2 * SPDX-License-Identifier: BSD-3-Clause
4 * Copyright (c) 1982, 1986, 1988, 1993
5 * The Regents of the University of California.
8 * Redistribution and use in source and binary forms, with or without
9 * modification, are permitted provided that the following conditions
11 * 1. Redistributions of source code must retain the above copyright
12 * notice, this list of conditions and the following disclaimer.
13 * 2. Redistributions in binary form must reproduce the above copyright
14 * notice, this list of conditions and the following disclaimer in the
15 * documentation and/or other materials provided with the distribution.
16 * 3. Neither the name of the University nor the names of its contributors
17 * may be used to endorse or promote products derived from this software
18 * without specific prior written permission.
20 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
21 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
22 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
23 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
24 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
25 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
26 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
27 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
28 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
29 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
32 * @(#)mbuf.h 8.5 (Berkeley) 2/19/95
39 /* XXX: These includes suck. Sorry! */
40 #include <sys/queue.h>
42 #include <sys/systm.h>
43 #include <sys/refcount.h>
53 #define MBUF_PROBE1(probe, arg0) \
54 SDT_PROBE1(sdt, , , probe, arg0)
55 #define MBUF_PROBE2(probe, arg0, arg1) \
56 SDT_PROBE2(sdt, , , probe, arg0, arg1)
57 #define MBUF_PROBE3(probe, arg0, arg1, arg2) \
58 SDT_PROBE3(sdt, , , probe, arg0, arg1, arg2)
59 #define MBUF_PROBE4(probe, arg0, arg1, arg2, arg3) \
60 SDT_PROBE4(sdt, , , probe, arg0, arg1, arg2, arg3)
61 #define MBUF_PROBE5(probe, arg0, arg1, arg2, arg3, arg4) \
62 SDT_PROBE5(sdt, , , probe, arg0, arg1, arg2, arg3, arg4)
64 SDT_PROBE_DECLARE(sdt, , , m__init);
65 SDT_PROBE_DECLARE(sdt, , , m__gethdr);
66 SDT_PROBE_DECLARE(sdt, , , m__get);
67 SDT_PROBE_DECLARE(sdt, , , m__getcl);
68 SDT_PROBE_DECLARE(sdt, , , m__clget);
69 SDT_PROBE_DECLARE(sdt, , , m__cljget);
70 SDT_PROBE_DECLARE(sdt, , , m__cljset);
71 SDT_PROBE_DECLARE(sdt, , , m__free);
72 SDT_PROBE_DECLARE(sdt, , , m__freem);
77 * Mbufs are of a single size, MSIZE (sys/param.h), which includes overhead.
78 * An mbuf may add a single "mbuf cluster" of size MCLBYTES (also in
79 * sys/param.h), which has no additional overhead and is used instead of the
80 * internal data area; this is done when at least MINCLSIZE of data must be
81 * stored. Additionally, it is possible to allocate a separate buffer
82 * externally and attach it to the mbuf in a way similar to that of mbuf
85 * NB: These calculation do not take actual compiler-induced alignment and
86 * padding inside the complete struct mbuf into account. Appropriate
87 * attention is required when changing members of struct mbuf.
89 * MLEN is data length in a normal mbuf.
90 * MHLEN is data length in an mbuf with pktheader.
91 * MINCLSIZE is a smallest amount of data that should be put into cluster.
93 * Compile-time assertions in uipc_mbuf.c test these values to ensure that
97 #define MHSIZE offsetof(struct mbuf, m_dat)
98 #define MPKTHSIZE offsetof(struct mbuf, m_pktdat)
99 #define MLEN ((int)(MSIZE - MHSIZE))
100 #define MHLEN ((int)(MSIZE - MPKTHSIZE))
101 #define MINCLSIZE (MHLEN + 1)
106 * Macro for type conversion: convert mbuf pointer to data pointer of correct
109 * mtod(m, t) -- Convert mbuf pointer to data pointer of correct type.
110 * mtodo(m, o) -- Same as above but with offset 'o' into data.
112 #define mtod(m, t) ((t)((m)->m_data))
113 #define mtodo(m, o) ((void *)(((m)->m_data) + (o)))
116 * Argument structure passed to UMA routines during mbuf and packet
120 int flags; /* Flags for mbuf being allocated */
121 short type; /* Type of mbuf being allocated */
126 * Packet tag structure (see below for details).
129 SLIST_ENTRY(m_tag) m_tag_link; /* List of packet tags */
130 u_int16_t m_tag_id; /* Tag ID */
131 u_int16_t m_tag_len; /* Length of data */
132 u_int32_t m_tag_cookie; /* ABI/Module ID */
133 void (*m_tag_free)(struct m_tag *);
137 * Static network interface owned tag.
138 * Allocated through ifp->if_snd_tag_alloc().
141 struct ifnet *ifp; /* network interface tag belongs to */
142 volatile u_int refcount;
146 * Record/packet header in first mbuf of chain; valid only if M_PKTHDR is set.
149 * Compile-time assertions in uipc_mbuf.c test these values to ensure that
154 struct m_snd_tag *snd_tag; /* send tag, if any */
155 struct ifnet *rcvif; /* rcv interface */
157 SLIST_HEAD(packet_tags, m_tag) tags; /* list of packet tags */
158 int32_t len; /* total packet length */
160 /* Layer crossing persistent information. */
161 uint32_t flowid; /* packet's 4-tuple system */
162 uint32_t csum_flags; /* checksum and offload features */
163 uint16_t fibnum; /* this packet should use this fib */
164 uint8_t numa_domain; /* NUMA domain of recvd pkt */
165 uint8_t rsstype; /* hash type */
167 uint64_t rcv_tstmp; /* timestamp in ns */
169 uint8_t l2hlen; /* layer 2 hdr len */
170 uint8_t l3hlen; /* layer 3 hdr len */
171 uint8_t l4hlen; /* layer 4 hdr len */
172 uint8_t l5hlen; /* layer 5 hdr len */
179 uint32_t thirtytwo[2];
180 uint64_t sixtyfour[1];
181 uintptr_t unintptr[1];
185 /* Layer specific non-persistent local storage for reassembly, etc. */
189 uint32_t thirtytwo[2];
190 uint64_t sixtyfour[1];
191 uintptr_t unintptr[1];
195 #define ether_vtag PH_per.sixteen[0]
197 #define vt_nrecs sixteen[0]
198 #define tso_segsz PH_per.sixteen[1]
199 #define lro_nsegs tso_segsz
200 #define csum_phsum PH_per.sixteen[2]
201 #define csum_data PH_per.thirtytwo[1]
202 #define pace_thoff PH_loc.sixteen[0]
203 #define pace_tlen PH_loc.sixteen[1]
204 #define pace_drphdrlen PH_loc.sixteen[2]
205 #define pace_tos PH_loc.eight[6]
206 #define pace_lock PH_loc.eight[7]
209 * Description of external storage mapped into mbuf; valid only if M_EXT is
213 * Compile-time assertions in uipc_mbuf.c test these values to ensure that
216 typedef void m_ext_free_t(struct mbuf *);
220 * If EXT_FLAG_EMBREF is set, then we use refcount in the
221 * mbuf, the 'ext_count' member. Otherwise, we have a
222 * shadow copy and we use pointer 'ext_cnt'. The original
223 * mbuf is responsible to carry the pointer to free routine
224 * and its arguments. They aren't copied into shadows in
225 * mb_dupcl() to avoid dereferencing next cachelines.
