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 #define MLEN (MSIZE - sizeof(struct m_hdr)) /* normal data len */
57 #define MHLEN (MLEN - sizeof(struct pkthdr)) /* data len w/pkthdr */
58 #define MINCLSIZE (MHLEN + 1) /* smallest amount to put in cluster */
59 #define M_MAXCOMPRESS (MHLEN / 2) /* max amount to copy for compression */
63 * Macro for type conversion: convert mbuf pointer to data pointer of correct
66 * mtod(m, t) -- Convert mbuf pointer to data pointer of correct type.
68 #define mtod(m, t) ((t)((m)->m_data))
71 * Argument structure passed to UMA routines during mbuf and packet
75 int flags; /* Flags for mbuf being allocated */
76 short type; /* Type of mbuf being allocated */
87 * Header present at the beginning of every mbuf.
90 struct mbuf *mh_next; /* next buffer in chain */
91 struct mbuf *mh_nextpkt; /* next chain in queue/record */
92 caddr_t mh_data; /* location of data */
93 int mh_len; /* amount of data in this mbuf */
94 int mh_flags; /* flags; see below */
95 short mh_type; /* type of data in this mbuf */
96 uint8_t pad[M_HDR_PAD];/* word align */
100 * Packet tag structure (see below for details).
103 SLIST_ENTRY(m_tag) m_tag_link; /* List of packet tags */
104 u_int16_t m_tag_id; /* Tag ID */
105 u_int16_t m_tag_len; /* Length of data */
106 u_int32_t m_tag_cookie; /* ABI/Module ID */
107 void (*m_tag_free)(struct m_tag *);
111 * Record/packet header in first mbuf of chain; valid only if M_PKTHDR is set.
114 struct ifnet *rcvif; /* rcv interface */
115 /* variables for ip and tcp reassembly */
116 void *header; /* pointer to packet header */
117 int len; /* total packet length */
118 uint32_t flowid; /* packet's 4-tuple system
121 /* variables for hardware checksum */
122 int csum_flags; /* flags regarding checksum */
123 int csum_data; /* data field used by csum routines */
124 u_int16_t tso_segsz; /* TSO segment size */
126 u_int16_t vt_vtag; /* Ethernet 802.1p+q vlan tag */
127 u_int16_t vt_nrecs; /* # of IGMPv3 records in this chain */
129 SLIST_HEAD(packet_tags, m_tag) tags; /* list of packet tags */
131 #define ether_vtag PH_vt.vt_vtag
134 * Description of external storage mapped into mbuf; valid only if M_EXT is
138 caddr_t ext_buf; /* start of buffer */
139 void (*ext_free) /* free routine if not the usual */
141 void *ext_arg1; /* optional argument pointer */
142 void *ext_arg2; /* optional argument pointer */
143 u_int ext_size; /* size of buffer, for ext_free */
144 volatile u_int *ref_cnt; /* pointer to ref count info */
145 int ext_type; /* type of external storage */
149 * The core of the mbuf object along with some shortcut defines for practical
156 struct pkthdr MH_pkthdr; /* M_PKTHDR set */
158 struct m_ext MH_ext; /* M_EXT set */
159 char MH_databuf[MHLEN];
162 char M_databuf[MLEN]; /* !M_PKTHDR, !M_EXT */
165 #define m_next m_hdr.mh_next
166 #define m_len m_hdr.mh_len
167 #define m_data m_hdr.mh_data
168 #define m_type m_hdr.mh_type
169 #define m_flags m_hdr.mh_flags
170 #define m_nextpkt m_hdr.mh_nextpkt
171 #define m_act m_nextpkt
172 #define m_pkthdr M_dat.MH.MH_pkthdr
173 #define m_ext M_dat.MH.MH_dat.MH_ext
174 #define m_pktdat M_dat.MH.MH_dat.MH_databuf
175 #define m_dat M_dat.M_databuf
180 #define M_EXT 0x00000001 /* has associated external storage */
181 #define M_PKTHDR 0x00000002 /* start of record */
182 #define M_EOR 0x00000004 /* end of record */
183 #define M_RDONLY 0x00000008 /* associated data is marked read-only */
184 #define M_PROTO1 0x00000010 /* protocol-specific */
185 #define M_PROTO2 0x00000020 /* protocol-specific */
186 #define M_PROTO3 0x00000040 /* protocol-specific */
187 #define M_PROTO4 0x00000080 /* protocol-specific */
188 #define M_PROTO5 0x00000100 /* protocol-specific */
189 #define M_BCAST 0x00000200 /* send/received as link-level broadcast */
190 #define M_MCAST 0x00000400 /* send/received as link-level multicast */
191 #define M_FRAG 0x00000800 /* packet is a fragment of a larger packet */
192 #define M_FIRSTFRAG 0x00001000 /* packet is first fragment */
193 #define M_LASTFRAG 0x00002000 /* packet is last fragment */
194 #define M_SKIP_FIREWALL 0x00004000 /* skip firewall processing */
195 #define M_FREELIST 0x00008000 /* mbuf is on the free list */
196 #define M_VLANTAG 0x00010000 /* ether_vtag is valid */
197 #define M_PROMISC 0x00020000 /* packet was not for us */
198 #define M_NOFREE 0x00040000 /* do not free mbuf, embedded in cluster */
199 #define M_PROTO6 0x00080000 /* protocol-specific */
200 #define M_PROTO7 0x00100000 /* protocol-specific */
201 #define M_PROTO8 0x00200000 /* protocol-specific */
202 #define M_FLOWID 0x00400000 /* deprecated: flowid is valid */
203 #define M_HASHTYPEBITS 0x0F000000 /* mask of bits holding flowid hash type */
206 * For RELENG_{6,7} steal these flags for limited multiple routing table
207 * support. In RELENG_8 and beyond, use just one flag and a tag.
