2 * Copyright (c) 1982, 1986, 1988, 1993
3 * The Regents of the University of California.
6 * Redistribution and use in source and binary forms, with or without
7 * modification, are permitted provided that the following conditions
9 * 1. Redistributions of source code must retain the above copyright
10 * notice, this list of conditions and the following disclaimer.
11 * 2. Redistributions in binary form must reproduce the above copyright
12 * notice, this list of conditions and the following disclaimer in the
13 * documentation and/or other materials provided with the distribution.
14 * 3. Neither the name of the University nor the names of its contributors
15 * may be used to endorse or promote products derived from this software
16 * without specific prior written permission.
18 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
19 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
20 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
21 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
22 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
23 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
24 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
25 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
26 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
27 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
30 * @(#)mbuf.h 8.5 (Berkeley) 2/19/95
37 /* XXX: These includes suck. Sorry! */
38 #include <sys/queue.h>
40 #include <sys/systm.h>
48 * Mbufs are of a single size, MSIZE (sys/param.h), which includes overhead.
49 * An mbuf may add a single "mbuf cluster" of size MCLBYTES (also in
50 * sys/param.h), which has no additional overhead and is used instead of the
51 * internal data area; this is done when at least MINCLSIZE of data must be
52 * stored. Additionally, it is possible to allocate a separate buffer
53 * externally and attach it to the mbuf in a way similar to that of mbuf
56 * NB: These calculation do not take actual compiler-induced alignment and
57 * padding inside the complete struct mbuf into account. Appropriate
58 * attention is required when changing members of struct mbuf.
60 * MLEN is data length in a normal mbuf.
61 * MHLEN is data length in an mbuf with pktheader.
62 * MINCLSIZE is a smallest amount of data that should be put into cluster.
64 #define MLEN ((int)(MSIZE - sizeof(struct m_hdr)))
65 #define MHLEN ((int)(MLEN - sizeof(struct pkthdr)))
66 #define MINCLSIZE (MHLEN + 1)
70 * Macro for type conversion: convert mbuf pointer to data pointer of correct
73 * mtod(m, t) -- Convert mbuf pointer to data pointer of correct type.
74 * mtodo(m, o) -- Same as above but with offset 'o' into data.
76 #define mtod(m, t) ((t)((m)->m_data))
77 #define mtodo(m, o) ((void *)(((m)->m_data) + (o)))
80 * Argument structure passed to UMA routines during mbuf and packet
84 int flags; /* Flags for mbuf being allocated */
85 short type; /* Type of mbuf being allocated */
90 * Header present at the beginning of every mbuf.
95 struct mbuf *mh_next; /* next buffer in chain */
96 struct mbuf *mh_nextpkt; /* next chain in queue/record */
97 caddr_t mh_data; /* location of data */
98 int32_t mh_len; /* amount of data in this mbuf */
99 uint32_t mh_type:8, /* type of data in this mbuf */
100 mh_flags:24; /* flags; see below */
101 #if !defined(__LP64__)
102 uint32_t mh_pad; /* pad for 64bit alignment */
107 * Packet tag structure (see below for details).
110 SLIST_ENTRY(m_tag) m_tag_link; /* List of packet tags */
111 u_int16_t m_tag_id; /* Tag ID */
112 u_int16_t m_tag_len; /* Length of data */
113 u_int32_t m_tag_cookie; /* ABI/Module ID */
114 void (*m_tag_free)(struct m_tag *);
118 * Record/packet header in first mbuf of chain; valid only if M_PKTHDR is set.
123 struct ifnet *rcvif; /* rcv interface */
124 SLIST_HEAD(packet_tags, m_tag) tags; /* list of packet tags */
125 int32_t len; /* total packet length */
127 /* Layer crossing persistent information. */
128 uint32_t flowid; /* packet's 4-tuple system */
129 uint64_t csum_flags; /* checksum and offload features */
130 uint16_t fibnum; /* this packet should use this fib */
131 uint8_t cosqos; /* class/quality of service */
132 uint8_t rsstype; /* hash type */
133 uint8_t l2hlen; /* layer 2 header length */
134 uint8_t l3hlen; /* layer 3 header length */
135 uint8_t l4hlen; /* layer 4 header length */
136 uint8_t l5hlen; /* layer 5 header length */
140 uint32_t thirtytwo[2];
141 uint64_t sixtyfour[1];
142 uintptr_t unintptr[1];
146 /* Layer specific non-persistent local storage for reassembly, etc. */
150 uint32_t thirtytwo[2];
151 uint64_t sixtyfour[1];
152 uintptr_t unintptr[1];
156 #define ether_vtag PH_per.sixteen[0]
158 #define vt_nrecs sixteen[0]
159 #define tso_segsz PH_per.sixteen[1]
160 #define csum_phsum PH_per.sixteen[2]
161 #define csum_data PH_per.thirtytwo[1]
162 #define pkt_tcphdr PH_loc.ptr
165 * Description of external storage mapped into mbuf; valid only if M_EXT is
171 volatile u_int *ext_cnt; /* pointer to ref count info */
172 caddr_t ext_buf; /* start of buffer */
173 uint32_t ext_size; /* size of buffer, for ext_free */
174 uint32_t ext_type:8, /* type of external storage */
175 ext_flags:24; /* external storage mbuf flags */
176 void (*ext_free) /* free routine if not the usual */
177 (struct mbuf *, void *, void *);
178 void *ext_arg1; /* optional argument pointer */
179 void *ext_arg2; /* optional argument pointer */
183 * The core of the mbuf object along with some shortcut defines for practical
190 struct pkthdr MH_pkthdr; /* M_PKTHDR set */
192 struct m_ext MH_ext; /* M_EXT set */
193 char MH_databuf[MHLEN];
196 char M_databuf[MLEN]; /* !M_PKTHDR, !M_EXT */
199 #define m_next m_hdr.mh_next
200 #define m_len m_hdr.mh_len
201 #define m_data m_hdr.mh_data
202 #define m_type m_hdr.mh_type
203 #define m_flags m_hdr.mh_flags
204 #define m_nextpkt m_hdr.mh_nextpkt
205 #define m_act m_nextpkt
206 #define m_pkthdr M_dat.MH.MH_pkthdr
207 #define m_ext M_dat.MH.MH_dat.MH_ext
208 #define m_pktdat M_dat.MH.MH_dat.MH_databuf
209 #define m_dat M_dat.M_databuf
212 * mbuf flags of global significance and layer crossing.
