mbuf.h revision 98612
1/*- 2 * Copyright (c) 1982, 1986, 1988, 1993 3 * The Regents of the University of California. All rights reserved. 4 * 5 * Redistribution and use in source and binary forms, with or without 6 * modification, are permitted provided that the following conditions 7 * are met: 8 * 1. Redistributions of source code must retain the above copyright 9 * notice, this list of conditions and the following disclaimer. 10 * 2. Redistributions in binary form must reproduce the above copyright 11 * notice, this list of conditions and the following disclaimer in the 12 * documentation and/or other materials provided with the distribution. 13 * 3. All advertising materials mentioning features or use of this software 14 * must display the following acknowledgement: 15 * This product includes software developed by the University of 16 * California, Berkeley and its contributors. 17 * 4. Neither the name of the University nor the names of its contributors 18 * may be used to endorse or promote products derived from this software 19 * without specific prior written permission. 20 * 21 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 22 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 23 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 24 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 25 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 26 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 27 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 28 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 29 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 30 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 31 * SUCH DAMAGE. 32 * 33 * @(#)mbuf.h 8.5 (Berkeley) 2/19/95 34 * $FreeBSD: head/sys/sys/mbuf.h 98612 2002-06-22 11:29:08Z luigi $ 35 */ 36 37#ifndef _SYS_MBUF_H_ 38#define _SYS_MBUF_H_ 39 40/* 41 * Mbufs are of a single size, MSIZE (machine/param.h), which 42 * includes overhead. An mbuf may add a single "mbuf cluster" of size 43 * MCLBYTES (also in machine/param.h), which has no additional overhead 44 * and is used instead of the internal data area; this is done when 45 * at least MINCLSIZE of data must be stored. Additionally, it is possible 46 * to allocate a separate buffer externally and attach it to the mbuf in 47 * a way similar to that of mbuf clusters. 48 */ 49#define MLEN (MSIZE - sizeof(struct m_hdr)) /* normal data len */ 50#define MHLEN (MLEN - sizeof(struct pkthdr)) /* data len w/pkthdr */ 51#define MINCLSIZE (MHLEN + 1) /* smallest amount to put in cluster */ 52#define M_MAXCOMPRESS (MHLEN / 2) /* max amount to copy for compression */ 53 54#ifdef _KERNEL 55/*- 56 * Macros for type conversion: 57 * mtod(m, t) -- Convert mbuf pointer to data pointer of correct type. 58 * dtom(x) -- Convert data pointer within mbuf to mbuf pointer (XXX). 