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_raw);
66 SDT_PROBE_DECLARE(sdt, , , m__gethdr);
67 SDT_PROBE_DECLARE(sdt, , , m__get_raw);
68 SDT_PROBE_DECLARE(sdt, , , m__get);
69 SDT_PROBE_DECLARE(sdt, , , m__getcl);
70 SDT_PROBE_DECLARE(sdt, , , m__getjcl);
71 SDT_PROBE_DECLARE(sdt, , , m__clget);
72 SDT_PROBE_DECLARE(sdt, , , m__cljget);
73 SDT_PROBE_DECLARE(sdt, , , m__cljset);
74 SDT_PROBE_DECLARE(sdt, , , m__free);
75 SDT_PROBE_DECLARE(sdt, , , m__freem);
80 * Mbufs are of a single size, MSIZE (sys/param.h), which includes overhead.
81 * An mbuf may add a single "mbuf cluster" of size MCLBYTES (also in
82 * sys/param.h), which has no additional overhead and is used instead of the
83 * internal data area; this is done when at least MINCLSIZE of data must be
84 * stored. Additionally, it is possible to allocate a separate buffer
85 * externally and attach it to the mbuf in a way similar to that of mbuf
88 * NB: These calculation do not take actual compiler-induced alignment and
89 * padding inside the complete struct mbuf into account. Appropriate
90 * attention is required when changing members of struct mbuf.
92 * MLEN is data length in a normal mbuf.
93 * MHLEN is data length in an mbuf with pktheader.
94 * MINCLSIZE is a smallest amount of data that should be put into cluster.
96 * Compile-time assertions in uipc_mbuf.c test these values to ensure that
100 #define MHSIZE offsetof(struct mbuf, m_dat)
101 #define MPKTHSIZE offsetof(struct mbuf, m_pktdat)
102 #define MLEN ((int)(MSIZE - MHSIZE))
103 #define MHLEN ((int)(MSIZE - MPKTHSIZE))
104 #define MINCLSIZE (MHLEN + 1)
109 * Macro for type conversion: convert mbuf pointer to data pointer of correct
112 * mtod(m, t) -- Convert mbuf pointer to data pointer of correct type.
113 * mtodo(m, o) -- Same as above but with offset 'o' into data.
115 #define mtod(m, t) ((t)((m)->m_data))
116 #define mtodo(m, o) ((void *)(((m)->m_data) + (o)))
119 * Argument structure passed to UMA routines during mbuf and packet
123 int flags; /* Flags for mbuf being allocated */
124 short type; /* Type of mbuf being allocated */
129 * Packet tag structure (see below for details).
132 SLIST_ENTRY(m_tag) m_tag_link; /* List of packet tags */
133 u_int16_t m_tag_id; /* Tag ID */
134 u_int16_t m_tag_len; /* Length of data */
135 u_int32_t m_tag_cookie; /* ABI/Module ID */
136 void (*m_tag_free)(struct m_tag *);
140 * Static network interface owned tag.
141 * Allocated through ifp->if_snd_tag_alloc().
144 struct ifnet *ifp; /* network interface tag belongs to */
145 volatile u_int refcount;
146 u_int type; /* One of IF_SND_TAG_TYPE_*. */
150 * Record/packet header in first mbuf of chain; valid only if M_PKTHDR is set.
153 * Compile-time assertions in uipc_mbuf.c test these values to ensure that
158 struct m_snd_tag *snd_tag; /* send tag, if any */
159 struct ifnet *rcvif; /* rcv interface */
161 SLIST_HEAD(packet_tags, m_tag) tags; /* list of packet tags */
162 int32_t len; /* total packet length */
164 /* Layer crossing persistent information. */
165 uint32_t flowid; /* packet's 4-tuple system */
166 uint32_t csum_flags; /* checksum and offload features */
167 uint16_t fibnum; /* this packet should use this fib */
168 uint8_t numa_domain; /* NUMA domain of recvd pkt */
169 uint8_t rsstype; /* hash type */
171 uint64_t rcv_tstmp; /* timestamp in ns */
173 uint8_t l2hlen; /* layer 2 hdr len */
174 uint8_t l3hlen; /* layer 3 hdr len */
175 uint8_t l4hlen; /* layer 4 hdr len */
176 uint8_t l5hlen; /* layer 5 hdr len */
177 uint8_t inner_l2hlen;
178 uint8_t inner_l3hlen;
179 uint8_t inner_l4hlen;
180 uint8_t inner_l5hlen;
186 uint32_t thirtytwo[2];
187 uint64_t sixtyfour[1];
188 uintptr_t unintptr[1];
192 /* Layer specific non-persistent local storage for reassembly, etc. */
196 uint32_t thirtytwo[2];
197 uint64_t sixtyfour[1];
198 uintptr_t unintptr[1];
202 #define ether_vtag PH_per.sixteen[0]
203 #define tcp_tun_port PH_per.sixteen[0] /* outbound */
205 #define vt_nrecs sixteen[0] /* mld and v6-ND */
206 #define tso_segsz PH_per.sixteen[1] /* inbound after LRO */
207 #define lro_nsegs tso_segsz /* inbound after LRO */
208 #define csum_data PH_per.thirtytwo[1] /* inbound from hardware up */
209 #define lro_tcp_d_len PH_loc.sixteen[0] /* inbound during LRO (no reassembly) */
210 #define lro_tcp_d_csum PH_loc.sixteen[1] /* inbound during LRO (no reassembly) */
211 #define lro_tcp_h_off PH_loc.sixteen[2] /* inbound during LRO (no reassembly) */
212 #define lro_etype PH_loc.sixteen[3] /* inbound during LRO (no reassembly) */
213 /* Note PH_loc is used during IP reassembly (all 8 bytes as a ptr) */
216 * TLS records for TLS 1.0-1.2 can have the following header lengths:
217 * - 5 (AES-CBC with implicit IV)
218 * - 21 (AES-CBC with explicit IV)
219 * - 13 (AES-GCM with 8 byte explicit IV)
221 #define MBUF_PEXT_HDR_LEN 23
224 * TLS records for TLS 1.0-1.2 can have the following maximum trailer
227 * - 36 (AES-CBC with SHA1 and up to 16 bytes of padding)
228 * - 48 (AES-CBC with SHA2-256 and up to 16 bytes of padding)
229 * - 64 (AES-CBC with SHA2-384 and up to 16 bytes of padding)
231 #define MBUF_PEXT_TRAIL_LEN 64
233 #if defined(__LP64__)
234 #define MBUF_PEXT_MAX_PGS (40 / sizeof(vm_paddr_t))
236 #define MBUF_PEXT_MAX_PGS (72 / sizeof(vm_paddr_t))
239 #define MBUF_PEXT_MAX_BYTES \
240 (MBUF_PEXT_MAX_PGS * PAGE_SIZE + MBUF_PEXT_HDR_LEN + MBUF_PEXT_TRAIL_LEN)
246 * Description of external storage mapped into mbuf; valid only if M_EXT is
250 * Compile-time assertions in uipc_mbuf.c test these values to ensure that
253 typedef void m_ext_free_t(struct mbuf *);
257 * If EXT_FLAG_EMBREF is set, then we use refcount in the
258 * mbuf, the 'ext_count' member. Otherwise, we have a
259 * shadow copy and we use pointer 'ext_cnt'. The original
260 * mbuf is responsible to carry the pointer to free routine
261 * and its arguments. They aren't copied into shadows in
262 * mb_dupcl() to avoid dereferencing next cachelines.
264 volatile u_int ext_count;
265 volatile u_int *ext_cnt;
267 uint32_t ext_size; /* size of buffer, for ext_free */
268 uint32_t ext_type:8, /* type of external storage */
269 ext_flags:24; /* external storage mbuf flags */
273 * Regular M_EXT mbuf:
274 * o ext_buf always points to the external buffer.
275 * o ext_free (below) and two optional arguments
276 * ext_arg1 and ext_arg2 store the free context for
277 * the external storage. They are set only in the
278 * refcount carrying mbuf, the one with
279 * EXT_FLAG_EMBREF flag, with exclusion for
280 * EXT_EXTREF type, where the free context is copied
281 * into all mbufs that use same external storage.