209 #define M_FIB 0xF0000000 /* steal some bits to store fib number. */
211 #define M_NOTIFICATION M_PROTO5 /* SCTP notification */
214 * Flags to purge when crossing layers.
216 #define M_PROTOFLAGS \
217 (M_PROTO1|M_PROTO2|M_PROTO3|M_PROTO4|M_PROTO5|M_PROTO6|M_PROTO7|M_PROTO8)
220 * Network interface cards are able to hash protocol fields (such as IPv4
221 * addresses and TCP port numbers) classify packets into flows. These flows
222 * can then be used to maintain ordering while delivering packets to the OS
223 * via parallel input queues, as well as to provide a stateless affinity
224 * model. NIC drivers can pass up the hash via m->m_pkthdr.flowid, and set
225 * m_flag fields to indicate how the hash should be interpreted by the
228 * Most NICs support RSS, which provides ordering and explicit affinity, and
229 * use the hash m_flag bits to indicate what header fields were covered by
230 * the hash. M_HASHTYPE_OPAQUE can be set by non-RSS cards or configurations
231 * that provide an opaque flow identifier, allowing for ordering and
232 * distribution without explicit affinity.
234 #define M_HASHTYPE_SHIFT 24
235 #define M_HASHTYPE_NONE 0x0
236 #define M_HASHTYPE_RSS_IPV4 0x1 /* IPv4 2-tuple */
237 #define M_HASHTYPE_RSS_TCP_IPV4 0x2 /* TCPv4 4-tuple */
238 #define M_HASHTYPE_RSS_IPV6 0x3 /* IPv6 2-tuple */
239 #define M_HASHTYPE_RSS_TCP_IPV6 0x4 /* TCPv6 4-tuple */
240 #define M_HASHTYPE_RSS_IPV6_EX 0x5 /* IPv6 2-tuple + ext hdrs */
241 #define M_HASHTYPE_RSS_TCP_IPV6_EX 0x6 /* TCPv6 4-tiple + ext hdrs */
242 #define M_HASHTYPE_OPAQUE 0xf /* ordering, not affinity */
244 #define M_HASHTYPE_CLEAR(m) (m)->m_flags &= ~(M_HASHTYPEBITS)
245 #define M_HASHTYPE_GET(m) (((m)->m_flags & M_HASHTYPEBITS) >> \
247 #define M_HASHTYPE_SET(m, v) do { \
248 (m)->m_flags &= ~M_HASHTYPEBITS; \
249 (m)->m_flags |= ((v) << M_HASHTYPE_SHIFT); \
251 #define M_HASHTYPE_TEST(m, v) (M_HASHTYPE_GET(m) == (v))
254 * Flags preserved when copying m_pkthdr.
256 #define M_COPYFLAGS \
257 (M_PKTHDR|M_EOR|M_RDONLY|M_PROTOFLAGS|M_SKIP_FIREWALL|M_BCAST|M_MCAST|\
258 M_FRAG|M_FIRSTFRAG|M_LASTFRAG|M_VLANTAG|M_PROMISC|M_FIB|M_HASHTYPEBITS)
261 * External buffer types: identify ext_buf type.
263 #define EXT_CLUSTER 1 /* mbuf cluster */
264 #define EXT_SFBUF 2 /* sendfile(2)'s sf_bufs */
265 #define EXT_JUMBOP 3 /* jumbo cluster 4096 bytes */
266 #define EXT_JUMBO9 4 /* jumbo cluster 9216 bytes */
267 #define EXT_JUMBO16 5 /* jumbo cluster 16184 bytes */
268 #define EXT_PACKET 6 /* mbuf+cluster from packet zone */
269 #define EXT_MBUF 7 /* external mbuf reference (M_IOVEC) */
270 #define EXT_NET_DRV 100 /* custom ext_buf provided by net driver(s) */
271 #define EXT_MOD_TYPE 200 /* custom module's ext_buf type */
272 #define EXT_DISPOSABLE 300 /* can throw this buffer away w/page flipping */
273 #define EXT_EXTREF 400 /* has externally maintained ref_cnt ptr */
276 * Flags indicating hw checksum support and sw checksum requirements. This
277 * field can be directly tested against if_data.ifi_hwassist.
