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
178 * NOTE: forwards compatibility definitions for mbuf(9)
180 * These aren't 1:1 with the macros in r277203; in particular they're exposed
181 * to both userland and kernel, whereas this is exposed to just _KERNEL -- to
182 * avoid disruption with existing KBI/KPIs
184 #define MHSIZE offsetof(struct mbuf, m_dat)
185 #define MPKTHSIZE offsetof(struct mbuf, m_pktdat)
190 #define M_EXT 0x00000001 /* has associated external storage */
191 #define M_PKTHDR 0x00000002 /* start of record */
192 #define M_EOR 0x00000004 /* end of record */
193 #define M_RDONLY 0x00000008 /* associated data is marked read-only */
194 #define M_PROTO1 0x00000010 /* protocol-specific */
195 #define M_PROTO2 0x00000020 /* protocol-specific */
196 #define M_PROTO3 0x00000040 /* protocol-specific */
197 #define M_PROTO4 0x00000080 /* protocol-specific */
198 #define M_PROTO5 0x00000100 /* protocol-specific */
199 #define M_BCAST 0x00000200 /* send/received as link-level broadcast */
200 #define M_MCAST 0x00000400 /* send/received as link-level multicast */
201 #define M_FRAG 0x00000800 /* packet is a fragment of a larger packet */
202 #define M_FIRSTFRAG 0x00001000 /* packet is first fragment */
203 #define M_LASTFRAG 0x00002000 /* packet is last fragment */
204 #define M_SKIP_FIREWALL 0x00004000 /* skip firewall processing */
205 #define M_FREELIST 0x00008000 /* mbuf is on the free list */
206 #define M_VLANTAG 0x00010000 /* ether_vtag is valid */
207 #define M_PROMISC 0x00020000 /* packet was not for us */
208 #define M_NOFREE 0x00040000 /* do not free mbuf, embedded in cluster */
209 #define M_PROTO6 0x00080000 /* protocol-specific */
210 #define M_PROTO7 0x00100000 /* protocol-specific */
211 #define M_PROTO8 0x00200000 /* protocol-specific */
212 #define M_FLOWID 0x00400000 /* deprecated: flowid is valid */
213 #define M_HASHTYPEBITS 0x0F000000 /* mask of bits holding flowid hash type */
216 * For RELENG_{6,7} steal these flags for limited multiple routing table
217 * support. In RELENG_8 and beyond, use just one flag and a tag.
219 #define M_FIB 0xF0000000 /* steal some bits to store fib number. */
221 #define M_NOTIFICATION M_PROTO5 /* SCTP notification */
224 * Flags to purge when crossing layers.
226 #define M_PROTOFLAGS \
227 (M_PROTO1|M_PROTO2|M_PROTO3|M_PROTO4|M_PROTO5|M_PROTO6|M_PROTO7|M_PROTO8)
230 * Network interface cards are able to hash protocol fields (such as IPv4
231 * addresses and TCP port numbers) classify packets into flows. These flows
232 * can then be used to maintain ordering while delivering packets to the OS
233 * via parallel input queues, as well as to provide a stateless affinity
234 * model. NIC drivers can pass up the hash via m->m_pkthdr.flowid, and set
235 * m_flag fields to indicate how the hash should be interpreted by the
238 * Most NICs support RSS, which provides ordering and explicit affinity, and
239 * use the hash m_flag bits to indicate what header fields were covered by
240 * the hash. M_HASHTYPE_OPAQUE can be set by non-RSS cards or configurations
241 * that provide an opaque flow identifier, allowing for ordering and
242 * distribution without explicit affinity.
244 #define M_HASHTYPE_SHIFT 24
245 #define M_HASHTYPE_NONE 0x0
246 #define M_HASHTYPE_RSS_IPV4 0x1 /* IPv4 2-tuple */
247 #define M_HASHTYPE_RSS_TCP_IPV4 0x2 /* TCPv4 4-tuple */
248 #define M_HASHTYPE_RSS_IPV6 0x3 /* IPv6 2-tuple */
249 #define M_HASHTYPE_RSS_TCP_IPV6 0x4 /* TCPv6 4-tuple */
250 #define M_HASHTYPE_RSS_IPV6_EX 0x5 /* IPv6 2-tuple + ext hdrs */
251 #define M_HASHTYPE_RSS_TCP_IPV6_EX 0x6 /* TCPv6 4-tiple + ext hdrs */
252 #define M_HASHTYPE_OPAQUE 0xf /* ordering, not affinity */
254 #define M_HASHTYPE_CLEAR(m) (m)->m_flags &= ~(M_HASHTYPEBITS)
255 #define M_HASHTYPE_GET(m) (((m)->m_flags & M_HASHTYPEBITS) >> \
257 #define M_HASHTYPE_SET(m, v) do { \
258 (m)->m_flags &= ~M_HASHTYPEBITS; \
259 (m)->m_flags |= ((v) << M_HASHTYPE_SHIFT); \
261 #define M_HASHTYPE_TEST(m, v) (M_HASHTYPE_GET(m) == (v))
264 * Flags preserved when copying m_pkthdr.