59 */ 60#define mtod(m, t) ((t)((m)->m_data)) 61#define dtom(x) ((struct mbuf *)((intptr_t)(x) & ~(MSIZE-1))) 62#endif /* _KERNEL */ 63 64/* 65 * Header present at the beginning of every mbuf. 66 */ 67struct m_hdr { 68 struct mbuf *mh_next; /* next buffer in chain */ 69 struct mbuf *mh_nextpkt; /* next chain in queue/record */ 70 caddr_t mh_data; /* location of data */ 71 int mh_len; /* amount of data in this mbuf */ 72 short mh_type; /* type of data in this mbuf */ 73 short mh_flags; /* flags; see below */ 74}; 75 76/* 77 * Record/packet header in first mbuf of chain; valid only if M_PKTHDR is set. 78 */ 79struct pkthdr { 80 struct ifnet *rcvif; /* rcv interface */ 81 int len; /* total packet length */ 82 /* variables for ip and tcp reassembly */ 83 void *header; /* pointer to packet header */ 84 /* variables for hardware checksum */ 85 int csum_flags; /* flags regarding checksum */ 86 int csum_data; /* data field used by csum routines */ 87 struct mbuf *aux; /* extra data buffer; ipsec/others */ 88}; 89 90/* 91 * Description of external storage mapped into mbuf; valid only if M_EXT is set. 92 */ 93struct m_ext { 94 caddr_t ext_buf; /* start of buffer */ 95 void (*ext_free) /* free routine if not the usual */ 96 (caddr_t, void *); 97 void *ext_args; /* optional argument pointer */ 98 u_int ext_size; /* size of buffer, for ext_free */ 99 u_int *ref_cnt; /* pointer to ref count info */ 100 int ext_type; /* type of external storage */ 101}; 102 103/* 104 * The core of the mbuf object along with some shortcut defines for 105 * practical purposes. 106 */ 107struct mbuf { 108 struct m_hdr m_hdr; 109 union { 110 struct { 111 struct pkthdr MH_pkthdr; /* M_PKTHDR set */ 112 union { 113 struct m_ext MH_ext; /* M_EXT set */ 114 char MH_databuf[MHLEN]; 115 } MH_dat; 116 } MH; 117 char M_databuf[MLEN]; /* !M_PKTHDR, !M_EXT */ 118 } M_dat; 119}; 120#define m_next m_hdr.mh_next 121#define m_len m_hdr.mh_len 122#define m_data m_hdr.mh_data 123#define m_type m_hdr.mh_type 124#define m_flags m_hdr.mh_flags 125#define m_nextpkt m_hdr.mh_nextpkt 126#define m_act m_nextpkt 127#define m_pkthdr M_dat.MH.MH_pkthdr 128#define m_ext M_dat.MH.MH_dat.MH_ext 129#define m_pktdat M_dat.MH.MH_dat.MH_databuf 130#define m_dat M_dat.M_databuf 131 132/* 133 * mbuf flags. 134 */ 135#define M_EXT 0x0001 /* has associated external storage */ 136#define M_PKTHDR 0x0002 /* start of record */ 137#define M_EOR 0x0004 /* end of record */ 138#define M_RDONLY 0x0008 /* associated data is marked read-only */ 139#define M_PROTO1 0x0010 /* protocol-specific */ 140#define M_PROTO2 0x0020 /* protocol-specific */ 141#define M_PROTO3 0x0040 /* protocol-specific */ 142#define M_PROTO4 0x0080 /* protocol-specific */ 143#define M_PROTO5 0x0100 /* protocol-specific */ 144 145/* 146 * mbuf pkthdr flags (also stored in m_flags). 147 */ 148#define M_BCAST 0x0200 /* send/received as link-level broadcast */ 149#define M_MCAST 0x0400 /* send/received as link-level multicast */ 150#define M_FRAG 0x0800 /* packet is a fragment of a larger packet */ 151#define M_FIRSTFRAG 0x1000 /* packet is first fragment */ 152#define M_LASTFRAG 0x2000 /* packet is last fragment */ 153 154/* 155 * External buffer types: identify ext_buf type. 156 */ 157#define EXT_CLUSTER 1 /* mbuf cluster */ 158#define EXT_SFBUF 2 /* sendfile(2)'s sf_bufs */ 159#define EXT_NET_DRV 100 /* custom ext_buf provided by net driver(s) */ 160#define EXT_MOD_TYPE 200 /* custom module's ext_buf type */ 161 162/* 163 * Flags copied when copying m_pkthdr. 