1/*-
2 * SPDX-License-Identifier: BSD-3-Clause
3 *
4 * Copyright (c) 1982, 1986, 1988, 1993
5 *	The Regents of the University of California.
6 * All rights reserved.
7 *
8 * Redistribution and use in source and binary forms, with or without
9 * modification, are permitted provided that the following conditions
10 * are met:
11 * 1. Redistributions of source code must retain the above copyright
12 *    notice, this list of conditions and the following disclaimer.
13 * 2. Redistributions in binary form must reproduce the above copyright
14 *    notice, this list of conditions and the following disclaimer in the
15 *    documentation and/or other materials provided with the distribution.
16 * 3. Neither the name of the University nor the names of its contributors
17 *    may be used to endorse or promote products derived from this software
18 *    without specific prior written permission.
19 *
20 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
21 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
22 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
23 * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
24 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
25 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
26 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
27 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
28 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
29 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
30 * SUCH DAMAGE.
31 *
32 *	@(#)mbuf.h	8.5 (Berkeley) 2/19/95
33 * $FreeBSD$
34 */
35
36#ifndef _SYS_MBUF_H_
37#define	_SYS_MBUF_H_
38
39/* XXX: These includes suck. Sorry! */
40#include <sys/queue.h>
41#ifdef _KERNEL
42#include <sys/systm.h>
43#include <vm/uma.h>
44#ifdef WITNESS
45#include <sys/lock.h>
46#endif
47#endif
48
49#ifdef _KERNEL
50#include <sys/sdt.h>
51
52#define	MBUF_PROBE1(probe, arg0)					\
53	SDT_PROBE1(sdt, , , probe, arg0)
54#define	MBUF_PROBE2(probe, arg0, arg1)					\
55	SDT_PROBE2(sdt, , , probe, arg0, arg1)
56#define	MBUF_PROBE3(probe, arg0, arg1, arg2)				\
57	SDT_PROBE3(sdt, , , probe, arg0, arg1, arg2)
58#define	MBUF_PROBE4(probe, arg0, arg1, arg2, arg3)			\
59	SDT_PROBE4(sdt, , , probe, arg0, arg1, arg2, arg3)
60#define	MBUF_PROBE5(probe, arg0, arg1, arg2, arg3, arg4)		\
61	SDT_PROBE5(sdt, , , probe, arg0, arg1, arg2, arg3, arg4)
62
63SDT_PROBE_DECLARE(sdt, , , m__init);
64SDT_PROBE_DECLARE(sdt, , , m__gethdr);
65SDT_PROBE_DECLARE(sdt, , , m__get);
66SDT_PROBE_DECLARE(sdt, , , m__getcl);
67SDT_PROBE_DECLARE(sdt, , , m__clget);
68SDT_PROBE_DECLARE(sdt, , , m__cljget);
69SDT_PROBE_DECLARE(sdt, , , m__cljset);
70SDT_PROBE_DECLARE(sdt, , , m__free);
71SDT_PROBE_DECLARE(sdt, , , m__freem);
72
73#endif /* _KERNEL */
74
75/*
76 * Mbufs are of a single size, MSIZE (sys/param.h), which includes overhead.
77 * An mbuf may add a single "mbuf cluster" of size MCLBYTES (also in
78 * sys/param.h), which has no additional overhead and is used instead of the
79 * internal data area; this is done when at least MINCLSIZE of data must be
80 * stored.  Additionally, it is possible to allocate a separate buffer
81 * externally and attach it to the mbuf in a way similar to that of mbuf
82 * clusters.
83 *
84 * NB: These calculation do not take actual compiler-induced alignment and
85 * padding inside the complete struct mbuf into account.  Appropriate
86 * attention is required when changing members of struct mbuf.
87 *
88 * MLEN is data length in a normal mbuf.
89 * MHLEN is data length in an mbuf with pktheader.
90 * MINCLSIZE is a smallest amount of data that should be put into cluster.
91 *
92 * Compile-time assertions in uipc_mbuf.c test these values to ensure that
93 * they are sensible.
94 */
95struct mbuf;
96#define	MHSIZE		offsetof(struct mbuf, m_dat)
97#define	MPKTHSIZE	offsetof(struct mbuf, m_pktdat)
98#define	MLEN		((int)(MSIZE - MHSIZE))
99#define	MHLEN		((int)(MSIZE - MPKTHSIZE))
100#define	MINCLSIZE	(MHLEN + 1)
101
102#ifdef _KERNEL
103/*-
104 * Macro for type conversion: convert mbuf pointer to data pointer of correct
105 * type:
106 *
107 * mtod(m, t)	-- Convert mbuf pointer to data pointer of correct type.
108 * mtodo(m, o) -- Same as above but with offset 'o' into data.
109 */
110#define	mtod(m, t)	((t)((m)->m_data))
111#define	mtodo(m, o)	((void *)(((m)->m_data) + (o)))
112
113/*
114 * Argument structure passed to UMA routines during mbuf and packet
115 * allocations.
116 */
117struct mb_args {
118	int	flags;	/* Flags for mbuf being allocated */
119	short	type;	/* Type of mbuf being allocated */
120};
121#endif /* _KERNEL */
122
123/*
124 * Packet tag structure (see below for details).
125 */
126struct m_tag {
127	SLIST_ENTRY(m_tag)	m_tag_link;	/* List of packet tags */
128	u_int16_t		m_tag_id;	/* Tag ID */
129	u_int16_t		m_tag_len;	/* Length of data */
130	u_int32_t		m_tag_cookie;	/* ABI/Module ID */
131	void			(*m_tag_free)(struct m_tag *);
132};
133
134/*
135 * Static network interface owned tag.
136 * Allocated through ifp->if_snd_tag_alloc().
137 */
138struct m_snd_tag {
139	struct ifnet *ifp;		/* network interface tag belongs to */
140};
141
142/*
143 * Record/packet header in first mbuf of chain; valid only if M_PKTHDR is set.
144 * Size ILP32: 48
145 *	 LP64: 56
146 * Compile-time assertions in uipc_mbuf.c test these values to ensure that
147 * they are correct.
148 */
149struct pkthdr {
150	union {
151		struct m_snd_tag *snd_tag;	/* send tag, if any */
152		struct ifnet	*rcvif;		/* rcv interface */
153	};
154	SLIST_HEAD(packet_tags, m_tag) tags; /* list of packet tags */
155	int32_t		 len;		/* total packet length */
156
157	/* Layer crossing persistent information. */
158	uint32_t	 flowid;	/* packet's 4-tuple system */
159	uint32_t	 csum_flags;	/* checksum and offload features */
160	uint16_t	 fibnum;	/* this packet should use this fib */
161	uint8_t		 cosqos;	/* class/quality of service */
162	uint8_t		 rsstype;	/* hash type */
163	union {
164		uint64_t	rcv_tstmp;	/* timestamp in ns */
165		struct {
166			uint8_t		 l2hlen;	/* layer 2 hdr len */
167			uint8_t		 l3hlen;	/* layer 3 hdr len */
168			uint8_t		 l4hlen;	/* layer 4 hdr len */
169			uint8_t		 l5hlen;	/* layer 5 hdr len */
170			uint32_t	 spare;
171		};
172	};
173	union {
174		uint8_t  eight[8];
175		uint16_t sixteen[4];
176		uint32_t thirtytwo[2];
177		uint64_t sixtyfour[1];
178		uintptr_t unintptr[1];
179		void	*ptr;
180	} PH_per;
181
182	/* Layer specific non-persistent local storage for reassembly, etc. */
183	union {
184		uint8_t  eight[8];
185		uint16_t sixteen[4];
186		uint32_t thirtytwo[2];
187		uint64_t sixtyfour[1];
188		uintptr_t unintptr[1];
189		void 	*ptr;
190	} PH_loc;
191};
192#define	ether_vtag	PH_per.sixteen[0]
193#define	PH_vt		PH_per
194#define	vt_nrecs	sixteen[0]
195#define	tso_segsz	PH_per.sixteen[1]
196#define	lro_nsegs	tso_segsz
197#define	csum_phsum	PH_per.sixteen[2]
198#define	csum_data	PH_per.thirtytwo[1]
199#define pace_thoff	PH_loc.sixteen[0]
200#define pace_tlen	PH_loc.sixteen[1]
201#define pace_drphdrlen	PH_loc.sixteen[2]
202#define pace_tos	PH_loc.eight[6]
203#define pace_lock	PH_loc.eight[7]
204
205/*
206 * Description of external storage mapped into mbuf; valid only if M_EXT is
207 * set.