266 #define M_COPYFLAGS \
267 (M_PKTHDR|M_EOR|M_RDONLY|M_PROTOFLAGS|M_SKIP_FIREWALL|M_BCAST|M_MCAST|\
268 M_FRAG|M_FIRSTFRAG|M_LASTFRAG|M_VLANTAG|M_PROMISC|M_FIB|M_HASHTYPEBITS)
271 * External buffer types: identify ext_buf type.
273 #define EXT_CLUSTER 1 /* mbuf cluster */
274 #define EXT_SFBUF 2 /* sendfile(2)'s sf_bufs */
275 #define EXT_JUMBOP 3 /* jumbo cluster 4096 bytes */
276 #define EXT_JUMBO9 4 /* jumbo cluster 9216 bytes */
277 #define EXT_JUMBO16 5 /* jumbo cluster 16184 bytes */
278 #define EXT_PACKET 6 /* mbuf+cluster from packet zone */
279 #define EXT_MBUF 7 /* external mbuf reference (M_IOVEC) */
280 #define EXT_NET_DRV 100 /* custom ext_buf provided by net driver(s) */
281 #define EXT_MOD_TYPE 200 /* custom module's ext_buf type */
282 #define EXT_DISPOSABLE 300 /* can throw this buffer away w/page flipping */
283 #define EXT_EXTREF 400 /* has externally maintained ref_cnt ptr */
286 * Flags indicating hw checksum support and sw checksum requirements. This
287 * field can be directly tested against if_data.ifi_hwassist.
289 #define CSUM_IP 0x0001 /* will csum IP */
290 #define CSUM_TCP 0x0002 /* will csum TCP */
291 #define CSUM_UDP 0x0004 /* will csum UDP */
292 #define CSUM_IP_FRAGS 0x0008 /* removed, left for compat */
293 #define CSUM_FRAGMENT 0x0010 /* will do IP fragmentation */
294 #define CSUM_TSO 0x0020 /* will do TSO */
295 #define CSUM_SCTP 0x0040 /* will csum SCTP */
296 #define CSUM_SCTP_IPV6 0x0080 /* will csum IPv6/SCTP */
298 #define CSUM_IP_CHECKED 0x0100 /* did csum IP */
299 #define CSUM_IP_VALID 0x0200 /* ... the csum is valid */
300 #define CSUM_DATA_VALID 0x0400 /* csum_data field is valid */
301 #define CSUM_PSEUDO_HDR 0x0800 /* csum_data has pseudo hdr */
302 #define CSUM_SCTP_VALID 0x1000 /* SCTP checksum is valid */
303 #define CSUM_UDP_IPV6 0x2000 /* will csum IPv6/UDP */
304 #define CSUM_TCP_IPV6 0x4000 /* will csum IPv6/TCP */
305 /* CSUM_TSO_IPV6 0x8000 will do IPv6/TSO */
307 /* CSUM_FRAGMENT_IPV6 0x10000 will do IPv6 fragementation */
309 #define CSUM_DELAY_DATA_IPV6 (CSUM_TCP_IPV6 | CSUM_UDP_IPV6)
310 #define CSUM_DATA_VALID_IPV6 CSUM_DATA_VALID
312 #define CSUM_DELAY_DATA (CSUM_TCP | CSUM_UDP)
313 #define CSUM_DELAY_IP (CSUM_IP) /* Only v4, no v6 IP hdr csum */
318 #define MT_NOTMBUF 0 /* USED INTERNALLY ONLY! Object is not mbuf */
319 #define MT_DATA 1 /* dynamic (data) allocation */
320 #define MT_HEADER MT_DATA /* packet header, use M_PKTHDR instead */
321 #define MT_SONAME 8 /* socket name */
322 #define MT_CONTROL 14 /* extra-data protocol message */
323 #define MT_OOBDATA 15 /* expedited data */
324 #define MT_NTYPES 16 /* number of mbuf types for mbtypes[] */
326 #define MT_NOINIT 255 /* Not a type but a flag to allocate
327 a non-initialized mbuf */
329 #define MB_NOTAGS 0x1UL /* no tags attached to mbuf */
332 * General mbuf allocator statistics structure.
