 593 int m_pkthdr_init(struct mbuf *, int); 594 struct mbuf *m_prepend(struct mbuf *, int, int); 595 void m_print(const struct mbuf *, int); 596 struct mbuf *m_pulldown(struct mbuf *, int, int, int *); 597 struct mbuf *m_pullup(struct mbuf *, int); 598 int m_sanity(struct mbuf *, int); 599 struct mbuf *m_split(struct mbuf *, int, int); 600 struct mbuf *m_uiotombuf(struct uio *, int, int, int, int); 601 struct mbuf *m_unshare(struct mbuf *, int); 602 603 static __inline int 604 m_gettype(int size) 605 { 606 int type; 607 608 switch (size) { 609 case MSIZE: 610 type = EXT_MBUF; 611 break; 612 case MCLBYTES: 613 type = EXT_CLUSTER; 614 break; 615 #if MJUMPAGESIZE != MCLBYTES 616 case MJUMPAGESIZE: 617 type = EXT_JUMBOP; 618 break; 619 #endif 620 case MJUM9BYTES: 621 type = EXT_JUMBO9; 622 break; 623 case MJUM16BYTES: 624 type = EXT_JUMBO16; 625 break; 626 default: 627 panic("%s: invalid cluster size %d", __func__, size); 628 } 629 630 return (type); 631 } 632 633 /* 634 * Associated an external reference counted buffer with an mbuf. 635 */ 636 static __inline void 637 m_extaddref(struct mbuf *m, caddr_t buf, u_int size, u_int *ref_cnt, 638 void (*freef)(struct mbuf *, void *, void *), void *arg1, void *arg2) 639 { 640 641 KASSERT(ref_cnt != NULL, ("%s: ref_cnt not provided", __func__)); 642 643 atomic_add_int(ref_cnt, 1); 644 m->m_flags |= M_EXT; 645 m->m_ext.ext_buf = buf; 646 m->m_ext.ext_cnt = ref_cnt; 647 m->m_data = m->m_ext.ext_buf; 648 m->m_ext.ext_size = size; 649 m->m_ext.ext_free = freef; 650 m->m_ext.ext_arg1 = arg1; 651 m->m_ext.ext_arg2 = arg2; 652 m->m_ext.ext_type = EXT_EXTREF; 653 m->m_ext.ext_flags = 0; 654 } 655 656 static __inline uma_zone_t 657 m_getzone(int size) 658 { 659 uma_zone_t zone; 660 661 switch (size) { 662 case MCLBYTES: 663 zone = zone_clust; 664 break; 665 #if MJUMPAGESIZE != MCLBYTES 666 case MJUMPAGESIZE: 667 zone = zone_jumbop; 668 break; 669 #endif 670 case MJUM9BYTES: 671 zone = zone_jumbo9; 672 break; 673 case MJUM16BYTES: 674 zone = zone_jumbo16; 675 break; 676 default: 677 panic("%s: invalid cluster size %d", __func__, size); 678 } 679 680 return (zone); 681 } 682 683 /* 684 * Initialize an mbuf with linear storage. 685 * 686 * Inline because the consumer text overhead will be roughly the same to 687 * initialize or call a function with this many parameters and M_PKTHDR 688 * should go away with constant propagation for !MGETHDR. 689 */ 690 static __inline int 691 m_init(struct mbuf *m, int how, short type, int flags) 692 { 693 int error; 694 695 m->m_next = NULL; 696 m->m_nextpkt = NULL; 697 m->m_data = m->m_dat; 698 m->m_len = 0; 699 m->m_flags = flags; 700 m->m_type = type; 701 if (flags & M_PKTHDR) 702 error = m_pkthdr_init(m, how); 703 else 704 error = 0; 705 706 MBUF_PROBE5(m__init, m, how, type, flags, error); 707 return (error); 708 } 709 710 static __inline struct mbuf * 711 m_get(int how, short type) 712 { 713 struct mbuf *m; 714 struct mb_args args; 715 716 args.flags = 0; 717 args.type = type; 718 m = uma_zalloc_arg(zone_mbuf, &args, how); 719 MBUF_PROBE3(m__get, how, type, m); 720 return (m); 721 } 722 723 static __inline struct mbuf * 724 m_gethdr(int how, short type) 725 { 726 struct mbuf *m; 727 struct mb_args args; 728 729 args.flags = M_PKTHDR; 730 args.type = type; 731 m = uma_zalloc_arg(zone_mbuf, &args, how); 732 MBUF_PROBE3(m__gethdr, how, type, m); 733 return (m); 734 } 735 736 static __inline struct mbuf * 737 m_getcl(int how, short type, int flags) 738 { 739 struct mbuf *m; 740 struct mb_args args; 741 742 args.flags = flags; 743 args.type = type; 744 m = uma_zalloc_arg(zone_pack, &args, how); 745 MBUF_PROBE4(m__getcl, how, type, flags, m); 746 return (m); 747 } 748 749 /* 750 * XXX: m_cljset() is a dangerous API. One must attach only a new, 751 * unreferenced cluster to an mbuf(9). It is not possible to assert 752 * that, so care can be taken only by users of the API. 753 */ 754 static __inline void 755 m_cljset(struct mbuf *m, void *cl, int type) 756 { 757 int size; 758 759 switch (type) { 760 case EXT_CLUSTER: 761 size = MCLBYTES; 762 break; 763 #if MJUMPAGESIZE != MCLBYTES 764 case EXT_JUMBOP: 765 size = MJUMPAGESIZE; 766 break; 767 #endif 768 case EXT_JUMBO9: 769 size = MJUM9BYTES; 770 break; 771 case EXT_JUMBO16: 772 size = MJUM16BYTES; 773 break; 774 default: 775 panic("%s: unknown cluster type %d", __func__, type); 776 break; 777 } 778 779 m->m_data = m->m_ext.ext_buf = cl; 780 m->m_ext.ext_free = m->m_ext.ext_arg1 = m->m_ext.ext_arg2 = NULL; 781 m->m_ext.ext_size = size; 782 m->m_ext.ext_type = type; 783 m->m_ext.ext_flags = EXT_FLAG_EMBREF; 784 m->m_ext.ext_count = 1; 785 m->m_flags |= M_EXT; 786 MBUF_PROBE3(m__cljset, m, cl, type); 787 } 788 789 static __inline void 790 m_chtype(struct mbuf *m, short new_type) 791 { 792 793 m->m_type = new_type; 794 } 795 796 static __inline void 797 m_clrprotoflags(struct mbuf *m) 798 { 799 800 while (m) { 801 m->m_flags &= ~M_PROTOFLAGS; 802 m = m->m_next; 803 } 804 } 805 806 static __inline struct mbuf * 807 m_last(struct mbuf *m) 808 { 809 810 while (m->m_next) 811 m = m->m_next; 812 return (m); 813 } 814 815 static inline u_int 816 m_extrefcnt(struct mbuf *m) 817 { 818 819 KASSERT(m->m_flags & M_EXT, ("%s: M_EXT missing", __func__)); 820 821 return ((m->m_ext.ext_flags & EXT_FLAG_EMBREF) ? m->m_ext.ext_count : 822 *m->m_ext.ext_cnt); 823 } 824 825 /* 826 * mbuf, cluster, and external object allocation macros (for compatibility 827 * purposes). 828 */ 829 #define M_MOVE_PKTHDR(to, from) m_move_pkthdr((to), (from)) 830 #define MGET(m, how, type) ((m) = m_get((how), (type))) 831 #define MGETHDR(m, how, type) ((m) = m_gethdr((how), (type))) 832 #define MCLGET(m, how) m_clget((m), (how)) 833 #define MEXTADD(m, buf, size, free, arg1, arg2, flags, type) \ 834 m_extadd((m), (caddr_t)(buf), (size), (free), (arg1), (arg2), \ 835 (flags), (type)) 836 #define m_getm(m, len, how, type) \ 837 m_getm2((m), (len), (how), (type), M_PKTHDR) 838 839 /* 840 * Evaluate TRUE if it's safe to write to the mbuf m's data region (this can 841 * be both the local data payload, or an external buffer area, depending on 842 * whether M_EXT is set). 