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__getjcl);
69 SDT_PROBE_DECLARE(sdt, , , m__clget);
70 SDT_PROBE_DECLARE(sdt, , , m__cljget);
71 SDT_PROBE_DECLARE(sdt, , , m__cljset);
72 SDT_PROBE_DECLARE(sdt, , , m__free);
73 SDT_PROBE_DECLARE(sdt, , , m__freem);
78 * Mbufs are of a single size, MSIZE (sys/param.h), which includes overhead.
79 * An mbuf may add a single "mbuf cluster" of size MCLBYTES (also in
80 * sys/param.h), which has no additional overhead and is used instead of the
81 * internal data area; this is done when at least MINCLSIZE of data must be
82 * stored. Additionally, it is possible to allocate a separate buffer
83 * externally and attach it to the mbuf in a way similar to that of mbuf
86 * NB: These calculation do not take actual compiler-induced alignment and
87 * padding inside the complete struct mbuf into account. Appropriate
88 * attention is required when changing members of struct mbuf.
90 * MLEN is data length in a normal mbuf.
91 * MHLEN is data length in an mbuf with pktheader.
92 * MINCLSIZE is a smallest amount of data that should be put into cluster.
94 * Compile-time assertions in uipc_mbuf.c test these values to ensure that
98 #define MHSIZE offsetof(struct mbuf, m_dat)
99 #define MPKTHSIZE offsetof(struct mbuf, m_pktdat)
100 #define MLEN ((int)(MSIZE - MHSIZE))
101 #define MHLEN ((int)(MSIZE - MPKTHSIZE))
102 #define MINCLSIZE (MHLEN + 1)
107 * Macro for type conversion: convert mbuf pointer to data pointer of correct
110 * mtod(m, t) -- Convert mbuf pointer to data pointer of correct type.
111 * mtodo(m, o) -- Same as above but with offset 'o' into data.
113 #define mtod(m, t) ((t)((m)->m_data))
114 #define mtodo(m, o) ((void *)(((m)->m_data) + (o)))
117 * Argument structure passed to UMA routines during mbuf and packet
121 int flags; /* Flags for mbuf being allocated */
122 short type; /* Type of mbuf being allocated */
127 * Packet tag structure (see below for details).
130 SLIST_ENTRY(m_tag) m_tag_link; /* List of packet tags */
131 u_int16_t m_tag_id; /* Tag ID */
132 u_int16_t m_tag_len; /* Length of data */
133 u_int32_t m_tag_cookie; /* ABI/Module ID */
134 void (*m_tag_free)(struct m_tag *);
138 * Static network interface owned tag.
139 * Allocated through ifp->if_snd_tag_alloc().
142 struct ifnet *ifp; /* network interface tag belongs to */
143 volatile u_int refcount;
144 u_int type; /* One of IF_SND_TAG_TYPE_*. */
148 * Record/packet header in first mbuf of chain; valid only if M_PKTHDR is set.
151 * Compile-time assertions in uipc_mbuf.c test these values to ensure that
156 struct m_snd_tag *snd_tag; /* send tag, if any */
157 struct ifnet *rcvif; /* rcv interface */
159 SLIST_HEAD(packet_tags, m_tag) tags; /* list of packet tags */
160 int32_t len; /* total packet length */
162 /* Layer crossing persistent information. */
163 uint32_t flowid; /* packet's 4-tuple system */
164 uint32_t csum_flags; /* checksum and offload features */
165 uint16_t fibnum; /* this packet should use this fib */
166 uint8_t numa_domain; /* NUMA domain of recvd pkt */
167 uint8_t rsstype; /* hash type */
169 uint64_t rcv_tstmp; /* timestamp in ns */
171 uint8_t l2hlen; /* layer 2 hdr len */
172 uint8_t l3hlen; /* layer 3 hdr len */
173 uint8_t l4hlen; /* layer 4 hdr len */
174 uint8_t l5hlen; /* layer 5 hdr len */
175 uint8_t inner_l2hlen;
176 uint8_t inner_l3hlen;
177 uint8_t inner_l4hlen;
178 uint8_t inner_l5hlen;
184 uint32_t thirtytwo[2];
185 uint64_t sixtyfour[1];
186 uintptr_t unintptr[1];
190 /* Layer specific non-persistent local storage for reassembly, etc. */
194 uint32_t thirtytwo[2];
195 uint64_t sixtyfour[1];
196 uintptr_t unintptr[1];
200 #define ether_vtag PH_per.sixteen[0]
201 #define tcp_tun_port PH_per.sixteen[0] /* outbound */
203 #define vt_nrecs sixteen[0] /* mld and v6-ND */
204 #define tso_segsz PH_per.sixteen[1] /* inbound after LRO */
205 #define lro_nsegs tso_segsz /* inbound after LRO */
206 #define csum_data PH_per.thirtytwo[1] /* inbound from hardware up */
207 #define lro_tcp_d_len PH_loc.sixteen[0] /* inbound during LRO (no reassembly) */
208 #define lro_tcp_d_csum PH_loc.sixteen[1] /* inbound during LRO (no reassembly) */
209 #define lro_tcp_h_off PH_loc.sixteen[2] /* inbound during LRO (no reassembly) */
210 #define lro_etype PH_loc.sixteen[3] /* inbound during LRO (no reassembly) */
211 /* Note PH_loc is used during IP reassembly (all 8 bytes as a ptr) */
214 * TLS records for TLS 1.0-1.2 can have the following header lengths:
215 * - 5 (AES-CBC with implicit IV)
216 * - 21 (AES-CBC with explicit IV)
217 * - 13 (AES-GCM with 8 byte explicit IV)
219 #define MBUF_PEXT_HDR_LEN 23
222 * TLS records for TLS 1.0-1.2 can have the following maximum trailer
225 * - 36 (AES-CBC with SHA1 and up to 16 bytes of padding)
226 * - 48 (AES-CBC with SHA2-256 and up to 16 bytes of padding)
227 * - 64 (AES-CBC with SHA2-384 and up to 16 bytes of padding)
229 #define MBUF_PEXT_TRAIL_LEN 64
231 #if defined(__LP64__)
232 #define MBUF_PEXT_MAX_PGS (40 / sizeof(vm_paddr_t))
234 #define MBUF_PEXT_MAX_PGS (72 / sizeof(vm_paddr_t))
237 #define MBUF_PEXT_MAX_BYTES \
238 (MBUF_PEXT_MAX_PGS * PAGE_SIZE + MBUF_PEXT_HDR_LEN + MBUF_PEXT_TRAIL_LEN)
244 * Description of external storage mapped into mbuf; valid only if M_EXT is
248 * Compile-time assertions in uipc_mbuf.c test these values to ensure that
251 typedef void m_ext_free_t(struct mbuf *);
255 * If EXT_FLAG_EMBREF is set, then we use refcount in the
256 * mbuf, the 'ext_count' member. Otherwise, we have a
257 * shadow copy and we use pointer 'ext_cnt'. The original
258 * mbuf is responsible to carry the pointer to free routine
259 * and its arguments. They aren't copied into shadows in
260 * mb_dupcl() to avoid dereferencing next cachelines.
262 volatile u_int ext_count;
263 volatile u_int *ext_cnt;
265 uint32_t ext_size; /* size of buffer, for ext_free */
266 uint32_t ext_type:8, /* type of external storage */
267 ext_flags:24; /* external storage mbuf flags */
271 * Regular M_EXT mbuf:
272 * o ext_buf always points to the external buffer.
273 * o ext_free (below) and two optional arguments
274 * ext_arg1 and ext_arg2 store the free context for
275 * the external storage. They are set only in the
276 * refcount carrying mbuf, the one with
277 * EXT_FLAG_EMBREF flag, with exclusion for
278 * EXT_EXTREF type, where the free context is copied
279 * into all mbufs that use same external storage.
