2 * SPDX-License-Identifier: BSD-3-Clause
4 * Copyright (c) 1982, 1986, 1988, 1993
5 * The Regents of the University of California.
8 * Redistribution and use in source and binary forms, with or without
9 * modification, are permitted provided that the following conditions
11 * 1. Redistributions of source code must retain the above copyright
12 * notice, this list of conditions and the following disclaimer.
13 * 2. Redistributions in binary form must reproduce the above copyright
14 * notice, this list of conditions and the following disclaimer in the
15 * documentation and/or other materials provided with the distribution.
16 * 3. Neither the name of the University nor the names of its contributors
17 * may be used to endorse or promote products derived from this software
18 * without specific prior written permission.
20 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
21 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
22 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
23 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
24 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
25 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
26 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
27 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
28 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
29 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
32 * @(#)mbuf.h 8.5 (Berkeley) 2/19/95
39 /* XXX: These includes suck. Sorry! */
40 #include <sys/queue.h>
42 #include <sys/systm.h>
43 #include <sys/refcount.h>
53 #define MBUF_PROBE1(probe, arg0) \
54 SDT_PROBE1(sdt, , , probe, arg0)
55 #define MBUF_PROBE2(probe, arg0, arg1) \
56 SDT_PROBE2(sdt, , , probe, arg0, arg1)
57 #define MBUF_PROBE3(probe, arg0, arg1, arg2) \
58 SDT_PROBE3(sdt, , , probe, arg0, arg1, arg2)
59 #define MBUF_PROBE4(probe, arg0, arg1, arg2, arg3) \
60 SDT_PROBE4(sdt, , , probe, arg0, arg1, arg2, arg3)
61 #define MBUF_PROBE5(probe, arg0, arg1, arg2, arg3, arg4) \
62 SDT_PROBE5(sdt, , , probe, arg0, arg1, arg2, arg3, arg4)
64 SDT_PROBE_DECLARE(sdt, , , m__init);
65 SDT_PROBE_DECLARE(sdt, , , m__gethdr);
66 SDT_PROBE_DECLARE(sdt, , , m__get);
67 SDT_PROBE_DECLARE(sdt, , , m__getcl);
68 SDT_PROBE_DECLARE(sdt, , , m__clget);
69 SDT_PROBE_DECLARE(sdt, , , m__cljget);
70 SDT_PROBE_DECLARE(sdt, , , m__cljset);
71 SDT_PROBE_DECLARE(sdt, , , m__free);
72 SDT_PROBE_DECLARE(sdt, , , m__freem);
77 * Mbufs are of a single size, MSIZE (sys/param.h), which includes overhead.
78 * An mbuf may add a single "mbuf cluster" of size MCLBYTES (also in
79 * sys/param.h), which has no additional overhead and is used instead of the
80 * internal data area; this is done when at least MINCLSIZE of data must be
81 * stored. Additionally, it is possible to allocate a separate buffer
82 * externally and attach it to the mbuf in a way similar to that of mbuf
85 * NB: These calculation do not take actual compiler-induced alignment and
86 * padding inside the complete struct mbuf into account. Appropriate
87 * attention is required when changing members of struct mbuf.
89 * MLEN is data length in a normal mbuf.
90 * MHLEN is data length in an mbuf with pktheader.
91 * MINCLSIZE is a smallest amount of data that should be put into cluster.
93 * Compile-time assertions in uipc_mbuf.c test these values to ensure that
97 #define MHSIZE offsetof(struct mbuf, m_dat)
98 #define MPKTHSIZE offsetof(struct mbuf, m_pktdat)
99 #define MLEN ((int)(MSIZE - MHSIZE))
100 #define MHLEN ((int)(MSIZE - MPKTHSIZE))
101 #define MINCLSIZE (MHLEN + 1)
106 * Macro for type conversion: convert mbuf pointer to data pointer of correct
109 * mtod(m, t) -- Convert mbuf pointer to data pointer of correct type.
110 * mtodo(m, o) -- Same as above but with offset 'o' into data.
112 #define mtod(m, t) ((t)((m)->m_data))
113 #define mtodo(m, o) ((void *)(((m)->m_data) + (o)))
116 * Argument structure passed to UMA routines during mbuf and packet
120 int flags; /* Flags for mbuf being allocated */
121 short type; /* Type of mbuf being allocated */
126 * Packet tag structure (see below for details).
129 SLIST_ENTRY(m_tag) m_tag_link; /* List of packet tags */
130 u_int16_t m_tag_id; /* Tag ID */
131 u_int16_t m_tag_len; /* Length of data */
132 u_int32_t m_tag_cookie; /* ABI/Module ID */
133 void (*m_tag_free)(struct m_tag *);
137 * Static network interface owned tag.
138 * Allocated through ifp->if_snd_tag_alloc().
141 struct ifnet *ifp; /* network interface tag belongs to */
142 volatile u_int refcount;
146 * Record/packet header in first mbuf of chain; valid only if M_PKTHDR is set.
149 * Compile-time assertions in uipc_mbuf.c test these values to ensure that
154 struct m_snd_tag *snd_tag; /* send tag, if any */
155 struct ifnet *rcvif; /* rcv interface */
157 SLIST_HEAD(packet_tags, m_tag) tags; /* list of packet tags */
158 int32_t len; /* total packet length */
160 /* Layer crossing persistent information. */
161 uint32_t flowid; /* packet's 4-tuple system */
162 uint32_t csum_flags; /* checksum and offload features */
163 uint16_t fibnum; /* this packet should use this fib */
164 uint8_t numa_domain; /* NUMA domain of recvd pkt */
165 uint8_t rsstype; /* hash type */
167 uint64_t rcv_tstmp; /* timestamp in ns */
169 uint8_t l2hlen; /* layer 2 hdr len */
170 uint8_t l3hlen; /* layer 3 hdr len */
171 uint8_t l4hlen; /* layer 4 hdr len */
172 uint8_t l5hlen; /* layer 5 hdr len */
179 uint32_t thirtytwo[2];
180 uint64_t sixtyfour[1];
181 uintptr_t unintptr[1];
185 /* Layer specific non-persistent local storage for reassembly, etc. */
189 uint32_t thirtytwo[2];
190 uint64_t sixtyfour[1];
191 uintptr_t unintptr[1];
195 #define ether_vtag PH_per.sixteen[0]
197 #define vt_nrecs sixteen[0]
198 #define tso_segsz PH_per.sixteen[1]
199 #define lro_nsegs tso_segsz
200 #define csum_phsum PH_per.sixteen[2]
201 #define csum_data PH_per.thirtytwo[1]
202 #define pace_thoff PH_loc.sixteen[0]
203 #define pace_tlen PH_loc.sixteen[1]
204 #define pace_drphdrlen PH_loc.sixteen[2]
205 #define pace_tos PH_loc.eight[6]
206 #define pace_lock PH_loc.eight[7]
209 * Description of external storage mapped into mbuf; valid only if M_EXT is
213 * Compile-time assertions in uipc_mbuf.c test these values to ensure that
216 typedef void m_ext_free_t(struct mbuf *);
220 * If EXT_FLAG_EMBREF is set, then we use refcount in the
221 * mbuf, the 'ext_count' member. Otherwise, we have a
222 * shadow copy and we use pointer 'ext_cnt'. The original
223 * mbuf is responsible to carry the pointer to free routine
224 * and its arguments. They aren't copied into shadows in
225 * mb_dupcl() to avoid dereferencing next cachelines.
