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 * Macro for type conversion: convert mbuf pointer to data pointer of correct
64 * type:
65 *
66 * mtod(m, t) -- Convert mbuf pointer to data pointer of correct type.
67 */
68 #define mtod(m, t) ((t)((m)->m_data))
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 uint32_t flowid; /* packet's 4-tuple system
119 * flow identifier
120 */
121 /* variables for hardware checksum */
122 int csum_flags; /* flags regarding checksum */
123 int csum_data; /* data field used by csum routines */
124 u_int16_t tso_segsz; /* TSO segment size */
125 union {
126 u_int16_t vt_vtag; /* Ethernet 802.1p+q vlan tag */
127 u_int16_t vt_nrecs; /* # of IGMPv3 records in this chain */
128 } PH_vt;
129 SLIST_HEAD(packet_tags, m_tag) tags; /* list of packet tags */
130 };
131 #define ether_vtag PH_vt.vt_vtag
132
133 /*
134 * Description of external storage mapped into mbuf; valid only if M_EXT is
135 * set.
136 */
137 struct m_ext {
138 caddr_t ext_buf; /* start of buffer */
139 void (*ext_free) /* free routine if not the usual */
140 (void *, void *);
141 void *ext_arg1; /* optional argument pointer */
142 void *ext_arg2; /* optional argument pointer */
143 u_int ext_size; /* size of buffer, for ext_free */
144 volatile u_int *ref_cnt; /* pointer to ref count info */
145 int ext_type; /* type of external storage */
146 };
147
148 /*
149 * The core of the mbuf object along with some shortcut defines for practical
150 * purposes.
151 */
152 struct mbuf {
153 struct m_hdr m_hdr;
154 union {
155 struct {
156 struct pkthdr MH_pkthdr; /* M_PKTHDR set */
157 union {
158 struct m_ext MH_ext; /* M_EXT set */
159 char MH_databuf[MHLEN];
160 } MH_dat;
161 } MH;
162 char M_databuf[MLEN]; /* !M_PKTHDR, !M_EXT */
163 } M_dat;
164 };
165 #define m_next m_hdr.mh_next
166 #define m_len m_hdr.mh_len
167 #define m_data m_hdr.mh_data
168 #define m_type m_hdr.mh_type
169 #define m_flags m_hdr.mh_flags
170 #define m_nextpkt m_hdr.mh_nextpkt
171 #define m_act m_nextpkt
172 #define m_pkthdr M_dat.MH.MH_pkthdr
173 #define m_ext M_dat.MH.MH_dat.MH_ext
174 #define m_pktdat M_dat.MH.MH_dat.MH_databuf
175 #define m_dat M_dat.M_databuf
176
177 /*
178 * mbuf flags.
179 */
180 #define M_EXT 0x00000001 /* has associated external storage */
181 #define M_PKTHDR 0x00000002 /* start of record */
182 #define M_EOR 0x00000004 /* end of record */
183 #define M_RDONLY 0x00000008 /* associated data is marked read-only */
184 #define M_PROTO1 0x00000010 /* protocol-specific */
185 #define M_PROTO2 0x00000020 /* protocol-specific */
186 #define M_PROTO3 0x00000040 /* protocol-specific */
187 #define M_PROTO4 0x00000080 /* protocol-specific */
188 #define M_PROTO5 0x00000100 /* protocol-specific */
189 #define M_BCAST 0x00000200 /* send/received as link-level broadcast */
190 #define M_MCAST 0x00000400 /* send/received as link-level multicast */
191 #define M_FRAG 0x00000800 /* packet is a fragment of a larger packet */
192 #define M_FIRSTFRAG 0x00001000 /* packet is first fragment */
193 #define M_LASTFRAG 0x00002000 /* packet is last fragment */
194 #define M_SKIP_FIREWALL 0x00004000 /* skip firewall processing */
195 #define M_FREELIST 0x00008000 /* mbuf is on the free list */
196 #define M_VLANTAG 0x00010000 /* ether_vtag is valid */
197 #define M_PROMISC 0x00020000 /* packet was not for us */
198 #define M_NOFREE 0x00040000 /* do not free mbuf, embedded in cluster */
199 #define M_PROTO6 0x00080000 /* protocol-specific */
200 #define M_PROTO7 0x00100000 /* protocol-specific */
201 #define M_PROTO8 0x00200000 /* protocol-specific */
202 #define M_FLOWID 0x00400000 /* flowid is valid */
203 /*
204 * For RELENG_{6,7} steal these flags for limited multiple routing table
205 * support. In RELENG_8 and beyond, use just one flag and a tag.
206 */
207 #define M_FIB 0xF0000000 /* steal some bits to store fib number. */
208
209 #define M_NOTIFICATION M_PROTO5 /* SCTP notification */
210
211 /*
212 * Flags to purge when crossing layers.
213 */
214 #define M_PROTOFLAGS \
215 (M_PROTO1|M_PROTO2|M_PROTO3|M_PROTO4|M_PROTO5|M_PROTO6|M_PROTO7|M_PROTO8)
216
217 /*
218 * Flags preserved when copying m_pkthdr.