227 volatile u_int ext_count;
228 volatile u_int *ext_cnt;
232 * If ext_type == EXT_PGS, 'ext_pgs' points to a
233 * structure describing the buffer. Otherwise,
234 * 'ext_buf' points to the start of the buffer.
236 struct mbuf_ext_pgs *ext_pgs;
239 uint32_t ext_size; /* size of buffer, for ext_free */
240 uint32_t ext_type:8, /* type of external storage */
241 ext_flags:24; /* external storage mbuf flags */
243 * Fields below store the free context for the external storage.
244 * They are valid only in the refcount carrying mbuf, the one with
245 * EXT_FLAG_EMBREF flag, with exclusion for EXT_EXTREF type, where
246 * the free context is copied into all mbufs that use same external
249 #define m_ext_copylen offsetof(struct m_ext, ext_free)
250 m_ext_free_t *ext_free; /* free routine if not the usual */
251 void *ext_arg1; /* optional argument pointer */
252 void *ext_arg2; /* optional argument pointer */
256 * The core of the mbuf object along with some shortcut defines for practical
261 * Header present at the beginning of every mbuf.
264 * Compile-time assertions in uipc_mbuf.c test these values to ensure
265 * that they are correct.
267 union { /* next buffer in chain */
269 SLIST_ENTRY(mbuf) m_slist;
270 STAILQ_ENTRY(mbuf) m_stailq;
272 union { /* next chain in queue/record */
273 struct mbuf *m_nextpkt;
274 SLIST_ENTRY(mbuf) m_slistpkt;
275 STAILQ_ENTRY(mbuf) m_stailqpkt;
277 caddr_t m_data; /* location of data */
278 int32_t m_len; /* amount of data in this mbuf */
279 uint32_t m_type:8, /* type of data in this mbuf */
280 m_flags:24; /* flags; see below */
281 #if !defined(__LP64__)
282 uint32_t m_pad; /* pad for 64bit alignment */
286 * A set of optional headers (packet header, external storage header)
287 * and internal data storage. Historically, these arrays were sized
288 * to MHLEN (space left after a packet header) and MLEN (space left
289 * after only a regular mbuf header); they are now variable size in
290 * order to support future work on variable-size mbufs.
294 struct pkthdr m_pkthdr; /* M_PKTHDR set */
296 struct m_ext m_ext; /* M_EXT set */
300 char m_dat[0]; /* !M_PKTHDR, !M_EXT */
307 * TLS records for TLS 1.0-1.2 can have the following header lengths:
308 * - 5 (AES-CBC with implicit IV)
309 * - 21 (AES-CBC with explicit IV)
310 * - 13 (AES-GCM with 8 byte explicit IV)
312 #define MBUF_PEXT_HDR_LEN 24
315 * TLS records for TLS 1.0-1.2 can have the following maximum trailer
318 * - 36 (AES-CBC with SHA1 and up to 16 bytes of padding)
319 * - 48 (AES-CBC with SHA2-256 and up to 16 bytes of padding)
320 * - 64 (AES-CBC with SHA2-384 and up to 16 bytes of padding)
322 #define MBUF_PEXT_TRAIL_LEN 64
325 #define MBUF_PEXT_MAX_PGS (152 / sizeof(vm_paddr_t))
327 #define MBUF_PEXT_MAX_PGS (156 / sizeof(vm_paddr_t))
330 #define MBUF_PEXT_MAX_BYTES \
331 (MBUF_PEXT_MAX_PGS * PAGE_SIZE + MBUF_PEXT_HDR_LEN + MBUF_PEXT_TRAIL_LEN)
334 * This struct is 256 bytes in size and is arranged so that the most
335 * common case (accessing the first 4 pages of a 16KB TLS record) will
336 * fit in a single 64 byte cacheline.
338 struct mbuf_ext_pgs {
339 uint8_t npgs; /* Number of attached pages */
340 uint8_t nrdy; /* Pages with I/O pending */
341 uint8_t hdr_len; /* TLS header length */
342 uint8_t trail_len; /* TLS trailer length */
343 uint16_t first_pg_off; /* Offset into 1st page */
344 uint16_t last_pg_len; /* Length of last page */
345 vm_paddr_t pa[MBUF_PEXT_MAX_PGS]; /* phys addrs of pages */
346 char hdr[MBUF_PEXT_HDR_LEN]; /* TLS header */
347 void *tls; /* TLS session */
348 #if defined(__i386__) || \
349 (defined(__powerpc__) && !defined(__powerpc64__) && defined(BOOKE))
351 * i386 and Book-E PowerPC have 64-bit vm_paddr_t, so there is
352 * a 4 byte remainder from the space allocated for pa[].
357 char trail[MBUF_PEXT_TRAIL_LEN]; /* TLS trailer */
362 STAILQ_ENTRY(mbuf_ext_pgs) stailq;
369 mbuf_ext_pg_len(struct mbuf_ext_pgs *ext_pgs, int pidx, int pgoff)
371 KASSERT(pgoff == 0 || pidx == 0,
372 ("page %d with non-zero offset %d in %p", pidx, pgoff, ext_pgs));
373 if (pidx == ext_pgs->npgs - 1) {
374 return (ext_pgs->last_pg_len);
376 return (PAGE_SIZE - pgoff);
380 #ifdef INVARIANT_SUPPORT
381 void mb_ext_pgs_check(struct mbuf_ext_pgs *ext_pgs);
384 #define MBUF_EXT_PGS_ASSERT_SANITY(ext_pgs) mb_ext_pgs_check((ext_pgs))
386 #define MBUF_EXT_PGS_ASSERT_SANITY(ext_pgs)
391 * mbuf flags of global significance and layer crossing.
392 * Those of only protocol/layer specific significance are to be mapped
393 * to M_PROTO[1-12] and cleared at layer handoff boundaries.
394 * NB: Limited to the lower 24 bits.
396 #define M_EXT 0x00000001 /* has associated external storage */
397 #define M_PKTHDR 0x00000002 /* start of record */
398 #define M_EOR 0x00000004 /* end of record */
399 #define M_RDONLY 0x00000008 /* associated data is marked read-only */
400 #define M_BCAST 0x00000010 /* send/received as link-level broadcast */
401 #define M_MCAST 0x00000020 /* send/received as link-level multicast */
402 #define M_PROMISC 0x00000040 /* packet was not for us */
403 #define M_VLANTAG 0x00000080 /* ether_vtag is valid */
404 #define M_NOMAP 0x00000100 /* mbuf data is unmapped */
405 #define M_NOFREE 0x00000200 /* do not free mbuf, embedded in cluster */
406 #define M_TSTMP 0x00000400 /* rcv_tstmp field is valid */
407 #define M_TSTMP_HPREC 0x00000800 /* rcv_tstmp is high-prec, typically
408 hw-stamped on port (useful for IEEE 1588
410 #define M_TSTMP_LRO 0x00001000 /* Time LRO pushed in pkt is valid in (PH_loc) */
412 #define M_PROTO1 0x00001000 /* protocol-specific */
413 #define M_PROTO2 0x00002000 /* protocol-specific */
414 #define M_PROTO3 0x00004000 /* protocol-specific */
415 #define M_PROTO4 0x00008000 /* protocol-specific */
416 #define M_PROTO5 0x00010000 /* protocol-specific */
417 #define M_PROTO6 0x00020000 /* protocol-specific */
418 #define M_PROTO7 0x00040000 /* protocol-specific */
419 #define M_PROTO8 0x00080000 /* protocol-specific */
420 #define M_PROTO9 0x00100000 /* protocol-specific */
421 #define M_PROTO10 0x00200000 /* protocol-specific */
422 #define M_PROTO11 0x00400000 /* protocol-specific */
423 #define M_PROTO12 0x00800000 /* protocol-specific */
425 #define MB_DTOR_SKIP 0x1 /* don't pollute the cache by touching a freed mbuf */
428 * Flags to purge when crossing layers.