334 * Many of these statistics are no longer used; we instead track many
335 * allocator statistics through UMA's built in statistics mechanism.
338 u_long m_mbufs; /* XXX */
339 u_long m_mclusts; /* XXX */
341 u_long m_drain; /* times drained protocols for space */
342 u_long m_mcfail; /* XXX: times m_copym failed */
343 u_long m_mpfail; /* XXX: times m_pullup failed */
344 u_long m_msize; /* length of an mbuf */
345 u_long m_mclbytes; /* length of an mbuf cluster */
346 u_long m_minclsize; /* min length of data to allocate a cluster */
347 u_long m_mlen; /* length of data in an mbuf */
348 u_long m_mhlen; /* length of data in a header mbuf */
350 /* Number of mbtypes (gives # elems in mbtypes[] array) */
353 /* XXX: Sendfile stats should eventually move to their own struct */
354 u_long sf_iocnt; /* times sendfile had to do disk I/O */
355 u_long sf_allocfail; /* times sfbuf allocation failed */
356 u_long sf_allocwait; /* times sfbuf allocation had to wait */
360 * Flags specifying how an allocation should be made.
362 * The flag to use is as follows:
363 * - M_NOWAIT (M_DONTWAIT) from an interrupt handler to not block allocation.
364 * - M_WAITOK (M_WAIT) from wherever it is safe to block.
366 * M_DONTWAIT/M_NOWAIT means that we will not block the thread explicitly and
367 * if we cannot allocate immediately we may return NULL, whereas
368 * M_WAIT/M_WAITOK means that if we cannot allocate resources we
369 * will block until they are available, and thus never return NULL.
371 * XXX Eventually just phase this out to use M_WAITOK/M_NOWAIT.
373 #define MBTOM(how) (how)
374 #define M_DONTWAIT M_NOWAIT
375 #define M_TRYWAIT M_WAITOK
376 #define M_WAIT M_WAITOK
379 * String names of mbuf-related UMA(9) and malloc(9) types. Exposed to
380 * !_KERNEL so that monitoring tools can look up the zones with
383 #define MBUF_MEM_NAME "mbuf"
384 #define MBUF_CLUSTER_MEM_NAME "mbuf_cluster"
385 #define MBUF_PACKET_MEM_NAME "mbuf_packet"
386 #define MBUF_JUMBOP_MEM_NAME "mbuf_jumbo_page"
387 #define MBUF_JUMBO9_MEM_NAME "mbuf_jumbo_9k"
388 #define MBUF_JUMBO16_MEM_NAME "mbuf_jumbo_16k"
389 #define MBUF_TAG_MEM_NAME "mbuf_tag"
390 #define MBUF_EXTREFCNT_MEM_NAME "mbuf_ext_refcnt"
395 #define MBUF_CHECKSLEEP(how) do { \
396 if (how == M_WAITOK) \
397 WITNESS_WARN(WARN_GIANTOK | WARN_SLEEPOK, NULL, \
398 "Sleeping in \"%s\"", __func__); \
401 #define MBUF_CHECKSLEEP(how)
405 * Network buffer allocation API
407 * The rest of it is defined in kern/kern_mbuf.c
409 extern uma_zone_t zone_mbuf;
410 extern uma_zone_t zone_clust;
411 extern uma_zone_t zone_pack;
412 extern uma_zone_t zone_jumbop;
413 extern uma_zone_t zone_jumbo9;
414 extern uma_zone_t zone_jumbo16;
415 extern uma_zone_t zone_ext_refcnt;
417 static __inline struct mbuf *m_getcl(int how, short type, int flags);
418 static __inline struct mbuf *m_get(int how, short type);
419 static __inline struct mbuf *m_gethdr(int how, short type);
420 static __inline struct mbuf *m_getjcl(int how, short type, int flags,
422 static __inline struct mbuf *m_getclr(int how, short type); /* XXX */
423 static __inline int m_init(struct mbuf *m, uma_zone_t zone,
424 int size, int how, short type, int flags);
425 static __inline struct mbuf *m_free(struct mbuf *m);
426 static __inline void m_clget(struct mbuf *m, int how);
427 static __inline void *m_cljget(struct mbuf *m, int how, int size);
428 static __inline void m_chtype(struct mbuf *m, short new_type);
429 void mb_free_ext(struct mbuf *);
430 static __inline struct mbuf *m_last(struct mbuf *m);
431 int m_pkthdr_init(struct mbuf *m, int how);
445 #if MJUMPAGESIZE != MCLBYTES
457 panic("%s: m_getjcl: invalid cluster size", __func__);
464 * Associated an external reference counted buffer with an mbuf.