279 #define CSUM_IP 0x0001 /* will csum IP */
280 #define CSUM_TCP 0x0002 /* will csum TCP */
281 #define CSUM_UDP 0x0004 /* will csum UDP */
282 #define CSUM_IP_FRAGS 0x0008 /* will csum IP fragments */
283 #define CSUM_FRAGMENT 0x0010 /* will do IP fragmentation */
284 #define CSUM_TSO 0x0020 /* will do TSO */
285 #define CSUM_SCTP 0x0040 /* will csum SCTP */
286 #define CSUM_SCTP_IPV6 0x0080 /* will csum IPv6/SCTP */
288 #define CSUM_IP_CHECKED 0x0100 /* did csum IP */
289 #define CSUM_IP_VALID 0x0200 /* ... the csum is valid */
290 #define CSUM_DATA_VALID 0x0400 /* csum_data field is valid */
291 #define CSUM_PSEUDO_HDR 0x0800 /* csum_data has pseudo hdr */
292 #define CSUM_SCTP_VALID 0x1000 /* SCTP checksum is valid */
293 #define CSUM_UDP_IPV6 0x2000 /* will csum IPv6/UDP */
294 #define CSUM_TCP_IPV6 0x4000 /* will csum IPv6/TCP */
295 /* CSUM_TSO_IPV6 0x8000 will do IPv6/TSO */
297 /* CSUM_FRAGMENT_IPV6 0x10000 will do IPv6 fragementation */
299 #define CSUM_DELAY_DATA_IPV6 (CSUM_TCP_IPV6 | CSUM_UDP_IPV6)
300 #define CSUM_DATA_VALID_IPV6 CSUM_DATA_VALID
302 #define CSUM_DELAY_DATA (CSUM_TCP | CSUM_UDP)
303 #define CSUM_DELAY_IP (CSUM_IP) /* Only v4, no v6 IP hdr csum */
308 #define MT_NOTMBUF 0 /* USED INTERNALLY ONLY! Object is not mbuf */
309 #define MT_DATA 1 /* dynamic (data) allocation */
310 #define MT_HEADER MT_DATA /* packet header, use M_PKTHDR instead */
311 #define MT_SONAME 8 /* socket name */
312 #define MT_CONTROL 14 /* extra-data protocol message */
313 #define MT_OOBDATA 15 /* expedited data */
314 #define MT_NTYPES 16 /* number of mbuf types for mbtypes[] */
316 #define MT_NOINIT 255 /* Not a type but a flag to allocate
317 a non-initialized mbuf */
319 #define MB_NOTAGS 0x1UL /* no tags attached to mbuf */
322 * General mbuf allocator statistics structure.
324 * Many of these statistics are no longer used; we instead track many
325 * allocator statistics through UMA's built in statistics mechanism.
328 u_long m_mbufs; /* XXX */
329 u_long m_mclusts; /* XXX */
331 u_long m_drain; /* times drained protocols for space */
332 u_long m_mcfail; /* XXX: times m_copym failed */
333 u_long m_mpfail; /* XXX: times m_pullup failed */
334 u_long m_msize; /* length of an mbuf */
335 u_long m_mclbytes; /* length of an mbuf cluster */
336 u_long m_minclsize; /* min length of data to allocate a cluster */
337 u_long m_mlen; /* length of data in an mbuf */
338 u_long m_mhlen; /* length of data in a header mbuf */
340 /* Number of mbtypes (gives # elems in mbtypes[] array) */
343 /* XXX: Sendfile stats should eventually move to their own struct */
344 u_long sf_iocnt; /* times sendfile had to do disk I/O */
345 u_long sf_allocfail; /* times sfbuf allocation failed */
346 u_long sf_allocwait; /* times sfbuf allocation had to wait */
350 * Flags specifying how an allocation should be made.
352 * The flag to use is as follows:
353 * - M_DONTWAIT or M_NOWAIT from an interrupt handler to not block allocation.
354 * - M_WAIT or M_WAITOK from wherever it is safe to block.
356 * M_DONTWAIT/M_NOWAIT means that we will not block the thread explicitly and
357 * if we cannot allocate immediately we may return NULL, whereas
358 * M_WAIT/M_WAITOK means that if we cannot allocate resources we
359 * will block until they are available, and thus never return NULL.
361 * XXX Eventually just phase this out to use M_WAITOK/M_NOWAIT.
