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]
202 #define vt_nrecs sixteen[0] /* mld and v6-ND */
203 #define tso_segsz PH_per.sixteen[1] /* inbound after LRO */
204 #define lro_nsegs tso_segsz /* inbound after LRO */
205 #define csum_data PH_per.thirtytwo[1] /* inbound from hardware up */
206 #define lro_len PH_loc.sixteen[0] /* inbound during LRO (no reassembly) */
207 #define lro_csum PH_loc.sixteen[1] /* inbound during LRO (no reassembly) */
208 /* Note PH_loc is used during IP reassembly (all 8 bytes as a ptr) */
211 * TLS records for TLS 1.0-1.2 can have the following header lengths:
212 * - 5 (AES-CBC with implicit IV)
213 * - 21 (AES-CBC with explicit IV)
214 * - 13 (AES-GCM with 8 byte explicit IV)
216 #define MBUF_PEXT_HDR_LEN 23
219 * TLS records for TLS 1.0-1.2 can have the following maximum trailer
222 * - 36 (AES-CBC with SHA1 and up to 16 bytes of padding)
223 * - 48 (AES-CBC with SHA2-256 and up to 16 bytes of padding)
224 * - 64 (AES-CBC with SHA2-384 and up to 16 bytes of padding)
226 #define MBUF_PEXT_TRAIL_LEN 64
228 #if defined(__LP64__)
229 #define MBUF_PEXT_MAX_PGS (40 / sizeof(vm_paddr_t))
231 #define MBUF_PEXT_MAX_PGS (72 / sizeof(vm_paddr_t))
234 #define MBUF_PEXT_MAX_BYTES \
235 (MBUF_PEXT_MAX_PGS * PAGE_SIZE + MBUF_PEXT_HDR_LEN + MBUF_PEXT_TRAIL_LEN)
241 * Description of external storage mapped into mbuf; valid only if M_EXT is
245 * Compile-time assertions in uipc_mbuf.c test these values to ensure that
248 typedef void m_ext_free_t(struct mbuf *);
252 * If EXT_FLAG_EMBREF is set, then we use refcount in the
253 * mbuf, the 'ext_count' member. Otherwise, we have a
254 * shadow copy and we use pointer 'ext_cnt'. The original
255 * mbuf is responsible to carry the pointer to free routine
256 * and its arguments. They aren't copied into shadows in
257 * mb_dupcl() to avoid dereferencing next cachelines.
259 volatile u_int ext_count;
260 volatile u_int *ext_cnt;
262 uint32_t ext_size; /* size of buffer, for ext_free */
263 uint32_t ext_type:8, /* type of external storage */
264 ext_flags:24; /* external storage mbuf flags */
268 * Regular M_EXT mbuf:
269 * o ext_buf always points to the external buffer.
270 * o ext_free (below) and two optional arguments
271 * ext_arg1 and ext_arg2 store the free context for
272 * the external storage. They are set only in the
273 * refcount carrying mbuf, the one with
274 * EXT_FLAG_EMBREF flag, with exclusion for
275 * EXT_EXTREF type, where the free context is copied
276 * into all mbufs that use same external storage.
278 char *ext_buf; /* start of buffer */
279 #define m_ext_copylen offsetof(struct m_ext, ext_arg2)
284 * Multi-page M_EXTPG mbuf:
285 * o extpg_pa - page vector.
286 * o extpg_trail and extpg_hdr - TLS trailer and
288 * Uses ext_free and may also use ext_arg1.
290 vm_paddr_t extpg_pa[MBUF_PEXT_MAX_PGS];
291 char extpg_trail[MBUF_PEXT_TRAIL_LEN];
292 char extpg_hdr[MBUF_PEXT_HDR_LEN];
293 /* Pretend these 3 fields are part of mbuf itself. */
294 #define m_epg_pa m_ext.extpg_pa
295 #define m_epg_trail m_ext.extpg_trail
296 #define m_epg_hdr m_ext.extpg_hdr
297 #define m_epg_ext_copylen offsetof(struct m_ext, ext_free)
301 * Free method and optional argument pointer, both
302 * used by M_EXT and M_EXTPG.
304 m_ext_free_t *ext_free;
309 * The core of the mbuf object along with some shortcut defines for practical
314 * Header present at the beginning of every mbuf.
317 * Compile-time assertions in uipc_mbuf.c test these values to ensure
318 * that they are correct.
320 union { /* next buffer in chain */
322 SLIST_ENTRY(mbuf) m_slist;
323 STAILQ_ENTRY(mbuf) m_stailq;
325 union { /* next chain in queue/record */
326 struct mbuf *m_nextpkt;
327 SLIST_ENTRY(mbuf) m_slistpkt;
328 STAILQ_ENTRY(mbuf) m_stailqpkt;
330 caddr_t m_data; /* location of data */
331 int32_t m_len; /* amount of data in this mbuf */
332 uint32_t m_type:8, /* type of data in this mbuf */
333 m_flags:24; /* flags; see below */
334 #if !defined(__LP64__)
335 uint32_t m_pad; /* pad for 64bit alignment */
339 * A set of optional headers (packet header, external storage header)
340 * and internal data storage. Historically, these arrays were sized
341 * to MHLEN (space left after a packet header) and MLEN (space left
342 * after only a regular mbuf header); they are now variable size in
343 * order to support future work on variable-size mbufs.
349 struct pkthdr m_pkthdr;
352 * Multi-page M_EXTPG mbuf has its meta data
353 * split between the below anonymous structure
354 * and m_ext. It carries vector of pages,
355 * optional header and trailer char vectors
356 * and pointers to socket/TLS data.
358 #define m_epg_startcopy m_epg_npgs
359 #define m_epg_endcopy m_epg_stailq
361 /* Overall count of pages and count of
362 * pages with I/O pending. */
365 /* TLS header and trailer lengths.
366 * The data itself resides in m_ext. */
367 uint8_t m_epg_hdrlen;
368 uint8_t m_epg_trllen;
369 /* Offset into 1st page and length of
370 * data in the last page. */
371 uint16_t m_epg_1st_off;
372 uint16_t m_epg_last_len;
374 #define EPG_FLAG_ANON 0x1 /* Data can be encrypted in place. */
375 #define EPG_FLAG_2FREE 0x2 /* Scheduled for free. */
376 uint8_t m_epg_record_type;
379 struct ktls_session *m_epg_tls;
380 struct socket *m_epg_so;
381 uint64_t m_epg_seqno;
382 STAILQ_ENTRY(mbuf) m_epg_stailq;
386 /* M_EXT or M_EXTPG set. */
388 /* M_PKTHDR set, neither M_EXT nor M_EXTPG. */
392 char m_dat[0]; /* !M_PKTHDR, !M_EXT */
398 m_epg_pagelen(const struct mbuf *m, int pidx, int pgoff)
401 KASSERT(pgoff == 0 || pidx == 0,
402 ("page %d with non-zero offset %d in %p", pidx, pgoff, m));
404 if (pidx == m->m_epg_npgs - 1) {
405 return (m->m_epg_last_len);
407 return (PAGE_SIZE - pgoff);
412 #define MCHECK(ex, msg) KASSERT((ex), \
413 ("Multi page mbuf %p with " #msg " at %s:%d", \
414 m, __FILE__, __LINE__))
416 * NB: This expects a non-empty buffer (npgs > 0 and
419 #define MBUF_EXT_PGS_ASSERT_SANITY(m) do { \
420 MCHECK(m->m_epg_npgs > 0, "no valid pages"); \
421 MCHECK(m->m_epg_npgs <= nitems(m->m_epg_pa), \
423 MCHECK(m->m_epg_nrdy <= m->m_epg_npgs, \
424 "too many ready pages"); \
425 MCHECK(m->m_epg_1st_off < PAGE_SIZE, \
426 "too large page offset"); \
427 MCHECK(m->m_epg_last_len > 0, "zero last page length"); \
428 MCHECK(m->m_epg_last_len <= PAGE_SIZE, \
429 "too large last page length"); \
430 if (m->m_epg_npgs == 1) \
431 MCHECK(m->m_epg_1st_off + \
432 m->m_epg_last_len <= PAGE_SIZE, \
433 "single page too large"); \
434 MCHECK(m->m_epg_hdrlen <= sizeof(m->m_epg_hdr), \
435 "too large header length"); \
436 MCHECK(m->m_epg_trllen <= sizeof(m->m_epg_trail), \
437 "too large header length"); \
440 #define MBUF_EXT_PGS_ASSERT_SANITY(m) do {} while (0);
445 * mbuf flags of global significance and layer crossing.