208 * Size ILP32: 28
209 *	 LP64: 48
210 * Compile-time assertions in uipc_mbuf.c test these values to ensure that
211 * they are correct.
212 */
213typedef	void m_ext_free_t(struct mbuf *);
214struct m_ext {
215	union {
216		/*
217		 * If EXT_FLAG_EMBREF is set, then we use refcount in the
218		 * mbuf, the 'ext_count' member.  Otherwise, we have a
219		 * shadow copy and we use pointer 'ext_cnt'.  The original
220		 * mbuf is responsible to carry the pointer to free routine
221		 * and its arguments.  They aren't copied into shadows in
222		 * mb_dupcl() to avoid dereferencing next cachelines.
223		 */
224		volatile u_int	 ext_count;
225		volatile u_int	*ext_cnt;
226	};
227	char		*ext_buf;	/* start of buffer */
228	uint32_t	 ext_size;	/* size of buffer, for ext_free */
229	uint32_t	 ext_type:8,	/* type of external storage */
230			 ext_flags:24;	/* external storage mbuf flags */
231	/*
232	 * Fields below store the free context for the external storage.
233	 * They are valid only in the refcount carrying mbuf, the one with
234	 * EXT_FLAG_EMBREF flag, with exclusion for EXT_EXTREF type, where
235	 * the free context is copied into all mbufs that use same external
236	 * storage.
237	 */
238#define	m_ext_copylen	offsetof(struct m_ext, ext_free)
239	m_ext_free_t	*ext_free;	/* free routine if not the usual */
240	void		*ext_arg1;	/* optional argument pointer */
241	void		*ext_arg2;	/* optional argument pointer */
242};
243
244/*
245 * The core of the mbuf object along with some shortcut defines for practical
246 * purposes.
247 */
248struct mbuf {
249	/*
250	 * Header present at the beginning of every mbuf.
251	 * Size ILP32: 24
252	 *      LP64: 32
253	 * Compile-time assertions in uipc_mbuf.c test these values to ensure
254	 * that they are correct.
255	 */
256	union {	/* next buffer in chain */
257		struct mbuf		*m_next;
258		SLIST_ENTRY(mbuf)	m_slist;
259		STAILQ_ENTRY(mbuf)	m_stailq;
260	};
261	union {	/* next chain in queue/record */
262		struct mbuf		*m_nextpkt;
263		SLIST_ENTRY(mbuf)	m_slistpkt;
264		STAILQ_ENTRY(mbuf)	m_stailqpkt;
265	};
266	caddr_t		 m_data;	/* location of data */
267	int32_t		 m_len;		/* amount of data in this mbuf */
268	uint32_t	 m_type:8,	/* type of data in this mbuf */
269			 m_flags:24;	/* flags; see below */
270#if !defined(__LP64__)
271	uint32_t	 m_pad;		/* pad for 64bit alignment */
272#endif
273
274	/*
275	 * A set of optional headers (packet header, external storage header)
276	 * and internal data storage.  Historically, these arrays were sized
277	 * to MHLEN (space left after a packet header) and MLEN (space left
278	 * after only a regular mbuf header); they are now variable size in
279	 * order to support future work on variable-size mbufs.
280	 */
281	union {
282		struct {
283			struct pkthdr	m_pkthdr;	/* M_PKTHDR set */
284			union {
285				struct m_ext	m_ext;	/* M_EXT set */
286				char		m_pktdat[0];
287			};
288		};
289		char	m_dat[0];			/* !M_PKTHDR, !M_EXT */
290	};
291};
292
293/*
294 * mbuf flags of global significance and layer crossing.
295 * Those of only protocol/layer specific significance are to be mapped
296 * to M_PROTO[1-12] and cleared at layer handoff boundaries.
297 * NB: Limited to the lower 24 bits.
298 */
299#define	M_EXT		0x00000001 /* has associated external storage */
300#define	M_PKTHDR	0x00000002 /* start of record */
301#define	M_EOR		0x00000004 /* end of record */
302#define	M_RDONLY	0x00000008 /* associated data is marked read-only */
303#define	M_BCAST		0x00000010 /* send/received as link-level broadcast */
304#define	M_MCAST		0x00000020 /* send/received as link-level multicast */
305#define	M_PROMISC	0x00000040 /* packet was not for us */
306#define	M_VLANTAG	0x00000080 /* ether_vtag is valid */
307#define	M_NOMAP		0x00000100 /* mbuf data is unmapped (soon from Drew) */
308#define	M_NOFREE	0x00000200 /* do not free mbuf, embedded in cluster */
309#define	M_TSTMP		0x00000400 /* rcv_tstmp field is valid */
310#define	M_TSTMP_HPREC	0x00000800 /* rcv_tstmp is high-prec, typically
311				      hw-stamped on port (useful for IEEE 1588
312				      and 802.1AS) */
313
314#define	M_PROTO1	0x00001000 /* protocol-specific */
315#define	M_PROTO2	0x00002000 /* protocol-specific */
316#define	M_PROTO3	0x00004000 /* protocol-specific */
317#define	M_PROTO4	0x00008000 /* protocol-specific */
318#define	M_PROTO5	0x00010000 /* protocol-specific */
319#define	M_PROTO6	0x00020000 /* protocol-specific */
320#define	M_PROTO7	0x00040000 /* protocol-specific */
321#define	M_PROTO8	0x00080000 /* protocol-specific */
322#define	M_PROTO9	0x00100000 /* protocol-specific */
323#define	M_PROTO10	0x00200000 /* protocol-specific */
324#define	M_PROTO11	0x00400000 /* protocol-specific */
325#define	M_PROTO12	0x00800000 /* protocol-specific */
326
327#define MB_DTOR_SKIP	0x1	/* don't pollute the cache by touching a freed mbuf */
328
329/*
330 * Flags to purge when crossing layers.
331 */
332#define	M_PROTOFLAGS \
333    (M_PROTO1|M_PROTO2|M_PROTO3|M_PROTO4|M_PROTO5|M_PROTO6|M_PROTO7|M_PROTO8|\
334     M_PROTO9|M_PROTO10|M_PROTO11|M_PROTO12)
335
336/*
337 * Flags preserved when copying m_pkthdr.