843 */ 844 #define M_WRITABLE(m) (!((m)->m_flags & M_RDONLY) && \ 845 (!(((m)->m_flags & M_EXT)) || \ 846 (m_extrefcnt(m) == 1))) 847 848 /* Check if the supplied mbuf has a packet header, or else panic. */ 849 #define M_ASSERTPKTHDR(m) \ 850 KASSERT((m) != NULL && (m)->m_flags & M_PKTHDR, \ 851 ("%s: no mbuf packet header!", __func__)) 852 853 /* 854 * Ensure that the supplied mbuf is a valid, non-free mbuf. 855 * 856 * XXX: Broken at the moment. Need some UMA magic to make it work again. 857 */ 858 #define M_ASSERTVALID(m) \ 859 KASSERT((((struct mbuf *)m)->m_flags & 0) == 0, \ 860 ("%s: attempted use of a free mbuf!", __func__)) 861 862 /* 863 * Return the address of the start of the buffer associated with an mbuf, 864 * handling external storage, packet-header mbufs, and regular data mbufs. 865 */ 866 #define M_START(m) \ 867 (((m)->m_flags & M_EXT) ? (m)->m_ext.ext_buf : \ 868 ((m)->m_flags & M_PKTHDR) ? &(m)->m_pktdat[0] : \ 869 &(m)->m_dat[0]) 870 871 /* 872 * Return the size of the buffer associated with an mbuf, handling external 873 * storage, packet-header mbufs, and regular data mbufs. 874 */ 875 #define M_SIZE(m) \ 876 (((m)->m_flags & M_EXT) ? (m)->m_ext.ext_size : \ 877 ((m)->m_flags & M_PKTHDR) ? MHLEN : \ 878 MLEN) 879 880 /* 881 * Set the m_data pointer of a newly allocated mbuf to place an object of the 882 * specified size at the end of the mbuf, longword aligned. 883 * 884 * NB: Historically, we had M_ALIGN(), MH_ALIGN(), and MEXT_ALIGN() as 885 * separate macros, each asserting that it was called at the proper moment. 886 * This required callers to themselves test the storage type and call the 887 * right one. Rather than require callers to be aware of those layout 888 * decisions, we centralize here. 889 */ 890 static __inline void 891 m_align(struct mbuf *m, int len) 892 { 893 #ifdef INVARIANTS 894 const char *msg = "%s: not a virgin mbuf"; 895 #endif 896 int adjust; 897 898 KASSERT(m->m_data == M_START(m), (msg, __func__)); 899 900 adjust = M_SIZE(m) - len; 901 m->m_data += adjust &~ (sizeof(long)-1); 902 } 903 904 #define M_ALIGN(m, len) m_align(m, len) 905 #define MH_ALIGN(m, len) m_align(m, len) 906 #define MEXT_ALIGN(m, len) m_align(m, len) 907 908 /* 909 * Compute the amount of space available before the current start of data in 910 * an mbuf. 911 * 912 * The M_WRITABLE() is a temporary, conservative safety measure: the burden 913 * of checking writability of the mbuf data area rests solely with the caller. 914 * 915 * NB: In previous versions, M_LEADINGSPACE() would only check M_WRITABLE() 916 * for mbufs with external storage. We now allow mbuf-embedded data to be 917 * read-only as well. 918 */ 919 #define M_LEADINGSPACE(m) \ 920 (M_WRITABLE(m) ? ((m)->m_data - M_START(m)) : 0) 921 922 /* 923 * Compute the amount of space available after the end of data in an mbuf. 924 * 925 * The M_WRITABLE() is a temporary, conservative safety measure: the burden 926 * of checking writability of the mbuf data area rests solely with the caller. 