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.
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 /*
48 * Mbufs are of a single size, MSIZE (sys/param.h), which includes overhead.
49 * An mbuf may add a single "mbuf cluster" of size MCLBYTES (also in
50 * sys/param.h), which has no additional overhead and is used instead of the
51 * internal data area; this is done when at least MINCLSIZE of data must be
52 * stored. Additionally, it is possible to allocate a separate buffer
53 * externally and attach it to the mbuf in a way similar to that of mbuf
54 * clusters.
55 */
56 #define MLEN (MSIZE - sizeof(struct m_hdr)) /* normal data len */
57 #define MHLEN (MLEN - sizeof(struct pkthdr)) /* data len w/pkthdr */
58 #define MINCLSIZE (MHLEN + 1) /* smallest amount to put in cluster */
59 #define M_MAXCOMPRESS (MHLEN / 2) /* max amount to copy for compression */
60
61 #ifdef _KERNEL
62 /*-
63 * Macros for type conversion:
64 * mtod(m, t) -- Convert mbuf pointer to data pointer of correct type.
65 * dtom(x) -- Convert data pointer within mbuf to mbuf pointer (XXX).
66 */
67 #define mtod(m, t) ((t)((m)->m_data))
68 #define dtom(x) ((struct mbuf *)((intptr_t)(x) & ~(MSIZE-1)))
69
70 /*
71 * Argument structure passed to UMA routines during mbuf and packet
72 * allocations.
73 */
74 struct mb_args {
75 int flags; /* Flags for mbuf being allocated */
76 short type; /* Type of mbuf being allocated */
77 };
78 #endif /* _KERNEL */
79
80 #if defined(__LP64__)
81 #define M_HDR_PAD 6
82 #else
83 #define M_HDR_PAD 2
84 #endif
85
86 /*
87 * Header present at the beginning of every mbuf.
88 */
89 struct m_hdr {
90 struct mbuf *mh_next; /* next buffer in chain */
91 struct mbuf *mh_nextpkt; /* next chain in queue/record */
92 caddr_t mh_data; /* location of data */
93 int mh_len; /* amount of data in this mbuf */
94 int mh_flags; /* flags; see below */
95 short mh_type; /* type of data in this mbuf */
96 uint8_t pad[M_HDR_PAD];/* word align */
97 };
98
99 /*
100 * Packet tag structure (see below for details).
101 */
102 struct m_tag {
103 SLIST_ENTRY(m_tag) m_tag_link; /* List of packet tags */
104 u_int16_t m_tag_id; /* Tag ID */
105 u_int16_t m_tag_len; /* Length of data */
106 u_int32_t m_tag_cookie; /* ABI/Module ID */
107 void (*m_tag_free)(struct m_tag *);
108 };
109
110 /*
111 * Record/packet header in first mbuf of chain; valid only if M_PKTHDR is set.
112 */
113 struct pkthdr {
114 struct ifnet *rcvif; /* rcv interface */
115 /* variables for ip and tcp reassembly */
116 void *header; /* pointer to packet header */
117 int len; /* total packet length */
118 /* variables for hardware checksum */
119 int csum_flags; /* flags regarding checksum */
120 int csum_data; /* data field used by csum routines */
121 u_int16_t tso_segsz; /* TSO segment size */
122 u_int16_t ether_vtag; /* Ethernet 802.1p+q vlan tag */
123 SLIST_HEAD(packet_tags, m_tag) tags; /* list of packet tags */
124 };
125
126 /*
127 * Description of external storage mapped into mbuf; valid only if M_EXT is
128 * set.
129 */
130 struct m_ext {
131 caddr_t ext_buf; /* start of buffer */
132 void (*ext_free) /* free routine if not the usual */
133 (void *, void *);
134 void *ext_args; /* optional argument pointer */
135 u_int ext_size; /* size of buffer, for ext_free */
136 volatile u_int *ref_cnt; /* pointer to ref count info */
137 int ext_type; /* type of external storage */
138 };
139
140 /*
141 * The core of the mbuf object along with some shortcut defines for practical
142 * purposes.
143 */
144 struct mbuf {
145 struct m_hdr m_hdr;
146 union {
147 struct {
148 struct pkthdr MH_pkthdr; /* M_PKTHDR set */
149 union {
150 struct m_ext MH_ext; /* M_EXT set */
151 char MH_databuf[MHLEN];
152 } MH_dat;
153 } MH;
154 char M_databuf[MLEN]; /* !M_PKTHDR, !M_EXT */
155 } M_dat;
156 };
157 #define m_next m_hdr.mh_next
158 #define m_len m_hdr.mh_len
159 #define m_data m_hdr.mh_data
160 #define m_type m_hdr.mh_type
161 #define m_flags m_hdr.mh_flags
162 #define m_nextpkt m_hdr.mh_nextpkt
163 #define m_act m_nextpkt
164 #define m_pkthdr M_dat.MH.MH_pkthdr
165 #define m_ext M_dat.MH.MH_dat.MH_ext
166 #define m_pktdat M_dat.MH.MH_dat.MH_databuf
167 #define m_dat M_dat.M_databuf
168
169 /*
170 * mbuf flags.