338 */
339#define M_COPYFLAGS \
340    (M_PKTHDR|M_EOR|M_RDONLY|M_BCAST|M_MCAST|M_PROMISC|M_VLANTAG|M_TSTMP| \
341     M_TSTMP_HPREC|M_PROTOFLAGS)
342
343/*
344 * Mbuf flag description for use with printf(9) %b identifier.
345 */
346#define	M_FLAG_BITS \
347    "\20\1M_EXT\2M_PKTHDR\3M_EOR\4M_RDONLY\5M_BCAST\6M_MCAST" \
348    "\7M_PROMISC\10M_VLANTAG\13M_TSTMP\14M_TSTMP_HPREC"
349#define	M_FLAG_PROTOBITS \
350    "\15M_PROTO1\16M_PROTO2\17M_PROTO3\20M_PROTO4\21M_PROTO5" \
351    "\22M_PROTO6\23M_PROTO7\24M_PROTO8\25M_PROTO9\26M_PROTO10" \
352    "\27M_PROTO11\30M_PROTO12"
353#define	M_FLAG_PRINTF (M_FLAG_BITS M_FLAG_PROTOBITS)
354
355/*
356 * Network interface cards are able to hash protocol fields (such as IPv4
357 * addresses and TCP port numbers) classify packets into flows.  These flows
358 * can then be used to maintain ordering while delivering packets to the OS
359 * via parallel input queues, as well as to provide a stateless affinity
360 * model.  NIC drivers can pass up the hash via m->m_pkthdr.flowid, and set
361 * m_flag fields to indicate how the hash should be interpreted by the
362 * network stack.
363 *
364 * Most NICs support RSS, which provides ordering and explicit affinity, and
365 * use the hash m_flag bits to indicate what header fields were covered by
366 * the hash.  M_HASHTYPE_OPAQUE and M_HASHTYPE_OPAQUE_HASH can be set by non-
367 * RSS cards or configurations that provide an opaque flow identifier, allowing
368 * for ordering and distribution without explicit affinity.  Additionally,
369 * M_HASHTYPE_OPAQUE_HASH indicates that the flow identifier has hash
370 * properties.
371 *
372 * The meaning of the IPV6_EX suffix:
373 * "o  Home address from the home address option in the IPv6 destination
374 *     options header.  If the extension header is not present, use the Source
375 *     IPv6 Address.
376 *  o  IPv6 address that is contained in the Routing-Header-Type-2 from the
377 *     associated extension header.  If the extension header is not present,
378 *     use the Destination IPv6 Address."
379 * Quoted from:
380 * https://docs.microsoft.com/en-us/windows-hardware/drivers/network/rss-hashing-types#ndishashipv6ex
381 */
382#define	M_HASHTYPE_HASHPROP		0x80	/* has hash properties */
383#define	M_HASHTYPE_HASH(t)		(M_HASHTYPE_HASHPROP | (t))
384/* Microsoft RSS standard hash types */
385#define	M_HASHTYPE_NONE			0
386#define	M_HASHTYPE_RSS_IPV4		M_HASHTYPE_HASH(1) /* IPv4 2-tuple */
387#define	M_HASHTYPE_RSS_TCP_IPV4		M_HASHTYPE_HASH(2) /* TCPv4 4-tuple */
388#define	M_HASHTYPE_RSS_IPV6		M_HASHTYPE_HASH(3) /* IPv6 2-tuple */
389#define	M_HASHTYPE_RSS_TCP_IPV6		M_HASHTYPE_HASH(4) /* TCPv6 4-tuple */
390#define	M_HASHTYPE_RSS_IPV6_EX		M_HASHTYPE_HASH(5) /* IPv6 2-tuple +
391							    * ext hdrs */
392#define	M_HASHTYPE_RSS_TCP_IPV6_EX	M_HASHTYPE_HASH(6) /* TCPv6 4-tuple +
393							    * ext hdrs */
394#define	M_HASHTYPE_RSS_UDP_IPV4		M_HASHTYPE_HASH(7) /* IPv4 UDP 4-tuple*/
395#define	M_HASHTYPE_RSS_UDP_IPV6		M_HASHTYPE_HASH(9) /* IPv6 UDP 4-tuple*/
396#define	M_HASHTYPE_RSS_UDP_IPV6_EX	M_HASHTYPE_HASH(10)/* IPv6 UDP 4-tuple +
397							    * ext hdrs */
398
399#define	M_HASHTYPE_OPAQUE		63	/* ordering, not affinity */
400#define	M_HASHTYPE_OPAQUE_HASH		M_HASHTYPE_HASH(M_HASHTYPE_OPAQUE)
401						/* ordering+hash, not affinity*/
402
403#define	M_HASHTYPE_CLEAR(m)	((m)->m_pkthdr.rsstype = 0)
404#define	M_HASHTYPE_GET(m)	((m)->m_pkthdr.rsstype)
405#define	M_HASHTYPE_SET(m, v)	((m)->m_pkthdr.rsstype = (v))
406#define	M_HASHTYPE_TEST(m, v)	(M_HASHTYPE_GET(m) == (v))
407#define	M_HASHTYPE_ISHASH(m)	(M_HASHTYPE_GET(m) & M_HASHTYPE_HASHPROP)
408
409/*
410 * COS/QOS class and quality of service tags.
411 * It uses DSCP code points as base.
412 */
413#define	QOS_DSCP_CS0		0x00
414#define	QOS_DSCP_DEF		QOS_DSCP_CS0
415#define	QOS_DSCP_CS1		0x20
416#define	QOS_DSCP_AF11		0x28
417#define	QOS_DSCP_AF12		0x30
418#define	QOS_DSCP_AF13		0x38
419#define	QOS_DSCP_CS2		0x40
420#define	QOS_DSCP_AF21		0x48
421#define	QOS_DSCP_AF22		0x50
422#define	QOS_DSCP_AF23		0x58
423#define	QOS_DSCP_CS3		0x60
424#define	QOS_DSCP_AF31		0x68
425#define	QOS_DSCP_AF32		0x70
426#define	QOS_DSCP_AF33		0x78
427#define	QOS_DSCP_CS4		0x80
428#define	QOS_DSCP_AF41		0x88
429#define	QOS_DSCP_AF42		0x90
430#define	QOS_DSCP_AF43		0x98
431#define	QOS_DSCP_CS5		0xa0
432#define	QOS_DSCP_EF		0xb8
433#define	QOS_DSCP_CS6		0xc0
434#define	QOS_DSCP_CS7		0xe0
435
436/*
437 * External mbuf storage buffer types.
438 */
439#define	EXT_CLUSTER	1	/* mbuf cluster */
440#define	EXT_SFBUF	2	/* sendfile(2)'s sf_buf */
441#define	EXT_JUMBOP	3	/* jumbo cluster page sized */
442#define	EXT_JUMBO9	4	/* jumbo cluster 9216 bytes */
443#define	EXT_JUMBO16	5	/* jumbo cluster 16184 bytes */
444#define	EXT_PACKET	6	/* mbuf+cluster from packet zone */
445#define	EXT_MBUF	7	/* external mbuf reference */
446
447#define	EXT_VENDOR1	224	/* for vendor-internal use */
448#define	EXT_VENDOR2	225	/* for vendor-internal use */
449#define	EXT_VENDOR3	226	/* for vendor-internal use */
450#define	EXT_VENDOR4	227	/* for vendor-internal use */
451
452#define	EXT_EXP1	244	/* for experimental use */
453#define	EXT_EXP2	245	/* for experimental use */
454#define	EXT_EXP3	246	/* for experimental use */
455#define	EXT_EXP4	247	/* for experimental use */
456
457#define	EXT_NET_DRV	252	/* custom ext_buf provided by net driver(s) */
458#define	EXT_MOD_TYPE	253	/* custom module's ext_buf type */
459#define	EXT_DISPOSABLE	254	/* can throw this buffer away w/page flipping */
460#define	EXT_EXTREF	255	/* has externally maintained ext_cnt ptr */
461
462/*
463 * Flags for external mbuf buffer types.