281 char *ext_buf; /* start of buffer */
282 #define m_ext_copylen offsetof(struct m_ext, ext_arg2)
287 * Multi-page M_EXTPG mbuf:
288 * o extpg_pa - page vector.
289 * o extpg_trail and extpg_hdr - TLS trailer and
291 * Uses ext_free and may also use ext_arg1.
293 vm_paddr_t extpg_pa[MBUF_PEXT_MAX_PGS];
294 char extpg_trail[MBUF_PEXT_TRAIL_LEN];
295 char extpg_hdr[MBUF_PEXT_HDR_LEN];
296 /* Pretend these 3 fields are part of mbuf itself. */
297 #define m_epg_pa m_ext.extpg_pa
298 #define m_epg_trail m_ext.extpg_trail
299 #define m_epg_hdr m_ext.extpg_hdr
300 #define m_epg_ext_copylen offsetof(struct m_ext, ext_free)
304 * Free method and optional argument pointer, both
305 * used by M_EXT and M_EXTPG.
307 m_ext_free_t *ext_free;
312 * The core of the mbuf object along with some shortcut defines for practical
317 * Header present at the beginning of every mbuf.
320 * Compile-time assertions in uipc_mbuf.c test these values to ensure
321 * that they are correct.
323 union { /* next buffer in chain */
325 SLIST_ENTRY(mbuf) m_slist;
326 STAILQ_ENTRY(mbuf) m_stailq;
328 union { /* next chain in queue/record */
329 struct mbuf *m_nextpkt;
330 SLIST_ENTRY(mbuf) m_slistpkt;
331 STAILQ_ENTRY(mbuf) m_stailqpkt;
333 caddr_t m_data; /* location of data */
334 int32_t m_len; /* amount of data in this mbuf */
335 uint32_t m_type:8, /* type of data in this mbuf */
336 m_flags:24; /* flags; see below */
337 #if !defined(__LP64__)
338 uint32_t m_pad; /* pad for 64bit alignment */
342 * A set of optional headers (packet header, external storage header)
343 * and internal data storage. Historically, these arrays were sized
344 * to MHLEN (space left after a packet header) and MLEN (space left
345 * after only a regular mbuf header); they are now variable size in
346 * order to support future work on variable-size mbufs.
352 struct pkthdr m_pkthdr;
355 * Multi-page M_EXTPG mbuf has its meta data
356 * split between the below anonymous structure
357 * and m_ext. It carries vector of pages,
358 * optional header and trailer char vectors
359 * and pointers to socket/TLS data.
361 #define m_epg_startcopy m_epg_npgs
362 #define m_epg_endcopy m_epg_stailq
364 /* Overall count of pages and count of
365 * pages with I/O pending. */
368 /* TLS header and trailer lengths.
369 * The data itself resides in m_ext. */
370 uint8_t m_epg_hdrlen;
371 uint8_t m_epg_trllen;
372 /* Offset into 1st page and length of
373 * data in the last page. */
374 uint16_t m_epg_1st_off;
375 uint16_t m_epg_last_len;
377 #define EPG_FLAG_ANON 0x1 /* Data can be encrypted in place. */
378 #define EPG_FLAG_2FREE 0x2 /* Scheduled for free. */
379 uint8_t m_epg_record_type;
382 struct ktls_session *m_epg_tls;
383 struct socket *m_epg_so;
384 uint64_t m_epg_seqno;
385 STAILQ_ENTRY(mbuf) m_epg_stailq;
389 /* M_EXT or M_EXTPG set. */
391 /* M_PKTHDR set, neither M_EXT nor M_EXTPG. */
395 char m_dat[0]; /* !M_PKTHDR, !M_EXT */
401 m_epg_pagelen(const struct mbuf *m, int pidx, int pgoff)
404 KASSERT(pgoff == 0 || pidx == 0,
405 ("page %d with non-zero offset %d in %p", pidx, pgoff, m));
407 if (pidx == m->m_epg_npgs - 1) {
408 return (m->m_epg_last_len);
410 return (PAGE_SIZE - pgoff);
415 #define MCHECK(ex, msg) KASSERT((ex), \
416 ("Multi page mbuf %p with " #msg " at %s:%d", \
417 m, __FILE__, __LINE__))
419 * NB: This expects a non-empty buffer (npgs > 0 and
422 #define MBUF_EXT_PGS_ASSERT_SANITY(m) do { \
423 MCHECK(m->m_epg_npgs > 0, "no valid pages"); \
424 MCHECK(m->m_epg_npgs <= nitems(m->m_epg_pa), \
426 MCHECK(m->m_epg_nrdy <= m->m_epg_npgs, \
427 "too many ready pages"); \
428 MCHECK(m->m_epg_1st_off < PAGE_SIZE, \
429 "too large page offset"); \
430 MCHECK(m->m_epg_last_len > 0, "zero last page length"); \
431 MCHECK(m->m_epg_last_len <= PAGE_SIZE, \
432 "too large last page length"); \
433 if (m->m_epg_npgs == 1) \
434 MCHECK(m->m_epg_1st_off + \
435 m->m_epg_last_len <= PAGE_SIZE, \
436 "single page too large"); \
437 MCHECK(m->m_epg_hdrlen <= sizeof(m->m_epg_hdr), \
438 "too large header length"); \
439 MCHECK(m->m_epg_trllen <= sizeof(m->m_epg_trail), \
440 "too large header length"); \
443 #define MBUF_EXT_PGS_ASSERT_SANITY(m) do {} while (0)
448 * mbuf flags of global significance and layer crossing.
449 * Those of only protocol/layer specific significance are to be mapped
450 * to M_PROTO[1-11] and cleared at layer handoff boundaries.
451 * NB: Limited to the lower 24 bits.
453 #define M_EXT 0x00000001 /* has associated external storage */
454 #define M_PKTHDR 0x00000002 /* start of record */
455 #define M_EOR 0x00000004 /* end of record */
456 #define M_RDONLY 0x00000008 /* associated data is marked read-only */
457 #define M_BCAST 0x00000010 /* send/received as link-level broadcast */
458 #define M_MCAST 0x00000020 /* send/received as link-level multicast */
459 #define M_PROMISC 0x00000040 /* packet was not for us */
460 #define M_VLANTAG 0x00000080 /* ether_vtag is valid */
461 #define M_EXTPG 0x00000100 /* has array of unmapped pages and TLS */
462 #define M_NOFREE 0x00000200 /* do not free mbuf, embedded in cluster */
463 #define M_TSTMP 0x00000400 /* rcv_tstmp field is valid */
464 #define M_TSTMP_HPREC 0x00000800 /* rcv_tstmp is high-prec, typically
465 hw-stamped on port (useful for IEEE 1588
467 #define M_TSTMP_LRO 0x00001000 /* Time LRO pushed in pkt is valid in (PH_loc) */
469 #define M_PROTO1 0x00002000 /* protocol-specific */
470 #define M_PROTO2 0x00004000 /* protocol-specific */
471 #define M_PROTO3 0x00008000 /* protocol-specific */
472 #define M_PROTO4 0x00010000 /* protocol-specific */
473 #define M_PROTO5 0x00020000 /* protocol-specific */
474 #define M_PROTO6 0x00040000 /* protocol-specific */
475 #define M_PROTO7 0x00080000 /* protocol-specific */
476 #define M_PROTO8 0x00100000 /* protocol-specific */
477 #define M_PROTO9 0x00200000 /* protocol-specific */
478 #define M_PROTO10 0x00400000 /* protocol-specific */
479 #define M_PROTO11 0x00800000 /* protocol-specific */
481 #define MB_DTOR_SKIP 0x1 /* don't pollute the cache by touching a freed mbuf */
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 * Flags preserved during demote.
500 #define M_DEMOTEFLAGS \
501 (M_EXT | M_RDONLY | M_NOFREE | M_EXTPG)
504 * Mbuf flag description for use with printf(9) %b identifier.