213 * Those of only protocol/layer specific significance are to be mapped
214 * to M_PROTO[1-12] and cleared at layer handoff boundaries.
215 * NB: Limited to the lower 24 bits.
217 #define M_EXT 0x00000001 /* has associated external storage */
218 #define M_PKTHDR 0x00000002 /* start of record */
219 #define M_EOR 0x00000004 /* end of record */
220 #define M_RDONLY 0x00000008 /* associated data is marked read-only */
221 #define M_BCAST 0x00000010 /* send/received as link-level broadcast */
222 #define M_MCAST 0x00000020 /* send/received as link-level multicast */
223 #define M_PROMISC 0x00000040 /* packet was not for us */
224 #define M_VLANTAG 0x00000080 /* ether_vtag is valid */
225 #define M_FLOWID 0x00000100 /* deprecated: flowid is valid */
226 #define M_NOFREE 0x00000200 /* do not free mbuf, embedded in cluster */
228 #define M_PROTO1 0x00001000 /* protocol-specific */
229 #define M_PROTO2 0x00002000 /* protocol-specific */
230 #define M_PROTO3 0x00004000 /* protocol-specific */
231 #define M_PROTO4 0x00008000 /* protocol-specific */
232 #define M_PROTO5 0x00010000 /* protocol-specific */
233 #define M_PROTO6 0x00020000 /* protocol-specific */
234 #define M_PROTO7 0x00040000 /* protocol-specific */
235 #define M_PROTO8 0x00080000 /* protocol-specific */
236 #define M_PROTO9 0x00100000 /* protocol-specific */
237 #define M_PROTO10 0x00200000 /* protocol-specific */
238 #define M_PROTO11 0x00400000 /* protocol-specific */
239 #define M_PROTO12 0x00800000 /* protocol-specific */
242 * Flags to purge when crossing layers.
244 #define M_PROTOFLAGS \
245 (M_PROTO1|M_PROTO2|M_PROTO3|M_PROTO4|M_PROTO5|M_PROTO6|M_PROTO7|M_PROTO8|\
246 M_PROTO9|M_PROTO10|M_PROTO11|M_PROTO12)
249 * Flags preserved when copying m_pkthdr.
251 #define M_COPYFLAGS \
252 (M_PKTHDR|M_EOR|M_RDONLY|M_BCAST|M_MCAST|M_VLANTAG|M_PROMISC| \
256 * Mbuf flag description for use with printf(9) %b identifier.
258 #define M_FLAG_BITS \
259 "\20\1M_EXT\2M_PKTHDR\3M_EOR\4M_RDONLY\5M_BCAST\6M_MCAST" \
260 "\7M_PROMISC\10M_VLANTAG\11M_FLOWID"
261 #define M_FLAG_PROTOBITS \
262 "\15M_PROTO1\16M_PROTO2\17M_PROTO3\20M_PROTO4\21M_PROTO5" \
263 "\22M_PROTO6\23M_PROTO7\24M_PROTO8\25M_PROTO9\26M_PROTO10" \
264 "\27M_PROTO11\30M_PROTO12"
265 #define M_FLAG_PRINTF (M_FLAG_BITS M_FLAG_PROTOBITS)
268 * Network interface cards are able to hash protocol fields (such as IPv4
269 * addresses and TCP port numbers) classify packets into flows. These flows
270 * can then be used to maintain ordering while delivering packets to the OS
271 * via parallel input queues, as well as to provide a stateless affinity
272 * model. NIC drivers can pass up the hash via m->m_pkthdr.flowid, and set
273 * m_flag fields to indicate how the hash should be interpreted by the
276 * Most NICs support RSS, which provides ordering and explicit affinity, and
277 * use the hash m_flag bits to indicate what header fields were covered by
278 * the hash. M_HASHTYPE_OPAQUE can be set by non-RSS cards or configurations
279 * that provide an opaque flow identifier, allowing for ordering and
280 * distribution without explicit affinity.
282 #define M_HASHTYPE_NONE 0
283 #define M_HASHTYPE_RSS_IPV4 1 /* IPv4 2-tuple */
284 #define M_HASHTYPE_RSS_TCP_IPV4 2 /* TCPv4 4-tuple */
285 #define M_HASHTYPE_RSS_IPV6 3 /* IPv6 2-tuple */
286 #define M_HASHTYPE_RSS_TCP_IPV6 4 /* TCPv6 4-tuple */
287 #define M_HASHTYPE_RSS_IPV6_EX 5 /* IPv6 2-tuple + ext hdrs */
288 #define M_HASHTYPE_RSS_TCP_IPV6_EX 6 /* TCPv6 4-tiple + ext hdrs */
289 #define M_HASHTYPE_OPAQUE 255 /* ordering, not affinity */
291 #define M_HASHTYPE_CLEAR(m) ((m)->m_pkthdr.rsstype = 0)
292 #define M_HASHTYPE_GET(m) ((m)->m_pkthdr.rsstype)
293 #define M_HASHTYPE_SET(m, v) ((m)->m_pkthdr.rsstype = (v))
294 #define M_HASHTYPE_TEST(m, v) (M_HASHTYPE_GET(m) == (v))
297 * COS/QOS class and quality of service tags.
