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 * Compile-time assertions in uipc_mbuf.c test these values to ensure that
68 #define MHSIZE offsetof(struct mbuf, m_dat)
69 #define MPKTHSIZE offsetof(struct mbuf, m_pktdat)
70 #define MLEN ((int)(MSIZE - MHSIZE))
71 #define MHLEN ((int)(MSIZE - MPKTHSIZE))
72 #define MINCLSIZE (MHLEN + 1)
76 * Macro for type conversion: convert mbuf pointer to data pointer of correct
79 * mtod(m, t) -- Convert mbuf pointer to data pointer of correct type.
80 * mtodo(m, o) -- Same as above but with offset 'o' into data.
82 #define mtod(m, t) ((t)((m)->m_data))
83 #define mtodo(m, o) ((void *)(((m)->m_data) + (o)))
86 * Argument structure passed to UMA routines during mbuf and packet
90 int flags; /* Flags for mbuf being allocated */
91 short type; /* Type of mbuf being allocated */
96 * Packet tag structure (see below for details).
99 SLIST_ENTRY(m_tag) m_tag_link; /* List of packet tags */
100 u_int16_t m_tag_id; /* Tag ID */
101 u_int16_t m_tag_len; /* Length of data */
102 u_int32_t m_tag_cookie; /* ABI/Module ID */
103 void (*m_tag_free)(struct m_tag *);
107 * Record/packet header in first mbuf of chain; valid only if M_PKTHDR is set.
110 * Compile-time assertions in uipc_mbuf.c test these values to ensure that
114 struct ifnet *rcvif; /* rcv interface */
115 SLIST_HEAD(packet_tags, m_tag) tags; /* list of packet tags */
116 int32_t len; /* total packet length */
118 /* Layer crossing persistent information. */
119 uint32_t flowid; /* packet's 4-tuple system */
120 uint64_t csum_flags; /* checksum and offload features */
121 uint16_t fibnum; /* this packet should use this fib */
122 uint8_t cosqos; /* class/quality of service */
123 uint8_t rsstype; /* hash type */
124 uint8_t l2hlen; /* layer 2 header length */
125 uint8_t l3hlen; /* layer 3 header length */
126 uint8_t l4hlen; /* layer 4 header length */
127 uint8_t l5hlen; /* layer 5 header length */
131 uint32_t thirtytwo[2];
132 uint64_t sixtyfour[1];
133 uintptr_t unintptr[1];
137 /* Layer specific non-persistent local storage for reassembly, etc. */
141 uint32_t thirtytwo[2];
142 uint64_t sixtyfour[1];
143 uintptr_t unintptr[1];
147 #define ether_vtag PH_per.sixteen[0]
149 #define vt_nrecs sixteen[0]
150 #define tso_segsz PH_per.sixteen[1]
151 #define csum_phsum PH_per.sixteen[2]
152 #define csum_data PH_per.thirtytwo[1]
153 #define pkt_tcphdr PH_loc.ptr
156 * Description of external storage mapped into mbuf; valid only if M_EXT is
160 * Compile-time assertions in uipc_mbuf.c test these values to ensure that
164 volatile u_int *ext_cnt; /* pointer to ref count info */
165 caddr_t ext_buf; /* start of buffer */
166 uint32_t ext_size; /* size of buffer, for ext_free */
167 uint32_t ext_type:8, /* type of external storage */
168 ext_flags:24; /* external storage mbuf flags */
169 void (*ext_free) /* free routine if not the usual */
170 (struct mbuf *, void *, void *);
171 void *ext_arg1; /* optional argument pointer */
172 void *ext_arg2; /* optional argument pointer */
176 * The core of the mbuf object along with some shortcut defines for practical
181 * Header present at the beginning of every mbuf.
184 * Compile-time assertions in uipc_mbuf.c test these values to ensure
185 * that they are correct.
187 union { /* next buffer in chain */
189 SLIST_ENTRY(mbuf) m_slist;
190 STAILQ_ENTRY(mbuf) m_stailq;
192 union { /* next chain in queue/record */
193 struct mbuf *m_nextpkt;
194 SLIST_ENTRY(mbuf) m_slistpkt;
195 STAILQ_ENTRY(mbuf) m_stailqpkt;
197 caddr_t m_data; /* location of data */
198 int32_t m_len; /* amount of data in this mbuf */
199 uint32_t m_type:8, /* type of data in this mbuf */
200 m_flags:24; /* flags; see below */
201 #if !defined(__LP64__)
202 uint32_t m_pad; /* pad for 64bit alignment */
206 * A set of optional headers (packet header, external storage header)
207 * and internal data storage. Historically, these arrays were sized
208 * to MHLEN (space left after a packet header) and MLEN (space left
209 * after only a regular mbuf header); they are now variable size in
210 * order to support future work on variable-size mbufs.
214 struct pkthdr m_pkthdr; /* M_PKTHDR set */
216 struct m_ext m_ext; /* M_EXT set */
220 char m_dat[0]; /* !M_PKTHDR, !M_EXT */
225 * mbuf flags of global significance and layer crossing.
226 * Those of only protocol/layer specific significance are to be mapped
227 * to M_PROTO[1-12] and cleared at layer handoff boundaries.
228 * NB: Limited to the lower 24 bits.