430 #define M_PROTOFLAGS \
431 (M_PROTO1|M_PROTO2|M_PROTO3|M_PROTO4|M_PROTO5|M_PROTO6|M_PROTO7|M_PROTO8|\
432 M_PROTO9|M_PROTO10|M_PROTO11|M_PROTO12)
435 * Flags preserved when copying m_pkthdr.
437 #define M_COPYFLAGS \
438 (M_PKTHDR|M_EOR|M_RDONLY|M_BCAST|M_MCAST|M_PROMISC|M_VLANTAG|M_TSTMP| \
439 M_TSTMP_HPREC|M_PROTOFLAGS)
442 * Mbuf flag description for use with printf(9) %b identifier.
444 #define M_FLAG_BITS \
445 "\20\1M_EXT\2M_PKTHDR\3M_EOR\4M_RDONLY\5M_BCAST\6M_MCAST" \
446 "\7M_PROMISC\10M_VLANTAG\11M_NOMAP\12M_NOFREE\13M_TSTMP\14M_TSTMP_HPREC"
447 #define M_FLAG_PROTOBITS \
448 "\15M_PROTO1\16M_PROTO2\17M_PROTO3\20M_PROTO4\21M_PROTO5" \
449 "\22M_PROTO6\23M_PROTO7\24M_PROTO8\25M_PROTO9\26M_PROTO10" \
450 "\27M_PROTO11\30M_PROTO12"
451 #define M_FLAG_PRINTF (M_FLAG_BITS M_FLAG_PROTOBITS)
454 * Network interface cards are able to hash protocol fields (such as IPv4
455 * addresses and TCP port numbers) classify packets into flows. These flows
456 * can then be used to maintain ordering while delivering packets to the OS
457 * via parallel input queues, as well as to provide a stateless affinity
458 * model. NIC drivers can pass up the hash via m->m_pkthdr.flowid, and set
459 * m_flag fields to indicate how the hash should be interpreted by the
462 * Most NICs support RSS, which provides ordering and explicit affinity, and
463 * use the hash m_flag bits to indicate what header fields were covered by
464 * the hash. M_HASHTYPE_OPAQUE and M_HASHTYPE_OPAQUE_HASH can be set by non-
465 * RSS cards or configurations that provide an opaque flow identifier, allowing
466 * for ordering and distribution without explicit affinity. Additionally,
467 * M_HASHTYPE_OPAQUE_HASH indicates that the flow identifier has hash
470 * The meaning of the IPV6_EX suffix:
471 * "o Home address from the home address option in the IPv6 destination
472 * options header. If the extension header is not present, use the Source
474 * o IPv6 address that is contained in the Routing-Header-Type-2 from the
475 * associated extension header. If the extension header is not present,
476 * use the Destination IPv6 Address."
478 * https://docs.microsoft.com/en-us/windows-hardware/drivers/network/rss-hashing-types#ndishashipv6ex
480 #define M_HASHTYPE_HASHPROP 0x80 /* has hash properties */
481 #define M_HASHTYPE_HASH(t) (M_HASHTYPE_HASHPROP | (t))
482 /* Microsoft RSS standard hash types */
483 #define M_HASHTYPE_NONE 0
484 #define M_HASHTYPE_RSS_IPV4 M_HASHTYPE_HASH(1) /* IPv4 2-tuple */
485 #define M_HASHTYPE_RSS_TCP_IPV4 M_HASHTYPE_HASH(2) /* TCPv4 4-tuple */
486 #define M_HASHTYPE_RSS_IPV6 M_HASHTYPE_HASH(3) /* IPv6 2-tuple */
487 #define M_HASHTYPE_RSS_TCP_IPV6 M_HASHTYPE_HASH(4) /* TCPv6 4-tuple */
488 #define M_HASHTYPE_RSS_IPV6_EX M_HASHTYPE_HASH(5) /* IPv6 2-tuple +
490 #define M_HASHTYPE_RSS_TCP_IPV6_EX M_HASHTYPE_HASH(6) /* TCPv6 4-tuple +
492 #define M_HASHTYPE_RSS_UDP_IPV4 M_HASHTYPE_HASH(7) /* IPv4 UDP 4-tuple*/
493 #define M_HASHTYPE_RSS_UDP_IPV6 M_HASHTYPE_HASH(9) /* IPv6 UDP 4-tuple*/
494 #define M_HASHTYPE_RSS_UDP_IPV6_EX M_HASHTYPE_HASH(10)/* IPv6 UDP 4-tuple +
497 #define M_HASHTYPE_OPAQUE 63 /* ordering, not affinity */
498 #define M_HASHTYPE_OPAQUE_HASH M_HASHTYPE_HASH(M_HASHTYPE_OPAQUE)
499 /* ordering+hash, not affinity*/
501 #define M_HASHTYPE_CLEAR(m) ((m)->m_pkthdr.rsstype = 0)
502 #define M_HASHTYPE_GET(m) ((m)->m_pkthdr.rsstype)
503 #define M_HASHTYPE_SET(m, v) ((m)->m_pkthdr.rsstype = (v))
504 #define M_HASHTYPE_TEST(m, v) (M_HASHTYPE_GET(m) == (v))
505 #define M_HASHTYPE_ISHASH(m) (M_HASHTYPE_GET(m) & M_HASHTYPE_HASHPROP)
508 * External mbuf storage buffer types.
510 #define EXT_CLUSTER 1 /* mbuf cluster */
511 #define EXT_SFBUF 2 /* sendfile(2)'s sf_buf */
512 #define EXT_JUMBOP 3 /* jumbo cluster page sized */
513 #define EXT_JUMBO9 4 /* jumbo cluster 9216 bytes */
514 #define EXT_JUMBO16 5 /* jumbo cluster 16184 bytes */
515 #define EXT_PACKET 6 /* mbuf+cluster from packet zone */
516 #define EXT_MBUF 7 /* external mbuf reference */
517 #define EXT_RXRING 8 /* data in NIC receive ring */
518 #define EXT_PGS 9 /* array of unmapped pages */
520 #define EXT_VENDOR1 224 /* for vendor-internal use */
521 #define EXT_VENDOR2 225 /* for vendor-internal use */
522 #define EXT_VENDOR3 226 /* for vendor-internal use */
523 #define EXT_VENDOR4 227 /* for vendor-internal use */
525 #define EXT_EXP1 244 /* for experimental use */
526 #define EXT_EXP2 245 /* for experimental use */
527 #define EXT_EXP3 246 /* for experimental use */
528 #define EXT_EXP4 247 /* for experimental use */
530 #define EXT_NET_DRV 252 /* custom ext_buf provided by net driver(s) */
531 #define EXT_MOD_TYPE 253 /* custom module's ext_buf type */
532 #define EXT_DISPOSABLE 254 /* can throw this buffer away w/page flipping */
533 #define EXT_EXTREF 255 /* has externally maintained ext_cnt ptr */
536 * Flags for external mbuf buffer types.