298 * It uses DSCP code points as base.
300 #define QOS_DSCP_CS0 0x00
301 #define QOS_DSCP_DEF QOS_DSCP_CS0
302 #define QOS_DSCP_CS1 0x20
303 #define QOS_DSCP_AF11 0x28
304 #define QOS_DSCP_AF12 0x30
305 #define QOS_DSCP_AF13 0x38
306 #define QOS_DSCP_CS2 0x40
307 #define QOS_DSCP_AF21 0x48
308 #define QOS_DSCP_AF22 0x50
309 #define QOS_DSCP_AF23 0x58
310 #define QOS_DSCP_CS3 0x60
311 #define QOS_DSCP_AF31 0x68
312 #define QOS_DSCP_AF32 0x70
313 #define QOS_DSCP_AF33 0x78
314 #define QOS_DSCP_CS4 0x80
315 #define QOS_DSCP_AF41 0x88
316 #define QOS_DSCP_AF42 0x90
317 #define QOS_DSCP_AF43 0x98
318 #define QOS_DSCP_CS5 0xa0
319 #define QOS_DSCP_EF 0xb8
320 #define QOS_DSCP_CS6 0xc0
321 #define QOS_DSCP_CS7 0xe0
324 * External mbuf storage buffer types.
326 #define EXT_CLUSTER 1 /* mbuf cluster */
327 #define EXT_SFBUF 2 /* sendfile(2)'s sf_bufs */
328 #define EXT_JUMBOP 3 /* jumbo cluster 4096 bytes */
329 #define EXT_JUMBO9 4 /* jumbo cluster 9216 bytes */
330 #define EXT_JUMBO16 5 /* jumbo cluster 16184 bytes */
331 #define EXT_PACKET 6 /* mbuf+cluster from packet zone */
332 #define EXT_MBUF 7 /* external mbuf reference (M_IOVEC) */
334 #define EXT_VENDOR1 224 /* for vendor-internal use */
335 #define EXT_VENDOR2 225 /* for vendor-internal use */
336 #define EXT_VENDOR3 226 /* for vendor-internal use */
337 #define EXT_VENDOR4 227 /* for vendor-internal use */
339 #define EXT_EXP1 244 /* for experimental use */
340 #define EXT_EXP2 245 /* for experimental use */
341 #define EXT_EXP3 246 /* for experimental use */
342 #define EXT_EXP4 247 /* for experimental use */
344 #define EXT_NET_DRV 252 /* custom ext_buf provided by net driver(s) */
345 #define EXT_MOD_TYPE 253 /* custom module's ext_buf type */
346 #define EXT_DISPOSABLE 254 /* can throw this buffer away w/page flipping */
347 #define EXT_EXTREF 255 /* has externally maintained ext_cnt ptr */
350 * Flags for external mbuf buffer types.
351 * NB: limited to the lower 24 bits.
353 #define EXT_FLAG_EMBREF 0x000001 /* embedded ext_cnt, notyet */
354 #define EXT_FLAG_EXTREF 0x000002 /* external ext_cnt, notyet */
355 #define EXT_FLAG_NOFREE 0x000010 /* don't free mbuf to pool, notyet */
357 #define EXT_FLAG_VENDOR1 0x010000 /* for vendor-internal use */
358 #define EXT_FLAG_VENDOR2 0x020000 /* for vendor-internal use */
359 #define EXT_FLAG_VENDOR3 0x040000 /* for vendor-internal use */
360 #define EXT_FLAG_VENDOR4 0x080000 /* for vendor-internal use */
362 #define EXT_FLAG_EXP1 0x100000 /* for experimental use */
363 #define EXT_FLAG_EXP2 0x200000 /* for experimental use */
364 #define EXT_FLAG_EXP3 0x400000 /* for experimental use */
365 #define EXT_FLAG_EXP4 0x800000 /* for experimental use */
368 * EXT flag description for use with printf(9) %b identifier.
370 #define EXT_FLAG_BITS \
371 "\20\1EXT_FLAG_EMBREF\2EXT_FLAG_EXTREF\5EXT_FLAG_NOFREE" \
372 "\21EXT_FLAG_VENDOR1\22EXT_FLAG_VENDOR2\23EXT_FLAG_VENDOR3" \
373 "\24EXT_FLAG_VENDOR4\25EXT_FLAG_EXP1\26EXT_FLAG_EXP2\27EXT_FLAG_EXP3" \
377 * External reference/free functions.
379 void sf_ext_ref(void *, void *);
380 void sf_ext_free(void *, void *);
383 * Flags indicating checksum, segmentation and other offload work to be
384 * done, or already done, by hardware or lower layers. It is split into
385 * separate inbound and outbound flags.
387 * Outbound flags that are set by upper protocol layers requesting lower
388 * layers, or ideally the hardware, to perform these offloading tasks.
389 * For outbound packets this field and its flags can be directly tested
390 * against if_data.ifi_hwassist.
