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 #define CSUM_SCTP 0x0040 /* will csum SCTP */
241
242 #define CSUM_IP_CHECKED 0x0100 /* did csum IP */
243 #define CSUM_IP_VALID 0x0200 /* ... the csum is valid */
244 #define CSUM_DATA_VALID 0x0400 /* csum_data field is valid */
245 #define CSUM_PSEUDO_HDR 0x0800 /* csum_data has pseudo hdr */
246 #define CSUM_SCTP_VALID 0x1000 /* SCTP checksum is valid */
247
248 #define CSUM_DELAY_DATA (CSUM_TCP | CSUM_UDP)
249 #define CSUM_DELAY_IP (CSUM_IP) /* XXX add ipv6 here too? */
250
251 /*
252 * mbuf types.
253 */
254 #define MT_NOTMBUF 0 /* USED INTERNALLY ONLY! Object is not mbuf */
255 #define MT_DATA 1 /* dynamic (data) allocation */
256 #define MT_HEADER MT_DATA /* packet header, use M_PKTHDR instead */
257 #define MT_SONAME 8 /* socket name */
258 #define MT_CONTROL 14 /* extra-data protocol message */
259 #define MT_OOBDATA 15 /* expedited data */
260 #define MT_NTYPES 16 /* number of mbuf types for mbtypes[] */
261
262 #define MT_NOINIT 255 /* Not a type but a flag to allocate
263 a non-initialized mbuf */
264
265 #define MB_NOTAGS 0x1UL /* no tags attached to mbuf */
266
267 /*
268 * General mbuf allocator statistics structure.
269 *
270 * Many of these statistics are no longer used; we instead track many
271 * allocator statistics through UMA's built in statistics mechanism.
272 */
273 struct mbstat {
274 u_long m_mbufs; /* XXX */
275 u_long m_mclusts; /* XXX */
276
277 u_long m_drain; /* times drained protocols for space */
278 u_long m_mcfail; /* XXX: times m_copym failed */
279 u_long m_mpfail; /* XXX: times m_pullup failed */
280 u_long m_msize; /* length of an mbuf */
281 u_long m_mclbytes; /* length of an mbuf cluster */
282 u_long m_minclsize; /* min length of data to allocate a cluster */
283 u_long m_mlen; /* length of data in an mbuf */
284 u_long m_mhlen; /* length of data in a header mbuf */
285
286 /* Number of mbtypes (gives # elems in mbtypes[] array) */
287 short m_numtypes;
288
289 /* XXX: Sendfile stats should eventually move to their own struct */
290 u_long sf_iocnt; /* times sendfile had to do disk I/O */
291 u_long sf_allocfail; /* times sfbuf allocation failed */
292 u_long sf_allocwait; /* times sfbuf allocation had to wait */
293 };
294
295 /*
296 * Flags specifying how an allocation should be made.
297 *
298 * The flag to use is as follows:
299 * - M_DONTWAIT or M_NOWAIT from an interrupt handler to not block allocation.
300 * - M_WAIT or M_WAITOK or M_TRYWAIT from wherever it is safe to block.
301 *
302 * M_DONTWAIT/M_NOWAIT means that we will not block the thread explicitly and
303 * if we cannot allocate immediately we may return NULL, whereas
304 * M_WAIT/M_WAITOK/M_TRYWAIT means that if we cannot allocate resources we
305 * will block until they are available, and thus never return NULL.
306 *
307 * XXX Eventually just phase this out to use M_WAITOK/M_NOWAIT.
308 */
309 #define MBTOM(how) (how)
310 #define M_DONTWAIT M_NOWAIT
311 #define M_TRYWAIT M_WAITOK
312 #define M_WAIT M_WAITOK
313
314 /*
315 * String names of mbuf-related UMA(9) and malloc(9) types. Exposed to
316 * !_KERNEL so that monitoring tools can look up the zones with
317 * libmemstat(3).
