2 * Copyright (c) 1982, 1986, 1988, 1993
3 * The Regents of the University of California.
6 * Redistribution and use in source and binary forms, with or without
7 * modification, are permitted provided that the following conditions
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.
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
30 * @(#)mbuf.h 8.5 (Berkeley) 2/19/95
37 /* XXX: These includes suck. Sorry! */
38 #include <sys/queue.h>
40 #include <sys/systm.h>
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
56 * NB: These calculation do not take actual compiler-induced alignment and
57 * padding inside the complete struct mbuf into account. Appropriate
58 * attention is required when changing members of struct mbuf.
60 * MLEN is data length in a normal mbuf.
61 * MHLEN is data length in an mbuf with pktheader.
62 * MINCLSIZE is a smallest amount of data that should be put into cluster.
64 #define MLEN ((int)(MSIZE - sizeof(struct m_hdr)))
65 #define MHLEN ((int)(MLEN - sizeof(struct pkthdr)))
66 #define MINCLSIZE (MHLEN + 1)
70 * Macro for type conversion: convert mbuf pointer to data pointer of correct
73 * mtod(m, t) -- Convert mbuf pointer to data pointer of correct type.
74 * mtodo(m, o) -- Same as above but with offset 'o' into data.
76 #define mtod(m, t) ((t)((m)->m_data))
77 #define mtodo(m, o) ((void *)(((m)->m_data) + (o)))
80 * Argument structure passed to UMA routines during mbuf and packet
84 int flags; /* Flags for mbuf being allocated */
85 short type; /* Type of mbuf being allocated */
90 * Header present at the beginning of every mbuf.
95 struct mbuf *mh_next; /* next buffer in chain */
96 struct mbuf *mh_nextpkt; /* next chain in queue/record */
97 caddr_t mh_data; /* location of data */
98 int32_t mh_len; /* amount of data in this mbuf */
99 uint32_t mh_type:8, /* type of data in this mbuf */
100 mh_flags:24; /* flags; see below */
101 #if !defined(__LP64__)
102 uint32_t mh_pad; /* pad for 64bit alignment */
107 * Packet tag structure (see below for details).
110 SLIST_ENTRY(m_tag) m_tag_link; /* List of packet tags */
111 u_int16_t m_tag_id; /* Tag ID */
112 u_int16_t m_tag_len; /* Length of data */
113 u_int32_t m_tag_cookie; /* ABI/Module ID */
114 void (*m_tag_free)(struct m_tag *);
118 * Record/packet header in first mbuf of chain; valid only if M_PKTHDR is set.
123 struct ifnet *rcvif; /* rcv interface */
124 SLIST_HEAD(packet_tags, m_tag) tags; /* list of packet tags */
125 int32_t len; /* total packet length */
127 /* Layer crossing persistent information. */
128 uint32_t flowid; /* packet's 4-tuple system */
129 uint64_t csum_flags; /* checksum and offload features */
130 uint16_t fibnum; /* this packet should use this fib */
131 uint8_t cosqos; /* class/quality of service */
132 uint8_t rsstype; /* hash type */
133 uint8_t l2hlen; /* layer 2 header length */
134 uint8_t l3hlen; /* layer 3 header length */
135 uint8_t l4hlen; /* layer 4 header length */
136 uint8_t l5hlen; /* layer 5 header length */
140 uint32_t thirtytwo[2];
141 uint64_t sixtyfour[1];
142 uintptr_t unintptr[1];
146 /* Layer specific non-persistent local storage for reassembly, etc. */
150 uint32_t thirtytwo[2];
151 uint64_t sixtyfour[1];
152 uintptr_t unintptr[1];
156 #define ether_vtag PH_per.sixteen[0]
158 #define vt_nrecs sixteen[0]
159 #define tso_segsz PH_per.sixteen[1]
160 #define csum_phsum PH_per.sixteen[2]
161 #define csum_data PH_per.thirtytwo[1]
164 * Description of external storage mapped into mbuf; valid only if M_EXT is
170 volatile u_int *ref_cnt; /* pointer to ref count info */
171 caddr_t ext_buf; /* start of buffer */
172 uint32_t ext_size; /* size of buffer, for ext_free */
173 uint32_t ext_type:8, /* type of external storage */
174 ext_flags:24; /* external storage mbuf flags */
175 int (*ext_free) /* free routine if not the usual */
176 (struct mbuf *, void *, void *);
177 void *ext_arg1; /* optional argument pointer */
178 void *ext_arg2; /* optional argument pointer */
182 * The core of the mbuf object along with some shortcut defines for practical
189 struct pkthdr MH_pkthdr; /* M_PKTHDR set */
191 struct m_ext MH_ext; /* M_EXT set */
192 char MH_databuf[MHLEN];
195 char M_databuf[MLEN]; /* !M_PKTHDR, !M_EXT */
198 #define m_next m_hdr.mh_next
199 #define m_len m_hdr.mh_len
200 #define m_data m_hdr.mh_data
201 #define m_type m_hdr.mh_type
202 #define m_flags m_hdr.mh_flags
203 #define m_nextpkt m_hdr.mh_nextpkt
204 #define m_pkthdr M_dat.MH.MH_pkthdr
205 #define m_ext M_dat.MH.MH_dat.MH_ext
206 #define m_pktdat M_dat.MH.MH_dat.MH_databuf
207 #define m_dat M_dat.M_databuf
210 * NOTE: forwards compatibility definitions for mbuf(9)
212 * These aren't 1:1 with the macros in r277203; in particular they're exposed
213 * to both userland and kernel, whereas this is exposed to just _KERNEL -- to
214 * avoid disruption with existing KBI/KPIs
216 #define MHSIZE offsetof(struct mbuf, m_dat)
217 #define MPKTHSIZE offsetof(struct mbuf, m_pktdat)
220 * mbuf flags of global significance and layer crossing.