363 #define MBTOM(how) (how)
364 #define M_DONTWAIT M_NOWAIT
365 #define M_TRYWAIT M_WAITOK
366 #define M_WAIT M_WAITOK
369 * String names of mbuf-related UMA(9) and malloc(9) types. Exposed to
370 * !_KERNEL so that monitoring tools can look up the zones with
373 #define MBUF_MEM_NAME "mbuf"
374 #define MBUF_CLUSTER_MEM_NAME "mbuf_cluster"
375 #define MBUF_PACKET_MEM_NAME "mbuf_packet"
376 #define MBUF_JUMBOP_MEM_NAME "mbuf_jumbo_page"
377 #define MBUF_JUMBO9_MEM_NAME "mbuf_jumbo_9k"
378 #define MBUF_JUMBO16_MEM_NAME "mbuf_jumbo_16k"
379 #define MBUF_TAG_MEM_NAME "mbuf_tag"
380 #define MBUF_EXTREFCNT_MEM_NAME "mbuf_ext_refcnt"
385 #define MBUF_CHECKSLEEP(how) do { \
386 if (how == M_WAITOK) \
387 WITNESS_WARN(WARN_GIANTOK | WARN_SLEEPOK, NULL, \
388 "Sleeping in \"%s\"", __func__); \
391 #define MBUF_CHECKSLEEP(how)
395 * Network buffer allocation API
397 * The rest of it is defined in kern/kern_mbuf.c
400 extern uma_zone_t zone_mbuf;
401 extern uma_zone_t zone_clust;
402 extern uma_zone_t zone_pack;
403 extern uma_zone_t zone_jumbop;
404 extern uma_zone_t zone_jumbo9;
405 extern uma_zone_t zone_jumbo16;
406 extern uma_zone_t zone_ext_refcnt;
408 static __inline struct mbuf *m_getcl(int how, short type, int flags);
409 static __inline struct mbuf *m_get(int how, short type);
410 static __inline struct mbuf *m_get2(int how, short type, int flags,
412 static __inline struct mbuf *m_gethdr(int how, short type);
413 static __inline struct mbuf *m_getjcl(int how, short type, int flags,
415 static __inline struct mbuf *m_getclr(int how, short type); /* XXX */
416 static __inline int m_init(struct mbuf *m, uma_zone_t zone,
417 int size, int how, short type, int flags);
418 static __inline struct mbuf *m_free(struct mbuf *m);
419 static __inline void m_clget(struct mbuf *m, int how);
420 static __inline void *m_cljget(struct mbuf *m, int how, int size);
421 static __inline void m_chtype(struct mbuf *m, short new_type);
422 void mb_free_ext(struct mbuf *);
423 static __inline struct mbuf *m_last(struct mbuf *m);
424 int m_pkthdr_init(struct mbuf *m, int how);
438 #if MJUMPAGESIZE != MCLBYTES
450 panic("%s: invalid cluster size", __func__);
456 static __inline uma_zone_t
465 #if MJUMPAGESIZE != MCLBYTES
477 panic("%s: invalid cluster size", __func__);
484 * Initialize an mbuf with linear storage.
486 * Inline because the consumer text overhead will be roughly the same to
487 * initialize or call a function with this many parameters and M_PKTHDR
488 * should go away with constant propagation for !MGETHDR.
491 m_init(struct mbuf *m, uma_zone_t zone, int size, int how, short type,
498 m->m_data = m->m_dat;
502 if (flags & M_PKTHDR) {
503 if ((error = m_pkthdr_init(m, how)) != 0)
510 static __inline struct mbuf *
511 m_get(int how, short type)
517 return ((struct mbuf *)(uma_zalloc_arg(zone_mbuf, &args, how)));
521 * XXX This should be deprecated, very little use.
523 static __inline struct mbuf *
524 m_getclr(int how, short type)
531 m = uma_zalloc_arg(zone_mbuf, &args, how);
533 bzero(m->m_data, MLEN);
537 static __inline struct mbuf *
538 m_gethdr(int how, short type)
542 args.flags = M_PKTHDR;
544 return ((struct mbuf *)(uma_zalloc_arg(zone_mbuf, &args, how)));
547 static __inline struct mbuf *
548 m_getcl(int how, short type, int flags)
554 return ((struct mbuf *)(uma_zalloc_arg(zone_pack, &args, how)));
558 * m_get2() allocates minimum mbuf that would fit "size" argument.
560 * XXX: This is rather large, should be real function maybe.
562 static __inline struct mbuf *
563 m_get2(int how, short type, int flags, int size)
572 if (size <= MHLEN || (size <= MLEN && (flags & M_PKTHDR) == 0))
573 return ((struct mbuf *)(uma_zalloc_arg(zone_mbuf, &args, how)));
574 if (size <= MCLBYTES)
575 return ((struct mbuf *)(uma_zalloc_arg(zone_pack, &args, how)));
577 if (size > MJUM16BYTES)
580 m = uma_zalloc_arg(zone_mbuf, &args, how);
584 #if MJUMPAGESIZE != MCLBYTES
585 if (size <= MJUMPAGESIZE)
589 if (size <= MJUM9BYTES)
594 n = uma_zalloc_arg(zone, m, how);
596 uma_zfree(zone_mbuf, m);
604 * m_getjcl() returns an mbuf with a cluster of the specified size attached.
605 * For size it takes MCLBYTES, MJUMPAGESIZE, MJUM9BYTES, MJUM16BYTES.
607 * XXX: This is rather large, should be real function maybe.