318 */
319 #define MBUF_MEM_NAME "mbuf"
320 #define MBUF_CLUSTER_MEM_NAME "mbuf_cluster"
321 #define MBUF_PACKET_MEM_NAME "mbuf_packet"
322 #define MBUF_JUMBOP_MEM_NAME "mbuf_jumbo_pagesize"
323 #define MBUF_JUMBO9_MEM_NAME "mbuf_jumbo_9k"
324 #define MBUF_JUMBO16_MEM_NAME "mbuf_jumbo_16k"
325 #define MBUF_TAG_MEM_NAME "mbuf_tag"
326 #define MBUF_EXTREFCNT_MEM_NAME "mbuf_ext_refcnt"
327
328 #ifdef _KERNEL
329
330 #ifdef WITNESS
331 #define MBUF_CHECKSLEEP(how) do { \
332 if (how == M_WAITOK) \
333 WITNESS_WARN(WARN_GIANTOK | WARN_SLEEPOK, NULL, \
334 "Sleeping in \"%s\"", __func__); \
335 } while (0)
336 #else
337 #define MBUF_CHECKSLEEP(how)
338 #endif
339
340 /*
341 * Network buffer allocation API
342 *
343 * The rest of it is defined in kern/kern_mbuf.c
344 */
345
346 extern uma_zone_t zone_mbuf;
347 extern uma_zone_t zone_clust;
348 extern uma_zone_t zone_pack;
349 extern uma_zone_t zone_jumbop;
350 extern uma_zone_t zone_jumbo9;
351 extern uma_zone_t zone_jumbo16;
352 extern uma_zone_t zone_ext_refcnt;
353
354 static __inline struct mbuf *m_getcl(int how, short type, int flags);
355 static __inline struct mbuf *m_get(int how, short type);
356 static __inline struct mbuf *m_gethdr(int how, short type);
357 static __inline struct mbuf *m_getjcl(int how, short type, int flags,
358 int size);
359 static __inline struct mbuf *m_getclr(int how, short type); /* XXX */
360 static __inline struct mbuf *m_free(struct mbuf *m);
361 static __inline void m_clget(struct mbuf *m, int how);
362 static __inline void *m_cljget(struct mbuf *m, int how, int size);
363 static __inline void m_chtype(struct mbuf *m, short new_type);
364 void mb_free_ext(struct mbuf *);
365 static __inline struct mbuf *m_last(struct mbuf *m);
366
367 static __inline int
368 m_gettype(int size)
369 {
370 int type;
371
372 switch (size) {
373 case MSIZE:
374 type = EXT_MBUF;
375 break;
376 case MCLBYTES:
377 type = EXT_CLUSTER;
378 break;
379 #if MJUMPAGESIZE != MCLBYTES
380 case MJUMPAGESIZE:
381 type = EXT_JUMBOP;
382 break;
383 #endif
384 case MJUM9BYTES:
385 type = EXT_JUMBO9;
386 break;
387 case MJUM16BYTES:
388 type = EXT_JUMBO16;
389 break;
390 default:
391 panic("%s: m_getjcl: invalid cluster size", __func__);
392 }
393
394 return (type);
395 }
396
397 static __inline uma_zone_t
398 m_getzone(int size)
399 {
400 uma_zone_t zone;
401
402 switch (size) {
403 case MSIZE:
404 zone = zone_mbuf;
405 break;
406 case MCLBYTES:
407 zone = zone_clust;
408 break;
409 #if MJUMPAGESIZE != MCLBYTES
410 case MJUMPAGESIZE:
411 zone = zone_jumbop;
412 break;
413 #endif
414 case MJUM9BYTES:
415 zone = zone_jumbo9;
416 break;
417 case MJUM16BYTES:
418 zone = zone_jumbo16;
419 break;
420 default:
421 panic("%s: m_getjcl: invalid cluster type", __func__);
422 }
423
424 return (zone);
425 }
426
427 static __inline struct mbuf *
428 m_get(int how, short type)
429 {
430 struct mb_args args;
431
432 args.flags = 0;
433 args.type = type;
434 return ((struct mbuf *)(uma_zalloc_arg(zone_mbuf, &args, how)));
435 }
436
437 /*
438 * XXX This should be deprecated, very little use.