537 * NB: limited to the lower 24 bits.
539 #define EXT_FLAG_EMBREF 0x000001 /* embedded ext_count */
540 #define EXT_FLAG_EXTREF 0x000002 /* external ext_cnt, notyet */
542 #define EXT_FLAG_NOFREE 0x000010 /* don't free mbuf to pool, notyet */
544 #define EXT_FLAG_VENDOR1 0x010000 /* These flags are vendor */
545 #define EXT_FLAG_VENDOR2 0x020000 /* or submodule specific, */
546 #define EXT_FLAG_VENDOR3 0x040000 /* not used by mbuf code. */
547 #define EXT_FLAG_VENDOR4 0x080000 /* Set/read by submodule. */
549 #define EXT_FLAG_EXP1 0x100000 /* for experimental use */
550 #define EXT_FLAG_EXP2 0x200000 /* for experimental use */
551 #define EXT_FLAG_EXP3 0x400000 /* for experimental use */
552 #define EXT_FLAG_EXP4 0x800000 /* for experimental use */
555 * EXT flag description for use with printf(9) %b identifier.
557 #define EXT_FLAG_BITS \
558 "\20\1EXT_FLAG_EMBREF\2EXT_FLAG_EXTREF\5EXT_FLAG_NOFREE" \
559 "\21EXT_FLAG_VENDOR1\22EXT_FLAG_VENDOR2\23EXT_FLAG_VENDOR3" \
560 "\24EXT_FLAG_VENDOR4\25EXT_FLAG_EXP1\26EXT_FLAG_EXP2\27EXT_FLAG_EXP3" \
563 #define MBUF_EXT_PGS_ASSERT(m) \
564 KASSERT((((m)->m_flags & M_EXT) != 0) && \
565 ((m)->m_ext.ext_type == EXT_PGS), \
566 ("%s: m %p !M_EXT or !EXT_PGS", __func__, m))
569 * Flags indicating checksum, segmentation and other offload work to be
570 * done, or already done, by hardware or lower layers. It is split into
571 * separate inbound and outbound flags.
573 * Outbound flags that are set by upper protocol layers requesting lower
574 * layers, or ideally the hardware, to perform these offloading tasks.
575 * For outbound packets this field and its flags can be directly tested
576 * against ifnet if_hwassist.
578 #define CSUM_IP 0x00000001 /* IP header checksum offload */
579 #define CSUM_IP_UDP 0x00000002 /* UDP checksum offload */
580 #define CSUM_IP_TCP 0x00000004 /* TCP checksum offload */
581 #define CSUM_IP_SCTP 0x00000008 /* SCTP checksum offload */
582 #define CSUM_IP_TSO 0x00000010 /* TCP segmentation offload */
583 #define CSUM_IP_ISCSI 0x00000020 /* iSCSI checksum offload */
585 #define CSUM_IP6_UDP 0x00000200 /* UDP checksum offload */
586 #define CSUM_IP6_TCP 0x00000400 /* TCP checksum offload */
587 #define CSUM_IP6_SCTP 0x00000800 /* SCTP checksum offload */
588 #define CSUM_IP6_TSO 0x00001000 /* TCP segmentation offload */
589 #define CSUM_IP6_ISCSI 0x00002000 /* iSCSI checksum offload */
591 /* Inbound checksum support where the checksum was verified by hardware. */
592 #define CSUM_L3_CALC 0x01000000 /* calculated layer 3 csum */
593 #define CSUM_L3_VALID 0x02000000 /* checksum is correct */
594 #define CSUM_L4_CALC 0x04000000 /* calculated layer 4 csum */
595 #define CSUM_L4_VALID 0x08000000 /* checksum is correct */
596 #define CSUM_L5_CALC 0x10000000 /* calculated layer 5 csum */
597 #define CSUM_L5_VALID 0x20000000 /* checksum is correct */
598 #define CSUM_COALESCED 0x40000000 /* contains merged segments */
600 #define CSUM_SND_TAG 0x80000000 /* Packet header has send tag */
603 * CSUM flag description for use with printf(9) %b identifier.
606 "\20\1CSUM_IP\2CSUM_IP_UDP\3CSUM_IP_TCP\4CSUM_IP_SCTP\5CSUM_IP_TSO" \
608 "\12CSUM_IP6_UDP\13CSUM_IP6_TCP\14CSUM_IP6_SCTP\15CSUM_IP6_TSO" \
609 "\16CSUM_IP6_ISCSI" \
610 "\31CSUM_L3_CALC\32CSUM_L3_VALID\33CSUM_L4_CALC\34CSUM_L4_VALID" \
611 "\35CSUM_L5_CALC\36CSUM_L5_VALID\37CSUM_COALESCED\40CSUM_SND_TAG"
613 /* CSUM flags compatibility mappings. */
614 #define CSUM_IP_CHECKED CSUM_L3_CALC
615 #define CSUM_IP_VALID CSUM_L3_VALID
616 #define CSUM_DATA_VALID CSUM_L4_VALID
617 #define CSUM_PSEUDO_HDR CSUM_L4_CALC
618 #define CSUM_SCTP_VALID CSUM_L4_VALID
619 #define CSUM_DELAY_DATA (CSUM_TCP|CSUM_UDP)
620 #define CSUM_DELAY_IP CSUM_IP /* Only v4, no v6 IP hdr csum */
621 #define CSUM_DELAY_DATA_IPV6 (CSUM_TCP_IPV6|CSUM_UDP_IPV6)
622 #define CSUM_DATA_VALID_IPV6 CSUM_DATA_VALID
623 #define CSUM_TCP CSUM_IP_TCP
624 #define CSUM_UDP CSUM_IP_UDP
625 #define CSUM_SCTP CSUM_IP_SCTP
626 #define CSUM_TSO (CSUM_IP_TSO|CSUM_IP6_TSO)
627 #define CSUM_UDP_IPV6 CSUM_IP6_UDP
628 #define CSUM_TCP_IPV6 CSUM_IP6_TCP
629 #define CSUM_SCTP_IPV6 CSUM_IP6_SCTP
632 * mbuf types describing the content of the mbuf (including external storage).