446 * Those of only protocol/layer specific significance are to be mapped
447 * to M_PROTO[1-11] and cleared at layer handoff boundaries.
448 * NB: Limited to the lower 24 bits.
450 #define M_EXT 0x00000001 /* has associated external storage */
451 #define M_PKTHDR 0x00000002 /* start of record */
452 #define M_EOR 0x00000004 /* end of record */
453 #define M_RDONLY 0x00000008 /* associated data is marked read-only */
454 #define M_BCAST 0x00000010 /* send/received as link-level broadcast */
455 #define M_MCAST 0x00000020 /* send/received as link-level multicast */
456 #define M_PROMISC 0x00000040 /* packet was not for us */
457 #define M_VLANTAG 0x00000080 /* ether_vtag is valid */
458 #define M_EXTPG 0x00000100 /* has array of unmapped pages and TLS */
459 #define M_NOFREE 0x00000200 /* do not free mbuf, embedded in cluster */
460 #define M_TSTMP 0x00000400 /* rcv_tstmp field is valid */
461 #define M_TSTMP_HPREC 0x00000800 /* rcv_tstmp is high-prec, typically
462 hw-stamped on port (useful for IEEE 1588
464 #define M_TSTMP_LRO 0x00001000 /* Time LRO pushed in pkt is valid in (PH_loc) */
466 #define M_PROTO1 0x00002000 /* protocol-specific */
467 #define M_PROTO2 0x00004000 /* protocol-specific */
468 #define M_PROTO3 0x00008000 /* protocol-specific */
469 #define M_PROTO4 0x00010000 /* protocol-specific */
470 #define M_PROTO5 0x00020000 /* protocol-specific */
471 #define M_PROTO6 0x00040000 /* protocol-specific */
472 #define M_PROTO7 0x00080000 /* protocol-specific */
473 #define M_PROTO8 0x00100000 /* protocol-specific */
474 #define M_PROTO9 0x00200000 /* protocol-specific */
475 #define M_PROTO10 0x00400000 /* protocol-specific */
476 #define M_PROTO11 0x00800000 /* protocol-specific */
478 #define MB_DTOR_SKIP 0x1 /* don't pollute the cache by touching a freed mbuf */
481 * Flags to purge when crossing layers.
483 #define M_PROTOFLAGS \
484 (M_PROTO1|M_PROTO2|M_PROTO3|M_PROTO4|M_PROTO5|M_PROTO6|M_PROTO7|M_PROTO8|\
485 M_PROTO9|M_PROTO10|M_PROTO11)
488 * Flags preserved when copying m_pkthdr.
490 #define M_COPYFLAGS \
491 (M_PKTHDR|M_EOR|M_RDONLY|M_BCAST|M_MCAST|M_PROMISC|M_VLANTAG|M_TSTMP| \
492 M_TSTMP_HPREC|M_TSTMP_LRO|M_PROTOFLAGS)
495 * Mbuf flag description for use with printf(9) %b identifier.
497 #define M_FLAG_BITS \
498 "\20\1M_EXT\2M_PKTHDR\3M_EOR\4M_RDONLY\5M_BCAST\6M_MCAST" \
499 "\7M_PROMISC\10M_VLANTAG\11M_EXTPG\12M_NOFREE\13M_TSTMP\14M_TSTMP_HPREC\15M_TSTMP_LRO"
500 #define M_FLAG_PROTOBITS \
501 "\16M_PROTO1\17M_PROTO2\20M_PROTO3\21M_PROTO4" \
502 "\22M_PROTO5\23M_PROTO6\24M_PROTO7\25M_PROTO8\26M_PROTO9" \
503 "\27M_PROTO10\28M_PROTO11"
504 #define M_FLAG_PRINTF (M_FLAG_BITS M_FLAG_PROTOBITS)
507 * Network interface cards are able to hash protocol fields (such as IPv4
508 * addresses and TCP port numbers) classify packets into flows. These flows
509 * can then be used to maintain ordering while delivering packets to the OS
510 * via parallel input queues, as well as to provide a stateless affinity
511 * model. NIC drivers can pass up the hash via m->m_pkthdr.flowid, and set
512 * m_flag fields to indicate how the hash should be interpreted by the
515 * Most NICs support RSS, which provides ordering and explicit affinity, and
516 * use the hash m_flag bits to indicate what header fields were covered by
517 * the hash. M_HASHTYPE_OPAQUE and M_HASHTYPE_OPAQUE_HASH can be set by non-
518 * RSS cards or configurations that provide an opaque flow identifier, allowing
519 * for ordering and distribution without explicit affinity. Additionally,
520 * M_HASHTYPE_OPAQUE_HASH indicates that the flow identifier has hash
523 * The meaning of the IPV6_EX suffix:
524 * "o Home address from the home address option in the IPv6 destination
525 * options header. If the extension header is not present, use the Source
527 * o IPv6 address that is contained in the Routing-Header-Type-2 from the
528 * associated extension header. If the extension header is not present,
529 * use the Destination IPv6 Address."