392 #define CSUM_IP 0x00000001 /* IP header checksum offload */
393 #define CSUM_IP_UDP 0x00000002 /* UDP checksum offload */
394 #define CSUM_IP_TCP 0x00000004 /* TCP checksum offload */
395 #define CSUM_IP_SCTP 0x00000008 /* SCTP checksum offload */
396 #define CSUM_IP_TSO 0x00000010 /* TCP segmentation offload */
397 #define CSUM_IP_ISCSI 0x00000020 /* iSCSI checksum offload */
399 #define CSUM_IP6_UDP 0x00000200 /* UDP checksum offload */
400 #define CSUM_IP6_TCP 0x00000400 /* TCP checksum offload */
401 #define CSUM_IP6_SCTP 0x00000800 /* SCTP checksum offload */
402 #define CSUM_IP6_TSO 0x00001000 /* TCP segmentation offload */
403 #define CSUM_IP6_ISCSI 0x00002000 /* iSCSI checksum offload */
405 /* Inbound checksum support where the checksum was verified by hardware. */
406 #define CSUM_L3_CALC 0x01000000 /* calculated layer 3 csum */
407 #define CSUM_L3_VALID 0x02000000 /* checksum is correct */
408 #define CSUM_L4_CALC 0x04000000 /* calculated layer 4 csum */
409 #define CSUM_L4_VALID 0x08000000 /* checksum is correct */
410 #define CSUM_L5_CALC 0x10000000 /* calculated layer 5 csum */
411 #define CSUM_L5_VALID 0x20000000 /* checksum is correct */
412 #define CSUM_COALESED 0x40000000 /* contains merged segments */
415 * CSUM flag description for use with printf(9) %b identifier.
418 "\20\1CSUM_IP\2CSUM_IP_UDP\3CSUM_IP_TCP\4CSUM_IP_SCTP\5CSUM_IP_TSO" \
420 "\12CSUM_IP6_UDP\13CSUM_IP6_TCP\14CSUM_IP6_SCTP\15CSUM_IP6_TSO" \
421 "\16CSUM_IP6_ISCSI" \
422 "\31CSUM_L3_CALC\32CSUM_L3_VALID\33CSUM_L4_CALC\34CSUM_L4_VALID" \
423 "\35CSUM_L5_CALC\36CSUM_L5_VALID\37CSUM_COALESED"
425 /* CSUM flags compatibility mappings. */
426 #define CSUM_IP_CHECKED CSUM_L3_CALC
427 #define CSUM_IP_VALID CSUM_L3_VALID
428 #define CSUM_DATA_VALID CSUM_L4_VALID
429 #define CSUM_PSEUDO_HDR CSUM_L4_CALC
430 #define CSUM_SCTP_VALID CSUM_L4_VALID
431 #define CSUM_DELAY_DATA (CSUM_TCP|CSUM_UDP)
432 #define CSUM_DELAY_IP CSUM_IP /* Only v4, no v6 IP hdr csum */
433 #define CSUM_DELAY_DATA_IPV6 (CSUM_TCP_IPV6|CSUM_UDP_IPV6)
434 #define CSUM_DATA_VALID_IPV6 CSUM_DATA_VALID
435 #define CSUM_TCP CSUM_IP_TCP
436 #define CSUM_UDP CSUM_IP_UDP
437 #define CSUM_SCTP CSUM_IP_SCTP
438 #define CSUM_TSO (CSUM_IP_TSO|CSUM_IP6_TSO)
439 #define CSUM_UDP_IPV6 CSUM_IP6_UDP
440 #define CSUM_TCP_IPV6 CSUM_IP6_TCP
441 #define CSUM_SCTP_IPV6 CSUM_IP6_SCTP
442 #define CSUM_FRAGMENT 0x0 /* Unused */
445 * mbuf types describing the content of the mbuf (including external storage).
447 #define MT_NOTMBUF 0 /* USED INTERNALLY ONLY! Object is not mbuf */
448 #define MT_DATA 1 /* dynamic (data) allocation */
449 #define MT_HEADER MT_DATA /* packet header, use M_PKTHDR instead */
451 #define MT_VENDOR1 4 /* for vendor-internal use */
452 #define MT_VENDOR2 5 /* for vendor-internal use */
453 #define MT_VENDOR3 6 /* for vendor-internal use */
454 #define MT_VENDOR4 7 /* for vendor-internal use */
456 #define MT_SONAME 8 /* socket name */
458 #define MT_EXP1 9 /* for experimental use */
459 #define MT_EXP2 10 /* for experimental use */
460 #define MT_EXP3 11 /* for experimental use */
461 #define MT_EXP4 12 /* for experimental use */
463 #define MT_CONTROL 14 /* extra-data protocol message */
464 #define MT_OOBDATA 15 /* expedited data */
465 #define MT_NTYPES 16 /* number of mbuf types for mbtypes[] */
467 #define MT_NOINIT 255 /* Not a type but a flag to allocate
468 a non-initialized mbuf */
471 * String names of mbuf-related UMA(9) and malloc(9) types. Exposed to
472 * !_KERNEL so that monitoring tools can look up the zones with
475 #define MBUF_MEM_NAME "mbuf"
476 #define MBUF_CLUSTER_MEM_NAME "mbuf_cluster"
477 #define MBUF_PACKET_MEM_NAME "mbuf_packet"
478 #define MBUF_JUMBOP_MEM_NAME "mbuf_jumbo_page"
479 #define MBUF_JUMBO9_MEM_NAME "mbuf_jumbo_9k"
480 #define MBUF_JUMBO16_MEM_NAME "mbuf_jumbo_16k"
481 #define MBUF_TAG_MEM_NAME "mbuf_tag"
482 #define MBUF_EXTREFCNT_MEM_NAME "mbuf_ext_refcnt"
487 #define MBUF_CHECKSLEEP(how) do { \
488 if (how == M_WAITOK) \
489 WITNESS_WARN(WARN_GIANTOK | WARN_SLEEPOK, NULL, \
490 "Sleeping in \"%s\"", __func__); \
493 #define MBUF_CHECKSLEEP(how)
497 * Network buffer allocation API
499 * The rest of it is defined in kern/kern_mbuf.c
501 extern uma_zone_t zone_mbuf;
502 extern uma_zone_t zone_clust;
503 extern uma_zone_t zone_pack;
504 extern uma_zone_t zone_jumbop;
505 extern uma_zone_t zone_jumbo9;
506 extern uma_zone_t zone_jumbo16;
507 extern uma_zone_t zone_ext_refcnt;
509 void mb_free_ext(struct mbuf *);
510 int m_pkthdr_init(struct mbuf *, int);
524 #if MJUMPAGESIZE != MCLBYTES
536 panic("%s: invalid cluster size %d", __func__, size);
543 * Associated an external reference counted buffer with an mbuf.