609 static __inline struct mbuf *
610 m_getjcl(int how, short type, int flags, int size)
616 if (size == MCLBYTES)
617 return m_getcl(how, type, flags);
622 m = uma_zalloc_arg(zone_mbuf, &args, how);
626 zone = m_getzone(size);
627 n = uma_zalloc_arg(zone, m, how);
629 uma_zfree(zone_mbuf, m);
636 m_free_fast(struct mbuf *m)
639 if (m->m_flags & M_PKTHDR)
640 KASSERT(SLIST_EMPTY(&m->m_pkthdr.tags), ("doing fast free of mbuf with tags"));
643 uma_zfree_arg(zone_mbuf, m, (void *)MB_NOTAGS);
646 static __inline struct mbuf *
647 m_free(struct mbuf *m)
649 struct mbuf *n = m->m_next;
651 if (m->m_flags & M_EXT)
653 else if ((m->m_flags & M_NOFREE) == 0)
654 uma_zfree(zone_mbuf, m);
659 m_clget(struct mbuf *m, int how)
662 if (m->m_flags & M_EXT)
663 printf("%s: %p mbuf already has cluster\n", __func__, m);
664 m->m_ext.ext_buf = (char *)NULL;
665 uma_zalloc_arg(zone_clust, m, how);
667 * On a cluster allocation failure, drain the packet zone and retry,
668 * we might be able to loosen a few clusters up on the drain.
670 if ((how & M_NOWAIT) && (m->m_ext.ext_buf == NULL)) {
671 zone_drain(zone_pack);
672 uma_zalloc_arg(zone_clust, m, how);
677 * m_cljget() is different from m_clget() as it can allocate clusters without
678 * attaching them to an mbuf. In that case the return value is the pointer
679 * to the cluster of the requested size. If an mbuf was specified, it gets
680 * the cluster attached to it and the return value can be safely ignored.
681 * For size it takes MCLBYTES, MJUMPAGESIZE, MJUM9BYTES, MJUM16BYTES.
683 static __inline void *
684 m_cljget(struct mbuf *m, int how, int size)
688 if (m && m->m_flags & M_EXT)
689 printf("%s: %p mbuf already has cluster\n", __func__, m);
691 m->m_ext.ext_buf = NULL;
693 zone = m_getzone(size);
694 return (uma_zalloc_arg(zone, m, how));
698 m_cljset(struct mbuf *m, void *cl, int type)
708 #if MJUMPAGESIZE != MCLBYTES
723 panic("%s: unknown cluster type", __func__);
727 m->m_data = m->m_ext.ext_buf = cl;
728 m->m_ext.ext_free = m->m_ext.ext_arg1 = m->m_ext.ext_arg2 = NULL;
729 m->m_ext.ext_size = size;
730 m->m_ext.ext_type = type;
731 m->m_ext.ref_cnt = uma_find_refcnt(zone, cl);
737 m_chtype(struct mbuf *m, short new_type)
740 m->m_type = new_type;
743 static __inline struct mbuf *
744 m_last(struct mbuf *m)
752 extern void (*m_addr_chg_pf_p)(struct mbuf *m);
755 m_addr_changed(struct mbuf *m)
763 * mbuf, cluster, and external object allocation macros (for compatibility
766 #define M_MOVE_PKTHDR(to, from) m_move_pkthdr((to), (from))
767 #define MGET(m, how, type) ((m) = m_get((how), (type)))
768 #define MGETHDR(m, how, type) ((m) = m_gethdr((how), (type)))
769 #define MCLGET(m, how) m_clget((m), (how))
770 #define MEXTADD(m, buf, size, free, arg1, arg2, flags, type) \
771 m_extadd((m), (caddr_t)(buf), (size), (free),(arg1),(arg2),(flags), (type))
772 #define m_getm(m, len, how, type) \
773 m_getm2((m), (len), (how), (type), M_PKTHDR)
776 * Evaluate TRUE if it's safe to write to the mbuf m's data region (this can
777 * be both the local data payload, or an external buffer area, depending on
778 * whether M_EXT is set).
780 #define M_WRITABLE(m) (!((m)->m_flags & M_RDONLY) && \
781 (!(((m)->m_flags & M_EXT)) || \
782 (*((m)->m_ext.ref_cnt) == 1)) ) \
784 /* Check if the supplied mbuf has a packet header, or else panic. */
785 #define M_ASSERTPKTHDR(m) \
786 KASSERT((m) != NULL && (m)->m_flags & M_PKTHDR, \
787 ("%s: no mbuf packet header!", __func__))
790 * Ensure that the supplied mbuf is a valid, non-free mbuf.
792 * XXX: Broken at the moment. Need some UMA magic to make it work again.
794 #define M_ASSERTVALID(m) \
795 KASSERT((((struct mbuf *)m)->m_flags & 0) == 0, \
796 ("%s: attempted use of a free mbuf!", __func__))
799 * Set the m_data pointer of a newly-allocated mbuf (m_get/MGET) to place an
800 * object of the specified size at the end of the mbuf, longword aligned.