531 * https://docs.microsoft.com/en-us/windows-hardware/drivers/network/rss-hashing-types#ndishashipv6ex
533 #define M_HASHTYPE_HASHPROP 0x80 /* has hash properties */
534 #define M_HASHTYPE_HASH(t) (M_HASHTYPE_HASHPROP | (t))
535 /* Microsoft RSS standard hash types */
536 #define M_HASHTYPE_NONE 0
537 #define M_HASHTYPE_RSS_IPV4 M_HASHTYPE_HASH(1) /* IPv4 2-tuple */
538 #define M_HASHTYPE_RSS_TCP_IPV4 M_HASHTYPE_HASH(2) /* TCPv4 4-tuple */
539 #define M_HASHTYPE_RSS_IPV6 M_HASHTYPE_HASH(3) /* IPv6 2-tuple */
540 #define M_HASHTYPE_RSS_TCP_IPV6 M_HASHTYPE_HASH(4) /* TCPv6 4-tuple */
541 #define M_HASHTYPE_RSS_IPV6_EX M_HASHTYPE_HASH(5) /* IPv6 2-tuple +
543 #define M_HASHTYPE_RSS_TCP_IPV6_EX M_HASHTYPE_HASH(6) /* TCPv6 4-tuple +
545 #define M_HASHTYPE_RSS_UDP_IPV4 M_HASHTYPE_HASH(7) /* IPv4 UDP 4-tuple*/
546 #define M_HASHTYPE_RSS_UDP_IPV6 M_HASHTYPE_HASH(9) /* IPv6 UDP 4-tuple*/
547 #define M_HASHTYPE_RSS_UDP_IPV6_EX M_HASHTYPE_HASH(10)/* IPv6 UDP 4-tuple +
550 #define M_HASHTYPE_OPAQUE 63 /* ordering, not affinity */
551 #define M_HASHTYPE_OPAQUE_HASH M_HASHTYPE_HASH(M_HASHTYPE_OPAQUE)
552 /* ordering+hash, not affinity*/
554 #define M_HASHTYPE_CLEAR(m) ((m)->m_pkthdr.rsstype = 0)
555 #define M_HASHTYPE_GET(m) ((m)->m_pkthdr.rsstype)
556 #define M_HASHTYPE_SET(m, v) ((m)->m_pkthdr.rsstype = (v))
557 #define M_HASHTYPE_TEST(m, v) (M_HASHTYPE_GET(m) == (v))
558 #define M_HASHTYPE_ISHASH(m) (M_HASHTYPE_GET(m) & M_HASHTYPE_HASHPROP)
561 * External mbuf storage buffer types.
563 #define EXT_CLUSTER 1 /* mbuf cluster */
564 #define EXT_SFBUF 2 /* sendfile(2)'s sf_buf */
565 #define EXT_JUMBOP 3 /* jumbo cluster page sized */
566 #define EXT_JUMBO9 4 /* jumbo cluster 9216 bytes */
567 #define EXT_JUMBO16 5 /* jumbo cluster 16184 bytes */
568 #define EXT_PACKET 6 /* mbuf+cluster from packet zone */
569 #define EXT_MBUF 7 /* external mbuf reference */
570 #define EXT_RXRING 8 /* data in NIC receive ring */
572 #define EXT_VENDOR1 224 /* for vendor-internal use */
573 #define EXT_VENDOR2 225 /* for vendor-internal use */
574 #define EXT_VENDOR3 226 /* for vendor-internal use */
575 #define EXT_VENDOR4 227 /* for vendor-internal use */
577 #define EXT_EXP1 244 /* for experimental use */
578 #define EXT_EXP2 245 /* for experimental use */
579 #define EXT_EXP3 246 /* for experimental use */
580 #define EXT_EXP4 247 /* for experimental use */
582 #define EXT_NET_DRV 252 /* custom ext_buf provided by net driver(s) */
583 #define EXT_MOD_TYPE 253 /* custom module's ext_buf type */
584 #define EXT_DISPOSABLE 254 /* can throw this buffer away w/page flipping */
585 #define EXT_EXTREF 255 /* has externally maintained ext_cnt ptr */
588 * Flags for external mbuf buffer types.
589 * NB: limited to the lower 24 bits.
591 #define EXT_FLAG_EMBREF 0x000001 /* embedded ext_count */
592 #define EXT_FLAG_EXTREF 0x000002 /* external ext_cnt, notyet */
594 #define EXT_FLAG_NOFREE 0x000010 /* don't free mbuf to pool, notyet */
596 #define EXT_FLAG_VENDOR1 0x010000 /* These flags are vendor */
597 #define EXT_FLAG_VENDOR2 0x020000 /* or submodule specific, */
598 #define EXT_FLAG_VENDOR3 0x040000 /* not used by mbuf code. */
599 #define EXT_FLAG_VENDOR4 0x080000 /* Set/read by submodule. */
601 #define EXT_FLAG_EXP1 0x100000 /* for experimental use */
602 #define EXT_FLAG_EXP2 0x200000 /* for experimental use */
603 #define EXT_FLAG_EXP3 0x400000 /* for experimental use */
604 #define EXT_FLAG_EXP4 0x800000 /* for experimental use */
607 * EXT flag description for use with printf(9) %b identifier.
609 #define EXT_FLAG_BITS \
610 "\20\1EXT_FLAG_EMBREF\2EXT_FLAG_EXTREF\5EXT_FLAG_NOFREE" \
611 "\21EXT_FLAG_VENDOR1\22EXT_FLAG_VENDOR2\23EXT_FLAG_VENDOR3" \
612 "\24EXT_FLAG_VENDOR4\25EXT_FLAG_EXP1\26EXT_FLAG_EXP2\27EXT_FLAG_EXP3" \
616 * Flags indicating checksum, segmentation and other offload work to be
617 * done, or already done, by hardware or lower layers. It is split into
618 * separate inbound and outbound flags.
620 * Outbound flags that are set by upper protocol layers requesting lower
621 * layers, or ideally the hardware, to perform these offloading tasks.
622 * For outbound packets this field and its flags can be directly tested
623 * against ifnet if_hwassist. Note that the outbound and the inbound flags do
624 * not collide right now but they could be allowed to (as long as the flags are
625 * scrubbed appropriately when the direction of an mbuf changes). CSUM_BITS
626 * would also have to split into CSUM_BITS_TX and CSUM_BITS_RX.
628 * CSUM_INNER_<x> is the same as CSUM_<x> but it applies to the inner frame.
629 * The CSUM_ENCAP_<x> bits identify the outer encapsulation.