506 #define M_FLAG_BITS \
507 "\20\1M_EXT\2M_PKTHDR\3M_EOR\4M_RDONLY\5M_BCAST\6M_MCAST" \
508 "\7M_PROMISC\10M_VLANTAG\11M_EXTPG\12M_NOFREE\13M_TSTMP\14M_TSTMP_HPREC\15M_TSTMP_LRO"
509 #define M_FLAG_PROTOBITS \
510 "\16M_PROTO1\17M_PROTO2\20M_PROTO3\21M_PROTO4" \
511 "\22M_PROTO5\23M_PROTO6\24M_PROTO7\25M_PROTO8\26M_PROTO9" \
512 "\27M_PROTO10\28M_PROTO11"
513 #define M_FLAG_PRINTF (M_FLAG_BITS M_FLAG_PROTOBITS)
516 * Network interface cards are able to hash protocol fields (such as IPv4
517 * addresses and TCP port numbers) classify packets into flows. These flows
518 * can then be used to maintain ordering while delivering packets to the OS
519 * via parallel input queues, as well as to provide a stateless affinity
520 * model. NIC drivers can pass up the hash via m->m_pkthdr.flowid, and set
521 * m_flag fields to indicate how the hash should be interpreted by the
524 * Most NICs support RSS, which provides ordering and explicit affinity, and
525 * use the hash m_flag bits to indicate what header fields were covered by
526 * the hash. M_HASHTYPE_OPAQUE and M_HASHTYPE_OPAQUE_HASH can be set by non-
527 * RSS cards or configurations that provide an opaque flow identifier, allowing
528 * for ordering and distribution without explicit affinity. Additionally,
529 * M_HASHTYPE_OPAQUE_HASH indicates that the flow identifier has hash
532 * The meaning of the IPV6_EX suffix:
533 * "o Home address from the home address option in the IPv6 destination
534 * options header. If the extension header is not present, use the Source
536 * o IPv6 address that is contained in the Routing-Header-Type-2 from the
537 * associated extension header. If the extension header is not present,
538 * use the Destination IPv6 Address."
540 * https://docs.microsoft.com/en-us/windows-hardware/drivers/network/rss-hashing-types#ndishashipv6ex
542 #define M_HASHTYPE_HASHPROP 0x80 /* has hash properties */
543 #define M_HASHTYPE_INNER 0x40 /* calculated from inner headers */
544 #define M_HASHTYPE_HASH(t) (M_HASHTYPE_HASHPROP | (t))
545 /* Microsoft RSS standard hash types */
546 #define M_HASHTYPE_NONE 0
547 #define M_HASHTYPE_RSS_IPV4 M_HASHTYPE_HASH(1) /* IPv4 2-tuple */
548 #define M_HASHTYPE_RSS_TCP_IPV4 M_HASHTYPE_HASH(2) /* TCPv4 4-tuple */
549 #define M_HASHTYPE_RSS_IPV6 M_HASHTYPE_HASH(3) /* IPv6 2-tuple */
550 #define M_HASHTYPE_RSS_TCP_IPV6 M_HASHTYPE_HASH(4) /* TCPv6 4-tuple */
551 #define M_HASHTYPE_RSS_IPV6_EX M_HASHTYPE_HASH(5) /* IPv6 2-tuple +
553 #define M_HASHTYPE_RSS_TCP_IPV6_EX M_HASHTYPE_HASH(6) /* TCPv6 4-tuple +
555 #define M_HASHTYPE_RSS_UDP_IPV4 M_HASHTYPE_HASH(7) /* IPv4 UDP 4-tuple*/
556 #define M_HASHTYPE_RSS_UDP_IPV6 M_HASHTYPE_HASH(9) /* IPv6 UDP 4-tuple*/
557 #define M_HASHTYPE_RSS_UDP_IPV6_EX M_HASHTYPE_HASH(10)/* IPv6 UDP 4-tuple +
560 #define M_HASHTYPE_OPAQUE 0x3f /* ordering, not affinity */
561 #define M_HASHTYPE_OPAQUE_HASH M_HASHTYPE_HASH(M_HASHTYPE_OPAQUE)
562 /* ordering+hash, not affinity*/
564 #define M_HASHTYPE_CLEAR(m) ((m)->m_pkthdr.rsstype = 0)
565 #define M_HASHTYPE_GET(m) ((m)->m_pkthdr.rsstype & ~M_HASHTYPE_INNER)
566 #define M_HASHTYPE_SET(m, v) ((m)->m_pkthdr.rsstype = (v))
567 #define M_HASHTYPE_TEST(m, v) (M_HASHTYPE_GET(m) == (v))
568 #define M_HASHTYPE_ISHASH(m) \
569 (((m)->m_pkthdr.rsstype & M_HASHTYPE_HASHPROP) != 0)
570 #define M_HASHTYPE_SETINNER(m) do { \
571 (m)->m_pkthdr.rsstype |= M_HASHTYPE_INNER; \
575 * External mbuf storage buffer types.
577 #define EXT_CLUSTER 1 /* mbuf cluster */
578 #define EXT_SFBUF 2 /* sendfile(2)'s sf_buf */
579 #define EXT_JUMBOP 3 /* jumbo cluster page sized */
580 #define EXT_JUMBO9 4 /* jumbo cluster 9216 bytes */
581 #define EXT_JUMBO16 5 /* jumbo cluster 16184 bytes */
582 #define EXT_PACKET 6 /* mbuf+cluster from packet zone */
583 #define EXT_MBUF 7 /* external mbuf reference */
584 #define EXT_RXRING 8 /* data in NIC receive ring */
586 #define EXT_VENDOR1 224 /* for vendor-internal use */
587 #define EXT_VENDOR2 225 /* for vendor-internal use */
588 #define EXT_VENDOR3 226 /* for vendor-internal use */
589 #define EXT_VENDOR4 227 /* for vendor-internal use */
591 #define EXT_EXP1 244 /* for experimental use */
592 #define EXT_EXP2 245 /* for experimental use */
593 #define EXT_EXP3 246 /* for experimental use */
594 #define EXT_EXP4 247 /* for experimental use */
596 #define EXT_NET_DRV 252 /* custom ext_buf provided by net driver(s) */
597 #define EXT_MOD_TYPE 253 /* custom module's ext_buf type */
598 #define EXT_DISPOSABLE 254 /* can throw this buffer away w/page flipping */
599 #define EXT_EXTREF 255 /* has externally maintained ext_cnt ptr */
602 * Flags for external mbuf buffer types.
603 * NB: limited to the lower 24 bits.
605 #define EXT_FLAG_EMBREF 0x000001 /* embedded ext_count */
606 #define EXT_FLAG_EXTREF 0x000002 /* external ext_cnt, notyet */
608 #define EXT_FLAG_NOFREE 0x000010 /* don't free mbuf to pool, notyet */
610 #define EXT_FLAG_VENDOR1 0x010000 /* These flags are vendor */
611 #define EXT_FLAG_VENDOR2 0x020000 /* or submodule specific, */
612 #define EXT_FLAG_VENDOR3 0x040000 /* not used by mbuf code. */
613 #define EXT_FLAG_VENDOR4 0x080000 /* Set/read by submodule. */
615 #define EXT_FLAG_EXP1 0x100000 /* for experimental use */
616 #define EXT_FLAG_EXP2 0x200000 /* for experimental use */
617 #define EXT_FLAG_EXP3 0x400000 /* for experimental use */
618 #define EXT_FLAG_EXP4 0x800000 /* for experimental use */
621 * EXT flag description for use with printf(9) %b identifier.
623 #define EXT_FLAG_BITS \
624 "\20\1EXT_FLAG_EMBREF\2EXT_FLAG_EXTREF\5EXT_FLAG_NOFREE" \
625 "\21EXT_FLAG_VENDOR1\22EXT_FLAG_VENDOR2\23EXT_FLAG_VENDOR3" \
626 "\24EXT_FLAG_VENDOR4\25EXT_FLAG_EXP1\26EXT_FLAG_EXP2\27EXT_FLAG_EXP3" \
630 * Flags indicating checksum, segmentation and other offload work to be
631 * done, or already done, by hardware or lower layers. It is split into
632 * separate inbound and outbound flags.