927 * 928 * NB: In previous versions, M_TRAILINGSPACE() would only check M_WRITABLE() 929 * for mbufs with external storage. We now allow mbuf-embedded data to be 930 * read-only as well. 931 */ 932 #define M_TRAILINGSPACE(m) \ 933 (M_WRITABLE(m) ? \ 934 ((M_START(m) + M_SIZE(m)) - ((m)->m_data + (m)->m_len)) : 0) 935 936 /* 937 * Arrange to prepend space of size plen to mbuf m. If a new mbuf must be 938 * allocated, how specifies whether to wait. If the allocation fails, the 939 * original mbuf chain is freed and m is set to NULL. 940 */ 941 #define M_PREPEND(m, plen, how) do { \ 942 struct mbuf **_mmp = &(m); \ 943 struct mbuf *_mm = *_mmp; \ 944 int _mplen = (plen); \ 945 int __mhow = (how); \ 946 \ 947 MBUF_CHECKSLEEP(how); \ 948 if (M_LEADINGSPACE(_mm) >= _mplen) { \ 949 _mm->m_data -= _mplen; \ 950 _mm->m_len += _mplen; \ 951 } else \ 952 _mm = m_prepend(_mm, _mplen, __mhow); \ 953 if (_mm != NULL && _mm->m_flags & M_PKTHDR) \ 954 _mm->m_pkthdr.len += _mplen; \ 955 *_mmp = _mm; \ 956 } while (0) 957 958 /* 959 * Change mbuf to new type. This is a relatively expensive operation and 960 * should be avoided. 961 */ 962 #define MCHTYPE(m, t) m_chtype((m), (t)) 963 964 /* Length to m_copy to copy all. */ 965 #define M_COPYALL 1000000000 966 967 /* Compatibility with 4.3. */ 968 #define m_copy(m, o, l) m_copym((m), (o), (l), M_NOWAIT) 969 970 extern int max_datalen; /* MHLEN - max_hdr */ 971 extern int max_hdr; /* Largest link + protocol header */ 972 extern int max_linkhdr; /* Largest link-level header */ 973 extern int max_protohdr; /* Largest protocol header */ 974 extern int nmbclusters; /* Maximum number of clusters */ 975 976 /*- 977 * Network packets may have annotations attached by affixing a list of 978 * "packet tags" to the pkthdr structure. Packet tags are dynamically 979 * allocated semi-opaque data structures that have a fixed header 980 * (struct m_tag) that specifies the size of the memory block and a 981 * <cookie,type> pair that identifies it. The cookie is a 32-bit unique 982 * unsigned value used to identify a module or ABI. By convention this value 983 * is chosen as the date+time that the module is created, expressed as the 984 * number of seconds since the epoch (e.g., using date -u +'%s'). The type 985 * value is an ABI/module-specific value that identifies a particular 986 * annotation and is private to the module. For compatibility with systems 987 * like OpenBSD that define packet tags w/o an ABI/module cookie, the value 988 * PACKET_ABI_COMPAT is used to implement m_tag_get and m_tag_find 989 * compatibility shim functions and several tag types are defined below. 990 * Users that do not require compatibility should use a private cookie value 991 * so that packet tag-related definitions can be maintained privately. 