171 */
172 #define M_EXT 0x00000001 /* has associated external storage */
173 #define M_PKTHDR 0x00000002 /* start of record */
174 #define M_EOR 0x00000004 /* end of record */
175 #define M_RDONLY 0x00000008 /* associated data is marked read-only */
176 #define M_PROTO1 0x00000010 /* protocol-specific */
177 #define M_PROTO2 0x00000020 /* protocol-specific */
178 #define M_PROTO3 0x00000040 /* protocol-specific */
179 #define M_PROTO4 0x00000080 /* protocol-specific */
180 #define M_PROTO5 0x00000100 /* protocol-specific */
181 #define M_BCAST 0x00000200 /* send/received as link-level broadcast */
182 #define M_MCAST 0x00000400 /* send/received as link-level multicast */
183 #define M_FRAG 0x00000800 /* packet is a fragment of a larger packet */
184 #define M_FIRSTFRAG 0x00001000 /* packet is first fragment */
185 #define M_LASTFRAG 0x00002000 /* packet is last fragment */
186 #define M_SKIP_FIREWALL 0x00004000 /* skip firewall processing */
187 #define M_FREELIST 0x00008000 /* mbuf is on the free list */
188 #define M_VLANTAG 0x00010000 /* ether_vtag is valid */
189 #define M_PROMISC 0x00020000 /* packet was not for us */
190 #define M_NOFREE 0x00040000 /* do not free mbuf, embedded in cluster */
191 #define M_PROTO6 0x00080000 /* protocol-specific */
192 #define M_PROTO7 0x00100000 /* protocol-specific */
193 #define M_PROTO8 0x00200000 /* protocol-specific */
194 /*
195 * For RELENG_{6,7} steal these flags for limited multiple routing table
196 * support. In RELENG_8 and beyond, use just one flag and a tag.
197 */
198 #define M_FIB 0xF0000000 /* steal some bits to store fib number. */
199
200 #define M_NOTIFICATION M_PROTO5 /* SCTP notification */
201
202 /*
203 * Flags to purge when crossing layers.
204 */
205 #define M_PROTOFLAGS \
206 (M_PROTO1|M_PROTO2|M_PROTO3|M_PROTO4|M_PROTO5|M_PROTO6|M_PROTO7|M_PROTO8)
207
208 /*
209 * Flags preserved when copying m_pkthdr.
210 */
211 #define M_COPYFLAGS \
212 (M_PKTHDR|M_EOR|M_RDONLY|M_PROTOFLAGS|M_SKIP_FIREWALL|M_BCAST|M_MCAST|\
213 M_FRAG|M_FIRSTFRAG|M_LASTFRAG|M_VLANTAG|M_PROMISC|M_FIB)
214
215 /*
216 * External buffer types: identify ext_buf type.
217 */
218 #define EXT_CLUSTER 1 /* mbuf cluster */
219 #define EXT_SFBUF 2 /* sendfile(2)'s sf_bufs */
220 #define EXT_JUMBOP 3 /* jumbo cluster 4096 bytes */
221 #define EXT_JUMBO9 4 /* jumbo cluster 9216 bytes */
222 #define EXT_JUMBO16 5 /* jumbo cluster 16184 bytes */
223 #define EXT_PACKET 6 /* mbuf+cluster from packet zone */
224 #define EXT_MBUF 7 /* external mbuf reference (M_IOVEC) */
225 #define EXT_NET_DRV 100 /* custom ext_buf provided by net driver(s) */
226 #define EXT_MOD_TYPE 200 /* custom module's ext_buf type */
227 #define EXT_DISPOSABLE 300 /* can throw this buffer away w/page flipping */
228 #define EXT_EXTREF 400 /* has externally maintained ref_cnt ptr */
229
230 /*
231 * Flags indicating hw checksum support and sw checksum requirements. This
232 * field can be directly tested against if_data.ifi_hwassist.
233 */
234 #define CSUM_IP 0x0001 /* will csum IP */
235 #define CSUM_TCP 0x0002 /* will csum TCP */
236 #define CSUM_UDP 0x0004 /* will csum UDP */
237 #define CSUM_IP_FRAGS 0x0008 /* will csum IP fragments */
238 #define CSUM_FRAGMENT 0x0010 /* will do IP fragmentation */
239 #define CSUM_TSO 0x0020 /* will do TSO */
240
241 #define CSUM_IP_CHECKED 0x0100 /* did csum IP */
242 #define CSUM_IP_VALID 0x0200 /* ... the csum is valid */
243 #define CSUM_DATA_VALID 0x0400 /* csum_data field is valid */
244 #define CSUM_PSEUDO_HDR 0x0800 /* csum_data has pseudo hdr */
245
246 #define CSUM_DELAY_DATA (CSUM_TCP | CSUM_UDP)
247 #define CSUM_DELAY_IP (CSUM_IP) /* XXX add ipv6 here too? */
248
249 /*
250 * mbuf types.