464 * NB: limited to the lower 24 bits.
465 */
466#define	EXT_FLAG_EMBREF		0x000001	/* embedded ext_count */
467#define	EXT_FLAG_EXTREF		0x000002	/* external ext_cnt, notyet */
468
469#define	EXT_FLAG_NOFREE		0x000010	/* don't free mbuf to pool, notyet */
470
471#define	EXT_FLAG_VENDOR1	0x010000	/* These flags are vendor */
472#define	EXT_FLAG_VENDOR2	0x020000	/* or submodule specific, */
473#define	EXT_FLAG_VENDOR3	0x040000	/* not used by mbuf code. */
474#define	EXT_FLAG_VENDOR4	0x080000	/* Set/read by submodule. */
475
476#define	EXT_FLAG_EXP1		0x100000	/* for experimental use */
477#define	EXT_FLAG_EXP2		0x200000	/* for experimental use */
478#define	EXT_FLAG_EXP3		0x400000	/* for experimental use */
479#define	EXT_FLAG_EXP4		0x800000	/* for experimental use */
480
481/*
482 * EXT flag description for use with printf(9) %b identifier.
483 */
484#define	EXT_FLAG_BITS \
485    "\20\1EXT_FLAG_EMBREF\2EXT_FLAG_EXTREF\5EXT_FLAG_NOFREE" \
486    "\21EXT_FLAG_VENDOR1\22EXT_FLAG_VENDOR2\23EXT_FLAG_VENDOR3" \
487    "\24EXT_FLAG_VENDOR4\25EXT_FLAG_EXP1\26EXT_FLAG_EXP2\27EXT_FLAG_EXP3" \
488    "\30EXT_FLAG_EXP4"
489
490/*
491 * Flags indicating checksum, segmentation and other offload work to be
492 * done, or already done, by hardware or lower layers.  It is split into
493 * separate inbound and outbound flags.
494 *
495 * Outbound flags that are set by upper protocol layers requesting lower
496 * layers, or ideally the hardware, to perform these offloading tasks.
497 * For outbound packets this field and its flags can be directly tested
498 * against ifnet if_hwassist.
499 */
500#define	CSUM_IP			0x00000001	/* IP header checksum offload */
501#define	CSUM_IP_UDP		0x00000002	/* UDP checksum offload */
502#define	CSUM_IP_TCP		0x00000004	/* TCP checksum offload */
503#define	CSUM_IP_SCTP		0x00000008	/* SCTP checksum offload */
504#define	CSUM_IP_TSO		0x00000010	/* TCP segmentation offload */
505#define	CSUM_IP_ISCSI		0x00000020	/* iSCSI checksum offload */
506
507#define	CSUM_IP6_UDP		0x00000200	/* UDP checksum offload */
508#define	CSUM_IP6_TCP		0x00000400	/* TCP checksum offload */
509#define	CSUM_IP6_SCTP		0x00000800	/* SCTP checksum offload */
510#define	CSUM_IP6_TSO		0x00001000	/* TCP segmentation offload */
511#define	CSUM_IP6_ISCSI		0x00002000	/* iSCSI checksum offload */
512
513/* Inbound checksum support where the checksum was verified by hardware. */
514#define	CSUM_L3_CALC		0x01000000	/* calculated layer 3 csum */
515#define	CSUM_L3_VALID		0x02000000	/* checksum is correct */
516#define	CSUM_L4_CALC		0x04000000	/* calculated layer 4 csum */
517#define	CSUM_L4_VALID		0x08000000	/* checksum is correct */
518#define	CSUM_L5_CALC		0x10000000	/* calculated layer 5 csum */
519#define	CSUM_L5_VALID		0x20000000	/* checksum is correct */
520#define	CSUM_COALESCED		0x40000000	/* contains merged segments */
521
522/*
523 * CSUM flag description for use with printf(9) %b identifier.
524 */
525#define	CSUM_BITS \
526    "\20\1CSUM_IP\2CSUM_IP_UDP\3CSUM_IP_TCP\4CSUM_IP_SCTP\5CSUM_IP_TSO" \
527    "\6CSUM_IP_ISCSI" \
528    "\12CSUM_IP6_UDP\13CSUM_IP6_TCP\14CSUM_IP6_SCTP\15CSUM_IP6_TSO" \
529    "\16CSUM_IP6_ISCSI" \
530    "\31CSUM_L3_CALC\32CSUM_L3_VALID\33CSUM_L4_CALC\34CSUM_L4_VALID" \
531    "\35CSUM_L5_CALC\36CSUM_L5_VALID\37CSUM_COALESCED"
532
533/* CSUM flags compatibility mappings. */
534#define	CSUM_IP_CHECKED		CSUM_L3_CALC
535#define	CSUM_IP_VALID		CSUM_L3_VALID
536#define	CSUM_DATA_VALID		CSUM_L4_VALID
537#define	CSUM_PSEUDO_HDR		CSUM_L4_CALC
538#define	CSUM_SCTP_VALID		CSUM_L4_VALID
539#define	CSUM_DELAY_DATA		(CSUM_TCP|CSUM_UDP)
540#define	CSUM_DELAY_IP		CSUM_IP		/* Only v4, no v6 IP hdr csum */
541#define	CSUM_DELAY_DATA_IPV6	(CSUM_TCP_IPV6|CSUM_UDP_IPV6)
542#define	CSUM_DATA_VALID_IPV6	CSUM_DATA_VALID
543#define	CSUM_TCP		CSUM_IP_TCP
544#define	CSUM_UDP		CSUM_IP_UDP
545#define	CSUM_SCTP		CSUM_IP_SCTP
546#define	CSUM_TSO		(CSUM_IP_TSO|CSUM_IP6_TSO)
547#define	CSUM_UDP_IPV6		CSUM_IP6_UDP
548#define	CSUM_TCP_IPV6		CSUM_IP6_TCP
549#define	CSUM_SCTP_IPV6		CSUM_IP6_SCTP
550
551/*
552 * mbuf types describing the content of the mbuf (including external storage).
553 */
554#define	MT_NOTMBUF	0	/* USED INTERNALLY ONLY! Object is not mbuf */
555#define	MT_DATA		1	/* dynamic (data) allocation */
556#define	MT_HEADER	MT_DATA	/* packet header, use M_PKTHDR instead */
557
558#define	MT_VENDOR1	4	/* for vendor-internal use */
559#define	MT_VENDOR2	5	/* for vendor-internal use */
560#define	MT_VENDOR3	6	/* for vendor-internal use */
561#define	MT_VENDOR4	7	/* for vendor-internal use */
562
563#define	MT_SONAME	8	/* socket name */
564
565#define	MT_EXP1		9	/* for experimental use */
566#define	MT_EXP2		10	/* for experimental use */
567#define	MT_EXP3		11	/* for experimental use */
568#define	MT_EXP4		12	/* for experimental use */
569
570#define	MT_CONTROL	14	/* extra-data protocol message */
571#define	MT_EXTCONTROL	15	/* control message with externalized contents */
572#define	MT_OOBDATA	16	/* expedited data  */
573
574#define	MT_NOINIT	255	/* Not a type but a flag to allocate
575				   a non-initialized mbuf */
576
577/*
578 * String names of mbuf-related UMA(9) and malloc(9) types.  Exposed to
579 * !_KERNEL so that monitoring tools can look up the zones with
580 * libmemstat(3).