283 char *ext_buf; /* start of buffer */
284 #define m_ext_copylen offsetof(struct m_ext, ext_arg2)
289 * Multi-page M_EXTPG mbuf:
290 * o extpg_pa - page vector.
291 * o extpg_trail and extpg_hdr - TLS trailer and
293 * Uses ext_free and may also use ext_arg1.
295 vm_paddr_t extpg_pa[MBUF_PEXT_MAX_PGS];
296 char extpg_trail[MBUF_PEXT_TRAIL_LEN];
297 char extpg_hdr[MBUF_PEXT_HDR_LEN];
298 /* Pretend these 3 fields are part of mbuf itself. */
299 #define m_epg_pa m_ext.extpg_pa
300 #define m_epg_trail m_ext.extpg_trail
301 #define m_epg_hdr m_ext.extpg_hdr
302 #define m_epg_ext_copylen offsetof(struct m_ext, ext_free)
306 * Free method and optional argument pointer, both
307 * used by M_EXT and M_EXTPG.
309 m_ext_free_t *ext_free;
314 * The core of the mbuf object along with some shortcut defines for practical
319 * Header present at the beginning of every mbuf.
322 * Compile-time assertions in uipc_mbuf.c test these values to ensure
323 * that they are correct.
325 union { /* next buffer in chain */
327 SLIST_ENTRY(mbuf) m_slist;
328 STAILQ_ENTRY(mbuf) m_stailq;
330 union { /* next chain in queue/record */
331 struct mbuf *m_nextpkt;
332 SLIST_ENTRY(mbuf) m_slistpkt;
333 STAILQ_ENTRY(mbuf) m_stailqpkt;
335 caddr_t m_data; /* location of data */
336 int32_t m_len; /* amount of data in this mbuf */
337 uint32_t m_type:8, /* type of data in this mbuf */
338 m_flags:24; /* flags; see below */
339 #if !defined(__LP64__)
340 uint32_t m_pad; /* pad for 64bit alignment */
344 * A set of optional headers (packet header, external storage header)
345 * and internal data storage. Historically, these arrays were sized
346 * to MHLEN (space left after a packet header) and MLEN (space left
347 * after only a regular mbuf header); they are now variable size in
348 * order to support future work on variable-size mbufs.
354 struct pkthdr m_pkthdr;
357 * Multi-page M_EXTPG mbuf has its meta data
358 * split between the below anonymous structure
359 * and m_ext. It carries vector of pages,
360 * optional header and trailer char vectors
361 * and pointers to socket/TLS data.
363 #define m_epg_startcopy m_epg_npgs
364 #define m_epg_endcopy m_epg_stailq
366 /* Overall count of pages and count of
367 * pages with I/O pending. */
370 /* TLS header and trailer lengths.
371 * The data itself resides in m_ext. */
372 uint8_t m_epg_hdrlen;
373 uint8_t m_epg_trllen;
374 /* Offset into 1st page and length of
375 * data in the last page. */
376 uint16_t m_epg_1st_off;
377 uint16_t m_epg_last_len;
379 #define EPG_FLAG_ANON 0x1 /* Data can be encrypted in place. */
380 #define EPG_FLAG_2FREE 0x2 /* Scheduled for free. */
381 uint8_t m_epg_record_type;
384 struct ktls_session *m_epg_tls;
385 struct socket *m_epg_so;
386 uint64_t m_epg_seqno;
387 STAILQ_ENTRY(mbuf) m_epg_stailq;
391 /* M_EXT or M_EXTPG set. */
393 /* M_PKTHDR set, neither M_EXT nor M_EXTPG. */
397 char m_dat[0]; /* !M_PKTHDR, !M_EXT */
403 m_epg_pagelen(const struct mbuf *m, int pidx, int pgoff)
406 KASSERT(pgoff == 0 || pidx == 0,
407 ("page %d with non-zero offset %d in %p", pidx, pgoff, m));
409 if (pidx == m->m_epg_npgs - 1) {
410 return (m->m_epg_last_len);
412 return (PAGE_SIZE - pgoff);
417 #define MCHECK(ex, msg) KASSERT((ex), \
418 ("Multi page mbuf %p with " #msg " at %s:%d", \
419 m, __FILE__, __LINE__))
421 * NB: This expects a non-empty buffer (npgs > 0 and
424 #define MBUF_EXT_PGS_ASSERT_SANITY(m) do { \
425 MCHECK(m->m_epg_npgs > 0, "no valid pages"); \
426 MCHECK(m->m_epg_npgs <= nitems(m->m_epg_pa), \
428 MCHECK(m->m_epg_nrdy <= m->m_epg_npgs, \
429 "too many ready pages"); \
430 MCHECK(m->m_epg_1st_off < PAGE_SIZE, \
431 "too large page offset"); \
432 MCHECK(m->m_epg_last_len > 0, "zero last page length"); \
433 MCHECK(m->m_epg_last_len <= PAGE_SIZE, \
434 "too large last page length"); \
435 if (m->m_epg_npgs == 1) \
436 MCHECK(m->m_epg_1st_off + \
437 m->m_epg_last_len <= PAGE_SIZE, \
438 "single page too large"); \
439 MCHECK(m->m_epg_hdrlen <= sizeof(m->m_epg_hdr), \
440 "too large header length"); \
441 MCHECK(m->m_epg_trllen <= sizeof(m->m_epg_trail), \
442 "too large header length"); \
445 #define MBUF_EXT_PGS_ASSERT_SANITY(m) do {} while (0);
450 * mbuf flags of global significance and layer crossing.
451 * Those of only protocol/layer specific significance are to be mapped
452 * to M_PROTO[1-11] and cleared at layer handoff boundaries.
453 * NB: Limited to the lower 24 bits.
455 #define M_EXT 0x00000001 /* has associated external storage */
456 #define M_PKTHDR 0x00000002 /* start of record */
457 #define M_EOR 0x00000004 /* end of record */
458 #define M_RDONLY 0x00000008 /* associated data is marked read-only */
459 #define M_BCAST 0x00000010 /* send/received as link-level broadcast */
460 #define M_MCAST 0x00000020 /* send/received as link-level multicast */
461 #define M_PROMISC 0x00000040 /* packet was not for us */
462 #define M_VLANTAG 0x00000080 /* ether_vtag is valid */
463 #define M_EXTPG 0x00000100 /* has array of unmapped pages and TLS */
464 #define M_NOFREE 0x00000200 /* do not free mbuf, embedded in cluster */
465 #define M_TSTMP 0x00000400 /* rcv_tstmp field is valid */
466 #define M_TSTMP_HPREC 0x00000800 /* rcv_tstmp is high-prec, typically
467 hw-stamped on port (useful for IEEE 1588
469 #define M_TSTMP_LRO 0x00001000 /* Time LRO pushed in pkt is valid in (PH_loc) */
471 #define M_PROTO1 0x00002000 /* protocol-specific */
472 #define M_PROTO2 0x00004000 /* protocol-specific */
473 #define M_PROTO3 0x00008000 /* protocol-specific */
474 #define M_PROTO4 0x00010000 /* protocol-specific */
475 #define M_PROTO5 0x00020000 /* protocol-specific */
476 #define M_PROTO6 0x00040000 /* protocol-specific */
477 #define M_PROTO7 0x00080000 /* protocol-specific */
478 #define M_PROTO8 0x00100000 /* protocol-specific */
479 #define M_PROTO9 0x00200000 /* protocol-specific */
480 #define M_PROTO10 0x00400000 /* protocol-specific */
481 #define M_PROTO11 0x00800000 /* protocol-specific */
484 * Flags to purge when crossing layers.
486 #define M_PROTOFLAGS \
487 (M_PROTO1|M_PROTO2|M_PROTO3|M_PROTO4|M_PROTO5|M_PROTO6|M_PROTO7|M_PROTO8|\
488 M_PROTO9|M_PROTO10|M_PROTO11)
491 * Flags preserved when copying m_pkthdr.