221 * Those of only protocol/layer specific significance are to be mapped
222 * to M_PROTO[1-12] and cleared at layer handoff boundaries.
223 * NB: Limited to the lower 24 bits.
225 #define M_EXT 0x00000001 /* has associated external storage */
226 #define M_PKTHDR 0x00000002 /* start of record */
227 #define M_EOR 0x00000004 /* end of record */
228 #define M_RDONLY 0x00000008 /* associated data is marked read-only */
229 #define M_BCAST 0x00000010 /* send/received as link-level broadcast */
230 #define M_MCAST 0x00000020 /* send/received as link-level multicast */
231 #define M_PROMISC 0x00000040 /* packet was not for us */
232 #define M_VLANTAG 0x00000080 /* ether_vtag is valid */
233 #define M_FLOWID 0x00000100 /* deprecated: flowid is valid */
234 #define M_NOFREE 0x00000200 /* do not free mbuf, embedded in cluster */
236 #define M_PROTO1 0x00001000 /* protocol-specific */
237 #define M_PROTO2 0x00002000 /* protocol-specific */
238 #define M_PROTO3 0x00004000 /* protocol-specific */
239 #define M_PROTO4 0x00008000 /* protocol-specific */
240 #define M_PROTO5 0x00010000 /* protocol-specific */
241 #define M_PROTO6 0x00020000 /* protocol-specific */
242 #define M_PROTO7 0x00040000 /* protocol-specific */
243 #define M_PROTO8 0x00080000 /* protocol-specific */
244 #define M_PROTO9 0x00100000 /* protocol-specific */
245 #define M_PROTO10 0x00200000 /* protocol-specific */
246 #define M_PROTO11 0x00400000 /* protocol-specific */
247 #define M_PROTO12 0x00800000 /* protocol-specific */
250 * Flags to purge when crossing layers.
252 #define M_PROTOFLAGS \
253 (M_PROTO1|M_PROTO2|M_PROTO3|M_PROTO4|M_PROTO5|M_PROTO6|M_PROTO7|M_PROTO8|\
254 M_PROTO9|M_PROTO10|M_PROTO11|M_PROTO12)
257 * Flags preserved when copying m_pkthdr.
259 #define M_COPYFLAGS \
260 (M_PKTHDR|M_EOR|M_RDONLY|M_BCAST|M_MCAST|M_PROMISC|M_VLANTAG|M_FLOWID| \
264 * Mbuf flag description for use with printf(9) %b identifier.
266 #define M_FLAG_BITS \
267 "\20\1M_EXT\2M_PKTHDR\3M_EOR\4M_RDONLY\5M_BCAST\6M_MCAST" \
268 "\7M_PROMISC\10M_VLANTAG\11M_FLOWID"
269 #define M_FLAG_PROTOBITS \
270 "\15M_PROTO1\16M_PROTO2\17M_PROTO3\20M_PROTO4\21M_PROTO5" \
271 "\22M_PROTO6\23M_PROTO7\24M_PROTO8\25M_PROTO9\26M_PROTO10" \
272 "\27M_PROTO11\30M_PROTO12"
273 #define M_FLAG_PRINTF (M_FLAG_BITS M_FLAG_PROTOBITS)
276 * Network interface cards are able to hash protocol fields (such as IPv4
277 * addresses and TCP port numbers) classify packets into flows. These flows
278 * can then be used to maintain ordering while delivering packets to the OS
279 * via parallel input queues, as well as to provide a stateless affinity
280 * model. NIC drivers can pass up the hash via m->m_pkthdr.flowid, and set
281 * m_flag fields to indicate how the hash should be interpreted by the
284 * Most NICs support RSS, which provides ordering and explicit affinity, and
285 * use the hash m_flag bits to indicate what header fields were covered by
286 * the hash. M_HASHTYPE_OPAQUE can be set by non-RSS cards or configurations
287 * that provide an opaque flow identifier, allowing for ordering and
288 * distribution without explicit affinity.
290 #define M_HASHTYPE_NONE 0
291 #define M_HASHTYPE_RSS_IPV4 1 /* IPv4 2-tuple */
292 #define M_HASHTYPE_RSS_TCP_IPV4 2 /* TCPv4 4-tuple */
293 #define M_HASHTYPE_RSS_IPV6 3 /* IPv6 2-tuple */
294 #define M_HASHTYPE_RSS_TCP_IPV6 4 /* TCPv6 4-tuple */
295 #define M_HASHTYPE_RSS_IPV6_EX 5 /* IPv6 2-tuple + ext hdrs */
296 #define M_HASHTYPE_RSS_TCP_IPV6_EX 6 /* TCPv6 4-tiple + ext hdrs */
297 #define M_HASHTYPE_OPAQUE 255 /* ordering, not affinity */
299 #define M_HASHTYPE_CLEAR(m) ((m)->m_pkthdr.rsstype = 0)
301 * Handle M_FLOWID for legacy drivers still using them.
303 #define M_HASHTYPE_GET(m) (((m->m_flags & M_FLOWID) && \
304 (m)->m_pkthdr.rsstype == M_HASHTYPE_NONE) ? M_HASHTYPE_OPAQUE : \
305 (m)->m_pkthdr.rsstype)
306 #define M_HASHTYPE_SET(m, v) do { \
307 if ((v) != M_HASHTYPE_NONE) \
308 m->m_flags |= M_FLOWID; \
309 (m)->m_pkthdr.rsstype = (v); \
311 #define M_HASHTYPE_TEST(m, v) (M_HASHTYPE_GET(m) == (v))
314 * COS/QOS class and quality of service tags.