230 #define M_EXT 0x00000001 /* has associated external storage */
231 #define M_PKTHDR 0x00000002 /* start of record */
232 #define M_EOR 0x00000004 /* end of record */
233 #define M_RDONLY 0x00000008 /* associated data is marked read-only */
234 #define M_BCAST 0x00000010 /* send/received as link-level broadcast */
235 #define M_MCAST 0x00000020 /* send/received as link-level multicast */
236 #define M_PROMISC 0x00000040 /* packet was not for us */
237 #define M_VLANTAG 0x00000080 /* ether_vtag is valid */
238 #define M_UNUSED_8 0x00000100 /* --available-- */
239 #define M_NOFREE 0x00000200 /* do not free mbuf, embedded in cluster */
241 #define M_PROTO1 0x00001000 /* protocol-specific */
242 #define M_PROTO2 0x00002000 /* protocol-specific */
243 #define M_PROTO3 0x00004000 /* protocol-specific */
244 #define M_PROTO4 0x00008000 /* protocol-specific */
245 #define M_PROTO5 0x00010000 /* protocol-specific */
246 #define M_PROTO6 0x00020000 /* protocol-specific */
247 #define M_PROTO7 0x00040000 /* protocol-specific */
248 #define M_PROTO8 0x00080000 /* protocol-specific */
249 #define M_PROTO9 0x00100000 /* protocol-specific */
250 #define M_PROTO10 0x00200000 /* protocol-specific */
251 #define M_PROTO11 0x00400000 /* protocol-specific */
252 #define M_PROTO12 0x00800000 /* protocol-specific */
255 * Flags to purge when crossing layers.
257 #define M_PROTOFLAGS \
258 (M_PROTO1|M_PROTO2|M_PROTO3|M_PROTO4|M_PROTO5|M_PROTO6|M_PROTO7|M_PROTO8|\
259 M_PROTO9|M_PROTO10|M_PROTO11|M_PROTO12)
262 * Flags preserved when copying m_pkthdr.
264 #define M_COPYFLAGS \
265 (M_PKTHDR|M_EOR|M_RDONLY|M_BCAST|M_MCAST|M_PROMISC|M_VLANTAG| \
269 * Mbuf flag description for use with printf(9) %b identifier.
271 #define M_FLAG_BITS \
272 "\20\1M_EXT\2M_PKTHDR\3M_EOR\4M_RDONLY\5M_BCAST\6M_MCAST" \
273 "\7M_PROMISC\10M_VLANTAG"
274 #define M_FLAG_PROTOBITS \
275 "\15M_PROTO1\16M_PROTO2\17M_PROTO3\20M_PROTO4\21M_PROTO5" \
276 "\22M_PROTO6\23M_PROTO7\24M_PROTO8\25M_PROTO9\26M_PROTO10" \
277 "\27M_PROTO11\30M_PROTO12"
278 #define M_FLAG_PRINTF (M_FLAG_BITS M_FLAG_PROTOBITS)
281 * Network interface cards are able to hash protocol fields (such as IPv4
282 * addresses and TCP port numbers) classify packets into flows. These flows
283 * can then be used to maintain ordering while delivering packets to the OS
284 * via parallel input queues, as well as to provide a stateless affinity
285 * model. NIC drivers can pass up the hash via m->m_pkthdr.flowid, and set
286 * m_flag fields to indicate how the hash should be interpreted by the
289 * Most NICs support RSS, which provides ordering and explicit affinity, and
290 * use the hash m_flag bits to indicate what header fields were covered by
291 * the hash. M_HASHTYPE_OPAQUE can be set by non-RSS cards or configurations
292 * that provide an opaque flow identifier, allowing for ordering and
293 * distribution without explicit affinity.
295 /* Microsoft RSS standard hash types */
296 #define M_HASHTYPE_NONE 0
297 #define M_HASHTYPE_RSS_IPV4 1 /* IPv4 2-tuple */
298 #define M_HASHTYPE_RSS_TCP_IPV4 2 /* TCPv4 4-tuple */
299 #define M_HASHTYPE_RSS_IPV6 3 /* IPv6 2-tuple */
300 #define M_HASHTYPE_RSS_TCP_IPV6 4 /* TCPv6 4-tuple */
301 #define M_HASHTYPE_RSS_IPV6_EX 5 /* IPv6 2-tuple + ext hdrs */
302 #define M_HASHTYPE_RSS_TCP_IPV6_EX 6 /* TCPv6 4-tiple + ext hdrs */
303 /* Non-standard RSS hash types */
304 #define M_HASHTYPE_RSS_UDP_IPV4 7 /* IPv4 UDP 4-tuple */
305 #define M_HASHTYPE_RSS_UDP_IPV4_EX 8 /* IPv4 UDP 4-tuple + ext hdrs */
306 #define M_HASHTYPE_RSS_UDP_IPV6 9 /* IPv6 UDP 4-tuple */
307 #define M_HASHTYPE_RSS_UDP_IPV6_EX 10 /* IPv6 UDP 4-tuple + ext hdrs */
309 #define M_HASHTYPE_OPAQUE 255 /* ordering, not affinity */
311 #define M_HASHTYPE_CLEAR(m) ((m)->m_pkthdr.rsstype = 0)
312 #define M_HASHTYPE_GET(m) ((m)->m_pkthdr.rsstype)
313 #define M_HASHTYPE_SET(m, v) ((m)->m_pkthdr.rsstype = (v))
314 #define M_HASHTYPE_TEST(m, v) (M_HASHTYPE_GET(m) == (v))
317 * COS/QOS class and quality of service tags.
318 * It uses DSCP code points as base.