493 #define M_COPYFLAGS \
494 (M_PKTHDR|M_EOR|M_RDONLY|M_BCAST|M_MCAST|M_PROMISC|M_VLANTAG|M_TSTMP| \
495 M_TSTMP_HPREC|M_TSTMP_LRO|M_PROTOFLAGS)
498 * Mbuf flag description for use with printf(9) %b identifier.
500 #define M_FLAG_BITS \
501 "\20\1M_EXT\2M_PKTHDR\3M_EOR\4M_RDONLY\5M_BCAST\6M_MCAST" \
502 "\7M_PROMISC\10M_VLANTAG\11M_EXTPG\12M_NOFREE\13M_TSTMP\14M_TSTMP_HPREC\15M_TSTMP_LRO"
503 #define M_FLAG_PROTOBITS \
504 "\16M_PROTO1\17M_PROTO2\20M_PROTO3\21M_PROTO4" \
505 "\22M_PROTO5\23M_PROTO6\24M_PROTO7\25M_PROTO8\26M_PROTO9" \
506 "\27M_PROTO10\28M_PROTO11"
507 #define M_FLAG_PRINTF (M_FLAG_BITS M_FLAG_PROTOBITS)
510 * Network interface cards are able to hash protocol fields (such as IPv4
511 * addresses and TCP port numbers) classify packets into flows. These flows
512 * can then be used to maintain ordering while delivering packets to the OS
513 * via parallel input queues, as well as to provide a stateless affinity
514 * model. NIC drivers can pass up the hash via m->m_pkthdr.flowid, and set
515 * m_flag fields to indicate how the hash should be interpreted by the
518 * Most NICs support RSS, which provides ordering and explicit affinity, and
519 * use the hash m_flag bits to indicate what header fields were covered by
520 * the hash. M_HASHTYPE_OPAQUE and M_HASHTYPE_OPAQUE_HASH can be set by non-
521 * RSS cards or configurations that provide an opaque flow identifier, allowing
522 * for ordering and distribution without explicit affinity. Additionally,
523 * M_HASHTYPE_OPAQUE_HASH indicates that the flow identifier has hash
526 * The meaning of the IPV6_EX suffix:
527 * "o Home address from the home address option in the IPv6 destination
528 * options header. If the extension header is not present, use the Source
530 * o IPv6 address that is contained in the Routing-Header-Type-2 from the
531 * associated extension header. If the extension header is not present,
532 * use the Destination IPv6 Address."
534 * https://docs.microsoft.com/en-us/windows-hardware/drivers/network/rss-hashing-types#ndishashipv6ex
536 #define M_HASHTYPE_HASHPROP 0x80 /* has hash properties */
537 #define M_HASHTYPE_INNER 0x40 /* calculated from inner headers */
538 #define M_HASHTYPE_HASH(t) (M_HASHTYPE_HASHPROP | (t))
539 /* Microsoft RSS standard hash types */
540 #define M_HASHTYPE_NONE 0
541 #define M_HASHTYPE_RSS_IPV4 M_HASHTYPE_HASH(1) /* IPv4 2-tuple */
542 #define M_HASHTYPE_RSS_TCP_IPV4 M_HASHTYPE_HASH(2) /* TCPv4 4-tuple */
543 #define M_HASHTYPE_RSS_IPV6 M_HASHTYPE_HASH(3) /* IPv6 2-tuple */
544 #define M_HASHTYPE_RSS_TCP_IPV6 M_HASHTYPE_HASH(4) /* TCPv6 4-tuple */
545 #define M_HASHTYPE_RSS_IPV6_EX M_HASHTYPE_HASH(5) /* IPv6 2-tuple +
547 #define M_HASHTYPE_RSS_TCP_IPV6_EX M_HASHTYPE_HASH(6) /* TCPv6 4-tuple +
549 #define M_HASHTYPE_RSS_UDP_IPV4 M_HASHTYPE_HASH(7) /* IPv4 UDP 4-tuple*/
550 #define M_HASHTYPE_RSS_UDP_IPV6 M_HASHTYPE_HASH(9) /* IPv6 UDP 4-tuple*/
551 #define M_HASHTYPE_RSS_UDP_IPV6_EX M_HASHTYPE_HASH(10)/* IPv6 UDP 4-tuple +
554 #define M_HASHTYPE_OPAQUE 0x3f /* ordering, not affinity */
555 #define M_HASHTYPE_OPAQUE_HASH M_HASHTYPE_HASH(M_HASHTYPE_OPAQUE)
556 /* ordering+hash, not affinity*/
558 #define M_HASHTYPE_CLEAR(m) ((m)->m_pkthdr.rsstype = 0)
559 #define M_HASHTYPE_GET(m) ((m)->m_pkthdr.rsstype & ~M_HASHTYPE_INNER)
560 #define M_HASHTYPE_SET(m, v) ((m)->m_pkthdr.rsstype = (v))
561 #define M_HASHTYPE_TEST(m, v) (M_HASHTYPE_GET(m) == (v))
562 #define M_HASHTYPE_ISHASH(m) \
563 (((m)->m_pkthdr.rsstype & M_HASHTYPE_HASHPROP) != 0)
564 #define M_HASHTYPE_SETINNER(m) do { \
565 (m)->m_pkthdr.rsstype |= M_HASHTYPE_INNER; \
569 * External mbuf storage buffer types.
571 #define EXT_CLUSTER 1 /* mbuf cluster */
572 #define EXT_SFBUF 2 /* sendfile(2)'s sf_buf */
573 #define EXT_JUMBOP 3 /* jumbo cluster page sized */
574 #define EXT_JUMBO9 4 /* jumbo cluster 9216 bytes */
575 #define EXT_JUMBO16 5 /* jumbo cluster 16184 bytes */
576 #define EXT_PACKET 6 /* mbuf+cluster from packet zone */
577 #define EXT_MBUF 7 /* external mbuf reference */
578 #define EXT_RXRING 8 /* data in NIC receive ring */
580 #define EXT_VENDOR1 224 /* for vendor-internal use */
581 #define EXT_VENDOR2 225 /* for vendor-internal use */
582 #define EXT_VENDOR3 226 /* for vendor-internal use */
583 #define EXT_VENDOR4 227 /* for vendor-internal use */
585 #define EXT_EXP1 244 /* for experimental use */
586 #define EXT_EXP2 245 /* for experimental use */
587 #define EXT_EXP3 246 /* for experimental use */
588 #define EXT_EXP4 247 /* for experimental use */
590 #define EXT_NET_DRV 252 /* custom ext_buf provided by net driver(s) */
591 #define EXT_MOD_TYPE 253 /* custom module's ext_buf type */
592 #define EXT_DISPOSABLE 254 /* can throw this buffer away w/page flipping */
593 #define EXT_EXTREF 255 /* has externally maintained ext_cnt ptr */
596 * Flags for external mbuf buffer types.
597 * NB: limited to the lower 24 bits.
599 #define EXT_FLAG_EMBREF 0x000001 /* embedded ext_count */
600 #define EXT_FLAG_EXTREF 0x000002 /* external ext_cnt, notyet */
602 #define EXT_FLAG_NOFREE 0x000010 /* don't free mbuf to pool, notyet */
604 #define EXT_FLAG_VENDOR1 0x010000 /* These flags are vendor */
605 #define EXT_FLAG_VENDOR2 0x020000 /* or submodule specific, */
606 #define EXT_FLAG_VENDOR3 0x040000 /* not used by mbuf code. */
607 #define EXT_FLAG_VENDOR4 0x080000 /* Set/read by submodule. */
609 #define EXT_FLAG_EXP1 0x100000 /* for experimental use */
610 #define EXT_FLAG_EXP2 0x200000 /* for experimental use */
611 #define EXT_FLAG_EXP3 0x400000 /* for experimental use */
612 #define EXT_FLAG_EXP4 0x800000 /* for experimental use */
615 * EXT flag description for use with printf(9) %b identifier.