315 * It uses DSCP code points as base.
317 #define QOS_DSCP_CS0 0x00
318 #define QOS_DSCP_DEF QOS_DSCP_CS0
319 #define QOS_DSCP_CS1 0x20
320 #define QOS_DSCP_AF11 0x28
321 #define QOS_DSCP_AF12 0x30
322 #define QOS_DSCP_AF13 0x38
323 #define QOS_DSCP_CS2 0x40
324 #define QOS_DSCP_AF21 0x48
325 #define QOS_DSCP_AF22 0x50
326 #define QOS_DSCP_AF23 0x58
327 #define QOS_DSCP_CS3 0x60
328 #define QOS_DSCP_AF31 0x68
329 #define QOS_DSCP_AF32 0x70
330 #define QOS_DSCP_AF33 0x78
331 #define QOS_DSCP_CS4 0x80
332 #define QOS_DSCP_AF41 0x88
333 #define QOS_DSCP_AF42 0x90
334 #define QOS_DSCP_AF43 0x98
335 #define QOS_DSCP_CS5 0xa0
336 #define QOS_DSCP_EF 0xb8
337 #define QOS_DSCP_CS6 0xc0
338 #define QOS_DSCP_CS7 0xe0
341 * External mbuf storage buffer types.
343 #define EXT_CLUSTER 1 /* mbuf cluster */
344 #define EXT_SFBUF 2 /* sendfile(2)'s sf_bufs */
345 #define EXT_JUMBOP 3 /* jumbo cluster 4096 bytes */
346 #define EXT_JUMBO9 4 /* jumbo cluster 9216 bytes */
347 #define EXT_JUMBO16 5 /* jumbo cluster 16184 bytes */
348 #define EXT_PACKET 6 /* mbuf+cluster from packet zone */
349 #define EXT_MBUF 7 /* external mbuf reference (M_IOVEC) */
351 #define EXT_VENDOR1 224 /* for vendor-internal use */
352 #define EXT_VENDOR2 225 /* for vendor-internal use */
353 #define EXT_VENDOR3 226 /* for vendor-internal use */
354 #define EXT_VENDOR4 227 /* for vendor-internal use */
356 #define EXT_EXP1 244 /* for experimental use */
357 #define EXT_EXP2 245 /* for experimental use */
358 #define EXT_EXP3 246 /* for experimental use */
359 #define EXT_EXP4 247 /* for experimental use */
361 #define EXT_NET_DRV 252 /* custom ext_buf provided by net driver(s) */
362 #define EXT_MOD_TYPE 253 /* custom module's ext_buf type */
363 #define EXT_DISPOSABLE 254 /* can throw this buffer away w/page flipping */
364 #define EXT_EXTREF 255 /* has externally maintained ref_cnt ptr */
367 * Flags for external mbuf buffer types.
368 * NB: limited to the lower 24 bits.
370 #define EXT_FLAG_EMBREF 0x000001 /* embedded ref_cnt, notyet */
371 #define EXT_FLAG_EXTREF 0x000002 /* external ref_cnt, notyet */
372 #define EXT_FLAG_NOFREE 0x000010 /* don't free mbuf to pool, notyet */
374 #define EXT_FLAG_VENDOR1 0x010000 /* for vendor-internal use */
375 #define EXT_FLAG_VENDOR2 0x020000 /* for vendor-internal use */
376 #define EXT_FLAG_VENDOR3 0x040000 /* for vendor-internal use */
377 #define EXT_FLAG_VENDOR4 0x080000 /* for vendor-internal use */
379 #define EXT_FLAG_EXP1 0x100000 /* for experimental use */
380 #define EXT_FLAG_EXP2 0x200000 /* for experimental use */
381 #define EXT_FLAG_EXP3 0x400000 /* for experimental use */
382 #define EXT_FLAG_EXP4 0x800000 /* for experimental use */
385 * EXT flag description for use with printf(9) %b identifier.
387 #define EXT_FLAG_BITS \
388 "\20\1EXT_FLAG_EMBREF\2EXT_FLAG_EXTREF\5EXT_FLAG_NOFREE" \
389 "\21EXT_FLAG_VENDOR1\22EXT_FLAG_VENDOR2\23EXT_FLAG_VENDOR3" \
390 "\24EXT_FLAG_VENDOR4\25EXT_FLAG_EXP1\26EXT_FLAG_EXP2\27EXT_FLAG_EXP3" \
394 * Return values for (*ext_free).
396 #define EXT_FREE_OK 0 /* Normal return */
399 * Flags indicating checksum, segmentation and other offload work to be
400 * done, or already done, by hardware or lower layers. It is split into
401 * separate inbound and outbound flags.
403 * Outbound flags that are set by upper protocol layers requesting lower
404 * layers, or ideally the hardware, to perform these offloading tasks.
405 * For outbound packets this field and its flags can be directly tested
406 * against if_data.ifi_hwassist.