802 #define M_ALIGN(m, len) do { \
803 KASSERT(!((m)->m_flags & (M_PKTHDR|M_EXT)), \
804 ("%s: M_ALIGN not normal mbuf", __func__)); \
805 KASSERT((m)->m_data == (m)->m_dat, \
806 ("%s: M_ALIGN not a virgin mbuf", __func__)); \
807 (m)->m_data += (MLEN - (len)) & ~(sizeof(long) - 1); \
811 * As above, for mbufs allocated with m_gethdr/MGETHDR or initialized by
814 #define MH_ALIGN(m, len) do { \
815 KASSERT((m)->m_flags & M_PKTHDR && !((m)->m_flags & M_EXT), \
816 ("%s: MH_ALIGN not PKTHDR mbuf", __func__)); \
817 KASSERT((m)->m_data == (m)->m_pktdat, \
818 ("%s: MH_ALIGN not a virgin mbuf", __func__)); \
819 (m)->m_data += (MHLEN - (len)) & ~(sizeof(long) - 1); \
823 * Compute the amount of space available before the current start of data in
826 * The M_WRITABLE() is a temporary, conservative safety measure: the burden
827 * of checking writability of the mbuf data area rests solely with the caller.
829 #define M_LEADINGSPACE(m) \
830 ((m)->m_flags & M_EXT ? \
831 (M_WRITABLE(m) ? (m)->m_data - (m)->m_ext.ext_buf : 0): \
832 (m)->m_flags & M_PKTHDR ? (m)->m_data - (m)->m_pktdat : \
833 (m)->m_data - (m)->m_dat)
836 * Compute the amount of space available after the end of data in an mbuf.
838 * The M_WRITABLE() is a temporary, conservative safety measure: the burden
839 * of checking writability of the mbuf data area rests solely with the caller.
841 #define M_TRAILINGSPACE(m) \
842 ((m)->m_flags & M_EXT ? \
843 (M_WRITABLE(m) ? (m)->m_ext.ext_buf + (m)->m_ext.ext_size \
844 - ((m)->m_data + (m)->m_len) : 0) : \
845 &(m)->m_dat[MLEN] - ((m)->m_data + (m)->m_len))
848 * Arrange to prepend space of size plen to mbuf m. If a new mbuf must be
849 * allocated, how specifies whether to wait. If the allocation fails, the
850 * original mbuf chain is freed and m is set to NULL.
852 #define M_PREPEND(m, plen, how) do { \
853 struct mbuf **_mmp = &(m); \
854 struct mbuf *_mm = *_mmp; \
855 int _mplen = (plen); \
856 int __mhow = (how); \
858 MBUF_CHECKSLEEP(how); \
859 if (M_LEADINGSPACE(_mm) >= _mplen) { \
860 _mm->m_data -= _mplen; \
861 _mm->m_len += _mplen; \
863 _mm = m_prepend(_mm, _mplen, __mhow); \
864 if (_mm != NULL && _mm->m_flags & M_PKTHDR) \
865 _mm->m_pkthdr.len += _mplen; \
870 * Change mbuf to new type. This is a relatively expensive operation and
873 #define MCHTYPE(m, t) m_chtype((m), (t))
875 /* Length to m_copy to copy all. */
876 #define M_COPYALL 1000000000
878 /* Compatibility with 4.3. */
879 #define m_copy(m, o, l) m_copym((m), (o), (l), M_DONTWAIT)
881 extern int max_datalen; /* MHLEN - max_hdr */
882 extern int max_hdr; /* Largest link + protocol header */
883 extern int max_linkhdr; /* Largest link-level header */
884 extern int max_protohdr; /* Largest protocol header */
885 extern struct mbstat mbstat; /* General mbuf stats/infos */
886 extern int nmbclusters; /* Maximum number of clusters */
890 void m_adj(struct mbuf *, int);
891 void m_align(struct mbuf *, int);
892 int m_apply(struct mbuf *, int, int,
893 int (*)(void *, void *, u_int), void *);
894 int m_append(struct mbuf *, int, c_caddr_t);
895 void m_cat(struct mbuf *, struct mbuf *);
896 void m_extadd(struct mbuf *, caddr_t, u_int,
897 void (*)(void *, void *), void *, void *, int, int);
898 struct mbuf *m_collapse(struct mbuf *, int, int);
899 void m_copyback(struct mbuf *, int, int, c_caddr_t);
900 void m_copydata(const struct mbuf *, int, int, caddr_t);
901 struct mbuf *m_copym(struct mbuf *, int, int, int);
902 struct mbuf *m_copymdata(struct mbuf *, struct mbuf *,
904 struct mbuf *m_copypacket(struct mbuf *, int);