631 #define CSUM_IP 0x00000001 /* IP header checksum offload */
632 #define CSUM_IP_UDP 0x00000002 /* UDP checksum offload */
633 #define CSUM_IP_TCP 0x00000004 /* TCP checksum offload */
634 #define CSUM_IP_SCTP 0x00000008 /* SCTP checksum offload */
635 #define CSUM_IP_TSO 0x00000010 /* TCP segmentation offload */
636 #define CSUM_IP_ISCSI 0x00000020 /* iSCSI checksum offload */
638 #define CSUM_INNER_IP6_UDP 0x00000040
639 #define CSUM_INNER_IP6_TCP 0x00000080
640 #define CSUM_INNER_IP6_TSO 0x00000100
641 #define CSUM_IP6_UDP 0x00000200 /* UDP checksum offload */
642 #define CSUM_IP6_TCP 0x00000400 /* TCP checksum offload */
643 #define CSUM_IP6_SCTP 0x00000800 /* SCTP checksum offload */
644 #define CSUM_IP6_TSO 0x00001000 /* TCP segmentation offload */
645 #define CSUM_IP6_ISCSI 0x00002000 /* iSCSI checksum offload */
647 #define CSUM_INNER_IP 0x00004000
648 #define CSUM_INNER_IP_UDP 0x00008000
649 #define CSUM_INNER_IP_TCP 0x00010000
650 #define CSUM_INNER_IP_TSO 0x00020000
652 #define CSUM_ENCAP_VXLAN 0x00040000 /* VXLAN outer encapsulation */
653 #define CSUM_ENCAP_RSVD1 0x00080000
655 /* Inbound checksum support where the checksum was verified by hardware. */
656 #define CSUM_INNER_L3_CALC 0x00100000
657 #define CSUM_INNER_L3_VALID 0x00200000
658 #define CSUM_INNER_L4_CALC 0x00400000
659 #define CSUM_INNER_L4_VALID 0x00800000
660 #define CSUM_L3_CALC 0x01000000 /* calculated layer 3 csum */
661 #define CSUM_L3_VALID 0x02000000 /* checksum is correct */
662 #define CSUM_L4_CALC 0x04000000 /* calculated layer 4 csum */
663 #define CSUM_L4_VALID 0x08000000 /* checksum is correct */
664 #define CSUM_L5_CALC 0x10000000 /* calculated layer 5 csum */
665 #define CSUM_L5_VALID 0x20000000 /* checksum is correct */
666 #define CSUM_COALESCED 0x40000000 /* contains merged segments */
668 #define CSUM_SND_TAG 0x80000000 /* Packet header has send tag */
670 #define CSUM_FLAGS_TX (CSUM_IP | CSUM_IP_UDP | CSUM_IP_TCP | CSUM_IP_SCTP | \
671 CSUM_IP_TSO | CSUM_IP_ISCSI | CSUM_INNER_IP6_UDP | CSUM_INNER_IP6_TCP | \
672 CSUM_INNER_IP6_TSO | CSUM_IP6_UDP | CSUM_IP6_TCP | CSUM_IP6_SCTP | \
673 CSUM_IP6_TSO | CSUM_IP6_ISCSI | CSUM_INNER_IP | CSUM_INNER_IP_UDP | \
674 CSUM_INNER_IP_TCP | CSUM_INNER_IP_TSO | CSUM_ENCAP_VXLAN | \
675 CSUM_ENCAP_RSVD1 | CSUM_SND_TAG)
677 #define CSUM_FLAGS_RX (CSUM_INNER_L3_CALC | CSUM_INNER_L3_VALID | \
678 CSUM_INNER_L4_CALC | CSUM_INNER_L4_VALID | CSUM_L3_CALC | CSUM_L3_VALID | \
679 CSUM_L4_CALC | CSUM_L4_VALID | CSUM_L5_CALC | CSUM_L5_VALID | \
683 * CSUM flag description for use with printf(9) %b identifier.
686 "\20\1CSUM_IP\2CSUM_IP_UDP\3CSUM_IP_TCP\4CSUM_IP_SCTP\5CSUM_IP_TSO" \
687 "\6CSUM_IP_ISCSI\7CSUM_INNER_IP6_UDP\10CSUM_INNER_IP6_TCP" \
688 "\11CSUM_INNER_IP6_TSO\12CSUM_IP6_UDP\13CSUM_IP6_TCP\14CSUM_IP6_SCTP" \
689 "\15CSUM_IP6_TSO\16CSUM_IP6_ISCSI\17CSUM_INNER_IP\20CSUM_INNER_IP_UDP" \
690 "\21CSUM_INNER_IP_TCP\22CSUM_INNER_IP_TSO\23CSUM_ENCAP_VXLAN" \
691 "\24CSUM_ENCAP_RSVD1\25CSUM_INNER_L3_CALC\26CSUM_INNER_L3_VALID" \
692 "\27CSUM_INNER_L4_CALC\30CSUM_INNER_L4_VALID\31CSUM_L3_CALC" \
693 "\32CSUM_L3_VALID\33CSUM_L4_CALC\34CSUM_L4_VALID\35CSUM_L5_CALC" \
694 "\36CSUM_L5_VALID\37CSUM_COALESCED\40CSUM_SND_TAG"
696 /* CSUM flags compatibility mappings. */
697 #define CSUM_IP_CHECKED CSUM_L3_CALC
698 #define CSUM_IP_VALID CSUM_L3_VALID
699 #define CSUM_DATA_VALID CSUM_L4_VALID
700 #define CSUM_PSEUDO_HDR CSUM_L4_CALC
701 #define CSUM_SCTP_VALID CSUM_L4_VALID
702 #define CSUM_DELAY_DATA (CSUM_TCP|CSUM_UDP)
703 #define CSUM_DELAY_IP CSUM_IP /* Only v4, no v6 IP hdr csum */
704 #define CSUM_DELAY_DATA_IPV6 (CSUM_TCP_IPV6|CSUM_UDP_IPV6)
705 #define CSUM_DATA_VALID_IPV6 CSUM_DATA_VALID
706 #define CSUM_TCP CSUM_IP_TCP
707 #define CSUM_UDP CSUM_IP_UDP
708 #define CSUM_SCTP CSUM_IP_SCTP
709 #define CSUM_TSO (CSUM_IP_TSO|CSUM_IP6_TSO)
710 #define CSUM_INNER_TSO (CSUM_INNER_IP_TSO|CSUM_INNER_IP6_TSO)
711 #define CSUM_UDP_IPV6 CSUM_IP6_UDP
712 #define CSUM_TCP_IPV6 CSUM_IP6_TCP
713 #define CSUM_SCTP_IPV6 CSUM_IP6_SCTP
716 * mbuf types describing the content of the mbuf (including external storage).