546 m_extaddref(struct mbuf *m, caddr_t buf, u_int size, u_int *ref_cnt,
547 void (*freef)(struct mbuf *, void *, void *), void *arg1, void *arg2)
550 KASSERT(ref_cnt != NULL, ("%s: ref_cnt not provided", __func__));
552 atomic_add_int(ref_cnt, 1);
554 m->m_ext.ext_buf = buf;
555 m->m_ext.ext_cnt = ref_cnt;
556 m->m_data = m->m_ext.ext_buf;
557 m->m_ext.ext_size = size;
558 m->m_ext.ext_free = freef;
559 m->m_ext.ext_arg1 = arg1;
560 m->m_ext.ext_arg2 = arg2;
561 m->m_ext.ext_type = EXT_EXTREF;
564 static __inline uma_zone_t
573 #if MJUMPAGESIZE != MCLBYTES
585 panic("%s: invalid cluster size %d", __func__, size);
592 * Initialize an mbuf with linear storage.
594 * Inline because the consumer text overhead will be roughly the same to
595 * initialize or call a function with this many parameters and M_PKTHDR
596 * should go away with constant propagation for !MGETHDR.
599 m_init(struct mbuf *m, uma_zone_t zone, int size, int how, short type,
606 m->m_data = m->m_dat;
610 if (flags & M_PKTHDR) {
611 if ((error = m_pkthdr_init(m, how)) != 0)
618 static __inline struct mbuf *
619 m_get(int how, short type)
625 return (uma_zalloc_arg(zone_mbuf, &args, how));
629 * XXX This should be deprecated, very little use.
631 static __inline struct mbuf *
632 m_getclr(int how, short type)
639 m = uma_zalloc_arg(zone_mbuf, &args, how);
641 bzero(m->m_data, MLEN);
645 static __inline struct mbuf *
646 m_gethdr(int how, short type)
650 args.flags = M_PKTHDR;
652 return (uma_zalloc_arg(zone_mbuf, &args, how));
655 static __inline struct mbuf *
656 m_getcl(int how, short type, int flags)
662 return (uma_zalloc_arg(zone_pack, &args, how));
666 m_clget(struct mbuf *m, int how)
669 if (m->m_flags & M_EXT)
670 printf("%s: %p mbuf already has cluster\n", __func__, m);
671 m->m_ext.ext_buf = (char *)NULL;
672 uma_zalloc_arg(zone_clust, m, how);
674 * On a cluster allocation failure, drain the packet zone and retry,
675 * we might be able to loosen a few clusters up on the drain.
677 if ((how & M_NOWAIT) && (m->m_ext.ext_buf == NULL)) {
678 zone_drain(zone_pack);
679 uma_zalloc_arg(zone_clust, m, how);
684 * m_cljget() is different from m_clget() as it can allocate clusters without
685 * attaching them to an mbuf. In that case the return value is the pointer
686 * to the cluster of the requested size. If an mbuf was specified, it gets
687 * the cluster attached to it and the return value can be safely ignored.
688 * For size it takes MCLBYTES, MJUMPAGESIZE, MJUM9BYTES, MJUM16BYTES.
690 static __inline void *
691 m_cljget(struct mbuf *m, int how, int size)
695 if (m && m->m_flags & M_EXT)
696 printf("%s: %p mbuf already has cluster\n", __func__, m);
698 m->m_ext.ext_buf = NULL;
700 zone = m_getzone(size);
701 return (uma_zalloc_arg(zone, m, how));
705 m_cljset(struct mbuf *m, void *cl, int type)
715 #if MJUMPAGESIZE != MCLBYTES
730 panic("%s: unknown cluster type %d", __func__, type);
734 m->m_data = m->m_ext.ext_buf = cl;
735 m->m_ext.ext_free = m->m_ext.ext_arg1 = m->m_ext.ext_arg2 = NULL;
736 m->m_ext.ext_size = size;
737 m->m_ext.ext_type = type;
738 m->m_ext.ext_flags = 0;
739 m->m_ext.ext_cnt = uma_find_refcnt(zone, cl);
745 m_chtype(struct mbuf *m, short new_type)
748 m->m_type = new_type;
752 m_clrprotoflags(struct mbuf *m)
755 m->m_flags &= ~M_PROTOFLAGS;
758 static __inline struct mbuf *
759 m_last(struct mbuf *m)
768 * mbuf, cluster, and external object allocation macros (for compatibility
771 #define M_MOVE_PKTHDR(to, from) m_move_pkthdr((to), (from))
772 #define MGET(m, how, type) ((m) = m_get((how), (type)))
773 #define MGETHDR(m, how, type) ((m) = m_gethdr((how), (type)))
774 #define MCLGET(m, how) m_clget((m), (how))
775 #define MEXTADD(m, buf, size, free, arg1, arg2, flags, type) \
776 (void )m_extadd((m), (caddr_t)(buf), (size), (free), (arg1), (arg2),\
777 (flags), (type), M_NOWAIT)
778 #define m_getm(m, len, how, type) \
779 m_getm2((m), (len), (how), (type), M_PKTHDR)
782 * Evaluate TRUE if it's safe to write to the mbuf m's data region (this can
783 * be both the local data payload, or an external buffer area, depending on
784 * whether M_EXT is set).