251 */
252 #define MT_NOTMBUF 0 /* USED INTERNALLY ONLY! Object is not mbuf */
253 #define MT_DATA 1 /* dynamic (data) allocation */
254 #define MT_HEADER MT_DATA /* packet header, use M_PKTHDR instead */
255 #define MT_SONAME 8 /* socket name */
256 #define MT_CONTROL 14 /* extra-data protocol message */
257 #define MT_OOBDATA 15 /* expedited data */
258 #define MT_NTYPES 16 /* number of mbuf types for mbtypes[] */
259
260 #define MT_NOINIT 255 /* Not a type but a flag to allocate
261 a non-initialized mbuf */
262
263 #define MB_NOTAGS 0x1UL /* no tags attached to mbuf */
264
265 /*
266 * General mbuf allocator statistics structure.
267 *
268 * Many of these statistics are no longer used; we instead track many
269 * allocator statistics through UMA's built in statistics mechanism.
270 */
271 struct mbstat {
272 u_long m_mbufs; /* XXX */
273 u_long m_mclusts; /* XXX */
274
275 u_long m_drain; /* times drained protocols for space */
276 u_long m_mcfail; /* XXX: times m_copym failed */
277 u_long m_mpfail; /* XXX: times m_pullup failed */
278 u_long m_msize; /* length of an mbuf */
279 u_long m_mclbytes; /* length of an mbuf cluster */
280 u_long m_minclsize; /* min length of data to allocate a cluster */
281 u_long m_mlen; /* length of data in an mbuf */
282 u_long m_mhlen; /* length of data in a header mbuf */
283
284 /* Number of mbtypes (gives # elems in mbtypes[] array) */
285 short m_numtypes;
286
287 /* XXX: Sendfile stats should eventually move to their own struct */
288 u_long sf_iocnt; /* times sendfile had to do disk I/O */
289 u_long sf_allocfail; /* times sfbuf allocation failed */
290 u_long sf_allocwait; /* times sfbuf allocation had to wait */
291 };
292
293 /*
294 * Flags specifying how an allocation should be made.
295 *
296 * The flag to use is as follows:
297 * - M_DONTWAIT or M_NOWAIT from an interrupt handler to not block allocation.
298 * - M_WAIT or M_WAITOK or M_TRYWAIT from wherever it is safe to block.
299 *
300 * M_DONTWAIT/M_NOWAIT means that we will not block the thread explicitly and
301 * if we cannot allocate immediately we may return NULL, whereas
302 * M_WAIT/M_WAITOK/M_TRYWAIT means that if we cannot allocate resources we
303 * will block until they are available, and thus never return NULL.
304 *
305 * XXX Eventually just phase this out to use M_WAITOK/M_NOWAIT.
306 */
307 #define MBTOM(how) (how)
308 #define M_DONTWAIT M_NOWAIT
309 #define M_TRYWAIT M_WAITOK
310 #define M_WAIT M_WAITOK
311
312 /*
313 * String names of mbuf-related UMA(9) and malloc(9) types. Exposed to
314 * !_KERNEL so that monitoring tools can look up the zones with
315 * libmemstat(3).
316 */
317 #define MBUF_MEM_NAME "mbuf"
318 #define MBUF_CLUSTER_MEM_NAME "mbuf_cluster"
319 #define MBUF_PACKET_MEM_NAME "mbuf_packet"
320 #define MBUF_JUMBOP_MEM_NAME "mbuf_jumbo_pagesize"
321 #define MBUF_JUMBO9_MEM_NAME "mbuf_jumbo_9k"
322 #define MBUF_JUMBO16_MEM_NAME "mbuf_jumbo_16k"
323 #define MBUF_TAG_MEM_NAME "mbuf_tag"
324 #define MBUF_EXTREFCNT_MEM_NAME "mbuf_ext_refcnt"
325
326 #ifdef _KERNEL
327
328 #ifdef WITNESS
329 #define MBUF_CHECKSLEEP(how) do { \
330 if (how == M_WAITOK) \
331 WITNESS_WARN(WARN_GIANTOK | WARN_SLEEPOK, NULL, \
332 "Sleeping in \"%s\"", __func__); \
333 } while (0)
334 #else
335 #define MBUF_CHECKSLEEP(how)
336 #endif
337
338 /*
339 * Network buffer allocation API
340 *
341 * The rest of it is defined in kern/kern_mbuf.c
342 */
343
344 extern uma_zone_t zone_mbuf;
345 extern uma_zone_t zone_clust;
346 extern uma_zone_t zone_pack;
347 extern uma_zone_t zone_jumbop;
348 extern uma_zone_t zone_jumbo9;
349 extern uma_zone_t zone_jumbo16;
350 extern uma_zone_t zone_ext_refcnt;
351
352 static __inline struct mbuf *m_getcl(int how, short type, int flags);
353 static __inline struct mbuf *m_get(int how, short type);
354 static __inline struct mbuf *m_gethdr(int how, short type);
355 static __inline struct mbuf *m_getjcl(int how, short type, int flags,
356 int size);
357 static __inline struct mbuf *m_getclr(int how, short type); /* XXX */
