FreeBSD/Linux Kernel Cross Reference
sys/sys/mbuf.h
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. Neither the name of the University nor the names of its contributors
14 * may be used to endorse or promote products derived from this software
15 * without specific prior written permission.
16 *
17 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
18 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
19 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
20 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
21 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
22 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
23 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
24 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
25 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
26 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
27 * SUCH DAMAGE.
28 *
29 * @(#)mbuf.h 8.5 (Berkeley) 2/19/95
30 * $FreeBSD: releng/6.1/sys/sys/mbuf.h 158179 2006-04-30 16:44:43Z cvs2svn $
31 */
32
33 #ifndef _SYS_MBUF_H_
34 #define _SYS_MBUF_H_
35
36 /* XXX: These includes suck. Sorry! */
37 #include <sys/queue.h>
38 #ifdef _KERNEL
39 #include <sys/systm.h>
40 #include <vm/uma.h>
41 #ifdef WITNESS
42 #include <sys/lock.h>
43 #endif
44 #endif
45
46 /*
47 * Mbufs are of a single size, MSIZE (sys/param.h), which
48 * includes overhead. An mbuf may add a single "mbuf cluster" of size
49 * MCLBYTES (also in sys/param.h), which has no additional overhead
50 * and is used instead of the internal data area; this is done when
51 * at least MINCLSIZE of data must be stored. Additionally, it is possible
52 * to allocate a separate buffer externally and attach it to the mbuf in
53 * a way similar to that of mbuf clusters.
54 */
55 #define MLEN (MSIZE - sizeof(struct m_hdr)) /* normal data len */
56 #define MHLEN (MLEN - sizeof(struct pkthdr)) /* data len w/pkthdr */
57 #define MINCLSIZE (MHLEN + 1) /* smallest amount to put in cluster */
58 #define M_MAXCOMPRESS (MHLEN / 2) /* max amount to copy for compression */
59
60 #ifdef _KERNEL
61 /*-
62 * Macros for type conversion:
63 * mtod(m, t) -- Convert mbuf pointer to data pointer of correct type.
64 * dtom(x) -- Convert data pointer within mbuf to mbuf pointer (XXX).
65 */
66 #define mtod(m, t) ((t)((m)->m_data))
67 #define dtom(x) ((struct mbuf *)((intptr_t)(x) & ~(MSIZE-1)))
68
69 /*
70 * Argument structure passed to UMA routines during mbuf and packet
71 * allocations.
72 */
73 struct mb_args {
74 int flags; /* Flags for mbuf being allocated */
75 short type; /* Type of mbuf being allocated */
76 };
77 #endif /* _KERNEL */
78
79 /*
80 * Header present at the beginning of every mbuf.
81 */
82 struct m_hdr {
83 struct mbuf *mh_next; /* next buffer in chain */
84 struct mbuf *mh_nextpkt; /* next chain in queue/record */
85 caddr_t mh_data; /* location of data */
86 int mh_len; /* amount of data in this mbuf */
87 int mh_flags; /* flags; see below */
88 short mh_type; /* type of data in this mbuf */
89 };
90
91 /*
92 * Packet tag structure (see below for details).
93 */
94 struct m_tag {
95 SLIST_ENTRY(m_tag) m_tag_link; /* List of packet tags */
96 u_int16_t m_tag_id; /* Tag ID */
97 u_int16_t m_tag_len; /* Length of data */
98 u_int32_t m_tag_cookie; /* ABI/Module ID */
99 void (*m_tag_free)(struct m_tag *);
100 };
101
102 /*
103 * Record/packet header in first mbuf of chain; valid only if M_PKTHDR is set.
104 */
105 struct pkthdr {
106 struct ifnet *rcvif; /* rcv interface */
107 int len; /* total packet length */
108 /* variables for ip and tcp reassembly */
109 void *header; /* pointer to packet header */
110 /* variables for hardware checksum */
111 int csum_flags; /* flags regarding checksum */
112 int csum_data; /* data field used by csum routines */
113 SLIST_HEAD(packet_tags, m_tag) tags; /* list of packet tags */
114 };
115
116 /*
117 * Description of external storage mapped into mbuf; valid only if M_EXT is set.
118 */
119 struct m_ext {
120 caddr_t ext_buf; /* start of buffer */
121 void (*ext_free) /* free routine if not the usual */
122 (void *, void *);
123 void *ext_args; /* optional argument pointer */
124 u_int ext_size; /* size of buffer, for ext_free */
125 volatile u_int *ref_cnt; /* pointer to ref count info */
126 int ext_type; /* type of external storage */
127 };
128
129 /*
130 * The core of the mbuf object along with some shortcut defines for
131 * practical purposes.
