2 * SPDX-License-Identifier: BSD-3-Clause
4 * Copyright (c) 1982, 1986, 1988, 1993
5 * The Regents of the University of California.
8 * Redistribution and use in source and binary forms, with or without
9 * modification, are permitted provided that the following conditions
11 * 1. Redistributions of source code must retain the above copyright
12 * notice, this list of conditions and the following disclaimer.
13 * 2. Redistributions in binary form must reproduce the above copyright
14 * notice, this list of conditions and the following disclaimer in the
15 * documentation and/or other materials provided with the distribution.
16 * 3. Neither the name of the University nor the names of its contributors
17 * may be used to endorse or promote products derived from this software
18 * without specific prior written permission.
20 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
21 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
22 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
23 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
24 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
25 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
26 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
27 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
28 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
29 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
32 * @(#)mbuf.h 8.5 (Berkeley) 2/19/95
39 /* XXX: These includes suck. Sorry! */
40 #include <sys/queue.h>
42 #include <sys/systm.h>
43 #include <sys/refcount.h>
53 #define MBUF_PROBE1(probe, arg0) \
54 SDT_PROBE1(sdt, , , probe, arg0)
55 #define MBUF_PROBE2(probe, arg0, arg1) \
56 SDT_PROBE2(sdt, , , probe, arg0, arg1)
57 #define MBUF_PROBE3(probe, arg0, arg1, arg2) \
58 SDT_PROBE3(sdt, , , probe, arg0, arg1, arg2)
59 #define MBUF_PROBE4(probe, arg0, arg1, arg2, arg3) \
60 SDT_PROBE4(sdt, , , probe, arg0, arg1, arg2, arg3)
61 #define MBUF_PROBE5(probe, arg0, arg1, arg2, arg3, arg4) \
62 SDT_PROBE5(sdt, , , probe, arg0, arg1, arg2, arg3, arg4)
64 SDT_PROBE_DECLARE(sdt, , , m__init);
65 SDT_PROBE_DECLARE(sdt, , , m__gethdr);
66 SDT_PROBE_DECLARE(sdt, , , m__get);
67 SDT_PROBE_DECLARE(sdt, , , m__getcl);
68 SDT_PROBE_DECLARE(sdt, , , m__clget);
69 SDT_PROBE_DECLARE(sdt, , , m__cljget);
70 SDT_PROBE_DECLARE(sdt, , , m__cljset);
71 SDT_PROBE_DECLARE(sdt, , , m__free);
72 SDT_PROBE_DECLARE(sdt, , , m__freem);
77 * Mbufs are of a single size, MSIZE (sys/param.h), which includes overhead.
78 * An mbuf may add a single "mbuf cluster" of size MCLBYTES (also in
79 * sys/param.h), which has no additional overhead and is used instead of the
80 * internal data area; this is done when at least MINCLSIZE of data must be
81 * stored. Additionally, it is possible to allocate a separate buffer
82 * externally and attach it to the mbuf in a way similar to that of mbuf
85 * NB: These calculation do not take actual compiler-induced alignment and
86 * padding inside the complete struct mbuf into account. Appropriate
87 * attention is required when changing members of struct mbuf.
89 * MLEN is data length in a normal mbuf.
90 * MHLEN is data length in an mbuf with pktheader.
91 * MINCLSIZE is a smallest amount of data that should be put into cluster.
93 * Compile-time assertions in uipc_mbuf.c test these values to ensure that
97 #define MHSIZE offsetof(struct mbuf, m_dat)
98 #define MPKTHSIZE offsetof(struct mbuf, m_pktdat)
99 #define MLEN ((int)(MSIZE - MHSIZE))
100 #define MHLEN ((int)(MSIZE - MPKTHSIZE))
101 #define MINCLSIZE (MHLEN + 1)
106 * Macro for type conversion: convert mbuf pointer to data pointer of correct
109 * mtod(m, t) -- Convert mbuf pointer to data pointer of correct type.
110 * mtodo(m, o) -- Same as above but with offset 'o' into data.
112 #define mtod(m, t) ((t)((m)->m_data))
113 #define mtodo(m, o) ((void *)(((m)->m_data) + (o)))
116 * Argument structure passed to UMA routines during mbuf and packet
120 int flags; /* Flags for mbuf being allocated */
121 short type; /* Type of mbuf being allocated */
126 * Packet tag structure (see below for details).
129 SLIST_ENTRY(m_tag) m_tag_link; /* List of packet tags */
130 u_int16_t m_tag_id; /* Tag ID */
131 u_int16_t m_tag_len; /* Length of data */
132 u_int32_t m_tag_cookie; /* ABI/Module ID */
133 void (*m_tag_free)(struct m_tag *);
137 * Static network interface owned tag.
138 * Allocated through ifp->if_snd_tag_alloc().
141 struct ifnet *ifp; /* network interface tag belongs to */
142 volatile u_int refcount;
146 * Record/packet header in first mbuf of chain; valid only if M_PKTHDR is set.
149 * Compile-time assertions in uipc_mbuf.c test these values to ensure that
154 struct m_snd_tag *snd_tag; /* send tag, if any */
155 struct ifnet *rcvif; /* rcv interface */
157 SLIST_HEAD(packet_tags, m_tag) tags; /* list of packet tags */
158 int32_t len; /* total packet length */
160 /* Layer crossing persistent information. */
161 uint32_t flowid; /* packet's 4-tuple system */
162 uint32_t csum_flags; /* checksum and offload features */
163 uint16_t fibnum; /* this packet should use this fib */
164 uint8_t numa_domain; /* NUMA domain of recvd pkt */
165 uint8_t rsstype; /* hash type */
167 uint64_t rcv_tstmp; /* timestamp in ns */
169 uint8_t l2hlen; /* layer 2 hdr len */
170 uint8_t l3hlen; /* layer 3 hdr len */
171 uint8_t l4hlen; /* layer 4 hdr len */
172 uint8_t l5hlen; /* layer 5 hdr len */
179 uint32_t thirtytwo[2];
180 uint64_t sixtyfour[1];
181 uintptr_t unintptr[1];
185 /* Layer specific non-persistent local storage for reassembly, etc. */
189 uint32_t thirtytwo[2];
190 uint64_t sixtyfour[1];
191 uintptr_t unintptr[1];
195 #define ether_vtag PH_per.sixteen[0]
197 #define vt_nrecs sixteen[0]
198 #define tso_segsz PH_per.sixteen[1]
199 #define lro_nsegs tso_segsz
200 #define csum_phsum PH_per.sixteen[2]
201 #define csum_data PH_per.thirtytwo[1]
202 #define lro_len PH_per.sixteen[0] /* inbound during LRO */
203 #define lro_csum PH_per.sixteen[1] /* inbound during LRO */
204 #define pace_thoff PH_loc.sixteen[0]
205 #define pace_tlen PH_loc.sixteen[1]
206 #define pace_drphdrlen PH_loc.sixteen[2]
207 #define pace_tos PH_loc.eight[6]
208 #define pace_lock PH_loc.eight[7]
211 * Description of external storage mapped into mbuf; valid only if M_EXT is
215 * Compile-time assertions in uipc_mbuf.c test these values to ensure that
218 typedef void m_ext_free_t(struct mbuf *);
222 * If EXT_FLAG_EMBREF is set, then we use refcount in the
223 * mbuf, the 'ext_count' member. Otherwise, we have a
224 * shadow copy and we use pointer 'ext_cnt'. The original
225 * mbuf is responsible to carry the pointer to free routine
226 * and its arguments. They aren't copied into shadows in
227 * mb_dupcl() to avoid dereferencing next cachelines.