634 #define MT_NOTMBUF 0 /* USED INTERNALLY ONLY! Object is not mbuf */
635 #define MT_DATA 1 /* dynamic (data) allocation */
636 #define MT_HEADER MT_DATA /* packet header, use M_PKTHDR instead */
638 #define MT_VENDOR1 4 /* for vendor-internal use */
639 #define MT_VENDOR2 5 /* for vendor-internal use */
640 #define MT_VENDOR3 6 /* for vendor-internal use */
641 #define MT_VENDOR4 7 /* for vendor-internal use */
643 #define MT_SONAME 8 /* socket name */
645 #define MT_EXP1 9 /* for experimental use */
646 #define MT_EXP2 10 /* for experimental use */
647 #define MT_EXP3 11 /* for experimental use */
648 #define MT_EXP4 12 /* for experimental use */
650 #define MT_CONTROL 14 /* extra-data protocol message */
651 #define MT_EXTCONTROL 15 /* control message with externalized contents */
652 #define MT_OOBDATA 16 /* expedited data */
654 #define MT_NOINIT 255 /* Not a type but a flag to allocate
655 a non-initialized mbuf */
658 * String names of mbuf-related UMA(9) and malloc(9) types. Exposed to
659 * !_KERNEL so that monitoring tools can look up the zones with
662 #define MBUF_MEM_NAME "mbuf"
663 #define MBUF_CLUSTER_MEM_NAME "mbuf_cluster"
664 #define MBUF_PACKET_MEM_NAME "mbuf_packet"
665 #define MBUF_JUMBOP_MEM_NAME "mbuf_jumbo_page"
666 #define MBUF_JUMBO9_MEM_NAME "mbuf_jumbo_9k"
667 #define MBUF_JUMBO16_MEM_NAME "mbuf_jumbo_16k"
668 #define MBUF_TAG_MEM_NAME "mbuf_tag"
669 #define MBUF_EXTREFCNT_MEM_NAME "mbuf_ext_refcnt"
670 #define MBUF_EXTPGS_MEM_NAME "mbuf_extpgs"
675 #define MBUF_CHECKSLEEP(how) do { \
676 if (how == M_WAITOK) \
677 WITNESS_WARN(WARN_GIANTOK | WARN_SLEEPOK, NULL, \
678 "Sleeping in \"%s\"", __func__); \
681 #define MBUF_CHECKSLEEP(how)
685 * Network buffer allocation API
687 * The rest of it is defined in kern/kern_mbuf.c
689 extern uma_zone_t zone_mbuf;
690 extern uma_zone_t zone_clust;
691 extern uma_zone_t zone_pack;
692 extern uma_zone_t zone_jumbop;
693 extern uma_zone_t zone_jumbo9;
694 extern uma_zone_t zone_jumbo16;
695 extern uma_zone_t zone_extpgs;
697 void mb_dupcl(struct mbuf *, struct mbuf *);
698 void mb_free_ext(struct mbuf *);
699 void mb_free_mext_pgs(struct mbuf *);
700 struct mbuf *mb_alloc_ext_pgs(int, bool, m_ext_free_t);
701 int mb_unmapped_compress(struct mbuf *m);
702 struct mbuf *mb_unmapped_to_ext(struct mbuf *m);
703 void mb_free_notready(struct mbuf *m, int count);
704 void m_adj(struct mbuf *, int);
705 int m_apply(struct mbuf *, int, int,
706 int (*)(void *, void *, u_int), void *);
707 int m_append(struct mbuf *, int, c_caddr_t);
708 void m_cat(struct mbuf *, struct mbuf *);
709 void m_catpkt(struct mbuf *, struct mbuf *);
710 int m_clget(struct mbuf *m, int how);
711 void *m_cljget(struct mbuf *m, int how, int size);
712 struct mbuf *m_collapse(struct mbuf *, int, int);
713 void m_copyback(struct mbuf *, int, int, c_caddr_t);
714 void m_copydata(const struct mbuf *, int, int, caddr_t);
715 struct mbuf *m_copym(struct mbuf *, int, int, int);
716 struct mbuf *m_copypacket(struct mbuf *, int);
717 void m_copy_pkthdr(struct mbuf *, struct mbuf *);
718 struct mbuf *m_copyup(struct mbuf *, int, int);
719 struct mbuf *m_defrag(struct mbuf *, int);
720 void m_demote_pkthdr(struct mbuf *);
721 void m_demote(struct mbuf *, int, int);
722 struct mbuf *m_devget(char *, int, int, struct ifnet *,
723 void (*)(char *, caddr_t, u_int));
724 void m_dispose_extcontrolm(struct mbuf *m);
725 struct mbuf *m_dup(const struct mbuf *, int);
726 int m_dup_pkthdr(struct mbuf *, const struct mbuf *, int);
727 void m_extadd(struct mbuf *, char *, u_int, m_ext_free_t,
728 void *, void *, int, int);
729 u_int m_fixhdr(struct mbuf *);
730 struct mbuf *m_fragment(struct mbuf *, int, int);
731 void m_freem(struct mbuf *);
732 struct mbuf *m_get2(int, int, short, int);
733 struct mbuf *m_getjcl(int, short, int, int);
734 struct mbuf *m_getm2(struct mbuf *, int, int, short, int);
735 struct mbuf *m_getptr(struct mbuf *, int, int *);
736 u_int m_length(struct mbuf *, struct mbuf **);
737 int m_mbuftouio(struct uio *, const struct mbuf *, int);
738 int m_unmappedtouio(const struct mbuf *, int, struct uio *, int);
739 void m_move_pkthdr(struct mbuf *, struct mbuf *);
740 int m_pkthdr_init(struct mbuf *, int);
741 struct mbuf *m_prepend(struct mbuf *, int, int);
742 void m_print(const struct mbuf *, int);
743 struct mbuf *m_pulldown(struct mbuf *, int, int, int *);
744 struct mbuf *m_pullup(struct mbuf *, int);
745 int m_sanity(struct mbuf *, int);
746 struct mbuf *m_split(struct mbuf *, int, int);
747 struct mbuf *m_uiotombuf(struct uio *, int, int, int, int);
748 struct mbuf *m_unshare(struct mbuf *, int);
749 void m_snd_tag_init(struct m_snd_tag *, struct ifnet *);
750 void m_snd_tag_destroy(struct m_snd_tag *);
764 #if MJUMPAGESIZE != MCLBYTES
776 panic("%s: invalid cluster size %d", __func__, size);
783 * Associated an external reference counted buffer with an mbuf.
786 m_extaddref(struct mbuf *m, char *buf, u_int size, u_int *ref_cnt,
787 m_ext_free_t freef, void *arg1, void *arg2)
790 KASSERT(ref_cnt != NULL, ("%s: ref_cnt not provided", __func__));
792 atomic_add_int(ref_cnt, 1);
794 m->m_ext.ext_buf = buf;
795 m->m_ext.ext_cnt = ref_cnt;
796 m->m_data = m->m_ext.ext_buf;
797 m->m_ext.ext_size = size;
798 m->m_ext.ext_free = freef;
799 m->m_ext.ext_arg1 = arg1;
800 m->m_ext.ext_arg2 = arg2;
801 m->m_ext.ext_type = EXT_EXTREF;
802 m->m_ext.ext_flags = 0;
805 static __inline uma_zone_t
814 #if MJUMPAGESIZE != MCLBYTES
826 panic("%s: invalid cluster size %d", __func__, size);
833 * Initialize an mbuf with linear storage.
835 * Inline because the consumer text overhead will be roughly the same to
836 * initialize or call a function with this many parameters and M_PKTHDR
837 * should go away with constant propagation for !MGETHDR.