617 #define EXT_FLAG_BITS \
618 "\20\1EXT_FLAG_EMBREF\2EXT_FLAG_EXTREF\5EXT_FLAG_NOFREE" \
619 "\21EXT_FLAG_VENDOR1\22EXT_FLAG_VENDOR2\23EXT_FLAG_VENDOR3" \
620 "\24EXT_FLAG_VENDOR4\25EXT_FLAG_EXP1\26EXT_FLAG_EXP2\27EXT_FLAG_EXP3" \
624 * Flags indicating checksum, segmentation and other offload work to be
625 * done, or already done, by hardware or lower layers. It is split into
626 * separate inbound and outbound flags.
628 * Outbound flags that are set by upper protocol layers requesting lower
629 * layers, or ideally the hardware, to perform these offloading tasks.
630 * For outbound packets this field and its flags can be directly tested
631 * against ifnet if_hwassist. Note that the outbound and the inbound flags do
632 * not collide right now but they could be allowed to (as long as the flags are
633 * scrubbed appropriately when the direction of an mbuf changes). CSUM_BITS
634 * would also have to split into CSUM_BITS_TX and CSUM_BITS_RX.
636 * CSUM_INNER_<x> is the same as CSUM_<x> but it applies to the inner frame.
637 * The CSUM_ENCAP_<x> bits identify the outer encapsulation.
639 #define CSUM_IP 0x00000001 /* IP header checksum offload */
640 #define CSUM_IP_UDP 0x00000002 /* UDP checksum offload */
641 #define CSUM_IP_TCP 0x00000004 /* TCP checksum offload */
642 #define CSUM_IP_SCTP 0x00000008 /* SCTP checksum offload */
643 #define CSUM_IP_TSO 0x00000010 /* TCP segmentation offload */
644 #define CSUM_IP_ISCSI 0x00000020 /* iSCSI checksum offload */
646 #define CSUM_INNER_IP6_UDP 0x00000040
647 #define CSUM_INNER_IP6_TCP 0x00000080
648 #define CSUM_INNER_IP6_TSO 0x00000100
649 #define CSUM_IP6_UDP 0x00000200 /* UDP checksum offload */
650 #define CSUM_IP6_TCP 0x00000400 /* TCP checksum offload */
651 #define CSUM_IP6_SCTP 0x00000800 /* SCTP checksum offload */
652 #define CSUM_IP6_TSO 0x00001000 /* TCP segmentation offload */
653 #define CSUM_IP6_ISCSI 0x00002000 /* iSCSI checksum offload */
655 #define CSUM_INNER_IP 0x00004000
656 #define CSUM_INNER_IP_UDP 0x00008000
657 #define CSUM_INNER_IP_TCP 0x00010000
658 #define CSUM_INNER_IP_TSO 0x00020000
660 #define CSUM_ENCAP_VXLAN 0x00040000 /* VXLAN outer encapsulation */
661 #define CSUM_ENCAP_RSVD1 0x00080000
663 /* Inbound checksum support where the checksum was verified by hardware. */
664 #define CSUM_INNER_L3_CALC 0x00100000
665 #define CSUM_INNER_L3_VALID 0x00200000
666 #define CSUM_INNER_L4_CALC 0x00400000
667 #define CSUM_INNER_L4_VALID 0x00800000
668 #define CSUM_L3_CALC 0x01000000 /* calculated layer 3 csum */
669 #define CSUM_L3_VALID 0x02000000 /* checksum is correct */
670 #define CSUM_L4_CALC 0x04000000 /* calculated layer 4 csum */
671 #define CSUM_L4_VALID 0x08000000 /* checksum is correct */
672 #define CSUM_L5_CALC 0x10000000 /* calculated layer 5 csum */
673 #define CSUM_L5_VALID 0x20000000 /* checksum is correct */
674 #define CSUM_COALESCED 0x40000000 /* contains merged segments */
676 #define CSUM_SND_TAG 0x80000000 /* Packet header has send tag */
678 #define CSUM_FLAGS_TX (CSUM_IP | CSUM_IP_UDP | CSUM_IP_TCP | CSUM_IP_SCTP | \
679 CSUM_IP_TSO | CSUM_IP_ISCSI | CSUM_INNER_IP6_UDP | CSUM_INNER_IP6_TCP | \
680 CSUM_INNER_IP6_TSO | CSUM_IP6_UDP | CSUM_IP6_TCP | CSUM_IP6_SCTP | \
681 CSUM_IP6_TSO | CSUM_IP6_ISCSI | CSUM_INNER_IP | CSUM_INNER_IP_UDP | \
682 CSUM_INNER_IP_TCP | CSUM_INNER_IP_TSO | CSUM_ENCAP_VXLAN | \
683 CSUM_ENCAP_RSVD1 | CSUM_SND_TAG)
685 #define CSUM_FLAGS_RX (CSUM_INNER_L3_CALC | CSUM_INNER_L3_VALID | \
686 CSUM_INNER_L4_CALC | CSUM_INNER_L4_VALID | CSUM_L3_CALC | CSUM_L3_VALID | \
687 CSUM_L4_CALC | CSUM_L4_VALID | CSUM_L5_CALC | CSUM_L5_VALID | \
691 * CSUM flag description for use with printf(9) %b identifier.
694 "\20\1CSUM_IP\2CSUM_IP_UDP\3CSUM_IP_TCP\4CSUM_IP_SCTP\5CSUM_IP_TSO" \
695 "\6CSUM_IP_ISCSI\7CSUM_INNER_IP6_UDP\10CSUM_INNER_IP6_TCP" \
696 "\11CSUM_INNER_IP6_TSO\12CSUM_IP6_UDP\13CSUM_IP6_TCP\14CSUM_IP6_SCTP" \
697 "\15CSUM_IP6_TSO\16CSUM_IP6_ISCSI\17CSUM_INNER_IP\20CSUM_INNER_IP_UDP" \
698 "\21CSUM_INNER_IP_TCP\22CSUM_INNER_IP_TSO\23CSUM_ENCAP_VXLAN" \
699 "\24CSUM_ENCAP_RSVD1\25CSUM_INNER_L3_CALC\26CSUM_INNER_L3_VALID" \
700 "\27CSUM_INNER_L4_CALC\30CSUM_INNER_L4_VALID\31CSUM_L3_CALC" \
701 "\32CSUM_L3_VALID\33CSUM_L4_CALC\34CSUM_L4_VALID\35CSUM_L5_CALC" \
702 "\36CSUM_L5_VALID\37CSUM_COALESCED\40CSUM_SND_TAG"
704 /* CSUM flags compatibility mappings. */
705 #define CSUM_IP_CHECKED CSUM_L3_CALC
706 #define CSUM_IP_VALID CSUM_L3_VALID
707 #define CSUM_DATA_VALID CSUM_L4_VALID
708 #define CSUM_PSEUDO_HDR CSUM_L4_CALC
709 #define CSUM_SCTP_VALID CSUM_L4_VALID
710 #define CSUM_DELAY_DATA (CSUM_TCP|CSUM_UDP)
711 #define CSUM_DELAY_IP CSUM_IP /* Only v4, no v6 IP hdr csum */
712 #define CSUM_DELAY_DATA_IPV6 (CSUM_TCP_IPV6|CSUM_UDP_IPV6)
713 #define CSUM_DATA_VALID_IPV6 CSUM_DATA_VALID
714 #define CSUM_TCP CSUM_IP_TCP
715 #define CSUM_UDP CSUM_IP_UDP
716 #define CSUM_SCTP CSUM_IP_SCTP
717 #define CSUM_TSO (CSUM_IP_TSO|CSUM_IP6_TSO)
718 #define CSUM_INNER_TSO (CSUM_INNER_IP_TSO|CSUM_INNER_IP6_TSO)
719 #define CSUM_UDP_IPV6 CSUM_IP6_UDP
720 #define CSUM_TCP_IPV6 CSUM_IP6_TCP
721 #define CSUM_SCTP_IPV6 CSUM_IP6_SCTP
724 * mbuf types describing the content of the mbuf (including external storage).