439 */
440 static __inline struct mbuf *
441 m_getclr(int how, short type)
442 {
443 struct mbuf *m;
444 struct mb_args args;
445
446 args.flags = 0;
447 args.type = type;
448 m = uma_zalloc_arg(zone_mbuf, &args, how);
449 if (m != NULL)
450 bzero(m->m_data, MLEN);
451 return (m);
452 }
453
454 static __inline struct mbuf *
455 m_gethdr(int how, short type)
456 {
457 struct mb_args args;
458
459 args.flags = M_PKTHDR;
460 args.type = type;
461 return ((struct mbuf *)(uma_zalloc_arg(zone_mbuf, &args, how)));
462 }
463
464 static __inline struct mbuf *
465 m_getcl(int how, short type, int flags)
466 {
467 struct mb_args args;
468
469 args.flags = flags;
470 args.type = type;
471 return ((struct mbuf *)(uma_zalloc_arg(zone_pack, &args, how)));
472 }
473
474 /*
475 * m_getjcl() returns an mbuf with a cluster of the specified size attached.
476 * For size it takes MCLBYTES, MJUMPAGESIZE, MJUM9BYTES, MJUM16BYTES.
477 *
478 * XXX: This is rather large, should be real function maybe.
479 */
480 static __inline struct mbuf *
481 m_getjcl(int how, short type, int flags, int size)
482 {
483 struct mb_args args;
484 struct mbuf *m, *n;
485 uma_zone_t zone;
486
487 if (size == MCLBYTES)
488 return m_getcl(how, type, flags);
489
490 args.flags = flags;
491 args.type = type;
492
493 m = uma_zalloc_arg(zone_mbuf, &args, how);
494 if (m == NULL)
495 return (NULL);
496
497 zone = m_getzone(size);
498 n = uma_zalloc_arg(zone, m, how);
499 if (n == NULL) {
500 uma_zfree(zone_mbuf, m);
501 return (NULL);
502 }
503 return (m);
504 }
505
506 static __inline void
507 m_free_fast(struct mbuf *m)
508 {
509 KASSERT(SLIST_EMPTY(&m->m_pkthdr.tags), ("doing fast free of mbuf with tags"));
510
511 uma_zfree_arg(zone_mbuf, m, (void *)MB_NOTAGS);
512 }
513
514 static __inline struct mbuf *
515 m_free(struct mbuf *m)
516 {
517 struct mbuf *n = m->m_next;
518
519 if (m->m_flags & M_EXT)
520 mb_free_ext(m);
521 else if ((m->m_flags & M_NOFREE) == 0)
522 uma_zfree(zone_mbuf, m);
523 return (n);
524 }
525
526 static __inline void
527 m_clget(struct mbuf *m, int how)
528 {
529
530 if (m->m_flags & M_EXT)
531 printf("%s: %p mbuf already has cluster\n", __func__, m);
532 m->m_ext.ext_buf = (char *)NULL;
533 uma_zalloc_arg(zone_clust, m, how);
534 /*
535 * On a cluster allocation failure, drain the packet zone and retry,
536 * we might be able to loosen a few clusters up on the drain.
537 */
538 if ((how & M_NOWAIT) && (m->m_ext.ext_buf == NULL)) {
539 zone_drain(zone_pack);
540 uma_zalloc_arg(zone_clust, m, how);
541 }
542 }
543
544 /*
545 * m_cljget() is different from m_clget() as it can allocate clusters without
546 * attaching them to an mbuf. In that case the return value is the pointer
547 * to the cluster of the requested size. If an mbuf was specified, it gets
548 * the cluster attached to it and the return value can be safely ignored.