164 */ 165#define M_COPYFLAGS (M_PKTHDR|M_EOR|M_PROTO1|M_PROTO1|M_PROTO2|M_PROTO3 | \ 166 M_PROTO4|M_PROTO5|M_BCAST|M_MCAST|M_FRAG|M_RDONLY) 167 168/* 169 * Flags indicating hw checksum support and sw checksum requirements. 170 */ 171#define CSUM_IP 0x0001 /* will csum IP */ 172#define CSUM_TCP 0x0002 /* will csum TCP */ 173#define CSUM_UDP 0x0004 /* will csum UDP */ 174#define CSUM_IP_FRAGS 0x0008 /* will csum IP fragments */ 175#define CSUM_FRAGMENT 0x0010 /* will do IP fragmentation */ 176 177#define CSUM_IP_CHECKED 0x0100 /* did csum IP */ 178#define CSUM_IP_VALID 0x0200 /* ... the csum is valid */ 179#define CSUM_DATA_VALID 0x0400 /* csum_data field is valid */ 180#define CSUM_PSEUDO_HDR 0x0800 /* csum_data has pseudo hdr */ 181 182#define CSUM_DELAY_DATA (CSUM_TCP | CSUM_UDP) 183#define CSUM_DELAY_IP (CSUM_IP) /* XXX add ipv6 here too? */ 184 185/* 186 * mbuf types. 187 */ 188#define MT_NOTMBUF 0 /* USED INTERNALLY ONLY! Object is not mbuf */ 189#define MT_DATA 1 /* dynamic (data) allocation */ 190#define MT_HEADER 2 /* packet header */ 191#if 0 192#define MT_SOCKET 3 /* socket structure */ 193#define MT_PCB 4 /* protocol control block */ 194#define MT_RTABLE 5 /* routing tables */ 195#define MT_HTABLE 6 /* IMP host tables */ 196#define MT_ATABLE 7 /* address resolution tables */ 197#endif 198#define MT_SONAME 8 /* socket name */ 199#if 0 200#define MT_SOOPTS 10 /* socket options */ 201#endif 202#define MT_FTABLE 11 /* fragment reassembly header */ 203#if 0 204#define MT_RIGHTS 12 /* access rights */ 205#define MT_IFADDR 13 /* interface address */ 206#endif 207#define MT_TAG 13 /* volatile metadata associated to pkts */ 208#define MT_CONTROL 14 /* extra-data protocol message */ 209#define MT_OOBDATA 15 /* expedited data */ 210#define MT_NTYPES 16 /* number of mbuf types for mbtypes[] */ 211 212/* 213 * Mbuf and cluster allocation statistics PCPU structure. 214 */ 215struct mbpstat { 216 u_long mb_mbfree; 217 u_long mb_mbpgs; 218 u_long mb_clfree; 219 u_long mb_clpgs; 220 long mb_mbtypes[MT_NTYPES]; 221 short mb_active; 222}; 223 224/* 225 * General mbuf allocator statistics structure. 226 * XXX: Modifications of these are not protected by any mutex locks nor by 227 * any atomic() manipulations. As a result, we may occasionally lose 228 * a count or two. Luckily, not all of these fields are modified at all 229 * and remain static, and those that are manipulated are only manipulated 230 * in failure situations, which do not occur (hopefully) very often. 231 */ 232struct mbstat { 233 u_long m_drops; /* times failed to allocate */ 234 u_long m_wait; /* times succesfully returned from wait */ 235 u_long m_drain; /* times drained protocols for space */ 236 u_long m_mcfail; /* XXX: times m_copym failed */ 237 u_long m_mpfail; /* XXX: times m_pullup failed */ 238 u_long m_msize; /* length of an mbuf */ 239 u_long m_mclbytes; /* length of an mbuf cluster */ 240 u_long m_minclsize; /* min length of data to allocate a cluster */ 241 u_long m_mlen; /* length of data in an mbuf */ 242 u_long m_mhlen; /* length of data in a header mbuf */ 243 /* Number of mbtypes (gives # elems in mbpstat's mb_mbtypes[] array: */ 244 short m_numtypes; 245}; 246 247/* 248 * Flags specifying how an allocation should be made. 249 * M_DONTWAIT means "don't block if nothing is available" whereas 250 * M_TRYWAIT means "block for mbuf_wait ticks at most if nothing is 251 * available." 