992 * 993 * Note that the packet tag returned by m_tag_alloc has the default memory 994 * alignment implemented by malloc. To reference private data one can use a 995 * construct like: 996 * 997 * struct m_tag *mtag = m_tag_alloc(...); 998 * struct foo *p = (struct foo *)(mtag+1); 999 * 1000 * if the alignment of struct m_tag is sufficient for referencing members of 1001 * struct foo. Otherwise it is necessary to embed struct m_tag within the 1002 * private data structure to insure proper alignment; e.g., 1003 * 1004 * struct foo { 1005 * struct m_tag tag; 1006 * ... 1007 * }; 1008 * struct foo *p = (struct foo *) m_tag_alloc(...); 1009 * struct m_tag *mtag = &p->tag; 1010 */ 1011 1012 /* 1013 * Persistent tags stay with an mbuf until the mbuf is reclaimed. Otherwise 1014 * tags are expected to ``vanish'' when they pass through a network 1015 * interface. For most interfaces this happens normally as the tags are 1016 * reclaimed when the mbuf is free'd. However in some special cases 1017 * reclaiming must be done manually. An example is packets that pass through 1018 * the loopback interface. Also, one must be careful to do this when 1019 * ``turning around'' packets (e.g., icmp_reflect). 1020 * 1021 * To mark a tag persistent bit-or this flag in when defining the tag id. 1022 * The tag will then be treated as described above. 1023 */ 1024 #define MTAG_PERSISTENT 0x800 1025 1026 #define PACKET_TAG_NONE 0 /* Nadda */ 1027 1028 /* Packet tags for use with PACKET_ABI_COMPAT. */ 1029 #define PACKET_TAG_IPSEC_IN_DONE 1 /* IPsec applied, in */ 1030 #define PACKET_TAG_IPSEC_OUT_DONE 2 /* IPsec applied, out */ 1031 #define PACKET_TAG_IPSEC_IN_CRYPTO_DONE 3 /* NIC IPsec crypto done */ 1032 #define PACKET_TAG_IPSEC_OUT_CRYPTO_NEEDED 4 /* NIC IPsec crypto req'ed */ 1033 #define PACKET_TAG_IPSEC_IN_COULD_DO_CRYPTO 5 /* NIC notifies IPsec */ 1034 #define PACKET_TAG_IPSEC_PENDING_TDB 6 /* Reminder to do IPsec */ 1035 #define PACKET_TAG_BRIDGE 7 /* Bridge processing done */ 1036 #define PACKET_TAG_GIF 8 /* GIF processing done */ 1037 #define PACKET_TAG_GRE 9 /* GRE processing done */ 1038 #define PACKET_TAG_IN_PACKET_CHECKSUM 10 /* NIC checksumming done */ 1039 #define PACKET_TAG_ENCAP 11 /* Encap. processing */ 1040 #define PACKET_TAG_IPSEC_SOCKET 12 /* IPSEC socket ref */ 1041 #define PACKET_TAG_IPSEC_HISTORY 13 /* IPSEC history */ 1042 #define PACKET_TAG_IPV6_INPUT 14 /* IPV6 input processing */ 1043 #define PACKET_TAG_DUMMYNET 15 /* dummynet info */ 1044 #define PACKET_TAG_DIVERT 17 /* divert info */ 1045 #define PACKET_TAG_IPFORWARD 18 /* ipforward info */ 1046 #define PACKET_TAG_MACLABEL (19 | MTAG_PERSISTENT) /* MAC label */ 1047 #define PACKET_TAG_PF (21 | MTAG_PERSISTENT) /* PF/ALTQ information */ 1048 #define PACKET_TAG_RTSOCKFAM 25 /* rtsock sa family */ 1049 #define PACKET_TAG_IPOPTIONS 27 /* Saved IP options */ 1050 #define PACKET_TAG_CARP 28 /* CARP info */ 1051 #define PACKET_TAG_IPSEC_NAT_T_PORTS 29 /* two uint16_t */ 1052 #define PACKET_TAG_ND_OUTGOING 30 /* ND outgoing */ 1053 1054 /* Specific cookies and tags. */ 1055 1056 /* Packet tag routines. */ 1057 struct m_tag *m_tag_alloc(u_int32_t, int, int, int); 