358 static __inline struct mbuf *m_free(struct mbuf *m);
359 static __inline void m_clget(struct mbuf *m, int how);
360 static __inline void *m_cljget(struct mbuf *m, int how, int size);
361 static __inline void m_chtype(struct mbuf *m, short new_type);
362 void mb_free_ext(struct mbuf *);
363 static __inline struct mbuf *m_last(struct mbuf *m);
364
365 static __inline int
366 m_gettype(int size)
367 {
368 int type;
369
370 switch (size) {
371 case MSIZE:
372 type = EXT_MBUF;
373 break;
374 case MCLBYTES:
375 type = EXT_CLUSTER;
376 break;
377 #if MJUMPAGESIZE != MCLBYTES
378 case MJUMPAGESIZE:
379 type = EXT_JUMBOP;
380 break;
381 #endif
382 case MJUM9BYTES:
383 type = EXT_JUMBO9;
384 break;
385 case MJUM16BYTES:
386 type = EXT_JUMBO16;
387 break;
388 default:
389 panic("%s: m_getjcl: invalid cluster size", __func__);
390 }
391
392 return (type);
393 }
394
395 static __inline uma_zone_t
396 m_getzone(int size)
397 {
398 uma_zone_t zone;
399
400 switch (size) {
401 case MSIZE:
402 zone = zone_mbuf;
403 break;
404 case MCLBYTES:
405 zone = zone_clust;
406 break;
407 #if MJUMPAGESIZE != MCLBYTES
408 case MJUMPAGESIZE:
409 zone = zone_jumbop;
410 break;
411 #endif
412 case MJUM9BYTES:
413 zone = zone_jumbo9;
414 break;
415 case MJUM16BYTES:
416 zone = zone_jumbo16;
417 break;
418 default:
419 panic("%s: m_getjcl: invalid cluster type", __func__);
420 }
421
422 return (zone);
423 }
424
425 static __inline struct mbuf *
426 m_get(int how, short type)
427 {
428 struct mb_args args;
429
430 args.flags = 0;
431 args.type = type;
432 return ((struct mbuf *)(uma_zalloc_arg(zone_mbuf, &args, how)));
433 }
434
435 /*
436 * XXX This should be deprecated, very little use.
437 */
438 static __inline struct mbuf *
439 m_getclr(int how, short type)
440 {
441 struct mbuf *m;
442 struct mb_args args;
443
444 args.flags = 0;
445 args.type = type;
446 m = uma_zalloc_arg(zone_mbuf, &args, how);
447 if (m != NULL)
448 bzero(m->m_data, MLEN);
449 return (m);
450 }
451
452 static __inline struct mbuf *
453 m_gethdr(int how, short type)
454 {
455 struct mb_args args;
456
457 args.flags = M_PKTHDR;
458 args.type = type;
459 return ((struct mbuf *)(uma_zalloc_arg(zone_mbuf, &args, how)));
460 }
461
462 static __inline struct mbuf *
463 m_getcl(int how, short type, int flags)
464 {
465 struct mb_args args;
466
467 args.flags = flags;
468 args.type = type;
469 return ((struct mbuf *)(uma_zalloc_arg(zone_pack, &args, how)));
470 }
471
472 /*
473 * m_getjcl() returns an mbuf with a cluster of the specified size attached.
474 * For size it takes MCLBYTES, MJUMPAGESIZE, MJUM9BYTES, MJUM16BYTES.
475 *
476 * XXX: This is rather large, should be real function maybe.
477 */
478 static __inline struct mbuf *
479 m_getjcl(int how, short type, int flags, int size)
480 {
481 struct mb_args args;
482 struct mbuf *m, *n;
483 uma_zone_t zone;
484
485 args.flags = flags;
486 args.type = type;
487
488 m = uma_zalloc_arg(zone_mbuf, &args, how);
489 if (m == NULL)
490 return (NULL);
491
492 zone = m_getzone(size);
493 n = uma_zalloc_arg(zone, m, how);
494 if (n == NULL) {
495 uma_zfree(zone_mbuf, m);
496 return (NULL);
497 }
498 return (m);
499 }
500
501 static __inline void
502 m_free_fast(struct mbuf *m)
503 {
504 KASSERT(SLIST_EMPTY(&m->m_pkthdr.tags), ("doing fast free of mbuf with tags"));
505
506 uma_zfree_arg(zone_mbuf, m, (void *)MB_NOTAGS);
507 }
508
509 static __inline struct mbuf *
510 m_free(struct mbuf *m)
511 {
512 struct mbuf *n = m->m_next;
513
514 if (m->m_flags & M_EXT)
515 mb_free_ext(m);
516 else if ((m->m_flags & M_NOFREE) == 0)
517 uma_zfree(zone_mbuf, m);
518 return (n);
519 }
520
521 static __inline void
522 m_clget(struct mbuf *m, int how)
523 {
524
525 if (m->m_flags & M_EXT)
526 printf("%s: %p mbuf already has cluster\n", __func__, m);
527 m->m_ext.ext_buf = (char *)NULL;
528 uma_zalloc_arg(zone_clust, m, how);
529 /*
530 * On a cluster allocation failure, drain the packet zone and retry,
531 * we might be able to loosen a few clusters up on the drain.