726 #define MT_NOTMBUF 0 /* USED INTERNALLY ONLY! Object is not mbuf */
727 #define MT_DATA 1 /* dynamic (data) allocation */
728 #define MT_HEADER MT_DATA /* packet header, use M_PKTHDR instead */
730 #define MT_VENDOR1 4 /* for vendor-internal use */
731 #define MT_VENDOR2 5 /* for vendor-internal use */
732 #define MT_VENDOR3 6 /* for vendor-internal use */
733 #define MT_VENDOR4 7 /* for vendor-internal use */
735 #define MT_SONAME 8 /* socket name */
737 #define MT_EXP1 9 /* for experimental use */
738 #define MT_EXP2 10 /* for experimental use */
739 #define MT_EXP3 11 /* for experimental use */
740 #define MT_EXP4 12 /* for experimental use */
742 #define MT_CONTROL 14 /* extra-data protocol message */
743 #define MT_EXTCONTROL 15 /* control message with externalized contents */
744 #define MT_OOBDATA 16 /* expedited data */
746 #define MT_NOINIT 255 /* Not a type but a flag to allocate
747 a non-initialized mbuf */
750 * String names of mbuf-related UMA(9) and malloc(9) types. Exposed to
751 * !_KERNEL so that monitoring tools can look up the zones with
754 #define MBUF_MEM_NAME "mbuf"
755 #define MBUF_CLUSTER_MEM_NAME "mbuf_cluster"
756 #define MBUF_PACKET_MEM_NAME "mbuf_packet"
757 #define MBUF_JUMBOP_MEM_NAME "mbuf_jumbo_page"
758 #define MBUF_JUMBO9_MEM_NAME "mbuf_jumbo_9k"
759 #define MBUF_JUMBO16_MEM_NAME "mbuf_jumbo_16k"
760 #define MBUF_TAG_MEM_NAME "mbuf_tag"
761 #define MBUF_EXTREFCNT_MEM_NAME "mbuf_ext_refcnt"
762 #define MBUF_EXTPGS_MEM_NAME "mbuf_extpgs"
765 union if_snd_tag_alloc_params;
768 #define MBUF_CHECKSLEEP(how) do { \
769 if (how == M_WAITOK) \
770 WITNESS_WARN(WARN_GIANTOK | WARN_SLEEPOK, NULL, \
771 "Sleeping in \"%s\"", __func__); \
774 #define MBUF_CHECKSLEEP(how)
778 * Network buffer allocation API
780 * The rest of it is defined in kern/kern_mbuf.c
782 extern uma_zone_t zone_mbuf;
783 extern uma_zone_t zone_clust;
784 extern uma_zone_t zone_pack;
785 extern uma_zone_t zone_jumbop;
786 extern uma_zone_t zone_jumbo9;
787 extern uma_zone_t zone_jumbo16;
788 extern uma_zone_t zone_extpgs;
790 void mb_dupcl(struct mbuf *, struct mbuf *);
791 void mb_free_ext(struct mbuf *);
792 void mb_free_extpg(struct mbuf *);
793 void mb_free_mext_pgs(struct mbuf *);
794 struct mbuf *mb_alloc_ext_pgs(int, m_ext_free_t);
795 struct mbuf *mb_alloc_ext_plus_pages(int, int);
796 struct mbuf *mb_mapped_to_unmapped(struct mbuf *, int, int, int,
798 int mb_unmapped_compress(struct mbuf *m);
799 struct mbuf *mb_unmapped_to_ext(struct mbuf *m);
800 void mb_free_notready(struct mbuf *m, int count);
801 void m_adj(struct mbuf *, int);
802 void m_adj_decap(struct mbuf *, int);
803 int m_apply(struct mbuf *, int, int,
804 int (*)(void *, void *, u_int), void *);
805 int m_append(struct mbuf *, int, c_caddr_t);
806 void m_cat(struct mbuf *, struct mbuf *);
807 void m_catpkt(struct mbuf *, struct mbuf *);
808 int m_clget(struct mbuf *m, int how);
809 void *m_cljget(struct mbuf *m, int how, int size);
810 struct mbuf *m_collapse(struct mbuf *, int, int);
811 void m_copyback(struct mbuf *, int, int, c_caddr_t);
812 void m_copydata(const struct mbuf *, int, int, caddr_t);
813 struct mbuf *m_copym(struct mbuf *, int, int, int);
814 struct mbuf *m_copypacket(struct mbuf *, int);
815 void m_copy_pkthdr(struct mbuf *, struct mbuf *);
816 struct mbuf *m_copyup(struct mbuf *, int, int);
817 struct mbuf *m_defrag(struct mbuf *, int);
818 void m_demote_pkthdr(struct mbuf *);
819 void m_demote(struct mbuf *, int, int);
820 struct mbuf *m_devget(char *, int, int, struct ifnet *,
821 void (*)(char *, caddr_t, u_int));
822 void m_dispose_extcontrolm(struct mbuf *m);
823 struct mbuf *m_dup(const struct mbuf *, int);
824 int m_dup_pkthdr(struct mbuf *, const struct mbuf *, int);
825 void m_extadd(struct mbuf *, char *, u_int, m_ext_free_t,
826 void *, void *, int, int);
827 u_int m_fixhdr(struct mbuf *);
828 struct mbuf *m_fragment(struct mbuf *, int, int);
829 void m_freem(struct mbuf *);
830 void m_free_raw(struct mbuf *);
831 struct mbuf *m_get2(int, int, short, int);
832 struct mbuf *m_getjcl(int, short, int, int);
833 struct mbuf *m_getm2(struct mbuf *, int, int, short, int);
834 struct mbuf *m_getptr(struct mbuf *, int, int *);
835 u_int m_length(struct mbuf *, struct mbuf **);
836 int m_mbuftouio(struct uio *, const struct mbuf *, int);
837 void m_move_pkthdr(struct mbuf *, struct mbuf *);
838 int m_pkthdr_init(struct mbuf *, int);
839 struct mbuf *m_prepend(struct mbuf *, int, int);
840 void m_print(const struct mbuf *, int);
841 struct mbuf *m_pulldown(struct mbuf *, int, int, int *);
842 struct mbuf *m_pullup(struct mbuf *, int);
843 int m_sanity(struct mbuf *, int);
844 struct mbuf *m_split(struct mbuf *, int, int);
845 struct mbuf *m_uiotombuf(struct uio *, int, int, int, int);
846 int m_unmapped_uiomove(const struct mbuf *, int, struct uio *,
848 struct mbuf *m_unshare(struct mbuf *, int);
849 int m_snd_tag_alloc(struct ifnet *,
850 union if_snd_tag_alloc_params *, struct m_snd_tag **);
851 void m_snd_tag_init(struct m_snd_tag *, struct ifnet *, u_int);
852 void m_snd_tag_destroy(struct m_snd_tag *);
866 #if MJUMPAGESIZE != MCLBYTES
878 panic("%s: invalid cluster size %d", __func__, size);
885 * Associated an external reference counted buffer with an mbuf.
888 m_extaddref(struct mbuf *m, char *buf, u_int size, u_int *ref_cnt,
889 m_ext_free_t freef, void *arg1, void *arg2)
892 KASSERT(ref_cnt != NULL, ("%s: ref_cnt not provided", __func__));
894 atomic_add_int(ref_cnt, 1);
896 m->m_ext.ext_buf = buf;
897 m->m_ext.ext_cnt = ref_cnt;
898 m->m_data = m->m_ext.ext_buf;
899 m->m_ext.ext_size = size;
900 m->m_ext.ext_free = freef;
901 m->m_ext.ext_arg1 = arg1;
902 m->m_ext.ext_arg2 = arg2;
903 m->m_ext.ext_type = EXT_EXTREF;
904 m->m_ext.ext_flags = 0;
907 static __inline uma_zone_t
916 #if MJUMPAGESIZE != MCLBYTES
928 panic("%s: invalid cluster size %d", __func__, size);
935 * Initialize an mbuf with linear storage.
937 * Inline because the consumer text overhead will be roughly the same to
938 * initialize or call a function with this many parameters and M_PKTHDR
939 * should go away with constant propagation for !MGETHDR.