227 volatile u_int ext_count;
228 volatile u_int *ext_cnt;
230 char *ext_buf; /* start of buffer */
231 uint32_t ext_size; /* size of buffer, for ext_free */
232 uint32_t ext_type:8, /* type of external storage */
233 ext_flags:24; /* external storage mbuf flags */
235 * Fields below store the free context for the external storage.
236 * They are valid only in the refcount carrying mbuf, the one with
237 * EXT_FLAG_EMBREF flag, with exclusion for EXT_EXTREF type, where
238 * the free context is copied into all mbufs that use same external
241 #define m_ext_copylen offsetof(struct m_ext, ext_free)
242 m_ext_free_t *ext_free; /* free routine if not the usual */
243 void *ext_arg1; /* optional argument pointer */
244 void *ext_arg2; /* optional argument pointer */
248 * The core of the mbuf object along with some shortcut defines for practical
253 * Header present at the beginning of every mbuf.
256 * Compile-time assertions in uipc_mbuf.c test these values to ensure
257 * that they are correct.
259 union { /* next buffer in chain */
261 SLIST_ENTRY(mbuf) m_slist;
262 STAILQ_ENTRY(mbuf) m_stailq;
264 union { /* next chain in queue/record */
265 struct mbuf *m_nextpkt;
266 SLIST_ENTRY(mbuf) m_slistpkt;
267 STAILQ_ENTRY(mbuf) m_stailqpkt;
269 caddr_t m_data; /* location of data */
270 int32_t m_len; /* amount of data in this mbuf */
271 uint32_t m_type:8, /* type of data in this mbuf */
272 m_flags:24; /* flags; see below */
273 #if !defined(__LP64__)
274 uint32_t m_pad; /* pad for 64bit alignment */
278 * A set of optional headers (packet header, external storage header)
279 * and internal data storage. Historically, these arrays were sized
280 * to MHLEN (space left after a packet header) and MLEN (space left
281 * after only a regular mbuf header); they are now variable size in
282 * order to support future work on variable-size mbufs.
286 struct pkthdr m_pkthdr; /* M_PKTHDR set */
288 struct m_ext m_ext; /* M_EXT set */
292 char m_dat[0]; /* !M_PKTHDR, !M_EXT */
297 * mbuf flags of global significance and layer crossing.
298 * Those of only protocol/layer specific significance are to be mapped
299 * to M_PROTO[1-12] and cleared at layer handoff boundaries.
300 * NB: Limited to the lower 24 bits.
302 #define M_EXT 0x00000001 /* has associated external storage */
303 #define M_PKTHDR 0x00000002 /* start of record */
304 #define M_EOR 0x00000004 /* end of record */
305 #define M_RDONLY 0x00000008 /* associated data is marked read-only */
306 #define M_BCAST 0x00000010 /* send/received as link-level broadcast */
307 #define M_MCAST 0x00000020 /* send/received as link-level multicast */
308 #define M_PROMISC 0x00000040 /* packet was not for us */
309 #define M_VLANTAG 0x00000080 /* ether_vtag is valid */
310 #define M_NOMAP 0x00000100 /* mbuf data is unmapped (soon from Drew) */
311 #define M_NOFREE 0x00000200 /* do not free mbuf, embedded in cluster */
312 #define M_TSTMP 0x00000400 /* rcv_tstmp field is valid */
313 #define M_TSTMP_HPREC 0x00000800 /* rcv_tstmp is high-prec, typically
314 hw-stamped on port (useful for IEEE 1588
317 #define M_PROTO1 0x00001000 /* protocol-specific */
318 #define M_PROTO2 0x00002000 /* protocol-specific */
319 #define M_PROTO3 0x00004000 /* protocol-specific */
320 #define M_PROTO4 0x00008000 /* protocol-specific */
321 #define M_PROTO5 0x00010000 /* protocol-specific */
322 #define M_PROTO6 0x00020000 /* protocol-specific */
323 #define M_PROTO7 0x00040000 /* protocol-specific */
324 #define M_PROTO8 0x00080000 /* protocol-specific */
325 #define M_PROTO9 0x00100000 /* protocol-specific */
326 #define M_PROTO10 0x00200000 /* protocol-specific */
327 #define M_PROTO11 0x00400000 /* protocol-specific */
328 #define M_PROTO12 0x00800000 /* protocol-specific */
330 #define MB_DTOR_SKIP 0x1 /* don't pollute the cache by touching a freed mbuf */
333 * Flags to purge when crossing layers.
335 #define M_PROTOFLAGS \
336 (M_PROTO1|M_PROTO2|M_PROTO3|M_PROTO4|M_PROTO5|M_PROTO6|M_PROTO7|M_PROTO8|\
337 M_PROTO9|M_PROTO10|M_PROTO11|M_PROTO12)
340 * Flags preserved when copying m_pkthdr.
342 #define M_COPYFLAGS \
343 (M_PKTHDR|M_EOR|M_RDONLY|M_BCAST|M_MCAST|M_PROMISC|M_VLANTAG|M_TSTMP| \
344 M_TSTMP_HPREC|M_PROTOFLAGS)
347 * Mbuf flag description for use with printf(9) %b identifier.