634 * Outbound flags that are set by upper protocol layers requesting lower
635 * layers, or ideally the hardware, to perform these offloading tasks.
636 * For outbound packets this field and its flags can be directly tested
637 * against ifnet if_hwassist. Note that the outbound and the inbound flags do
638 * not collide right now but they could be allowed to (as long as the flags are
639 * scrubbed appropriately when the direction of an mbuf changes). CSUM_BITS
640 * would also have to split into CSUM_BITS_TX and CSUM_BITS_RX.
642 * CSUM_INNER_<x> is the same as CSUM_<x> but it applies to the inner frame.
643 * The CSUM_ENCAP_<x> bits identify the outer encapsulation.
645 #define CSUM_IP 0x00000001 /* IP header checksum offload */
646 #define CSUM_IP_UDP 0x00000002 /* UDP checksum offload */
647 #define CSUM_IP_TCP 0x00000004 /* TCP checksum offload */
648 #define CSUM_IP_SCTP 0x00000008 /* SCTP checksum offload */
649 #define CSUM_IP_TSO 0x00000010 /* TCP segmentation offload */
650 #define CSUM_IP_ISCSI 0x00000020 /* iSCSI checksum offload */
652 #define CSUM_INNER_IP6_UDP 0x00000040
653 #define CSUM_INNER_IP6_TCP 0x00000080
654 #define CSUM_INNER_IP6_TSO 0x00000100
655 #define CSUM_IP6_UDP 0x00000200 /* UDP checksum offload */
656 #define CSUM_IP6_TCP 0x00000400 /* TCP checksum offload */
657 #define CSUM_IP6_SCTP 0x00000800 /* SCTP checksum offload */
658 #define CSUM_IP6_TSO 0x00001000 /* TCP segmentation offload */
659 #define CSUM_IP6_ISCSI 0x00002000 /* iSCSI checksum offload */
661 #define CSUM_INNER_IP 0x00004000
662 #define CSUM_INNER_IP_UDP 0x00008000
663 #define CSUM_INNER_IP_TCP 0x00010000
664 #define CSUM_INNER_IP_TSO 0x00020000
666 #define CSUM_ENCAP_VXLAN 0x00040000 /* VXLAN outer encapsulation */
667 #define CSUM_ENCAP_RSVD1 0x00080000
669 /* Inbound checksum support where the checksum was verified by hardware. */
670 #define CSUM_INNER_L3_CALC 0x00100000
671 #define CSUM_INNER_L3_VALID 0x00200000
672 #define CSUM_INNER_L4_CALC 0x00400000
673 #define CSUM_INNER_L4_VALID 0x00800000
674 #define CSUM_L3_CALC 0x01000000 /* calculated layer 3 csum */
675 #define CSUM_L3_VALID 0x02000000 /* checksum is correct */
676 #define CSUM_L4_CALC 0x04000000 /* calculated layer 4 csum */
677 #define CSUM_L4_VALID 0x08000000 /* checksum is correct */
678 #define CSUM_L5_CALC 0x10000000 /* calculated layer 5 csum */
679 #define CSUM_L5_VALID 0x20000000 /* checksum is correct */
680 #define CSUM_COALESCED 0x40000000 /* contains merged segments */
682 #define CSUM_SND_TAG 0x80000000 /* Packet header has send tag */
684 #define CSUM_FLAGS_TX (CSUM_IP | CSUM_IP_UDP | CSUM_IP_TCP | CSUM_IP_SCTP | \
685 CSUM_IP_TSO | CSUM_IP_ISCSI | CSUM_INNER_IP6_UDP | CSUM_INNER_IP6_TCP | \
686 CSUM_INNER_IP6_TSO | CSUM_IP6_UDP | CSUM_IP6_TCP | CSUM_IP6_SCTP | \
687 CSUM_IP6_TSO | CSUM_IP6_ISCSI | CSUM_INNER_IP | CSUM_INNER_IP_UDP | \
688 CSUM_INNER_IP_TCP | CSUM_INNER_IP_TSO | CSUM_ENCAP_VXLAN | \
689 CSUM_ENCAP_RSVD1 | CSUM_SND_TAG)
691 #define CSUM_FLAGS_RX (CSUM_INNER_L3_CALC | CSUM_INNER_L3_VALID | \
692 CSUM_INNER_L4_CALC | CSUM_INNER_L4_VALID | CSUM_L3_CALC | CSUM_L3_VALID | \
693 CSUM_L4_CALC | CSUM_L4_VALID | CSUM_L5_CALC | CSUM_L5_VALID | \
697 * CSUM flag description for use with printf(9) %b identifier.
700 "\20\1CSUM_IP\2CSUM_IP_UDP\3CSUM_IP_TCP\4CSUM_IP_SCTP\5CSUM_IP_TSO" \
701 "\6CSUM_IP_ISCSI\7CSUM_INNER_IP6_UDP\10CSUM_INNER_IP6_TCP" \
702 "\11CSUM_INNER_IP6_TSO\12CSUM_IP6_UDP\13CSUM_IP6_TCP\14CSUM_IP6_SCTP" \
703 "\15CSUM_IP6_TSO\16CSUM_IP6_ISCSI\17CSUM_INNER_IP\20CSUM_INNER_IP_UDP" \
704 "\21CSUM_INNER_IP_TCP\22CSUM_INNER_IP_TSO\23CSUM_ENCAP_VXLAN" \
705 "\24CSUM_ENCAP_RSVD1\25CSUM_INNER_L3_CALC\26CSUM_INNER_L3_VALID" \
706 "\27CSUM_INNER_L4_CALC\30CSUM_INNER_L4_VALID\31CSUM_L3_CALC" \
707 "\32CSUM_L3_VALID\33CSUM_L4_CALC\34CSUM_L4_VALID\35CSUM_L5_CALC" \
708 "\36CSUM_L5_VALID\37CSUM_COALESCED\40CSUM_SND_TAG"
710 /* CSUM flags compatibility mappings. */
711 #define CSUM_IP_CHECKED CSUM_L3_CALC
712 #define CSUM_IP_VALID CSUM_L3_VALID
713 #define CSUM_DATA_VALID CSUM_L4_VALID
714 #define CSUM_PSEUDO_HDR CSUM_L4_CALC
715 #define CSUM_SCTP_VALID CSUM_L4_VALID
716 #define CSUM_DELAY_DATA (CSUM_TCP|CSUM_UDP)
717 #define CSUM_DELAY_IP CSUM_IP /* Only v4, no v6 IP hdr csum */
718 #define CSUM_DELAY_DATA_IPV6 (CSUM_TCP_IPV6|CSUM_UDP_IPV6)
719 #define CSUM_DATA_VALID_IPV6 CSUM_DATA_VALID
720 #define CSUM_TCP CSUM_IP_TCP
721 #define CSUM_UDP CSUM_IP_UDP
722 #define CSUM_SCTP CSUM_IP_SCTP
723 #define CSUM_TSO (CSUM_IP_TSO|CSUM_IP6_TSO)
724 #define CSUM_INNER_TSO (CSUM_INNER_IP_TSO|CSUM_INNER_IP6_TSO)
725 #define CSUM_UDP_IPV6 CSUM_IP6_UDP
726 #define CSUM_TCP_IPV6 CSUM_IP6_TCP
727 #define CSUM_SCTP_IPV6 CSUM_IP6_SCTP
728 #define CSUM_TLS_MASK (CSUM_L5_CALC|CSUM_L5_VALID)
729 #define CSUM_TLS_DECRYPTED CSUM_L5_CALC
732 * mbuf types describing the content of the mbuf (including external storage).