581 */
582#define	MBUF_MEM_NAME		"mbuf"
583#define	MBUF_CLUSTER_MEM_NAME	"mbuf_cluster"
584#define	MBUF_PACKET_MEM_NAME	"mbuf_packet"
585#define	MBUF_JUMBOP_MEM_NAME	"mbuf_jumbo_page"
586#define	MBUF_JUMBO9_MEM_NAME	"mbuf_jumbo_9k"
587#define	MBUF_JUMBO16_MEM_NAME	"mbuf_jumbo_16k"
588#define	MBUF_TAG_MEM_NAME	"mbuf_tag"
589#define	MBUF_EXTREFCNT_MEM_NAME	"mbuf_ext_refcnt"
590
591#ifdef _KERNEL
592
593#ifdef WITNESS
594#define	MBUF_CHECKSLEEP(how) do {					\
595	if (how == M_WAITOK)						\
596		WITNESS_WARN(WARN_GIANTOK | WARN_SLEEPOK, NULL,		\
597		    "Sleeping in \"%s\"", __func__);			\
598} while (0)
599#else
600#define	MBUF_CHECKSLEEP(how)
601#endif
602
603/*
604 * Network buffer allocation API
605 *
606 * The rest of it is defined in kern/kern_mbuf.c
607 */
608extern uma_zone_t	zone_mbuf;
609extern uma_zone_t	zone_clust;
610extern uma_zone_t	zone_pack;
611extern uma_zone_t	zone_jumbop;
612extern uma_zone_t	zone_jumbo9;
613extern uma_zone_t	zone_jumbo16;
614
615void		 mb_dupcl(struct mbuf *, struct mbuf *);
616void		 mb_free_ext(struct mbuf *);
617void		 m_adj(struct mbuf *, int);
618int		 m_apply(struct mbuf *, int, int,
619		    int (*)(void *, void *, u_int), void *);
620int		 m_append(struct mbuf *, int, c_caddr_t);
621void		 m_cat(struct mbuf *, struct mbuf *);
622void		 m_catpkt(struct mbuf *, struct mbuf *);
623int		 m_clget(struct mbuf *m, int how);
624void 		*m_cljget(struct mbuf *m, int how, int size);
625struct mbuf	*m_collapse(struct mbuf *, int, int);
626void		 m_copyback(struct mbuf *, int, int, c_caddr_t);
627void		 m_copydata(const struct mbuf *, int, int, caddr_t);
628struct mbuf	*m_copym(struct mbuf *, int, int, int);
629struct mbuf	*m_copypacket(struct mbuf *, int);
630void		 m_copy_pkthdr(struct mbuf *, struct mbuf *);
631struct mbuf	*m_copyup(struct mbuf *, int, int);
632struct mbuf	*m_defrag(struct mbuf *, int);
633void		 m_demote_pkthdr(struct mbuf *);
634void		 m_demote(struct mbuf *, int, int);
635struct mbuf	*m_devget(char *, int, int, struct ifnet *,
636		    void (*)(char *, caddr_t, u_int));
637void		 m_dispose_extcontrolm(struct mbuf *m);
638struct mbuf	*m_dup(const struct mbuf *, int);
639int		 m_dup_pkthdr(struct mbuf *, const struct mbuf *, int);
640void		 m_extadd(struct mbuf *, char *, u_int, m_ext_free_t,
641		    void *, void *, int, int);
642u_int		 m_fixhdr(struct mbuf *);
643struct mbuf	*m_fragment(struct mbuf *, int, int);
644void		 m_freem(struct mbuf *);
645struct mbuf	*m_get2(int, int, short, int);
646struct mbuf	*m_getjcl(int, short, int, int);
647struct mbuf	*m_getm2(struct mbuf *, int, int, short, int);
648struct mbuf	*m_getptr(struct mbuf *, int, int *);
649u_int		 m_length(struct mbuf *, struct mbuf **);
650int		 m_mbuftouio(struct uio *, const struct mbuf *, int);
651void		 m_move_pkthdr(struct mbuf *, struct mbuf *);
652int		 m_pkthdr_init(struct mbuf *, int);
653struct mbuf	*m_prepend(struct mbuf *, int, int);
654void		 m_print(const struct mbuf *, int);
655struct mbuf	*m_pulldown(struct mbuf *, int, int, int *);
656struct mbuf	*m_pullup(struct mbuf *, int);
657int		 m_sanity(struct mbuf *, int);
658struct mbuf	*m_split(struct mbuf *, int, int);
659struct mbuf	*m_uiotombuf(struct uio *, int, int, int, int);
660struct mbuf	*m_unshare(struct mbuf *, int);
661
662static __inline int
663m_gettype(int size)
664{
665	int type;
666
667	switch (size) {
668	case MSIZE:
669		type = EXT_MBUF;
670		break;
671	case MCLBYTES:
672		type = EXT_CLUSTER;
673		break;
674#if MJUMPAGESIZE != MCLBYTES
675	case MJUMPAGESIZE:
676		type = EXT_JUMBOP;
677		break;
678#endif
679	case MJUM9BYTES:
680		type = EXT_JUMBO9;
681		break;
682	case MJUM16BYTES:
683		type = EXT_JUMBO16;
684		break;
685	default:
686		panic("%s: invalid cluster size %d", __func__, size);
687	}
688
689	return (type);
690}
691
692/*
693 * Associated an external reference counted buffer with an mbuf.
694 */
695static __inline void
696m_extaddref(struct mbuf *m, char *buf, u_int size, u_int *ref_cnt,
697    m_ext_free_t freef, void *arg1, void *arg2)
698{
699
700	KASSERT(ref_cnt != NULL, ("%s: ref_cnt not provided", __func__));
701
702	atomic_add_int(ref_cnt, 1);
703	m->m_flags |= M_EXT;
704	m->m_ext.ext_buf = buf;
705	m->m_ext.ext_cnt = ref_cnt;
706	m->m_data = m->m_ext.ext_buf;
707	m->m_ext.ext_size = size;
708	m->m_ext.ext_free = freef;
709	m->m_ext.ext_arg1 = arg1;
710	m->m_ext.ext_arg2 = arg2;
711	m->m_ext.ext_type = EXT_EXTREF;
712	m->m_ext.ext_flags = 0;
713}
714
715static __inline uma_zone_t
716m_getzone(int size)
717{
718	uma_zone_t zone;
719
720	switch (size) {
721	case MCLBYTES:
722		zone = zone_clust;
723		break;
724#if MJUMPAGESIZE != MCLBYTES
725	case MJUMPAGESIZE:
726		zone = zone_jumbop;
727		break;
728#endif
729	case MJUM9BYTES:
730		zone = zone_jumbo9;
731		break;
732	case MJUM16BYTES:
733		zone = zone_jumbo16;
734		break;
735	default:
736		panic("%s: invalid cluster size %d", __func__, size);
737	}
738
739	return (zone);
740}
741
742/*
743 * Initialize an mbuf with linear storage.