252 */ 253#define M_DONTWAIT 1 254#define M_TRYWAIT 0 255#define M_WAIT M_TRYWAIT /* XXX: Deprecated. */ 256 257#ifdef _KERNEL 258/*- 259 * mbuf external reference count management macros. 260 * 261 * MEXT_IS_REF(m): true if (m) is not the only mbuf referencing 262 * the external buffer ext_buf. 263 * 264 * MEXT_REM_REF(m): remove reference to m_ext object. 265 * 266 * MEXT_ADD_REF(m): add reference to m_ext object already 267 * referred to by (m). 268 */ 269#define MEXT_IS_REF(m) (*((m)->m_ext.ref_cnt) > 1) 270 271#define MEXT_REM_REF(m) do { \ 272 KASSERT(*((m)->m_ext.ref_cnt) > 0, ("m_ext refcnt < 0")); \ 273 atomic_subtract_int((m)->m_ext.ref_cnt, 1); \ 274} while(0) 275 276#define MEXT_ADD_REF(m) atomic_add_int((m)->m_ext.ref_cnt, 1) 277 278/* 279 * mbuf, cluster, and external object allocation macros 280 * (for compatibility purposes). 281 */ 282#define m_getclr m_get_clrd 283#define MGET(m, how, type) (m) = m_get((how), (type)) 284#define MGETHDR(m, how, type) (m) = m_gethdr((how), (type)) 285#define MCLGET(m, how) m_clget((m), (how)) 286#define MEXTADD(m, buf, size, free, args, flags, type) \ 287 m_extadd((m), (caddr_t)(buf), (size), (free), (args), (flags), (type)) 288 289/* 290 * MEXTFREE(m): disassociate (and possibly free) an external object from (m). 291 * 292 * If the atomic_cmpset_int() returns 0, then we effectively do nothing 293 * in terms of "cleaning up" (freeing the ext buf and ref. counter) as 294 * this means that either there are still references, or another thread 295 * is taking care of the clean-up. 296 */ 297#define MEXTFREE(m) do { \ 298 struct mbuf *_mb = (m); \ 299 \ 300 MEXT_REM_REF(_mb); \ 301 if (atomic_cmpset_int(_mb->m_ext.ref_cnt, 0, 1)) \ 302 _mext_free(_mb); \ 303 _mb->m_flags &= ~M_EXT; \ 304} while (0) 305 306/* 307 * Evaluate TRUE if it's safe to write to the mbuf m's data region (this 308 * can be both the local data payload, or an external buffer area, 309 * depending on whether M_EXT is set). 310 */ 311#define M_WRITABLE(m) (!((m)->m_flags & M_RDONLY) && (!((m)->m_flags \ 312 & M_EXT) || !MEXT_IS_REF(m))) 313 314/*- 315 * Copy mbuf pkthdr from "from" to "to". 316 * "from" must have M_PKTHDR set, and "to" must be empty. 317 * aux pointer will be moved to "to". 318 */ 319#define M_COPY_PKTHDR(to, from) do { \ 320 struct mbuf *_mfrom = (from); \ 321 struct mbuf *_mto = (to); \ 322 \ 323 _mto->m_data = _mto->m_pktdat; \ 324 _mto->m_flags = _mfrom->m_flags & M_COPYFLAGS; \ 325 _mto->m_pkthdr = _mfrom->m_pkthdr; \ 326 _mfrom->m_pkthdr.aux = NULL; \ 327} while (0) 328 329/* 330 * Set the m_data pointer of a newly-allocated mbuf (m_get/MGET) to place 331 * an object of the specified size at the end of the mbuf, longword aligned. 332 */ 333#define M_ALIGN(m, len) do { \ 334 (m)->m_data += (MLEN - (len)) & ~(sizeof(long) - 1); \ 335} while (0) 336 337/* 338 * As above, for mbufs allocated with m_gethdr/MGETHDR 339 * or initialized by M_COPY_PKTHDR. 