734 #define MT_NOTMBUF 0 /* USED INTERNALLY ONLY! Object is not mbuf */
735 #define MT_DATA 1 /* dynamic (data) allocation */
736 #define MT_HEADER MT_DATA /* packet header, use M_PKTHDR instead */
738 #define MT_VENDOR1 4 /* for vendor-internal use */
739 #define MT_VENDOR2 5 /* for vendor-internal use */
740 #define MT_VENDOR3 6 /* for vendor-internal use */
741 #define MT_VENDOR4 7 /* for vendor-internal use */
743 #define MT_SONAME 8 /* socket name */
745 #define MT_EXP1 9 /* for experimental use */
746 #define MT_EXP2 10 /* for experimental use */
747 #define MT_EXP3 11 /* for experimental use */
748 #define MT_EXP4 12 /* for experimental use */
750 #define MT_CONTROL 14 /* extra-data protocol message */
751 #define MT_EXTCONTROL 15 /* control message with externalized contents */
752 #define MT_OOBDATA 16 /* expedited data */
754 #define MT_NOINIT 255 /* Not a type but a flag to allocate
755 a non-initialized mbuf */
758 * String names of mbuf-related UMA(9) and malloc(9) types. Exposed to
759 * !_KERNEL so that monitoring tools can look up the zones with
762 #define MBUF_MEM_NAME "mbuf"
763 #define MBUF_CLUSTER_MEM_NAME "mbuf_cluster"
764 #define MBUF_PACKET_MEM_NAME "mbuf_packet"
765 #define MBUF_JUMBOP_MEM_NAME "mbuf_jumbo_page"
766 #define MBUF_JUMBO9_MEM_NAME "mbuf_jumbo_9k"
767 #define MBUF_JUMBO16_MEM_NAME "mbuf_jumbo_16k"
768 #define MBUF_TAG_MEM_NAME "mbuf_tag"
769 #define MBUF_EXTREFCNT_MEM_NAME "mbuf_ext_refcnt"
770 #define MBUF_EXTPGS_MEM_NAME "mbuf_extpgs"
773 union if_snd_tag_alloc_params;
776 #define MBUF_CHECKSLEEP(how) do { \
777 if (how == M_WAITOK) \
778 WITNESS_WARN(WARN_GIANTOK | WARN_SLEEPOK, NULL, \
779 "Sleeping in \"%s\"", __func__); \
782 #define MBUF_CHECKSLEEP(how) do {} while (0)
786 * Network buffer allocation API
788 * The rest of it is defined in kern/kern_mbuf.c
790 extern uma_zone_t zone_mbuf;
791 extern uma_zone_t zone_clust;
792 extern uma_zone_t zone_pack;
793 extern uma_zone_t zone_jumbop;
794 extern uma_zone_t zone_jumbo9;
795 extern uma_zone_t zone_jumbo16;
796 extern uma_zone_t zone_extpgs;
798 void mb_dupcl(struct mbuf *, struct mbuf *);
799 void mb_free_ext(struct mbuf *);
800 void mb_free_extpg(struct mbuf *);
801 void mb_free_mext_pgs(struct mbuf *);
802 struct mbuf *mb_alloc_ext_pgs(int, m_ext_free_t);
803 struct mbuf *mb_alloc_ext_plus_pages(int, int);
804 struct mbuf *mb_mapped_to_unmapped(struct mbuf *, int, int, int,
806 int mb_unmapped_compress(struct mbuf *m);
807 struct mbuf *mb_unmapped_to_ext(struct mbuf *m);
808 void mb_free_notready(struct mbuf *m, int count);
809 void m_adj(struct mbuf *, int);
810 void m_adj_decap(struct mbuf *, int);
811 int m_apply(struct mbuf *, int, int,
812 int (*)(void *, void *, u_int), void *);
813 int m_append(struct mbuf *, int, c_caddr_t);
814 void m_cat(struct mbuf *, struct mbuf *);
815 void m_catpkt(struct mbuf *, struct mbuf *);
816 int m_clget(struct mbuf *m, int how);
817 void *m_cljget(struct mbuf *m, int how, int size);
818 struct mbuf *m_collapse(struct mbuf *, int, int);
819 void m_copyback(struct mbuf *, int, int, c_caddr_t);
820 void m_copydata(const struct mbuf *, int, int, caddr_t);
821 struct mbuf *m_copym(struct mbuf *, int, int, int);
822 struct mbuf *m_copypacket(struct mbuf *, int);
823 void m_copy_pkthdr(struct mbuf *, struct mbuf *);
824 struct mbuf *m_copyup(struct mbuf *, int, int);
825 struct mbuf *m_defrag(struct mbuf *, int);
826 void m_demote_pkthdr(struct mbuf *);
827 void m_demote(struct mbuf *, int, int);
828 struct mbuf *m_devget(char *, int, int, struct ifnet *,
829 void (*)(char *, caddr_t, u_int));
830 void m_dispose_extcontrolm(struct mbuf *m);
831 struct mbuf *m_dup(const struct mbuf *, int);
832 int m_dup_pkthdr(struct mbuf *, const struct mbuf *, int);
833 void m_extadd(struct mbuf *, char *, u_int, m_ext_free_t,
834 void *, void *, int, int);
835 u_int m_fixhdr(struct mbuf *);
836 struct mbuf *m_fragment(struct mbuf *, int, int);
837 void m_freem(struct mbuf *);
838 void m_free_raw(struct mbuf *);
839 struct mbuf *m_get2(int, int, short, int);
840 struct mbuf *m_getjcl(int, short, int, int);
841 struct mbuf *m_getm2(struct mbuf *, int, int, short, int);
842 struct mbuf *m_getptr(struct mbuf *, int, int *);
843 u_int m_length(struct mbuf *, struct mbuf **);
844 int m_mbuftouio(struct uio *, const struct mbuf *, int);
845 void m_move_pkthdr(struct mbuf *, struct mbuf *);
846 int m_pkthdr_init(struct mbuf *, int);
847 struct mbuf *m_prepend(struct mbuf *, int, int);
848 void m_print(const struct mbuf *, int);
849 struct mbuf *m_pulldown(struct mbuf *, int, int, int *);
850 struct mbuf *m_pullup(struct mbuf *, int);
851 int m_sanity(struct mbuf *, int);
852 struct mbuf *m_split(struct mbuf *, int, int);
853 struct mbuf *m_uiotombuf(struct uio *, int, int, int, int);
854 int m_unmapped_uiomove(const struct mbuf *, int, struct uio *,
856 struct mbuf *m_unshare(struct mbuf *, int);
857 int m_snd_tag_alloc(struct ifnet *,
858 union if_snd_tag_alloc_params *, struct m_snd_tag **);
859 void m_snd_tag_init(struct m_snd_tag *, struct ifnet *, u_int);
860 void m_snd_tag_destroy(struct m_snd_tag *);
874 #if MJUMPAGESIZE != MCLBYTES
886 panic("%s: invalid cluster size %d", __func__, size);
893 * Associated an external reference counted buffer with an mbuf.
896 m_extaddref(struct mbuf *m, char *buf, u_int size, u_int *ref_cnt,
897 m_ext_free_t freef, void *arg1, void *arg2)
900 KASSERT(ref_cnt != NULL, ("%s: ref_cnt not provided", __func__));
902 atomic_add_int(ref_cnt, 1);
904 m->m_ext.ext_buf = buf;
905 m->m_ext.ext_cnt = ref_cnt;
906 m->m_data = m->m_ext.ext_buf;
907 m->m_ext.ext_size = size;
908 m->m_ext.ext_free = freef;
909 m->m_ext.ext_arg1 = arg1;
910 m->m_ext.ext_arg2 = arg2;
911 m->m_ext.ext_type = EXT_EXTREF;
912 m->m_ext.ext_flags = 0;
915 static __inline uma_zone_t
924 #if MJUMPAGESIZE != MCLBYTES
936 panic("%s: invalid cluster size %d", __func__, size);
943 * Initialize an mbuf with linear storage.