132 */
133 struct mbuf {
134 struct m_hdr m_hdr;
135 union {
136 struct {
137 struct pkthdr MH_pkthdr; /* M_PKTHDR set */
138 union {
139 struct m_ext MH_ext; /* M_EXT set */
140 char MH_databuf[MHLEN];
141 } MH_dat;
142 } MH;
143 char M_databuf[MLEN]; /* !M_PKTHDR, !M_EXT */
144 } M_dat;
145 };
146 #define m_next m_hdr.mh_next
147 #define m_len m_hdr.mh_len
148 #define m_data m_hdr.mh_data
149 #define m_type m_hdr.mh_type
150 #define m_flags m_hdr.mh_flags
151 #define m_nextpkt m_hdr.mh_nextpkt
152 #define m_act m_nextpkt
153 #define m_pkthdr M_dat.MH.MH_pkthdr
154 #define m_ext M_dat.MH.MH_dat.MH_ext
155 #define m_pktdat M_dat.MH.MH_dat.MH_databuf
156 #define m_dat M_dat.M_databuf
157
158 /*
159 * mbuf flags.
160 */
161 #define M_EXT 0x0001 /* has associated external storage */
162 #define M_PKTHDR 0x0002 /* start of record */
163 #define M_EOR 0x0004 /* end of record */
164 #define M_RDONLY 0x0008 /* associated data is marked read-only */
165 #define M_PROTO1 0x0010 /* protocol-specific */
166 #define M_PROTO2 0x0020 /* protocol-specific */
167 #define M_PROTO3 0x0040 /* protocol-specific */
168 #define M_PROTO4 0x0080 /* protocol-specific */
169 #define M_PROTO5 0x0100 /* protocol-specific */
170 #define M_SKIP_FIREWALL 0x4000 /* skip firewall processing */
171 #define M_FREELIST 0x8000 /* mbuf is on the free list */
172
173 /*
174 * mbuf pkthdr flags (also stored in m_flags).
175 */
176 #define M_BCAST 0x0200 /* send/received as link-level broadcast */
177 #define M_MCAST 0x0400 /* send/received as link-level multicast */
178 #define M_FRAG 0x0800 /* packet is a fragment of a larger packet */
179 #define M_FIRSTFRAG 0x1000 /* packet is first fragment */
180 #define M_LASTFRAG 0x2000 /* packet is last fragment */
181 #define M_VLANTAG 0x10000 /* packet has VLAN tag attached */
182
183 /*
184 * External buffer types: identify ext_buf type.
185 */
186 #define EXT_CLUSTER 1 /* mbuf cluster */
187 #define EXT_SFBUF 2 /* sendfile(2)'s sf_bufs */
188 #define EXT_PACKET 3 /* came out of Packet zone */
189 #define EXT_JUMBOP 4 /* jumbo cluster 4096 bytes */
190 #define EXT_JUMBO9 5 /* jumbo cluster 9216 bytes */
191 #define EXT_JUMBO16 6 /* jumbo cluster 16184 bytes */
192 #define EXT_NET_DRV 100 /* custom ext_buf provided by net driver(s) */
193 #define EXT_MOD_TYPE 200 /* custom module's ext_buf type */
194 #define EXT_DISPOSABLE 300 /* can throw this buffer away w/page flipping */
195 #define EXT_EXTREF 400 /* has externally maintained ref_cnt ptr */
196
197 /*
198 * Flags copied when copying m_pkthdr.
199 */
200 #define M_COPYFLAGS (M_PKTHDR|M_EOR|M_RDONLY|M_PROTO1|M_PROTO1|M_PROTO2|\
201 M_PROTO3|M_PROTO4|M_PROTO5|M_SKIP_FIREWALL|\
202 M_BCAST|M_MCAST|M_FRAG|M_FIRSTFRAG|M_LASTFRAG|\
203 M_VLANTAG)
204
205 /*
206 * Flags indicating hw checksum support and sw checksum requirements.
207 */
208 #define CSUM_IP 0x0001 /* will csum IP */
209 #define CSUM_TCP 0x0002 /* will csum TCP */
210 #define CSUM_UDP 0x0004 /* will csum UDP */
211 #define CSUM_IP_FRAGS 0x0008 /* will csum IP fragments */
212 #define CSUM_FRAGMENT 0x0010 /* will do IP fragmentation */
213
214 #define CSUM_IP_CHECKED 0x0100 /* did csum IP */
215 #define CSUM_IP_VALID 0x0200 /* ... the csum is valid */
216 #define CSUM_DATA_VALID 0x0400 /* csum_data field is valid */
217 #define CSUM_PSEUDO_HDR 0x0800 /* csum_data has pseudo hdr */
218
219 #define CSUM_DELAY_DATA (CSUM_TCP | CSUM_UDP)
220 #define CSUM_DELAY_IP (CSUM_IP) /* XXX add ipv6 here too? */
221
222 /*
223 * mbuf types.