532 */
533 if ((how & M_NOWAIT) && (m->m_ext.ext_buf == NULL)) {
534 zone_drain(zone_pack);
535 uma_zalloc_arg(zone_clust, m, how);
536 }
537 }
538
539 /*
540 * m_cljget() is different from m_clget() as it can allocate clusters without
541 * attaching them to an mbuf. In that case the return value is the pointer
542 * to the cluster of the requested size. If an mbuf was specified, it gets
543 * the cluster attached to it and the return value can be safely ignored.
544 * For size it takes MCLBYTES, MJUMPAGESIZE, MJUM9BYTES, MJUM16BYTES.
545 */
546 static __inline void *
547 m_cljget(struct mbuf *m, int how, int size)
548 {
549 uma_zone_t zone;
550
551 if (m && m->m_flags & M_EXT)
552 printf("%s: %p mbuf already has cluster\n", __func__, m);
553 if (m != NULL)
554 m->m_ext.ext_buf = NULL;
555
556 zone = m_getzone(size);
557 return (uma_zalloc_arg(zone, m, how));
558 }
559
560 static __inline void
561 m_cljset(struct mbuf *m, void *cl, int type)
562 {
563 uma_zone_t zone;
564 int size;
565
566 switch (type) {
567 case EXT_CLUSTER:
568 size = MCLBYTES;
569 zone = zone_clust;
570 break;
571 #if MJUMPAGESIZE != MCLBYTES
572 case EXT_JUMBOP:
573 size = MJUMPAGESIZE;
574 zone = zone_jumbop;
575 break;
576 #endif
577 case EXT_JUMBO9:
578 size = MJUM9BYTES;
579 zone = zone_jumbo9;
580 break;
581 case EXT_JUMBO16:
582 size = MJUM16BYTES;
583 zone = zone_jumbo16;
584 break;
585 default:
586 panic("unknown cluster type");
587 break;
588 }
589
590 m->m_data = m->m_ext.ext_buf = cl;
591 m->m_ext.ext_free = m->m_ext.ext_args = NULL;
592 m->m_ext.ext_size = size;
593 m->m_ext.ext_type = type;
594 m->m_ext.ref_cnt = uma_find_refcnt(zone, cl);
595 m->m_flags |= M_EXT;
596
597 }
598
599 static __inline void
600 m_chtype(struct mbuf *m, short new_type)
601 {
602
603 m->m_type = new_type;
604 }
605
606 static __inline struct mbuf *
607 m_last(struct mbuf *m)
608 {
609
610 while (m->m_next)
611 m = m->m_next;
612 return (m);
613 }
614
615 /*
616 * mbuf, cluster, and external object allocation macros (for compatibility
617 * purposes).
618 */
619 #define M_MOVE_PKTHDR(to, from) m_move_pkthdr((to), (from))
620 #define MGET(m, how, type) ((m) = m_get((how), (type)))
621 #define MGETHDR(m, how, type) ((m) = m_gethdr((how), (type)))
622 #define MCLGET(m, how) m_clget((m), (how))
623 #define MEXTADD(m, buf, size, free, args, flags, type) \
624 m_extadd((m), (caddr_t)(buf), (size), (free), (args), (flags), (type))
625 #define m_getm(m, len, how, type) \
626 m_getm2((m), (len), (how), (type), M_PKTHDR)
627
628 /*
629 * Evaluate TRUE if it's safe to write to the mbuf m's data region (this can
630 * be both the local data payload, or an external buffer area, depending on
631 * whether M_EXT is set).
632 */
633 #define M_WRITABLE(m) (!((m)->m_flags & M_RDONLY) && \
634 (!(((m)->m_flags & M_EXT)) || \
635 (*((m)->m_ext.ref_cnt) == 1)) ) \
636
637 /* Check if the supplied mbuf has a packet header, or else panic. */
638 #define M_ASSERTPKTHDR(m) \
639 KASSERT(m != NULL && m->m_flags & M_PKTHDR, \
640 ("%s: no mbuf packet header!", __func__))
641
642 /*
643 * Ensure that the supplied mbuf is a valid, non-free mbuf.
644 *
645 * XXX: Broken at the moment. Need some UMA magic to make it work again.
646 */
647 #define M_ASSERTVALID(m) \
648 KASSERT((((struct mbuf *)m)->m_flags & 0) == 0, \
649 ("%s: attempted use of a free mbuf!", __func__))
650
651 /*
652 * Set the m_data pointer of a newly-allocated mbuf (m_get/MGET) to place an
653 * object of the specified size at the end of the mbuf, longword aligned.
654 */
655 #define M_ALIGN(m, len) do { \
656 KASSERT(!((m)->m_flags & (M_PKTHDR|M_EXT)), \
657 ("%s: M_ALIGN not normal mbuf", __func__)); \
658 KASSERT((m)->m_data == (m)->m_dat, \
659 ("%s: M_ALIGN not a virgin mbuf", __func__)); \
660 (m)->m_data += (MLEN - (len)) & ~(sizeof(long) - 1); \
661 } while (0)
662
663 /*
664 * As above, for mbufs allocated with m_gethdr/MGETHDR or initialized by
665 * M_DUP/MOVE_PKTHDR.