942 m_init(struct mbuf *m, int how, short type, int flags)
948 m->m_data = m->m_dat;
952 if (flags & M_PKTHDR)
953 error = m_pkthdr_init(m, how);
957 MBUF_PROBE5(m__init, m, how, type, flags, error);
961 static __inline struct mbuf *
962 m_get_raw(int how, short type)
968 args.type = type | MT_NOINIT;
969 m = uma_zalloc_arg(zone_mbuf, &args, how);
970 MBUF_PROBE3(m__get_raw, how, type, m);
974 static __inline struct mbuf *
975 m_get(int how, short type)
982 m = uma_zalloc_arg(zone_mbuf, &args, how);
983 MBUF_PROBE3(m__get, how, type, m);
987 static __inline struct mbuf *
988 m_gethdr_raw(int how, short type)
993 args.flags = M_PKTHDR;
994 args.type = type | MT_NOINIT;
995 m = uma_zalloc_arg(zone_mbuf, &args, how);
996 MBUF_PROBE3(m__gethdr_raw, how, type, m);
1000 static __inline struct mbuf *
1001 m_gethdr(int how, short type)
1004 struct mb_args args;
1006 args.flags = M_PKTHDR;
1008 m = uma_zalloc_arg(zone_mbuf, &args, how);
1009 MBUF_PROBE3(m__gethdr, how, type, m);
1013 static __inline struct mbuf *
1014 m_getcl(int how, short type, int flags)
1017 struct mb_args args;
1021 m = uma_zalloc_arg(zone_pack, &args, how);
1022 MBUF_PROBE4(m__getcl, how, type, flags, m);
1027 * XXX: m_cljset() is a dangerous API. One must attach only a new,
1028 * unreferenced cluster to an mbuf(9). It is not possible to assert
1029 * that, so care can be taken only by users of the API.
1031 static __inline void
1032 m_cljset(struct mbuf *m, void *cl, int type)
1040 #if MJUMPAGESIZE != MCLBYTES
1042 size = MJUMPAGESIZE;
1052 panic("%s: unknown cluster type %d", __func__, type);
1056 m->m_data = m->m_ext.ext_buf = cl;
1057 m->m_ext.ext_free = m->m_ext.ext_arg1 = m->m_ext.ext_arg2 = NULL;
1058 m->m_ext.ext_size = size;
1059 m->m_ext.ext_type = type;
1060 m->m_ext.ext_flags = EXT_FLAG_EMBREF;
1061 m->m_ext.ext_count = 1;
1062 m->m_flags |= M_EXT;
1063 MBUF_PROBE3(m__cljset, m, cl, type);
1066 static __inline void
1067 m_chtype(struct mbuf *m, short new_type)
1070 m->m_type = new_type;
1073 static __inline void
1074 m_clrprotoflags(struct mbuf *m)
1078 m->m_flags &= ~M_PROTOFLAGS;
1083 static __inline struct mbuf *
1084 m_last(struct mbuf *m)
1093 m_extrefcnt(struct mbuf *m)
1096 KASSERT(m->m_flags & M_EXT, ("%s: M_EXT missing", __func__));
1098 return ((m->m_ext.ext_flags & EXT_FLAG_EMBREF) ? m->m_ext.ext_count :
1103 * mbuf, cluster, and external object allocation macros (for compatibility
1106 #define M_MOVE_PKTHDR(to, from) m_move_pkthdr((to), (from))
1107 #define MGET(m, how, type) ((m) = m_get((how), (type)))
1108 #define MGETHDR(m, how, type) ((m) = m_gethdr((how), (type)))
1109 #define MCLGET(m, how) m_clget((m), (how))
1110 #define MEXTADD(m, buf, size, free, arg1, arg2, flags, type) \
1111 m_extadd((m), (char *)(buf), (size), (free), (arg1), (arg2), \
1113 #define m_getm(m, len, how, type) \
1114 m_getm2((m), (len), (how), (type), M_PKTHDR)
1117 * Evaluate TRUE if it's safe to write to the mbuf m's data region (this can
1118 * be both the local data payload, or an external buffer area, depending on
1119 * whether M_EXT is set).
1121 #define M_WRITABLE(m) (((m)->m_flags & (M_RDONLY | M_EXTPG)) == 0 && \
1122 (!(((m)->m_flags & M_EXT)) || \
1123 (m_extrefcnt(m) == 1)))
1125 /* Check if the supplied mbuf has a packet header, or else panic. */
1126 #define M_ASSERTPKTHDR(m) \
1127 KASSERT((m) != NULL && (m)->m_flags & M_PKTHDR, \
1128 ("%s: no mbuf packet header!", __func__))
1130 /* Check if mbuf is multipage. */
1131 #define M_ASSERTEXTPG(m) \
1132 KASSERT(((m)->m_flags & (M_EXTPG|M_PKTHDR)) == M_EXTPG, \
1133 ("%s: m %p is not multipage!", __func__, m))
1136 * Ensure that the supplied mbuf is a valid, non-free mbuf.
1138 * XXX: Broken at the moment. Need some UMA magic to make it work again.
1140 #define M_ASSERTVALID(m) \
1141 KASSERT((((struct mbuf *)m)->m_flags & 0) == 0, \
1142 ("%s: attempted use of a free mbuf!", __func__))
1144 /* Check whether any mbuf in the chain is unmapped. */
1146 #define M_ASSERTMAPPED(m) do { \
1147 for (struct mbuf *__m = (m); __m != NULL; __m = __m->m_next) \
1148 KASSERT((__m->m_flags & M_EXTPG) == 0, \
1149 ("%s: chain %p contains an unmapped mbuf", __func__, (m)));\
1152 #define M_ASSERTMAPPED(m)
1156 * Return the address of the start of the buffer associated with an mbuf,
1157 * handling external storage, packet-header mbufs, and regular data mbufs.
1159 #define M_START(m) \
1160 (((m)->m_flags & M_EXTPG) ? NULL : \
1161 ((m)->m_flags & M_EXT) ? (m)->m_ext.ext_buf : \
1162 ((m)->m_flags & M_PKTHDR) ? &(m)->m_pktdat[0] : \
1166 * Return the size of the buffer associated with an mbuf, handling external
1167 * storage, packet-header mbufs, and regular data mbufs.
1170 (((m)->m_flags & M_EXT) ? (m)->m_ext.ext_size : \
1171 ((m)->m_flags & M_PKTHDR) ? MHLEN : \
1175 * Set the m_data pointer of a newly allocated mbuf to place an object of the
1176 * specified size at the end of the mbuf, longword aligned.
1178 * NB: Historically, we had M_ALIGN(), MH_ALIGN(), and MEXT_ALIGN() as
1179 * separate macros, each asserting that it was called at the proper moment.
1180 * This required callers to themselves test the storage type and call the
1181 * right one. Rather than require callers to be aware of those layout
1182 * decisions, we centralize here.
1184 static __inline void
1185 m_align(struct mbuf *m, int len)
1188 const char *msg = "%s: not a virgin mbuf";
1192 KASSERT(m->m_data == M_START(m), (msg, __func__));
1194 adjust = M_SIZE(m) - len;
1195 m->m_data += adjust &~ (sizeof(long)-1);
1198 #define M_ALIGN(m, len) m_align(m, len)
1199 #define MH_ALIGN(m, len) m_align(m, len)
1200 #define MEXT_ALIGN(m, len) m_align(m, len)
1203 * Compute the amount of space available before the current start of data in
1206 * The M_WRITABLE() is a temporary, conservative safety measure: the burden
1207 * of checking writability of the mbuf data area rests solely with the caller.
1209 * NB: In previous versions, M_LEADINGSPACE() would only check M_WRITABLE()
1210 * for mbufs with external storage. We now allow mbuf-embedded data to be
1211 * read-only as well.
1213 #define M_LEADINGSPACE(m) \
1214 (M_WRITABLE(m) ? ((m)->m_data - M_START(m)) : 0)
1217 * Compute the amount of space available after the end of data in an mbuf.