340 */ 341#define MH_ALIGN(m, len) do { \ 342 (m)->m_data += (MHLEN - (len)) & ~(sizeof(long) - 1); \ 343} while (0) 344 345/* 346 * Compute the amount of space available 347 * before the current start of data in an mbuf. 348 * 349 * The M_WRITABLE() is a temporary, conservative safety measure: the burden 350 * of checking writability of the mbuf data area rests solely with the caller. 351 */ 352#define M_LEADINGSPACE(m) \ 353 ((m)->m_flags & M_EXT ? \ 354 (M_WRITABLE(m) ? (m)->m_data - (m)->m_ext.ext_buf : 0): \ 355 (m)->m_flags & M_PKTHDR ? (m)->m_data - (m)->m_pktdat : \ 356 (m)->m_data - (m)->m_dat) 357 358/* 359 * Compute the amount of space available 360 * after the end of data in an mbuf. 361 * 362 * The M_WRITABLE() is a temporary, conservative safety measure: the burden 363 * of checking writability of the mbuf data area rests solely with the caller. 364 */ 365#define M_TRAILINGSPACE(m) \ 366 ((m)->m_flags & M_EXT ? \ 367 (M_WRITABLE(m) ? (m)->m_ext.ext_buf + (m)->m_ext.ext_size \ 368 - ((m)->m_data + (m)->m_len) : 0) : \ 369 &(m)->m_dat[MLEN] - ((m)->m_data + (m)->m_len)) 370 371/* 372 * Arrange to prepend space of size plen to mbuf m. 373 * If a new mbuf must be allocated, how specifies whether to wait. 374 * If the allocation fails, the original mbuf chain is freed and m is 375 * set to NULL. 376 */ 377#define M_PREPEND(m, plen, how) do { \ 378 struct mbuf **_mmp = &(m); \ 379 struct mbuf *_mm = *_mmp; \ 380 int _mplen = (plen); \ 381 int __mhow = (how); \ 382 \ 383 if (M_LEADINGSPACE(_mm) >= _mplen) { \ 384 _mm->m_data -= _mplen; \ 385 _mm->m_len += _mplen; \ 386 } else \ 387 _mm = m_prepend(_mm, _mplen, __mhow); \ 388 if (_mm != NULL && _mm->m_flags & M_PKTHDR) \ 389 _mm->m_pkthdr.len += _mplen; \ 390 *_mmp = _mm; \ 391} while (0) 392 393/* 394 * Change mbuf to new type. 395 * This is a relatively expensive operation and should be avoided. 396 */ 397#define MCHTYPE(m, t) m_chtype((m), (t)) 398 399/* Length to m_copy to copy all. */ 400#define M_COPYALL 1000000000 401 402/* Compatibility with 4.3 */ 403#define m_copy(m, o, l) m_copym((m), (o), (l), M_DONTWAIT) 404 405/* 406 * pkthdr.aux type tags. 407 */ 408struct mauxtag { 409 int af; 410 int type; 411 void *p; 412}; 413 414/* 415 * Some packet tags to identify different mbuf annotations. 416 * 417 * Eventually, these annotations will end up in an appropriate chain 418 * (struct m_tag or similar, e.g. as in NetBSD) properly managed by 419 * the mbuf handling routines. 420 * 421 * As a temporary and low impact solution to replace the even uglier 422 * approach used so far in some parts of the network stack (which relies 423 * on global variables), these annotations are stored in MT_TAG 424 * mbufs (or lookalikes) prepended to the actual mbuf chain. 425 * 426 * m_type = MT_TAG 427 * m_flags = m_tag_id 428 * m_next = next buffer in chain. 429 * 430 * BE VERY CAREFUL not to pass these blocks to the mbuf handling routines. 