224 */
225 #define MT_NOTMBUF 0 /* USED INTERNALLY ONLY! Object is not mbuf */
226 #define MT_DATA 1 /* dynamic (data) allocation */
227 #define MT_HEADER 2 /* packet header */
228 #if 0
229 #define MT_SOCKET 3 /* socket structure */
230 #define MT_PCB 4 /* protocol control block */
231 #define MT_RTABLE 5 /* routing tables */
232 #define MT_HTABLE 6 /* IMP host tables */
233 #define MT_ATABLE 7 /* address resolution tables */
234 #endif
235 #define MT_SONAME 8 /* socket name */
236 #if 0
237 #define MT_SOOPTS 10 /* socket options */
238 #endif
239 #define MT_FTABLE 11 /* fragment reassembly header */
240 #if 0
241 #define MT_RIGHTS 12 /* access rights */
242 #define MT_IFADDR 13 /* interface address */
243 #endif
244 #define MT_CONTROL 14 /* extra-data protocol message */
245 #define MT_OOBDATA 15 /* expedited data */
246 #define MT_NTYPES 16 /* number of mbuf types for mbtypes[] */
247
248 #define MT_NOINIT 255 /* Not a type but a flag to allocate
249 a non-initialized mbuf */
250
251 /*
252 * General mbuf allocator statistics structure.
253 */
254 struct mbstat {
255 u_long m_mbufs; /* XXX */
256 u_long m_mclusts; /* XXX */
257
258 u_long m_drain; /* times drained protocols for space */
259 u_long m_mcfail; /* XXX: times m_copym failed */
260 u_long m_mpfail; /* XXX: times m_pullup failed */
261 u_long m_msize; /* length of an mbuf */
262 u_long m_mclbytes; /* length of an mbuf cluster */
263 u_long m_minclsize; /* min length of data to allocate a cluster */
264 u_long m_mlen; /* length of data in an mbuf */
265 u_long m_mhlen; /* length of data in a header mbuf */
266
267 /* Number of mbtypes (gives # elems in mbtypes[] array: */
268 short m_numtypes;
269
270 /* XXX: Sendfile stats should eventually move to their own struct */
271 u_long sf_iocnt; /* times sendfile had to do disk I/O */
272 u_long sf_allocfail; /* times sfbuf allocation failed */
273 u_long sf_allocwait; /* times sfbuf allocation had to wait */
274 };
275
276 /*
277 * Flags specifying how an allocation should be made.
278 *
279 * The flag to use is as follows:
280 * - M_DONTWAIT or M_NOWAIT from an interrupt handler to not block allocation.
281 * - M_WAIT or M_WAITOK or M_TRYWAIT from wherever it is safe to block.
282 *
283 * M_DONTWAIT/M_NOWAIT means that we will not block the thread explicitly
284 * and if we cannot allocate immediately we may return NULL,
285 * whereas M_WAIT/M_WAITOK/M_TRYWAIT means that if we cannot allocate
286 * resources we will block until they are available, and thus never
287 * return NULL.
288 *
289 * XXX Eventually just phase this out to use M_WAITOK/M_NOWAIT.
290 */
291 #define MBTOM(how) (how)
292 #define M_DONTWAIT M_NOWAIT
293 #define M_TRYWAIT M_WAITOK
294 #define M_WAIT M_WAITOK
295
296 /*
297 * String names of mbuf-related UMA(9) and malloc(9) types. Exposed to
298 * !_KERNEL so that monitoring tools can look up the zones with
299 * libmemstat(3).
300 */
301 #define MBUF_MEM_NAME "mbuf"
302 #define MBUF_CLUSTER_MEM_NAME "mbuf_cluster"
303 #define MBUF_JUMBOP_MEM_NAME "mbuf_jumbo_pagesize"
304 #define MBUF_JUMBO9_MEM_NAME "mbuf_jumbo_9k"
305 #define MBUF_JUMBO16_MEM_NAME "mbuf_jumbo_16k"
306 #define MBUF_PACKET_MEM_NAME "mbuf_packet"
307 #define MBUF_TAG_MEM_NAME "mbuf_tag"
308
309 #ifdef _KERNEL
310 /*-
311 * mbuf external reference count management macros.
312 *
313 * MEXT_IS_REF(m): true if (m) is not the only mbuf referencing
314 * the external buffer ext_buf.
315 *
316 * MEXT_REM_REF(m): remove reference to m_ext object.
317 *
318 * MEXT_ADD_REF(m): add reference to m_ext object already
319 * referred to by (m). XXX Note that it is VERY important that you
320 * always set the second mbuf's m_ext.ref_cnt to point to the first
321 * one's (i.e., n->m_ext.ref_cnt = m->m_ext.ref_cnt) AFTER you run
322 * MEXT_ADD_REF(m). This is because m might have a lazy initialized
323 * ref_cnt (NULL) before this is run and it will only be looked up
324 * from here. We should make MEXT_ADD_REF() always take two mbufs
325 * as arguments so that it can take care of this itself.