744 *
745 * Inline because the consumer text overhead will be roughly the same to
746 * initialize or call a function with this many parameters and M_PKTHDR
747 * should go away with constant propagation for !MGETHDR.
748 */
749static __inline int
750m_init(struct mbuf *m, int how, short type, int flags)
751{
752	int error;
753
754	m->m_next = NULL;
755	m->m_nextpkt = NULL;
756	m->m_data = m->m_dat;
757	m->m_len = 0;
758	m->m_flags = flags;
759	m->m_type = type;
760	if (flags & M_PKTHDR)
761		error = m_pkthdr_init(m, how);
762	else
763		error = 0;
764
765	MBUF_PROBE5(m__init, m, how, type, flags, error);
766	return (error);
767}
768
769static __inline struct mbuf *
770m_get(int how, short type)
771{
772	struct mbuf *m;
773	struct mb_args args;
774
775	args.flags = 0;
776	args.type = type;
777	m = uma_zalloc_arg(zone_mbuf, &args, how);
778	MBUF_PROBE3(m__get, how, type, m);
779	return (m);
780}
781
782static __inline struct mbuf *
783m_gethdr(int how, short type)
784{
785	struct mbuf *m;
786	struct mb_args args;
787
788	args.flags = M_PKTHDR;
789	args.type = type;
790	m = uma_zalloc_arg(zone_mbuf, &args, how);
791	MBUF_PROBE3(m__gethdr, how, type, m);
792	return (m);
793}
794
795static __inline struct mbuf *
796m_getcl(int how, short type, int flags)
797{
798	struct mbuf *m;
799	struct mb_args args;
800
801	args.flags = flags;
802	args.type = type;
803	m = uma_zalloc_arg(zone_pack, &args, how);
804	MBUF_PROBE4(m__getcl, how, type, flags, m);
805	return (m);
806}
807
808/*
809 * XXX: m_cljset() is a dangerous API.  One must attach only a new,
810 * unreferenced cluster to an mbuf(9).  It is not possible to assert
811 * that, so care can be taken only by users of the API.
812 */
813static __inline void
814m_cljset(struct mbuf *m, void *cl, int type)
815{
816	int size;
817
818	switch (type) {
819	case EXT_CLUSTER:
820		size = MCLBYTES;
821		break;
822#if MJUMPAGESIZE != MCLBYTES
823	case EXT_JUMBOP:
824		size = MJUMPAGESIZE;
825		break;
826#endif
827	case EXT_JUMBO9:
828		size = MJUM9BYTES;
829		break;
830	case EXT_JUMBO16:
831		size = MJUM16BYTES;
832		break;
833	default:
834		panic("%s: unknown cluster type %d", __func__, type);
835		break;
836	}
837
838	m->m_data = m->m_ext.ext_buf = cl;
839	m->m_ext.ext_free = m->m_ext.ext_arg1 = m->m_ext.ext_arg2 = NULL;
840	m->m_ext.ext_size = size;
841	m->m_ext.ext_type = type;
842	m->m_ext.ext_flags = EXT_FLAG_EMBREF;
843	m->m_ext.ext_count = 1;
844	m->m_flags |= M_EXT;
845	MBUF_PROBE3(m__cljset, m, cl, type);
846}
847
848static __inline void
849m_chtype(struct mbuf *m, short new_type)
850{
851
852	m->m_type = new_type;
853}
854
855static __inline void
856m_clrprotoflags(struct mbuf *m)
857{
858
859	while (m) {
860		m->m_flags &= ~M_PROTOFLAGS;
861		m = m->m_next;
862	}
863}
864
865static __inline struct mbuf *
866m_last(struct mbuf *m)
867{
868
869	while (m->m_next)
870		m = m->m_next;
871	return (m);
872}
873
874static inline u_int
875m_extrefcnt(struct mbuf *m)
876{
877
878	KASSERT(m->m_flags & M_EXT, ("%s: M_EXT missing", __func__));
879
880	return ((m->m_ext.ext_flags & EXT_FLAG_EMBREF) ? m->m_ext.ext_count :
881	    *m->m_ext.ext_cnt);
882}
883
884/*
885 * mbuf, cluster, and external object allocation macros (for compatibility
886 * purposes).
887 */
888#define	M_MOVE_PKTHDR(to, from)	m_move_pkthdr((to), (from))
889#define	MGET(m, how, type)	((m) = m_get((how), (type)))
890#define	MGETHDR(m, how, type)	((m) = m_gethdr((how), (type)))
891#define	MCLGET(m, how)		m_clget((m), (how))
892#define	MEXTADD(m, buf, size, free, arg1, arg2, flags, type)		\
893    m_extadd((m), (char *)(buf), (size), (free), (arg1), (arg2),	\
894    (flags), (type))
895#define	m_getm(m, len, how, type)					\
896    m_getm2((m), (len), (how), (type), M_PKTHDR)
897
898/*
899 * Evaluate TRUE if it's safe to write to the mbuf m's data region (this can
900 * be both the local data payload, or an external buffer area, depending on
901 * whether M_EXT is set).
902 */
903#define	M_WRITABLE(m)	(!((m)->m_flags & M_RDONLY) &&			\
904			 (!(((m)->m_flags & M_EXT)) ||			\
905			 (m_extrefcnt(m) == 1)))
906
907/* Check if the supplied mbuf has a packet header, or else panic. */
908#define	M_ASSERTPKTHDR(m)						\
909	KASSERT((m) != NULL && (m)->m_flags & M_PKTHDR,			\
910	    ("%s: no mbuf packet header!", __func__))
911
912/*
913 * Ensure that the supplied mbuf is a valid, non-free mbuf.
914 *
915 * XXX: Broken at the moment.  Need some UMA magic to make it work again.
916 */
917#define	M_ASSERTVALID(m)						\
918	KASSERT((((struct mbuf *)m)->m_flags & 0) == 0,			\
919	    ("%s: attempted use of a free mbuf!", __func__))
920
921/*
922 * Return the address of the start of the buffer associated with an mbuf,
923 * handling external storage, packet-header mbufs, and regular data mbufs.
924 */
925#define	M_START(m)							\
926	(((m)->m_flags & M_EXT) ? (m)->m_ext.ext_buf :			\
927	 ((m)->m_flags & M_PKTHDR) ? &(m)->m_pktdat[0] :		\
928	 &(m)->m_dat[0])
929
930/*
931 * Return the size of the buffer associated with an mbuf, handling external
932 * storage, packet-header mbufs, and regular data mbufs.
933 */
934#define	M_SIZE(m)							\
935	(((m)->m_flags & M_EXT) ? (m)->m_ext.ext_size :			\
936	 ((m)->m_flags & M_PKTHDR) ? MHLEN :				\
937	 MLEN)
938
939/*
940 * Set the m_data pointer of a newly allocated mbuf to place an object of the
941 * specified size at the end of the mbuf, longword aligned.
942 *
943 * NB: Historically, we had M_ALIGN(), MH_ALIGN(), and MEXT_ALIGN() as
944 * separate macros, each asserting that it was called at the proper moment.
945 * This required callers to themselves test the storage type and call the
946 * right one.  Rather than require callers to be aware of those layout
947 * decisions, we centralize here.