467 m_extaddref(struct mbuf *m, caddr_t buf, u_int size, u_int *ref_cnt,
468 void (*freef)(void *, void *), void *arg1, void *arg2)
471 KASSERT(ref_cnt != NULL, ("%s: ref_cnt not provided", __func__));
473 atomic_add_int(ref_cnt, 1);
475 m->m_ext.ext_buf = buf;
476 m->m_ext.ref_cnt = ref_cnt;
477 m->m_data = m->m_ext.ext_buf;
478 m->m_ext.ext_size = size;
479 m->m_ext.ext_free = freef;
480 m->m_ext.ext_arg1 = arg1;
481 m->m_ext.ext_arg2 = arg2;
482 m->m_ext.ext_type = EXT_EXTREF;
485 static __inline uma_zone_t
497 #if MJUMPAGESIZE != MCLBYTES
509 panic("%s: m_getjcl: invalid cluster type", __func__);
516 * Initialize an mbuf with linear storage.
518 * Inline because the consumer text overhead will be roughly the same to
519 * initialize or call a function with this many parameters and M_PKTHDR
520 * should go away with constant propagation for !MGETHDR.
523 m_init(struct mbuf *m, uma_zone_t zone, int size, int how, short type,
530 m->m_data = m->m_dat;
534 if (flags & M_PKTHDR) {
535 if ((error = m_pkthdr_init(m, how)) != 0)
542 static __inline struct mbuf *
543 m_get(int how, short type)
549 return ((struct mbuf *)(uma_zalloc_arg(zone_mbuf, &args, how)));
553 * XXX This should be deprecated, very little use.
555 static __inline struct mbuf *
556 m_getclr(int how, short type)
563 m = uma_zalloc_arg(zone_mbuf, &args, how);
565 bzero(m->m_data, MLEN);
569 static __inline struct mbuf *
570 m_gethdr(int how, short type)
574 args.flags = M_PKTHDR;
576 return ((struct mbuf *)(uma_zalloc_arg(zone_mbuf, &args, how)));
579 static __inline struct mbuf *
580 m_getcl(int how, short type, int flags)
586 return ((struct mbuf *)(uma_zalloc_arg(zone_pack, &args, how)));
590 * m_getjcl() returns an mbuf with a cluster of the specified size attached.
591 * For size it takes MCLBYTES, MJUMPAGESIZE, MJUM9BYTES, MJUM16BYTES.
593 * XXX: This is rather large, should be real function maybe.
595 static __inline struct mbuf *
596 m_getjcl(int how, short type, int flags, int size)
602 if (size == MCLBYTES)
603 return m_getcl(how, type, flags);
608 m = uma_zalloc_arg(zone_mbuf, &args, how);
612 zone = m_getzone(size);
613 n = uma_zalloc_arg(zone, m, how);
615 uma_zfree(zone_mbuf, m);
622 m_free_fast(struct mbuf *m)
625 if (m->m_flags & M_PKTHDR)
626 KASSERT(SLIST_EMPTY(&m->m_pkthdr.tags), ("doing fast free of mbuf with tags"));
629 uma_zfree_arg(zone_mbuf, m, (void *)MB_NOTAGS);
632 static __inline struct mbuf *
633 m_free(struct mbuf *m)
635 struct mbuf *n = m->m_next;
637 if (m->m_flags & M_EXT)
639 else if ((m->m_flags & M_NOFREE) == 0)
640 uma_zfree(zone_mbuf, m);
645 m_clget(struct mbuf *m, int how)
648 if (m->m_flags & M_EXT)
649 printf("%s: %p mbuf already has cluster\n", __func__, m);
650 m->m_ext.ext_buf = (char *)NULL;
651 uma_zalloc_arg(zone_clust, m, how);
653 * On a cluster allocation failure, drain the packet zone and retry,
654 * we might be able to loosen a few clusters up on the drain.
656 if ((how & M_NOWAIT) && (m->m_ext.ext_buf == NULL)) {
657 zone_drain(zone_pack);
658 uma_zalloc_arg(zone_clust, m, how);
663 * m_cljget() is different from m_clget() as it can allocate clusters without
664 * attaching them to an mbuf. In that case the return value is the pointer
665 * to the cluster of the requested size. If an mbuf was specified, it gets
666 * the cluster attached to it and the return value can be safely ignored.