349 #define M_FLAG_BITS \
350 "\20\1M_EXT\2M_PKTHDR\3M_EOR\4M_RDONLY\5M_BCAST\6M_MCAST" \
351 "\7M_PROMISC\10M_VLANTAG\12M_NOFREE\13M_TSTMP\14M_TSTMP_HPREC"
352 #define M_FLAG_PROTOBITS \
353 "\15M_PROTO1\16M_PROTO2\17M_PROTO3\20M_PROTO4\21M_PROTO5" \
354 "\22M_PROTO6\23M_PROTO7\24M_PROTO8\25M_PROTO9\26M_PROTO10" \
355 "\27M_PROTO11\30M_PROTO12"
356 #define M_FLAG_PRINTF (M_FLAG_BITS M_FLAG_PROTOBITS)
359 * Network interface cards are able to hash protocol fields (such as IPv4
360 * addresses and TCP port numbers) classify packets into flows. These flows
361 * can then be used to maintain ordering while delivering packets to the OS
362 * via parallel input queues, as well as to provide a stateless affinity
363 * model. NIC drivers can pass up the hash via m->m_pkthdr.flowid, and set
364 * m_flag fields to indicate how the hash should be interpreted by the
367 * Most NICs support RSS, which provides ordering and explicit affinity, and
368 * use the hash m_flag bits to indicate what header fields were covered by
369 * the hash. M_HASHTYPE_OPAQUE and M_HASHTYPE_OPAQUE_HASH can be set by non-
370 * RSS cards or configurations that provide an opaque flow identifier, allowing
371 * for ordering and distribution without explicit affinity. Additionally,
372 * M_HASHTYPE_OPAQUE_HASH indicates that the flow identifier has hash
375 * The meaning of the IPV6_EX suffix:
376 * "o Home address from the home address option in the IPv6 destination
377 * options header. If the extension header is not present, use the Source
379 * o IPv6 address that is contained in the Routing-Header-Type-2 from the
380 * associated extension header. If the extension header is not present,
381 * use the Destination IPv6 Address."
383 * https://docs.microsoft.com/en-us/windows-hardware/drivers/network/rss-hashing-types#ndishashipv6ex
385 #define M_HASHTYPE_HASHPROP 0x80 /* has hash properties */
386 #define M_HASHTYPE_HASH(t) (M_HASHTYPE_HASHPROP | (t))
387 /* Microsoft RSS standard hash types */
388 #define M_HASHTYPE_NONE 0
389 #define M_HASHTYPE_RSS_IPV4 M_HASHTYPE_HASH(1) /* IPv4 2-tuple */
390 #define M_HASHTYPE_RSS_TCP_IPV4 M_HASHTYPE_HASH(2) /* TCPv4 4-tuple */
391 #define M_HASHTYPE_RSS_IPV6 M_HASHTYPE_HASH(3) /* IPv6 2-tuple */
392 #define M_HASHTYPE_RSS_TCP_IPV6 M_HASHTYPE_HASH(4) /* TCPv6 4-tuple */
393 #define M_HASHTYPE_RSS_IPV6_EX M_HASHTYPE_HASH(5) /* IPv6 2-tuple +
395 #define M_HASHTYPE_RSS_TCP_IPV6_EX M_HASHTYPE_HASH(6) /* TCPv6 4-tuple +
397 #define M_HASHTYPE_RSS_UDP_IPV4 M_HASHTYPE_HASH(7) /* IPv4 UDP 4-tuple*/
398 #define M_HASHTYPE_RSS_UDP_IPV6 M_HASHTYPE_HASH(9) /* IPv6 UDP 4-tuple*/
399 #define M_HASHTYPE_RSS_UDP_IPV6_EX M_HASHTYPE_HASH(10)/* IPv6 UDP 4-tuple +
402 #define M_HASHTYPE_OPAQUE 63 /* ordering, not affinity */
403 #define M_HASHTYPE_OPAQUE_HASH M_HASHTYPE_HASH(M_HASHTYPE_OPAQUE)
404 /* ordering+hash, not affinity*/
406 #define M_HASHTYPE_CLEAR(m) ((m)->m_pkthdr.rsstype = 0)
407 #define M_HASHTYPE_GET(m) ((m)->m_pkthdr.rsstype)
408 #define M_HASHTYPE_SET(m, v) ((m)->m_pkthdr.rsstype = (v))
409 #define M_HASHTYPE_TEST(m, v) (M_HASHTYPE_GET(m) == (v))
410 #define M_HASHTYPE_ISHASH(m) (M_HASHTYPE_GET(m) & M_HASHTYPE_HASHPROP)
413 * External mbuf storage buffer types.
415 #define EXT_CLUSTER 1 /* mbuf cluster */
416 #define EXT_SFBUF 2 /* sendfile(2)'s sf_buf */
417 #define EXT_JUMBOP 3 /* jumbo cluster page sized */
418 #define EXT_JUMBO9 4 /* jumbo cluster 9216 bytes */
419 #define EXT_JUMBO16 5 /* jumbo cluster 16184 bytes */
420 #define EXT_PACKET 6 /* mbuf+cluster from packet zone */
421 #define EXT_MBUF 7 /* external mbuf reference */
422 #define EXT_RXRING 8 /* data in NIC receive ring */
424 #define EXT_VENDOR1 224 /* for vendor-internal use */
425 #define EXT_VENDOR2 225 /* for vendor-internal use */
426 #define EXT_VENDOR3 226 /* for vendor-internal use */
427 #define EXT_VENDOR4 227 /* for vendor-internal use */
429 #define EXT_EXP1 244 /* for experimental use */
430 #define EXT_EXP2 245 /* for experimental use */
431 #define EXT_EXP3 246 /* for experimental use */
432 #define EXT_EXP4 247 /* for experimental use */
434 #define EXT_NET_DRV 252 /* custom ext_buf provided by net driver(s) */
435 #define EXT_MOD_TYPE 253 /* custom module's ext_buf type */
436 #define EXT_DISPOSABLE 254 /* can throw this buffer away w/page flipping */
437 #define EXT_EXTREF 255 /* has externally maintained ext_cnt ptr */
440 * Flags for external mbuf buffer types.
441 * NB: limited to the lower 24 bits.
443 #define EXT_FLAG_EMBREF 0x000001 /* embedded ext_count */
444 #define EXT_FLAG_EXTREF 0x000002 /* external ext_cnt, notyet */
446 #define EXT_FLAG_NOFREE 0x000010 /* don't free mbuf to pool, notyet */
448 #define EXT_FLAG_VENDOR1 0x010000 /* These flags are vendor */
449 #define EXT_FLAG_VENDOR2 0x020000 /* or submodule specific, */
450 #define EXT_FLAG_VENDOR3 0x040000 /* not used by mbuf code. */
451 #define EXT_FLAG_VENDOR4 0x080000 /* Set/read by submodule. */
453 #define EXT_FLAG_EXP1 0x100000 /* for experimental use */
454 #define EXT_FLAG_EXP2 0x200000 /* for experimental use */
455 #define EXT_FLAG_EXP3 0x400000 /* for experimental use */
456 #define EXT_FLAG_EXP4 0x800000 /* for experimental use */
459 * EXT flag description for use with printf(9) %b identifier.