666 */
667 #define MH_ALIGN(m, len) do { \
668 KASSERT((m)->m_flags & M_PKTHDR && !((m)->m_flags & M_EXT), \
669 ("%s: MH_ALIGN not PKTHDR mbuf", __func__)); \
670 KASSERT((m)->m_data == (m)->m_pktdat, \
671 ("%s: MH_ALIGN not a virgin mbuf", __func__)); \
672 (m)->m_data += (MHLEN - (len)) & ~(sizeof(long) - 1); \
673 } while (0)
674
675 /*
676 * Compute the amount of space available before the current start of data in
677 * an mbuf.
678 *
679 * The M_WRITABLE() is a temporary, conservative safety measure: the burden
680 * of checking writability of the mbuf data area rests solely with the caller.
681 */
682 #define M_LEADINGSPACE(m) \
683 ((m)->m_flags & M_EXT ? \
684 (M_WRITABLE(m) ? (m)->m_data - (m)->m_ext.ext_buf : 0): \
685 (m)->m_flags & M_PKTHDR ? (m)->m_data - (m)->m_pktdat : \
686 (m)->m_data - (m)->m_dat)
687
688 /*
689 * Compute the amount of space available after the end of data in an mbuf.
690 *
691 * The M_WRITABLE() is a temporary, conservative safety measure: the burden
692 * of checking writability of the mbuf data area rests solely with the caller.
693 */
694 #define M_TRAILINGSPACE(m) \
695 ((m)->m_flags & M_EXT ? \
696 (M_WRITABLE(m) ? (m)->m_ext.ext_buf + (m)->m_ext.ext_size \
697 - ((m)->m_data + (m)->m_len) : 0) : \
698 &(m)->m_dat[MLEN] - ((m)->m_data + (m)->m_len))
699
700 /*
701 * Arrange to prepend space of size plen to mbuf m. If a new mbuf must be
702 * allocated, how specifies whether to wait. If the allocation fails, the
703 * original mbuf chain is freed and m is set to NULL.
704 */
705 #define M_PREPEND(m, plen, how) do { \
706 struct mbuf **_mmp = &(m); \
707 struct mbuf *_mm = *_mmp; \
708 int _mplen = (plen); \
709 int __mhow = (how); \
710 \
711 MBUF_CHECKSLEEP(how); \
712 if (M_LEADINGSPACE(_mm) >= _mplen) { \
713 _mm->m_data -= _mplen; \
714 _mm->m_len += _mplen; \
715 } else \
716 _mm = m_prepend(_mm, _mplen, __mhow); \
717 if (_mm != NULL && _mm->m_flags & M_PKTHDR) \
718 _mm->m_pkthdr.len += _mplen; \
719 *_mmp = _mm; \
720 } while (0)
721
722 /*
723 * Change mbuf to new type. This is a relatively expensive operation and
724 * should be avoided.
725 */
726 #define MCHTYPE(m, t) m_chtype((m), (t))
727
728 /* Length to m_copy to copy all. */
729 #define M_COPYALL 1000000000
730
731 /* Compatibility with 4.3. */
732 #define m_copy(m, o, l) m_copym((m), (o), (l), M_DONTWAIT)
733
734 extern int max_datalen; /* MHLEN - max_hdr */
735 extern int max_hdr; /* Largest link + protocol header */
736 extern int max_linkhdr; /* Largest link-level header */
737 extern int max_protohdr; /* Largest protocol header */
738 extern struct mbstat mbstat; /* General mbuf stats/infos */
739 extern int nmbclusters; /* Maximum number of clusters */
740
741 struct uio;
742
743 void m_adj(struct mbuf *, int);
744 void m_align(struct mbuf *, int);
745 int m_apply(struct mbuf *, int, int,
746 int (*)(void *, void *, u_int), void *);
747 int m_append(struct mbuf *, int, c_caddr_t);
748 void m_cat(struct mbuf *, struct mbuf *);
749 void m_extadd(struct mbuf *, caddr_t, u_int,
750 void (*)(void *, void *), void *, int, int);
751 struct mbuf *m_collapse(struct mbuf *, int, int);
752 void m_copyback(struct mbuf *, int, int, c_caddr_t);
753 void m_copydata(const struct mbuf *, int, int, caddr_t);
754 struct mbuf *m_copym(struct mbuf *, int, int, int);
755 struct mbuf *m_copymdata(struct mbuf *, struct mbuf *,
756 int, int, int, int);
757 struct mbuf *m_copypacket(struct mbuf *, int);
758 void m_copy_pkthdr(struct mbuf *, struct mbuf *);