945 * Inline because the consumer text overhead will be roughly the same to
946 * initialize or call a function with this many parameters and M_PKTHDR
947 * should go away with constant propagation for !MGETHDR.
950 m_init(struct mbuf *m, int how, short type, int flags)
956 m->m_data = m->m_dat;
960 if (flags & M_PKTHDR)
961 error = m_pkthdr_init(m, how);
965 MBUF_PROBE5(m__init, m, how, type, flags, error);
969 static __inline struct mbuf *
970 m_get(int how, short type)
977 m = uma_zalloc_arg(zone_mbuf, &args, how);
978 MBUF_PROBE3(m__get, how, type, m);
982 static __inline struct mbuf *
983 m_gethdr(int how, short type)
988 args.flags = M_PKTHDR;
990 m = uma_zalloc_arg(zone_mbuf, &args, how);
991 MBUF_PROBE3(m__gethdr, how, type, m);
995 static __inline struct mbuf *
996 m_getcl(int how, short type, int flags)
1003 m = uma_zalloc_arg(zone_pack, &args, how);
1004 MBUF_PROBE4(m__getcl, how, type, flags, m);
1009 * XXX: m_cljset() is a dangerous API. One must attach only a new,
1010 * unreferenced cluster to an mbuf(9). It is not possible to assert
1011 * that, so care can be taken only by users of the API.
1013 static __inline void
1014 m_cljset(struct mbuf *m, void *cl, int type)
1022 #if MJUMPAGESIZE != MCLBYTES
1024 size = MJUMPAGESIZE;
1034 panic("%s: unknown cluster type %d", __func__, type);
1038 m->m_data = m->m_ext.ext_buf = cl;
1039 m->m_ext.ext_free = m->m_ext.ext_arg1 = m->m_ext.ext_arg2 = NULL;
1040 m->m_ext.ext_size = size;
1041 m->m_ext.ext_type = type;
1042 m->m_ext.ext_flags = EXT_FLAG_EMBREF;
1043 m->m_ext.ext_count = 1;
1044 m->m_flags |= M_EXT;
1045 MBUF_PROBE3(m__cljset, m, cl, type);
1048 static __inline void
1049 m_chtype(struct mbuf *m, short new_type)
1052 m->m_type = new_type;
1055 static __inline void
1056 m_clrprotoflags(struct mbuf *m)
1060 m->m_flags &= ~M_PROTOFLAGS;
1065 static __inline struct mbuf *
1066 m_last(struct mbuf *m)
1075 m_extrefcnt(struct mbuf *m)
1078 KASSERT(m->m_flags & M_EXT, ("%s: M_EXT missing", __func__));
1080 return ((m->m_ext.ext_flags & EXT_FLAG_EMBREF) ? m->m_ext.ext_count :
1085 * mbuf, cluster, and external object allocation macros (for compatibility
1088 #define M_MOVE_PKTHDR(to, from) m_move_pkthdr((to), (from))
1089 #define MGET(m, how, type) ((m) = m_get((how), (type)))
1090 #define MGETHDR(m, how, type) ((m) = m_gethdr((how), (type)))
1091 #define MCLGET(m, how) m_clget((m), (how))
1092 #define MEXTADD(m, buf, size, free, arg1, arg2, flags, type) \
1093 m_extadd((m), (char *)(buf), (size), (free), (arg1), (arg2), \
1095 #define m_getm(m, len, how, type) \
1096 m_getm2((m), (len), (how), (type), M_PKTHDR)
1099 * Evaluate TRUE if it's safe to write to the mbuf m's data region (this can
1100 * be both the local data payload, or an external buffer area, depending on
1101 * whether M_EXT is set).
1103 #define M_WRITABLE(m) (((m)->m_flags & (M_RDONLY | M_EXTPG)) == 0 && \
1104 (!(((m)->m_flags & M_EXT)) || \
1105 (m_extrefcnt(m) == 1)))
1107 /* Check if the supplied mbuf has a packet header, or else panic. */
1108 #define M_ASSERTPKTHDR(m) \
1109 KASSERT((m) != NULL && (m)->m_flags & M_PKTHDR, \
1110 ("%s: no mbuf packet header!", __func__))
1112 /* Check if the supplied mbuf has no send tag, or else panic. */
1113 #define M_ASSERT_NO_SND_TAG(m) \
1114 KASSERT((m) != NULL && (m)->m_flags & M_PKTHDR && \
1115 ((m)->m_pkthdr.csum_flags & CSUM_SND_TAG) == 0, \
1116 ("%s: receive mbuf has send tag!", __func__))
1118 /* Check if mbuf is multipage. */
1119 #define M_ASSERTEXTPG(m) \
1120 KASSERT(((m)->m_flags & (M_EXTPG|M_PKTHDR)) == M_EXTPG, \
1121 ("%s: m %p is not multipage!", __func__, m))
1124 * Ensure that the supplied mbuf is a valid, non-free mbuf.
1126 * XXX: Broken at the moment. Need some UMA magic to make it work again.
1128 #define M_ASSERTVALID(m) \
1129 KASSERT((((struct mbuf *)m)->m_flags & 0) == 0, \
1130 ("%s: attempted use of a free mbuf!", __func__))
1132 /* Check whether any mbuf in the chain is unmapped. */
1134 #define M_ASSERTMAPPED(m) do { \
1135 for (struct mbuf *__m = (m); __m != NULL; __m = __m->m_next) \
1136 KASSERT((__m->m_flags & M_EXTPG) == 0, \
1137 ("%s: chain %p contains an unmapped mbuf", __func__, (m)));\
1140 #define M_ASSERTMAPPED(m) do {} while (0)
1144 * Return the address of the start of the buffer associated with an mbuf,
1145 * handling external storage, packet-header mbufs, and regular data mbufs.
1147 #define M_START(m) \
1148 (((m)->m_flags & M_EXTPG) ? NULL : \
1149 ((m)->m_flags & M_EXT) ? (m)->m_ext.ext_buf : \
1150 ((m)->m_flags & M_PKTHDR) ? &(m)->m_pktdat[0] : \
1154 * Return the size of the buffer associated with an mbuf, handling external
1155 * storage, packet-header mbufs, and regular data mbufs.
1158 (((m)->m_flags & M_EXT) ? (m)->m_ext.ext_size : \
1159 ((m)->m_flags & M_PKTHDR) ? MHLEN : \
1163 * Set the m_data pointer of a newly allocated mbuf to place an object of the
1164 * specified size at the end of the mbuf, longword aligned.
1166 * NB: Historically, we had M_ALIGN(), MH_ALIGN(), and MEXT_ALIGN() as
1167 * separate macros, each asserting that it was called at the proper moment.
1168 * This required callers to themselves test the storage type and call the
1169 * right one. Rather than require callers to be aware of those layout
1170 * decisions, we centralize here.
1172 static __inline void
1173 m_align(struct mbuf *m, int len)
1176 const char *msg = "%s: not a virgin mbuf";
1180 KASSERT(m->m_data == M_START(m), (msg, __func__));
1182 adjust = M_SIZE(m) - len;
1183 m->m_data += adjust &~ (sizeof(long)-1);
1186 #define M_ALIGN(m, len) m_align(m, len)
1187 #define MH_ALIGN(m, len) m_align(m, len)
1188 #define MEXT_ALIGN(m, len) m_align(m, len)
1191 * Compute the amount of space available before the current start of data in
1194 * The M_WRITABLE() is a temporary, conservative safety measure: the burden
1195 * of checking writability of the mbuf data area rests solely with the caller.
1197 * NB: In previous versions, M_LEADINGSPACE() would only check M_WRITABLE()
1198 * for mbufs with external storage. We now allow mbuf-embedded data to be
1199 * read-only as well.
1201 #define M_LEADINGSPACE(m) \
1202 (M_WRITABLE(m) ? ((m)->m_data - M_START(m)) : 0)
1205 * Compute the amount of space available after the end of data in an mbuf.