667 * For size it takes MCLBYTES, MJUMPAGESIZE, MJUM9BYTES, MJUM16BYTES.
669 static __inline void *
670 m_cljget(struct mbuf *m, int how, int size)
674 if (m && m->m_flags & M_EXT)
675 printf("%s: %p mbuf already has cluster\n", __func__, m);
677 m->m_ext.ext_buf = NULL;
679 zone = m_getzone(size);
680 return (uma_zalloc_arg(zone, m, how));
684 m_cljset(struct mbuf *m, void *cl, int type)
694 #if MJUMPAGESIZE != MCLBYTES
709 panic("unknown cluster type");
713 m->m_data = m->m_ext.ext_buf = cl;
714 m->m_ext.ext_free = m->m_ext.ext_arg1 = m->m_ext.ext_arg2 = NULL;
715 m->m_ext.ext_size = size;
716 m->m_ext.ext_type = type;
717 m->m_ext.ref_cnt = uma_find_refcnt(zone, cl);
723 m_chtype(struct mbuf *m, short new_type)
726 m->m_type = new_type;
729 static __inline struct mbuf *
730 m_last(struct mbuf *m)
738 extern void (*m_addr_chg_pf_p)(struct mbuf *m);
741 m_addr_changed(struct mbuf *m)
749 * mbuf, cluster, and external object allocation macros (for compatibility
752 #define M_MOVE_PKTHDR(to, from) m_move_pkthdr((to), (from))
753 #define MGET(m, how, type) ((m) = m_get((how), (type)))
754 #define MGETHDR(m, how, type) ((m) = m_gethdr((how), (type)))
755 #define MCLGET(m, how) m_clget((m), (how))
756 #define MEXTADD(m, buf, size, free, arg1, arg2, flags, type) \
757 m_extadd((m), (caddr_t)(buf), (size), (free),(arg1),(arg2),(flags), (type))
758 #define m_getm(m, len, how, type) \
759 m_getm2((m), (len), (how), (type), M_PKTHDR)
762 * Evaluate TRUE if it's safe to write to the mbuf m's data region (this can
763 * be both the local data payload, or an external buffer area, depending on
764 * whether M_EXT is set).
766 #define M_WRITABLE(m) (!((m)->m_flags & M_RDONLY) && \
767 (!(((m)->m_flags & M_EXT)) || \
768 (*((m)->m_ext.ref_cnt) == 1)) ) \
770 /* Check if the supplied mbuf has a packet header, or else panic. */
771 #define M_ASSERTPKTHDR(m) \
772 KASSERT((m) != NULL && (m)->m_flags & M_PKTHDR, \
773 ("%s: no mbuf packet header!", __func__))
776 * Ensure that the supplied mbuf is a valid, non-free mbuf.
778 * XXX: Broken at the moment. Need some UMA magic to make it work again.
780 #define M_ASSERTVALID(m) \
781 KASSERT((((struct mbuf *)m)->m_flags & 0) == 0, \
782 ("%s: attempted use of a free mbuf!", __func__))
785 * Set the m_data pointer of a newly-allocated mbuf (m_get/MGET) to place an
786 * object of the specified size at the end of the mbuf, longword aligned.
788 #define M_ALIGN(m, len) do { \
789 KASSERT(!((m)->m_flags & (M_PKTHDR|M_EXT)), \
790 ("%s: M_ALIGN not normal mbuf", __func__)); \
791 KASSERT((m)->m_data == (m)->m_dat, \
792 ("%s: M_ALIGN not a virgin mbuf", __func__)); \
793 (m)->m_data += (MLEN - (len)) & ~(sizeof(long) - 1); \
797 * As above, for mbufs allocated with m_gethdr/MGETHDR or initialized by
800 #define MH_ALIGN(m, len) do { \
801 KASSERT((m)->m_flags & M_PKTHDR && !((m)->m_flags & M_EXT), \
802 ("%s: MH_ALIGN not PKTHDR mbuf", __func__)); \
803 KASSERT((m)->m_data == (m)->m_pktdat, \
804 ("%s: MH_ALIGN not a virgin mbuf", __func__)); \
805 (m)->m_data += (MHLEN - (len)) & ~(sizeof(long) - 1); \
809 * Compute the amount of space available before the current start of data in
812 * The M_WRITABLE() is a temporary, conservative safety measure: the burden
813 * of checking writability of the mbuf data area rests solely with the caller.