229 volatile u_int ext_count;
230 volatile u_int *ext_cnt;
234 * If ext_type == EXT_PGS, 'ext_pgs' points to a
235 * structure describing the buffer. Otherwise,
236 * 'ext_buf' points to the start of the buffer.
238 struct mbuf_ext_pgs *ext_pgs;
241 uint32_t ext_size; /* size of buffer, for ext_free */
242 uint32_t ext_type:8, /* type of external storage */
243 ext_flags:24; /* external storage mbuf flags */
245 * Fields below store the free context for the external storage.
246 * They are valid only in the refcount carrying mbuf, the one with
247 * EXT_FLAG_EMBREF flag, with exclusion for EXT_EXTREF type, where
248 * the free context is copied into all mbufs that use same external
251 #define m_ext_copylen offsetof(struct m_ext, ext_free)
252 m_ext_free_t *ext_free; /* free routine if not the usual */
253 void *ext_arg1; /* optional argument pointer */
254 void *ext_arg2; /* optional argument pointer */
258 * The core of the mbuf object along with some shortcut defines for practical
263 * Header present at the beginning of every mbuf.
266 * Compile-time assertions in uipc_mbuf.c test these values to ensure
267 * that they are correct.
269 union { /* next buffer in chain */
271 SLIST_ENTRY(mbuf) m_slist;
272 STAILQ_ENTRY(mbuf) m_stailq;
274 union { /* next chain in queue/record */
275 struct mbuf *m_nextpkt;
276 SLIST_ENTRY(mbuf) m_slistpkt;
277 STAILQ_ENTRY(mbuf) m_stailqpkt;
279 caddr_t m_data; /* location of data */
280 int32_t m_len; /* amount of data in this mbuf */
281 uint32_t m_type:8, /* type of data in this mbuf */
282 m_flags:24; /* flags; see below */
283 #if !defined(__LP64__)
284 uint32_t m_pad; /* pad for 64bit alignment */
288 * A set of optional headers (packet header, external storage header)
289 * and internal data storage. Historically, these arrays were sized
290 * to MHLEN (space left after a packet header) and MLEN (space left
291 * after only a regular mbuf header); they are now variable size in
292 * order to support future work on variable-size mbufs.
296 struct pkthdr m_pkthdr; /* M_PKTHDR set */
298 struct m_ext m_ext; /* M_EXT set */
302 char m_dat[0]; /* !M_PKTHDR, !M_EXT */
310 * TLS records for TLS 1.0-1.2 can have the following header lengths:
311 * - 5 (AES-CBC with implicit IV)
312 * - 21 (AES-CBC with explicit IV)
313 * - 13 (AES-GCM with 8 byte explicit IV)
315 #define MBUF_PEXT_HDR_LEN 24
318 * TLS records for TLS 1.0-1.2 can have the following maximum trailer
321 * - 36 (AES-CBC with SHA1 and up to 16 bytes of padding)
322 * - 48 (AES-CBC with SHA2-256 and up to 16 bytes of padding)
323 * - 64 (AES-CBC with SHA2-384 and up to 16 bytes of padding)
325 #define MBUF_PEXT_TRAIL_LEN 64
328 #define MBUF_PEXT_MAX_PGS (152 / sizeof(vm_paddr_t))
330 #define MBUF_PEXT_MAX_PGS (156 / sizeof(vm_paddr_t))
333 #define MBUF_PEXT_MAX_BYTES \
334 (MBUF_PEXT_MAX_PGS * PAGE_SIZE + MBUF_PEXT_HDR_LEN + MBUF_PEXT_TRAIL_LEN)
337 * This struct is 256 bytes in size and is arranged so that the most
338 * common case (accessing the first 4 pages of a 16KB TLS record) will
339 * fit in a single 64 byte cacheline.
341 struct mbuf_ext_pgs {
342 uint8_t npgs; /* Number of attached pages */
343 uint8_t nrdy; /* Pages with I/O pending */
344 uint8_t hdr_len; /* TLS header length */
345 uint8_t trail_len; /* TLS trailer length */
346 uint16_t first_pg_off; /* Offset into 1st page */
347 uint16_t last_pg_len; /* Length of last page */
348 vm_paddr_t pa[MBUF_PEXT_MAX_PGS]; /* phys addrs of pages */
349 char hdr[MBUF_PEXT_HDR_LEN]; /* TLS header */
350 struct ktls_session *tls; /* TLS session */
351 #if defined(__i386__) || \
352 (defined(__powerpc__) && !defined(__powerpc64__) && defined(BOOKE))
354 * i386 and Book-E PowerPC have 64-bit vm_paddr_t, so there is
355 * a 4 byte remainder from the space allocated for pa[].
360 char trail[MBUF_PEXT_TRAIL_LEN]; /* TLS trailer */
365 STAILQ_ENTRY(mbuf_ext_pgs) stailq;
373 mbuf_ext_pg_len(struct mbuf_ext_pgs *ext_pgs, int pidx, int pgoff)
375 KASSERT(pgoff == 0 || pidx == 0,
376 ("page %d with non-zero offset %d in %p", pidx, pgoff, ext_pgs));
377 if (pidx == ext_pgs->npgs - 1) {
378 return (ext_pgs->last_pg_len);
380 return (PAGE_SIZE - pgoff);
384 #ifdef INVARIANT_SUPPORT
385 void mb_ext_pgs_check(struct mbuf_ext_pgs *ext_pgs);
388 #define MBUF_EXT_PGS_ASSERT_SANITY(ext_pgs) mb_ext_pgs_check((ext_pgs))
390 #define MBUF_EXT_PGS_ASSERT_SANITY(ext_pgs)
395 * mbuf flags of global significance and layer crossing.
396 * Those of only protocol/layer specific significance are to be mapped
397 * to M_PROTO[1-11] and cleared at layer handoff boundaries.
398 * NB: Limited to the lower 24 bits.
400 #define M_EXT 0x00000001 /* has associated external storage */
401 #define M_PKTHDR 0x00000002 /* start of record */
402 #define M_EOR 0x00000004 /* end of record */
403 #define M_RDONLY 0x00000008 /* associated data is marked read-only */
404 #define M_BCAST 0x00000010 /* send/received as link-level broadcast */
405 #define M_MCAST 0x00000020 /* send/received as link-level multicast */
406 #define M_PROMISC 0x00000040 /* packet was not for us */
407 #define M_VLANTAG 0x00000080 /* ether_vtag is valid */
408 #define M_NOMAP 0x00000100 /* mbuf data is unmapped */
409 #define M_NOFREE 0x00000200 /* do not free mbuf, embedded in cluster */
410 #define M_TSTMP 0x00000400 /* rcv_tstmp field is valid */
411 #define M_TSTMP_HPREC 0x00000800 /* rcv_tstmp is high-prec, typically
412 hw-stamped on port (useful for IEEE 1588
414 #define M_TSTMP_LRO 0x00001000 /* Time LRO pushed in pkt is valid in (PH_loc) */
416 #define M_PROTO1 0x00002000 /* protocol-specific */
417 #define M_PROTO2 0x00004000 /* protocol-specific */
418 #define M_PROTO3 0x00008000 /* protocol-specific */
419 #define M_PROTO4 0x00010000 /* protocol-specific */
420 #define M_PROTO5 0x00020000 /* protocol-specific */
421 #define M_PROTO6 0x00040000 /* protocol-specific */
422 #define M_PROTO7 0x00080000 /* protocol-specific */
423 #define M_PROTO8 0x00100000 /* protocol-specific */
424 #define M_PROTO9 0x00200000 /* protocol-specific */
425 #define M_PROTO10 0x00400000 /* protocol-specific */
426 #define M_PROTO11 0x00800000 /* protocol-specific */
428 #define MB_DTOR_SKIP 0x1 /* don't pollute the cache by touching a freed mbuf */
431 * Flags to purge when crossing layers.