320 #define QOS_DSCP_CS0 0x00
321 #define QOS_DSCP_DEF QOS_DSCP_CS0
322 #define QOS_DSCP_CS1 0x20
323 #define QOS_DSCP_AF11 0x28
324 #define QOS_DSCP_AF12 0x30
325 #define QOS_DSCP_AF13 0x38
326 #define QOS_DSCP_CS2 0x40
327 #define QOS_DSCP_AF21 0x48
328 #define QOS_DSCP_AF22 0x50
329 #define QOS_DSCP_AF23 0x58
330 #define QOS_DSCP_CS3 0x60
331 #define QOS_DSCP_AF31 0x68
332 #define QOS_DSCP_AF32 0x70
333 #define QOS_DSCP_AF33 0x78
334 #define QOS_DSCP_CS4 0x80
335 #define QOS_DSCP_AF41 0x88
336 #define QOS_DSCP_AF42 0x90
337 #define QOS_DSCP_AF43 0x98
338 #define QOS_DSCP_CS5 0xa0
339 #define QOS_DSCP_EF 0xb8
340 #define QOS_DSCP_CS6 0xc0
341 #define QOS_DSCP_CS7 0xe0
344 * External mbuf storage buffer types.
346 #define EXT_CLUSTER 1 /* mbuf cluster */
347 #define EXT_SFBUF 2 /* sendfile(2)'s sf_bufs */
348 #define EXT_JUMBOP 3 /* jumbo cluster 4096 bytes */
349 #define EXT_JUMBO9 4 /* jumbo cluster 9216 bytes */
350 #define EXT_JUMBO16 5 /* jumbo cluster 16184 bytes */
351 #define EXT_PACKET 6 /* mbuf+cluster from packet zone */
352 #define EXT_MBUF 7 /* external mbuf reference (M_IOVEC) */
354 #define EXT_VENDOR1 224 /* for vendor-internal use */
355 #define EXT_VENDOR2 225 /* for vendor-internal use */
356 #define EXT_VENDOR3 226 /* for vendor-internal use */
357 #define EXT_VENDOR4 227 /* for vendor-internal use */
359 #define EXT_EXP1 244 /* for experimental use */
360 #define EXT_EXP2 245 /* for experimental use */
361 #define EXT_EXP3 246 /* for experimental use */
362 #define EXT_EXP4 247 /* for experimental use */
364 #define EXT_NET_DRV 252 /* custom ext_buf provided by net driver(s) */
365 #define EXT_MOD_TYPE 253 /* custom module's ext_buf type */
366 #define EXT_DISPOSABLE 254 /* can throw this buffer away w/page flipping */
367 #define EXT_EXTREF 255 /* has externally maintained ext_cnt ptr */
370 * Flags for external mbuf buffer types.
371 * NB: limited to the lower 24 bits.
373 #define EXT_FLAG_EMBREF 0x000001 /* embedded ext_cnt, notyet */
374 #define EXT_FLAG_EXTREF 0x000002 /* external ext_cnt, notyet */
375 #define EXT_FLAG_NOFREE 0x000010 /* don't free mbuf to pool, notyet */
377 #define EXT_FLAG_VENDOR1 0x010000 /* for vendor-internal use */
378 #define EXT_FLAG_VENDOR2 0x020000 /* for vendor-internal use */
379 #define EXT_FLAG_VENDOR3 0x040000 /* for vendor-internal use */
380 #define EXT_FLAG_VENDOR4 0x080000 /* for vendor-internal use */
382 #define EXT_FLAG_EXP1 0x100000 /* for experimental use */
383 #define EXT_FLAG_EXP2 0x200000 /* for experimental use */
384 #define EXT_FLAG_EXP3 0x400000 /* for experimental use */
385 #define EXT_FLAG_EXP4 0x800000 /* for experimental use */
388 * EXT flag description for use with printf(9) %b identifier.
390 #define EXT_FLAG_BITS \
391 "\20\1EXT_FLAG_EMBREF\2EXT_FLAG_EXTREF\5EXT_FLAG_NOFREE" \
392 "\21EXT_FLAG_VENDOR1\22EXT_FLAG_VENDOR2\23EXT_FLAG_VENDOR3" \
393 "\24EXT_FLAG_VENDOR4\25EXT_FLAG_EXP1\26EXT_FLAG_EXP2\27EXT_FLAG_EXP3" \
397 * External reference/free functions.
399 void sf_ext_ref(void *, void *);
400 void sf_ext_free(void *, void *);
403 * Flags indicating checksum, segmentation and other offload work to be
404 * done, or already done, by hardware or lower layers. It is split into
405 * separate inbound and outbound flags.
407 * Outbound flags that are set by upper protocol layers requesting lower
408 * layers, or ideally the hardware, to perform these offloading tasks.
409 * For outbound packets this field and its flags can be directly tested
410 * against ifnet if_hwassist.
412 #define CSUM_IP 0x00000001 /* IP header checksum offload */
413 #define CSUM_IP_UDP 0x00000002 /* UDP checksum offload */
414 #define CSUM_IP_TCP 0x00000004 /* TCP checksum offload */
415 #define CSUM_IP_SCTP 0x00000008 /* SCTP checksum offload */
416 #define CSUM_IP_TSO 0x00000010 /* TCP segmentation offload */
417 #define CSUM_IP_ISCSI 0x00000020 /* iSCSI checksum offload */
419 #define CSUM_IP6_UDP 0x00000200 /* UDP checksum offload */
420 #define CSUM_IP6_TCP 0x00000400 /* TCP checksum offload */
421 #define CSUM_IP6_SCTP 0x00000800 /* SCTP checksum offload */
422 #define CSUM_IP6_TSO 0x00001000 /* TCP segmentation offload */
423 #define CSUM_IP6_ISCSI 0x00002000 /* iSCSI checksum offload */
425 /* Inbound checksum support where the checksum was verified by hardware. */
426 #define CSUM_L3_CALC 0x01000000 /* calculated layer 3 csum */
427 #define CSUM_L3_VALID 0x02000000 /* checksum is correct */
428 #define CSUM_L4_CALC 0x04000000 /* calculated layer 4 csum */
429 #define CSUM_L4_VALID 0x08000000 /* checksum is correct */
430 #define CSUM_L5_CALC 0x10000000 /* calculated layer 5 csum */
431 #define CSUM_L5_VALID 0x20000000 /* checksum is correct */
432 #define CSUM_COALESED 0x40000000 /* contains merged segments */
435 * CSUM flag description for use with printf(9) %b identifier.