840 m_init(struct mbuf *m, int how, short type, int flags)
846 m->m_data = m->m_dat;
850 if (flags & M_PKTHDR)
851 error = m_pkthdr_init(m, how);
855 MBUF_PROBE5(m__init, m, how, type, flags, error);
859 static __inline struct mbuf *
860 m_get(int how, short type)
867 m = uma_zalloc_arg(zone_mbuf, &args, how);
868 MBUF_PROBE3(m__get, how, type, m);
872 static __inline struct mbuf *
873 m_gethdr(int how, short type)
878 args.flags = M_PKTHDR;
880 m = uma_zalloc_arg(zone_mbuf, &args, how);
881 MBUF_PROBE3(m__gethdr, how, type, m);
885 static __inline struct mbuf *
886 m_getcl(int how, short type, int flags)
893 m = uma_zalloc_arg(zone_pack, &args, how);
894 MBUF_PROBE4(m__getcl, how, type, flags, m);
899 * XXX: m_cljset() is a dangerous API. One must attach only a new,
900 * unreferenced cluster to an mbuf(9). It is not possible to assert
901 * that, so care can be taken only by users of the API.
904 m_cljset(struct mbuf *m, void *cl, int type)
912 #if MJUMPAGESIZE != MCLBYTES
924 panic("%s: unknown cluster type %d", __func__, type);
928 m->m_data = m->m_ext.ext_buf = cl;
929 m->m_ext.ext_free = m->m_ext.ext_arg1 = m->m_ext.ext_arg2 = NULL;
930 m->m_ext.ext_size = size;
931 m->m_ext.ext_type = type;
932 m->m_ext.ext_flags = EXT_FLAG_EMBREF;
933 m->m_ext.ext_count = 1;
935 MBUF_PROBE3(m__cljset, m, cl, type);
939 m_chtype(struct mbuf *m, short new_type)
942 m->m_type = new_type;
946 m_clrprotoflags(struct mbuf *m)
950 m->m_flags &= ~M_PROTOFLAGS;
955 static __inline struct mbuf *
956 m_last(struct mbuf *m)
965 m_extrefcnt(struct mbuf *m)
968 KASSERT(m->m_flags & M_EXT, ("%s: M_EXT missing", __func__));
970 return ((m->m_ext.ext_flags & EXT_FLAG_EMBREF) ? m->m_ext.ext_count :
975 * mbuf, cluster, and external object allocation macros (for compatibility
978 #define M_MOVE_PKTHDR(to, from) m_move_pkthdr((to), (from))
979 #define MGET(m, how, type) ((m) = m_get((how), (type)))
980 #define MGETHDR(m, how, type) ((m) = m_gethdr((how), (type)))
981 #define MCLGET(m, how) m_clget((m), (how))
982 #define MEXTADD(m, buf, size, free, arg1, arg2, flags, type) \
983 m_extadd((m), (char *)(buf), (size), (free), (arg1), (arg2), \
985 #define m_getm(m, len, how, type) \
986 m_getm2((m), (len), (how), (type), M_PKTHDR)
989 * Evaluate TRUE if it's safe to write to the mbuf m's data region (this can
990 * be both the local data payload, or an external buffer area, depending on
991 * whether M_EXT is set).
993 #define M_WRITABLE(m) (((m)->m_flags & (M_RDONLY | M_NOMAP)) == 0 && \
994 (!(((m)->m_flags & M_EXT)) || \
995 (m_extrefcnt(m) == 1)))
997 /* Check if the supplied mbuf has a packet header, or else panic. */
998 #define M_ASSERTPKTHDR(m) \
999 KASSERT((m) != NULL && (m)->m_flags & M_PKTHDR, \
1000 ("%s: no mbuf packet header!", __func__))
1003 * Ensure that the supplied mbuf is a valid, non-free mbuf.
1005 * XXX: Broken at the moment. Need some UMA magic to make it work again.
1007 #define M_ASSERTVALID(m) \
1008 KASSERT((((struct mbuf *)m)->m_flags & 0) == 0, \
1009 ("%s: attempted use of a free mbuf!", __func__))
1012 * Return the address of the start of the buffer associated with an mbuf,
1013 * handling external storage, packet-header mbufs, and regular data mbufs.
1015 #define M_START(m) \
1016 (((m)->m_flags & M_NOMAP) ? NULL : \
1017 ((m)->m_flags & M_EXT) ? (m)->m_ext.ext_buf : \
1018 ((m)->m_flags & M_PKTHDR) ? &(m)->m_pktdat[0] : \
1022 * Return the size of the buffer associated with an mbuf, handling external
1023 * storage, packet-header mbufs, and regular data mbufs.
1026 (((m)->m_flags & M_EXT) ? (m)->m_ext.ext_size : \
1027 ((m)->m_flags & M_PKTHDR) ? MHLEN : \
1031 * Set the m_data pointer of a newly allocated mbuf to place an object of the
1032 * specified size at the end of the mbuf, longword aligned.
1034 * NB: Historically, we had M_ALIGN(), MH_ALIGN(), and MEXT_ALIGN() as
1035 * separate macros, each asserting that it was called at the proper moment.
1036 * This required callers to themselves test the storage type and call the
1037 * right one. Rather than require callers to be aware of those layout
1038 * decisions, we centralize here.
1040 static __inline void
1041 m_align(struct mbuf *m, int len)
1044 const char *msg = "%s: not a virgin mbuf";
1048 KASSERT(m->m_data == M_START(m), (msg, __func__));
1050 adjust = M_SIZE(m) - len;
1051 m->m_data += adjust &~ (sizeof(long)-1);
1054 #define M_ALIGN(m, len) m_align(m, len)
1055 #define MH_ALIGN(m, len) m_align(m, len)
1056 #define MEXT_ALIGN(m, len) m_align(m, len)
1059 * Compute the amount of space available before the current start of data in
1062 * The M_WRITABLE() is a temporary, conservative safety measure: the burden
1063 * of checking writability of the mbuf data area rests solely with the caller.
1065 * NB: In previous versions, M_LEADINGSPACE() would only check M_WRITABLE()
1066 * for mbufs with external storage. We now allow mbuf-embedded data to be
1067 * read-only as well.
1069 #define M_LEADINGSPACE(m) \
1070 (M_WRITABLE(m) ? ((m)->m_data - M_START(m)) : 0)
1073 * Compute the amount of space available after the end of data in an mbuf.
1075 * The M_WRITABLE() is a temporary, conservative safety measure: the burden
1076 * of checking writability of the mbuf data area rests solely with the caller.
1078 * NB: In previous versions, M_TRAILINGSPACE() would only check M_WRITABLE()
1079 * for mbufs with external storage. We now allow mbuf-embedded data to be
1080 * read-only as well.
1082 #define M_TRAILINGSPACE(m) \
1084 ((M_START(m) + M_SIZE(m)) - ((m)->m_data + (m)->m_len)) : 0)
1087 * Arrange to prepend space of size plen to mbuf m. If a new mbuf must be
1088 * allocated, how specifies whether to wait. If the allocation fails, the
1089 * original mbuf chain is freed and m is set to NULL.