219 */
220 #define M_COPYFLAGS \
221 (M_PKTHDR|M_EOR|M_RDONLY|M_PROTOFLAGS|M_SKIP_FIREWALL|M_BCAST|M_MCAST|\
222 M_FRAG|M_FIRSTFRAG|M_LASTFRAG|M_VLANTAG|M_PROMISC|M_FIB)
223
224 /*
225 * External buffer types: identify ext_buf type.
226 */
227 #define EXT_CLUSTER 1 /* mbuf cluster */
228 #define EXT_SFBUF 2 /* sendfile(2)'s sf_bufs */
229 #define EXT_JUMBOP 3 /* jumbo cluster 4096 bytes */
230 #define EXT_JUMBO9 4 /* jumbo cluster 9216 bytes */
231 #define EXT_JUMBO16 5 /* jumbo cluster 16184 bytes */
232 #define EXT_PACKET 6 /* mbuf+cluster from packet zone */
233 #define EXT_MBUF 7 /* external mbuf reference (M_IOVEC) */
234 #define EXT_NET_DRV 100 /* custom ext_buf provided by net driver(s) */
235 #define EXT_MOD_TYPE 200 /* custom module's ext_buf type */
236 #define EXT_DISPOSABLE 300 /* can throw this buffer away w/page flipping */
237 #define EXT_EXTREF 400 /* has externally maintained ref_cnt ptr */
238
239 /*
240 * Flags indicating hw checksum support and sw checksum requirements. This
241 * field can be directly tested against if_data.ifi_hwassist.
242 */
243 #define CSUM_IP 0x0001 /* will csum IP */
244 #define CSUM_TCP 0x0002 /* will csum TCP */
245 #define CSUM_UDP 0x0004 /* will csum UDP */
246 #define CSUM_IP_FRAGS 0x0008 /* will csum IP fragments */
247 #define CSUM_FRAGMENT 0x0010 /* will do IP fragmentation */
248 #define CSUM_TSO 0x0020 /* will do TSO */
249 #define CSUM_SCTP 0x0040 /* will csum SCTP */
250
251 #define CSUM_IP_CHECKED 0x0100 /* did csum IP */
252 #define CSUM_IP_VALID 0x0200 /* ... the csum is valid */
253 #define CSUM_DATA_VALID 0x0400 /* csum_data field is valid */
254 #define CSUM_PSEUDO_HDR 0x0800 /* csum_data has pseudo hdr */
255 #define CSUM_SCTP_VALID 0x1000 /* SCTP checksum is valid */
256
257 #define CSUM_DELAY_DATA (CSUM_TCP | CSUM_UDP)
258 #define CSUM_DELAY_IP (CSUM_IP) /* XXX add ipv6 here too? */
259
260 /*
261 * mbuf types.
262 */
263 #define MT_NOTMBUF 0 /* USED INTERNALLY ONLY! Object is not mbuf */
264 #define MT_DATA 1 /* dynamic (data) allocation */
265 #define MT_HEADER MT_DATA /* packet header, use M_PKTHDR instead */
266 #define MT_SONAME 8 /* socket name */
267 #define MT_CONTROL 14 /* extra-data protocol message */
268 #define MT_OOBDATA 15 /* expedited data */
269 #define MT_NTYPES 16 /* number of mbuf types for mbtypes[] */
270
271 #define MT_NOINIT 255 /* Not a type but a flag to allocate
272 a non-initialized mbuf */
273
274 #define MB_NOTAGS 0x1UL /* no tags attached to mbuf */
275
276 /*
277 * General mbuf allocator statistics structure.
278 *
279 * Many of these statistics are no longer used; we instead track many
280 * allocator statistics through UMA's built in statistics mechanism.
281 */
282 struct mbstat {
283 u_long m_mbufs; /* XXX */
284 u_long m_mclusts; /* XXX */
285
286 u_long m_drain; /* times drained protocols for space */
287 u_long m_mcfail; /* XXX: times m_copym failed */
288 u_long m_mpfail; /* XXX: times m_pullup failed */
289 u_long m_msize; /* length of an mbuf */
290 u_long m_mclbytes; /* length of an mbuf cluster */
291 u_long m_minclsize; /* min length of data to allocate a cluster */
292 u_long m_mlen; /* length of data in an mbuf */
293 u_long m_mhlen; /* length of data in a header mbuf */
294
295 /* Number of mbtypes (gives # elems in mbtypes[] array) */
296 short m_numtypes;
297
298 /* XXX: Sendfile stats should eventually move to their own struct */
299 u_long sf_iocnt; /* times sendfile had to do disk I/O */
300 u_long sf_allocfail; /* times sfbuf allocation failed */
301 u_long sf_allocwait; /* times sfbuf allocation had to wait */
302 };
303
304 /*
305 * Flags specifying how an allocation should be made.
306 *
307 * The flag to use is as follows:
308 * - M_DONTWAIT or M_NOWAIT from an interrupt handler to not block allocation.
309 * - M_WAIT or M_WAITOK from wherever it is safe to block.