326 */
327 #define MEXT_IS_REF(m) (((m)->m_ext.ref_cnt != NULL) \
328 && (*((m)->m_ext.ref_cnt) > 1))
329
330 #define MEXT_REM_REF(m) do { \
331 KASSERT((m)->m_ext.ref_cnt != NULL, ("m_ext refcnt lazy NULL")); \
332 KASSERT(*((m)->m_ext.ref_cnt) > 0, ("m_ext refcnt < 0")); \
333 atomic_subtract_int((m)->m_ext.ref_cnt, 1); \
334 } while(0)
335
336 #define MEXT_ADD_REF(m) do { \
337 if ((m)->m_ext.ref_cnt == NULL) { \
338 KASSERT((m)->m_ext.ext_type == EXT_CLUSTER || \
339 (m)->m_ext.ext_type == EXT_PACKET, \
340 ("Unexpected mbuf type has lazy refcnt")); \
341 (m)->m_ext.ref_cnt = (u_int *)uma_find_refcnt( \
342 zone_clust, (m)->m_ext.ext_buf); \
343 *((m)->m_ext.ref_cnt) = 2; \
344 } else \
345 atomic_add_int((m)->m_ext.ref_cnt, 1); \
346 } while (0)
347
348 #ifdef WITNESS
349 #define MBUF_CHECKSLEEP(how) do { \
350 if (how == M_WAITOK) \
351 WITNESS_WARN(WARN_GIANTOK | WARN_SLEEPOK, NULL, \
352 "Sleeping in \"%s\"", __func__); \
353 } while(0)
354 #else
355 #define MBUF_CHECKSLEEP(how)
356 #endif
357
358 /*
359 * Network buffer allocation API
360 *
361 * The rest of it is defined in kern/kern_mbuf.c
362 */
363
364 extern uma_zone_t zone_mbuf;
365 extern uma_zone_t zone_clust;
366 extern uma_zone_t zone_pack;
367 extern uma_zone_t zone_jumbop;
368 extern uma_zone_t zone_jumbo9;
369 extern uma_zone_t zone_jumbo16;
370
371 static __inline struct mbuf *m_get(int how, short type);
372 static __inline struct mbuf *m_gethdr(int how, short type);
373 static __inline struct mbuf *m_getcl(int how, short type, int flags);
374 static __inline struct mbuf *m_getjcl(int how, short type, int flags, int size);
375 static __inline struct mbuf *m_getclr(int how, short type); /* XXX */
376 static __inline struct mbuf *m_free(struct mbuf *m);
377 static __inline void m_clget(struct mbuf *m, int how);
378 static __inline void *m_cljget(struct mbuf *m, int how, int size);
379 static __inline void m_chtype(struct mbuf *m, short new_type);
380 void mb_free_ext(struct mbuf *);
381
382 static __inline
383 struct mbuf *
384 m_get(int how, short type)
385 {
386 struct mb_args args;
387
388 args.flags = 0;
389 args.type = type;
390 return (uma_zalloc_arg(zone_mbuf, &args, how));
391 }
392
393 /* XXX This should be depracated, very little use */
394 static __inline
395 struct mbuf *
396 m_getclr(int how, short type)
397 {
398 struct mbuf *m;
399 struct mb_args args;
400
401 args.flags = 0;
402 args.type = type;
403 m = uma_zalloc_arg(zone_mbuf, &args, how);
404 if (m != NULL)
405 bzero(m->m_data, MLEN);
406 return m;
407 }
408
409 static __inline
410 struct mbuf *
411 m_gethdr(int how, short type)
412 {
413 struct mb_args args;
414
415 args.flags = M_PKTHDR;
416 args.type = type;
417 return (uma_zalloc_arg(zone_mbuf, &args, how));
418 }
419
420 static __inline
421 struct mbuf *
422 m_getcl(int how, short type, int flags)
423 {
424 struct mb_args args;
425
426 args.flags = flags;
427 args.type = type;
428 return (uma_zalloc_arg(zone_pack, &args, how));
429 }
430
431 /*
432 * m_getjcl() returns an mbuf with a cluster of the specified size attached.
433 * For size it takes MCLBYTES, MJUMPAGESIZE, MJUM9BYTES, MJUM16BYTES.