1091 #define M_PREPEND(m, plen, how) do { \
1092 struct mbuf **_mmp = &(m); \
1093 struct mbuf *_mm = *_mmp; \
1094 int _mplen = (plen); \
1095 int __mhow = (how); \
1097 MBUF_CHECKSLEEP(how); \
1098 if (M_LEADINGSPACE(_mm) >= _mplen) { \
1099 _mm->m_data -= _mplen; \
1100 _mm->m_len += _mplen; \
1102 _mm = m_prepend(_mm, _mplen, __mhow); \
1103 if (_mm != NULL && _mm->m_flags & M_PKTHDR) \
1104 _mm->m_pkthdr.len += _mplen; \
1109 * Change mbuf to new type. This is a relatively expensive operation and
1110 * should be avoided.
1112 #define MCHTYPE(m, t) m_chtype((m), (t))
1114 /* Return the rcvif of a packet header. */
1115 static __inline struct ifnet *
1116 m_rcvif(struct mbuf *m)
1120 if (m->m_pkthdr.csum_flags & CSUM_SND_TAG)
1122 return (m->m_pkthdr.rcvif);
1125 /* Length to m_copy to copy all. */
1126 #define M_COPYALL 1000000000
1128 extern int max_datalen; /* MHLEN - max_hdr */
1129 extern int max_hdr; /* Largest link + protocol header */
1130 extern int max_linkhdr; /* Largest link-level header */
1131 extern int max_protohdr; /* Largest protocol header */
1132 extern int nmbclusters; /* Maximum number of clusters */
1133 extern bool mb_use_ext_pgs; /* Use ext_pgs for sendfile */
1136 * Network packets may have annotations attached by affixing a list of
1137 * "packet tags" to the pkthdr structure. Packet tags are dynamically
1138 * allocated semi-opaque data structures that have a fixed header
1139 * (struct m_tag) that specifies the size of the memory block and a
1140 * <cookie,type> pair that identifies it. The cookie is a 32-bit unique
1141 * unsigned value used to identify a module or ABI. By convention this value
1142 * is chosen as the date+time that the module is created, expressed as the
1143 * number of seconds since the epoch (e.g., using date -u +'%s'). The type
1144 * value is an ABI/module-specific value that identifies a particular
1145 * annotation and is private to the module. For compatibility with systems
1146 * like OpenBSD that define packet tags w/o an ABI/module cookie, the value
1147 * PACKET_ABI_COMPAT is used to implement m_tag_get and m_tag_find
1148 * compatibility shim functions and several tag types are defined below.
1149 * Users that do not require compatibility should use a private cookie value
1150 * so that packet tag-related definitions can be maintained privately.
1152 * Note that the packet tag returned by m_tag_alloc has the default memory
1153 * alignment implemented by malloc. To reference private data one can use a
1156 * struct m_tag *mtag = m_tag_alloc(...);
1157 * struct foo *p = (struct foo *)(mtag+1);
1159 * if the alignment of struct m_tag is sufficient for referencing members of
1160 * struct foo. Otherwise it is necessary to embed struct m_tag within the
1161 * private data structure to insure proper alignment; e.g.,
1167 * struct foo *p = (struct foo *) m_tag_alloc(...);
1168 * struct m_tag *mtag = &p->tag;
1172 * Persistent tags stay with an mbuf until the mbuf is reclaimed. Otherwise
1173 * tags are expected to ``vanish'' when they pass through a network
1174 * interface. For most interfaces this happens normally as the tags are
1175 * reclaimed when the mbuf is free'd. However in some special cases
1176 * reclaiming must be done manually. An example is packets that pass through
1177 * the loopback interface. Also, one must be careful to do this when
1178 * ``turning around'' packets (e.g., icmp_reflect).
1180 * To mark a tag persistent bit-or this flag in when defining the tag id.
1181 * The tag will then be treated as described above.
1183 #define MTAG_PERSISTENT 0x800
1185 #define PACKET_TAG_NONE 0 /* Nadda */
1187 /* Packet tags for use with PACKET_ABI_COMPAT. */
1188 #define PACKET_TAG_IPSEC_IN_DONE 1 /* IPsec applied, in */
1189 #define PACKET_TAG_IPSEC_OUT_DONE 2 /* IPsec applied, out */
1190 #define PACKET_TAG_IPSEC_IN_CRYPTO_DONE 3 /* NIC IPsec crypto done */
1191 #define PACKET_TAG_IPSEC_OUT_CRYPTO_NEEDED 4 /* NIC IPsec crypto req'ed */
1192 #define PACKET_TAG_IPSEC_IN_COULD_DO_CRYPTO 5 /* NIC notifies IPsec */
1193 #define PACKET_TAG_IPSEC_PENDING_TDB 6 /* Reminder to do IPsec */
1194 #define PACKET_TAG_BRIDGE 7 /* Bridge processing done */
1195 #define PACKET_TAG_GIF 8 /* GIF processing done */
1196 #define PACKET_TAG_GRE 9 /* GRE processing done */
1197 #define PACKET_TAG_IN_PACKET_CHECKSUM 10 /* NIC checksumming done */
1198 #define PACKET_TAG_ENCAP 11 /* Encap. processing */
1199 #define PACKET_TAG_IPSEC_SOCKET 12 /* IPSEC socket ref */
1200 #define PACKET_TAG_IPSEC_HISTORY 13 /* IPSEC history */
1201 #define PACKET_TAG_IPV6_INPUT 14 /* IPV6 input processing */
1202 #define PACKET_TAG_DUMMYNET 15 /* dummynet info */
1203 #define PACKET_TAG_DIVERT 17 /* divert info */
1204 #define PACKET_TAG_IPFORWARD 18 /* ipforward info */
1205 #define PACKET_TAG_MACLABEL (19 | MTAG_PERSISTENT) /* MAC label */
1206 #define PACKET_TAG_PF (21 | MTAG_PERSISTENT) /* PF/ALTQ information */
1207 #define PACKET_TAG_RTSOCKFAM 25 /* rtsock sa family */
1208 #define PACKET_TAG_IPOPTIONS 27 /* Saved IP options */
1209 #define PACKET_TAG_CARP 28 /* CARP info */
1210 #define PACKET_TAG_IPSEC_NAT_T_PORTS 29 /* two uint16_t */
1211 #define PACKET_TAG_ND_OUTGOING 30 /* ND outgoing */
1213 /* Specific cookies and tags. */
1215 /* Packet tag routines. */
1216 struct m_tag *m_tag_alloc(u_int32_t, int, int, int);
1217 void m_tag_delete(struct mbuf *, struct m_tag *);
1218 void m_tag_delete_chain(struct mbuf *, struct m_tag *);
1219 void m_tag_free_default(struct m_tag *);
1220 struct m_tag *m_tag_locate(struct mbuf *, u_int32_t, int, struct m_tag *);
1221 struct m_tag *m_tag_copy(struct m_tag *, int);
1222 int m_tag_copy_chain(struct mbuf *, const struct mbuf *, int);
1223 void m_tag_delete_nonpersistent(struct mbuf *);
1226 * Initialize the list of tags associated with an mbuf.
1228 static __inline void
1229 m_tag_init(struct mbuf *m)
1232 SLIST_INIT(&m->m_pkthdr.tags);
1236 * Set up the contents of a tag. Note that this does not fill in the free
1237 * method; the caller is expected to do that.
1239 * XXX probably should be called m_tag_init, but that was already taken.
1241 static __inline void
1242 m_tag_setup(struct m_tag *t, u_int32_t cookie, int type, int len)
1247 t->m_tag_cookie = cookie;
1251 * Reclaim resources associated with a tag.