310 *
311 * M_DONTWAIT/M_NOWAIT means that we will not block the thread explicitly and
312 * if we cannot allocate immediately we may return NULL, whereas
313 * M_WAIT/M_WAITOK means that if we cannot allocate resources we
314 * will block until they are available, and thus never return NULL.
315 *
316 * XXX Eventually just phase this out to use M_WAITOK/M_NOWAIT.
317 */
318 #define MBTOM(how) (how)
319 #define M_DONTWAIT M_NOWAIT
320 #define M_TRYWAIT M_WAITOK
321 #define M_WAIT M_WAITOK
322
323 /*
324 * String names of mbuf-related UMA(9) and malloc(9) types. Exposed to
325 * !_KERNEL so that monitoring tools can look up the zones with
326 * libmemstat(3).
327 */
328 #define MBUF_MEM_NAME "mbuf"
329 #define MBUF_CLUSTER_MEM_NAME "mbuf_cluster"
330 #define MBUF_PACKET_MEM_NAME "mbuf_packet"
331 #define MBUF_JUMBOP_MEM_NAME "mbuf_jumbo_page"
332 #define MBUF_JUMBO9_MEM_NAME "mbuf_jumbo_9k"
333 #define MBUF_JUMBO16_MEM_NAME "mbuf_jumbo_16k"
334 #define MBUF_TAG_MEM_NAME "mbuf_tag"
335 #define MBUF_EXTREFCNT_MEM_NAME "mbuf_ext_refcnt"
336
337 #ifdef _KERNEL
338
339 #ifdef WITNESS
340 #define MBUF_CHECKSLEEP(how) do { \
341 if (how == M_WAITOK) \
342 WITNESS_WARN(WARN_GIANTOK | WARN_SLEEPOK, NULL, \
343 "Sleeping in \"%s\"", __func__); \
344 } while (0)
345 #else
346 #define MBUF_CHECKSLEEP(how)
347 #endif
348
349 /*
350 * Network buffer allocation API
351 *
352 * The rest of it is defined in kern/kern_mbuf.c
353 */
354
355 extern uma_zone_t zone_mbuf;
356 extern uma_zone_t zone_clust;
357 extern uma_zone_t zone_pack;
358 extern uma_zone_t zone_jumbop;
359 extern uma_zone_t zone_jumbo9;
360 extern uma_zone_t zone_jumbo16;
361 extern uma_zone_t zone_ext_refcnt;
362
363 static __inline struct mbuf *m_getcl(int how, short type, int flags);
364 static __inline struct mbuf *m_get(int how, short type);
365 static __inline struct mbuf *m_gethdr(int how, short type);
366 static __inline struct mbuf *m_getjcl(int how, short type, int flags,
367 int size);
368 static __inline struct mbuf *m_getclr(int how, short type); /* XXX */
369 static __inline int m_init(struct mbuf *m, uma_zone_t zone,
370 int size, int how, short type, int flags);
371 static __inline struct mbuf *m_free(struct mbuf *m);
372 static __inline void m_clget(struct mbuf *m, int how);
373 static __inline void *m_cljget(struct mbuf *m, int how, int size);
374 static __inline void m_chtype(struct mbuf *m, short new_type);
375 void mb_free_ext(struct mbuf *);
376 static __inline struct mbuf *m_last(struct mbuf *m);
377 int m_pkthdr_init(struct mbuf *m, int how);
378
379 static __inline int
380 m_gettype(int size)
381 {
382 int type;
383
384 switch (size) {
385 case MSIZE:
386 type = EXT_MBUF;
387 break;
388 case MCLBYTES:
389 type = EXT_CLUSTER;
390 break;
391 #if MJUMPAGESIZE != MCLBYTES
392 case MJUMPAGESIZE:
393 type = EXT_JUMBOP;
394 break;
395 #endif
396 case MJUM9BYTES:
397 type = EXT_JUMBO9;
398 break;
399 case MJUM16BYTES:
400 type = EXT_JUMBO16;
401 break;
402 default:
403 panic("%s: m_getjcl: invalid cluster size", __func__);
404 }
405
406 return (type);
407 }
408
409 /*
410 * Associated an external reference counted buffer with an mbuf.
411 */
412 static __inline void
413 m_extaddref(struct mbuf *m, caddr_t buf, u_int size, u_int *ref_cnt,
414 void (*freef)(void *, void *), void *arg1, void *arg2)
415 {
416
417 KASSERT(ref_cnt != NULL, ("%s: ref_cnt not provided", __func__));
418
419 atomic_add_int(ref_cnt, 1);
420 m->m_flags |= M_EXT;
421 m->m_ext.ext_buf = buf;
422 m->m_ext.ref_cnt = ref_cnt;
423 m->m_data = m->m_ext.ext_buf;
424 m->m_ext.ext_size = size;
425 m->m_ext.ext_free = freef;
426 m->m_ext.ext_arg1 = arg1;
427 m->m_ext.ext_arg2 = arg2;
428 m->m_ext.ext_type = EXT_EXTREF;
429 }
430
431 static __inline uma_zone_t
432 m_getzone(int size)
433 {
434 uma_zone_t zone;
435
436 switch (size) {
437 case MSIZE:
438 zone = zone_mbuf;
439 break;
440 case MCLBYTES:
441 zone = zone_clust;
442 break;
443 #if MJUMPAGESIZE != MCLBYTES
444 case MJUMPAGESIZE:
445 zone = zone_jumbop;
446 break;
447 #endif
448 case MJUM9BYTES:
449 zone = zone_jumbo9;
450 break;
451 case MJUM16BYTES:
452 zone = zone_jumbo16;
453 break;
454 default:
455 panic("%s: m_getjcl: invalid cluster type", __func__);
456 }
457
458 return (zone);
459 }
460
461 /*
462 * Initialize an mbuf with linear storage.