549 * For size it takes MCLBYTES, MJUMPAGESIZE, MJUM9BYTES, MJUM16BYTES.
550 */
551 static __inline void *
552 m_cljget(struct mbuf *m, int how, int size)
553 {
554 uma_zone_t zone;
555
556 if (m && m->m_flags & M_EXT)
557 printf("%s: %p mbuf already has cluster\n", __func__, m);
558 if (m != NULL)
559 m->m_ext.ext_buf = NULL;
560
561 zone = m_getzone(size);
562 return (uma_zalloc_arg(zone, m, how));
563 }
564
565 static __inline void
566 m_cljset(struct mbuf *m, void *cl, int type)
567 {
568 uma_zone_t zone;
569 int size;
570
571 switch (type) {
572 case EXT_CLUSTER:
573 size = MCLBYTES;
574 zone = zone_clust;
575 break;
576 #if MJUMPAGESIZE != MCLBYTES
577 case EXT_JUMBOP:
578 size = MJUMPAGESIZE;
579 zone = zone_jumbop;
580 break;
581 #endif
582 case EXT_JUMBO9:
583 size = MJUM9BYTES;
584 zone = zone_jumbo9;
585 break;
586 case EXT_JUMBO16:
587 size = MJUM16BYTES;
588 zone = zone_jumbo16;
589 break;
590 default:
591 panic("unknown cluster type");
592 break;
593 }
594
595 m->m_data = m->m_ext.ext_buf = cl;
596 m->m_ext.ext_free = m->m_ext.ext_args = NULL;
597 m->m_ext.ext_size = size;
598 m->m_ext.ext_type = type;
599 m->m_ext.ref_cnt = uma_find_refcnt(zone, cl);
600 m->m_flags |= M_EXT;
601
602 }
603
604 static __inline void
605 m_chtype(struct mbuf *m, short new_type)
606 {
607
608 m->m_type = new_type;
609 }
610
611 static __inline struct mbuf *
612 m_last(struct mbuf *m)
613 {
614
615 while (m->m_next)
616 m = m->m_next;
617 return (m);
618 }
619
620 /*
621 * mbuf, cluster, and external object allocation macros (for compatibility
622 * purposes).
623 */
624 #define M_MOVE_PKTHDR(to, from) m_move_pkthdr((to), (from))
625 #define MGET(m, how, type) ((m) = m_get((how), (type)))
626 #define MGETHDR(m, how, type) ((m) = m_gethdr((how), (type)))
627 #define MCLGET(m, how) m_clget((m), (how))
628 #define MEXTADD(m, buf, size, free, args, flags, type) \
629 m_extadd((m), (caddr_t)(buf), (size), (free), (args), (flags), (type))
630 #define m_getm(m, len, how, type) \
631 m_getm2((m), (len), (how), (type), M_PKTHDR)
632
633 /*
634 * Evaluate TRUE if it's safe to write to the mbuf m's data region (this can
635 * be both the local data payload, or an external buffer area, depending on
636 * whether M_EXT is set).
637 */
638 #define M_WRITABLE(m) (!((m)->m_flags & M_RDONLY) && \
639 (!(((m)->m_flags & M_EXT)) || \
640 (*((m)->m_ext.ref_cnt) == 1)) ) \
641
642 /* Check if the supplied mbuf has a packet header, or else panic. */
643 #define M_ASSERTPKTHDR(m) \
644 KASSERT(m != NULL && m->m_flags & M_PKTHDR, \
645 ("%s: no mbuf packet header!", __func__))
646
647 /*
648 * Ensure that the supplied mbuf is a valid, non-free mbuf.
649 *
650 * XXX: Broken at the moment. Need some UMA magic to make it work again.
651 */
652 #define M_ASSERTVALID(m) \
653 KASSERT((((struct mbuf *)m)->m_flags & 0) == 0, \
654 ("%s: attempted use of a free mbuf!", __func__))
655
656 /*
657 * Set the m_data pointer of a newly-allocated mbuf (m_get/MGET) to place an
658 * object of the specified size at the end of the mbuf, longword aligned.