433 #define M_PROTOFLAGS \
434 (M_PROTO1|M_PROTO2|M_PROTO3|M_PROTO4|M_PROTO5|M_PROTO6|M_PROTO7|M_PROTO8|\
435 M_PROTO9|M_PROTO10|M_PROTO11)
438 * Flags preserved when copying m_pkthdr.
440 #define M_COPYFLAGS \
441 (M_PKTHDR|M_EOR|M_RDONLY|M_BCAST|M_MCAST|M_PROMISC|M_VLANTAG|M_TSTMP| \
442 M_TSTMP_HPREC|M_TSTMP_LRO|M_PROTOFLAGS)
445 * Mbuf flag description for use with printf(9) %b identifier.
447 #define M_FLAG_BITS \
448 "\20\1M_EXT\2M_PKTHDR\3M_EOR\4M_RDONLY\5M_BCAST\6M_MCAST" \
449 "\7M_PROMISC\10M_VLANTAG\11M_NOMAP\12M_NOFREE\13M_TSTMP\14M_TSTMP_HPREC\15M_TSTMP_LRO"
450 #define M_FLAG_PROTOBITS \
451 "\16M_PROTO1\17M_PROTO2\20M_PROTO3\21M_PROTO4" \
452 "\22M_PROTO5\23M_PROTO6\24M_PROTO7\25M_PROTO8\26M_PROTO9" \
453 "\27M_PROTO10\28M_PROTO11"
454 #define M_FLAG_PRINTF (M_FLAG_BITS M_FLAG_PROTOBITS)
457 * Network interface cards are able to hash protocol fields (such as IPv4
458 * addresses and TCP port numbers) classify packets into flows. These flows
459 * can then be used to maintain ordering while delivering packets to the OS
460 * via parallel input queues, as well as to provide a stateless affinity
461 * model. NIC drivers can pass up the hash via m->m_pkthdr.flowid, and set
462 * m_flag fields to indicate how the hash should be interpreted by the
465 * Most NICs support RSS, which provides ordering and explicit affinity, and
466 * use the hash m_flag bits to indicate what header fields were covered by
467 * the hash. M_HASHTYPE_OPAQUE and M_HASHTYPE_OPAQUE_HASH can be set by non-
468 * RSS cards or configurations that provide an opaque flow identifier, allowing
469 * for ordering and distribution without explicit affinity. Additionally,
470 * M_HASHTYPE_OPAQUE_HASH indicates that the flow identifier has hash
473 * The meaning of the IPV6_EX suffix:
474 * "o Home address from the home address option in the IPv6 destination
475 * options header. If the extension header is not present, use the Source
477 * o IPv6 address that is contained in the Routing-Header-Type-2 from the
478 * associated extension header. If the extension header is not present,
479 * use the Destination IPv6 Address."
481 * https://docs.microsoft.com/en-us/windows-hardware/drivers/network/rss-hashing-types#ndishashipv6ex
483 #define M_HASHTYPE_HASHPROP 0x80 /* has hash properties */
484 #define M_HASHTYPE_HASH(t) (M_HASHTYPE_HASHPROP | (t))
485 /* Microsoft RSS standard hash types */
486 #define M_HASHTYPE_NONE 0
487 #define M_HASHTYPE_RSS_IPV4 M_HASHTYPE_HASH(1) /* IPv4 2-tuple */
488 #define M_HASHTYPE_RSS_TCP_IPV4 M_HASHTYPE_HASH(2) /* TCPv4 4-tuple */
489 #define M_HASHTYPE_RSS_IPV6 M_HASHTYPE_HASH(3) /* IPv6 2-tuple */
490 #define M_HASHTYPE_RSS_TCP_IPV6 M_HASHTYPE_HASH(4) /* TCPv6 4-tuple */
491 #define M_HASHTYPE_RSS_IPV6_EX M_HASHTYPE_HASH(5) /* IPv6 2-tuple +
493 #define M_HASHTYPE_RSS_TCP_IPV6_EX M_HASHTYPE_HASH(6) /* TCPv6 4-tuple +
495 #define M_HASHTYPE_RSS_UDP_IPV4 M_HASHTYPE_HASH(7) /* IPv4 UDP 4-tuple*/
496 #define M_HASHTYPE_RSS_UDP_IPV6 M_HASHTYPE_HASH(9) /* IPv6 UDP 4-tuple*/
497 #define M_HASHTYPE_RSS_UDP_IPV6_EX M_HASHTYPE_HASH(10)/* IPv6 UDP 4-tuple +
500 #define M_HASHTYPE_OPAQUE 63 /* ordering, not affinity */
501 #define M_HASHTYPE_OPAQUE_HASH M_HASHTYPE_HASH(M_HASHTYPE_OPAQUE)
502 /* ordering+hash, not affinity*/
504 #define M_HASHTYPE_CLEAR(m) ((m)->m_pkthdr.rsstype = 0)
505 #define M_HASHTYPE_GET(m) ((m)->m_pkthdr.rsstype)
506 #define M_HASHTYPE_SET(m, v) ((m)->m_pkthdr.rsstype = (v))
507 #define M_HASHTYPE_TEST(m, v) (M_HASHTYPE_GET(m) == (v))
508 #define M_HASHTYPE_ISHASH(m) (M_HASHTYPE_GET(m) & M_HASHTYPE_HASHPROP)
511 * External mbuf storage buffer types.
513 #define EXT_CLUSTER 1 /* mbuf cluster */
514 #define EXT_SFBUF 2 /* sendfile(2)'s sf_buf */
515 #define EXT_JUMBOP 3 /* jumbo cluster page sized */
516 #define EXT_JUMBO9 4 /* jumbo cluster 9216 bytes */
517 #define EXT_JUMBO16 5 /* jumbo cluster 16184 bytes */
518 #define EXT_PACKET 6 /* mbuf+cluster from packet zone */
519 #define EXT_MBUF 7 /* external mbuf reference */
520 #define EXT_RXRING 8 /* data in NIC receive ring */
521 #define EXT_PGS 9 /* array of unmapped pages */
523 #define EXT_VENDOR1 224 /* for vendor-internal use */
524 #define EXT_VENDOR2 225 /* for vendor-internal use */
525 #define EXT_VENDOR3 226 /* for vendor-internal use */
526 #define EXT_VENDOR4 227 /* for vendor-internal use */
528 #define EXT_EXP1 244 /* for experimental use */
529 #define EXT_EXP2 245 /* for experimental use */
530 #define EXT_EXP3 246 /* for experimental use */
531 #define EXT_EXP4 247 /* for experimental use */
533 #define EXT_NET_DRV 252 /* custom ext_buf provided by net driver(s) */
534 #define EXT_MOD_TYPE 253 /* custom module's ext_buf type */
535 #define EXT_DISPOSABLE 254 /* can throw this buffer away w/page flipping */
536 #define EXT_EXTREF 255 /* has externally maintained ext_cnt ptr */
539 * Flags for external mbuf buffer types.