463 *
464 * Inline because the consumer text overhead will be roughly the same to
465 * initialize or call a function with this many parameters and M_PKTHDR
466 * should go away with constant propagation for !MGETHDR.
467 */
468 static __inline int
469 m_init(struct mbuf *m, uma_zone_t zone, int size, int how, short type,
470 int flags)
471 {
472 int error;
473
474 m->m_next = NULL;
475 m->m_nextpkt = NULL;
476 m->m_data = m->m_dat;
477 m->m_len = 0;
478 m->m_flags = flags;
479 m->m_type = type;
480 if (flags & M_PKTHDR) {
481 if ((error = m_pkthdr_init(m, how)) != 0)
482 return (error);
483 }
484
485 return (0);
486 }
487
488 static __inline struct mbuf *
489 m_get(int how, short type)
490 {
491 struct mb_args args;
492
493 args.flags = 0;
494 args.type = type;
495 return ((struct mbuf *)(uma_zalloc_arg(zone_mbuf, &args, how)));
496 }
497
498 /*
499 * XXX This should be deprecated, very little use.
500 */
501 static __inline struct mbuf *
502 m_getclr(int how, short type)
503 {
504 struct mbuf *m;
505 struct mb_args args;
506
507 args.flags = 0;
508 args.type = type;
509 m = uma_zalloc_arg(zone_mbuf, &args, how);
510 if (m != NULL)
511 bzero(m->m_data, MLEN);
512 return (m);
513 }
514
515 static __inline struct mbuf *
516 m_gethdr(int how, short type)
517 {
518 struct mb_args args;
519
520 args.flags = M_PKTHDR;
521 args.type = type;
522 return ((struct mbuf *)(uma_zalloc_arg(zone_mbuf, &args, how)));
523 }
524
525 static __inline struct mbuf *
526 m_getcl(int how, short type, int flags)
527 {
528 struct mb_args args;
529
530 args.flags = flags;
531 args.type = type;
532 return ((struct mbuf *)(uma_zalloc_arg(zone_pack, &args, how)));
533 }
534
535 /*
536 * m_getjcl() returns an mbuf with a cluster of the specified size attached.
537 * For size it takes MCLBYTES, MJUMPAGESIZE, MJUM9BYTES, MJUM16BYTES.
538 *
539 * XXX: This is rather large, should be real function maybe.
540 */
541 static __inline struct mbuf *
542 m_getjcl(int how, short type, int flags, int size)
543 {
544 struct mb_args args;
545 struct mbuf *m, *n;
546 uma_zone_t zone;
547
548 if (size == MCLBYTES)
549 return m_getcl(how, type, flags);
550
551 args.flags = flags;
552 args.type = type;
553
554 m = uma_zalloc_arg(zone_mbuf, &args, how);
555 if (m == NULL)
556 return (NULL);
557
558 zone = m_getzone(size);
559 n = uma_zalloc_arg(zone, m, how);
560 if (n == NULL) {
561 uma_zfree(zone_mbuf, m);
562 return (NULL);
563 }
564 return (m);
565 }
566
567 static __inline void
568 m_free_fast(struct mbuf *m)
569 {
570 #ifdef INVARIANTS
571 if (m->m_flags & M_PKTHDR)
572 KASSERT(SLIST_EMPTY(&m->m_pkthdr.tags), ("doing fast free of mbuf with tags"));
573 #endif
574
575 uma_zfree_arg(zone_mbuf, m, (void *)MB_NOTAGS);
576 }
577
578 static __inline struct mbuf *
579 m_free(struct mbuf *m)
580 {
581 struct mbuf *n = m->m_next;
582
583 if (m->m_flags & M_EXT)
584 mb_free_ext(m);
585 else if ((m->m_flags & M_NOFREE) == 0)
586 uma_zfree(zone_mbuf, m);
587 return (n);
588 }
589
590 static __inline void
591 m_clget(struct mbuf *m, int how)
592 {
593
594 if (m->m_flags & M_EXT)
595 printf("%s: %p mbuf already has cluster\n", __func__, m);
596 m->m_ext.ext_buf = (char *)NULL;
597 uma_zalloc_arg(zone_clust, m, how);
598 /*
599 * On a cluster allocation failure, drain the packet zone and retry,
600 * we might be able to loosen a few clusters up on the drain.