905 void m_copy_pkthdr(struct mbuf *, struct mbuf *);
906 struct mbuf *m_copyup(struct mbuf *n, int len, int dstoff);
907 struct mbuf *m_defrag(struct mbuf *, int);
908 void m_demote(struct mbuf *, int);
909 struct mbuf *m_devget(char *, int, int, struct ifnet *,
910 void (*)(char *, caddr_t, u_int));
911 struct mbuf *m_dup(struct mbuf *, int);
912 int m_dup_pkthdr(struct mbuf *, struct mbuf *, int);
913 u_int m_fixhdr(struct mbuf *);
914 struct mbuf *m_fragment(struct mbuf *, int, int);
915 void m_freem(struct mbuf *);
916 struct mbuf *m_getm2(struct mbuf *, int, int, short, int);
917 struct mbuf *m_getptr(struct mbuf *, int, int *);
918 u_int m_length(struct mbuf *, struct mbuf **);
919 int m_mbuftouio(struct uio *, struct mbuf *, int);
920 void m_move_pkthdr(struct mbuf *, struct mbuf *);
921 struct mbuf *m_prepend(struct mbuf *, int, int);
922 void m_print(const struct mbuf *, int);
923 struct mbuf *m_pulldown(struct mbuf *, int, int, int *);
924 struct mbuf *m_pullup(struct mbuf *, int);
925 int m_sanity(struct mbuf *, int);
926 struct mbuf *m_split(struct mbuf *, int, int);
927 struct mbuf *m_uiotombuf(struct uio *, int, int, int, int);
928 struct mbuf *m_unshare(struct mbuf *, int how);
931 * Network packets may have annotations attached by affixing a list of
932 * "packet tags" to the pkthdr structure. Packet tags are dynamically
933 * allocated semi-opaque data structures that have a fixed header
934 * (struct m_tag) that specifies the size of the memory block and a
935 * <cookie,type> pair that identifies it. The cookie is a 32-bit unique
936 * unsigned value used to identify a module or ABI. By convention this value
937 * is chosen as the date+time that the module is created, expressed as the
938 * number of seconds since the epoch (e.g., using date -u +'%s'). The type
939 * value is an ABI/module-specific value that identifies a particular
940 * annotation and is private to the module. For compatibility with systems
941 * like OpenBSD that define packet tags w/o an ABI/module cookie, the value
942 * PACKET_ABI_COMPAT is used to implement m_tag_get and m_tag_find
943 * compatibility shim functions and several tag types are defined below.
944 * Users that do not require compatibility should use a private cookie value
945 * so that packet tag-related definitions can be maintained privately.
947 * Note that the packet tag returned by m_tag_alloc has the default memory
948 * alignment implemented by malloc. To reference private data one can use a
951 * struct m_tag *mtag = m_tag_alloc(...);
952 * struct foo *p = (struct foo *)(mtag+1);
954 * if the alignment of struct m_tag is sufficient for referencing members of
955 * struct foo. Otherwise it is necessary to embed struct m_tag within the
956 * private data structure to insure proper alignment; e.g.,
962 * struct foo *p = (struct foo *) m_tag_alloc(...);
963 * struct m_tag *mtag = &p->tag;
967 * Persistent tags stay with an mbuf until the mbuf is reclaimed. Otherwise
968 * tags are expected to ``vanish'' when they pass through a network
969 * interface. For most interfaces this happens normally as the tags are
970 * reclaimed when the mbuf is free'd. However in some special cases
971 * reclaiming must be done manually. An example is packets that pass through
972 * the loopback interface. Also, one must be careful to do this when
973 * ``turning around'' packets (e.g., icmp_reflect).
975 * To mark a tag persistent bit-or this flag in when defining the tag id.
976 * The tag will then be treated as described above.