759 struct mbuf *m_copyup(struct mbuf *n, int len, int dstoff);
760 struct mbuf *m_defrag(struct mbuf *, int);
761 void m_demote(struct mbuf *, int);
762 struct mbuf *m_devget(char *, int, int, struct ifnet *,
763 void (*)(char *, caddr_t, u_int));
764 struct mbuf *m_dup(struct mbuf *, int);
765 int m_dup_pkthdr(struct mbuf *, struct mbuf *, int);
766 u_int m_fixhdr(struct mbuf *);
767 struct mbuf *m_fragment(struct mbuf *, int, int);
768 void m_freem(struct mbuf *);
769 struct mbuf *m_getm2(struct mbuf *, int, int, short, int);
770 struct mbuf *m_getptr(struct mbuf *, int, int *);
771 u_int m_length(struct mbuf *, struct mbuf **);
772 void m_move_pkthdr(struct mbuf *, struct mbuf *);
773 struct mbuf *m_prepend(struct mbuf *, int, int);
774 void m_print(const struct mbuf *, int);
775 struct mbuf *m_pulldown(struct mbuf *, int, int, int *);
776 struct mbuf *m_pullup(struct mbuf *, int);
777 int m_sanity(struct mbuf *, int);
778 struct mbuf *m_split(struct mbuf *, int, int);
779 struct mbuf *m_uiotombuf(struct uio *, int, int, int, int);
780 struct mbuf *m_unshare(struct mbuf *, int how);
781
782 /*-
783 * Network packets may have annotations attached by affixing a list of
784 * "packet tags" to the pkthdr structure. Packet tags are dynamically
785 * allocated semi-opaque data structures that have a fixed header
786 * (struct m_tag) that specifies the size of the memory block and a
787 * <cookie,type> pair that identifies it. The cookie is a 32-bit unique
788 * unsigned value used to identify a module or ABI. By convention this value
789 * is chosen as the date+time that the module is created, expressed as the
790 * number of seconds since the epoch (e.g., using date -u +'%s'). The type
791 * value is an ABI/module-specific value that identifies a particular
792 * annotation and is private to the module. For compatibility with systems
793 * like OpenBSD that define packet tags w/o an ABI/module cookie, the value
794 * PACKET_ABI_COMPAT is used to implement m_tag_get and m_tag_find
795 * compatibility shim functions and several tag types are defined below.
796 * Users that do not require compatibility should use a private cookie value
797 * so that packet tag-related definitions can be maintained privately.
798 *
799 * Note that the packet tag returned by m_tag_alloc has the default memory
800 * alignment implemented by malloc. To reference private data one can use a
801 * construct like:
802 *
803 * struct m_tag *mtag = m_tag_alloc(...);
804 * struct foo *p = (struct foo *)(mtag+1);
805 *
806 * if the alignment of struct m_tag is sufficient for referencing members of
807 * struct foo. Otherwise it is necessary to embed struct m_tag within the
808 * private data structure to insure proper alignment; e.g.,
809 *
810 * struct foo {
811 * struct m_tag tag;
812 * ...
813 * };
814 * struct foo *p = (struct foo *) m_tag_alloc(...);
815 * struct m_tag *mtag = &p->tag;
816 */
817
818 /*
819 * Persistent tags stay with an mbuf until the mbuf is reclaimed. Otherwise
820 * tags are expected to ``vanish'' when they pass through a network
821 * interface. For most interfaces this happens normally as the tags are
822 * reclaimed when the mbuf is free'd. However in some special cases
823 * reclaiming must be done manually. An example is packets that pass through
824 * the loopback interface. Also, one must be careful to do this when
825 * ``turning around'' packets (e.g., icmp_reflect).
826 *
827 * To mark a tag persistent bit-or this flag in when defining the tag id.
828 * The tag will then be treated as described above.