434 */
435 static __inline /* XXX: This is rather large, should be real function maybe. */
436 struct mbuf *
437 m_getjcl(int how, short type, int flags, int size)
438 {
439 struct mb_args args;
440 struct mbuf *m, *n;
441 uma_zone_t zone;
442
443 args.flags = flags;
444 args.type = type;
445
446 m = uma_zalloc_arg(zone_mbuf, &args, how);
447 if (m == NULL)
448 return NULL;
449
450 switch (size) {
451 case MCLBYTES:
452 zone = zone_clust;
453 break;
454 #if MJUMPAGESIZE != MCLBYTES
455 case MJUMPAGESIZE:
456 zone = zone_jumbop;
457 break;
458 #endif
459 case MJUM9BYTES:
460 zone = zone_jumbo9;
461 break;
462 case MJUM16BYTES:
463 zone = zone_jumbo16;
464 break;
465 default:
466 panic("%s: m_getjcl: invalid cluster type", __func__);
467 }
468 n = uma_zalloc_arg(zone, m, how);
469 if (n == NULL) {
470 uma_zfree(zone_mbuf, m);
471 return NULL;
472 }
473 return m;
474 }
475
476 static __inline
477 struct mbuf *
478 m_free(struct mbuf *m)
479 {
480 struct mbuf *n = m->m_next;
481
482 #ifdef INVARIANTS
483 m->m_flags |= M_FREELIST;
484 #endif
485 if (m->m_flags & M_EXT)
486 mb_free_ext(m);
487 else
488 uma_zfree(zone_mbuf, m);
489 return n;
490 }
491
492 static __inline
493 void
494 m_clget(struct mbuf *m, int how)
495 {
496
497 m->m_ext.ext_buf = NULL;
498 uma_zalloc_arg(zone_clust, m, how);
499 }
500
501 /*
502 * m_cljget() is different from m_clget() as it can allocate clusters
503 * without attaching them to an mbuf. In that case the return value
504 * is the pointer to the cluster of the requested size. If an mbuf was
505 * specified, it gets the cluster attached to it and the return value
506 * can be safely ignored.
507 * For size it takes MCLBYTES, MJUMPAGESIZE, MJUM9BYTES, MJUM16BYTES.
508 */
509 static __inline
510 void *
511 m_cljget(struct mbuf *m, int how, int size)
512 {
513 uma_zone_t zone;
514
515 if (m && m->m_flags & M_EXT)
516 printf("%s: %p mbuf already has cluster\n", __func__, m);
517 if (m != NULL)
518 m->m_ext.ext_buf = NULL;
519
520 switch (size) {
521 case MCLBYTES:
522 zone = zone_clust;
523 break;
524 #if MJUMPAGESIZE != MCLBYTES
525 case MJUMPAGESIZE:
526 zone = zone_jumbop;
527 break;
528 #endif
529 case MJUM9BYTES:
530 zone = zone_jumbo9;
531 break;
532 case MJUM16BYTES:
533 zone = zone_jumbo16;
534 break;
535 default:
536 panic("%s: m_getjcl: invalid cluster type", __func__);
537 }
538
539 return (uma_zalloc_arg(zone, m, how));
540 }
541
542 static __inline
543 void
544 m_chtype(struct mbuf *m, short new_type)
545 {
546 m->m_type = new_type;
547 }
548
549 /*
550 * mbuf, cluster, and external object allocation macros
551 * (for compatibility purposes).
552 */
553 /* NB: M_COPY_PKTHDR is deprecated. Use M_MOVE_PKTHDR or m_dup_pktdr. */
554 #define M_MOVE_PKTHDR(to, from) m_move_pkthdr((to), (from))
555 #define MGET(m, how, type) ((m) = m_get((how), (type)))
556 #define MGETHDR(m, how, type) ((m) = m_gethdr((how), (type)))
557 #define MCLGET(m, how) m_clget((m), (how))
558 #define MEXTADD(m, buf, size, free, args, flags, type) \
559 m_extadd((m), (caddr_t)(buf), (size), (free), (args), (flags), (type))
560
561 /*
562 * Evaluate TRUE if it's safe to write to the mbuf m's data region (this
563 * can be both the local data payload, or an external buffer area,
564 * depending on whether M_EXT is set).
565 */
566 #define M_WRITABLE(m) (!((m)->m_flags & M_RDONLY) && (!((m)->m_flags \
567 & M_EXT) || !MEXT_IS_REF(m)))
568
569 /* Check if the supplied mbuf has a packet header, or else panic. */
570 #define M_ASSERTPKTHDR(m) \
571 KASSERT(m != NULL && m->m_flags & M_PKTHDR, \
572 ("%s: no mbuf packet header!", __func__))
573
574 /* Ensure that the supplied mbuf is a valid, non-free mbuf. */
575 #define M_ASSERTVALID(m) \
576 KASSERT((((struct mbuf *)m)->m_flags & M_FREELIST) == 0, \
577 ("%s: attempted use of a free mbuf!", __func__))
578
579 /*
580 * Set the m_data pointer of a newly-allocated mbuf (m_get/MGET) to place
581 * an object of the specified size at the end of the mbuf, longword aligned.