438 "\20\1CSUM_IP\2CSUM_IP_UDP\3CSUM_IP_TCP\4CSUM_IP_SCTP\5CSUM_IP_TSO" \
440 "\12CSUM_IP6_UDP\13CSUM_IP6_TCP\14CSUM_IP6_SCTP\15CSUM_IP6_TSO" \
441 "\16CSUM_IP6_ISCSI" \
442 "\31CSUM_L3_CALC\32CSUM_L3_VALID\33CSUM_L4_CALC\34CSUM_L4_VALID" \
443 "\35CSUM_L5_CALC\36CSUM_L5_VALID\37CSUM_COALESED"
445 /* CSUM flags compatibility mappings. */
446 #define CSUM_IP_CHECKED CSUM_L3_CALC
447 #define CSUM_IP_VALID CSUM_L3_VALID
448 #define CSUM_DATA_VALID CSUM_L4_VALID
449 #define CSUM_PSEUDO_HDR CSUM_L4_CALC
450 #define CSUM_SCTP_VALID CSUM_L4_VALID
451 #define CSUM_DELAY_DATA (CSUM_TCP|CSUM_UDP)
452 #define CSUM_DELAY_IP CSUM_IP /* Only v4, no v6 IP hdr csum */
453 #define CSUM_DELAY_DATA_IPV6 (CSUM_TCP_IPV6|CSUM_UDP_IPV6)
454 #define CSUM_DATA_VALID_IPV6 CSUM_DATA_VALID
455 #define CSUM_TCP CSUM_IP_TCP
456 #define CSUM_UDP CSUM_IP_UDP
457 #define CSUM_SCTP CSUM_IP_SCTP
458 #define CSUM_TSO (CSUM_IP_TSO|CSUM_IP6_TSO)
459 #define CSUM_UDP_IPV6 CSUM_IP6_UDP
460 #define CSUM_TCP_IPV6 CSUM_IP6_TCP
461 #define CSUM_SCTP_IPV6 CSUM_IP6_SCTP
464 * mbuf types describing the content of the mbuf (including external storage).
466 #define MT_NOTMBUF 0 /* USED INTERNALLY ONLY! Object is not mbuf */
467 #define MT_DATA 1 /* dynamic (data) allocation */
468 #define MT_HEADER MT_DATA /* packet header, use M_PKTHDR instead */
470 #define MT_VENDOR1 4 /* for vendor-internal use */
471 #define MT_VENDOR2 5 /* for vendor-internal use */
472 #define MT_VENDOR3 6 /* for vendor-internal use */
473 #define MT_VENDOR4 7 /* for vendor-internal use */
475 #define MT_SONAME 8 /* socket name */
477 #define MT_EXP1 9 /* for experimental use */
478 #define MT_EXP2 10 /* for experimental use */
479 #define MT_EXP3 11 /* for experimental use */
480 #define MT_EXP4 12 /* for experimental use */
482 #define MT_CONTROL 14 /* extra-data protocol message */
483 #define MT_OOBDATA 15 /* expedited data */
484 #define MT_NTYPES 16 /* number of mbuf types for mbtypes[] */
486 #define MT_NOINIT 255 /* Not a type but a flag to allocate
487 a non-initialized mbuf */
490 * String names of mbuf-related UMA(9) and malloc(9) types. Exposed to
491 * !_KERNEL so that monitoring tools can look up the zones with
494 #define MBUF_MEM_NAME "mbuf"
495 #define MBUF_CLUSTER_MEM_NAME "mbuf_cluster"
496 #define MBUF_PACKET_MEM_NAME "mbuf_packet"
497 #define MBUF_JUMBOP_MEM_NAME "mbuf_jumbo_page"
498 #define MBUF_JUMBO9_MEM_NAME "mbuf_jumbo_9k"
499 #define MBUF_JUMBO16_MEM_NAME "mbuf_jumbo_16k"
500 #define MBUF_TAG_MEM_NAME "mbuf_tag"
501 #define MBUF_EXTREFCNT_MEM_NAME "mbuf_ext_refcnt"
506 #define MBUF_CHECKSLEEP(how) do { \
507 if (how == M_WAITOK) \
508 WITNESS_WARN(WARN_GIANTOK | WARN_SLEEPOK, NULL, \
509 "Sleeping in \"%s\"", __func__); \
512 #define MBUF_CHECKSLEEP(how)
516 * Network buffer allocation API
518 * The rest of it is defined in kern/kern_mbuf.c
520 extern uma_zone_t zone_mbuf;
521 extern uma_zone_t zone_clust;
522 extern uma_zone_t zone_pack;
523 extern uma_zone_t zone_jumbop;
524 extern uma_zone_t zone_jumbo9;
525 extern uma_zone_t zone_jumbo16;
526 extern uma_zone_t zone_ext_refcnt;
528 void mb_free_ext(struct mbuf *);
529 int m_pkthdr_init(struct mbuf *, int);
543 #if MJUMPAGESIZE != MCLBYTES
555 panic("%s: invalid cluster size %d", __func__, size);
562 * Associated an external reference counted buffer with an mbuf.