461 #define EXT_FLAG_BITS \
462 "\20\1EXT_FLAG_EMBREF\2EXT_FLAG_EXTREF\5EXT_FLAG_NOFREE" \
463 "\21EXT_FLAG_VENDOR1\22EXT_FLAG_VENDOR2\23EXT_FLAG_VENDOR3" \
464 "\24EXT_FLAG_VENDOR4\25EXT_FLAG_EXP1\26EXT_FLAG_EXP2\27EXT_FLAG_EXP3" \
468 * Flags indicating checksum, segmentation and other offload work to be
469 * done, or already done, by hardware or lower layers. It is split into
470 * separate inbound and outbound flags.
472 * Outbound flags that are set by upper protocol layers requesting lower
473 * layers, or ideally the hardware, to perform these offloading tasks.
474 * For outbound packets this field and its flags can be directly tested
475 * against ifnet if_hwassist.
477 #define CSUM_IP 0x00000001 /* IP header checksum offload */
478 #define CSUM_IP_UDP 0x00000002 /* UDP checksum offload */
479 #define CSUM_IP_TCP 0x00000004 /* TCP checksum offload */
480 #define CSUM_IP_SCTP 0x00000008 /* SCTP checksum offload */
481 #define CSUM_IP_TSO 0x00000010 /* TCP segmentation offload */
482 #define CSUM_IP_ISCSI 0x00000020 /* iSCSI checksum offload */
484 #define CSUM_IP6_UDP 0x00000200 /* UDP checksum offload */
485 #define CSUM_IP6_TCP 0x00000400 /* TCP checksum offload */
486 #define CSUM_IP6_SCTP 0x00000800 /* SCTP checksum offload */
487 #define CSUM_IP6_TSO 0x00001000 /* TCP segmentation offload */
488 #define CSUM_IP6_ISCSI 0x00002000 /* iSCSI checksum offload */
490 /* Inbound checksum support where the checksum was verified by hardware. */
491 #define CSUM_L3_CALC 0x01000000 /* calculated layer 3 csum */
492 #define CSUM_L3_VALID 0x02000000 /* checksum is correct */
493 #define CSUM_L4_CALC 0x04000000 /* calculated layer 4 csum */
494 #define CSUM_L4_VALID 0x08000000 /* checksum is correct */
495 #define CSUM_L5_CALC 0x10000000 /* calculated layer 5 csum */
496 #define CSUM_L5_VALID 0x20000000 /* checksum is correct */
497 #define CSUM_COALESCED 0x40000000 /* contains merged segments */
499 #define CSUM_SND_TAG 0x80000000 /* Packet header has send tag */
502 * CSUM flag description for use with printf(9) %b identifier.
505 "\20\1CSUM_IP\2CSUM_IP_UDP\3CSUM_IP_TCP\4CSUM_IP_SCTP\5CSUM_IP_TSO" \
507 "\12CSUM_IP6_UDP\13CSUM_IP6_TCP\14CSUM_IP6_SCTP\15CSUM_IP6_TSO" \
508 "\16CSUM_IP6_ISCSI" \
509 "\31CSUM_L3_CALC\32CSUM_L3_VALID\33CSUM_L4_CALC\34CSUM_L4_VALID" \
510 "\35CSUM_L5_CALC\36CSUM_L5_VALID\37CSUM_COALESCED\40CSUM_SND_TAG"
512 /* CSUM flags compatibility mappings. */
513 #define CSUM_IP_CHECKED CSUM_L3_CALC
514 #define CSUM_IP_VALID CSUM_L3_VALID
515 #define CSUM_DATA_VALID CSUM_L4_VALID
516 #define CSUM_PSEUDO_HDR CSUM_L4_CALC
517 #define CSUM_SCTP_VALID CSUM_L4_VALID
518 #define CSUM_DELAY_DATA (CSUM_TCP|CSUM_UDP)
519 #define CSUM_DELAY_IP CSUM_IP /* Only v4, no v6 IP hdr csum */
520 #define CSUM_DELAY_DATA_IPV6 (CSUM_TCP_IPV6|CSUM_UDP_IPV6)
521 #define CSUM_DATA_VALID_IPV6 CSUM_DATA_VALID
522 #define CSUM_TCP CSUM_IP_TCP
523 #define CSUM_UDP CSUM_IP_UDP
524 #define CSUM_SCTP CSUM_IP_SCTP
525 #define CSUM_TSO (CSUM_IP_TSO|CSUM_IP6_TSO)
526 #define CSUM_UDP_IPV6 CSUM_IP6_UDP
527 #define CSUM_TCP_IPV6 CSUM_IP6_TCP
528 #define CSUM_SCTP_IPV6 CSUM_IP6_SCTP
531 * mbuf types describing the content of the mbuf (including external storage).
533 #define MT_NOTMBUF 0 /* USED INTERNALLY ONLY! Object is not mbuf */
534 #define MT_DATA 1 /* dynamic (data) allocation */
535 #define MT_HEADER MT_DATA /* packet header, use M_PKTHDR instead */
537 #define MT_VENDOR1 4 /* for vendor-internal use */
538 #define MT_VENDOR2 5 /* for vendor-internal use */
539 #define MT_VENDOR3 6 /* for vendor-internal use */
540 #define MT_VENDOR4 7 /* for vendor-internal use */
542 #define MT_SONAME 8 /* socket name */
544 #define MT_EXP1 9 /* for experimental use */
545 #define MT_EXP2 10 /* for experimental use */
546 #define MT_EXP3 11 /* for experimental use */
547 #define MT_EXP4 12 /* for experimental use */
549 #define MT_CONTROL 14 /* extra-data protocol message */
550 #define MT_EXTCONTROL 15 /* control message with externalized contents */
551 #define MT_OOBDATA 16 /* expedited data */
553 #define MT_NOINIT 255 /* Not a type but a flag to allocate
554 a non-initialized mbuf */
557 * String names of mbuf-related UMA(9) and malloc(9) types. Exposed to
558 * !_KERNEL so that monitoring tools can look up the zones with
561 #define MBUF_MEM_NAME "mbuf"
562 #define MBUF_CLUSTER_MEM_NAME "mbuf_cluster"
563 #define MBUF_PACKET_MEM_NAME "mbuf_packet"
564 #define MBUF_JUMBOP_MEM_NAME "mbuf_jumbo_page"
565 #define MBUF_JUMBO9_MEM_NAME "mbuf_jumbo_9k"
566 #define MBUF_JUMBO16_MEM_NAME "mbuf_jumbo_16k"
567 #define MBUF_TAG_MEM_NAME "mbuf_tag"
568 #define MBUF_EXTREFCNT_MEM_NAME "mbuf_ext_refcnt"
573 #define MBUF_CHECKSLEEP(how) do { \
574 if (how == M_WAITOK) \