540 * NB: limited to the lower 24 bits.
542 #define EXT_FLAG_EMBREF 0x000001 /* embedded ext_count */
543 #define EXT_FLAG_EXTREF 0x000002 /* external ext_cnt, notyet */
545 #define EXT_FLAG_NOFREE 0x000010 /* don't free mbuf to pool, notyet */
547 #define EXT_FLAG_VENDOR1 0x010000 /* These flags are vendor */
548 #define EXT_FLAG_VENDOR2 0x020000 /* or submodule specific, */
549 #define EXT_FLAG_VENDOR3 0x040000 /* not used by mbuf code. */
550 #define EXT_FLAG_VENDOR4 0x080000 /* Set/read by submodule. */
552 #define EXT_FLAG_EXP1 0x100000 /* for experimental use */
553 #define EXT_FLAG_EXP2 0x200000 /* for experimental use */
554 #define EXT_FLAG_EXP3 0x400000 /* for experimental use */
555 #define EXT_FLAG_EXP4 0x800000 /* for experimental use */
558 * EXT flag description for use with printf(9) %b identifier.
560 #define EXT_FLAG_BITS \
561 "\20\1EXT_FLAG_EMBREF\2EXT_FLAG_EXTREF\5EXT_FLAG_NOFREE" \
562 "\21EXT_FLAG_VENDOR1\22EXT_FLAG_VENDOR2\23EXT_FLAG_VENDOR3" \
563 "\24EXT_FLAG_VENDOR4\25EXT_FLAG_EXP1\26EXT_FLAG_EXP2\27EXT_FLAG_EXP3" \
566 #define MBUF_EXT_PGS_ASSERT(m) \
567 KASSERT((((m)->m_flags & M_EXT) != 0) && \
568 ((m)->m_ext.ext_type == EXT_PGS), \
569 ("%s: m %p !M_EXT or !EXT_PGS", __func__, m))
572 * Flags indicating checksum, segmentation and other offload work to be
573 * done, or already done, by hardware or lower layers. It is split into
574 * separate inbound and outbound flags.
576 * Outbound flags that are set by upper protocol layers requesting lower
577 * layers, or ideally the hardware, to perform these offloading tasks.
578 * For outbound packets this field and its flags can be directly tested
579 * against ifnet if_hwassist.
581 #define CSUM_IP 0x00000001 /* IP header checksum offload */
582 #define CSUM_IP_UDP 0x00000002 /* UDP checksum offload */
583 #define CSUM_IP_TCP 0x00000004 /* TCP checksum offload */
584 #define CSUM_IP_SCTP 0x00000008 /* SCTP checksum offload */
585 #define CSUM_IP_TSO 0x00000010 /* TCP segmentation offload */
586 #define CSUM_IP_ISCSI 0x00000020 /* iSCSI checksum offload */
588 #define CSUM_IP6_UDP 0x00000200 /* UDP checksum offload */
589 #define CSUM_IP6_TCP 0x00000400 /* TCP checksum offload */
590 #define CSUM_IP6_SCTP 0x00000800 /* SCTP checksum offload */
591 #define CSUM_IP6_TSO 0x00001000 /* TCP segmentation offload */
592 #define CSUM_IP6_ISCSI 0x00002000 /* iSCSI checksum offload */
594 /* Inbound checksum support where the checksum was verified by hardware. */
595 #define CSUM_L3_CALC 0x01000000 /* calculated layer 3 csum */
596 #define CSUM_L3_VALID 0x02000000 /* checksum is correct */
597 #define CSUM_L4_CALC 0x04000000 /* calculated layer 4 csum */
598 #define CSUM_L4_VALID 0x08000000 /* checksum is correct */
599 #define CSUM_L5_CALC 0x10000000 /* calculated layer 5 csum */
600 #define CSUM_L5_VALID 0x20000000 /* checksum is correct */
601 #define CSUM_COALESCED 0x40000000 /* contains merged segments */
603 #define CSUM_SND_TAG 0x80000000 /* Packet header has send tag */
606 * CSUM flag description for use with printf(9) %b identifier.
609 "\20\1CSUM_IP\2CSUM_IP_UDP\3CSUM_IP_TCP\4CSUM_IP_SCTP\5CSUM_IP_TSO" \
611 "\12CSUM_IP6_UDP\13CSUM_IP6_TCP\14CSUM_IP6_SCTP\15CSUM_IP6_TSO" \
612 "\16CSUM_IP6_ISCSI" \
613 "\31CSUM_L3_CALC\32CSUM_L3_VALID\33CSUM_L4_CALC\34CSUM_L4_VALID" \
614 "\35CSUM_L5_CALC\36CSUM_L5_VALID\37CSUM_COALESCED\40CSUM_SND_TAG"
616 /* CSUM flags compatibility mappings. */
617 #define CSUM_IP_CHECKED CSUM_L3_CALC
618 #define CSUM_IP_VALID CSUM_L3_VALID
619 #define CSUM_DATA_VALID CSUM_L4_VALID
620 #define CSUM_PSEUDO_HDR CSUM_L4_CALC
621 #define CSUM_SCTP_VALID CSUM_L4_VALID
622 #define CSUM_DELAY_DATA (CSUM_TCP|CSUM_UDP)
623 #define CSUM_DELAY_IP CSUM_IP /* Only v4, no v6 IP hdr csum */
624 #define CSUM_DELAY_DATA_IPV6 (CSUM_TCP_IPV6|CSUM_UDP_IPV6)
625 #define CSUM_DATA_VALID_IPV6 CSUM_DATA_VALID
626 #define CSUM_TCP CSUM_IP_TCP
627 #define CSUM_UDP CSUM_IP_UDP
628 #define CSUM_SCTP CSUM_IP_SCTP
629 #define CSUM_TSO (CSUM_IP_TSO|CSUM_IP6_TSO)
630 #define CSUM_UDP_IPV6 CSUM_IP6_UDP
631 #define CSUM_TCP_IPV6 CSUM_IP6_TCP
632 #define CSUM_SCTP_IPV6 CSUM_IP6_SCTP
635 * mbuf types describing the content of the mbuf (including external storage).