948 */
949static __inline void
950m_align(struct mbuf *m, int len)
951{
952#ifdef INVARIANTS
953	const char *msg = "%s: not a virgin mbuf";
954#endif
955	int adjust;
956
957	KASSERT(m->m_data == M_START(m), (msg, __func__));
958
959	adjust = M_SIZE(m) - len;
960	m->m_data += adjust &~ (sizeof(long)-1);
961}
962
963#define	M_ALIGN(m, len)		m_align(m, len)
964#define	MH_ALIGN(m, len)	m_align(m, len)
965#define	MEXT_ALIGN(m, len)	m_align(m, len)
966
967/*
968 * Compute the amount of space available before the current start of data in
969 * an mbuf.
970 *
971 * The M_WRITABLE() is a temporary, conservative safety measure: the burden
972 * of checking writability of the mbuf data area rests solely with the caller.
973 *
974 * NB: In previous versions, M_LEADINGSPACE() would only check M_WRITABLE()
975 * for mbufs with external storage.  We now allow mbuf-embedded data to be
976 * read-only as well.
977 */
978#define	M_LEADINGSPACE(m)						\
979	(M_WRITABLE(m) ? ((m)->m_data - M_START(m)) : 0)
980
981/*
982 * Compute the amount of space available after the end of data in an mbuf.
983 *
984 * The M_WRITABLE() is a temporary, conservative safety measure: the burden
985 * of checking writability of the mbuf data area rests solely with the caller.
986 *
987 * NB: In previous versions, M_TRAILINGSPACE() would only check M_WRITABLE()
988 * for mbufs with external storage.  We now allow mbuf-embedded data to be
989 * read-only as well.
990 */
991#define	M_TRAILINGSPACE(m)						\
992	(M_WRITABLE(m) ?						\
993	    ((M_START(m) + M_SIZE(m)) - ((m)->m_data + (m)->m_len)) : 0)
994
995/*
996 * Arrange to prepend space of size plen to mbuf m.  If a new mbuf must be
997 * allocated, how specifies whether to wait.  If the allocation fails, the
998 * original mbuf chain is freed and m is set to NULL.
999 */
1000#define	M_PREPEND(m, plen, how) do {					\
1001	struct mbuf **_mmp = &(m);					\
1002	struct mbuf *_mm = *_mmp;					\
1003	int _mplen = (plen);						\
1004	int __mhow = (how);						\
1005									\
1006	MBUF_CHECKSLEEP(how);						\
1007	if (M_LEADINGSPACE(_mm) >= _mplen) {				\
1008		_mm->m_data -= _mplen;					\
1009		_mm->m_len += _mplen;					\
1010	} else								\
1011		_mm = m_prepend(_mm, _mplen, __mhow);			\
1012	if (_mm != NULL && _mm->m_flags & M_PKTHDR)			\
1013		_mm->m_pkthdr.len += _mplen;				\
1014	*_mmp = _mm;							\
1015} while (0)
1016
1017/*
1018 * Change mbuf to new type.  This is a relatively expensive operation and
1019 * should be avoided.
1020 */
1021#define	MCHTYPE(m, t)	m_chtype((m), (t))
1022
1023/* Length to m_copy to copy all. */
1024#define	M_COPYALL	1000000000
1025
1026extern int		max_datalen;	/* MHLEN - max_hdr */
1027extern int		max_hdr;	/* Largest link + protocol header */
1028extern int		max_linkhdr;	/* Largest link-level header */
1029extern int		max_protohdr;	/* Largest protocol header */
1030extern int		nmbclusters;	/* Maximum number of clusters */
1031
1032/*-
1033 * Network packets may have annotations attached by affixing a list of
1034 * "packet tags" to the pkthdr structure.  Packet tags are dynamically
1035 * allocated semi-opaque data structures that have a fixed header
1036 * (struct m_tag) that specifies the size of the memory block and a
1037 * <cookie,type> pair that identifies it.  The cookie is a 32-bit unique
1038 * unsigned value used to identify a module or ABI.  By convention this value
1039 * is chosen as the date+time that the module is created, expressed as the
1040 * number of seconds since the epoch (e.g., using date -u +'%s').  The type
1041 * value is an ABI/module-specific value that identifies a particular
1042 * annotation and is private to the module.  For compatibility with systems
1043 * like OpenBSD that define packet tags w/o an ABI/module cookie, the value
1044 * PACKET_ABI_COMPAT is used to implement m_tag_get and m_tag_find
1045 * compatibility shim functions and several tag types are defined below.
1046 * Users that do not require compatibility should use a private cookie value
1047 * so that packet tag-related definitions can be maintained privately.
1048 *
1049 * Note that the packet tag returned by m_tag_alloc has the default memory
1050 * alignment implemented by malloc.  To reference private data one can use a
1051 * construct like:
1052 *
1053 *	struct m_tag *mtag = m_tag_alloc(...);
1054 *	struct foo *p = (struct foo *)(mtag+1);
1055 *
1056 * if the alignment of struct m_tag is sufficient for referencing members of
1057 * struct foo.  Otherwise it is necessary to embed struct m_tag within the
1058 * private data structure to insure proper alignment; e.g.,
1059 *
1060 *	struct foo {
1061 *		struct m_tag	tag;
1062 *		...
1063 *	};
1064 *	struct foo *p = (struct foo *) m_tag_alloc(...);
1065 *	struct m_tag *mtag = &p->tag;
1066 */
1067
1068/*
1069 * Persistent tags stay with an mbuf until the mbuf is reclaimed.  Otherwise
1070 * tags are expected to ``vanish'' when they pass through a network
1071 * interface.  For most interfaces this happens normally as the tags are
1072 * reclaimed when the mbuf is free'd.  However in some special cases
1073 * reclaiming must be done manually.  An example is packets that pass through
1074 * the loopback interface.  Also, one must be careful to do this when
1075 * ``turning around'' packets (e.g., icmp_reflect).
1076 *
1077 * To mark a tag persistent bit-or this flag in when defining the tag id.
1078 * The tag will then be treated as described above.