1219 * The M_WRITABLE() is a temporary, conservative safety measure: the burden
1220 * of checking writability of the mbuf data area rests solely with the caller.
1222 * NB: In previous versions, M_TRAILINGSPACE() would only check M_WRITABLE()
1223 * for mbufs with external storage. We now allow mbuf-embedded data to be
1224 * read-only as well.
1226 #define M_TRAILINGSPACE(m) \
1228 ((M_START(m) + M_SIZE(m)) - ((m)->m_data + (m)->m_len)) : 0)
1231 * Arrange to prepend space of size plen to mbuf m. If a new mbuf must be
1232 * allocated, how specifies whether to wait. If the allocation fails, the
1233 * original mbuf chain is freed and m is set to NULL.
1235 #define M_PREPEND(m, plen, how) do { \
1236 struct mbuf **_mmp = &(m); \
1237 struct mbuf *_mm = *_mmp; \
1238 int _mplen = (plen); \
1239 int __mhow = (how); \
1241 MBUF_CHECKSLEEP(how); \
1242 if (M_LEADINGSPACE(_mm) >= _mplen) { \
1243 _mm->m_data -= _mplen; \
1244 _mm->m_len += _mplen; \
1246 _mm = m_prepend(_mm, _mplen, __mhow); \
1247 if (_mm != NULL && _mm->m_flags & M_PKTHDR) \
1248 _mm->m_pkthdr.len += _mplen; \
1253 * Change mbuf to new type. This is a relatively expensive operation and
1254 * should be avoided.
1256 #define MCHTYPE(m, t) m_chtype((m), (t))
1258 /* Return the rcvif of a packet header. */
1259 static __inline struct ifnet *
1260 m_rcvif(struct mbuf *m)
1264 if (m->m_pkthdr.csum_flags & CSUM_SND_TAG)
1266 return (m->m_pkthdr.rcvif);
1269 /* Length to m_copy to copy all. */
1270 #define M_COPYALL 1000000000
1272 extern int max_datalen; /* MHLEN - max_hdr */
1273 extern int max_hdr; /* Largest link + protocol header */
1274 extern int max_linkhdr; /* Largest link-level header */
1275 extern int max_protohdr; /* Largest protocol header */
1276 extern int nmbclusters; /* Maximum number of clusters */
1277 extern bool mb_use_ext_pgs; /* Use ext_pgs for sendfile */
1280 * Network packets may have annotations attached by affixing a list of
1281 * "packet tags" to the pkthdr structure. Packet tags are dynamically
1282 * allocated semi-opaque data structures that have a fixed header
1283 * (struct m_tag) that specifies the size of the memory block and a
1284 * <cookie,type> pair that identifies it. The cookie is a 32-bit unique
1285 * unsigned value used to identify a module or ABI. By convention this value
1286 * is chosen as the date+time that the module is created, expressed as the
1287 * number of seconds since the epoch (e.g., using date -u +'%s'). The type
1288 * value is an ABI/module-specific value that identifies a particular
1289 * annotation and is private to the module. For compatibility with systems
1290 * like OpenBSD that define packet tags w/o an ABI/module cookie, the value
1291 * PACKET_ABI_COMPAT is used to implement m_tag_get and m_tag_find
1292 * compatibility shim functions and several tag types are defined below.
1293 * Users that do not require compatibility should use a private cookie value
1294 * so that packet tag-related definitions can be maintained privately.
1296 * Note that the packet tag returned by m_tag_alloc has the default memory
1297 * alignment implemented by malloc. To reference private data one can use a
1300 * struct m_tag *mtag = m_tag_alloc(...);
1301 * struct foo *p = (struct foo *)(mtag+1);
1303 * if the alignment of struct m_tag is sufficient for referencing members of
1304 * struct foo. Otherwise it is necessary to embed struct m_tag within the
1305 * private data structure to insure proper alignment; e.g.,
1311 * struct foo *p = (struct foo *) m_tag_alloc(...);
1312 * struct m_tag *mtag = &p->tag;
1316 * Persistent tags stay with an mbuf until the mbuf is reclaimed. Otherwise
1317 * tags are expected to ``vanish'' when they pass through a network
1318 * interface. For most interfaces this happens normally as the tags are
1319 * reclaimed when the mbuf is free'd. However in some special cases
1320 * reclaiming must be done manually. An example is packets that pass through
1321 * the loopback interface. Also, one must be careful to do this when
1322 * ``turning around'' packets (e.g., icmp_reflect).
1324 * To mark a tag persistent bit-or this flag in when defining the tag id.
1325 * The tag will then be treated as described above.
1327 #define MTAG_PERSISTENT 0x800
1329 #define PACKET_TAG_NONE 0 /* Nadda */
1331 /* Packet tags for use with PACKET_ABI_COMPAT. */
1332 #define PACKET_TAG_IPSEC_IN_DONE 1 /* IPsec applied, in */
1333 #define PACKET_TAG_IPSEC_OUT_DONE 2 /* IPsec applied, out */
1334 #define PACKET_TAG_IPSEC_IN_CRYPTO_DONE 3 /* NIC IPsec crypto done */
1335 #define PACKET_TAG_IPSEC_OUT_CRYPTO_NEEDED 4 /* NIC IPsec crypto req'ed */
1336 #define PACKET_TAG_IPSEC_IN_COULD_DO_CRYPTO 5 /* NIC notifies IPsec */
1337 #define PACKET_TAG_IPSEC_PENDING_TDB 6 /* Reminder to do IPsec */
1338 #define PACKET_TAG_BRIDGE 7 /* Bridge processing done */
1339 #define PACKET_TAG_GIF 8 /* GIF processing done */
1340 #define PACKET_TAG_GRE 9 /* GRE processing done */
1341 #define PACKET_TAG_IN_PACKET_CHECKSUM 10 /* NIC checksumming done */
1342 #define PACKET_TAG_ENCAP 11 /* Encap. processing */
1343 #define PACKET_TAG_IPSEC_SOCKET 12 /* IPSEC socket ref */
1344 #define PACKET_TAG_IPSEC_HISTORY 13 /* IPSEC history */
1345 #define PACKET_TAG_IPV6_INPUT 14 /* IPV6 input processing */
1346 #define PACKET_TAG_DUMMYNET 15 /* dummynet info */
1347 #define PACKET_TAG_DIVERT 17 /* divert info */
1348 #define PACKET_TAG_IPFORWARD 18 /* ipforward info */
1349 #define PACKET_TAG_MACLABEL (19 | MTAG_PERSISTENT) /* MAC label */
1350 #define PACKET_TAG_PF (21 | MTAG_PERSISTENT) /* PF/ALTQ information */
1351 #define PACKET_TAG_RTSOCKFAM 25 /* rtsock sa family */
1352 #define PACKET_TAG_IPOPTIONS 27 /* Saved IP options */
1353 #define PACKET_TAG_CARP 28 /* CARP info */
1354 #define PACKET_TAG_IPSEC_NAT_T_PORTS 29 /* two uint16_t */
1355 #define PACKET_TAG_ND_OUTGOING 30 /* ND outgoing */
1357 /* Specific cookies and tags. */
1359 /* Packet tag routines. */
1360 struct m_tag *m_tag_alloc(u_int32_t, int, int, int);
1361 void m_tag_delete(struct mbuf *, struct m_tag *);
1362 void m_tag_delete_chain(struct mbuf *, struct m_tag *);
1363 void m_tag_free_default(struct m_tag *);
1364 struct m_tag *m_tag_locate(struct mbuf *, u_int32_t, int, struct m_tag *);
1365 struct m_tag *m_tag_copy(struct m_tag *, int);
1366 int m_tag_copy_chain(struct mbuf *, const struct mbuf *, int);
1367 void m_tag_delete_nonpersistent(struct mbuf *);
1370 * Initialize the list of tags associated with an mbuf.
1372 static __inline void
1373 m_tag_init(struct mbuf *m)
1376 SLIST_INIT(&m->m_pkthdr.tags);
1380 * Set up the contents of a tag. Note that this does not fill in the free
1381 * method; the caller is expected to do that.