431 * 432 */ 433 434#define m_tag_id m_hdr.mh_flags 435 436/* Packet tag types -- first ones are from NetBSD */ 437 438#define PACKET_TAG_NONE 0 /* Nadda */ 439#define PACKET_TAG_IPSEC_IN_DONE 1 /* IPsec applied, in */ 440#define PACKET_TAG_IPSEC_OUT_DONE 2 /* IPsec applied, out */ 441#define PACKET_TAG_IPSEC_IN_CRYPTO_DONE 3 /* NIC IPsec crypto done */ 442#define PACKET_TAG_IPSEC_OUT_CRYPTO_NEEDED 4 /* NIC IPsec crypto req'ed */ 443#define PACKET_TAG_IPSEC_IN_COULD_DO_CRYPTO 5 /* NIC notifies IPsec */ 444#define PACKET_TAG_IPSEC_PENDING_TDB 6 /* Reminder to do IPsec */ 445#define PACKET_TAG_BRIDGE 7 /* Bridge processing done */ 446#define PACKET_TAG_GIF 8 /* GIF processing done */ 447#define PACKET_TAG_GRE 9 /* GRE processing done */ 448#define PACKET_TAG_IN_PACKET_CHECKSUM 10 /* NIC checksumming done */ 449#define PACKET_TAG_ENCAP 11 /* Encap. processing */ 450#define PACKET_TAG_IPSEC_SOCKET 12 /* IPSEC socket ref */ 451#define PACKET_TAG_IPSEC_HISTORY 13 /* IPSEC history */ 452#define PACKET_TAG_IPV6_INPUT 14 /* IPV6 input processing */ 453 454/* Packet tags used in the FreeBSD network stack */ 455#define PACKET_TAG_DUMMYNET 15 /* dummynet info */ 456#define PACKET_TAG_IPFW 16 /* ipfw classification */ 457#define PACKET_TAG_DIVERT 17 /* divert info */ 458#define PACKET_TAG_IPFORWARD 18 /* ipforward info */ 459 460#define PACKET_TAG_MAX 19 461 462extern int max_datalen; /* MHLEN - max_hdr */ 463extern int max_hdr; /* largest link + protocol header */ 464extern int max_linkhdr; /* largest link-level header */ 465extern int max_protohdr; /* largest protocol header */ 466extern struct mbpstat mb_statpcpu[]; /* Per-CPU allocation stats. */ 467extern struct mbstat mbstat; /* General mbuf stats/infos. */ 468extern int nmbclusters; /* Maximum number of clusters */ 469extern int nmbcnt; /* Scale kmem_map for counter space */ 470extern int nmbufs; /* Maximum number of mbufs */ 471extern int nsfbufs; /* Number of sendfile(2) bufs */ 472 473void _mext_free(struct mbuf *); 474void m_adj(struct mbuf *, int); 475struct mbuf *m_aux_add(struct mbuf *, int, int); 476struct mbuf *m_aux_add2(struct mbuf *, int, int, void *); 477void m_aux_delete(struct mbuf *, struct mbuf *); 478struct mbuf *m_aux_find(struct mbuf *, int, int); 479struct mbuf *m_aux_find2(struct mbuf *, int, int, void *); 480void m_cat(struct mbuf *, struct mbuf *); 481void m_chtype(struct mbuf *, short); 482void m_clget(struct mbuf *, int); 483void m_extadd(struct mbuf *, caddr_t, u_int, 484 void (*free)(caddr_t, void *), void *, short, int); 485void m_copyback(struct mbuf *, int, int, caddr_t); 486void m_copydata(const struct mbuf *, int, int, caddr_t); 487struct mbuf *m_copym(struct mbuf *, int, int, int); 488struct mbuf *m_copypacket(struct mbuf *, int); 489struct mbuf *m_devget(char *, int, int, struct ifnet *, 490 void (*copy)(char *, caddr_t, u_int)); 491struct mbuf *m_dup(struct mbuf *, int); 492struct mbuf *m_free(struct mbuf *); 493void m_freem(struct mbuf *); 494struct mbuf *m_get(int, int); 495struct mbuf *m_get_clrd(int, int); 496struct mbuf *m_gethdr(int, int); 497struct mbuf *m_gethdr_clrd(int, int); 498struct mbuf *m_getm(struct mbuf *, int, int, int); 499struct mbuf *m_prepend(struct mbuf *, int, int); 500void m_print(const struct mbuf *m); 501struct mbuf *m_pulldown(struct mbuf *, int, int, int *); 502struct mbuf *m_pullup(struct mbuf *, int); 503struct mbuf *m_split(struct mbuf *, int, int); 504#endif /* _KERNEL */ 505 506#endif /* !_SYS_MBUF_H_ */ 507