1207 * The M_WRITABLE() is a temporary, conservative safety measure: the burden
1208 * of checking writability of the mbuf data area rests solely with the caller.
1210 * NB: In previous versions, M_TRAILINGSPACE() would only check M_WRITABLE()
1211 * for mbufs with external storage. We now allow mbuf-embedded data to be
1212 * read-only as well.
1214 #define M_TRAILINGSPACE(m) \
1216 ((M_START(m) + M_SIZE(m)) - ((m)->m_data + (m)->m_len)) : 0)
1219 * Arrange to prepend space of size plen to mbuf m. If a new mbuf must be
1220 * allocated, how specifies whether to wait. If the allocation fails, the
1221 * original mbuf chain is freed and m is set to NULL.
1223 #define M_PREPEND(m, plen, how) do { \
1224 struct mbuf **_mmp = &(m); \
1225 struct mbuf *_mm = *_mmp; \
1226 int _mplen = (plen); \
1227 int __mhow = (how); \
1229 MBUF_CHECKSLEEP(how); \
1230 if (M_LEADINGSPACE(_mm) >= _mplen) { \
1231 _mm->m_data -= _mplen; \
1232 _mm->m_len += _mplen; \
1234 _mm = m_prepend(_mm, _mplen, __mhow); \
1235 if (_mm != NULL && _mm->m_flags & M_PKTHDR) \
1236 _mm->m_pkthdr.len += _mplen; \
1241 * Change mbuf to new type. This is a relatively expensive operation and
1242 * should be avoided.
1244 #define MCHTYPE(m, t) m_chtype((m), (t))
1246 /* Return the rcvif of a packet header. */
1247 static __inline struct ifnet *
1248 m_rcvif(struct mbuf *m)
1252 if (m->m_pkthdr.csum_flags & CSUM_SND_TAG)
1254 return (m->m_pkthdr.rcvif);
1257 /* Length to m_copy to copy all. */
1258 #define M_COPYALL 1000000000
1260 extern int max_datalen; /* MHLEN - max_hdr */
1261 extern int max_hdr; /* Largest link + protocol header */
1262 extern int max_linkhdr; /* Largest link-level header */
1263 extern int max_protohdr; /* Largest protocol header */
1264 extern int nmbclusters; /* Maximum number of clusters */
1265 extern bool mb_use_ext_pgs; /* Use ext_pgs for sendfile */
1268 * Network packets may have annotations attached by affixing a list of
1269 * "packet tags" to the pkthdr structure. Packet tags are dynamically
1270 * allocated semi-opaque data structures that have a fixed header
1271 * (struct m_tag) that specifies the size of the memory block and a
1272 * <cookie,type> pair that identifies it. The cookie is a 32-bit unique
1273 * unsigned value used to identify a module or ABI. By convention this value
1274 * is chosen as the date+time that the module is created, expressed as the
1275 * number of seconds since the epoch (e.g., using date -u +'%s'). The type
1276 * value is an ABI/module-specific value that identifies a particular
1277 * annotation and is private to the module. For compatibility with systems
1278 * like OpenBSD that define packet tags w/o an ABI/module cookie, the value
1279 * PACKET_ABI_COMPAT is used to implement m_tag_get and m_tag_find
1280 * compatibility shim functions and several tag types are defined below.
1281 * Users that do not require compatibility should use a private cookie value
1282 * so that packet tag-related definitions can be maintained privately.
1284 * Note that the packet tag returned by m_tag_alloc has the default memory
1285 * alignment implemented by malloc. To reference private data one can use a
1288 * struct m_tag *mtag = m_tag_alloc(...);
1289 * struct foo *p = (struct foo *)(mtag+1);
1291 * if the alignment of struct m_tag is sufficient for referencing members of
1292 * struct foo. Otherwise it is necessary to embed struct m_tag within the
1293 * private data structure to insure proper alignment; e.g.,
1299 * struct foo *p = (struct foo *) m_tag_alloc(...);
1300 * struct m_tag *mtag = &p->tag;
1304 * Persistent tags stay with an mbuf until the mbuf is reclaimed. Otherwise
1305 * tags are expected to ``vanish'' when they pass through a network
1306 * interface. For most interfaces this happens normally as the tags are
1307 * reclaimed when the mbuf is free'd. However in some special cases
1308 * reclaiming must be done manually. An example is packets that pass through
1309 * the loopback interface. Also, one must be careful to do this when
1310 * ``turning around'' packets (e.g., icmp_reflect).
1312 * To mark a tag persistent bit-or this flag in when defining the tag id.
1313 * The tag will then be treated as described above.
1315 #define MTAG_PERSISTENT 0x800
1317 #define PACKET_TAG_NONE 0 /* Nadda */
1319 /* Packet tags for use with PACKET_ABI_COMPAT. */
1320 #define PACKET_TAG_IPSEC_IN_DONE 1 /* IPsec applied, in */
1321 #define PACKET_TAG_IPSEC_OUT_DONE 2 /* IPsec applied, out */
1322 #define PACKET_TAG_IPSEC_IN_CRYPTO_DONE 3 /* NIC IPsec crypto done */
1323 #define PACKET_TAG_IPSEC_OUT_CRYPTO_NEEDED 4 /* NIC IPsec crypto req'ed */
1324 #define PACKET_TAG_IPSEC_IN_COULD_DO_CRYPTO 5 /* NIC notifies IPsec */
1325 #define PACKET_TAG_IPSEC_PENDING_TDB 6 /* Reminder to do IPsec */
1326 #define PACKET_TAG_BRIDGE 7 /* Bridge processing done */
1327 #define PACKET_TAG_GIF 8 /* GIF processing done */
1328 #define PACKET_TAG_GRE 9 /* GRE processing done */
1329 #define PACKET_TAG_IN_PACKET_CHECKSUM 10 /* NIC checksumming done */
1330 #define PACKET_TAG_ENCAP 11 /* Encap. processing */
1331 #define PACKET_TAG_IPSEC_SOCKET 12 /* IPSEC socket ref */
1332 #define PACKET_TAG_IPSEC_HISTORY 13 /* IPSEC history */
1333 #define PACKET_TAG_IPV6_INPUT 14 /* IPV6 input processing */
1334 #define PACKET_TAG_DUMMYNET 15 /* dummynet info */
1335 #define PACKET_TAG_DIVERT 17 /* divert info */
1336 #define PACKET_TAG_IPFORWARD 18 /* ipforward info */
1337 #define PACKET_TAG_MACLABEL (19 | MTAG_PERSISTENT) /* MAC label */
1338 #define PACKET_TAG_PF 21 /* PF/ALTQ information */
1339 #define PACKET_TAG_RTSOCKFAM 25 /* rtsock sa family */
1340 #define PACKET_TAG_IPOPTIONS 27 /* Saved IP options */
1341 #define PACKET_TAG_CARP 28 /* CARP info */
1342 #define PACKET_TAG_IPSEC_NAT_T_PORTS 29 /* two uint16_t */
1343 #define PACKET_TAG_ND_OUTGOING 30 /* ND outgoing */
1345 /* Specific cookies and tags. */
1347 /* Packet tag routines. */
1348 struct m_tag *m_tag_alloc(u_int32_t, int, int, int);
1349 void m_tag_delete(struct mbuf *, struct m_tag *);
1350 void m_tag_delete_chain(struct mbuf *, struct m_tag *);
1351 void m_tag_free_default(struct m_tag *);
1352 struct m_tag *m_tag_locate(struct mbuf *, u_int32_t, int, struct m_tag *);
1353 struct m_tag *m_tag_copy(struct m_tag *, int);
1354 int m_tag_copy_chain(struct mbuf *, const struct mbuf *, int);
1355 void m_tag_delete_nonpersistent(struct mbuf *);
1358 * Initialize the list of tags associated with an mbuf.
1360 static __inline void
1361 m_tag_init(struct mbuf *m)
1364 SLIST_INIT(&m->m_pkthdr.tags);
1368 * Set up the contents of a tag. Note that this does not fill in the free
1369 * method; the caller is expected to do that.