601 */
602 if ((how & M_NOWAIT) && (m->m_ext.ext_buf == NULL)) {
603 zone_drain(zone_pack);
604 uma_zalloc_arg(zone_clust, m, how);
605 }
606 }
607
608 /*
609 * m_cljget() is different from m_clget() as it can allocate clusters without
610 * attaching them to an mbuf. In that case the return value is the pointer
611 * to the cluster of the requested size. If an mbuf was specified, it gets
612 * the cluster attached to it and the return value can be safely ignored.
613 * For size it takes MCLBYTES, MJUMPAGESIZE, MJUM9BYTES, MJUM16BYTES.
614 */
615 static __inline void *
616 m_cljget(struct mbuf *m, int how, int size)
617 {
618 uma_zone_t zone;
619
620 if (m && m->m_flags & M_EXT)
621 printf("%s: %p mbuf already has cluster\n", __func__, m);
622 if (m != NULL)
623 m->m_ext.ext_buf = NULL;
624
625 zone = m_getzone(size);
626 return (uma_zalloc_arg(zone, m, how));
627 }
628
629 static __inline void
630 m_cljset(struct mbuf *m, void *cl, int type)
631 {
632 uma_zone_t zone;
633 int size;
634
635 switch (type) {
636 case EXT_CLUSTER:
637 size = MCLBYTES;
638 zone = zone_clust;
639 break;
640 #if MJUMPAGESIZE != MCLBYTES
641 case EXT_JUMBOP:
642 size = MJUMPAGESIZE;
643 zone = zone_jumbop;
644 break;
645 #endif
646 case EXT_JUMBO9:
647 size = MJUM9BYTES;
648 zone = zone_jumbo9;
649 break;
650 case EXT_JUMBO16:
651 size = MJUM16BYTES;
652 zone = zone_jumbo16;
653 break;
654 default:
655 panic("unknown cluster type");
656 break;
657 }
658
659 m->m_data = m->m_ext.ext_buf = cl;
660 m->m_ext.ext_free = m->m_ext.ext_arg1 = m->m_ext.ext_arg2 = NULL;
661 m->m_ext.ext_size = size;
662 m->m_ext.ext_type = type;
663 m->m_ext.ref_cnt = uma_find_refcnt(zone, cl);
664 m->m_flags |= M_EXT;
665
666 }
667
668 static __inline void
669 m_chtype(struct mbuf *m, short new_type)
670 {
671
672 m->m_type = new_type;
673 }
674
675 static __inline struct mbuf *
676 m_last(struct mbuf *m)
677 {
678
679 while (m->m_next)
680 m = m->m_next;
681 return (m);
682 }
683
684 /*
685 * mbuf, cluster, and external object allocation macros (for compatibility
686 * purposes).
687 */
688 #define M_MOVE_PKTHDR(to, from) m_move_pkthdr((to), (from))
689 #define MGET(m, how, type) ((m) = m_get((how), (type)))
690 #define MGETHDR(m, how, type) ((m) = m_gethdr((how), (type)))
691 #define MCLGET(m, how) m_clget((m), (how))
692 #define MEXTADD(m, buf, size, free, arg1, arg2, flags, type) \
693 m_extadd((m), (caddr_t)(buf), (size), (free),(arg1),(arg2),(flags), (type))
694 #define m_getm(m, len, how, type) \
695 m_getm2((m), (len), (how), (type), M_PKTHDR)
696
697 /*
698 * Evaluate TRUE if it's safe to write to the mbuf m's data region (this can
699 * be both the local data payload, or an external buffer area, depending on
700 * whether M_EXT is set).
701 */
702 #define M_WRITABLE(m) (!((m)->m_flags & M_RDONLY) && \
703 (!(((m)->m_flags & M_EXT)) || \
704 (*((m)->m_ext.ref_cnt) == 1)) ) \
705
706 /* Check if the supplied mbuf has a packet header, or else panic. */
707 #define M_ASSERTPKTHDR(m) \
708 KASSERT((m) != NULL && (m)->m_flags & M_PKTHDR, \
709 ("%s: no mbuf packet header!", __func__))
710
711 /*
712 * Ensure that the supplied mbuf is a valid, non-free mbuf.
713 *
714 * XXX: Broken at the moment. Need some UMA magic to make it work again.
715 */
716 #define M_ASSERTVALID(m) \
717 KASSERT((((struct mbuf *)m)->m_flags & 0) == 0, \
718 ("%s: attempted use of a free mbuf!", __func__))
719
720 /*
721 * Set the m_data pointer of a newly-allocated mbuf (m_get/MGET) to place an
722 * object of the specified size at the end of the mbuf, longword aligned.
723 */
724 #define M_ALIGN(m, len) do { \
725 KASSERT(!((m)->m_flags & (M_PKTHDR|M_EXT)), \
726 ("%s: M_ALIGN not normal mbuf", __func__)); \
727 KASSERT((m)->m_data == (m)->m_dat, \
728 ("%s: M_ALIGN not a virgin mbuf", __func__)); \
729 (m)->m_data += (MLEN - (len)) & ~(sizeof(long) - 1); \
730 } while (0)
731
732 /*
733 * As above, for mbufs allocated with m_gethdr/MGETHDR or initialized by
734 * M_DUP/MOVE_PKTHDR.