637 #define MT_NOTMBUF 0 /* USED INTERNALLY ONLY! Object is not mbuf */
638 #define MT_DATA 1 /* dynamic (data) allocation */
639 #define MT_HEADER MT_DATA /* packet header, use M_PKTHDR instead */
641 #define MT_VENDOR1 4 /* for vendor-internal use */
642 #define MT_VENDOR2 5 /* for vendor-internal use */
643 #define MT_VENDOR3 6 /* for vendor-internal use */
644 #define MT_VENDOR4 7 /* for vendor-internal use */
646 #define MT_SONAME 8 /* socket name */
648 #define MT_EXP1 9 /* for experimental use */
649 #define MT_EXP2 10 /* for experimental use */
650 #define MT_EXP3 11 /* for experimental use */
651 #define MT_EXP4 12 /* for experimental use */
653 #define MT_CONTROL 14 /* extra-data protocol message */
654 #define MT_EXTCONTROL 15 /* control message with externalized contents */
655 #define MT_OOBDATA 16 /* expedited data */
657 #define MT_NOINIT 255 /* Not a type but a flag to allocate
658 a non-initialized mbuf */
661 * String names of mbuf-related UMA(9) and malloc(9) types. Exposed to
662 * !_KERNEL so that monitoring tools can look up the zones with
665 #define MBUF_MEM_NAME "mbuf"
666 #define MBUF_CLUSTER_MEM_NAME "mbuf_cluster"
667 #define MBUF_PACKET_MEM_NAME "mbuf_packet"
668 #define MBUF_JUMBOP_MEM_NAME "mbuf_jumbo_page"
669 #define MBUF_JUMBO9_MEM_NAME "mbuf_jumbo_9k"
670 #define MBUF_JUMBO16_MEM_NAME "mbuf_jumbo_16k"
671 #define MBUF_TAG_MEM_NAME "mbuf_tag"
672 #define MBUF_EXTREFCNT_MEM_NAME "mbuf_ext_refcnt"
673 #define MBUF_EXTPGS_MEM_NAME "mbuf_extpgs"
678 #define MBUF_CHECKSLEEP(how) do { \
679 if (how == M_WAITOK) \
680 WITNESS_WARN(WARN_GIANTOK | WARN_SLEEPOK, NULL, \
681 "Sleeping in \"%s\"", __func__); \
684 #define MBUF_CHECKSLEEP(how)
688 * Network buffer allocation API
690 * The rest of it is defined in kern/kern_mbuf.c
692 extern uma_zone_t zone_mbuf;
693 extern uma_zone_t zone_clust;
694 extern uma_zone_t zone_pack;
695 extern uma_zone_t zone_jumbop;
696 extern uma_zone_t zone_jumbo9;
697 extern uma_zone_t zone_jumbo16;
698 extern uma_zone_t zone_extpgs;
700 void mb_dupcl(struct mbuf *, struct mbuf *);
701 void mb_free_ext(struct mbuf *);
702 void mb_free_mext_pgs(struct mbuf *);
703 struct mbuf *mb_alloc_ext_pgs(int, bool, m_ext_free_t);
704 int mb_unmapped_compress(struct mbuf *m);
705 struct mbuf *mb_unmapped_to_ext(struct mbuf *m);
706 void mb_free_notready(struct mbuf *m, int count);
707 void m_adj(struct mbuf *, int);
708 int m_apply(struct mbuf *, int, int,
709 int (*)(void *, void *, u_int), void *);
710 int m_append(struct mbuf *, int, c_caddr_t);
711 void m_cat(struct mbuf *, struct mbuf *);
712 void m_catpkt(struct mbuf *, struct mbuf *);
713 int m_clget(struct mbuf *m, int how);
714 void *m_cljget(struct mbuf *m, int how, int size);
715 struct mbuf *m_collapse(struct mbuf *, int, int);
716 void m_copyback(struct mbuf *, int, int, c_caddr_t);
717 void m_copydata(const struct mbuf *, int, int, caddr_t);
718 struct mbuf *m_copym(struct mbuf *, int, int, int);
719 struct mbuf *m_copypacket(struct mbuf *, int);
720 void m_copy_pkthdr(struct mbuf *, struct mbuf *);
721 struct mbuf *m_copyup(struct mbuf *, int, int);
722 struct mbuf *m_defrag(struct mbuf *, int);
723 void m_demote_pkthdr(struct mbuf *);
724 void m_demote(struct mbuf *, int, int);
725 struct mbuf *m_devget(char *, int, int, struct ifnet *,
726 void (*)(char *, caddr_t, u_int));
727 void m_dispose_extcontrolm(struct mbuf *m);
728 struct mbuf *m_dup(const struct mbuf *, int);
729 int m_dup_pkthdr(struct mbuf *, const struct mbuf *, int);
730 void m_extadd(struct mbuf *, char *, u_int, m_ext_free_t,
731 void *, void *, int, int);
732 u_int m_fixhdr(struct mbuf *);
733 struct mbuf *m_fragment(struct mbuf *, int, int);
734 void m_freem(struct mbuf *);
735 struct mbuf *m_get2(int, int, short, int);
736 struct mbuf *m_getjcl(int, short, int, int);
737 struct mbuf *m_getm2(struct mbuf *, int, int, short, int);
738 struct mbuf *m_getptr(struct mbuf *, int, int *);
739 u_int m_length(struct mbuf *, struct mbuf **);
740 int m_mbuftouio(struct uio *, const struct mbuf *, int);
741 int m_unmappedtouio(const struct mbuf *, int, struct uio *, int);
742 void m_move_pkthdr(struct mbuf *, struct mbuf *);
743 int m_pkthdr_init(struct mbuf *, int);
744 struct mbuf *m_prepend(struct mbuf *, int, int);
745 void m_print(const struct mbuf *, int);
746 struct mbuf *m_pulldown(struct mbuf *, int, int, int *);
747 struct mbuf *m_pullup(struct mbuf *, int);
748 int m_sanity(struct mbuf *, int);
749 struct mbuf *m_split(struct mbuf *, int, int);
750 struct mbuf *m_uiotombuf(struct uio *, int, int, int, int);
751 struct mbuf *m_unshare(struct mbuf *, int);
752 void m_snd_tag_init(struct m_snd_tag *, struct ifnet *);
753 void m_snd_tag_destroy(struct m_snd_tag *);
767 #if MJUMPAGESIZE != MCLBYTES
779 panic("%s: invalid cluster size %d", __func__, size);
786 * Associated an external reference counted buffer with an mbuf.
789 m_extaddref(struct mbuf *m, char *buf, u_int size, u_int *ref_cnt,
790 m_ext_free_t freef, void *arg1, void *arg2)
793 KASSERT(ref_cnt != NULL, ("%s: ref_cnt not provided", __func__));
795 atomic_add_int(ref_cnt, 1);
797 m->m_ext.ext_buf = buf;
798 m->m_ext.ext_cnt = ref_cnt;
799 m->m_data = m->m_ext.ext_buf;
800 m->m_ext.ext_size = size;
801 m->m_ext.ext_free = freef;
802 m->m_ext.ext_arg1 = arg1;
803 m->m_ext.ext_arg2 = arg2;
804 m->m_ext.ext_type = EXT_EXTREF;
805 m->m_ext.ext_flags = 0;
808 static __inline uma_zone_t
817 #if MJUMPAGESIZE != MCLBYTES
829 panic("%s: invalid cluster size %d", __func__, size);
836 * Initialize an mbuf with linear storage.
838 * Inline because the consumer text overhead will be roughly the same to
839 * initialize or call a function with this many parameters and M_PKTHDR
840 * should go away with constant propagation for !MGETHDR.