1371 * XXX probably should be called m_tag_init, but that was already taken.
1373 static __inline void
1374 m_tag_setup(struct m_tag *t, u_int32_t cookie, int type, int len)
1379 t->m_tag_cookie = cookie;
1383 * Reclaim resources associated with a tag.
1385 static __inline void
1386 m_tag_free(struct m_tag *t)
1389 (*t->m_tag_free)(t);
1393 * Return the first tag associated with an mbuf.
1395 static __inline struct m_tag *
1396 m_tag_first(struct mbuf *m)
1399 return (SLIST_FIRST(&m->m_pkthdr.tags));
1403 * Return the next tag in the list of tags associated with an mbuf.
1405 static __inline struct m_tag *
1406 m_tag_next(struct mbuf *m __unused, struct m_tag *t)
1409 return (SLIST_NEXT(t, m_tag_link));
1413 * Prepend a tag to the list of tags associated with an mbuf.
1415 static __inline void
1416 m_tag_prepend(struct mbuf *m, struct m_tag *t)
1419 SLIST_INSERT_HEAD(&m->m_pkthdr.tags, t, m_tag_link);
1423 * Unlink a tag from the list of tags associated with an mbuf.
1425 static __inline void
1426 m_tag_unlink(struct mbuf *m, struct m_tag *t)
1429 SLIST_REMOVE(&m->m_pkthdr.tags, t, m_tag, m_tag_link);
1432 /* These are for OpenBSD compatibility. */
1433 #define MTAG_ABI_COMPAT 0 /* compatibility ABI */
1435 static __inline struct m_tag *
1436 m_tag_get(int type, int length, int wait)
1438 return (m_tag_alloc(MTAG_ABI_COMPAT, type, length, wait));
1441 static __inline struct m_tag *
1442 m_tag_find(struct mbuf *m, int type, struct m_tag *start)
1444 return (SLIST_EMPTY(&m->m_pkthdr.tags) ? (struct m_tag *)NULL :
1445 m_tag_locate(m, MTAG_ABI_COMPAT, type, start));
1448 static inline struct m_snd_tag *
1449 m_snd_tag_ref(struct m_snd_tag *mst)
1452 refcount_acquire(&mst->refcount);
1457 m_snd_tag_rele(struct m_snd_tag *mst)
1460 if (refcount_release(&mst->refcount))
1461 m_snd_tag_destroy(mst);
1464 static __inline struct mbuf *
1465 m_free(struct mbuf *m)
1467 struct mbuf *n = m->m_next;
1469 MBUF_PROBE1(m__free, m);
1470 if ((m->m_flags & (M_PKTHDR|M_NOFREE)) == (M_PKTHDR|M_NOFREE))
1471 m_tag_delete_chain(m, NULL);
1472 if (m->m_flags & M_PKTHDR && m->m_pkthdr.csum_flags & CSUM_SND_TAG)
1473 m_snd_tag_rele(m->m_pkthdr.snd_tag);
1474 if (m->m_flags & M_EXTPG)
1476 else if (m->m_flags & M_EXT)
1478 else if ((m->m_flags & M_NOFREE) == 0)
1479 uma_zfree(zone_mbuf, m);
1484 rt_m_getfib(struct mbuf *m)
1486 KASSERT(m->m_flags & M_PKTHDR , ("Attempt to get FIB from non header mbuf."));
1487 return (m->m_pkthdr.fibnum);
1490 #define M_GETFIB(_m) rt_m_getfib(_m)
1492 #define M_SETFIB(_m, _fib) do { \
1493 KASSERT((_m)->m_flags & M_PKTHDR, ("Attempt to set FIB on non header mbuf.")); \
1494 ((_m)->m_pkthdr.fibnum) = (_fib); \
1497 /* flags passed as first argument for "m_xxx_tcpip_hash()" */
1498 #define MBUF_HASHFLAG_L2 (1 << 2)
1499 #define MBUF_HASHFLAG_L3 (1 << 3)
1500 #define MBUF_HASHFLAG_L4 (1 << 4)
1502 /* mbuf hashing helper routines */
1503 uint32_t m_ether_tcpip_hash_init(void);
1504 uint32_t m_ether_tcpip_hash(const uint32_t, const struct mbuf *, uint32_t);
1505 uint32_t m_infiniband_tcpip_hash_init(void);
1506 uint32_t m_infiniband_tcpip_hash(const uint32_t, const struct mbuf *, uint32_t);
1508 #ifdef MBUF_PROFILING
1509 void m_profile(struct mbuf *m);
1510 #define M_PROFILE(m) m_profile(m)
1512 #define M_PROFILE(m)
1516 STAILQ_HEAD(, mbuf) mq_head;
1522 mbufq_init(struct mbufq *mq, int maxlen)
1525 STAILQ_INIT(&mq->mq_head);
1526 mq->mq_maxlen = maxlen;
1530 static inline struct mbuf *
1531 mbufq_flush(struct mbufq *mq)
1535 m = STAILQ_FIRST(&mq->mq_head);
1536 STAILQ_INIT(&mq->mq_head);
1542 mbufq_drain(struct mbufq *mq)
1546 n = mbufq_flush(mq);
1547 while ((m = n) != NULL) {
1548 n = STAILQ_NEXT(m, m_stailqpkt);
1553 static inline struct mbuf *
1554 mbufq_first(const struct mbufq *mq)
1557 return (STAILQ_FIRST(&mq->mq_head));
1560 static inline struct mbuf *
1561 mbufq_last(const struct mbufq *mq)
1564 return (STAILQ_LAST(&mq->mq_head, mbuf, m_stailqpkt));
1568 mbufq_full(const struct mbufq *mq)
1571 return (mq->mq_maxlen > 0 && mq->mq_len >= mq->mq_maxlen);
1575 mbufq_len(const struct mbufq *mq)
1578 return (mq->mq_len);
1582 mbufq_enqueue(struct mbufq *mq, struct mbuf *m)
1587 STAILQ_INSERT_TAIL(&mq->mq_head, m, m_stailqpkt);
1592 static inline struct mbuf *
1593 mbufq_dequeue(struct mbufq *mq)
1597 m = STAILQ_FIRST(&mq->mq_head);
1599 STAILQ_REMOVE_HEAD(&mq->mq_head, m_stailqpkt);
1600 m->m_nextpkt = NULL;
1607 mbufq_prepend(struct mbufq *mq, struct mbuf *m)
1610 STAILQ_INSERT_HEAD(&mq->mq_head, m, m_stailqpkt);
1615 * Note: this doesn't enforce the maximum list size for dst.
1618 mbufq_concat(struct mbufq *mq_dst, struct mbufq *mq_src)
1621 mq_dst->mq_len += mq_src->mq_len;
1622 STAILQ_CONCAT(&mq_dst->mq_head, &mq_src->mq_head);
1626 #ifdef _SYS_TIMESPEC_H_
1628 mbuf_tstmp2timespec(struct mbuf *m, struct timespec *ts)
1631 KASSERT((m->m_flags & M_PKTHDR) != 0, ("mbuf %p no M_PKTHDR", m));
1632 KASSERT((m->m_flags & (M_TSTMP|M_TSTMP_LRO)) != 0,
1633 ("mbuf %p no M_TSTMP or M_TSTMP_LRO", m));
1634 ts->tv_sec = m->m_pkthdr.rcv_tstmp / 1000000000;
1635 ts->tv_nsec = m->m_pkthdr.rcv_tstmp % 1000000000;
1640 mbuf_tstmp2timeval(struct mbuf *m, struct timeval *tv)
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,
1645 ("mbuf %p no M_TSTMP or M_TSTMP_LRO", m));
1646 tv->tv_sec = m->m_pkthdr.rcv_tstmp / 1000000000;
1647 tv->tv_usec = (m->m_pkthdr.rcv_tstmp % 1000000000) / 1000;
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_ */