1058 void m_tag_delete(struct mbuf *, struct m_tag *); 1059 void m_tag_delete_chain(struct mbuf *, struct m_tag *); 1060 void m_tag_free_default(struct m_tag *); 1061 struct m_tag *m_tag_locate(struct mbuf *, u_int32_t, int, struct m_tag *); 1062 struct m_tag *m_tag_copy(struct m_tag *, int); 1063 int m_tag_copy_chain(struct mbuf *, const struct mbuf *, int); 1064 void m_tag_delete_nonpersistent(struct mbuf *); 1065 1066 /* 1067 * Initialize the list of tags associated with an mbuf. 1068 */ 1069 static __inline void 1070 m_tag_init(struct mbuf *m) 1071 { 1072 1073 SLIST_INIT(&m->m_pkthdr.tags); 1074 } 1075 1076 /* 1077 * Set up the contents of a tag. Note that this does not fill in the free 1078 * method; the caller is expected to do that. 1079 * 1080 * XXX probably should be called m_tag_init, but that was already taken. 1081 */ 1082 static __inline void 1083 m_tag_setup(struct m_tag *t, u_int32_t cookie, int type, int len) 1084 { 1085 1086 t->m_tag_id = type; 1087 t->m_tag_len = len; 1088 t->m_tag_cookie = cookie; 1089 } 1090 1091 /* 1092 * Reclaim resources associated with a tag. 1093 */ 1094 static __inline void 1095 m_tag_free(struct m_tag *t) 1096 { 1097 1098 (*t->m_tag_free)(t); 1099 } 1100 1101 /* 1102 * Return the first tag associated with an mbuf. 1103 */ 1104 static __inline struct m_tag * 1105 m_tag_first(struct mbuf *m) 1106 { 1107 1108 return (SLIST_FIRST(&m->m_pkthdr.tags)); 1109 } 1110 1111 /* 1112 * Return the next tag in the list of tags associated with an mbuf. 1113 */ 1114 static __inline struct m_tag * 1115 m_tag_next(struct mbuf *m __unused, struct m_tag *t) 1116 { 1117 1118 return (SLIST_NEXT(t, m_tag_link)); 1119 } 1120 1121 /* 1122 * Prepend a tag to the list of tags associated with an mbuf. 1123 */ 1124 static __inline void 1125 m_tag_prepend(struct mbuf *m, struct m_tag *t) 1126 { 1127 1128 SLIST_INSERT_HEAD(&m->m_pkthdr.tags, t, m_tag_link); 1129 } 1130 1131 /* 1132 * Unlink a tag from the list of tags associated with an mbuf. 1133 */ 1134 static __inline void 1135 m_tag_unlink(struct mbuf *m, struct m_tag *t) 1136 { 1137 1138 SLIST_REMOVE(&m->m_pkthdr.tags, t, m_tag, m_tag_link); 1139 } 1140 1141 /* These are for OpenBSD compatibility. */ 1142 #define MTAG_ABI_COMPAT 0 /* compatibility ABI */ 1143 1144 static __inline struct m_tag * 1145 m_tag_get(int type, int length, int wait) 1146 { 1147 return (m_tag_alloc(MTAG_ABI_COMPAT, type, length, wait)); 1148 } 1149 1150 static __inline struct m_tag * 1151 m_tag_find(struct mbuf *m, int type, struct m_tag *start) 1152 { 1153 return (SLIST_EMPTY(&m->m_pkthdr.tags) ? (struct m_tag *)NULL : 1154 m_tag_locate(m, MTAG_ABI_COMPAT, type, start)); 1155 } 1156 1157 static __inline struct mbuf * 1158 m_free(struct mbuf *m) 1159 { 1160 struct mbuf *n = m->m_next; 1161 1162 MBUF_PROBE1(m__free, m); 1163 if ((m->m_flags & (M_PKTHDR|M_NOFREE)) == (M_PKTHDR|M_NOFREE)) 1164 m_tag_delete_chain(m, NULL); 1165 if (m->m_flags & M_EXT) 1166 mb_free_ext(m); 1167 else if ((m->m_flags & M_NOFREE) == 0) 1168 