659 */
660 #define M_ALIGN(m, len) do { \
661 KASSERT(!((m)->m_flags & (M_PKTHDR|M_EXT)), \
662 ("%s: M_ALIGN not normal mbuf", __func__)); \
663 KASSERT((m)->m_data == (m)->m_dat, \
664 ("%s: M_ALIGN not a virgin mbuf", __func__)); \
665 (m)->m_data += (MLEN - (len)) & ~(sizeof(long) - 1); \
666 } while (0)
667
668 /*
669 * As above, for mbufs allocated with m_gethdr/MGETHDR or initialized by
670 * M_DUP/MOVE_PKTHDR.
671 */
672 #define MH_ALIGN(m, len) do { \
673 KASSERT((m)->m_flags & M_PKTHDR && !((m)->m_flags & M_EXT), \
674 ("%s: MH_ALIGN not PKTHDR mbuf", __func__)); \
675 KASSERT((m)->m_data == (m)->m_pktdat, \
676 ("%s: MH_ALIGN not a virgin mbuf", __func__)); \
677 (m)->m_data += (MHLEN - (len)) & ~(sizeof(long) - 1); \
678 } while (0)
679
680 /*
681 * Compute the amount of space available before the current start of data in
682 * an mbuf.
683 *
684 * The M_WRITABLE() is a temporary, conservative safety measure: the burden
685 * of checking writability of the mbuf data area rests solely with the caller.
686 */
687 #define M_LEADINGSPACE(m) \
688 ((m)->m_flags & M_EXT ? \
689 (M_WRITABLE(m) ? (m)->m_data - (m)->m_ext.ext_buf : 0): \
690 (m)->m_flags & M_PKTHDR ? (m)->m_data - (m)->m_pktdat : \
691 (m)->m_data - (m)->m_dat)
692
693 /*
694 * Compute the amount of space available after the end of data in an mbuf.
695 *
696 * The M_WRITABLE() is a temporary, conservative safety measure: the burden
697 * of checking writability of the mbuf data area rests solely with the caller.
698 */
699 #define M_TRAILINGSPACE(m) \
700 ((m)->m_flags & M_EXT ? \
701 (M_WRITABLE(m) ? (m)->m_ext.ext_buf + (m)->m_ext.ext_size \
702 - ((m)->m_data + (m)->m_len) : 0) : \
703 &(m)->m_dat[MLEN] - ((m)->m_data + (m)->m_len))
704
705 /*
706 * Arrange to prepend space of size plen to mbuf m. If a new mbuf must be
707 * allocated, how specifies whether to wait. If the allocation fails, the
708 * original mbuf chain is freed and m is set to NULL.
709 */
710 #define M_PREPEND(m, plen, how) do { \
711 struct mbuf **_mmp = &(m); \
712 struct mbuf *_mm = *_mmp; \
713 int _mplen = (plen); \
714 int __mhow = (how); \
715 \
716 MBUF_CHECKSLEEP(how); \
717 if (M_LEADINGSPACE(_mm) >= _mplen) { \
718 _mm->m_data -= _mplen; \
719 _mm->m_len += _mplen; \
720 } else \
721 _mm = m_prepend(_mm, _mplen, __mhow); \
722 if (_mm != NULL && _mm->m_flags & M_PKTHDR) \
723 _mm->m_pkthdr.len += _mplen; \
724 *_mmp = _mm; \
725 } while (0)
726
727 /*
728 * Change mbuf to new type. This is a relatively expensive operation and
729 * should be avoided.