786 #define M_WRITABLE(m) (!((m)->m_flags & M_RDONLY) && \
787 (!(((m)->m_flags & M_EXT)) || \
788 (*((m)->m_ext.ext_cnt) == 1)) ) \
790 /* Check if the supplied mbuf has a packet header, or else panic. */
791 #define M_ASSERTPKTHDR(m) \
792 KASSERT((m) != NULL && (m)->m_flags & M_PKTHDR, \
793 ("%s: no mbuf packet header!", __func__))
796 * Ensure that the supplied mbuf is a valid, non-free mbuf.
798 * XXX: Broken at the moment. Need some UMA magic to make it work again.
800 #define M_ASSERTVALID(m) \
801 KASSERT((((struct mbuf *)m)->m_flags & 0) == 0, \
802 ("%s: attempted use of a free mbuf!", __func__))
805 * Set the m_data pointer of a newly-allocated mbuf (m_get/MGET) to place an
806 * object of the specified size at the end of the mbuf, longword aligned.
808 #define M_ALIGN(m, len) do { \
809 KASSERT(!((m)->m_flags & (M_PKTHDR|M_EXT)), \
810 ("%s: M_ALIGN not normal mbuf", __func__)); \
811 KASSERT((m)->m_data == (m)->m_dat, \
812 ("%s: M_ALIGN not a virgin mbuf", __func__)); \
813 (m)->m_data += (MLEN - (len)) & ~(sizeof(long) - 1); \
817 * As above, for mbufs allocated with m_gethdr/MGETHDR or initialized by
820 #define MH_ALIGN(m, len) do { \
821 KASSERT((m)->m_flags & M_PKTHDR && !((m)->m_flags & M_EXT), \
822 ("%s: MH_ALIGN not PKTHDR mbuf", __func__)); \
823 KASSERT((m)->m_data == (m)->m_pktdat, \
824 ("%s: MH_ALIGN not a virgin mbuf", __func__)); \
825 (m)->m_data += (MHLEN - (len)) & ~(sizeof(long) - 1); \
829 * As above, for mbuf with external storage.
831 #define MEXT_ALIGN(m, len) do { \
832 KASSERT((m)->m_flags & M_EXT, \
833 ("%s: MEXT_ALIGN not an M_EXT mbuf", __func__)); \
834 KASSERT((m)->m_data == (m)->m_ext.ext_buf, \
835 ("%s: MEXT_ALIGN not a virgin mbuf", __func__)); \
836 (m)->m_data += ((m)->m_ext.ext_size - (len)) & \
837 ~(sizeof(long) - 1); \
841 * Compute the amount of space available before the current start of data in
844 * The M_WRITABLE() is a temporary, conservative safety measure: the burden
845 * of checking writability of the mbuf data area rests solely with the caller.
847 #define M_LEADINGSPACE(m) \
848 ((m)->m_flags & M_EXT ? \
849 (M_WRITABLE(m) ? (m)->m_data - (m)->m_ext.ext_buf : 0): \
850 (m)->m_flags & M_PKTHDR ? (m)->m_data - (m)->m_pktdat : \
851 (m)->m_data - (m)->m_dat)
854 * Compute the amount of space available after the end of data in an mbuf.
856 * The M_WRITABLE() is a temporary, conservative safety measure: the burden
857 * of checking writability of the mbuf data area rests solely with the caller.
859 #define M_TRAILINGSPACE(m) \
860 ((m)->m_flags & M_EXT ? \
861 (M_WRITABLE(m) ? (m)->m_ext.ext_buf + (m)->m_ext.ext_size \
862 - ((m)->m_data + (m)->m_len) : 0) : \
863 &(m)->m_dat[MLEN] - ((m)->m_data + (m)->m_len))
866 * Arrange to prepend space of size plen to mbuf m. If a new mbuf must be
867 * allocated, how specifies whether to wait. If the allocation fails, the
868 * original mbuf chain is freed and m is set to NULL.