843 m_init(struct mbuf *m, int how, short type, int flags)
849 m->m_data = m->m_dat;
853 if (flags & M_PKTHDR)
854 error = m_pkthdr_init(m, how);
858 MBUF_PROBE5(m__init, m, how, type, flags, error);
862 static __inline struct mbuf *
863 m_get(int how, short type)
870 m = uma_zalloc_arg(zone_mbuf, &args, how);
871 MBUF_PROBE3(m__get, how, type, m);
875 static __inline struct mbuf *
876 m_gethdr(int how, short type)
881 args.flags = M_PKTHDR;
883 m = uma_zalloc_arg(zone_mbuf, &args, how);
884 MBUF_PROBE3(m__gethdr, how, type, m);
888 static __inline struct mbuf *
889 m_getcl(int how, short type, int flags)
896 m = uma_zalloc_arg(zone_pack, &args, how);
897 MBUF_PROBE4(m__getcl, how, type, flags, m);
902 * XXX: m_cljset() is a dangerous API. One must attach only a new,
903 * unreferenced cluster to an mbuf(9). It is not possible to assert
904 * that, so care can be taken only by users of the API.
907 m_cljset(struct mbuf *m, void *cl, int type)
915 #if MJUMPAGESIZE != MCLBYTES
927 panic("%s: unknown cluster type %d", __func__, type);
931 m->m_data = m->m_ext.ext_buf = cl;
932 m->m_ext.ext_free = m->m_ext.ext_arg1 = m->m_ext.ext_arg2 = NULL;
933 m->m_ext.ext_size = size;
934 m->m_ext.ext_type = type;
935 m->m_ext.ext_flags = EXT_FLAG_EMBREF;
936 m->m_ext.ext_count = 1;
938 MBUF_PROBE3(m__cljset, m, cl, type);
942 m_chtype(struct mbuf *m, short new_type)
945 m->m_type = new_type;
949 m_clrprotoflags(struct mbuf *m)
953 m->m_flags &= ~M_PROTOFLAGS;
958 static __inline struct mbuf *
959 m_last(struct mbuf *m)
968 m_extrefcnt(struct mbuf *m)
971 KASSERT(m->m_flags & M_EXT, ("%s: M_EXT missing", __func__));
973 return ((m->m_ext.ext_flags & EXT_FLAG_EMBREF) ? m->m_ext.ext_count :
978 * mbuf, cluster, and external object allocation macros (for compatibility
981 #define M_MOVE_PKTHDR(to, from) m_move_pkthdr((to), (from))
982 #define MGET(m, how, type) ((m) = m_get((how), (type)))
983 #define MGETHDR(m, how, type) ((m) = m_gethdr((how), (type)))
984 #define MCLGET(m, how) m_clget((m), (how))
985 #define MEXTADD(m, buf, size, free, arg1, arg2, flags, type) \
986 m_extadd((m), (char *)(buf), (size), (free), (arg1), (arg2), \
988 #define m_getm(m, len, how, type) \
989 m_getm2((m), (len), (how), (type), M_PKTHDR)
992 * Evaluate TRUE if it's safe to write to the mbuf m's data region (this can
993 * be both the local data payload, or an external buffer area, depending on
994 * whether M_EXT is set).
996 #define M_WRITABLE(m) (((m)->m_flags & (M_RDONLY | M_NOMAP)) == 0 && \
997 (!(((m)->m_flags & M_EXT)) || \
998 (m_extrefcnt(m) == 1)))
1000 /* Check if the supplied mbuf has a packet header, or else panic. */
1001 #define M_ASSERTPKTHDR(m) \
1002 KASSERT((m) != NULL && (m)->m_flags & M_PKTHDR, \
1003 ("%s: no mbuf packet header!", __func__))
1006 * Ensure that the supplied mbuf is a valid, non-free mbuf.
1008 * XXX: Broken at the moment. Need some UMA magic to make it work again.
1010 #define M_ASSERTVALID(m) \
1011 KASSERT((((struct mbuf *)m)->m_flags & 0) == 0, \
1012 ("%s: attempted use of a free mbuf!", __func__))
1015 * Return the address of the start of the buffer associated with an mbuf,
1016 * handling external storage, packet-header mbufs, and regular data mbufs.
1018 #define M_START(m) \
1019 (((m)->m_flags & M_NOMAP) ? NULL : \
1020 ((m)->m_flags & M_EXT) ? (m)->m_ext.ext_buf : \
1021 ((m)->m_flags & M_PKTHDR) ? &(m)->m_pktdat[0] : \
1025 * Return the size of the buffer associated with an mbuf, handling external
1026 * storage, packet-header mbufs, and regular data mbufs.
1029 (((m)->m_flags & M_EXT) ? (m)->m_ext.ext_size : \
1030 ((m)->m_flags & M_PKTHDR) ? MHLEN : \
1034 * Set the m_data pointer of a newly allocated mbuf to place an object of the
1035 * specified size at the end of the mbuf, longword aligned.
1037 * NB: Historically, we had M_ALIGN(), MH_ALIGN(), and MEXT_ALIGN() as
1038 * separate macros, each asserting that it was called at the proper moment.
1039 * This required callers to themselves test the storage type and call the
1040 * right one. Rather than require callers to be aware of those layout
1041 * decisions, we centralize here.
1043 static __inline void
1044 m_align(struct mbuf *m, int len)
1047 const char *msg = "%s: not a virgin mbuf";
1051 KASSERT(m->m_data == M_START(m), (msg, __func__));
1053 adjust = M_SIZE(m) - len;
1054 m->m_data += adjust &~ (sizeof(long)-1);
1057 #define M_ALIGN(m, len) m_align(m, len)
1058 #define MH_ALIGN(m, len) m_align(m, len)
1059 #define MEXT_ALIGN(m, len) m_align(m, len)
1062 * Compute the amount of space available before the current start of data in
1065 * The M_WRITABLE() is a temporary, conservative safety measure: the burden
1066 * of checking writability of the mbuf data area rests solely with the caller.
1068 * NB: In previous versions, M_LEADINGSPACE() would only check M_WRITABLE()
1069 * for mbufs with external storage. We now allow mbuf-embedded data to be
1070 * read-only as well.
1072 #define M_LEADINGSPACE(m) \
1073 (M_WRITABLE(m) ? ((m)->m_data - M_START(m)) : 0)
1076 * Compute the amount of space available after the end of data in an mbuf.
1078 * The M_WRITABLE() is a temporary, conservative safety measure: the burden
1079 * of checking writability of the mbuf data area rests solely with the caller.
1081 * NB: In previous versions, M_TRAILINGSPACE() would only check M_WRITABLE()
1082 * for mbufs with external storage. We now allow mbuf-embedded data to be
1083 * read-only as well.
1085 #define M_TRAILINGSPACE(m) \
1087 ((M_START(m) + M_SIZE(m)) - ((m)->m_data + (m)->m_len)) : 0)
1090 * Arrange to prepend space of size plen to mbuf m. If a new mbuf must be
1091 * allocated, how specifies whether to wait. If the allocation fails, the
1092 * original mbuf chain is freed and m is set to NULL.
1094 #define M_PREPEND(m, plen, how) do { \
1095 struct mbuf **_mmp = &(m); \
1096 struct mbuf *_mm = *_mmp; \
1097 int _mplen = (plen); \
1098 int __mhow = (how); \
1100 MBUF_CHECKSLEEP(how); \
1101 if (M_LEADINGSPACE(_mm) >= _mplen) { \
1102 _mm->m_data -= _mplen; \
1103 _mm->m_len += _mplen; \
1105 _mm = m_prepend(_mm, _mplen, __mhow); \
1106 if (_mm != NULL && _mm->m_flags & M_PKTHDR) \
1107 _mm->m_pkthdr.len += _mplen; \
1112 * Change mbuf to new type. This is a relatively expensive operation and
1113 * should be avoided.