730 */
731 #define MCHTYPE(m, t) m_chtype((m), (t))
732
733 /* Length to m_copy to copy all. */
734 #define M_COPYALL 1000000000
735
736 /* Compatibility with 4.3. */
737 #define m_copy(m, o, l) m_copym((m), (o), (l), M_DONTWAIT)
738
739 extern int max_datalen; /* MHLEN - max_hdr */
740 extern int max_hdr; /* Largest link + protocol header */
741 extern int max_linkhdr; /* Largest link-level header */
742 extern int max_protohdr; /* Largest protocol header */
743 extern struct mbstat mbstat; /* General mbuf stats/infos */
744 extern int nmbclusters; /* Maximum number of clusters */
745
746 struct uio;
747
748 void m_adj(struct mbuf *, int);
749 void m_align(struct mbuf *, int);
750 int m_apply(struct mbuf *, int, int,
751 int (*)(void *, void *, u_int), void *);
752 int m_append(struct mbuf *, int, c_caddr_t);
753 void m_cat(struct mbuf *, struct mbuf *);
754 void m_extadd(struct mbuf *, caddr_t, u_int,
755 void (*)(void *, void *), void *, int, int);
756 struct mbuf *m_collapse(struct mbuf *, int, int);
757 void m_copyback(struct mbuf *, int, int, c_caddr_t);
758 void m_copydata(const struct mbuf *, int, int, caddr_t);
759 struct mbuf *m_copym(struct mbuf *, int, int, int);
760 struct mbuf *m_copymdata(struct mbuf *, struct mbuf *,
761 int, int, int, int);
762 struct mbuf *m_copypacket(struct mbuf *, int);
763 void m_copy_pkthdr(struct mbuf *, struct mbuf *);
764 struct mbuf *m_copyup(struct mbuf *n, int len, int dstoff);
765 struct mbuf *m_defrag(struct mbuf *, int);
766 void m_demote(struct mbuf *, int);
767 struct mbuf *m_devget(char *, int, int, struct ifnet *,
768 void (*)(char *, caddr_t, u_int));
769 struct mbuf *m_dup(struct mbuf *, int);
770 int m_dup_pkthdr(struct mbuf *, struct mbuf *, int);
771 u_int m_fixhdr(struct mbuf *);
772 struct mbuf *m_fragment(struct mbuf *, int, int);
773 void m_freem(struct mbuf *);
774 struct mbuf *m_getm2(struct mbuf *, int, int, short, int);
775 struct mbuf *m_getptr(struct mbuf *, int, int *);
776 u_int m_length(struct mbuf *, struct mbuf **);
777 void m_move_pkthdr(struct mbuf *, struct mbuf *);
778 struct mbuf *m_prepend(struct mbuf *, int, int);
779 void m_print(const struct mbuf *, int);
780 struct mbuf *m_pulldown(struct mbuf *, int, int, int *);
781 struct mbuf *m_pullup(struct mbuf *, int);
782 int m_sanity(struct mbuf *, int);
783 struct mbuf *m_split(struct mbuf *, int, int);
784 struct mbuf *m_uiotombuf(struct uio *, int, int, int, int);
785 struct mbuf *m_unshare(struct mbuf *, int how);
786
787 /*-
788 * Network packets may have annotations attached by affixing a list of
789 * "packet tags" to the pkthdr structure. Packet tags are dynamically
790 * allocated semi-opaque data structures that have a fixed header
791 * (struct m_tag) that specifies the size of the memory block and a
792 * <cookie,type> pair that identifies it. The cookie is a 32-bit unique
793 * unsigned value used to identify a module or ABI. By convention this value
794 * is chosen as the date+time that the module is created, expressed as the
795 * number of seconds since the epoch (e.g., using date -u +'%s'). The type
796 * value is an ABI/module-specific value that identifies a particular
797 * annotation and is private to the module. For compatibility with systems
798 * like OpenBSD that define packet tags w/o an ABI/module cookie, the value
799 * PACKET_ABI_COMPAT is used to implement m_tag_get and m_tag_find
800 * compatibility shim functions and several tag types are defined below.
801 * Users that do not require compatibility should use a private cookie value
802 * so that packet tag-related definitions can be maintained privately.