815 #define M_LEADINGSPACE(m) \
816 ((m)->m_flags & M_EXT ? \
817 (M_WRITABLE(m) ? (m)->m_data - (m)->m_ext.ext_buf : 0): \
818 (m)->m_flags & M_PKTHDR ? (m)->m_data - (m)->m_pktdat : \
819 (m)->m_data - (m)->m_dat)
822 * Compute the amount of space available after the end of data in an mbuf.
824 * The M_WRITABLE() is a temporary, conservative safety measure: the burden
825 * of checking writability of the mbuf data area rests solely with the caller.
827 #define M_TRAILINGSPACE(m) \
828 ((m)->m_flags & M_EXT ? \
829 (M_WRITABLE(m) ? (m)->m_ext.ext_buf + (m)->m_ext.ext_size \
830 - ((m)->m_data + (m)->m_len) : 0) : \
831 &(m)->m_dat[MLEN] - ((m)->m_data + (m)->m_len))
834 * Arrange to prepend space of size plen to mbuf m. If a new mbuf must be
835 * allocated, how specifies whether to wait. If the allocation fails, the
836 * original mbuf chain is freed and m is set to NULL.
838 #define M_PREPEND(m, plen, how) do { \
839 struct mbuf **_mmp = &(m); \
840 struct mbuf *_mm = *_mmp; \
841 int _mplen = (plen); \
842 int __mhow = (how); \
844 MBUF_CHECKSLEEP(how); \
845 if (M_LEADINGSPACE(_mm) >= _mplen) { \
846 _mm->m_data -= _mplen; \
847 _mm->m_len += _mplen; \
849 _mm = m_prepend(_mm, _mplen, __mhow); \
850 if (_mm != NULL && _mm->m_flags & M_PKTHDR) \
851 _mm->m_pkthdr.len += _mplen; \
856 * Change mbuf to new type. This is a relatively expensive operation and
859 #define MCHTYPE(m, t) m_chtype((m), (t))
861 /* Length to m_copy to copy all. */
862 #define M_COPYALL 1000000000
864 /* Compatibility with 4.3. */
865 #define m_copy(m, o, l) m_copym((m), (o), (l), M_DONTWAIT)
867 extern int max_datalen; /* MHLEN - max_hdr */
868 extern int max_hdr; /* Largest link + protocol header */
869 extern int max_linkhdr; /* Largest link-level header */
870 extern int max_protohdr; /* Largest protocol header */
871 extern struct mbstat mbstat; /* General mbuf stats/infos */
872 extern int nmbclusters; /* Maximum number of clusters */
876 void m_adj(struct mbuf *, int);
877 void m_align(struct mbuf *, int);
878 int m_apply(struct mbuf *, int, int,
879 int (*)(void *, void *, u_int), void *);
880 int m_append(struct mbuf *, int, c_caddr_t);
881 void m_cat(struct mbuf *, struct mbuf *);
882 void m_extadd(struct mbuf *, caddr_t, u_int,
883 void (*)(void *, void *), void *, void *, int, int);
884 struct mbuf *m_collapse(struct mbuf *, int, int);
885 void m_copyback(struct mbuf *, int, int, c_caddr_t);
886 void m_copydata(const struct mbuf *, int, int, caddr_t);
887 struct mbuf *m_copym(struct mbuf *, int, int, int);
888 struct mbuf *m_copymdata(struct mbuf *, struct mbuf *,
890 struct mbuf *m_copypacket(struct mbuf *, int);
891 void m_copy_pkthdr(struct mbuf *, struct mbuf *);
892 struct mbuf *m_copyup(struct mbuf *n, int len, int dstoff);
893 struct mbuf *m_defrag(struct mbuf *, int);
894 void m_demote(struct mbuf *, int);
895 struct mbuf *m_devget(char *, int, int, struct ifnet *,
896 void (*)(char *, caddr_t, u_int));
897 struct mbuf *m_dup(struct mbuf *, int);
898 int m_dup_pkthdr(struct mbuf *, struct mbuf *, int);
899 u_int m_fixhdr(struct mbuf *);
900 struct mbuf *m_fragment(struct mbuf *, int, int);
901 void m_freem(struct mbuf *);
902 struct mbuf *m_getm2(struct mbuf *, int, int, short, int);
903 struct mbuf *m_getptr(struct mbuf *, int, int *);
904 u_int m_length(struct mbuf *, struct mbuf **);
905 int m_mbuftouio(struct uio *, struct mbuf *, int);
906 void m_move_pkthdr(struct mbuf *, struct mbuf *);
907 struct mbuf *m_prepend(struct mbuf *, int, int);
908 void m_print(const struct mbuf *, int);
909 struct mbuf *m_pulldown(struct mbuf *, int, int, int *);
910 struct mbuf *m_pullup(struct mbuf *, int);
911 int m_sanity(struct mbuf *, int);
912 struct mbuf *m_split(struct mbuf *, int, int);
913 struct mbuf *m_uiotombuf(struct uio *, int, int, int, int);
914 struct mbuf *m_unshare(struct mbuf *, int how);
917 * Network packets may have annotations attached by affixing a list of
918 * "packet tags" to the pkthdr structure. Packet tags are dynamically
919 * allocated semi-opaque data structures that have a fixed header
920 * (struct m_tag) that specifies the size of the memory block and a
921 * <cookie,type> pair that identifies it. The cookie is a 32-bit unique