1115 #define MCHTYPE(m, t) m_chtype((m), (t))
1117 /* Return the rcvif of a packet header. */
1118 static __inline struct ifnet *
1119 m_rcvif(struct mbuf *m)
1123 if (m->m_pkthdr.csum_flags & CSUM_SND_TAG)
1125 return (m->m_pkthdr.rcvif);
1128 /* Length to m_copy to copy all. */
1129 #define M_COPYALL 1000000000
1131 extern int max_datalen; /* MHLEN - max_hdr */
1132 extern int max_hdr; /* Largest link + protocol header */
1133 extern int max_linkhdr; /* Largest link-level header */
1134 extern int max_protohdr; /* Largest protocol header */
1135 extern int nmbclusters; /* Maximum number of clusters */
1136 extern bool mb_use_ext_pgs; /* Use ext_pgs for sendfile */
1139 * Network packets may have annotations attached by affixing a list of
1140 * "packet tags" to the pkthdr structure. Packet tags are dynamically
1141 * allocated semi-opaque data structures that have a fixed header
1142 * (struct m_tag) that specifies the size of the memory block and a
1143 * <cookie,type> pair that identifies it. The cookie is a 32-bit unique
1144 * unsigned value used to identify a module or ABI. By convention this value
1145 * is chosen as the date+time that the module is created, expressed as the
1146 * number of seconds since the epoch (e.g., using date -u +'%s'). The type
1147 * value is an ABI/module-specific value that identifies a particular
1148 * annotation and is private to the module. For compatibility with systems
1149 * like OpenBSD that define packet tags w/o an ABI/module cookie, the value
1150 * PACKET_ABI_COMPAT is used to implement m_tag_get and m_tag_find
1151 * compatibility shim functions and several tag types are defined below.
1152 * Users that do not require compatibility should use a private cookie value
1153 * so that packet tag-related definitions can be maintained privately.
1155 * Note that the packet tag returned by m_tag_alloc has the default memory
1156 * alignment implemented by malloc. To reference private data one can use a
1159 * struct m_tag *mtag = m_tag_alloc(...);
1160 * struct foo *p = (struct foo *)(mtag+1);
1162 * if the alignment of struct m_tag is sufficient for referencing members of
1163 * struct foo. Otherwise it is necessary to embed struct m_tag within the
1164 * private data structure to insure proper alignment; e.g.,
1170 * struct foo *p = (struct foo *) m_tag_alloc(...);
1171 * struct m_tag *mtag = &p->tag;
1175 * Persistent tags stay with an mbuf until the mbuf is reclaimed. Otherwise
1176 * tags are expected to ``vanish'' when they pass through a network
1177 * interface. For most interfaces this happens normally as the tags are
1178 * reclaimed when the mbuf is free'd. However in some special cases
1179 * reclaiming must be done manually. An example is packets that pass through
1180 * the loopback interface. Also, one must be careful to do this when
1181 * ``turning around'' packets (e.g., icmp_reflect).
1183 * To mark a tag persistent bit-or this flag in when defining the tag id.
1184 * The tag will then be treated as described above.
1186 #define MTAG_PERSISTENT 0x800
1188 #define PACKET_TAG_NONE 0 /* Nadda */
1190 /* Packet tags for use with PACKET_ABI_COMPAT. */
1191 #define PACKET_TAG_IPSEC_IN_DONE 1 /* IPsec applied, in */
1192 #define PACKET_TAG_IPSEC_OUT_DONE 2 /* IPsec applied, out */
1193 #define PACKET_TAG_IPSEC_IN_CRYPTO_DONE 3 /* NIC IPsec crypto done */
1194 #define PACKET_TAG_IPSEC_OUT_CRYPTO_NEEDED 4 /* NIC IPsec crypto req'ed */
1195 #define PACKET_TAG_IPSEC_IN_COULD_DO_CRYPTO 5 /* NIC notifies IPsec */
1196 #define PACKET_TAG_IPSEC_PENDING_TDB 6 /* Reminder to do IPsec */
1197 #define PACKET_TAG_BRIDGE 7 /* Bridge processing done */
1198 #define PACKET_TAG_GIF 8 /* GIF processing done */
1199 #define PACKET_TAG_GRE 9 /* GRE processing done */
1200 #define PACKET_TAG_IN_PACKET_CHECKSUM 10 /* NIC checksumming done */
1201 #define PACKET_TAG_ENCAP 11 /* Encap. processing */
1202 #define PACKET_TAG_IPSEC_SOCKET 12 /* IPSEC socket ref */
1203 #define PACKET_TAG_IPSEC_HISTORY 13 /* IPSEC history */
1204 #define PACKET_TAG_IPV6_INPUT 14 /* IPV6 input processing */
1205 #define PACKET_TAG_DUMMYNET 15 /* dummynet info */
1206 #define PACKET_TAG_DIVERT 17 /* divert info */
1207 #define PACKET_TAG_IPFORWARD 18 /* ipforward info */
1208 #define PACKET_TAG_MACLABEL (19 | MTAG_PERSISTENT) /* MAC label */
1209 #define PACKET_TAG_PF (21 | MTAG_PERSISTENT) /* PF/ALTQ information */
1210 #define PACKET_TAG_RTSOCKFAM 25 /* rtsock sa family */
1211 #define PACKET_TAG_IPOPTIONS 27 /* Saved IP options */
1212 #define PACKET_TAG_CARP 28 /* CARP info */
1213 #define PACKET_TAG_IPSEC_NAT_T_PORTS 29 /* two uint16_t */
1214 #define PACKET_TAG_ND_OUTGOING 30 /* ND outgoing */
1216 /* Specific cookies and tags. */
1218 /* Packet tag routines. */
1219 struct m_tag *m_tag_alloc(u_int32_t, int, int, int);
1220 void m_tag_delete(struct mbuf *, struct m_tag *);
1221 void m_tag_delete_chain(struct mbuf *, struct m_tag *);
1222 void m_tag_free_default(struct m_tag *);
1223 struct m_tag *m_tag_locate(struct mbuf *, u_int32_t, int, struct m_tag *);
1224 struct m_tag *m_tag_copy(struct m_tag *, int);
1225 int m_tag_copy_chain(struct mbuf *, const struct mbuf *, int);
1226 void m_tag_delete_nonpersistent(struct mbuf *);
1229 * Initialize the list of tags associated with an mbuf.
1231 static __inline void
1232 m_tag_init(struct mbuf *m)
1235 SLIST_INIT(&m->m_pkthdr.tags);
1239 * Set up the contents of a tag. Note that this does not fill in the free
1240 * method; the caller is expected to do that.
1242 * XXX probably should be called m_tag_init, but that was already taken.
1244 static __inline void
1245 m_tag_setup(struct m_tag *t, u_int32_t cookie, int type, int len)
1250 t->m_tag_cookie = cookie;
1254 * Reclaim resources associated with a tag.
1256 static __inline void
1257 m_tag_free(struct m_tag *t)
1260 (*t->m_tag_free)(t);
1264 * Return the first tag associated with an mbuf.