575 WITNESS_WARN(WARN_GIANTOK | WARN_SLEEPOK, NULL, \
576 "Sleeping in \"%s\"", __func__); \
579 #define MBUF_CHECKSLEEP(how)
583 * Network buffer allocation API
585 * The rest of it is defined in kern/kern_mbuf.c
587 extern uma_zone_t zone_mbuf;
588 extern uma_zone_t zone_clust;
589 extern uma_zone_t zone_pack;
590 extern uma_zone_t zone_jumbop;
591 extern uma_zone_t zone_jumbo9;
592 extern uma_zone_t zone_jumbo16;
594 void mb_dupcl(struct mbuf *, struct mbuf *);
595 void mb_free_ext(struct mbuf *);
596 void m_adj(struct mbuf *, int);
597 int m_apply(struct mbuf *, int, int,
598 int (*)(void *, void *, u_int), void *);
599 int m_append(struct mbuf *, int, c_caddr_t);
600 void m_cat(struct mbuf *, struct mbuf *);
601 void m_catpkt(struct mbuf *, struct mbuf *);
602 int m_clget(struct mbuf *m, int how);
603 void *m_cljget(struct mbuf *m, int how, int size);
604 struct mbuf *m_collapse(struct mbuf *, int, int);
605 void m_copyback(struct mbuf *, int, int, c_caddr_t);
606 void m_copydata(const struct mbuf *, int, int, caddr_t);
607 struct mbuf *m_copym(struct mbuf *, int, int, int);
608 struct mbuf *m_copypacket(struct mbuf *, int);
609 void m_copy_pkthdr(struct mbuf *, struct mbuf *);
610 struct mbuf *m_copyup(struct mbuf *, int, int);
611 struct mbuf *m_defrag(struct mbuf *, int);
612 void m_demote_pkthdr(struct mbuf *);
613 void m_demote(struct mbuf *, int, int);
614 struct mbuf *m_devget(char *, int, int, struct ifnet *,
615 void (*)(char *, caddr_t, u_int));
616 void m_dispose_extcontrolm(struct mbuf *m);
617 struct mbuf *m_dup(const struct mbuf *, int);
618 int m_dup_pkthdr(struct mbuf *, const struct mbuf *, int);
619 void m_extadd(struct mbuf *, char *, u_int, m_ext_free_t,
620 void *, void *, int, int);
621 u_int m_fixhdr(struct mbuf *);
622 struct mbuf *m_fragment(struct mbuf *, int, int);
623 void m_freem(struct mbuf *);
624 struct mbuf *m_get2(int, int, short, int);
625 struct mbuf *m_getjcl(int, short, int, int);
626 struct mbuf *m_getm2(struct mbuf *, int, int, short, int);
627 struct mbuf *m_getptr(struct mbuf *, int, int *);
628 u_int m_length(struct mbuf *, struct mbuf **);
629 int m_mbuftouio(struct uio *, const struct mbuf *, int);
630 void m_move_pkthdr(struct mbuf *, struct mbuf *);
631 int m_pkthdr_init(struct mbuf *, int);
632 struct mbuf *m_prepend(struct mbuf *, int, int);
633 void m_print(const struct mbuf *, int);
634 struct mbuf *m_pulldown(struct mbuf *, int, int, int *);
635 struct mbuf *m_pullup(struct mbuf *, int);
636 int m_sanity(struct mbuf *, int);
637 struct mbuf *m_split(struct mbuf *, int, int);
638 struct mbuf *m_uiotombuf(struct uio *, int, int, int, int);
639 struct mbuf *m_unshare(struct mbuf *, int);
640 void m_snd_tag_init(struct m_snd_tag *, struct ifnet *);
641 void m_snd_tag_destroy(struct m_snd_tag *);
655 #if MJUMPAGESIZE != MCLBYTES
667 panic("%s: invalid cluster size %d", __func__, size);
674 * Associated an external reference counted buffer with an mbuf.
677 m_extaddref(struct mbuf *m, char *buf, u_int size, u_int *ref_cnt,
678 m_ext_free_t freef, void *arg1, void *arg2)
681 KASSERT(ref_cnt != NULL, ("%s: ref_cnt not provided", __func__));
683 atomic_add_int(ref_cnt, 1);
685 m->m_ext.ext_buf = buf;
686 m->m_ext.ext_cnt = ref_cnt;
687 m->m_data = m->m_ext.ext_buf;
688 m->m_ext.ext_size = size;
689 m->m_ext.ext_free = freef;
690 m->m_ext.ext_arg1 = arg1;
691 m->m_ext.ext_arg2 = arg2;
692 m->m_ext.ext_type = EXT_EXTREF;
693 m->m_ext.ext_flags = 0;
696 static __inline uma_zone_t
705 #if MJUMPAGESIZE != MCLBYTES
717 panic("%s: invalid cluster size %d", __func__, size);
724 * Initialize an mbuf with linear storage.
726 * Inline because the consumer text overhead will be roughly the same to
727 * initialize or call a function with this many parameters and M_PKTHDR
728 * should go away with constant propagation for !MGETHDR.
731 m_init(struct mbuf *m, int how, short type, int flags)
737 m->m_data = m->m_dat;
741 if (flags & M_PKTHDR)
742 error = m_pkthdr_init(m, how);
746 MBUF_PROBE5(m__init, m, how, type, flags, error);
750 static __inline struct mbuf *
751 m_get(int how, short type)
758 m = uma_zalloc_arg(zone_mbuf, &args, how);
759 MBUF_PROBE3(m__get, how, type, m);
763 static __inline struct mbuf *
764 m_gethdr(int how, short type)
769 args.flags = M_PKTHDR;
771 m = uma_zalloc_arg(zone_mbuf, &args, how);
772 MBUF_PROBE3(m__gethdr, how, type, m);
776 static __inline struct mbuf *
777 m_getcl(int how, short type, int flags)
784 m = uma_zalloc_arg(zone_pack, &args, how);
785 MBUF_PROBE4(m__getcl, how, type, flags, m);
790 * XXX: m_cljset() is a dangerous API. One must attach only a new,
791 * unreferenced cluster to an mbuf(9). It is not possible to assert
792 * that, so care can be taken only by users of the API.