829 */
830 #define MTAG_PERSISTENT 0x800
831
832 #define PACKET_TAG_NONE 0 /* Nadda */
833
834 /* Packet tags for use with PACKET_ABI_COMPAT. */
835 #define PACKET_TAG_IPSEC_IN_DONE 1 /* IPsec applied, in */
836 #define PACKET_TAG_IPSEC_OUT_DONE 2 /* IPsec applied, out */
837 #define PACKET_TAG_IPSEC_IN_CRYPTO_DONE 3 /* NIC IPsec crypto done */
838 #define PACKET_TAG_IPSEC_OUT_CRYPTO_NEEDED 4 /* NIC IPsec crypto req'ed */
839 #define PACKET_TAG_IPSEC_IN_COULD_DO_CRYPTO 5 /* NIC notifies IPsec */
840 #define PACKET_TAG_IPSEC_PENDING_TDB 6 /* Reminder to do IPsec */
841 #define PACKET_TAG_BRIDGE 7 /* Bridge processing done */
842 #define PACKET_TAG_GIF 8 /* GIF processing done */
843 #define PACKET_TAG_GRE 9 /* GRE processing done */
844 #define PACKET_TAG_IN_PACKET_CHECKSUM 10 /* NIC checksumming done */
845 #define PACKET_TAG_ENCAP 11 /* Encap. processing */
846 #define PACKET_TAG_IPSEC_SOCKET 12 /* IPSEC socket ref */
847 #define PACKET_TAG_IPSEC_HISTORY 13 /* IPSEC history */
848 #define PACKET_TAG_IPV6_INPUT 14 /* IPV6 input processing */
849 #define PACKET_TAG_DUMMYNET 15 /* dummynet info */
850 #define PACKET_TAG_DIVERT 17 /* divert info */
851 #define PACKET_TAG_IPFORWARD 18 /* ipforward info */
852 #define PACKET_TAG_MACLABEL (19 | MTAG_PERSISTENT) /* MAC label */
853 #define PACKET_TAG_PF 21 /* PF + ALTQ information */
854 #define PACKET_TAG_RTSOCKFAM 25 /* rtsock sa family */
855 #define PACKET_TAG_IPOPTIONS 27 /* Saved IP options */
856 #define PACKET_TAG_CARP 28 /* CARP info */
857
858 /* Specific cookies and tags. */
859
860 /* Packet tag routines. */
861 struct m_tag *m_tag_alloc(u_int32_t, int, int, int);
862 void m_tag_delete(struct mbuf *, struct m_tag *);
863 void m_tag_delete_chain(struct mbuf *, struct m_tag *);
864 void m_tag_free_default(struct m_tag *);
865 struct m_tag *m_tag_locate(struct mbuf *, u_int32_t, int, struct m_tag *);
866 struct m_tag *m_tag_copy(struct m_tag *, int);
867 int m_tag_copy_chain(struct mbuf *, struct mbuf *, int);
868 void m_tag_delete_nonpersistent(struct mbuf *);
869
870 /*
871 * Initialize the list of tags associated with an mbuf.
872 */
873 static __inline void
874 m_tag_init(struct mbuf *m)
875 {
876
877 SLIST_INIT(&m->m_pkthdr.tags);
878 }
879
880 /*
881 * Set up the contents of a tag. Note that this does not fill in the free
882 * method; the caller is expected to do that.
883 *
884 * XXX probably should be called m_tag_init, but that was already taken.
885 */
886 static __inline void
887 m_tag_setup(struct m_tag *t, u_int32_t cookie, int type, int len)
888 {
889
890 t->m_tag_id = type;
891 t->m_tag_len = len;
892 t->m_tag_cookie = cookie;
893 }
894
895 /*
896 * Reclaim resources associated with a tag.
897 */
898 static __inline void
899 m_tag_free(struct m_tag *t)
900 {
901
902 (*t->m_tag_free)(t);
903 }
904
905 /*
906 * Return the first tag associated with an mbuf.
907 */
908 static __inline struct m_tag *
909 m_tag_first(struct mbuf *m)
910 {
911
912 return (SLIST_FIRST(&m->m_pkthdr.tags));
913 }
914
915 /*
916 * Return the next tag in the list of tags associated with an mbuf.
917 */
918 static __inline struct m_tag *
919 m_tag_next(struct mbuf *m, struct m_tag *t)
920 {
921
922 return (SLIST_NEXT(t, m_tag_link));
923 }
924
925 /*
926 * Prepend a tag to the list of tags associated with an mbuf.
927 */
928 static __inline void
929 m_tag_prepend(struct mbuf *m, struct m_tag *t)
930 {
931
932 SLIST_INSERT_HEAD(&m->m_pkthdr.tags, t, m_tag_link);
933 }
934
935 /*
936 * Unlink a tag from the list of tags associated with an mbuf.
937 */
938 static __inline void
939 m_tag_unlink(struct mbuf *m, struct m_tag *t)
940 {
941
942 SLIST_REMOVE(&m->m_pkthdr.tags, t, m_tag, m_tag_link);
943 }
944
945 /* These are for OpenBSD compatibility. */
946 #define MTAG_ABI_COMPAT 0 /* compatibility ABI */
947
948 static __inline struct m_tag *
949 m_tag_get(int type, int length, int wait)
950 {
951 return (m_tag_alloc(MTAG_ABI_COMPAT, type, length, wait));
952 }
953
954 static __inline struct m_tag *
955 m_tag_find(struct mbuf *m, int type, struct m_tag *start)
956 {
957 return (SLIST_EMPTY(&m->m_pkthdr.tags) ? (struct m_tag *)NULL :
958 m_tag_locate(m, MTAG_ABI_COMPAT, type, start));
959 }
960
961 /* XXX temporary FIB methods probably eventually use tags.*/
962 #define M_FIBSHIFT 28
963 #define M_FIBMASK 0x0F
964
965 /* get the fib from an mbuf and if it is not set, return the default */
966 #define M_GETFIB(_m) \
967 ((((_m)->m_flags & M_FIB) >> M_FIBSHIFT) & M_FIBMASK)
968
969 #define M_SETFIB(_m, _fib) do { \
970 _m->m_flags &= ~M_FIB; \
971 _m->m_flags |= (((_fib) << M_FIBSHIFT) & M_FIB); \
972 } while (0)
973
974 #endif /* _KERNEL */
975
976 #endif /* !_SYS_MBUF_H_ */
Cache object: 21a233cfc52ed7dc2004e68d2b39c6fd
|