735 */
736 #define MH_ALIGN(m, len) do { \
737 KASSERT((m)->m_flags & M_PKTHDR && !((m)->m_flags & M_EXT), \
738 ("%s: MH_ALIGN not PKTHDR mbuf", __func__)); \
739 KASSERT((m)->m_data == (m)->m_pktdat, \
740 ("%s: MH_ALIGN not a virgin mbuf", __func__)); \
741 (m)->m_data += (MHLEN - (len)) & ~(sizeof(long) - 1); \
742 } while (0)
743
744 /*
745 * Compute the amount of space available before the current start of data in
746 * an mbuf.
747 *
748 * The M_WRITABLE() is a temporary, conservative safety measure: the burden
749 * of checking writability of the mbuf data area rests solely with the caller.
750 */
751 #define M_LEADINGSPACE(m) \
752 ((m)->m_flags & M_EXT ? \
753 (M_WRITABLE(m) ? (m)->m_data - (m)->m_ext.ext_buf : 0): \
754 (m)->m_flags & M_PKTHDR ? (m)->m_data - (m)->m_pktdat : \
755 (m)->m_data - (m)->m_dat)
756
757 /*
758 * Compute the amount of space available after the end of data in an mbuf.
759 *
760 * The M_WRITABLE() is a temporary, conservative safety measure: the burden
761 * of checking writability of the mbuf data area rests solely with the caller.
762 */
763 #define M_TRAILINGSPACE(m) \
764 ((m)->m_flags & M_EXT ? \
765 (M_WRITABLE(m) ? (m)->m_ext.ext_buf + (m)->m_ext.ext_size \
766 - ((m)->m_data + (m)->m_len) : 0) : \
767 &(m)->m_dat[MLEN] - ((m)->m_data + (m)->m_len))
768
769 /*
770 * Arrange to prepend space of size plen to mbuf m. If a new mbuf must be
771 * allocated, how specifies whether to wait. If the allocation fails, the
772 * original mbuf chain is freed and m is set to NULL.
773 */
774 #define M_PREPEND(m, plen, how) do { \
775 struct mbuf **_mmp = &(m); \
776 struct mbuf *_mm = *_mmp; \
777 int _mplen = (plen); \
778 int __mhow = (how); \
779 \
780 MBUF_CHECKSLEEP(how); \
781 if (M_LEADINGSPACE(_mm) >= _mplen) { \
782 _mm->m_data -= _mplen; \
783 _mm->m_len += _mplen; \
784 } else \
785 _mm = m_prepend(_mm, _mplen, __mhow); \
786 if (_mm != NULL && _mm->m_flags & M_PKTHDR) \
787 _mm->m_pkthdr.len += _mplen; \
788 *_mmp = _mm; \
789 } while (0)
790
791 /*
792 * Change mbuf to new type. This is a relatively expensive operation and
793 * should be avoided.
794 */
795 #define MCHTYPE(m, t) m_chtype((m), (t))
796
797 /* Length to m_copy to copy all. */
798 #define M_COPYALL 1000000000
799
800 /* Compatibility with 4.3. */
801 #define m_copy(m, o, l) m_copym((m), (o), (l), M_DONTWAIT)
802
803 extern int max_datalen; /* MHLEN - max_hdr */
804 extern int max_hdr; /* Largest link + protocol header */
805 extern int max_linkhdr; /* Largest link-level header */
806 extern int max_protohdr; /* Largest protocol header */
807 extern struct mbstat mbstat; /* General mbuf stats/infos */
808 extern int nmbclusters; /* Maximum number of clusters */
809
810 struct uio;
811
812 void m_adj(struct mbuf *, int);
813 void m_align(struct mbuf *, int);
814 int m_apply(struct mbuf *, int, int,
815 int (*)(void *, void *, u_int), void *);
816 int m_append(struct mbuf *, int, c_caddr_t);
817 void m_cat(struct mbuf *, struct mbuf *);
818 void m_extadd(struct mbuf *, caddr_t, u_int,
819 void (*)(void *, void *), void *, void *, int, int);
820 struct mbuf *m_collapse(struct mbuf *, int, int);
821 void m_copyback(struct mbuf *, int, int, c_caddr_t);
822 void m_copydata(const struct mbuf *, int, int, caddr_t);
823 struct mbuf *m_copym(struct mbuf *, int, int, int);
824 struct mbuf *m_copymdata(struct mbuf *, struct mbuf *,
825 int, int, int, int);
826 struct mbuf *m_copypacket(struct mbuf *, int);
827 void m_copy_pkthdr(struct mbuf *, struct mbuf *);
828 struct mbuf *m_copyup(struct mbuf *n, int len, int dstoff);
829 struct mbuf *m_defrag(struct mbuf *, int);
830 void m_demote(struct mbuf *, int);
831 struct mbuf *m_devget(char *, int, int, struct ifnet *,
832 void (*)(char *, caddr_t, u_int));
833 struct mbuf *m_dup(struct mbuf *, int);
834 int m_dup_pkthdr(struct mbuf *, struct mbuf *, int);
835 u_int m_fixhdr(struct mbuf *);
836 struct mbuf *m_fragment(struct mbuf *, int, int);
837 void m_freem(struct mbuf *);
838 struct mbuf *m_getm2(struct mbuf *, int, int, short, int);
839 struct mbuf *m_getptr(struct mbuf *, int, int *);
840 u_int m_length(struct mbuf *, struct mbuf **);
841 int m_mbuftouio(struct uio *, struct mbuf *, int);
842 void m_move_pkthdr(struct mbuf *, struct mbuf *);
843 struct mbuf *m_prepend(struct mbuf *, int, int);
844 void m_print(const struct mbuf *, int);
845 struct mbuf *m_pulldown(struct mbuf *, int, int, int *);
846 struct mbuf *m_pullup(struct mbuf *, int);
847 int m_sanity(struct mbuf *, int);
848 struct mbuf *m_split(struct mbuf *, int, int);
849 struct mbuf *m_uiotombuf(struct uio *, int, int, int, int);
850 struct mbuf *m_unshare(struct mbuf *, int how);