582 */
583 #define M_ALIGN(m, len) do { \
584 (m)->m_data += (MLEN - (len)) & ~(sizeof(long) - 1); \
585 } while (0)
586
587 /*
588 * As above, for mbufs allocated with m_gethdr/MGETHDR
589 * or initialized by M_COPY_PKTHDR.
590 */
591 #define MH_ALIGN(m, len) do { \
592 (m)->m_data += (MHLEN - (len)) & ~(sizeof(long) - 1); \
593 } while (0)
594
595 /*
596 * Compute the amount of space available
597 * before the current start of data in an mbuf.
598 *
599 * The M_WRITABLE() is a temporary, conservative safety measure: the burden
600 * of checking writability of the mbuf data area rests solely with the caller.
601 */
602 #define M_LEADINGSPACE(m) \
603 ((m)->m_flags & M_EXT ? \
604 (M_WRITABLE(m) ? (m)->m_data - (m)->m_ext.ext_buf : 0): \
605 (m)->m_flags & M_PKTHDR ? (m)->m_data - (m)->m_pktdat : \
606 (m)->m_data - (m)->m_dat)
607
608 /*
609 * Compute the amount of space available
610 * after the end of data in an mbuf.
611 *
612 * The M_WRITABLE() is a temporary, conservative safety measure: the burden
613 * of checking writability of the mbuf data area rests solely with the caller.
614 */
615 #define M_TRAILINGSPACE(m) \
616 ((m)->m_flags & M_EXT ? \
617 (M_WRITABLE(m) ? (m)->m_ext.ext_buf + (m)->m_ext.ext_size \
618 - ((m)->m_data + (m)->m_len) : 0) : \
619 &(m)->m_dat[MLEN] - ((m)->m_data + (m)->m_len))
620
621 /*
622 * Arrange to prepend space of size plen to mbuf m.
623 * If a new mbuf must be allocated, how specifies whether to wait.
624 * If the allocation fails, the original mbuf chain is freed and m is
625 * set to NULL.
626 */
627 #define M_PREPEND(m, plen, how) do { \
628 struct mbuf **_mmp = &(m); \
629 struct mbuf *_mm = *_mmp; \
630 int _mplen = (plen); \
631 int __mhow = (how); \
632 \
633 MBUF_CHECKSLEEP(how); \
634 if (M_LEADINGSPACE(_mm) >= _mplen) { \
635 _mm->m_data -= _mplen; \
636 _mm->m_len += _mplen; \
637 } else \
638 _mm = m_prepend(_mm, _mplen, __mhow); \
639 if (_mm != NULL && _mm->m_flags & M_PKTHDR) \
640 _mm->m_pkthdr.len += _mplen; \
641 *_mmp = _mm; \
642 } while (0)
643
644 /*
645 * Change mbuf to new type.
646 * This is a relatively expensive operation and should be avoided.
647 */
648 #define MCHTYPE(m, t) m_chtype((m), (t))
649
650 /* Length to m_copy to copy all. */
651 #define M_COPYALL 1000000000
652
653 /* Compatibility with 4.3. */
654 #define m_copy(m, o, l) m_copym((m), (o), (l), M_DONTWAIT)
655
656 extern int max_datalen; /* MHLEN - max_hdr */
657 extern int max_hdr; /* Largest link + protocol header */
658 extern int max_linkhdr; /* Largest link-level header */
659 extern int max_protohdr; /* Largest protocol header */
660 extern struct mbstat mbstat; /* General mbuf stats/infos */
661 extern int nmbclusters; /* Maximum number of clusters */
662
663 struct uio;
664
665 void m_adj(struct mbuf *, int);
666 void m_align(struct mbuf *, int);
667 int m_apply(struct mbuf *, int, int,
668 int (*)(void *, void *, u_int), void *);
669 int m_append(struct mbuf *, int, c_caddr_t);
670 void m_cat(struct mbuf *, struct mbuf *);
671 void m_extadd(struct mbuf *, caddr_t, u_int,
672 void (*)(void *, void *), void *, int, int);
673 void m_copyback(struct mbuf *, int, int, c_caddr_t);
674 void m_copydata(const struct mbuf *, int, int, caddr_t);
675 struct mbuf *m_copym(struct mbuf *, int, int, int);
676 struct mbuf *m_copypacket(struct mbuf *, int);
677 void m_copy_pkthdr(struct mbuf *, struct mbuf *);
678 struct mbuf *m_copyup(struct mbuf *n, int len, int dstoff);
679 struct mbuf *m_defrag(struct mbuf *, int);
680 struct mbuf *m_devget(char *, int, int, struct ifnet *,
681 void (*)(char *, caddr_t, u_int));
682 struct mbuf *m_dup(struct mbuf *, int);
683 int m_dup_pkthdr(struct mbuf *, struct mbuf *, int);
684 u_int m_fixhdr(struct mbuf *);
685 struct mbuf *m_fragment(struct mbuf *, int, int);
686 void m_freem(struct mbuf *);
687 struct mbuf *m_getm(struct mbuf *, int, int, short);
688 struct mbuf *m_getptr(struct mbuf *, int, int *);
689 u_int m_length(struct mbuf *, struct mbuf **);
690 void m_move_pkthdr(struct mbuf *, struct mbuf *);
691 struct mbuf *m_prepend(struct mbuf *, int, int);
692 void m_print(const struct mbuf *, int);
693 struct mbuf *m_pulldown(struct mbuf *, int, int, int *);
694 struct mbuf *m_pullup(struct mbuf *, int);
695 struct mbuf *m_split(struct mbuf *, int, int);
696 struct mbuf *m_uiotombuf(struct uio *, int, int, int);
697 struct mbuf *m_unshare(struct mbuf *, int how);
698
699 /*-
700 * Network packets may have annotations attached by affixing a list
701 * of "packet tags" to the pkthdr structure. Packet tags are
702 * dynamically allocated semi-opaque data structures that have
703 * a fixed header (struct m_tag) that specifies the size of the
704 * memory block and a <cookie,type> pair that identifies it.