795 m_cljset(struct mbuf *m, void *cl, int type)
803 #if MJUMPAGESIZE != MCLBYTES
815 panic("%s: unknown cluster type %d", __func__, type);
819 m->m_data = m->m_ext.ext_buf = cl;
820 m->m_ext.ext_free = m->m_ext.ext_arg1 = m->m_ext.ext_arg2 = NULL;
821 m->m_ext.ext_size = size;
822 m->m_ext.ext_type = type;
823 m->m_ext.ext_flags = EXT_FLAG_EMBREF;
824 m->m_ext.ext_count = 1;
826 MBUF_PROBE3(m__cljset, m, cl, type);
830 m_chtype(struct mbuf *m, short new_type)
833 m->m_type = new_type;
837 m_clrprotoflags(struct mbuf *m)
841 m->m_flags &= ~M_PROTOFLAGS;
846 static __inline struct mbuf *
847 m_last(struct mbuf *m)
856 m_extrefcnt(struct mbuf *m)
859 KASSERT(m->m_flags & M_EXT, ("%s: M_EXT missing", __func__));
861 return ((m->m_ext.ext_flags & EXT_FLAG_EMBREF) ? m->m_ext.ext_count :
866 * mbuf, cluster, and external object allocation macros (for compatibility
869 #define M_MOVE_PKTHDR(to, from) m_move_pkthdr((to), (from))
870 #define MGET(m, how, type) ((m) = m_get((how), (type)))
871 #define MGETHDR(m, how, type) ((m) = m_gethdr((how), (type)))
872 #define MCLGET(m, how) m_clget((m), (how))
873 #define MEXTADD(m, buf, size, free, arg1, arg2, flags, type) \
874 m_extadd((m), (char *)(buf), (size), (free), (arg1), (arg2), \
876 #define m_getm(m, len, how, type) \
877 m_getm2((m), (len), (how), (type), M_PKTHDR)
880 * Evaluate TRUE if it's safe to write to the mbuf m's data region (this can
881 * be both the local data payload, or an external buffer area, depending on
882 * whether M_EXT is set).
884 #define M_WRITABLE(m) (!((m)->m_flags & M_RDONLY) && \
885 (!(((m)->m_flags & M_EXT)) || \
886 (m_extrefcnt(m) == 1)))
888 /* Check if the supplied mbuf has a packet header, or else panic. */
889 #define M_ASSERTPKTHDR(m) \
890 KASSERT((m) != NULL && (m)->m_flags & M_PKTHDR, \
891 ("%s: no mbuf packet header!", __func__))
894 * Ensure that the supplied mbuf is a valid, non-free mbuf.
896 * XXX: Broken at the moment. Need some UMA magic to make it work again.
898 #define M_ASSERTVALID(m) \
899 KASSERT((((struct mbuf *)m)->m_flags & 0) == 0, \
900 ("%s: attempted use of a free mbuf!", __func__))
903 * Return the address of the start of the buffer associated with an mbuf,
904 * handling external storage, packet-header mbufs, and regular data mbufs.
907 (((m)->m_flags & M_EXT) ? (m)->m_ext.ext_buf : \
908 ((m)->m_flags & M_PKTHDR) ? &(m)->m_pktdat[0] : \
912 * Return the size of the buffer associated with an mbuf, handling external
913 * storage, packet-header mbufs, and regular data mbufs.
916 (((m)->m_flags & M_EXT) ? (m)->m_ext.ext_size : \
917 ((m)->m_flags & M_PKTHDR) ? MHLEN : \
921 * Set the m_data pointer of a newly allocated mbuf to place an object of the
922 * specified size at the end of the mbuf, longword aligned.
924 * NB: Historically, we had M_ALIGN(), MH_ALIGN(), and MEXT_ALIGN() as
925 * separate macros, each asserting that it was called at the proper moment.
926 * This required callers to themselves test the storage type and call the
927 * right one. Rather than require callers to be aware of those layout
928 * decisions, we centralize here.
931 m_align(struct mbuf *m, int len)
934 const char *msg = "%s: not a virgin mbuf";
938 KASSERT(m->m_data == M_START(m), (msg, __func__));
940 adjust = M_SIZE(m) - len;
941 m->m_data += adjust &~ (sizeof(long)-1);
944 #define M_ALIGN(m, len) m_align(m, len)
945 #define MH_ALIGN(m, len) m_align(m, len)
946 #define MEXT_ALIGN(m, len) m_align(m, len)
949 * Compute the amount of space available before the current start of data in
952 * The M_WRITABLE() is a temporary, conservative safety measure: the burden
953 * of checking writability of the mbuf data area rests solely with the caller.
955 * NB: In previous versions, M_LEADINGSPACE() would only check M_WRITABLE()
956 * for mbufs with external storage. We now allow mbuf-embedded data to be
959 #define M_LEADINGSPACE(m) \
960 (M_WRITABLE(m) ? ((m)->m_data - M_START(m)) : 0)
963 * Compute the amount of space available after the end of data in an mbuf.
965 * The M_WRITABLE() is a temporary, conservative safety measure: the burden
966 * of checking writability of the mbuf data area rests solely with the caller.
968 * NB: In previous versions, M_TRAILINGSPACE() would only check M_WRITABLE()
969 * for mbufs with external storage. We now allow mbuf-embedded data to be
972 #define M_TRAILINGSPACE(m) \
974 ((M_START(m) + M_SIZE(m)) - ((m)->m_data + (m)->m_len)) : 0)
977 * Arrange to prepend space of size plen to mbuf m. If a new mbuf must be
978 * allocated, how specifies whether to wait. If the allocation fails, the
979 * original mbuf chain is freed and m is set to NULL.
981 #define M_PREPEND(m, plen, how) do { \
982 struct mbuf **_mmp = &(m); \
983 struct mbuf *_mm = *_mmp; \
984 int _mplen = (plen); \
985 int __mhow = (how); \
987 MBUF_CHECKSLEEP(how); \
988 if (M_LEADINGSPACE(_mm) >= _mplen) { \
989 _mm->m_data -= _mplen; \
990 _mm->m_len += _mplen; \
992 _mm = m_prepend(_mm, _mplen, __mhow); \
993 if (_mm != NULL && _mm->m_flags & M_PKTHDR) \
994 _mm->m_pkthdr.len += _mplen; \
999 * Change mbuf to new type. This is a relatively expensive operation and
1000 * should be avoided.