565 m_extaddref(struct mbuf *m, caddr_t buf, u_int size, u_int *ref_cnt,
566 void (*freef)(struct mbuf *, void *, void *), void *arg1, void *arg2)
569 KASSERT(ref_cnt != NULL, ("%s: ref_cnt not provided", __func__));
571 atomic_add_int(ref_cnt, 1);
573 m->m_ext.ext_buf = buf;
574 m->m_ext.ext_cnt = ref_cnt;
575 m->m_data = m->m_ext.ext_buf;
576 m->m_ext.ext_size = size;
577 m->m_ext.ext_free = freef;
578 m->m_ext.ext_arg1 = arg1;
579 m->m_ext.ext_arg2 = arg2;
580 m->m_ext.ext_type = EXT_EXTREF;
581 m->m_ext.ext_flags = 0;
584 static __inline uma_zone_t
593 #if MJUMPAGESIZE != MCLBYTES
605 panic("%s: invalid cluster size %d", __func__, size);
612 * Initialize an mbuf with linear storage.
614 * Inline because the consumer text overhead will be roughly the same to
615 * initialize or call a function with this many parameters and M_PKTHDR
616 * should go away with constant propagation for !MGETHDR.
619 m_init(struct mbuf *m, uma_zone_t zone __unused, int size __unused, int how,
620 short type, int flags)
626 m->m_data = m->m_dat;
630 if (flags & M_PKTHDR) {
631 if ((error = m_pkthdr_init(m, how)) != 0)
638 static __inline struct mbuf *
639 m_get(int how, short type)
645 return (uma_zalloc_arg(zone_mbuf, &args, how));
649 * XXX This should be deprecated, very little use.
651 static __inline struct mbuf *
652 m_getclr(int how, short type)
659 m = uma_zalloc_arg(zone_mbuf, &args, how);
661 bzero(m->m_data, MLEN);
665 static __inline struct mbuf *
666 m_gethdr(int how, short type)
670 args.flags = M_PKTHDR;
672 return (uma_zalloc_arg(zone_mbuf, &args, how));
675 static __inline struct mbuf *
676 m_getcl(int how, short type, int flags)
682 return (uma_zalloc_arg(zone_pack, &args, how));
686 m_clget(struct mbuf *m, int how)
689 KASSERT((m->m_flags & M_EXT) == 0, ("%s: mbuf %p has M_EXT",
691 m->m_ext.ext_buf = (char *)NULL;
692 uma_zalloc_arg(zone_clust, m, how);
694 * On a cluster allocation failure, drain the packet zone and retry,
695 * we might be able to loosen a few clusters up on the drain.
697 if ((how & M_NOWAIT) && (m->m_ext.ext_buf == NULL)) {
698 zone_drain(zone_pack);
699 uma_zalloc_arg(zone_clust, m, how);
701 return (m->m_flags & M_EXT);
705 * m_cljget() is different from m_clget() as it can allocate clusters without
706 * attaching them to an mbuf. In that case the return value is the pointer
707 * to the cluster of the requested size. If an mbuf was specified, it gets
708 * the cluster attached to it and the return value can be safely ignored.
709 * For size it takes MCLBYTES, MJUMPAGESIZE, MJUM9BYTES, MJUM16BYTES.
711 static __inline void *
712 m_cljget(struct mbuf *m, int how, int size)
717 KASSERT((m->m_flags & M_EXT) == 0, ("%s: mbuf %p has M_EXT",
719 m->m_ext.ext_buf = NULL;
722 zone = m_getzone(size);
723 return (uma_zalloc_arg(zone, m, how));
727 m_cljset(struct mbuf *m, void *cl, int type)
737 #if MJUMPAGESIZE != MCLBYTES
752 panic("%s: unknown cluster type %d", __func__, type);
756 m->m_data = m->m_ext.ext_buf = cl;
757 m->m_ext.ext_free = m->m_ext.ext_arg1 = m->m_ext.ext_arg2 = NULL;
758 m->m_ext.ext_size = size;
759 m->m_ext.ext_type = type;
760 m->m_ext.ext_flags = 0;
761 m->m_ext.ext_cnt = uma_find_refcnt(zone, cl);
767 m_chtype(struct mbuf *m, short new_type)
770 m->m_type = new_type;
774 m_clrprotoflags(struct mbuf *m)
778 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.ext_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 * Return the address of the start of the buffer associated with an mbuf,
831 * handling external storage, packet-header mbufs, and regular data mbufs.
834 (((m)->m_flags & M_EXT) ? (m)->m_ext.ext_buf : \
835 ((m)->m_flags & M_PKTHDR) ? &(m)->m_pktdat[0] : \
839 * Return the size of the buffer associated with an mbuf, handling external
840 * storage, packet-header mbufs, and regular data mbufs.
843 (((m)->m_flags & M_EXT) ? (m)->m_ext.ext_size : \
844 ((m)->m_flags & M_PKTHDR) ? MHLEN : \
848 * Set the m_data pointer of a newly allocated mbuf to place an object of the
849 * specified size at the end of the mbuf, longword aligned.
851 * NB: Historically, we had M_ALIGN(), MH_ALIGN(), and MEXT_ALIGN() as
852 * separate macros, each asserting that it was called at the proper moment.
853 * This required callers to themselves test the storage type and call the
854 * right one. Rather than require callers to be aware of those layout
855 * decisions, we centralize here.