408 #define CSUM_IP 0x00000001 /* IP header checksum offload */
409 #define CSUM_IP_UDP 0x00000002 /* UDP checksum offload */
410 #define CSUM_IP_TCP 0x00000004 /* TCP checksum offload */
411 #define CSUM_IP_SCTP 0x00000008 /* SCTP checksum offload */
412 #define CSUM_IP_TSO 0x00000010 /* TCP segmentation offload */
413 #define CSUM_IP_ISCSI 0x00000020 /* iSCSI checksum offload */
415 #define CSUM_IP6_UDP 0x00000200 /* UDP checksum offload */
416 #define CSUM_IP6_TCP 0x00000400 /* TCP checksum offload */
417 #define CSUM_IP6_SCTP 0x00000800 /* SCTP checksum offload */
418 #define CSUM_IP6_TSO 0x00001000 /* TCP segmentation offload */
419 #define CSUM_IP6_ISCSI 0x00002000 /* iSCSI checksum offload */
421 /* Inbound checksum support where the checksum was verified by hardware. */
422 #define CSUM_L3_CALC 0x01000000 /* calculated layer 3 csum */
423 #define CSUM_L3_VALID 0x02000000 /* checksum is correct */
424 #define CSUM_L4_CALC 0x04000000 /* calculated layer 4 csum */
425 #define CSUM_L4_VALID 0x08000000 /* checksum is correct */
426 #define CSUM_L5_CALC 0x10000000 /* calculated layer 5 csum */
427 #define CSUM_L5_VALID 0x20000000 /* checksum is correct */
428 #define CSUM_COALESED 0x40000000 /* contains merged segments */
431 * CSUM flag description for use with printf(9) %b identifier.
434 "\20\1CSUM_IP\2CSUM_IP_UDP\3CSUM_IP_TCP\4CSUM_IP_SCTP\5CSUM_IP_TSO" \
436 "\12CSUM_IP6_UDP\13CSUM_IP6_TCP\14CSUM_IP6_SCTP\15CSUM_IP6_TSO" \
437 "\16CSUM_IP6_ISCSI" \
438 "\31CSUM_L3_CALC\32CSUM_L3_VALID\33CSUM_L4_CALC\34CSUM_L4_VALID" \
439 "\35CSUM_L5_CALC\36CSUM_L5_VALID\37CSUM_COALESED"
441 /* CSUM flags compatibility mappings. */
442 #define CSUM_IP_CHECKED CSUM_L3_CALC
443 #define CSUM_IP_VALID CSUM_L3_VALID
444 #define CSUM_DATA_VALID CSUM_L4_VALID
445 #define CSUM_PSEUDO_HDR CSUM_L4_CALC
446 #define CSUM_SCTP_VALID CSUM_L4_VALID
447 #define CSUM_DELAY_DATA (CSUM_TCP|CSUM_UDP)
448 #define CSUM_DELAY_IP CSUM_IP /* Only v4, no v6 IP hdr csum */
449 #define CSUM_DELAY_DATA_IPV6 (CSUM_TCP_IPV6|CSUM_UDP_IPV6)
450 #define CSUM_DATA_VALID_IPV6 CSUM_DATA_VALID
451 #define CSUM_TCP CSUM_IP_TCP
452 #define CSUM_UDP CSUM_IP_UDP
453 #define CSUM_SCTP CSUM_IP_SCTP
454 #define CSUM_TSO (CSUM_IP_TSO|CSUM_IP6_TSO)
455 #define CSUM_UDP_IPV6 CSUM_IP6_UDP
456 #define CSUM_TCP_IPV6 CSUM_IP6_TCP
457 #define CSUM_SCTP_IPV6 CSUM_IP6_SCTP
458 #define CSUM_FRAGMENT 0x0 /* Unused */
461 * mbuf types describing the content of the mbuf (including external storage).
463 #define MT_NOTMBUF 0 /* USED INTERNALLY ONLY! Object is not mbuf */
464 #define MT_DATA 1 /* dynamic (data) allocation */
465 #define MT_HEADER MT_DATA /* packet header, use M_PKTHDR instead */
467 #define MT_VENDOR1 4 /* for vendor-internal use */
468 #define MT_VENDOR2 5 /* for vendor-internal use */
469 #define MT_VENDOR3 6 /* for vendor-internal use */
470 #define MT_VENDOR4 7 /* for vendor-internal use */
472 #define MT_SONAME 8 /* socket name */
474 #define MT_EXP1 9 /* for experimental use */
475 #define MT_EXP2 10 /* for experimental use */
476 #define MT_EXP3 11 /* for experimental use */
477 #define MT_EXP4 12 /* for experimental use */
479 #define MT_CONTROL 14 /* extra-data protocol message */
480 #define MT_OOBDATA 15 /* expedited data */
481 #define MT_NTYPES 16 /* number of mbuf types for mbtypes[] */
483 #define MT_NOINIT 255 /* Not a type but a flag to allocate
484 a non-initialized mbuf */
487 * Compatibility with historic mbuf allocator.
489 #define MBTOM(how) (how)
490 #define M_DONTWAIT M_NOWAIT
491 #define M_TRYWAIT M_WAITOK
492 #define M_WAIT M_WAITOK
495 * String names of mbuf-related UMA(9) and malloc(9) types. Exposed to
496 * !_KERNEL so that monitoring tools can look up the zones with
499 #define MBUF_MEM_NAME "mbuf"
500 #define MBUF_CLUSTER_MEM_NAME "mbuf_cluster"
501 #define MBUF_PACKET_MEM_NAME "mbuf_packet"
502 #define MBUF_JUMBOP_MEM_NAME "mbuf_jumbo_page"
503 #define MBUF_JUMBO9_MEM_NAME "mbuf_jumbo_9k"
504 #define MBUF_JUMBO16_MEM_NAME "mbuf_jumbo_16k"
505 #define MBUF_TAG_MEM_NAME "mbuf_tag"
506 #define MBUF_EXTREFCNT_MEM_NAME "mbuf_ext_refcnt"
511 #define MBUF_CHECKSLEEP(how) do { \
512 if (how == M_WAITOK) \
513 WITNESS_WARN(WARN_GIANTOK | WARN_SLEEPOK, NULL, \
514 "Sleeping in \"%s\"", __func__); \
517 #define MBUF_CHECKSLEEP(how)
521 * Network buffer allocation API
523 * The rest of it is defined in kern/kern_mbuf.c
525 extern uma_zone_t zone_mbuf;
526 extern uma_zone_t zone_clust;
527 extern uma_zone_t zone_pack;
528 extern uma_zone_t zone_jumbop;
529 extern uma_zone_t zone_jumbo9;
530 extern uma_zone_t zone_jumbo16;
531 extern uma_zone_t zone_ext_refcnt;
533 void mb_free_ext(struct mbuf *);
534 int m_pkthdr_init(struct mbuf *, int);
548 #if MJUMPAGESIZE != MCLBYTES
560 panic("%s: invalid cluster size %d", __func__, size);
567 * Associated an external reference counted buffer with an mbuf.