1383 * XXX probably should be called m_tag_init, but that was already taken.
1385 static __inline void
1386 m_tag_setup(struct m_tag *t, u_int32_t cookie, int type, int len)
1391 t->m_tag_cookie = cookie;
1395 * Reclaim resources associated with a tag.
1397 static __inline void
1398 m_tag_free(struct m_tag *t)
1401 (*t->m_tag_free)(t);
1405 * Return the first tag associated with an mbuf.
1407 static __inline struct m_tag *
1408 m_tag_first(struct mbuf *m)
1411 return (SLIST_FIRST(&m->m_pkthdr.tags));
1415 * Return the next tag in the list of tags associated with an mbuf.
1417 static __inline struct m_tag *
1418 m_tag_next(struct mbuf *m __unused, struct m_tag *t)
1421 return (SLIST_NEXT(t, m_tag_link));
1425 * Prepend a tag to the list of tags associated with an mbuf.
1427 static __inline void
1428 m_tag_prepend(struct mbuf *m, struct m_tag *t)
1431 SLIST_INSERT_HEAD(&m->m_pkthdr.tags, t, m_tag_link);
1435 * Unlink a tag from the list of tags associated with an mbuf.
1437 static __inline void
1438 m_tag_unlink(struct mbuf *m, struct m_tag *t)
1441 SLIST_REMOVE(&m->m_pkthdr.tags, t, m_tag, m_tag_link);
1444 /* These are for OpenBSD compatibility. */
1445 #define MTAG_ABI_COMPAT 0 /* compatibility ABI */
1447 static __inline struct m_tag *
1448 m_tag_get(int type, int length, int wait)
1450 return (m_tag_alloc(MTAG_ABI_COMPAT, type, length, wait));
1453 static __inline struct m_tag *
1454 m_tag_find(struct mbuf *m, int type, struct m_tag *start)
1456 return (SLIST_EMPTY(&m->m_pkthdr.tags) ? (struct m_tag *)NULL :
1457 m_tag_locate(m, MTAG_ABI_COMPAT, type, start));
1460 static inline struct m_snd_tag *
1461 m_snd_tag_ref(struct m_snd_tag *mst)
1464 refcount_acquire(&mst->refcount);
1469 m_snd_tag_rele(struct m_snd_tag *mst)
1472 if (refcount_release(&mst->refcount))
1473 m_snd_tag_destroy(mst);
1476 static __inline struct mbuf *
1477 m_free(struct mbuf *m)
1479 struct mbuf *n = m->m_next;
1481 MBUF_PROBE1(m__free, m);
1482 if ((m->m_flags & (M_PKTHDR|M_NOFREE)) == (M_PKTHDR|M_NOFREE))
1483 m_tag_delete_chain(m, NULL);
1484 if (m->m_flags & M_PKTHDR && m->m_pkthdr.csum_flags & CSUM_SND_TAG)
1485 m_snd_tag_rele(m->m_pkthdr.snd_tag);
1486 if (m->m_flags & M_EXTPG)
1488 else if (m->m_flags & M_EXT)
1490 else if ((m->m_flags & M_NOFREE) == 0)
1491 uma_zfree(zone_mbuf, m);
1496 rt_m_getfib(struct mbuf *m)
1498 KASSERT(m->m_flags & M_PKTHDR , ("Attempt to get FIB from non header mbuf."));
1499 return (m->m_pkthdr.fibnum);
1502 #define M_GETFIB(_m) rt_m_getfib(_m)
1504 #define M_SETFIB(_m, _fib) do { \
1505 KASSERT((_m)->m_flags & M_PKTHDR, ("Attempt to set FIB on non header mbuf.")); \
1506 ((_m)->m_pkthdr.fibnum) = (_fib); \
1509 /* flags passed as first argument for "m_xxx_tcpip_hash()" */
1510 #define MBUF_HASHFLAG_L2 (1 << 2)
1511 #define MBUF_HASHFLAG_L3 (1 << 3)
1512 #define MBUF_HASHFLAG_L4 (1 << 4)
1514 /* mbuf hashing helper routines */
1515 uint32_t m_ether_tcpip_hash_init(void);
1516 uint32_t m_ether_tcpip_hash(const uint32_t, const struct mbuf *, uint32_t);
1517 uint32_t m_infiniband_tcpip_hash_init(void);
1518 uint32_t m_infiniband_tcpip_hash(const uint32_t, const struct mbuf *, uint32_t);
1520 #ifdef MBUF_PROFILING
1521 void m_profile(struct mbuf *m);
1522 #define M_PROFILE(m) m_profile(m)
1524 #define M_PROFILE(m)
1528 STAILQ_HEAD(, mbuf) mq_head;
1534 mbufq_init(struct mbufq *mq, int maxlen)
1537 STAILQ_INIT(&mq->mq_head);
1538 mq->mq_maxlen = maxlen;
1542 static inline struct mbuf *
1543 mbufq_flush(struct mbufq *mq)
1547 m = STAILQ_FIRST(&mq->mq_head);
1548 STAILQ_INIT(&mq->mq_head);
1554 mbufq_drain(struct mbufq *mq)
1558 n = mbufq_flush(mq);
1559 while ((m = n) != NULL) {
1560 n = STAILQ_NEXT(m, m_stailqpkt);
1565 static inline struct mbuf *
1566 mbufq_first(const struct mbufq *mq)
1569 return (STAILQ_FIRST(&mq->mq_head));
1572 static inline struct mbuf *
1573 mbufq_last(const struct mbufq *mq)
1576 return (STAILQ_LAST(&mq->mq_head, mbuf, m_stailqpkt));
1580 mbufq_full(const struct mbufq *mq)
1583 return (mq->mq_maxlen > 0 && mq->mq_len >= mq->mq_maxlen);
1587 mbufq_len(const struct mbufq *mq)
1590 return (mq->mq_len);
1594 mbufq_enqueue(struct mbufq *mq, struct mbuf *m)
1599 STAILQ_INSERT_TAIL(&mq->mq_head, m, m_stailqpkt);
1604 static inline struct mbuf *
1605 mbufq_dequeue(struct mbufq *mq)
1609 m = STAILQ_FIRST(&mq->mq_head);
1611 STAILQ_REMOVE_HEAD(&mq->mq_head, m_stailqpkt);
1612 m->m_nextpkt = NULL;
1619 mbufq_prepend(struct mbufq *mq, struct mbuf *m)
1622 STAILQ_INSERT_HEAD(&mq->mq_head, m, m_stailqpkt);
1627 * Note: this doesn't enforce the maximum list size for dst.
1630 mbufq_concat(struct mbufq *mq_dst, struct mbufq *mq_src)
1633 mq_dst->mq_len += mq_src->mq_len;
1634 STAILQ_CONCAT(&mq_dst->mq_head, &mq_src->mq_head);
1638 #ifdef _SYS_TIMESPEC_H_
1640 mbuf_tstmp2timespec(struct mbuf *m, struct timespec *ts)
1643 KASSERT((m->m_flags & M_PKTHDR) != 0, ("mbuf %p no M_PKTHDR", m));
1644 KASSERT((m->m_flags & (M_TSTMP|M_TSTMP_LRO)) != 0, ("mbuf %p no M_TSTMP or M_TSTMP_LRO", m));
1645 ts->tv_sec = m->m_pkthdr.rcv_tstmp / 1000000000;
1646 ts->tv_nsec = m->m_pkthdr.rcv_tstmp % 1000000000;
1651 /* Invoked from the debugnet client code. */
1652 void debugnet_mbuf_drain(void);
1653 void debugnet_mbuf_start(void);
1654 void debugnet_mbuf_finish(void);
1655 void debugnet_mbuf_reinit(int nmbuf, int nclust, int clsize);
1659 mbuf_has_tls_session(struct mbuf *m)
1662 if (m->m_flags & M_EXTPG) {
1663 if (m->m_epg_tls != NULL) {
1670 #endif /* _KERNEL */
1671 #endif /* !_SYS_MBUF_H_ */