858 m_align(struct mbuf *m, int len)
861 const char *msg = "%s: not a virgin mbuf";
865 KASSERT(m->m_data == M_START(m), (msg, __func__));
867 adjust = M_SIZE(m) - len;
868 m->m_data += adjust &~ (sizeof(long)-1);
871 #define M_ALIGN(m, len) m_align(m, len)
872 #define MH_ALIGN(m, len) m_align(m, len)
873 #define MEXT_ALIGN(m, len) m_align(m, len)
876 * Compute the amount of space available before the current start of data in
879 * The M_WRITABLE() is a temporary, conservative safety measure: the burden
880 * of checking writability of the mbuf data area rests solely with the caller.
882 * NB: In previous versions, M_LEADINGSPACE() would only check M_WRITABLE()
883 * for mbufs with external storage. We now allow mbuf-embedded data to be
886 #define M_LEADINGSPACE(m) \
887 (M_WRITABLE(m) ? ((m)->m_data - M_START(m)) : 0)
890 * Compute the amount of space available after the end of data in an mbuf.
892 * The M_WRITABLE() is a temporary, conservative safety measure: the burden
893 * of checking writability of the mbuf data area rests solely with the caller.
895 * NB: In previous versions, M_TRAILINGSPACE() would only check M_WRITABLE()
896 * for mbufs with external storage. We now allow mbuf-embedded data to be
899 #define M_TRAILINGSPACE(m) \
901 ((M_START(m) + M_SIZE(m)) - ((m)->m_data + (m)->m_len)) : 0)
904 * Arrange to prepend space of size plen to mbuf m. If a new mbuf must be
905 * allocated, how specifies whether to wait. If the allocation fails, the
906 * original mbuf chain is freed and m is set to NULL.
908 #define M_PREPEND(m, plen, how) do { \
909 struct mbuf **_mmp = &(m); \
910 struct mbuf *_mm = *_mmp; \
911 int _mplen = (plen); \
912 int __mhow = (how); \
914 MBUF_CHECKSLEEP(how); \
915 if (M_LEADINGSPACE(_mm) >= _mplen) { \
916 _mm->m_data -= _mplen; \
917 _mm->m_len += _mplen; \
919 _mm = m_prepend(_mm, _mplen, __mhow); \
920 if (_mm != NULL && _mm->m_flags & M_PKTHDR) \
921 _mm->m_pkthdr.len += _mplen; \
926 * Change mbuf to new type. This is a relatively expensive operation and
929 #define MCHTYPE(m, t) m_chtype((m), (t))
931 /* Length to m_copy to copy all. */
932 #define M_COPYALL 1000000000
934 /* Compatibility with 4.3. */
935 #define m_copy(m, o, l) m_copym((m), (o), (l), M_NOWAIT)
937 extern int max_datalen; /* MHLEN - max_hdr */
938 extern int max_hdr; /* Largest link + protocol header */
939 extern int max_linkhdr; /* Largest link-level header */
940 extern int max_protohdr; /* Largest protocol header */
941 extern int nmbclusters; /* Maximum number of clusters */
945 void m_adj(struct mbuf *, int);
946 int m_apply(struct mbuf *, int, int,
947 int (*)(void *, void *, u_int), void *);
948 int m_append(struct mbuf *, int, c_caddr_t);
949 void m_cat(struct mbuf *, struct mbuf *);
950 void m_catpkt(struct mbuf *, struct mbuf *);
951 int m_extadd(struct mbuf *, caddr_t, u_int,
952 void (*)(struct mbuf *, void *, void *), void *, void *,
954 struct mbuf *m_collapse(struct mbuf *, int, int);
955 void m_copyback(struct mbuf *, int, int, c_caddr_t);
956 void m_copydata(const struct mbuf *, int, int, caddr_t);
957 struct mbuf *m_copym(struct mbuf *, int, int, int);
958 struct mbuf *m_copypacket(struct mbuf *, int);
959 void m_copy_pkthdr(struct mbuf *, struct mbuf *);
960 struct mbuf *m_copyup(struct mbuf *, int, int);
961 struct mbuf *m_defrag(struct mbuf *, int);
962 void m_demote_pkthdr(struct mbuf *);
963 void m_demote(struct mbuf *, int, int);
964 struct mbuf *m_devget(char *, int, int, struct ifnet *,
965 void (*)(char *, caddr_t, u_int));
966 struct mbuf *m_dup(struct mbuf *, int);
967 int m_dup_pkthdr(struct mbuf *, struct mbuf *, int);
968 u_int m_fixhdr(struct mbuf *);
969 struct mbuf *m_fragment(struct mbuf *, int, int);
970 void m_freem(struct mbuf *);
971 struct mbuf *m_get2(int, int, short, int);
972 struct mbuf *m_getjcl(int, short, int, int);
973 struct mbuf *m_getm2(struct mbuf *, int, int, short, int);
974 struct mbuf *m_getptr(struct mbuf *, int, int *);
975 u_int m_length(struct mbuf *, struct mbuf **);
976 int m_mbuftouio(struct uio *, struct mbuf *, int);
977 void m_move_pkthdr(struct mbuf *, struct mbuf *);
978 struct mbuf *m_prepend(struct mbuf *, int, int);
979 void m_print(const struct mbuf *, int);
980 struct mbuf *m_pulldown(struct mbuf *, int, int, int *);
981 struct mbuf *m_pullup(struct mbuf *, int);
982 int m_sanity(struct mbuf *, int);
983 struct mbuf *m_split(struct mbuf *, int, int);
984 struct mbuf *m_uiotombuf(struct uio *, int, int, int, int);
985 struct mbuf *m_unshare(struct mbuf *, int);
988 * Network packets may have annotations attached by affixing a list of
989 * "packet tags" to the pkthdr structure. Packet tags are dynamically
990 * allocated semi-opaque data structures that have a fixed header
991 * (struct m_tag) that specifies the size of the memory block and a
992 * <cookie,type> pair that identifies it. The cookie is a 32-bit unique
993 * unsigned value used to identify a module or ABI. By convention this value
994 * is chosen as the date+time that the module is created, expressed as the
995 * number of seconds since the epoch (e.g., using date -u +'%s'). The type
996 * value is an ABI/module-specific value that identifies a particular
997 * annotation and is private to the module. For compatibility with systems
998 * like OpenBSD that define packet tags w/o an ABI/module cookie, the value
999 * PACKET_ABI_COMPAT is used to implement m_tag_get and m_tag_find
1000 * compatibility shim functions and several tag types are defined below.