718 #define MT_NOTMBUF 0 /* USED INTERNALLY ONLY! Object is not mbuf */
719 #define MT_DATA 1 /* dynamic (data) allocation */
720 #define MT_HEADER MT_DATA /* packet header, use M_PKTHDR instead */
722 #define MT_VENDOR1 4 /* for vendor-internal use */
723 #define MT_VENDOR2 5 /* for vendor-internal use */
724 #define MT_VENDOR3 6 /* for vendor-internal use */
725 #define MT_VENDOR4 7 /* for vendor-internal use */
727 #define MT_SONAME 8 /* socket name */
729 #define MT_EXP1 9 /* for experimental use */
730 #define MT_EXP2 10 /* for experimental use */
731 #define MT_EXP3 11 /* for experimental use */
732 #define MT_EXP4 12 /* for experimental use */
734 #define MT_CONTROL 14 /* extra-data protocol message */
735 #define MT_EXTCONTROL 15 /* control message with externalized contents */
736 #define MT_OOBDATA 16 /* expedited data */
738 #define MT_NOINIT 255 /* Not a type but a flag to allocate
739 a non-initialized mbuf */
742 * String names of mbuf-related UMA(9) and malloc(9) types. Exposed to
743 * !_KERNEL so that monitoring tools can look up the zones with
746 #define MBUF_MEM_NAME "mbuf"
747 #define MBUF_CLUSTER_MEM_NAME "mbuf_cluster"
748 #define MBUF_PACKET_MEM_NAME "mbuf_packet"
749 #define MBUF_JUMBOP_MEM_NAME "mbuf_jumbo_page"
750 #define MBUF_JUMBO9_MEM_NAME "mbuf_jumbo_9k"
751 #define MBUF_JUMBO16_MEM_NAME "mbuf_jumbo_16k"
752 #define MBUF_TAG_MEM_NAME "mbuf_tag"
753 #define MBUF_EXTREFCNT_MEM_NAME "mbuf_ext_refcnt"
754 #define MBUF_EXTPGS_MEM_NAME "mbuf_extpgs"
759 #define MBUF_CHECKSLEEP(how) do { \
760 if (how == M_WAITOK) \
761 WITNESS_WARN(WARN_GIANTOK | WARN_SLEEPOK, NULL, \
762 "Sleeping in \"%s\"", __func__); \
765 #define MBUF_CHECKSLEEP(how)
769 * Network buffer allocation API
771 * The rest of it is defined in kern/kern_mbuf.c
773 extern uma_zone_t zone_mbuf;
774 extern uma_zone_t zone_clust;
775 extern uma_zone_t zone_pack;
776 extern uma_zone_t zone_jumbop;
777 extern uma_zone_t zone_jumbo9;
778 extern uma_zone_t zone_jumbo16;
779 extern uma_zone_t zone_extpgs;
781 void mb_dupcl(struct mbuf *, struct mbuf *);
782 void mb_free_ext(struct mbuf *);
783 void mb_free_extpg(struct mbuf *);
784 void mb_free_mext_pgs(struct mbuf *);
785 struct mbuf *mb_alloc_ext_pgs(int, m_ext_free_t);
786 struct mbuf *mb_alloc_ext_plus_pages(int, int);
787 struct mbuf *mb_mapped_to_unmapped(struct mbuf *, int, int, int,
789 int mb_unmapped_compress(struct mbuf *m);
790 struct mbuf *mb_unmapped_to_ext(struct mbuf *m);
791 void mb_free_notready(struct mbuf *m, int count);
792 void m_adj(struct mbuf *, int);
793 int m_apply(struct mbuf *, int, int,
794 int (*)(void *, void *, u_int), void *);
795 int m_append(struct mbuf *, int, c_caddr_t);
796 void m_cat(struct mbuf *, struct mbuf *);
797 void m_catpkt(struct mbuf *, struct mbuf *);
798 int m_clget(struct mbuf *m, int how);
799 void *m_cljget(struct mbuf *m, int how, int size);
800 struct mbuf *m_collapse(struct mbuf *, int, int);
801 void m_copyback(struct mbuf *, int, int, c_caddr_t);
802 void m_copydata(const struct mbuf *, int, int, caddr_t);
803 struct mbuf *m_copym(struct mbuf *, int, int, int);
804 struct mbuf *m_copypacket(struct mbuf *, int);
805 void m_copy_pkthdr(struct mbuf *, struct mbuf *);
806 struct mbuf *m_copyup(struct mbuf *, int, int);
807 struct mbuf *m_defrag(struct mbuf *, int);
808 void m_demote_pkthdr(struct mbuf *);
809 void m_demote(struct mbuf *, int, int);
810 struct mbuf *m_devget(char *, int, int, struct ifnet *,
811 void (*)(char *, caddr_t, u_int));
812 void m_dispose_extcontrolm(struct mbuf *m);
813 struct mbuf *m_dup(const struct mbuf *, int);
814 int m_dup_pkthdr(struct mbuf *, const struct mbuf *, int);
815 void m_extadd(struct mbuf *, char *, u_int, m_ext_free_t,
816 void *, void *, int, int);
817 u_int m_fixhdr(struct mbuf *);
818 struct mbuf *m_fragment(struct mbuf *, int, int);
819 void m_freem(struct mbuf *);
820 struct mbuf *m_get2(int, int, short, int);
821 struct mbuf *m_getjcl(int, short, int, int);
822 struct mbuf *m_getm2(struct mbuf *, int, int, short, int);
823 struct mbuf *m_getptr(struct mbuf *, int, int *);
824 u_int m_length(struct mbuf *, struct mbuf **);
825 int m_mbuftouio(struct uio *, const struct mbuf *, int);
826 int m_unmappedtouio(const struct mbuf *, int, struct uio *, int);
827 void m_move_pkthdr(struct mbuf *, struct mbuf *);
828 int m_pkthdr_init(struct mbuf *, int);
829 struct mbuf *m_prepend(struct mbuf *, int, int);
830 void m_print(const struct mbuf *, int);
831 struct mbuf *m_pulldown(struct mbuf *, int, int, int *);
832 struct mbuf *m_pullup(struct mbuf *, int);
833 int m_sanity(struct mbuf *, int);
834 struct mbuf *m_split(struct mbuf *, int, int);
835 struct mbuf *m_uiotombuf(struct uio *, int, int, int, int);
836 struct mbuf *m_unshare(struct mbuf *, int);
837 void m_snd_tag_init(struct m_snd_tag *, struct ifnet *, u_int);
838 void m_snd_tag_destroy(struct m_snd_tag *);
852 #if MJUMPAGESIZE != MCLBYTES
864 panic("%s: invalid cluster size %d", __func__, size);
871 * Associated an external reference counted buffer with an mbuf.
874 m_extaddref(struct mbuf *m, char *buf, u_int size, u_int *ref_cnt,
875 m_ext_free_t freef, void *arg1, void *arg2)
878 KASSERT(ref_cnt != NULL, ("%s: ref_cnt not provided", __func__));
880 atomic_add_int(ref_cnt, 1);
882 m->m_ext.ext_buf = buf;
883 m->m_ext.ext_cnt = ref_cnt;
884 m->m_data = m->m_ext.ext_buf;
885 m->m_ext.ext_size = size;
886 m->m_ext.ext_free = freef;
887 m->m_ext.ext_arg1 = arg1;
888 m->m_ext.ext_arg2 = arg2;
889 m->m_ext.ext_type = EXT_EXTREF;
890 m->m_ext.ext_flags = 0;
893 static __inline uma_zone_t
902 #if MJUMPAGESIZE != MCLBYTES
914 panic("%s: invalid cluster size %d", __func__, size);
921 * Initialize an mbuf with linear storage.
923 * Inline because the consumer text overhead will be roughly the same to
924 * initialize or call a function with this many parameters and M_PKTHDR
925 * should go away with constant propagation for !MGETHDR.
928 m_init(struct mbuf *m, int how, short type, int flags)
934 m->m_data = m->m_dat;
938 if (flags & M_PKTHDR)
939 error = m_pkthdr_init(m, how);
943 MBUF_PROBE5(m__init, m, how, type, flags, error);
947 static __inline struct mbuf *
948 m_get(int how, short type)
955 m = uma_zalloc_arg(zone_mbuf, &args, how);
956 MBUF_PROBE3(m__get, how, type, m);
960 static __inline struct mbuf *
961 m_gethdr(int how, short type)
966 args.flags = M_PKTHDR;
968 m = uma_zalloc_arg(zone_mbuf, &args, how);
969 MBUF_PROBE3(m__gethdr, how, type, m);
973 static __inline struct mbuf *
974 m_getcl(int how, short type, int flags)
981 m = uma_zalloc_arg(zone_pack, &args, how);
982 MBUF_PROBE4(m__getcl, how, type, flags, m);
987 * XXX: m_cljset() is a dangerous API. One must attach only a new,
988 * unreferenced cluster to an mbuf(9). It is not possible to assert
989 * that, so care can be taken only by users of the API.