570 m_extaddref(struct mbuf *m, caddr_t buf, u_int size, u_int *ref_cnt,
571 int (*freef)(struct mbuf *, void *, void *), void *arg1, void *arg2)
574 KASSERT(ref_cnt != NULL, ("%s: ref_cnt not provided", __func__));
576 atomic_add_int(ref_cnt, 1);
578 m->m_ext.ext_buf = buf;
579 m->m_ext.ref_cnt = ref_cnt;
580 m->m_data = m->m_ext.ext_buf;
581 m->m_ext.ext_size = size;
582 m->m_ext.ext_free = freef;
583 m->m_ext.ext_arg1 = arg1;
584 m->m_ext.ext_arg2 = arg2;
585 m->m_ext.ext_type = EXT_EXTREF;
586 m->m_ext.ext_flags = 0;
589 static __inline uma_zone_t
598 #if MJUMPAGESIZE != MCLBYTES
610 panic("%s: invalid cluster size %d", __func__, size);
617 * Initialize an mbuf with linear storage.
619 * Inline because the consumer text overhead will be roughly the same to
620 * initialize or call a function with this many parameters and M_PKTHDR
621 * should go away with constant propagation for !MGETHDR.
624 m_init(struct mbuf *m, uma_zone_t zone, int size, int how, short type,
631 m->m_data = m->m_dat;
635 if (flags & M_PKTHDR) {
636 if ((error = m_pkthdr_init(m, how)) != 0)
643 static __inline struct mbuf *
644 m_get(int how, short type)
650 return (uma_zalloc_arg(zone_mbuf, &args, how));
654 * XXX This should be deprecated, very little use.
656 static __inline struct mbuf *
657 m_getclr(int how, short type)
664 m = uma_zalloc_arg(zone_mbuf, &args, how);
666 bzero(m->m_data, MLEN);
670 static __inline struct mbuf *
671 m_gethdr(int how, short type)
675 args.flags = M_PKTHDR;
677 return (uma_zalloc_arg(zone_mbuf, &args, how));
680 static __inline struct mbuf *
681 m_getcl(int how, short type, int flags)
687 return (uma_zalloc_arg(zone_pack, &args, how));
691 m_clget(struct mbuf *m, int how)
694 if (m->m_flags & M_EXT)
695 printf("%s: %p mbuf already has cluster\n", __func__, m);
696 m->m_ext.ext_buf = (char *)NULL;
697 uma_zalloc_arg(zone_clust, m, how);
699 * On a cluster allocation failure, drain the packet zone and retry,
700 * we might be able to loosen a few clusters up on the drain.
702 if ((how & M_NOWAIT) && (m->m_ext.ext_buf == NULL)) {
703 zone_drain(zone_pack);
704 uma_zalloc_arg(zone_clust, m, how);
709 * m_cljget() is different from m_clget() as it can allocate clusters without
710 * attaching them to an mbuf. In that case the return value is the pointer
711 * to the cluster of the requested size. If an mbuf was specified, it gets
712 * the cluster attached to it and the return value can be safely ignored.
713 * For size it takes MCLBYTES, MJUMPAGESIZE, MJUM9BYTES, MJUM16BYTES.
715 static __inline void *
716 m_cljget(struct mbuf *m, int how, int size)
720 if (m && m->m_flags & M_EXT)
721 printf("%s: %p mbuf already has cluster\n", __func__, m);
723 m->m_ext.ext_buf = NULL;
725 zone = m_getzone(size);
726 return (uma_zalloc_arg(zone, m, how));
730 m_cljset(struct mbuf *m, void *cl, int type)
740 #if MJUMPAGESIZE != MCLBYTES
755 panic("%s: unknown cluster type %d", __func__, type);
759 m->m_data = m->m_ext.ext_buf = cl;
760 m->m_ext.ext_free = m->m_ext.ext_arg1 = m->m_ext.ext_arg2 = NULL;
761 m->m_ext.ext_size = size;
762 m->m_ext.ext_type = type;
763 m->m_ext.ext_flags = 0;
764 m->m_ext.ref_cnt = uma_find_refcnt(zone, cl);
770 m_chtype(struct mbuf *m, short new_type)
773 m->m_type = new_type;
777 m_clrprotoflags(struct mbuf *m)
780 m->m_flags &= ~M_PROTOFLAGS;
783 static __inline struct mbuf *
784 m_last(struct mbuf *m)
793 * mbuf, cluster, and external object allocation macros (for compatibility
796 #define M_MOVE_PKTHDR(to, from) m_move_pkthdr((to), (from))
797 #define MGET(m, how, type) ((m) = m_get((how), (type)))
798 #define MGETHDR(m, how, type) ((m) = m_gethdr((how), (type)))
799 #define MCLGET(m, how) m_clget((m), (how))
800 #define MEXTADD(m, buf, size, free, arg1, arg2, flags, type) \
801 (void )m_extadd((m), (caddr_t)(buf), (size), (free), (arg1), (arg2),\
802 (flags), (type), M_NOWAIT)
803 #define m_getm(m, len, how, type) \
804 m_getm2((m), (len), (how), (type), M_PKTHDR)
807 * Evaluate TRUE if it's safe to write to the mbuf m's data region (this can
808 * be both the local data payload, or an external buffer area, depending on
809 * whether M_EXT is set).