705 * The cookie is a 32-bit unique unsigned value used to identify
706 * a module or ABI. By convention this value is chose as the
707 * date+time that the module is created, expressed as the number of
708 * seconds since the epoch (e.g., using date -u +'%s'). The type value
709 * is an ABI/module-specific value that identifies a particular annotation
710 * and is private to the module. For compatibility with systems
711 * like OpenBSD that define packet tags w/o an ABI/module cookie,
712 * the value PACKET_ABI_COMPAT is used to implement m_tag_get and
713 * m_tag_find compatibility shim functions and several tag types are
714 * defined below. Users that do not require compatibility should use
715 * a private cookie value so that packet tag-related definitions
716 * can be maintained privately.
717 *
718 * Note that the packet tag returned by m_tag_alloc has the default
719 * memory alignment implemented by malloc. To reference private data
720 * one can use a construct like:
721 *
722 * struct m_tag *mtag = m_tag_alloc(...);
723 * struct foo *p = (struct foo *)(mtag+1);
724 *
725 * if the alignment of struct m_tag is sufficient for referencing members
726 * of struct foo. Otherwise it is necessary to embed struct m_tag within
727 * the private data structure to insure proper alignment; e.g.,
728 *
729 * struct foo {
730 * struct m_tag tag;
731 * ...
732 * };
733 * struct foo *p = (struct foo *) m_tag_alloc(...);
734 * struct m_tag *mtag = &p->tag;
735 */
736
737 /*
738 * Persistent tags stay with an mbuf until the mbuf is reclaimed.
739 * Otherwise tags are expected to ``vanish'' when they pass through
740 * a network interface. For most interfaces this happens normally
741 * as the tags are reclaimed when the mbuf is free'd. However in
742 * some special cases reclaiming must be done manually. An example
743 * is packets that pass through the loopback interface. Also, one
744 * must be careful to do this when ``turning around'' packets (e.g.,
745 * icmp_reflect).
746 *
747 * To mark a tag persistent bit-or this flag in when defining the
748 * tag id. The tag will then be treated as described above.