803 *
804 * Note that the packet tag returned by m_tag_alloc has the default memory
805 * alignment implemented by malloc. To reference private data one can use a
806 * construct like:
807 *
808 * struct m_tag *mtag = m_tag_alloc(...);
809 * struct foo *p = (struct foo *)(mtag+1);
810 *
811 * if the alignment of struct m_tag is sufficient for referencing members of
812 * struct foo. Otherwise it is necessary to embed struct m_tag within the
813 * private data structure to insure proper alignment; e.g.,
814 *
815 * struct foo {
816 * struct m_tag tag;
817 * ...
818 * };
819 * struct foo *p = (struct foo *) m_tag_alloc(...);
820 * struct m_tag *mtag = &p->tag;
821 */
822
823 /*
824 * Persistent tags stay with an mbuf until the mbuf is reclaimed. Otherwise
825 * tags are expected to ``vanish'' when they pass through a network
826 * interface. For most interfaces this happens normally as the tags are
827 * reclaimed when the mbuf is free'd. However in some special cases
828 * reclaiming must be done manually. An example is packets that pass through
829 * the loopback interface. Also, one must be careful to do this when
830 * ``turning around'' packets (e.g., icmp_reflect).
831 *
832 * To mark a tag persistent bit-or this flag in when defining the tag id.
833 * The tag will then be treated as described above.
834 */
835 #define MTAG_PERSISTENT 0x800
836
837 #define PACKET_TAG_NONE 0 /* Nadda */
838
839 /* Packet tags for use with PACKET_ABI_COMPAT. */
840 #define PACKET_TAG_IPSEC_IN_DONE 1 /* IPsec applied, in */
841 #define PACKET_TAG_IPSEC_OUT_DONE 2 /* IPsec applied, out */
842 #define PACKET_TAG_IPSEC_IN_CRYPTO_DONE 3 /* NIC IPsec crypto done */
843 #define PACKET_TAG_IPSEC_OUT_CRYPTO_NEEDED 4 /* NIC IPsec crypto req'ed */
844 #define PACKET_TAG_IPSEC_IN_COULD_DO_CRYPTO 5 /* NIC notifies IPsec */
845 #define PACKET_TAG_IPSEC_PENDING_TDB 6 /* Reminder to do IPsec */
846 #define PACKET_TAG_BRIDGE 7 /* Bridge processing done */
847 #define PACKET_TAG_GIF 8 /* GIF processing done */
848 #define PACKET_TAG_GRE 9 /* GRE processing done */
849 #define PACKET_TAG_IN_PACKET_CHECKSUM 10 /* NIC checksumming done */
850 #define PACKET_TAG_ENCAP 11 /* Encap. processing */
851 #define PACKET_TAG_IPSEC_SOCKET 12 /* IPSEC socket ref */
852 #define PACKET_TAG_IPSEC_HISTORY 13 /* IPSEC history */
853 #define PACKET_TAG_IPV6_INPUT 14 /* IPV6 input processing */
854 #define PACKET_TAG_DUMMYNET 15 /* dummynet info */
855 #define PACKET_TAG_DIVERT 17 /* divert info */
856 #define PACKET_TAG_IPFORWARD 18 /* ipforward info */
857 #define PACKET_TAG_MACLABEL (19 | MTAG_PERSISTENT) /* MAC label */
858 #define PACKET_TAG_PF 21 /* PF + ALTQ information */
859 #define PACKET_TAG_RTSOCKFAM 25 /* rtsock sa family */
860 #define PACKET_TAG_IPOPTIONS 27 /* Saved IP options */
861 #define PACKET_TAG_CARP 28 /* CARP info */
862
863 /* Specific cookies and tags. */
864
865 /* Packet tag routines. */
866 struct m_tag *m_tag_alloc(u_int32_t, int, int, int);
867 void m_tag_delete(struct mbuf *, struct m_tag *);
868 void m_tag_delete_chain(struct mbuf *, struct m_tag *);
869 void m_tag_free_default(struct m_tag *);
870 struct m_tag *m_tag_locate(struct mbuf *, u_int32_t, int, struct m_tag *);
871 struct m_tag *m_tag_copy(struct m_tag *, int);
872 int m_tag_copy_chain(struct mbuf *, struct mbuf *, int);
873 void m_tag_delete_nonpersistent(struct mbuf *);
874
875 /*
876 * Initialize the list of tags associated with an mbuf.