811 #define M_WRITABLE(m) (!((m)->m_flags & M_RDONLY) && \
812 (!(((m)->m_flags & M_EXT)) || \
813 (*((m)->m_ext.ref_cnt) == 1)) ) \
815 /* Check if the supplied mbuf has a packet header, or else panic. */
816 #define M_ASSERTPKTHDR(m) \
817 KASSERT((m) != NULL && (m)->m_flags & M_PKTHDR, \
818 ("%s: no mbuf packet header!", __func__))
821 * Ensure that the supplied mbuf is a valid, non-free mbuf.
823 * XXX: Broken at the moment. Need some UMA magic to make it work again.
825 #define M_ASSERTVALID(m) \
826 KASSERT((((struct mbuf *)m)->m_flags & 0) == 0, \
827 ("%s: attempted use of a free mbuf!", __func__))
830 * Set the m_data pointer of a newly-allocated mbuf (m_get/MGET) to place an
831 * object of the specified size at the end of the mbuf, longword aligned.
833 #define M_ALIGN(m, len) do { \
834 KASSERT(!((m)->m_flags & (M_PKTHDR|M_EXT)), \
835 ("%s: M_ALIGN not normal mbuf", __func__)); \
836 KASSERT((m)->m_data == (m)->m_dat, \
837 ("%s: M_ALIGN not a virgin mbuf", __func__)); \
838 (m)->m_data += (MLEN - (len)) & ~(sizeof(long) - 1); \
842 * As above, for mbufs allocated with m_gethdr/MGETHDR or initialized by
845 #define MH_ALIGN(m, len) do { \
846 KASSERT((m)->m_flags & M_PKTHDR && !((m)->m_flags & M_EXT), \
847 ("%s: MH_ALIGN not PKTHDR mbuf", __func__)); \
848 KASSERT((m)->m_data == (m)->m_pktdat, \
849 ("%s: MH_ALIGN not a virgin mbuf", __func__)); \
850 (m)->m_data += (MHLEN - (len)) & ~(sizeof(long) - 1); \
854 * As above, for mbuf with external storage.
856 #define MEXT_ALIGN(m, len) do { \
857 KASSERT((m)->m_flags & M_EXT, \
858 ("%s: MEXT_ALIGN not an M_EXT mbuf", __func__)); \
859 KASSERT((m)->m_data == (m)->m_ext.ext_buf, \
860 ("%s: MEXT_ALIGN not a virgin mbuf", __func__)); \
861 (m)->m_data += ((m)->m_ext.ext_size - (len)) & \
862 ~(sizeof(long) - 1); \
866 * Compute the amount of space available before the current start of data in
869 * The M_WRITABLE() is a temporary, conservative safety measure: the burden
870 * of checking writability of the mbuf data area rests solely with the caller.
872 #define M_LEADINGSPACE(m) \
873 ((m)->m_flags & M_EXT ? \
874 (M_WRITABLE(m) ? (m)->m_data - (m)->m_ext.ext_buf : 0): \
875 (m)->m_flags & M_PKTHDR ? (m)->m_data - (m)->m_pktdat : \
876 (m)->m_data - (m)->m_dat)
879 * Compute the amount of space available after the end of data in an mbuf.
881 * The M_WRITABLE() is a temporary, conservative safety measure: the burden
882 * of checking writability of the mbuf data area rests solely with the caller.
884 #define M_TRAILINGSPACE(m) \
885 ((m)->m_flags & M_EXT ? \
886 (M_WRITABLE(m) ? (m)->m_ext.ext_buf + (m)->m_ext.ext_size \
887 - ((m)->m_data + (m)->m_len) : 0) : \
888 &(m)->m_dat[MLEN] - ((m)->m_data + (m)->m_len))
891 * Arrange to prepend space of size plen to mbuf m. If a new mbuf must be
892 * allocated, how specifies whether to wait. If the allocation fails, the
893 * original mbuf chain is freed and m is set to NULL.