749 */
750 #define MTAG_PERSISTENT 0x800
751
752 #define PACKET_TAG_NONE 0 /* Nadda */
753
754 /* Packet tags for use with PACKET_ABI_COMPAT. */
755 #define PACKET_TAG_IPSEC_IN_DONE 1 /* IPsec applied, in */
756 #define PACKET_TAG_IPSEC_OUT_DONE 2 /* IPsec applied, out */
757 #define PACKET_TAG_IPSEC_IN_CRYPTO_DONE 3 /* NIC IPsec crypto done */
758 #define PACKET_TAG_IPSEC_OUT_CRYPTO_NEEDED 4 /* NIC IPsec crypto req'ed */
759 #define PACKET_TAG_IPSEC_IN_COULD_DO_CRYPTO 5 /* NIC notifies IPsec */
760 #define PACKET_TAG_IPSEC_PENDING_TDB 6 /* Reminder to do IPsec */
761 #define PACKET_TAG_BRIDGE 7 /* Bridge processing done */
762 #define PACKET_TAG_GIF 8 /* GIF processing done */
763 #define PACKET_TAG_GRE 9 /* GRE processing done */
764 #define PACKET_TAG_IN_PACKET_CHECKSUM 10 /* NIC checksumming done */
765 #define PACKET_TAG_ENCAP 11 /* Encap. processing */
766 #define PACKET_TAG_IPSEC_SOCKET 12 /* IPSEC socket ref */
767 #define PACKET_TAG_IPSEC_HISTORY 13 /* IPSEC history */
768 #define PACKET_TAG_IPV6_INPUT 14 /* IPV6 input processing */
769 #define PACKET_TAG_DUMMYNET 15 /* dummynet info */
770 #define PACKET_TAG_DIVERT 17 /* divert info */
771 #define PACKET_TAG_IPFORWARD 18 /* ipforward info */
772 #define PACKET_TAG_MACLABEL (19 | MTAG_PERSISTENT) /* MAC label */
773 #define PACKET_TAG_PF_ROUTED 21 /* PF routed, avoid loops */
774 #define PACKET_TAG_PF_FRAGCACHE 22 /* PF fragment cached */
775 #define PACKET_TAG_PF_QID 23 /* PF ALTQ queue id */
776 #define PACKET_TAG_PF_TAG 24 /* PF tagged */
777 #define PACKET_TAG_RTSOCKFAM 25 /* rtsock sa family */
778 #define PACKET_TAG_PF_TRANSLATE_LOCALHOST 26 /* PF translate localhost */
779 #define PACKET_TAG_IPOPTIONS 27 /* Saved IP options */
780 #define PACKET_TAG_CARP 28 /* CARP info */
781
782 /* Packet tag routines. */
783 struct m_tag *m_tag_alloc(u_int32_t, int, int, int);
784 void m_tag_delete(struct mbuf *, struct m_tag *);
785 void m_tag_delete_chain(struct mbuf *, struct m_tag *);
786 void m_tag_free_default(struct m_tag *);
787 struct m_tag *m_tag_locate(struct mbuf *, u_int32_t, int, struct m_tag *);
788 struct m_tag *m_tag_copy(struct m_tag *, int);
789 int m_tag_copy_chain(struct mbuf *, struct mbuf *, int);
790 void m_tag_delete_nonpersistent(struct mbuf *);
791
792 /*
793 * Initialize the list of tags associated with an mbuf.
794 */
795 static __inline void
796 m_tag_init(struct mbuf *m)
797 {
798 SLIST_INIT(&m->m_pkthdr.tags);
799 }
800
801 /*
802 * Set up the contents of a tag. Note that this does not
803 * fill in the free method; the caller is expected to do that.
804 *
805 * XXX probably should be called m_tag_init, but that was
806 * already taken.
807 */
808 static __inline void
809 m_tag_setup(struct m_tag *t, u_int32_t cookie, int type, int len)
810 {
811 t->m_tag_id = type;
812 t->m_tag_len = len;
813 t->m_tag_cookie = cookie;
814 }
815
816 /*
817 * Reclaim resources associated with a tag.
818 */
819 static __inline void
820 m_tag_free(struct m_tag *t)
821 {
822 (*t->m_tag_free)(t);
823 }
824
825 /*
826 * Return the first tag associated with an mbuf.
827 */
828 static __inline struct m_tag *
829 m_tag_first(struct mbuf *m)
830 {
831 return (SLIST_FIRST(&m->m_pkthdr.tags));
832 }
833
834 /*
835 * Return the next tag in the list of tags associated with an mbuf.
836 */
837 static __inline struct m_tag *
838 m_tag_next(struct mbuf *m, struct m_tag *t)
839 {
840 return (SLIST_NEXT(t, m_tag_link));
841 }
842
843 /*
844 * Prepend a tag to the list of tags associated with an mbuf.
845 */
846 static __inline void
847 m_tag_prepend(struct mbuf *m, struct m_tag *t)
848 {
849 SLIST_INSERT_HEAD(&m->m_pkthdr.tags, t, m_tag_link);
850 }
851
852 /*
853 * Unlink a tag from the list of tags associated with an mbuf.
854 */
855 static __inline void
856 m_tag_unlink(struct mbuf *m, struct m_tag *t)
857 {
858 SLIST_REMOVE(&m->m_pkthdr.tags, t, m_tag, m_tag_link);
859 }
860
861 /* These are for OpenBSD compatibility. */
862 #define MTAG_ABI_COMPAT 0 /* compatibility ABI */
863
864 static __inline struct m_tag *
865 m_tag_get(int type, int length, int wait)
866 {
867 return (m_tag_alloc(MTAG_ABI_COMPAT, type, length, wait));
868 }
869
870 static __inline struct m_tag *
871 m_tag_find(struct mbuf *m, int type, struct m_tag *start)
872 {
873 return (SLIST_EMPTY(&m->m_pkthdr.tags) ?
874 NULL : m_tag_locate(m, MTAG_ABI_COMPAT, type, start));
875 }
876
877 #endif /* _KERNEL */
878
879 #endif /* !_SYS_MBUF_H_ */
Cache object: 23e7a4d5d5f703dc0b48dfd3a18f5eca
|