uma_zfree(zone_mbuf, m); 1169 return (n); 1170 } 1171 1172 static __inline int 1173 rt_m_getfib(struct mbuf *m) 1174 { 1175 KASSERT(m->m_flags & M_PKTHDR , ("Attempt to get FIB from non header mbuf.")); 1176 return (m->m_pkthdr.fibnum); 1177 } 1178 1179 #define M_GETFIB(_m) rt_m_getfib(_m) 1180 1181 #define M_SETFIB(_m, _fib) do { \ 1182 KASSERT((_m)->m_flags & M_PKTHDR, ("Attempt to set FIB on non header mbuf.")); \ 1183 ((_m)->m_pkthdr.fibnum) = (_fib); \ 1184 } while (0) 1185 1186 /* flags passed as first argument for "m_ether_tcpip_hash()" */ 1187 #define MBUF_HASHFLAG_L2 (1 << 2) 1188 #define MBUF_HASHFLAG_L3 (1 << 3) 1189 #define MBUF_HASHFLAG_L4 (1 << 4) 1190 1191 /* mbuf hashing helper routines */ 1192 uint32_t m_ether_tcpip_hash_init(void); 1193 uint32_t m_ether_tcpip_hash(const uint32_t, const struct mbuf *, const uint32_t); 1194 1195 #ifdef MBUF_PROFILING 1196 void m_profile(struct mbuf *m); 1197 #define M_PROFILE(m) m_profile(m) 1198 #else 1199 #define M_PROFILE(m) 1200 #endif 1201 1202 struct mbufq { 1203 STAILQ_HEAD(, mbuf) mq_head; 1204 int mq_len; 1205 int mq_maxlen; 1206 }; 1207 1208 static inline void 1209 mbufq_init(struct mbufq *mq, int maxlen) 1210 { 1211 1212 STAILQ_INIT(&mq->mq_head); 1213 mq->mq_maxlen = maxlen; 1214 mq->mq_len = 0; 1215 } 1216 1217 static inline struct mbuf * 1218 mbufq_flush(struct mbufq *mq) 1219 { 1220 struct mbuf *m; 1221 1222 m = STAILQ_FIRST(&mq->mq_head); 1223 STAILQ_INIT(&mq->mq_head); 1224 mq->mq_len = 0; 1225 return (m); 1226 } 1227 1228 static inline void 1229 mbufq_drain(struct mbufq *mq) 1230 { 1231 struct mbuf *m, *n; 1232 1233 n = mbufq_flush(mq); 1234 while ((m = n) != NULL) { 1235 n = STAILQ_NEXT(m, m_stailqpkt); 1236 m_freem(m); 1237 } 1238 } 1239 1240 static inline struct mbuf * 1241 mbufq_first(const struct mbufq *mq) 1242 { 1243 1244 return (STAILQ_FIRST(&mq->mq_head)); 1245 } 1246 1247 static inline struct mbuf * 1248 mbufq_last(const struct mbufq *mq) 1249 { 1250 1251 return (STAILQ_LAST(&mq->mq_head, mbuf, m_stailqpkt)); 1252 } 1253 1254 static inline int 1255 mbufq_full(const struct mbufq *mq) 1256 { 1257 1258 return (mq->mq_len >= mq->mq_maxlen); 1259 } 1260 1261 static inline int 1262 mbufq_len(const struct mbufq *mq) 1263 { 1264 1265 return (mq->mq_len); 1266 } 1267 1268 static inline int 1269 mbufq_enqueue(struct mbufq *mq, struct mbuf *m) 1270 { 1271 1272 if (mbufq_full(mq)) 1273 return (ENOBUFS); 1274 STAILQ_INSERT_TAIL(&mq->mq_head, m, m_stailqpkt); 1275 mq->mq_len++; 1276 return (0); 1277 } 1278 1279 static inline struct mbuf * 1280 mbufq_dequeue(struct mbufq *mq) 1281 { 1282 struct mbuf *m; 1283 1284 m = STAILQ_FIRST(&mq->mq_head); 1285 if (m) { 1286 STAILQ_REMOVE_HEAD(&mq->mq_head, m_stailqpkt); 1287 m->m_nextpkt = NULL; 1288 mq->mq_len--; 1289 } 1290 return (m); 1291 } 1292 1293 static inline void 1294 mbufq_prepend(struct mbufq *mq, struct mbuf *m) 1295 { 1296 1297 STAILQ_INSERT_HEAD(&mq->mq_head, m, m_stailqpkt); 1298 mq->mq_len++; 1299 } 1300 #endif /* _KERNEL */ 1301 #endif /* !_SYS_MBUF_H_ */ 1302