992 m_cljset(struct mbuf *m, void *cl, int type)
1000 #if MJUMPAGESIZE != MCLBYTES
1002 size = MJUMPAGESIZE;
1012 panic("%s: unknown cluster type %d", __func__, type);
1016 m->m_data = m->m_ext.ext_buf = cl;
1017 m->m_ext.ext_free = m->m_ext.ext_arg1 = m->m_ext.ext_arg2 = NULL;
1018 m->m_ext.ext_size = size;
1019 m->m_ext.ext_type = type;
1020 m->m_ext.ext_flags = EXT_FLAG_EMBREF;
1021 m->m_ext.ext_count = 1;
1022 m->m_flags |= M_EXT;
1023 MBUF_PROBE3(m__cljset, m, cl, type);
1026 static __inline void
1027 m_chtype(struct mbuf *m, short new_type)
1030 m->m_type = new_type;
1033 static __inline void
1034 m_clrprotoflags(struct mbuf *m)
1038 m->m_flags &= ~M_PROTOFLAGS;
1043 static __inline struct mbuf *
1044 m_last(struct mbuf *m)
1053 m_extrefcnt(struct mbuf *m)
1056 KASSERT(m->m_flags & M_EXT, ("%s: M_EXT missing", __func__));
1058 return ((m->m_ext.ext_flags & EXT_FLAG_EMBREF) ? m->m_ext.ext_count :
1063 * mbuf, cluster, and external object allocation macros (for compatibility
1066 #define M_MOVE_PKTHDR(to, from) m_move_pkthdr((to), (from))
1067 #define MGET(m, how, type) ((m) = m_get((how), (type)))
1068 #define MGETHDR(m, how, type) ((m) = m_gethdr((how), (type)))
1069 #define MCLGET(m, how) m_clget((m), (how))
1070 #define MEXTADD(m, buf, size, free, arg1, arg2, flags, type) \
1071 m_extadd((m), (char *)(buf), (size), (free), (arg1), (arg2), \
1073 #define m_getm(m, len, how, type) \
1074 m_getm2((m), (len), (how), (type), M_PKTHDR)
1077 * Evaluate TRUE if it's safe to write to the mbuf m's data region (this can
1078 * be both the local data payload, or an external buffer area, depending on
1079 * whether M_EXT is set).
1081 #define M_WRITABLE(m) (((m)->m_flags & (M_RDONLY | M_EXTPG)) == 0 && \
1082 (!(((m)->m_flags & M_EXT)) || \
1083 (m_extrefcnt(m) == 1)))
1085 /* Check if the supplied mbuf has a packet header, or else panic. */
1086 #define M_ASSERTPKTHDR(m) \
1087 KASSERT((m) != NULL && (m)->m_flags & M_PKTHDR, \
1088 ("%s: no mbuf packet header!", __func__))
1090 /* Check if mbuf is multipage. */
1091 #define M_ASSERTEXTPG(m) \
1092 KASSERT(((m)->m_flags & (M_EXTPG|M_PKTHDR)) == M_EXTPG, \
1093 ("%s: m %p is not multipage!", __func__, m))
1096 * Ensure that the supplied mbuf is a valid, non-free mbuf.
1098 * XXX: Broken at the moment. Need some UMA magic to make it work again.
1100 #define M_ASSERTVALID(m) \
1101 KASSERT((((struct mbuf *)m)->m_flags & 0) == 0, \
1102 ("%s: attempted use of a free mbuf!", __func__))
1105 * Return the address of the start of the buffer associated with an mbuf,
1106 * handling external storage, packet-header mbufs, and regular data mbufs.
1108 #define M_START(m) \
1109 (((m)->m_flags & M_EXTPG) ? NULL : \
1110 ((m)->m_flags & M_EXT) ? (m)->m_ext.ext_buf : \
1111 ((m)->m_flags & M_PKTHDR) ? &(m)->m_pktdat[0] : \
1115 * Return the size of the buffer associated with an mbuf, handling external
1116 * storage, packet-header mbufs, and regular data mbufs.
1119 (((m)->m_flags & M_EXT) ? (m)->m_ext.ext_size : \
1120 ((m)->m_flags & M_PKTHDR) ? MHLEN : \
1124 * Set the m_data pointer of a newly allocated mbuf to place an object of the
1125 * specified size at the end of the mbuf, longword aligned.
1127 * NB: Historically, we had M_ALIGN(), MH_ALIGN(), and MEXT_ALIGN() as
1128 * separate macros, each asserting that it was called at the proper moment.
1129 * This required callers to themselves test the storage type and call the
1130 * right one. Rather than require callers to be aware of those layout
1131 * decisions, we centralize here.
1133 static __inline void
1134 m_align(struct mbuf *m, int len)
1137 const char *msg = "%s: not a virgin mbuf";
1141 KASSERT(m->m_data == M_START(m), (msg, __func__));
1143 adjust = M_SIZE(m) - len;
1144 m->m_data += adjust &~ (sizeof(long)-1);
1147 #define M_ALIGN(m, len) m_align(m, len)
1148 #define MH_ALIGN(m, len) m_align(m, len)
1149 #define MEXT_ALIGN(m, len) m_align(m, len)
1152 * Compute the amount of space available before the current start of data in
1155 * The M_WRITABLE() is a temporary, conservative safety measure: the burden
1156 * of checking writability of the mbuf data area rests solely with the caller.
1158 * NB: In previous versions, M_LEADINGSPACE() would only check M_WRITABLE()
1159 * for mbufs with external storage. We now allow mbuf-embedded data to be
1160 * read-only as well.
1162 #define M_LEADINGSPACE(m) \
1163 (M_WRITABLE(m) ? ((m)->m_data - M_START(m)) : 0)
1166 * Compute the amount of space available after the end of data in an mbuf.
1168 * The M_WRITABLE() is a temporary, conservative safety measure: the burden
1169 * of checking writability of the mbuf data area rests solely with the caller.
1171 * NB: In previous versions, M_TRAILINGSPACE() would only check M_WRITABLE()
1172 * for mbufs with external storage. We now allow mbuf-embedded data to be
1173 * read-only as well.
1175 #define M_TRAILINGSPACE(m) \
1177 ((M_START(m) + M_SIZE(m)) - ((m)->m_data + (m)->m_len)) : 0)
1180 * Arrange to prepend space of size plen to mbuf m. If a new mbuf must be
1181 * allocated, how specifies whether to wait. If the allocation fails, the
1182 * original mbuf chain is freed and m is set to NULL.
1184 #define M_PREPEND(m, plen, how) do { \
1185 struct mbuf **_mmp = &(m); \
1186 struct mbuf *_mm = *_mmp; \
1187 int _mplen = (plen); \
1188 int __mhow = (how); \
1190 MBUF_CHECKSLEEP(how); \
1191 if (M_LEADINGSPACE(_mm) >= _mplen) { \
1192 _mm->m_data -= _mplen; \
1193 _mm->m_len += _mplen; \
1195 _mm = m_prepend(_mm, _mplen, __mhow); \
1196 if (_mm != NULL && _mm->m_flags & M_PKTHDR) \
1197 _mm->m_pkthdr.len += _mplen; \
1202 * Change mbuf to new type. This is a relatively expensive operation and
1203 * should be avoided.