1002 #define MCHTYPE(m, t) m_chtype((m), (t))
1004 /* Return the rcvif of a packet header. */
1005 static __inline struct ifnet *
1006 m_rcvif(struct mbuf *m)
1010 if (m->m_pkthdr.csum_flags & CSUM_SND_TAG)
1012 return (m->m_pkthdr.rcvif);
1015 /* Length to m_copy to copy all. */
1016 #define M_COPYALL 1000000000
1018 extern int max_datalen; /* MHLEN - max_hdr */
1019 extern int max_hdr; /* Largest link + protocol header */
1020 extern int max_linkhdr; /* Largest link-level header */
1021 extern int max_protohdr; /* Largest protocol header */
1022 extern int nmbclusters; /* Maximum number of clusters */
1025 * Network packets may have annotations attached by affixing a list of
1026 * "packet tags" to the pkthdr structure. Packet tags are dynamically
1027 * allocated semi-opaque data structures that have a fixed header
1028 * (struct m_tag) that specifies the size of the memory block and a
1029 * <cookie,type> pair that identifies it. The cookie is a 32-bit unique
1030 * unsigned value used to identify a module or ABI. By convention this value
1031 * is chosen as the date+time that the module is created, expressed as the
1032 * number of seconds since the epoch (e.g., using date -u +'%s'). The type
1033 * value is an ABI/module-specific value that identifies a particular
1034 * annotation and is private to the module. For compatibility with systems
1035 * like OpenBSD that define packet tags w/o an ABI/module cookie, the value
1036 * PACKET_ABI_COMPAT is used to implement m_tag_get and m_tag_find
1037 * compatibility shim functions and several tag types are defined below.
1038 * Users that do not require compatibility should use a private cookie value
1039 * so that packet tag-related definitions can be maintained privately.
1041 * Note that the packet tag returned by m_tag_alloc has the default memory
1042 * alignment implemented by malloc. To reference private data one can use a
1045 * struct m_tag *mtag = m_tag_alloc(...);
1046 * struct foo *p = (struct foo *)(mtag+1);
1048 * if the alignment of struct m_tag is sufficient for referencing members of
1049 * struct foo. Otherwise it is necessary to embed struct m_tag within the
1050 * private data structure to insure proper alignment; e.g.,
1056 * struct foo *p = (struct foo *) m_tag_alloc(...);
1057 * struct m_tag *mtag = &p->tag;
1061 * Persistent tags stay with an mbuf until the mbuf is reclaimed. Otherwise
1062 * tags are expected to ``vanish'' when they pass through a network
1063 * interface. For most interfaces this happens normally as the tags are
1064 * reclaimed when the mbuf is free'd. However in some special cases
1065 * reclaiming must be done manually. An example is packets that pass through
1066 * the loopback interface. Also, one must be careful to do this when
1067 * ``turning around'' packets (e.g., icmp_reflect).
1069 * To mark a tag persistent bit-or this flag in when defining the tag id.
1070 * The tag will then be treated as described above.
1072 #define MTAG_PERSISTENT 0x800
1074 #define PACKET_TAG_NONE 0 /* Nadda */
1076 /* Packet tags for use with PACKET_ABI_COMPAT. */
1077 #define PACKET_TAG_IPSEC_IN_DONE 1 /* IPsec applied, in */
1078 #define PACKET_TAG_IPSEC_OUT_DONE 2 /* IPsec applied, out */
1079 #define PACKET_TAG_IPSEC_IN_CRYPTO_DONE 3 /* NIC IPsec crypto done */
1080 #define PACKET_TAG_IPSEC_OUT_CRYPTO_NEEDED 4 /* NIC IPsec crypto req'ed */
1081 #define PACKET_TAG_IPSEC_IN_COULD_DO_CRYPTO 5 /* NIC notifies IPsec */
1082 #define PACKET_TAG_IPSEC_PENDING_TDB 6 /* Reminder to do IPsec */
1083 #define PACKET_TAG_BRIDGE 7 /* Bridge processing done */
1084 #define PACKET_TAG_GIF 8 /* GIF processing done */
1085 #define PACKET_TAG_GRE 9 /* GRE processing done */
1086 #define PACKET_TAG_IN_PACKET_CHECKSUM 10 /* NIC checksumming done */
1087 #define PACKET_TAG_ENCAP 11 /* Encap. processing */
1088 #define PACKET_TAG_IPSEC_SOCKET 12 /* IPSEC socket ref */
1089 #define PACKET_TAG_IPSEC_HISTORY 13 /* IPSEC history */
1090 #define PACKET_TAG_IPV6_INPUT 14 /* IPV6 input processing */
1091 #define PACKET_TAG_DUMMYNET 15 /* dummynet info */
1092 #define PACKET_TAG_DIVERT 17 /* divert info */
1093 #define PACKET_TAG_IPFORWARD 18 /* ipforward info */
1094 #define PACKET_TAG_MACLABEL (19 | MTAG_PERSISTENT) /* MAC label */
1095 #define PACKET_TAG_PF (21 | MTAG_PERSISTENT) /* PF/ALTQ information */
1096 #define PACKET_TAG_RTSOCKFAM 25 /* rtsock sa family */
1097 #define PACKET_TAG_IPOPTIONS 27 /* Saved IP options */
1098 #define PACKET_TAG_CARP 28 /* CARP info */
1099 #define PACKET_TAG_IPSEC_NAT_T_PORTS 29 /* two uint16_t */
1100 #define PACKET_TAG_ND_OUTGOING 30 /* ND outgoing */
1102 /* Specific cookies and tags. */
1104 /* Packet tag routines. */
1105 struct m_tag *m_tag_alloc(u_int32_t, int, int, int);
1106 void m_tag_delete(struct mbuf *, struct m_tag *);
1107 void m_tag_delete_chain(struct mbuf *, struct m_tag *);
1108 void m_tag_free_default(struct m_tag *);
1109 struct m_tag *m_tag_locate(struct mbuf *, u_int32_t, int, struct m_tag *);
1110 struct m_tag *m_tag_copy(struct m_tag *, int);
1111 int m_tag_copy_chain(struct mbuf *, const struct mbuf *, int);
1112 void m_tag_delete_nonpersistent(struct mbuf *);
1115 * Initialize the list of tags associated with an mbuf.
1117 static __inline void
1118 m_tag_init(struct mbuf *m)
1121 SLIST_INIT(&m->m_pkthdr.tags);
1125 * Set up the contents of a tag. Note that this does not fill in the free
1126 * method; the caller is expected to do that.
1128 * XXX probably should be called m_tag_init, but that was already taken.
1130 static __inline void
1131 m_tag_setup(struct m_tag *t, u_int32_t cookie, int type, int len)
1136 t->m_tag_cookie = cookie;
1140 * Reclaim resources associated with a tag.
1142 static __inline void
1143 m_tag_free(struct m_tag *t)
1146 (*t->m_tag_free)(t);
1150 * Return the first tag associated with an mbuf.
1152 static __inline struct m_tag *
1153 m_tag_first(struct mbuf *m)
1156 return (SLIST_FIRST(&m->m_pkthdr.tags));
1160 * Return the next tag in the list of tags associated with an mbuf.
1162 static __inline struct m_tag *
1163 m_tag_next(struct mbuf *m __unused, struct m_tag *t)
1166 return (SLIST_NEXT(t, m_tag_link));
1170 * Prepend a tag to the list of tags associated with an mbuf.