851
852 /*-
853 * Network packets may have annotations attached by affixing a list of
854 * "packet tags" to the pkthdr structure. Packet tags are dynamically
855 * allocated semi-opaque data structures that have a fixed header
856 * (struct m_tag) that specifies the size of the memory block and a
857 * <cookie,type> pair that identifies it. The cookie is a 32-bit unique
858 * unsigned value used to identify a module or ABI. By convention this value
859 * is chosen as the date+time that the module is created, expressed as the
860 * number of seconds since the epoch (e.g., using date -u +'%s'). The type
861 * value is an ABI/module-specific value that identifies a particular
862 * annotation and is private to the module. For compatibility with systems
863 * like OpenBSD that define packet tags w/o an ABI/module cookie, the value
864 * PACKET_ABI_COMPAT is used to implement m_tag_get and m_tag_find
865 * compatibility shim functions and several tag types are defined below.
866 * Users that do not require compatibility should use a private cookie value
867 * so that packet tag-related definitions can be maintained privately.
868 *
869 * Note that the packet tag returned by m_tag_alloc has the default memory
870 * alignment implemented by malloc. To reference private data one can use a
871 * construct like:
872 *
873 * struct m_tag *mtag = m_tag_alloc(...);
874 * struct foo *p = (struct foo *)(mtag+1);
875 *
876 * if the alignment of struct m_tag is sufficient for referencing members of
877 * struct foo. Otherwise it is necessary to embed struct m_tag within the
878 * private data structure to insure proper alignment; e.g.,
879 *
880 * struct foo {
881 * struct m_tag tag;
882 * ...
883 * };
884 * struct foo *p = (struct foo *) m_tag_alloc(...);
885 * struct m_tag *mtag = &p->tag;
886 */
887
888 /*
889 * Persistent tags stay with an mbuf until the mbuf is reclaimed. Otherwise
890 * tags are expected to ``vanish'' when they pass through a network
891 * interface. For most interfaces this happens normally as the tags are
892 * reclaimed when the mbuf is free'd. However in some special cases
893 * reclaiming must be done manually. An example is packets that pass through
894 * the loopback interface. Also, one must be careful to do this when
895 * ``turning around'' packets (e.g., icmp_reflect).
896 *
897 * To mark a tag persistent bit-or this flag in when defining the tag id.
898 * The tag will then be treated as described above.
899 */
900 #define MTAG_PERSISTENT 0x800
901
902 #define PACKET_TAG_NONE 0 /* Nadda */
903
904 /* Packet tags for use with PACKET_ABI_COMPAT. */
905 #define PACKET_TAG_IPSEC_IN_DONE 1 /* IPsec applied, in */
906 #define PACKET_TAG_IPSEC_OUT_DONE 2 /* IPsec applied, out */
907 #define PACKET_TAG_IPSEC_IN_CRYPTO_DONE 3 /* NIC IPsec crypto done */
908 #define PACKET_TAG_IPSEC_OUT_CRYPTO_NEEDED 4 /* NIC IPsec crypto req'ed */
909 #define PACKET_TAG_IPSEC_IN_COULD_DO_CRYPTO 5 /* NIC notifies IPsec */
910 #define PACKET_TAG_IPSEC_PENDING_TDB 6 /* Reminder to do IPsec */
911 #define PACKET_TAG_BRIDGE 7 /* Bridge processing done */
912 #define PACKET_TAG_GIF 8 /* GIF processing done */
913 #define PACKET_TAG_GRE 9 /* GRE processing done */
914 #define PACKET_TAG_IN_PACKET_CHECKSUM 10 /* NIC checksumming done */
915 #define PACKET_TAG_ENCAP 11 /* Encap. processing */
916 #define PACKET_TAG_IPSEC_SOCKET 12 /* IPSEC socket ref */
917 #define PACKET_TAG_IPSEC_HISTORY 13 /* IPSEC history */
918 #define PACKET_TAG_IPV6_INPUT 14 /* IPV6 input processing */
919 #define PACKET_TAG_DUMMYNET 15 /* dummynet info */
920 #define PACKET_TAG_DIVERT 17 /* divert info */
921 #define PACKET_TAG_IPFORWARD 18 /* ipforward info */
922 #define PACKET_TAG_MACLABEL (19 | MTAG_PERSISTENT) /* MAC label */
923 #define PACKET_TAG_PF 21 /* PF + ALTQ information */
924 #define PACKET_TAG_RTSOCKFAM 25 /* rtsock sa family */
925 #define PACKET_TAG_IPOPTIONS 27 /* Saved IP options */
926 #define PACKET_TAG_CARP 28 /* CARP info */
927 #define PACKET_TAG_IPSEC_NAT_T_PORTS 29 /* two uint16_t */
928
929 /* Specific cookies and tags. */
930
931 /* Packet tag routines. */
932 struct m_tag *m_tag_alloc(u_int32_t, int, int, int);
933 void m_tag_delete(struct mbuf *, struct m_tag *);
934 void m_tag_delete_chain(struct mbuf *, struct m_tag *);
935 void m_tag_free_default(struct m_tag *);
936 struct m_tag *m_tag_locate(struct mbuf *, u_int32_t, int, struct m_tag *);
937 struct m_tag *m_tag_copy(struct m_tag *, int);
938 int m_tag_copy_chain(struct mbuf *, struct mbuf *, int);
939 void m_tag_delete_nonpersistent(struct mbuf *);
940
941 /*
942 * Initialize the list of tags associated with an mbuf.