1205 #define MCHTYPE(m, t) m_chtype((m), (t))
1207 /* Return the rcvif of a packet header. */
1208 static __inline struct ifnet *
1209 m_rcvif(struct mbuf *m)
1213 if (m->m_pkthdr.csum_flags & CSUM_SND_TAG)
1215 return (m->m_pkthdr.rcvif);
1218 /* Length to m_copy to copy all. */
1219 #define M_COPYALL 1000000000
1221 extern int max_datalen; /* MHLEN - max_hdr */
1222 extern int max_hdr; /* Largest link + protocol header */
1223 extern int max_linkhdr; /* Largest link-level header */
1224 extern int max_protohdr; /* Largest protocol header */
1225 extern int nmbclusters; /* Maximum number of clusters */
1226 extern bool mb_use_ext_pgs; /* Use ext_pgs for sendfile */
1229 * Network packets may have annotations attached by affixing a list of
1230 * "packet tags" to the pkthdr structure. Packet tags are dynamically
1231 * allocated semi-opaque data structures that have a fixed header
1232 * (struct m_tag) that specifies the size of the memory block and a
1233 * <cookie,type> pair that identifies it. The cookie is a 32-bit unique
1234 * unsigned value used to identify a module or ABI. By convention this value
1235 * is chosen as the date+time that the module is created, expressed as the
1236 * number of seconds since the epoch (e.g., using date -u +'%s'). The type
1237 * value is an ABI/module-specific value that identifies a particular
1238 * annotation and is private to the module. For compatibility with systems
1239 * like OpenBSD that define packet tags w/o an ABI/module cookie, the value
1240 * PACKET_ABI_COMPAT is used to implement m_tag_get and m_tag_find
1241 * compatibility shim functions and several tag types are defined below.
1242 * Users that do not require compatibility should use a private cookie value
1243 * so that packet tag-related definitions can be maintained privately.
1245 * Note that the packet tag returned by m_tag_alloc has the default memory
1246 * alignment implemented by malloc. To reference private data one can use a
1249 * struct m_tag *mtag = m_tag_alloc(...);
1250 * struct foo *p = (struct foo *)(mtag+1);
1252 * if the alignment of struct m_tag is sufficient for referencing members of
1253 * struct foo. Otherwise it is necessary to embed struct m_tag within the
1254 * private data structure to insure proper alignment; e.g.,
1260 * struct foo *p = (struct foo *) m_tag_alloc(...);
1261 * struct m_tag *mtag = &p->tag;
1265 * Persistent tags stay with an mbuf until the mbuf is reclaimed. Otherwise
1266 * tags are expected to ``vanish'' when they pass through a network
1267 * interface. For most interfaces this happens normally as the tags are
1268 * reclaimed when the mbuf is free'd. However in some special cases
1269 * reclaiming must be done manually. An example is packets that pass through
1270 * the loopback interface. Also, one must be careful to do this when
1271 * ``turning around'' packets (e.g., icmp_reflect).
1273 * To mark a tag persistent bit-or this flag in when defining the tag id.
1274 * The tag will then be treated as described above.
1276 #define MTAG_PERSISTENT 0x800
1278 #define PACKET_TAG_NONE 0 /* Nadda */
1280 /* Packet tags for use with PACKET_ABI_COMPAT. */
1281 #define PACKET_TAG_IPSEC_IN_DONE 1 /* IPsec applied, in */
1282 #define PACKET_TAG_IPSEC_OUT_DONE 2 /* IPsec applied, out */
1283 #define PACKET_TAG_IPSEC_IN_CRYPTO_DONE 3 /* NIC IPsec crypto done */
1284 #define PACKET_TAG_IPSEC_OUT_CRYPTO_NEEDED 4 /* NIC IPsec crypto req'ed */
1285 #define PACKET_TAG_IPSEC_IN_COULD_DO_CRYPTO 5 /* NIC notifies IPsec */
1286 #define PACKET_TAG_IPSEC_PENDING_TDB 6 /* Reminder to do IPsec */
1287 #define PACKET_TAG_BRIDGE 7 /* Bridge processing done */
1288 #define PACKET_TAG_GIF 8 /* GIF processing done */
1289 #define PACKET_TAG_GRE 9 /* GRE processing done */
1290 #define PACKET_TAG_IN_PACKET_CHECKSUM 10 /* NIC checksumming done */
1291 #define PACKET_TAG_ENCAP 11 /* Encap. processing */
1292 #define PACKET_TAG_IPSEC_SOCKET 12 /* IPSEC socket ref */
1293 #define PACKET_TAG_IPSEC_HISTORY 13 /* IPSEC history */
1294 #define PACKET_TAG_IPV6_INPUT 14 /* IPV6 input processing */
1295 #define PACKET_TAG_DUMMYNET 15 /* dummynet info */
1296 #define PACKET_TAG_DIVERT 17 /* divert info */
1297 #define PACKET_TAG_IPFORWARD 18 /* ipforward info */
1298 #define PACKET_TAG_MACLABEL (19 | MTAG_PERSISTENT) /* MAC label */
1299 #define PACKET_TAG_PF (21 | MTAG_PERSISTENT) /* PF/ALTQ information */
1300 #define PACKET_TAG_RTSOCKFAM 25 /* rtsock sa family */
1301 #define PACKET_TAG_IPOPTIONS 27 /* Saved IP options */
1302 #define PACKET_TAG_CARP 28 /* CARP info */
1303 #define PACKET_TAG_IPSEC_NAT_T_PORTS 29 /* two uint16_t */
1304 #define PACKET_TAG_ND_OUTGOING 30 /* ND outgoing */
1306 /* Specific cookies and tags. */
1308 /* Packet tag routines. */
1309 struct m_tag *m_tag_alloc(u_int32_t, int, int, int);
1310 void m_tag_delete(struct mbuf *, struct m_tag *);
1311 void m_tag_delete_chain(struct mbuf *, struct m_tag *);
1312 void m_tag_free_default(struct m_tag *);
1313 struct m_tag *m_tag_locate(struct mbuf *, u_int32_t, int, struct m_tag *);
1314 struct m_tag *m_tag_copy(struct m_tag *, int);
1315 int m_tag_copy_chain(struct mbuf *, const struct mbuf *, int);
1316 void m_tag_delete_nonpersistent(struct mbuf *);
1319 * Initialize the list of tags associated with an mbuf.
1321 static __inline void
1322 m_tag_init(struct mbuf *m)
1325 SLIST_INIT(&m->m_pkthdr.tags);
1329 * Set up the contents of a tag. Note that this does not fill in the free
1330 * method; the caller is expected to do that.
1332 * XXX probably should be called m_tag_init, but that was already taken.
1334 static __inline void
1335 m_tag_setup(struct m_tag *t, u_int32_t cookie, int type, int len)
1340 t->m_tag_cookie = cookie;
1344 * Reclaim resources associated with a tag.
1346 static __inline void
1347 m_tag_free(struct m_tag *t)
1350 (*t->m_tag_free)(t);
1354 * Return the first tag associated with an mbuf.
1356 static __inline struct m_tag *
1357 m_tag_first(struct mbuf *m)
1360 return (SLIST_FIRST(&m->m_pkthdr.tags));
1364 * Return the next tag in the list of tags associated with an mbuf.