1001 * Users that do not require compatibility should use a private cookie value
1002 * so that packet tag-related definitions can be maintained privately.
1004 * Note that the packet tag returned by m_tag_alloc has the default memory
1005 * alignment implemented by malloc. To reference private data one can use a
1008 * struct m_tag *mtag = m_tag_alloc(...);
1009 * struct foo *p = (struct foo *)(mtag+1);
1011 * if the alignment of struct m_tag is sufficient for referencing members of
1012 * struct foo. Otherwise it is necessary to embed struct m_tag within the
1013 * private data structure to insure proper alignment; e.g.,
1019 * struct foo *p = (struct foo *) m_tag_alloc(...);
1020 * struct m_tag *mtag = &p->tag;
1024 * Persistent tags stay with an mbuf until the mbuf is reclaimed. Otherwise
1025 * tags are expected to ``vanish'' when they pass through a network
1026 * interface. For most interfaces this happens normally as the tags are
1027 * reclaimed when the mbuf is free'd. However in some special cases
1028 * reclaiming must be done manually. An example is packets that pass through
1029 * the loopback interface. Also, one must be careful to do this when
1030 * ``turning around'' packets (e.g., icmp_reflect).
1032 * To mark a tag persistent bit-or this flag in when defining the tag id.
1033 * The tag will then be treated as described above.
1035 #define MTAG_PERSISTENT 0x800
1037 #define PACKET_TAG_NONE 0 /* Nadda */
1039 /* Packet tags for use with PACKET_ABI_COMPAT. */
1040 #define PACKET_TAG_IPSEC_IN_DONE 1 /* IPsec applied, in */
1041 #define PACKET_TAG_IPSEC_OUT_DONE 2 /* IPsec applied, out */
1042 #define PACKET_TAG_IPSEC_IN_CRYPTO_DONE 3 /* NIC IPsec crypto done */
1043 #define PACKET_TAG_IPSEC_OUT_CRYPTO_NEEDED 4 /* NIC IPsec crypto req'ed */
1044 #define PACKET_TAG_IPSEC_IN_COULD_DO_CRYPTO 5 /* NIC notifies IPsec */
1045 #define PACKET_TAG_IPSEC_PENDING_TDB 6 /* Reminder to do IPsec */
1046 #define PACKET_TAG_BRIDGE 7 /* Bridge processing done */
1047 #define PACKET_TAG_GIF 8 /* GIF processing done */
1048 #define PACKET_TAG_GRE 9 /* GRE processing done */
1049 #define PACKET_TAG_IN_PACKET_CHECKSUM 10 /* NIC checksumming done */
1050 #define PACKET_TAG_ENCAP 11 /* Encap. processing */
1051 #define PACKET_TAG_IPSEC_SOCKET 12 /* IPSEC socket ref */
1052 #define PACKET_TAG_IPSEC_HISTORY 13 /* IPSEC history */
1053 #define PACKET_TAG_IPV6_INPUT 14 /* IPV6 input processing */
1054 #define PACKET_TAG_DUMMYNET 15 /* dummynet info */
1055 #define PACKET_TAG_DIVERT 17 /* divert info */
1056 #define PACKET_TAG_IPFORWARD 18 /* ipforward info */
1057 #define PACKET_TAG_MACLABEL (19 | MTAG_PERSISTENT) /* MAC label */
1058 #define PACKET_TAG_PF (21 | MTAG_PERSISTENT) /* PF/ALTQ information */
1059 #define PACKET_TAG_RTSOCKFAM 25 /* rtsock sa family */
1060 #define PACKET_TAG_IPOPTIONS 27 /* Saved IP options */
1061 #define PACKET_TAG_CARP 28 /* CARP info */
1062 #define PACKET_TAG_IPSEC_NAT_T_PORTS 29 /* two uint16_t */
1063 #define PACKET_TAG_ND_OUTGOING 30 /* ND outgoing */
1065 /* Specific cookies and tags. */
1067 /* Packet tag routines. */
1068 struct m_tag *m_tag_alloc(u_int32_t, int, int, int);
1069 void m_tag_delete(struct mbuf *, struct m_tag *);
1070 void m_tag_delete_chain(struct mbuf *, struct m_tag *);
1071 void m_tag_free_default(struct m_tag *);
1072 struct m_tag *m_tag_locate(struct mbuf *, u_int32_t, int, struct m_tag *);
1073 struct m_tag *m_tag_copy(struct m_tag *, int);
1074 int m_tag_copy_chain(struct mbuf *, struct mbuf *, int);
1075 void m_tag_delete_nonpersistent(struct mbuf *);
1078 * Initialize the list of tags associated with an mbuf.
1080 static __inline void
1081 m_tag_init(struct mbuf *m)
1084 SLIST_INIT(&m->m_pkthdr.tags);
1088 * Set up the contents of a tag. Note that this does not fill in the free
1089 * method; the caller is expected to do that.
1091 * XXX probably should be called m_tag_init, but that was already taken.