1266 static __inline struct m_tag *
1267 m_tag_first(struct mbuf *m)
1270 return (SLIST_FIRST(&m->m_pkthdr.tags));
1274 * Return the next tag in the list of tags associated with an mbuf.
1276 static __inline struct m_tag *
1277 m_tag_next(struct mbuf *m __unused, struct m_tag *t)
1280 return (SLIST_NEXT(t, m_tag_link));
1284 * Prepend a tag to the list of tags associated with an mbuf.
1286 static __inline void
1287 m_tag_prepend(struct mbuf *m, struct m_tag *t)
1290 SLIST_INSERT_HEAD(&m->m_pkthdr.tags, t, m_tag_link);
1294 * Unlink a tag from the list of tags associated with an mbuf.
1296 static __inline void
1297 m_tag_unlink(struct mbuf *m, struct m_tag *t)
1300 SLIST_REMOVE(&m->m_pkthdr.tags, t, m_tag, m_tag_link);
1303 /* These are for OpenBSD compatibility. */
1304 #define MTAG_ABI_COMPAT 0 /* compatibility ABI */
1306 static __inline struct m_tag *
1307 m_tag_get(int type, int length, int wait)
1309 return (m_tag_alloc(MTAG_ABI_COMPAT, type, length, wait));
1312 static __inline struct m_tag *
1313 m_tag_find(struct mbuf *m, int type, struct m_tag *start)
1315 return (SLIST_EMPTY(&m->m_pkthdr.tags) ? (struct m_tag *)NULL :
1316 m_tag_locate(m, MTAG_ABI_COMPAT, type, start));
1319 static inline struct m_snd_tag *
1320 m_snd_tag_ref(struct m_snd_tag *mst)
1323 refcount_acquire(&mst->refcount);
1328 m_snd_tag_rele(struct m_snd_tag *mst)
1331 if (refcount_release(&mst->refcount))
1332 m_snd_tag_destroy(mst);
1335 static __inline struct mbuf *
1336 m_free(struct mbuf *m)
1338 struct mbuf *n = m->m_next;
1340 MBUF_PROBE1(m__free, m);
1341 if ((m->m_flags & (M_PKTHDR|M_NOFREE)) == (M_PKTHDR|M_NOFREE))
1342 m_tag_delete_chain(m, NULL);
1343 if (m->m_flags & M_PKTHDR && m->m_pkthdr.csum_flags & CSUM_SND_TAG)
1344 m_snd_tag_rele(m->m_pkthdr.snd_tag);
1345 if (m->m_flags & M_EXT)
1347 else if ((m->m_flags & M_NOFREE) == 0)
1348 uma_zfree(zone_mbuf, m);
1353 rt_m_getfib(struct mbuf *m)
1355 KASSERT(m->m_flags & M_PKTHDR , ("Attempt to get FIB from non header mbuf."));
1356 return (m->m_pkthdr.fibnum);
1359 #define M_GETFIB(_m) rt_m_getfib(_m)
1361 #define M_SETFIB(_m, _fib) do { \
1362 KASSERT((_m)->m_flags & M_PKTHDR, ("Attempt to set FIB on non header mbuf.")); \
1363 ((_m)->m_pkthdr.fibnum) = (_fib); \
1366 /* flags passed as first argument for "m_ether_tcpip_hash()" */
1367 #define MBUF_HASHFLAG_L2 (1 << 2)
1368 #define MBUF_HASHFLAG_L3 (1 << 3)
1369 #define MBUF_HASHFLAG_L4 (1 << 4)
1371 /* mbuf hashing helper routines */
1372 uint32_t m_ether_tcpip_hash_init(void);
1373 uint32_t m_ether_tcpip_hash(const uint32_t, const struct mbuf *, const uint32_t);
1375 #ifdef MBUF_PROFILING
1376 void m_profile(struct mbuf *m);
1377 #define M_PROFILE(m) m_profile(m)
1379 #define M_PROFILE(m)
1383 STAILQ_HEAD(, mbuf) mq_head;
1389 mbufq_init(struct mbufq *mq, int maxlen)
1392 STAILQ_INIT(&mq->mq_head);
1393 mq->mq_maxlen = maxlen;
1397 static inline struct mbuf *
1398 mbufq_flush(struct mbufq *mq)
1402 m = STAILQ_FIRST(&mq->mq_head);
1403 STAILQ_INIT(&mq->mq_head);
1409 mbufq_drain(struct mbufq *mq)
1413 n = mbufq_flush(mq);
1414 while ((m = n) != NULL) {
1415 n = STAILQ_NEXT(m, m_stailqpkt);
1420 static inline struct mbuf *
1421 mbufq_first(const struct mbufq *mq)
1424 return (STAILQ_FIRST(&mq->mq_head));
1427 static inline struct mbuf *
1428 mbufq_last(const struct mbufq *mq)
1431 return (STAILQ_LAST(&mq->mq_head, mbuf, m_stailqpkt));
1435 mbufq_full(const struct mbufq *mq)
1438 return (mq->mq_maxlen > 0 && mq->mq_len >= mq->mq_maxlen);
1442 mbufq_len(const struct mbufq *mq)
1445 return (mq->mq_len);
1449 mbufq_enqueue(struct mbufq *mq, struct mbuf *m)
1454 STAILQ_INSERT_TAIL(&mq->mq_head, m, m_stailqpkt);
1459 static inline struct mbuf *
1460 mbufq_dequeue(struct mbufq *mq)
1464 m = STAILQ_FIRST(&mq->mq_head);
1466 STAILQ_REMOVE_HEAD(&mq->mq_head, m_stailqpkt);
1467 m->m_nextpkt = NULL;
1474 mbufq_prepend(struct mbufq *mq, struct mbuf *m)
1477 STAILQ_INSERT_HEAD(&mq->mq_head, m, m_stailqpkt);
1482 * Note: this doesn't enforce the maximum list size for dst.
1485 mbufq_concat(struct mbufq *mq_dst, struct mbufq *mq_src)
1488 mq_dst->mq_len += mq_src->mq_len;
1489 STAILQ_CONCAT(&mq_dst->mq_head, &mq_src->mq_head);
1493 #ifdef _SYS_TIMESPEC_H_
1495 mbuf_tstmp2timespec(struct mbuf *m, struct timespec *ts)
1498 KASSERT((m->m_flags & M_PKTHDR) != 0, ("mbuf %p no M_PKTHDR", m));
1499 KASSERT((m->m_flags & (M_TSTMP|M_TSTMP_LRO)) != 0, ("mbuf %p no M_TSTMP or M_TSTMP_LRO", m));
1500 ts->tv_sec = m->m_pkthdr.rcv_tstmp / 1000000000;
1501 ts->tv_nsec = m->m_pkthdr.rcv_tstmp % 1000000000;
1506 /* Invoked from the netdump client code. */
1507 void netdump_mbuf_drain(void);
1508 void netdump_mbuf_dump(void);
1509 void netdump_mbuf_reinit(int nmbuf, int nclust, int clsize);
1513 mbuf_has_tls_session(struct mbuf *m)
1516 if (m->m_flags & M_NOMAP) {
1517 MBUF_EXT_PGS_ASSERT(m);
1518 if (m->m_ext.ext_pgs->tls != NULL) {
1525 #endif /* _KERNEL */
1526 #endif /* !_SYS_MBUF_H_ */