922 * unsigned value used to identify a module or ABI. By convention this value
923 * is chosen as the date+time that the module is created, expressed as the
924 * number of seconds since the epoch (e.g., using date -u +'%s'). The type
925 * value is an ABI/module-specific value that identifies a particular
926 * annotation and is private to the module. For compatibility with systems
927 * like OpenBSD that define packet tags w/o an ABI/module cookie, the value
928 * PACKET_ABI_COMPAT is used to implement m_tag_get and m_tag_find
929 * compatibility shim functions and several tag types are defined below.
930 * Users that do not require compatibility should use a private cookie value
931 * so that packet tag-related definitions can be maintained privately.
933 * Note that the packet tag returned by m_tag_alloc has the default memory
934 * alignment implemented by malloc. To reference private data one can use a
937 * struct m_tag *mtag = m_tag_alloc(...);
938 * struct foo *p = (struct foo *)(mtag+1);
940 * if the alignment of struct m_tag is sufficient for referencing members of
941 * struct foo. Otherwise it is necessary to embed struct m_tag within the
942 * private data structure to insure proper alignment; e.g.,
948 * struct foo *p = (struct foo *) m_tag_alloc(...);
949 * struct m_tag *mtag = &p->tag;
953 * Persistent tags stay with an mbuf until the mbuf is reclaimed. Otherwise
954 * tags are expected to ``vanish'' when they pass through a network
955 * interface. For most interfaces this happens normally as the tags are
956 * reclaimed when the mbuf is free'd. However in some special cases
957 * reclaiming must be done manually. An example is packets that pass through
958 * the loopback interface. Also, one must be careful to do this when
959 * ``turning around'' packets (e.g., icmp_reflect).
961 * To mark a tag persistent bit-or this flag in when defining the tag id.
962 * The tag will then be treated as described above.
964 #define MTAG_PERSISTENT 0x800
966 #define PACKET_TAG_NONE 0 /* Nadda */
968 /* Packet tags for use with PACKET_ABI_COMPAT. */
969 #define PACKET_TAG_IPSEC_IN_DONE 1 /* IPsec applied, in */
970 #define PACKET_TAG_IPSEC_OUT_DONE 2 /* IPsec applied, out */
971 #define PACKET_TAG_IPSEC_IN_CRYPTO_DONE 3 /* NIC IPsec crypto done */
972 #define PACKET_TAG_IPSEC_OUT_CRYPTO_NEEDED 4 /* NIC IPsec crypto req'ed */
973 #define PACKET_TAG_IPSEC_IN_COULD_DO_CRYPTO 5 /* NIC notifies IPsec */
974 #define PACKET_TAG_IPSEC_PENDING_TDB 6 /* Reminder to do IPsec */
975 #define PACKET_TAG_BRIDGE 7 /* Bridge processing done */
976 #define PACKET_TAG_GIF 8 /* GIF processing done */
977 #define PACKET_TAG_GRE 9 /* GRE processing done */
978 #define PACKET_TAG_IN_PACKET_CHECKSUM 10 /* NIC checksumming done */
979 #define PACKET_TAG_ENCAP 11 /* Encap. processing */
980 #define PACKET_TAG_IPSEC_SOCKET 12 /* IPSEC socket ref */
981 #define PACKET_TAG_IPSEC_HISTORY 13 /* IPSEC history */
982 #define PACKET_TAG_IPV6_INPUT 14 /* IPV6 input processing */
983 #define PACKET_TAG_DUMMYNET 15 /* dummynet info */
984 #define PACKET_TAG_DIVERT 17 /* divert info */
985 #define PACKET_TAG_IPFORWARD 18 /* ipforward info */
986 #define PACKET_TAG_MACLABEL (19 | MTAG_PERSISTENT) /* MAC label */
987 #define PACKET_TAG_PF 21 /* PF + ALTQ information */
988 #define PACKET_TAG_RTSOCKFAM 25 /* rtsock sa family */
989 #define PACKET_TAG_IPOPTIONS 27 /* Saved IP options */
990 #define PACKET_TAG_CARP 28 /* CARP info */
991 #define PACKET_TAG_IPSEC_NAT_T_PORTS 29 /* two uint16_t */
992 #define PACKET_TAG_ND_OUTGOING 30 /* ND outgoing */
994 /* Specific cookies and tags. */
996 /* Packet tag routines. */
997 struct m_tag *m_tag_alloc(u_int32_t, int, int, int);
998 void m_tag_delete(struct mbuf *, struct m_tag *);
999 void m_tag_delete_chain(struct mbuf *, struct m_tag *);
1000 void m_tag_free_default(struct m_tag *);
1001 struct m_tag *m_tag_locate(struct mbuf *, u_int32_t, int, struct m_tag *);
1002 struct m_tag *m_tag_copy(struct m_tag *, int);
1003 int m_tag_copy_chain(struct mbuf *, struct mbuf *, int);
1004 void m_tag_delete_nonpersistent(struct mbuf *);
1007 * Initialize the list of tags associated with an mbuf.