877 */
878 static __inline void
879 m_tag_init(struct mbuf *m)
880 {
881
882 SLIST_INIT(&m->m_pkthdr.tags);
883 }
884
885 /*
886 * Set up the contents of a tag. Note that this does not fill in the free
887 * method; the caller is expected to do that.
888 *
889 * XXX probably should be called m_tag_init, but that was already taken.
890 */
891 static __inline void
892 m_tag_setup(struct m_tag *t, u_int32_t cookie, int type, int len)
893 {
894
895 t->m_tag_id = type;
896 t->m_tag_len = len;
897 t->m_tag_cookie = cookie;
898 }
899
900 /*
901 * Reclaim resources associated with a tag.
902 */
903 static __inline void
904 m_tag_free(struct m_tag *t)
905 {
906
907 (*t->m_tag_free)(t);
908 }
909
910 /*
911 * Return the first tag associated with an mbuf.
912 */
913 static __inline struct m_tag *
914 m_tag_first(struct mbuf *m)
915 {
916
917 return (SLIST_FIRST(&m->m_pkthdr.tags));
918 }
919
920 /*
921 * Return the next tag in the list of tags associated with an mbuf.
922 */
923 static __inline struct m_tag *
924 m_tag_next(struct mbuf *m, struct m_tag *t)
925 {
926
927 return (SLIST_NEXT(t, m_tag_link));
928 }
929
930 /*
931 * Prepend a tag to the list of tags associated with an mbuf.
932 */
933 static __inline void
934 m_tag_prepend(struct mbuf *m, struct m_tag *t)
935 {
936
937 SLIST_INSERT_HEAD(&m->m_pkthdr.tags, t, m_tag_link);
938 }
939
940 /*
941 * Unlink a tag from the list of tags associated with an mbuf.
942 */
943 static __inline void
944 m_tag_unlink(struct mbuf *m, struct m_tag *t)
945 {
946
947 SLIST_REMOVE(&m->m_pkthdr.tags, t, m_tag, m_tag_link);
948 }
949
950 /* These are for OpenBSD compatibility. */
951 #define MTAG_ABI_COMPAT 0 /* compatibility ABI */
952
953 static __inline struct m_tag *
954 m_tag_get(int type, int length, int wait)
955 {
956 return (m_tag_alloc(MTAG_ABI_COMPAT, type, length, wait));
957 }
958
959 static __inline struct m_tag *
960 m_tag_find(struct mbuf *m, int type, struct m_tag *start)
961 {
962 return (SLIST_EMPTY(&m->m_pkthdr.tags) ? (struct m_tag *)NULL :
963 m_tag_locate(m, MTAG_ABI_COMPAT, type, start));
964 }
965
966 /* XXX temporary FIB methods probably eventually use tags.*/
967 #define M_FIBSHIFT 28
968 #define M_FIBMASK 0x0F
969
970 /* get the fib from an mbuf and if it is not set, return the default */
971 #define M_GETFIB(_m) \
972 ((((_m)->m_flags & M_FIB) >> M_FIBSHIFT) & M_FIBMASK)
973
974 #define M_SETFIB(_m, _fib) do { \
975 _m->m_flags &= ~M_FIB; \
976 _m->m_flags |= (((_fib) << M_FIBSHIFT) & M_FIB); \
977 } while (0)
978
979 #endif /* _KERNEL */
980
981 #endif /* !_SYS_MBUF_H_ */
Cache object: 3f7f8ba644bc7c4321fe04e93940c056
|