1093 static __inline void
1094 m_tag_setup(struct m_tag *t, u_int32_t cookie, int type, int len)
1099 t->m_tag_cookie = cookie;
1103 * Reclaim resources associated with a tag.
1105 static __inline void
1106 m_tag_free(struct m_tag *t)
1109 (*t->m_tag_free)(t);
1113 * Return the first tag associated with an mbuf.
1115 static __inline struct m_tag *
1116 m_tag_first(struct mbuf *m)
1119 return (SLIST_FIRST(&m->m_pkthdr.tags));
1123 * Return the next tag in the list of tags associated with an mbuf.
1125 static __inline struct m_tag *
1126 m_tag_next(struct mbuf *m __unused, struct m_tag *t)
1129 return (SLIST_NEXT(t, m_tag_link));
1133 * Prepend a tag to the list of tags associated with an mbuf.
1135 static __inline void
1136 m_tag_prepend(struct mbuf *m, struct m_tag *t)
1139 SLIST_INSERT_HEAD(&m->m_pkthdr.tags, t, m_tag_link);
1143 * Unlink a tag from the list of tags associated with an mbuf.
1145 static __inline void
1146 m_tag_unlink(struct mbuf *m, struct m_tag *t)
1149 SLIST_REMOVE(&m->m_pkthdr.tags, t, m_tag, m_tag_link);
1152 /* These are for OpenBSD compatibility. */
1153 #define MTAG_ABI_COMPAT 0 /* compatibility ABI */
1155 static __inline struct m_tag *
1156 m_tag_get(int type, int length, int wait)
1158 return (m_tag_alloc(MTAG_ABI_COMPAT, type, length, wait));
1161 static __inline struct m_tag *
1162 m_tag_find(struct mbuf *m, int type, struct m_tag *start)
1164 return (SLIST_EMPTY(&m->m_pkthdr.tags) ? (struct m_tag *)NULL :
1165 m_tag_locate(m, MTAG_ABI_COMPAT, type, start));
1168 static __inline struct mbuf *
1169 m_free(struct mbuf *m)
1171 struct mbuf *n = m->m_next;
1173 if ((m->m_flags & (M_PKTHDR|M_NOFREE)) == (M_PKTHDR|M_NOFREE))
1174 m_tag_delete_chain(m, NULL);
1175 if (m->m_flags & M_EXT)
1177 else if ((m->m_flags & M_NOFREE) == 0)
1178 uma_zfree(zone_mbuf, m);
1183 rt_m_getfib(struct mbuf *m)
1185 KASSERT(m->m_flags & M_PKTHDR , ("Attempt to get FIB from non header mbuf."));
1186 return (m->m_pkthdr.fibnum);
1189 #define M_GETFIB(_m) rt_m_getfib(_m)
1191 #define M_SETFIB(_m, _fib) do { \
1192 KASSERT((_m)->m_flags & M_PKTHDR, ("Attempt to set FIB on non header mbuf.")); \
1193 ((_m)->m_pkthdr.fibnum) = (_fib); \
1196 /* flags passed as first argument for "m_ether_tcpip_hash()" */
1197 #define MBUF_HASHFLAG_L2 (1 << 2)
1198 #define MBUF_HASHFLAG_L3 (1 << 3)
1199 #define MBUF_HASHFLAG_L4 (1 << 4)
1201 /* mbuf hashing helper routines */
1202 uint32_t m_ether_tcpip_hash_init(void);
1203 uint32_t m_ether_tcpip_hash(const uint32_t, const struct mbuf *, const uint32_t);
1205 #ifdef MBUF_PROFILING
1206 void m_profile(struct mbuf *m);
1207 #define M_PROFILE(m) m_profile(m)
1209 #define M_PROFILE(m)
1213 STAILQ_HEAD(, mbuf) mq_head;
1219 mbufq_init(struct mbufq *mq, int maxlen)
1222 STAILQ_INIT(&mq->mq_head);
1223 mq->mq_maxlen = maxlen;
1227 static inline struct mbuf *
1228 mbufq_flush(struct mbufq *mq)
1232 m = STAILQ_FIRST(&mq->mq_head);
1233 STAILQ_INIT(&mq->mq_head);
1239 mbufq_drain(struct mbufq *mq)
1243 n = mbufq_flush(mq);
1244 while ((m = n) != NULL) {
1245 n = STAILQ_NEXT(m, m_stailqpkt);
1250 static inline struct mbuf *
1251 mbufq_first(const struct mbufq *mq)
1254 return (STAILQ_FIRST(&mq->mq_head));
1257 static inline struct mbuf *
1258 mbufq_last(const struct mbufq *mq)
1261 return (STAILQ_LAST(&mq->mq_head, mbuf, m_stailqpkt));
1265 mbufq_full(const struct mbufq *mq)
1268 return (mq->mq_len >= mq->mq_maxlen);
1272 mbufq_len(const struct mbufq *mq)
1275 return (mq->mq_len);
1279 mbufq_enqueue(struct mbufq *mq, struct mbuf *m)
1284 STAILQ_INSERT_TAIL(&mq->mq_head, m, m_stailqpkt);
1289 static inline struct mbuf *
1290 mbufq_dequeue(struct mbufq *mq)
1294 m = STAILQ_FIRST(&mq->mq_head);
1296 STAILQ_REMOVE_HEAD(&mq->mq_head, m_stailqpkt);
1297 m->m_nextpkt = NULL;
1304 mbufq_prepend(struct mbufq *mq, struct mbuf *m)
1307 STAILQ_INSERT_HEAD(&mq->mq_head, m, m_stailqpkt);
1310 #endif /* _KERNEL */
1311 #endif /* !_SYS_MBUF_H_ */