943 */
944 static __inline void
945 m_tag_init(struct mbuf *m)
946 {
947
948 SLIST_INIT(&m->m_pkthdr.tags);
949 }
950
951 /*
952 * Set up the contents of a tag. Note that this does not fill in the free
953 * method; the caller is expected to do that.
954 *
955 * XXX probably should be called m_tag_init, but that was already taken.
956 */
957 static __inline void
958 m_tag_setup(struct m_tag *t, u_int32_t cookie, int type, int len)
959 {
960
961 t->m_tag_id = type;
962 t->m_tag_len = len;
963 t->m_tag_cookie = cookie;
964 }
965
966 /*
967 * Reclaim resources associated with a tag.
968 */
969 static __inline void
970 m_tag_free(struct m_tag *t)
971 {
972
973 (*t->m_tag_free)(t);
974 }
975
976 /*
977 * Return the first tag associated with an mbuf.
978 */
979 static __inline struct m_tag *
980 m_tag_first(struct mbuf *m)
981 {
982
983 return (SLIST_FIRST(&m->m_pkthdr.tags));
984 }
985
986 /*
987 * Return the next tag in the list of tags associated with an mbuf.
988 */
989 static __inline struct m_tag *
990 m_tag_next(struct mbuf *m, struct m_tag *t)
991 {
992
993 return (SLIST_NEXT(t, m_tag_link));
994 }
995
996 /*
997 * Prepend a tag to the list of tags associated with an mbuf.
998 */
999 static __inline void
1000 m_tag_prepend(struct mbuf *m, struct m_tag *t)
1001 {
1002
1003 SLIST_INSERT_HEAD(&m->m_pkthdr.tags, t, m_tag_link);
1004 }
1005
1006 /*
1007 * Unlink a tag from the list of tags associated with an mbuf.
1008 */
1009 static __inline void
1010 m_tag_unlink(struct mbuf *m, struct m_tag *t)
1011 {
1012
1013 SLIST_REMOVE(&m->m_pkthdr.tags, t, m_tag, m_tag_link);
1014 }
1015
1016 /* These are for OpenBSD compatibility. */
1017 #define MTAG_ABI_COMPAT 0 /* compatibility ABI */
1018
1019 static __inline struct m_tag *
1020 m_tag_get(int type, int length, int wait)
1021 {
1022 return (m_tag_alloc(MTAG_ABI_COMPAT, type, length, wait));
1023 }
1024
1025 static __inline struct m_tag *
1026 m_tag_find(struct mbuf *m, int type, struct m_tag *start)
1027 {
1028 return (SLIST_EMPTY(&m->m_pkthdr.tags) ? (struct m_tag *)NULL :
1029 m_tag_locate(m, MTAG_ABI_COMPAT, type, start));
1030 }
1031
1032 /* XXX temporary FIB methods probably eventually use tags.*/
1033 #define M_FIBSHIFT 28
1034 #define M_FIBMASK 0x0F
1035
1036 /* get the fib from an mbuf and if it is not set, return the default */
1037 #define M_GETFIB(_m) \
1038 ((((_m)->m_flags & M_FIB) >> M_FIBSHIFT) & M_FIBMASK)
1039
1040 #define M_SETFIB(_m, _fib) do { \
1041 _m->m_flags &= ~M_FIB; \
1042 _m->m_flags |= (((_fib) << M_FIBSHIFT) & M_FIB); \
1043 } while (0)
1044
1045 #endif /* _KERNEL */
1046
1047 #ifdef MBUF_PROFILING
1048 void m_profile(struct mbuf *m);
1049 #define M_PROFILE(m) m_profile(m)
1050 #else
1051 #define M_PROFILE(m)
1052 #endif
1053
1054
1055 #endif /* !_SYS_MBUF_H_ */
Cache object: 1fe3a6beb188e21244e2904d797d1873
|