1009 static __inline void
1010 m_tag_init(struct mbuf *m)
1013 SLIST_INIT(&m->m_pkthdr.tags);
1017 * Set up the contents of a tag. Note that this does not fill in the free
1018 * method; the caller is expected to do that.
1020 * XXX probably should be called m_tag_init, but that was already taken.
1022 static __inline void
1023 m_tag_setup(struct m_tag *t, u_int32_t cookie, int type, int len)
1028 t->m_tag_cookie = cookie;
1032 * Reclaim resources associated with a tag.
1034 static __inline void
1035 m_tag_free(struct m_tag *t)
1038 (*t->m_tag_free)(t);
1042 * Return the first tag associated with an mbuf.
1044 static __inline struct m_tag *
1045 m_tag_first(struct mbuf *m)
1048 return (SLIST_FIRST(&m->m_pkthdr.tags));
1052 * Return the next tag in the list of tags associated with an mbuf.
1054 static __inline struct m_tag *
1055 m_tag_next(struct mbuf *m, struct m_tag *t)
1058 return (SLIST_NEXT(t, m_tag_link));
1062 * Prepend a tag to the list of tags associated with an mbuf.
1064 static __inline void
1065 m_tag_prepend(struct mbuf *m, struct m_tag *t)
1068 SLIST_INSERT_HEAD(&m->m_pkthdr.tags, t, m_tag_link);
1072 * Unlink a tag from the list of tags associated with an mbuf.
1074 static __inline void
1075 m_tag_unlink(struct mbuf *m, struct m_tag *t)
1078 SLIST_REMOVE(&m->m_pkthdr.tags, t, m_tag, m_tag_link);
1081 /* These are for OpenBSD compatibility. */
1082 #define MTAG_ABI_COMPAT 0 /* compatibility ABI */
1084 static __inline struct m_tag *
1085 m_tag_get(int type, int length, int wait)
1087 return (m_tag_alloc(MTAG_ABI_COMPAT, type, length, wait));
1090 static __inline struct m_tag *
1091 m_tag_find(struct mbuf *m, int type, struct m_tag *start)
1093 return (SLIST_EMPTY(&m->m_pkthdr.tags) ? (struct m_tag *)NULL :
1094 m_tag_locate(m, MTAG_ABI_COMPAT, type, start));
1097 /* XXX temporary FIB methods probably eventually use tags.*/
1098 #define M_FIBSHIFT 28
1099 #define M_FIBMASK 0x0F
1101 /* get the fib from an mbuf and if it is not set, return the default */
1102 #define M_GETFIB(_m) \
1103 ((((_m)->m_flags & M_FIB) >> M_FIBSHIFT) & M_FIBMASK)
1105 #define M_SETFIB(_m, _fib) do { \
1106 _m->m_flags &= ~M_FIB; \
1107 _m->m_flags |= (((_fib) << M_FIBSHIFT) & M_FIB); \
1110 #endif /* _KERNEL */
1112 #ifdef MBUF_PROFILING
1113 void m_profile(struct mbuf *m);
1114 #define M_PROFILE(m) m_profile(m)
1116 #define M_PROFILE(m)
1120 #endif /* !_SYS_MBUF_H_ */