978 #define MTAG_PERSISTENT 0x800
980 #define PACKET_TAG_NONE 0 /* Nadda */
982 /* Packet tags for use with PACKET_ABI_COMPAT. */
983 #define PACKET_TAG_IPSEC_IN_DONE 1 /* IPsec applied, in */
984 #define PACKET_TAG_IPSEC_OUT_DONE 2 /* IPsec applied, out */
985 #define PACKET_TAG_IPSEC_IN_CRYPTO_DONE 3 /* NIC IPsec crypto done */
986 #define PACKET_TAG_IPSEC_OUT_CRYPTO_NEEDED 4 /* NIC IPsec crypto req'ed */
987 #define PACKET_TAG_IPSEC_IN_COULD_DO_CRYPTO 5 /* NIC notifies IPsec */
988 #define PACKET_TAG_IPSEC_PENDING_TDB 6 /* Reminder to do IPsec */
989 #define PACKET_TAG_BRIDGE 7 /* Bridge processing done */
990 #define PACKET_TAG_GIF 8 /* GIF processing done */
991 #define PACKET_TAG_GRE 9 /* GRE processing done */
992 #define PACKET_TAG_IN_PACKET_CHECKSUM 10 /* NIC checksumming done */
993 #define PACKET_TAG_ENCAP 11 /* Encap. processing */
994 #define PACKET_TAG_IPSEC_SOCKET 12 /* IPSEC socket ref */
995 #define PACKET_TAG_IPSEC_HISTORY 13 /* IPSEC history */
996 #define PACKET_TAG_IPV6_INPUT 14 /* IPV6 input processing */
997 #define PACKET_TAG_DUMMYNET 15 /* dummynet info */
998 #define PACKET_TAG_DIVERT 17 /* divert info */
999 #define PACKET_TAG_IPFORWARD 18 /* ipforward info */
1000 #define PACKET_TAG_MACLABEL (19 | MTAG_PERSISTENT) /* MAC label */
1001 #define PACKET_TAG_PF 21 /* PF + ALTQ information */
1002 #define PACKET_TAG_RTSOCKFAM 25 /* rtsock sa family */
1003 #define PACKET_TAG_IPOPTIONS 27 /* Saved IP options */
1004 #define PACKET_TAG_CARP 28 /* CARP info */
1005 #define PACKET_TAG_IPSEC_NAT_T_PORTS 29 /* two uint16_t */
1006 #define PACKET_TAG_ND_OUTGOING 30 /* ND outgoing */
1008 /* Specific cookies and tags. */
1010 /* Packet tag routines. */
1011 struct m_tag *m_tag_alloc(u_int32_t, int, int, int);
1012 void m_tag_delete(struct mbuf *, struct m_tag *);
1013 void m_tag_delete_chain(struct mbuf *, struct m_tag *);
1014 void m_tag_free_default(struct m_tag *);
1015 struct m_tag *m_tag_locate(struct mbuf *, u_int32_t, int, struct m_tag *);
1016 struct m_tag *m_tag_copy(struct m_tag *, int);
1017 int m_tag_copy_chain(struct mbuf *, struct mbuf *, int);
1018 void m_tag_delete_nonpersistent(struct mbuf *);
1021 * Initialize the list of tags associated with an mbuf.
1023 static __inline void
1024 m_tag_init(struct mbuf *m)
1027 SLIST_INIT(&m->m_pkthdr.tags);
1031 * Set up the contents of a tag. Note that this does not fill in the free
1032 * method; the caller is expected to do that.
1034 * XXX probably should be called m_tag_init, but that was already taken.
1036 static __inline void
1037 m_tag_setup(struct m_tag *t, u_int32_t cookie, int type, int len)
1042 t->m_tag_cookie = cookie;
1046 * Reclaim resources associated with a tag.
1048 static __inline void
1049 m_tag_free(struct m_tag *t)
1052 (*t->m_tag_free)(t);
1056 * Return the first tag associated with an mbuf.
1058 static __inline struct m_tag *
1059 m_tag_first(struct mbuf *m)
1062 return (SLIST_FIRST(&m->m_pkthdr.tags));
1066 * Return the next tag in the list of tags associated with an mbuf.
1068 static __inline struct m_tag *
1069 m_tag_next(struct mbuf *m, struct m_tag *t)
1072 return (SLIST_NEXT(t, m_tag_link));
1076 * Prepend a tag to the list of tags associated with an mbuf.
1078 static __inline void
1079 m_tag_prepend(struct mbuf *m, struct m_tag *t)
1082 SLIST_INSERT_HEAD(&m->m_pkthdr.tags, t, m_tag_link);
1086 * Unlink a tag from the list of tags associated with an mbuf.
1088 static __inline void
1089 m_tag_unlink(struct mbuf *m, struct m_tag *t)
1092 SLIST_REMOVE(&m->m_pkthdr.tags, t, m_tag, m_tag_link);
1095 /* These are for OpenBSD compatibility. */
1096 #define MTAG_ABI_COMPAT 0 /* compatibility ABI */
1098 static __inline struct m_tag *
1099 m_tag_get(int type, int length, int wait)
1101 return (m_tag_alloc(MTAG_ABI_COMPAT, type, length, wait));
1104 static __inline struct m_tag *
1105 m_tag_find(struct mbuf *m, int type, struct m_tag *start)
1107 return (SLIST_EMPTY(&m->m_pkthdr.tags) ? (struct m_tag *)NULL :
1108 m_tag_locate(m, MTAG_ABI_COMPAT, type, start));
1111 /* XXX temporary FIB methods probably eventually use tags.*/
1112 #define M_FIBSHIFT 28
1113 #define M_FIBMASK 0x0F
1115 /* get the fib from an mbuf and if it is not set, return the default */
1116 #define M_GETFIB(_m) \
1117 ((((_m)->m_flags & M_FIB) >> M_FIBSHIFT) & M_FIBMASK)
1119 #define M_SETFIB(_m, _fib) do { \
1120 _m->m_flags &= ~M_FIB; \
1121 _m->m_flags |= (((_fib) << M_FIBSHIFT) & M_FIB); \
1124 #endif /* _KERNEL */
1126 #ifdef MBUF_PROFILING
1127 void m_profile(struct mbuf *m);
1128 #define M_PROFILE(m) m_profile(m)
1130 #define M_PROFILE(m)
1134 #endif /* !_SYS_MBUF_H_ */