1253 static __inline void
1254 m_tag_free(struct m_tag *t)
1257 (*t->m_tag_free)(t);
1261 * Return the first tag associated with an mbuf.
1263 static __inline struct m_tag *
1264 m_tag_first(struct mbuf *m)
1267 return (SLIST_FIRST(&m->m_pkthdr.tags));
1271 * Return the next tag in the list of tags associated with an mbuf.
1273 static __inline struct m_tag *
1274 m_tag_next(struct mbuf *m __unused, struct m_tag *t)
1277 return (SLIST_NEXT(t, m_tag_link));
1281 * Prepend a tag to the list of tags associated with an mbuf.
1283 static __inline void
1284 m_tag_prepend(struct mbuf *m, struct m_tag *t)
1287 SLIST_INSERT_HEAD(&m->m_pkthdr.tags, t, m_tag_link);
1291 * Unlink a tag from the list of tags associated with an mbuf.
1293 static __inline void
1294 m_tag_unlink(struct mbuf *m, struct m_tag *t)
1297 SLIST_REMOVE(&m->m_pkthdr.tags, t, m_tag, m_tag_link);
1300 /* These are for OpenBSD compatibility. */
1301 #define MTAG_ABI_COMPAT 0 /* compatibility ABI */
1303 static __inline struct m_tag *
1304 m_tag_get(int type, int length, int wait)
1306 return (m_tag_alloc(MTAG_ABI_COMPAT, type, length, wait));
1309 static __inline struct m_tag *
1310 m_tag_find(struct mbuf *m, int type, struct m_tag *start)
1312 return (SLIST_EMPTY(&m->m_pkthdr.tags) ? (struct m_tag *)NULL :
1313 m_tag_locate(m, MTAG_ABI_COMPAT, type, start));
1316 static inline struct m_snd_tag *
1317 m_snd_tag_ref(struct m_snd_tag *mst)
1320 refcount_acquire(&mst->refcount);
1325 m_snd_tag_rele(struct m_snd_tag *mst)
1328 if (refcount_release(&mst->refcount))
1329 m_snd_tag_destroy(mst);
1332 static __inline struct mbuf *
1333 m_free(struct mbuf *m)
1335 struct mbuf *n = m->m_next;
1337 MBUF_PROBE1(m__free, m);
1338 if ((m->m_flags & (M_PKTHDR|M_NOFREE)) == (M_PKTHDR|M_NOFREE))
1339 m_tag_delete_chain(m, NULL);
1340 if (m->m_flags & M_PKTHDR && m->m_pkthdr.csum_flags & CSUM_SND_TAG)
1341 m_snd_tag_rele(m->m_pkthdr.snd_tag);
1342 if (m->m_flags & M_EXT)
1344 else if ((m->m_flags & M_NOFREE) == 0)
1345 uma_zfree(zone_mbuf, m);
1350 rt_m_getfib(struct mbuf *m)
1352 KASSERT(m->m_flags & M_PKTHDR , ("Attempt to get FIB from non header mbuf."));
1353 return (m->m_pkthdr.fibnum);
1356 #define M_GETFIB(_m) rt_m_getfib(_m)
1358 #define M_SETFIB(_m, _fib) do { \
1359 KASSERT((_m)->m_flags & M_PKTHDR, ("Attempt to set FIB on non header mbuf.")); \
1360 ((_m)->m_pkthdr.fibnum) = (_fib); \
1363 /* flags passed as first argument for "m_ether_tcpip_hash()" */
1364 #define MBUF_HASHFLAG_L2 (1 << 2)
1365 #define MBUF_HASHFLAG_L3 (1 << 3)
1366 #define MBUF_HASHFLAG_L4 (1 << 4)
1368 /* mbuf hashing helper routines */
1369 uint32_t m_ether_tcpip_hash_init(void);
1370 uint32_t m_ether_tcpip_hash(const uint32_t, const struct mbuf *, const uint32_t);
1372 #ifdef MBUF_PROFILING
1373 void m_profile(struct mbuf *m);
1374 #define M_PROFILE(m) m_profile(m)
1376 #define M_PROFILE(m)
1380 STAILQ_HEAD(, mbuf) mq_head;
1386 mbufq_init(struct mbufq *mq, int maxlen)
1389 STAILQ_INIT(&mq->mq_head);
1390 mq->mq_maxlen = maxlen;
1394 static inline struct mbuf *
1395 mbufq_flush(struct mbufq *mq)
1399 m = STAILQ_FIRST(&mq->mq_head);
1400 STAILQ_INIT(&mq->mq_head);
1406 mbufq_drain(struct mbufq *mq)
1410 n = mbufq_flush(mq);
1411 while ((m = n) != NULL) {
1412 n = STAILQ_NEXT(m, m_stailqpkt);
1417 static inline struct mbuf *
1418 mbufq_first(const struct mbufq *mq)
1421 return (STAILQ_FIRST(&mq->mq_head));
1424 static inline struct mbuf *
1425 mbufq_last(const struct mbufq *mq)
1428 return (STAILQ_LAST(&mq->mq_head, mbuf, m_stailqpkt));
1432 mbufq_full(const struct mbufq *mq)
1435 return (mq->mq_len >= mq->mq_maxlen);
1439 mbufq_len(const struct mbufq *mq)
1442 return (mq->mq_len);
1446 mbufq_enqueue(struct mbufq *mq, struct mbuf *m)
1451 STAILQ_INSERT_TAIL(&mq->mq_head, m, m_stailqpkt);
1456 static inline struct mbuf *
1457 mbufq_dequeue(struct mbufq *mq)
1461 m = STAILQ_FIRST(&mq->mq_head);
1463 STAILQ_REMOVE_HEAD(&mq->mq_head, m_stailqpkt);
1464 m->m_nextpkt = NULL;
1471 mbufq_prepend(struct mbufq *mq, struct mbuf *m)
1474 STAILQ_INSERT_HEAD(&mq->mq_head, m, m_stailqpkt);
1479 * Note: this doesn't enforce the maximum list size for dst.
1482 mbufq_concat(struct mbufq *mq_dst, struct mbufq *mq_src)
1485 mq_dst->mq_len += mq_src->mq_len;
1486 STAILQ_CONCAT(&mq_dst->mq_head, &mq_src->mq_head);
1490 #ifdef _SYS_TIMESPEC_H_
1492 mbuf_tstmp2timespec(struct mbuf *m, struct timespec *ts)
1495 KASSERT((m->m_flags & M_PKTHDR) != 0, ("mbuf %p no M_PKTHDR", m));
1496 KASSERT((m->m_flags & M_TSTMP) != 0, ("mbuf %p no M_TSTMP", m));
1497 ts->tv_sec = m->m_pkthdr.rcv_tstmp / 1000000000;
1498 ts->tv_nsec = m->m_pkthdr.rcv_tstmp % 1000000000;
1503 /* Invoked from the netdump client code. */
1504 void netdump_mbuf_drain(void);
1505 void netdump_mbuf_dump(void);
1506 void netdump_mbuf_reinit(int nmbuf, int nclust, int clsize);
1509 #endif /* _KERNEL */
1510 #endif /* !_SYS_MBUF_H_ */