1366 static __inline struct m_tag *
1367 m_tag_next(struct mbuf *m __unused, struct m_tag *t)
1370 return (SLIST_NEXT(t, m_tag_link));
1374 * Prepend a tag to the list of tags associated with an mbuf.
1376 static __inline void
1377 m_tag_prepend(struct mbuf *m, struct m_tag *t)
1380 SLIST_INSERT_HEAD(&m->m_pkthdr.tags, t, m_tag_link);
1384 * Unlink a tag from the list of tags associated with an mbuf.
1386 static __inline void
1387 m_tag_unlink(struct mbuf *m, struct m_tag *t)
1390 SLIST_REMOVE(&m->m_pkthdr.tags, t, m_tag, m_tag_link);
1393 /* These are for OpenBSD compatibility. */
1394 #define MTAG_ABI_COMPAT 0 /* compatibility ABI */
1396 static __inline struct m_tag *
1397 m_tag_get(int type, int length, int wait)
1399 return (m_tag_alloc(MTAG_ABI_COMPAT, type, length, wait));
1402 static __inline struct m_tag *
1403 m_tag_find(struct mbuf *m, int type, struct m_tag *start)
1405 return (SLIST_EMPTY(&m->m_pkthdr.tags) ? (struct m_tag *)NULL :
1406 m_tag_locate(m, MTAG_ABI_COMPAT, type, start));
1409 static inline struct m_snd_tag *
1410 m_snd_tag_ref(struct m_snd_tag *mst)
1413 refcount_acquire(&mst->refcount);
1418 m_snd_tag_rele(struct m_snd_tag *mst)
1421 if (refcount_release(&mst->refcount))
1422 m_snd_tag_destroy(mst);
1425 static __inline struct mbuf *
1426 m_free(struct mbuf *m)
1428 struct mbuf *n = m->m_next;
1430 MBUF_PROBE1(m__free, m);
1431 if ((m->m_flags & (M_PKTHDR|M_NOFREE)) == (M_PKTHDR|M_NOFREE))
1432 m_tag_delete_chain(m, NULL);
1433 if (m->m_flags & M_PKTHDR && m->m_pkthdr.csum_flags & CSUM_SND_TAG)
1434 m_snd_tag_rele(m->m_pkthdr.snd_tag);
1435 if (m->m_flags & M_EXTPG)
1437 else if (m->m_flags & M_EXT)
1439 else if ((m->m_flags & M_NOFREE) == 0)
1440 uma_zfree(zone_mbuf, m);
1445 rt_m_getfib(struct mbuf *m)
1447 KASSERT(m->m_flags & M_PKTHDR , ("Attempt to get FIB from non header mbuf."));
1448 return (m->m_pkthdr.fibnum);
1451 #define M_GETFIB(_m) rt_m_getfib(_m)
1453 #define M_SETFIB(_m, _fib) do { \
1454 KASSERT((_m)->m_flags & M_PKTHDR, ("Attempt to set FIB on non header mbuf.")); \
1455 ((_m)->m_pkthdr.fibnum) = (_fib); \
1458 /* flags passed as first argument for "m_xxx_tcpip_hash()" */
1459 #define MBUF_HASHFLAG_L2 (1 << 2)
1460 #define MBUF_HASHFLAG_L3 (1 << 3)
1461 #define MBUF_HASHFLAG_L4 (1 << 4)
1463 /* mbuf hashing helper routines */
1464 uint32_t m_ether_tcpip_hash_init(void);
1465 uint32_t m_ether_tcpip_hash(const uint32_t, const struct mbuf *, uint32_t);
1466 uint32_t m_infiniband_tcpip_hash_init(void);
1467 uint32_t m_infiniband_tcpip_hash(const uint32_t, const struct mbuf *, uint32_t);
1469 #ifdef MBUF_PROFILING
1470 void m_profile(struct mbuf *m);
1471 #define M_PROFILE(m) m_profile(m)
1473 #define M_PROFILE(m)
1477 STAILQ_HEAD(, mbuf) mq_head;
1483 mbufq_init(struct mbufq *mq, int maxlen)
1486 STAILQ_INIT(&mq->mq_head);
1487 mq->mq_maxlen = maxlen;
1491 static inline struct mbuf *
1492 mbufq_flush(struct mbufq *mq)
1496 m = STAILQ_FIRST(&mq->mq_head);
1497 STAILQ_INIT(&mq->mq_head);
1503 mbufq_drain(struct mbufq *mq)
1507 n = mbufq_flush(mq);
1508 while ((m = n) != NULL) {
1509 n = STAILQ_NEXT(m, m_stailqpkt);
1514 static inline struct mbuf *
1515 mbufq_first(const struct mbufq *mq)
1518 return (STAILQ_FIRST(&mq->mq_head));
1521 static inline struct mbuf *
1522 mbufq_last(const struct mbufq *mq)
1525 return (STAILQ_LAST(&mq->mq_head, mbuf, m_stailqpkt));
1529 mbufq_full(const struct mbufq *mq)
1532 return (mq->mq_maxlen > 0 && mq->mq_len >= mq->mq_maxlen);
1536 mbufq_len(const struct mbufq *mq)
1539 return (mq->mq_len);
1543 mbufq_enqueue(struct mbufq *mq, struct mbuf *m)
1548 STAILQ_INSERT_TAIL(&mq->mq_head, m, m_stailqpkt);
1553 static inline struct mbuf *
1554 mbufq_dequeue(struct mbufq *mq)
1558 m = STAILQ_FIRST(&mq->mq_head);
1560 STAILQ_REMOVE_HEAD(&mq->mq_head, m_stailqpkt);
1561 m->m_nextpkt = NULL;
1568 mbufq_prepend(struct mbufq *mq, struct mbuf *m)
1571 STAILQ_INSERT_HEAD(&mq->mq_head, m, m_stailqpkt);
1576 * Note: this doesn't enforce the maximum list size for dst.
1579 mbufq_concat(struct mbufq *mq_dst, struct mbufq *mq_src)
1582 mq_dst->mq_len += mq_src->mq_len;
1583 STAILQ_CONCAT(&mq_dst->mq_head, &mq_src->mq_head);
1587 #ifdef _SYS_TIMESPEC_H_
1589 mbuf_tstmp2timespec(struct mbuf *m, struct timespec *ts)
1592 KASSERT((m->m_flags & M_PKTHDR) != 0, ("mbuf %p no M_PKTHDR", m));
1593 KASSERT((m->m_flags & (M_TSTMP|M_TSTMP_LRO)) != 0, ("mbuf %p no M_TSTMP or M_TSTMP_LRO", m));
1594 ts->tv_sec = m->m_pkthdr.rcv_tstmp / 1000000000;
1595 ts->tv_nsec = m->m_pkthdr.rcv_tstmp % 1000000000;
1600 /* Invoked from the debugnet client code. */
1601 void debugnet_mbuf_drain(void);
1602 void debugnet_mbuf_start(void);
1603 void debugnet_mbuf_finish(void);
1604 void debugnet_mbuf_reinit(int nmbuf, int nclust, int clsize);
1608 mbuf_has_tls_session(struct mbuf *m)
1611 if (m->m_flags & M_EXTPG) {
1612 if (m->m_epg_tls != NULL) {
1619 #endif /* _KERNEL */
1620 #endif /* !_SYS_MBUF_H_ */