895 #define M_PREPEND(m, plen, how) do { \
896 struct mbuf **_mmp = &(m); \
897 struct mbuf *_mm = *_mmp; \
898 int _mplen = (plen); \
899 int __mhow = (how); \
901 MBUF_CHECKSLEEP(how); \
902 if (M_LEADINGSPACE(_mm) >= _mplen) { \
903 _mm->m_data -= _mplen; \
904 _mm->m_len += _mplen; \
906 _mm = m_prepend(_mm, _mplen, __mhow); \
907 if (_mm != NULL && _mm->m_flags & M_PKTHDR) \
908 _mm->m_pkthdr.len += _mplen; \
913 * Change mbuf to new type. This is a relatively expensive operation and
916 #define MCHTYPE(m, t) m_chtype((m), (t))
918 /* Length to m_copy to copy all. */
919 #define M_COPYALL 1000000000
921 /* Compatibility with 4.3. */
922 #define m_copy(m, o, l) m_copym((m), (o), (l), M_NOWAIT)
924 extern int max_datalen; /* MHLEN - max_hdr */
925 extern int max_hdr; /* Largest link + protocol header */
926 extern int max_linkhdr; /* Largest link-level header */
927 extern int max_protohdr; /* Largest protocol header */
928 extern int nmbclusters; /* Maximum number of clusters */
932 void m_adj(struct mbuf *, int);
933 void m_align(struct mbuf *, int);
934 int m_apply(struct mbuf *, int, int,
935 int (*)(void *, void *, u_int), void *);
936 int m_append(struct mbuf *, int, c_caddr_t);
937 void m_cat(struct mbuf *, struct mbuf *);
938 int m_extadd(struct mbuf *, caddr_t, u_int,
939 int (*)(struct mbuf *, void *, void *), void *, void *,
941 struct mbuf *m_collapse(struct mbuf *, int, int);
942 void m_copyback(struct mbuf *, int, int, c_caddr_t);
943 void m_copydata(const struct mbuf *, int, int, caddr_t);
944 struct mbuf *m_copym(struct mbuf *, int, int, int);
945 struct mbuf *m_copymdata(struct mbuf *, struct mbuf *,
947 struct mbuf *m_copypacket(struct mbuf *, int);
948 void m_copy_pkthdr(struct mbuf *, struct mbuf *);
949 struct mbuf *m_copyup(struct mbuf *, int, int);
950 struct mbuf *m_defrag(struct mbuf *, int);
951 void m_demote(struct mbuf *, int);
952 struct mbuf *m_devget(char *, int, int, struct ifnet *,
953 void (*)(char *, caddr_t, u_int));
954 struct mbuf *m_dup(struct mbuf *, int);
955 int m_dup_pkthdr(struct mbuf *, struct mbuf *, int);
956 u_int m_fixhdr(struct mbuf *);
957 struct mbuf *m_fragment(struct mbuf *, int, int);
958 void m_freem(struct mbuf *);
959 struct mbuf *m_get2(int, int, short, int);
960 struct mbuf *m_getjcl(int, short, int, int);
961 struct mbuf *m_getm2(struct mbuf *, int, int, short, int);
962 struct mbuf *m_getptr(struct mbuf *, int, int *);
963 u_int m_length(struct mbuf *, struct mbuf **);
964 int m_mbuftouio(struct uio *, struct mbuf *, int);
965 void m_move_pkthdr(struct mbuf *, struct mbuf *);
966 struct mbuf *m_prepend(struct mbuf *, int, int);
967 void m_print(const struct mbuf *, int);
968 struct mbuf *m_pulldown(struct mbuf *, int, int, int *);
969 struct mbuf *m_pullup(struct mbuf *, int);
970 int m_sanity(struct mbuf *, int);
971 struct mbuf *m_split(struct mbuf *, int, int);
972 struct mbuf *m_uiotombuf(struct uio *, int, int, int, int);
973 struct mbuf *m_unshare(struct mbuf *, int);
976 * Network packets may have annotations attached by affixing a list of
977 * "packet tags" to the pkthdr structure. Packet tags are dynamically
978 * allocated semi-opaque data structures that have a fixed header
979 * (struct m_tag) that specifies the size of the memory block and a
980 * <cookie,type> pair that identifies it. The cookie is a 32-bit unique
981 * unsigned value used to identify a module or ABI. By convention this value
982 * is chosen as the date+time that the module is created, expressed as the
983 * number of seconds since the epoch (e.g., using date -u +'%s'). The type
984 * value is an ABI/module-specific value that identifies a particular
985 * annotation and is private to the module. For compatibility with systems
986 * like OpenBSD that define packet tags w/o an ABI/module cookie, the value
987 * PACKET_ABI_COMPAT is used to implement m_tag_get and m_tag_find
988 * compatibility shim functions and several tag types are defined below.
989 * Users that do not require compatibility should use a private cookie value
990 * so that packet tag-related definitions can be maintained privately.
992 * Note that the packet tag returned by m_tag_alloc has the default memory
993 * alignment implemented by malloc. To reference private data one can use a
996 * struct m_tag *mtag = m_tag_alloc(...);
997 * struct foo *p = (struct foo *)(mtag+1);
999 * if the alignment of struct m_tag is sufficient for referencing members of
1000 * struct foo. Otherwise it is necessary to embed struct m_tag within the
1001 * private data structure to insure proper alignment; e.g.,
1007 * struct foo *p = (struct foo *) m_tag_alloc(...);
1008 * struct m_tag *mtag = &p->tag;
1012 * Persistent tags stay with an mbuf until the mbuf is reclaimed. Otherwise
1013 * tags are expected to ``vanish'' when they pass through a network
1014 * interface. For most interfaces this happens normally as the tags are
1015 * reclaimed when the mbuf is free'd. However in some special cases
1016 * reclaiming must be done manually. An example is packets that pass through
1017 * the loopback interface. Also, one must be careful to do this when
1018 * ``turning around'' packets (e.g., icmp_reflect).
1020 * To mark a tag persistent bit-or this flag in when defining the tag id.
1021 * The tag will then be treated as described above.