1172 static __inline void
1173 m_tag_prepend(struct mbuf *m, struct m_tag *t)
1176 SLIST_INSERT_HEAD(&m->m_pkthdr.tags, t, m_tag_link);
1180 * Unlink a tag from the list of tags associated with an mbuf.
1182 static __inline void
1183 m_tag_unlink(struct mbuf *m, struct m_tag *t)
1186 SLIST_REMOVE(&m->m_pkthdr.tags, t, m_tag, m_tag_link);
1189 /* These are for OpenBSD compatibility. */
1190 #define MTAG_ABI_COMPAT 0 /* compatibility ABI */
1192 static __inline struct m_tag *
1193 m_tag_get(int type, int length, int wait)
1195 return (m_tag_alloc(MTAG_ABI_COMPAT, type, length, wait));
1198 static __inline struct m_tag *
1199 m_tag_find(struct mbuf *m, int type, struct m_tag *start)
1201 return (SLIST_EMPTY(&m->m_pkthdr.tags) ? (struct m_tag *)NULL :
1202 m_tag_locate(m, MTAG_ABI_COMPAT, type, start));
1205 static inline struct m_snd_tag *
1206 m_snd_tag_ref(struct m_snd_tag *mst)
1209 refcount_acquire(&mst->refcount);
1214 m_snd_tag_rele(struct m_snd_tag *mst)
1217 if (refcount_release(&mst->refcount))
1218 m_snd_tag_destroy(mst);
1221 static __inline struct mbuf *
1222 m_free(struct mbuf *m)
1224 struct mbuf *n = m->m_next;
1226 MBUF_PROBE1(m__free, m);
1227 if ((m->m_flags & (M_PKTHDR|M_NOFREE)) == (M_PKTHDR|M_NOFREE))
1228 m_tag_delete_chain(m, NULL);
1229 if (m->m_flags & M_PKTHDR && m->m_pkthdr.csum_flags & CSUM_SND_TAG)
1230 m_snd_tag_rele(m->m_pkthdr.snd_tag);
1231 if (m->m_flags & M_EXT)
1233 else if ((m->m_flags & M_NOFREE) == 0)
1234 uma_zfree(zone_mbuf, m);
1239 rt_m_getfib(struct mbuf *m)
1241 KASSERT(m->m_flags & M_PKTHDR , ("Attempt to get FIB from non header mbuf."));
1242 return (m->m_pkthdr.fibnum);
1245 #define M_GETFIB(_m) rt_m_getfib(_m)
1247 #define M_SETFIB(_m, _fib) do { \
1248 KASSERT((_m)->m_flags & M_PKTHDR, ("Attempt to set FIB on non header mbuf.")); \
1249 ((_m)->m_pkthdr.fibnum) = (_fib); \
1252 /* flags passed as first argument for "m_ether_tcpip_hash()" */
1253 #define MBUF_HASHFLAG_L2 (1 << 2)
1254 #define MBUF_HASHFLAG_L3 (1 << 3)
1255 #define MBUF_HASHFLAG_L4 (1 << 4)
1257 /* mbuf hashing helper routines */
1258 uint32_t m_ether_tcpip_hash_init(void);
1259 uint32_t m_ether_tcpip_hash(const uint32_t, const struct mbuf *, const uint32_t);
1261 #ifdef MBUF_PROFILING
1262 void m_profile(struct mbuf *m);
1263 #define M_PROFILE(m) m_profile(m)
1265 #define M_PROFILE(m)
1269 STAILQ_HEAD(, mbuf) mq_head;
1275 mbufq_init(struct mbufq *mq, int maxlen)
1278 STAILQ_INIT(&mq->mq_head);
1279 mq->mq_maxlen = maxlen;
1283 static inline struct mbuf *
1284 mbufq_flush(struct mbufq *mq)
1288 m = STAILQ_FIRST(&mq->mq_head);
1289 STAILQ_INIT(&mq->mq_head);
1295 mbufq_drain(struct mbufq *mq)
1299 n = mbufq_flush(mq);
1300 while ((m = n) != NULL) {
1301 n = STAILQ_NEXT(m, m_stailqpkt);
1306 static inline struct mbuf *
1307 mbufq_first(const struct mbufq *mq)
1310 return (STAILQ_FIRST(&mq->mq_head));
1313 static inline struct mbuf *
1314 mbufq_last(const struct mbufq *mq)
1317 return (STAILQ_LAST(&mq->mq_head, mbuf, m_stailqpkt));
1321 mbufq_full(const struct mbufq *mq)
1324 return (mq->mq_len >= mq->mq_maxlen);
1328 mbufq_len(const struct mbufq *mq)
1331 return (mq->mq_len);
1335 mbufq_enqueue(struct mbufq *mq, struct mbuf *m)
1340 STAILQ_INSERT_TAIL(&mq->mq_head, m, m_stailqpkt);
1345 static inline struct mbuf *
1346 mbufq_dequeue(struct mbufq *mq)
1350 m = STAILQ_FIRST(&mq->mq_head);
1352 STAILQ_REMOVE_HEAD(&mq->mq_head, m_stailqpkt);
1353 m->m_nextpkt = NULL;
1360 mbufq_prepend(struct mbufq *mq, struct mbuf *m)
1363 STAILQ_INSERT_HEAD(&mq->mq_head, m, m_stailqpkt);
1368 * Note: this doesn't enforce the maximum list size for dst.
1371 mbufq_concat(struct mbufq *mq_dst, struct mbufq *mq_src)
1374 mq_dst->mq_len += mq_src->mq_len;
1375 STAILQ_CONCAT(&mq_dst->mq_head, &mq_src->mq_head);
1379 #ifdef _SYS_TIMESPEC_H_
1381 mbuf_tstmp2timespec(struct mbuf *m, struct timespec *ts)
1384 KASSERT((m->m_flags & M_PKTHDR) != 0, ("mbuf %p no M_PKTHDR", m));
1385 KASSERT((m->m_flags & M_TSTMP) != 0, ("mbuf %p no M_TSTMP", m));
1386 ts->tv_sec = m->m_pkthdr.rcv_tstmp / 1000000000;
1387 ts->tv_nsec = m->m_pkthdr.rcv_tstmp % 1000000000;
1392 /* Invoked from the netdump client code. */
1393 void netdump_mbuf_drain(void);
1394 void netdump_mbuf_dump(void);
1395 void netdump_mbuf_reinit(int nmbuf, int nclust, int clsize);
1398 #endif /* _KERNEL */
1399 #endif /* !_SYS_MBUF_H_ */