870 #define M_PREPEND(m, plen, how) do { \
871 struct mbuf **_mmp = &(m); \
872 struct mbuf *_mm = *_mmp; \
873 int _mplen = (plen); \
874 int __mhow = (how); \
876 MBUF_CHECKSLEEP(how); \
877 if (M_LEADINGSPACE(_mm) >= _mplen) { \
878 _mm->m_data -= _mplen; \
879 _mm->m_len += _mplen; \
881 _mm = m_prepend(_mm, _mplen, __mhow); \
882 if (_mm != NULL && _mm->m_flags & M_PKTHDR) \
883 _mm->m_pkthdr.len += _mplen; \
888 * Change mbuf to new type. This is a relatively expensive operation and
891 #define MCHTYPE(m, t) m_chtype((m), (t))
893 /* Length to m_copy to copy all. */
894 #define M_COPYALL 1000000000
896 /* Compatibility with 4.3. */
897 #define m_copy(m, o, l) m_copym((m), (o), (l), M_NOWAIT)
899 extern int max_datalen; /* MHLEN - max_hdr */
900 extern int max_hdr; /* Largest link + protocol header */
901 extern int max_linkhdr; /* Largest link-level header */
902 extern int max_protohdr; /* Largest protocol header */
903 extern int nmbclusters; /* Maximum number of clusters */
907 void m_adj(struct mbuf *, int);
908 void m_align(struct mbuf *, int);
909 int m_apply(struct mbuf *, int, int,
910 int (*)(void *, void *, u_int), void *);
911 int m_append(struct mbuf *, int, c_caddr_t);
912 void m_cat(struct mbuf *, struct mbuf *);
913 int m_extadd(struct mbuf *, caddr_t, u_int,
914 void (*)(struct mbuf *, void *, void *), void *, void *,
916 struct mbuf *m_collapse(struct mbuf *, int, int);
917 void m_copyback(struct mbuf *, int, int, c_caddr_t);
918 void m_copydata(const struct mbuf *, int, int, caddr_t);
919 struct mbuf *m_copym(struct mbuf *, int, int, int);
920 struct mbuf *m_copymdata(struct mbuf *, struct mbuf *,
922 struct mbuf *m_copypacket(struct mbuf *, int);
923 void m_copy_pkthdr(struct mbuf *, struct mbuf *);
924 struct mbuf *m_copyup(struct mbuf *, int, int);
925 struct mbuf *m_defrag(struct mbuf *, int);
926 void m_demote(struct mbuf *, int);
927 struct mbuf *m_devget(char *, int, int, struct ifnet *,
928 void (*)(char *, caddr_t, u_int));
929 struct mbuf *m_dup(struct mbuf *, int);
930 int m_dup_pkthdr(struct mbuf *, struct mbuf *, int);
931 u_int m_fixhdr(struct mbuf *);
932 struct mbuf *m_fragment(struct mbuf *, int, int);
933 void m_freem(struct mbuf *);
934 struct mbuf *m_get2(int, int, short, int);
935 struct mbuf *m_getjcl(int, short, int, int);
936 struct mbuf *m_getm2(struct mbuf *, int, int, short, int);
937 struct mbuf *m_getptr(struct mbuf *, int, int *);
938 u_int m_length(struct mbuf *, struct mbuf **);
939 int m_mbuftouio(struct uio *, struct mbuf *, int);
940 void m_move_pkthdr(struct mbuf *, struct mbuf *);
941 struct mbuf *m_prepend(struct mbuf *, int, int);
942 void m_print(const struct mbuf *, int);
943 struct mbuf *m_pulldown(struct mbuf *, int, int, int *);
944 struct mbuf *m_pullup(struct mbuf *, int);
945 int m_sanity(struct mbuf *, int);
946 struct mbuf *m_split(struct mbuf *, int, int);
947 struct mbuf *m_uiotombuf(struct uio *, int, int, int, int);
948 struct mbuf *m_unshare(struct mbuf *, int);
951 * Network packets may have annotations attached by affixing a list of
952 * "packet tags" to the pkthdr structure. Packet tags are dynamically
953 * allocated semi-opaque data structures that have a fixed header
954 * (struct m_tag) that specifies the size of the memory block and a
955 * <cookie,type> pair that identifies it. The cookie is a 32-bit unique
956 * unsigned value used to identify a module or ABI. By convention this value
957 * is chosen as the date+time that the module is created, expressed as the
958 * number of seconds since the epoch (e.g., using date -u +'%s'). The type
959 * value is an ABI/module-specific value that identifies a particular
960 * annotation and is private to the module. For compatibility with systems
961 * like OpenBSD that define packet tags w/o an ABI/module cookie, the value
962 * PACKET_ABI_COMPAT is used to implement m_tag_get and m_tag_find
963 * compatibility shim functions and several tag types are defined below.
964 * Users that do not require compatibility should use a private cookie value
965 * so that packet tag-related definitions can be maintained privately.
967 * Note that the packet tag returned by m_tag_alloc has the default memory
968 * alignment implemented by malloc. To reference private data one can use a
971 * struct m_tag *mtag = m_tag_alloc(...);
972 * struct foo *p = (struct foo *)(mtag+1);
974 * if the alignment of struct m_tag is sufficient for referencing members of
975 * struct foo. Otherwise it is necessary to embed struct m_tag within the
976 * private data structure to insure proper alignment; e.g.,
982 * struct foo *p = (struct foo *) m_tag_alloc(...);
983 * struct m_tag *mtag = &p->tag;
987 * Persistent tags stay with an mbuf until the mbuf is reclaimed. Otherwise
988 * tags are expected to ``vanish'' when they pass through a network
989 * interface. For most interfaces this happens normally as the tags are
990 * reclaimed when the mbuf is free'd. However in some special cases
991 * reclaiming must be done manually. An example is packets that pass through
992 * the loopback interface. Also, one must be careful to do this when
993 * ``turning around'' packets (e.g., icmp_reflect).
995 * To mark a tag persistent bit-or this flag in when defining the tag id.
996 * The tag will then be treated as described above.