1023 #define MTAG_PERSISTENT 0x800
1025 #define PACKET_TAG_NONE 0 /* Nadda */
1027 /* Packet tags for use with PACKET_ABI_COMPAT. */
1028 #define PACKET_TAG_IPSEC_IN_DONE 1 /* IPsec applied, in */
1029 #define PACKET_TAG_IPSEC_OUT_DONE 2 /* IPsec applied, out */
1030 #define PACKET_TAG_IPSEC_IN_CRYPTO_DONE 3 /* NIC IPsec crypto done */
1031 #define PACKET_TAG_IPSEC_OUT_CRYPTO_NEEDED 4 /* NIC IPsec crypto req'ed */
1032 #define PACKET_TAG_IPSEC_IN_COULD_DO_CRYPTO 5 /* NIC notifies IPsec */
1033 #define PACKET_TAG_IPSEC_PENDING_TDB 6 /* Reminder to do IPsec */
1034 #define PACKET_TAG_BRIDGE 7 /* Bridge processing done */
1035 #define PACKET_TAG_GIF 8 /* GIF processing done */
1036 #define PACKET_TAG_GRE 9 /* GRE processing done */
1037 #define PACKET_TAG_IN_PACKET_CHECKSUM 10 /* NIC checksumming done */
1038 #define PACKET_TAG_ENCAP 11 /* Encap. processing */
1039 #define PACKET_TAG_IPSEC_SOCKET 12 /* IPSEC socket ref */
1040 #define PACKET_TAG_IPSEC_HISTORY 13 /* IPSEC history */
1041 #define PACKET_TAG_IPV6_INPUT 14 /* IPV6 input processing */
1042 #define PACKET_TAG_DUMMYNET 15 /* dummynet info */
1043 #define PACKET_TAG_DIVERT 17 /* divert info */
1044 #define PACKET_TAG_IPFORWARD 18 /* ipforward info */
1045 #define PACKET_TAG_MACLABEL (19 | MTAG_PERSISTENT) /* MAC label */
1046 #define PACKET_TAG_PF (21 | MTAG_PERSISTENT) /* PF/ALTQ information */
1047 #define PACKET_TAG_RTSOCKFAM 25 /* rtsock sa family */
1048 #define PACKET_TAG_IPOPTIONS 27 /* Saved IP options */
1049 #define PACKET_TAG_CARP 28 /* CARP info */
1050 #define PACKET_TAG_IPSEC_NAT_T_PORTS 29 /* two uint16_t */
1051 #define PACKET_TAG_ND_OUTGOING 30 /* ND outgoing */
1053 /* Specific cookies and tags. */
1055 /* Packet tag routines. */
1056 struct m_tag *m_tag_alloc(u_int32_t, int, int, int);
1057 void m_tag_delete(struct mbuf *, struct m_tag *);
1058 void m_tag_delete_chain(struct mbuf *, struct m_tag *);
1059 void m_tag_free_default(struct m_tag *);
1060 struct m_tag *m_tag_locate(struct mbuf *, u_int32_t, int, struct m_tag *);
1061 struct m_tag *m_tag_copy(struct m_tag *, int);
1062 int m_tag_copy_chain(struct mbuf *, struct mbuf *, int);
1063 void m_tag_delete_nonpersistent(struct mbuf *);
1066 * Initialize the list of tags associated with an mbuf.
1068 static __inline void
1069 m_tag_init(struct mbuf *m)
1072 SLIST_INIT(&m->m_pkthdr.tags);
1076 * Set up the contents of a tag. Note that this does not fill in the free
1077 * method; the caller is expected to do that.
1079 * XXX probably should be called m_tag_init, but that was already taken.
1081 static __inline void
1082 m_tag_setup(struct m_tag *t, u_int32_t cookie, int type, int len)
1087 t->m_tag_cookie = cookie;
1091 * Reclaim resources associated with a tag.
1093 static __inline void
1094 m_tag_free(struct m_tag *t)
1097 (*t->m_tag_free)(t);
1101 * Return the first tag associated with an mbuf.
1103 static __inline struct m_tag *
1104 m_tag_first(struct mbuf *m)
1107 return (SLIST_FIRST(&m->m_pkthdr.tags));
1111 * Return the next tag in the list of tags associated with an mbuf.
1113 static __inline struct m_tag *
1114 m_tag_next(struct mbuf *m, struct m_tag *t)
1117 return (SLIST_NEXT(t, m_tag_link));
1121 * Prepend a tag to the list of tags associated with an mbuf.
1123 static __inline void
1124 m_tag_prepend(struct mbuf *m, struct m_tag *t)
1127 SLIST_INSERT_HEAD(&m->m_pkthdr.tags, t, m_tag_link);
1131 * Unlink a tag from the list of tags associated with an mbuf.
1133 static __inline void
1134 m_tag_unlink(struct mbuf *m, struct m_tag *t)
1137 SLIST_REMOVE(&m->m_pkthdr.tags, t, m_tag, m_tag_link);
1140 /* These are for OpenBSD compatibility. */
1141 #define MTAG_ABI_COMPAT 0 /* compatibility ABI */
1143 static __inline struct m_tag *
1144 m_tag_get(int type, int length, int wait)
1146 return (m_tag_alloc(MTAG_ABI_COMPAT, type, length, wait));
1149 static __inline struct m_tag *
1150 m_tag_find(struct mbuf *m, int type, struct m_tag *start)
1152 return (SLIST_EMPTY(&m->m_pkthdr.tags) ? (struct m_tag *)NULL :
1153 m_tag_locate(m, MTAG_ABI_COMPAT, type, start));
1156 static __inline struct mbuf *
1157 m_free(struct mbuf *m)
1159 struct mbuf *n = m->m_next;
1161 if ((m->m_flags & (M_PKTHDR|M_NOFREE)) == (M_PKTHDR|M_NOFREE))
1162 m_tag_delete_chain(m, NULL);
1163 if (m->m_flags & M_EXT)
1165 else if ((m->m_flags & M_NOFREE) == 0)
1166 uma_zfree(zone_mbuf, m);
1171 rt_m_getfib(struct mbuf *m)
1173 KASSERT(m->m_flags & M_PKTHDR , ("Attempt to get FIB from non header mbuf."));
1174 return (m->m_pkthdr.fibnum);
1177 #define M_GETFIB(_m) rt_m_getfib(_m)
1179 #define M_SETFIB(_m, _fib) do { \
1180 KASSERT((_m)->m_flags & M_PKTHDR, ("Attempt to set FIB on non header mbuf.")); \
1181 ((_m)->m_pkthdr.fibnum) = (_fib); \
1184 #endif /* _KERNEL */
1186 #ifdef MBUF_PROFILING
1187 void m_profile(struct mbuf *m);
1188 #define M_PROFILE(m) m_profile(m)
1190 #define M_PROFILE(m)
1194 #endif /* !_SYS_MBUF_H_ */