998 #define MTAG_PERSISTENT 0x800
1000 #define PACKET_TAG_NONE 0 /* Nadda */
1002 /* Packet tags for use with PACKET_ABI_COMPAT. */
1003 #define PACKET_TAG_IPSEC_IN_DONE 1 /* IPsec applied, in */
1004 #define PACKET_TAG_IPSEC_OUT_DONE 2 /* IPsec applied, out */
1005 #define PACKET_TAG_IPSEC_IN_CRYPTO_DONE 3 /* NIC IPsec crypto done */
1006 #define PACKET_TAG_IPSEC_OUT_CRYPTO_NEEDED 4 /* NIC IPsec crypto req'ed */
1007 #define PACKET_TAG_IPSEC_IN_COULD_DO_CRYPTO 5 /* NIC notifies IPsec */
1008 #define PACKET_TAG_IPSEC_PENDING_TDB 6 /* Reminder to do IPsec */
1009 #define PACKET_TAG_BRIDGE 7 /* Bridge processing done */
1010 #define PACKET_TAG_GIF 8 /* GIF processing done */
1011 #define PACKET_TAG_GRE 9 /* GRE processing done */
1012 #define PACKET_TAG_IN_PACKET_CHECKSUM 10 /* NIC checksumming done */
1013 #define PACKET_TAG_ENCAP 11 /* Encap. processing */
1014 #define PACKET_TAG_IPSEC_SOCKET 12 /* IPSEC socket ref */
1015 #define PACKET_TAG_IPSEC_HISTORY 13 /* IPSEC history */
1016 #define PACKET_TAG_IPV6_INPUT 14 /* IPV6 input processing */
1017 #define PACKET_TAG_DUMMYNET 15 /* dummynet info */
1018 #define PACKET_TAG_DIVERT 17 /* divert info */
1019 #define PACKET_TAG_IPFORWARD 18 /* ipforward info */
1020 #define PACKET_TAG_MACLABEL (19 | MTAG_PERSISTENT) /* MAC label */
1021 #define PACKET_TAG_PF (21 | MTAG_PERSISTENT) /* PF/ALTQ information */
1022 #define PACKET_TAG_RTSOCKFAM 25 /* rtsock sa family */
1023 #define PACKET_TAG_IPOPTIONS 27 /* Saved IP options */
1024 #define PACKET_TAG_CARP 28 /* CARP info */
1025 #define PACKET_TAG_IPSEC_NAT_T_PORTS 29 /* two uint16_t */
1026 #define PACKET_TAG_ND_OUTGOING 30 /* ND outgoing */
1028 /* Specific cookies and tags. */
1030 /* Packet tag routines. */
1031 struct m_tag *m_tag_alloc(u_int32_t, int, int, int);
1032 void m_tag_delete(struct mbuf *, struct m_tag *);
1033 void m_tag_delete_chain(struct mbuf *, struct m_tag *);
1034 void m_tag_free_default(struct m_tag *);
1035 struct m_tag *m_tag_locate(struct mbuf *, u_int32_t, int, struct m_tag *);
1036 struct m_tag *m_tag_copy(struct m_tag *, int);
1037 int m_tag_copy_chain(struct mbuf *, struct mbuf *, int);
1038 void m_tag_delete_nonpersistent(struct mbuf *);
1041 * Initialize the list of tags associated with an mbuf.
1043 static __inline void
1044 m_tag_init(struct mbuf *m)
1047 SLIST_INIT(&m->m_pkthdr.tags);
1051 * Set up the contents of a tag. Note that this does not fill in the free
1052 * method; the caller is expected to do that.
1054 * XXX probably should be called m_tag_init, but that was already taken.
1056 static __inline void
1057 m_tag_setup(struct m_tag *t, u_int32_t cookie, int type, int len)
1062 t->m_tag_cookie = cookie;
1066 * Reclaim resources associated with a tag.
1068 static __inline void
1069 m_tag_free(struct m_tag *t)
1072 (*t->m_tag_free)(t);
1076 * Return the first tag associated with an mbuf.
1078 static __inline struct m_tag *
1079 m_tag_first(struct mbuf *m)
1082 return (SLIST_FIRST(&m->m_pkthdr.tags));
1086 * Return the next tag in the list of tags associated with an mbuf.
1088 static __inline struct m_tag *
1089 m_tag_next(struct mbuf *m, struct m_tag *t)
1092 return (SLIST_NEXT(t, m_tag_link));
1096 * Prepend a tag to the list of tags associated with an mbuf.
1098 static __inline void
1099 m_tag_prepend(struct mbuf *m, struct m_tag *t)
1102 SLIST_INSERT_HEAD(&m->m_pkthdr.tags, t, m_tag_link);
1106 * Unlink a tag from the list of tags associated with an mbuf.
1108 static __inline void
1109 m_tag_unlink(struct mbuf *m, struct m_tag *t)
1112 SLIST_REMOVE(&m->m_pkthdr.tags, t, m_tag, m_tag_link);
1115 /* These are for OpenBSD compatibility. */
1116 #define MTAG_ABI_COMPAT 0 /* compatibility ABI */
1118 static __inline struct m_tag *
1119 m_tag_get(int type, int length, int wait)
1121 return (m_tag_alloc(MTAG_ABI_COMPAT, type, length, wait));
1124 static __inline struct m_tag *
1125 m_tag_find(struct mbuf *m, int type, struct m_tag *start)
1127 return (SLIST_EMPTY(&m->m_pkthdr.tags) ? (struct m_tag *)NULL :
1128 m_tag_locate(m, MTAG_ABI_COMPAT, type, start));
1131 static __inline struct mbuf *
1132 m_free(struct mbuf *m)
1134 struct mbuf *n = m->m_next;
1136 if ((m->m_flags & (M_PKTHDR|M_NOFREE)) == (M_PKTHDR|M_NOFREE))
1137 m_tag_delete_chain(m, NULL);
1138 if (m->m_flags & M_EXT)
1140 else if ((m->m_flags & M_NOFREE) == 0)
1141 uma_zfree(zone_mbuf, m);
1146 rt_m_getfib(struct mbuf *m)
1148 KASSERT(m->m_flags & M_PKTHDR , ("Attempt to get FIB from non header mbuf."));
1149 return (m->m_pkthdr.fibnum);
1152 #define M_GETFIB(_m) rt_m_getfib(_m)
1154 #define M_SETFIB(_m, _fib) do { \
1155 KASSERT((_m)->m_flags & M_PKTHDR, ("Attempt to set FIB on non header mbuf.")); \
1156 ((_m)->m_pkthdr.fibnum) = (_fib); \
1159 #endif /* _KERNEL */
1161 #ifdef MBUF_PROFILING
1162 void m_profile(struct mbuf *m);
1163 #define M_PROFILE(m) m_profile(m)
1165 #define M_PROFILE(m)
1169 #endif /* !_SYS_MBUF_H_ */