1079 */
1080#define	MTAG_PERSISTENT				0x800
1081
1082#define	PACKET_TAG_NONE				0  /* Nadda */
1083
1084/* Packet tags for use with PACKET_ABI_COMPAT. */
1085#define	PACKET_TAG_IPSEC_IN_DONE		1  /* IPsec applied, in */
1086#define	PACKET_TAG_IPSEC_OUT_DONE		2  /* IPsec applied, out */
1087#define	PACKET_TAG_IPSEC_IN_CRYPTO_DONE		3  /* NIC IPsec crypto done */
1088#define	PACKET_TAG_IPSEC_OUT_CRYPTO_NEEDED	4  /* NIC IPsec crypto req'ed */
1089#define	PACKET_TAG_IPSEC_IN_COULD_DO_CRYPTO	5  /* NIC notifies IPsec */
1090#define	PACKET_TAG_IPSEC_PENDING_TDB		6  /* Reminder to do IPsec */
1091#define	PACKET_TAG_BRIDGE			7  /* Bridge processing done */
1092#define	PACKET_TAG_GIF				8  /* GIF processing done */
1093#define	PACKET_TAG_GRE				9  /* GRE processing done */
1094#define	PACKET_TAG_IN_PACKET_CHECKSUM		10 /* NIC checksumming done */
1095#define	PACKET_TAG_ENCAP			11 /* Encap.  processing */
1096#define	PACKET_TAG_IPSEC_SOCKET			12 /* IPSEC socket ref */
1097#define	PACKET_TAG_IPSEC_HISTORY		13 /* IPSEC history */
1098#define	PACKET_TAG_IPV6_INPUT			14 /* IPV6 input processing */
1099#define	PACKET_TAG_DUMMYNET			15 /* dummynet info */
1100#define	PACKET_TAG_DIVERT			17 /* divert info */
1101#define	PACKET_TAG_IPFORWARD			18 /* ipforward info */
1102#define	PACKET_TAG_MACLABEL	(19 | MTAG_PERSISTENT) /* MAC label */
1103#define	PACKET_TAG_PF		(21 | MTAG_PERSISTENT) /* PF/ALTQ information */
1104#define	PACKET_TAG_RTSOCKFAM			25 /* rtsock sa family */
1105#define	PACKET_TAG_IPOPTIONS			27 /* Saved IP options */
1106#define	PACKET_TAG_CARP				28 /* CARP info */
1107#define	PACKET_TAG_IPSEC_NAT_T_PORTS		29 /* two uint16_t */
1108#define	PACKET_TAG_ND_OUTGOING			30 /* ND outgoing */
1109
1110/* Specific cookies and tags. */
1111
1112/* Packet tag routines. */
1113struct m_tag	*m_tag_alloc(u_int32_t, int, int, int);
1114void		 m_tag_delete(struct mbuf *, struct m_tag *);
1115void		 m_tag_delete_chain(struct mbuf *, struct m_tag *);
1116void		 m_tag_free_default(struct m_tag *);
1117struct m_tag	*m_tag_locate(struct mbuf *, u_int32_t, int, struct m_tag *);
1118struct m_tag	*m_tag_copy(struct m_tag *, int);
1119int		 m_tag_copy_chain(struct mbuf *, const struct mbuf *, int);
1120void		 m_tag_delete_nonpersistent(struct mbuf *);
1121
1122/*
1123 * Initialize the list of tags associated with an mbuf.
1124 */
1125static __inline void
1126m_tag_init(struct mbuf *m)
1127{
1128
1129	SLIST_INIT(&m->m_pkthdr.tags);
1130}
1131
1132/*
1133 * Set up the contents of a tag.  Note that this does not fill in the free
1134 * method; the caller is expected to do that.
1135 *
1136 * XXX probably should be called m_tag_init, but that was already taken.
1137 */
1138static __inline void
1139m_tag_setup(struct m_tag *t, u_int32_t cookie, int type, int len)
1140{
1141
1142	t->m_tag_id = type;
1143	t->m_tag_len = len;
1144	t->m_tag_cookie = cookie;
1145}
1146
1147/*
1148 * Reclaim resources associated with a tag.
1149 */
1150static __inline void
1151m_tag_free(struct m_tag *t)
1152{
1153
1154	(*t->m_tag_free)(t);
1155}
1156
1157/*
1158 * Return the first tag associated with an mbuf.
1159 */
1160static __inline struct m_tag *
1161m_tag_first(struct mbuf *m)
1162{
1163
1164	return (SLIST_FIRST(&m->m_pkthdr.tags));
1165}
1166
1167/*
1168 * Return the next tag in the list of tags associated with an mbuf.
1169 */
1170static __inline struct m_tag *
1171m_tag_next(struct mbuf *m __unused, struct m_tag *t)
1172{
1173
1174	return (SLIST_NEXT(t, m_tag_link));
1175}
1176
1177/*
1178 * Prepend a tag to the list of tags associated with an mbuf.
1179 */
1180static __inline void
1181m_tag_prepend(struct mbuf *m, struct m_tag *t)
1182{
1183
1184	SLIST_INSERT_HEAD(&m->m_pkthdr.tags, t, m_tag_link);
1185}
1186
1187/*
1188 * Unlink a tag from the list of tags associated with an mbuf.
1189 */
1190static __inline void
1191m_tag_unlink(struct mbuf *m, struct m_tag *t)
1192{
1193
1194	SLIST_REMOVE(&m->m_pkthdr.tags, t, m_tag, m_tag_link);
1195}
1196
1197/* These are for OpenBSD compatibility. */
1198#define	MTAG_ABI_COMPAT		0		/* compatibility ABI */
1199
1200static __inline struct m_tag *
1201m_tag_get(int type, int length, int wait)
1202{
1203	return (m_tag_alloc(MTAG_ABI_COMPAT, type, length, wait));
1204}
1205
1206static __inline struct m_tag *
1207m_tag_find(struct mbuf *m, int type, struct m_tag *start)
1208{
1209	return (SLIST_EMPTY(&m->m_pkthdr.tags) ? (struct m_tag *)NULL :
1210	    m_tag_locate(m, MTAG_ABI_COMPAT, type, start));
1211}
1212
1213static __inline struct mbuf *
1214m_free(struct mbuf *m)
1215{
1216	struct mbuf *n = m->m_next;
1217
1218	MBUF_PROBE1(m__free, m);
1219	if ((m->m_flags & (M_PKTHDR|M_NOFREE)) == (M_PKTHDR|M_NOFREE))
1220		m_tag_delete_chain(m, NULL);
1221	if (m->m_flags & M_EXT)
1222		mb_free_ext(m);
1223	else if ((m->m_flags & M_NOFREE) == 0)
1224		uma_zfree(zone_mbuf, m);
1225	return (n);
1226}
1227
1228static __inline int
1229rt_m_getfib(struct mbuf *m)
1230{
1231	KASSERT(m->m_flags & M_PKTHDR , ("Attempt to get FIB from non header mbuf."));
1232	return (m->m_pkthdr.fibnum);
1233}
1234
1235#define M_GETFIB(_m)   rt_m_getfib(_m)
1236
1237#define M_SETFIB(_m, _fib) do {						\
1238        KASSERT((_m)->m_flags & M_PKTHDR, ("Attempt to set FIB on non header mbuf."));	\
1239	((_m)->m_pkthdr.fibnum) = (_fib);				\
1240} while (0)
1241
1242/* flags passed as first argument for "m_ether_tcpip_hash()" */
1243#define	MBUF_HASHFLAG_L2	(1 << 2)
1244#define	MBUF_HASHFLAG_L3	(1 << 3)
1245#define	MBUF_HASHFLAG_L4	(1 << 4)
1246
1247/* mbuf hashing helper routines */
1248uint32_t	m_ether_tcpip_hash_init(void);
1249uint32_t	m_ether_tcpip_hash(const uint32_t, const struct mbuf *, const uint32_t);
1250
1251#ifdef MBUF_PROFILING
1252 void m_profile(struct mbuf *m);
1253 #define M_PROFILE(m) m_profile(m)
1254#else
1255 #define M_PROFILE(m)
1256#endif
1257
1258struct mbufq {
1259	STAILQ_HEAD(, mbuf)	mq_head;
1260	int			mq_len;
1261	int			mq_maxlen;
1262};
1263
1264static inline void
1265mbufq_init(struct mbufq *mq, int maxlen)
1266{
1267
1268	STAILQ_INIT(&mq->mq_head);
1269	mq->mq_maxlen = maxlen;
1270	mq->mq_len = 0;
1271}
1272
1273static inline struct mbuf *
1274mbufq_flush(struct mbufq *mq)
1275{
1276	struct mbuf *m;
1277
1278	m = STAILQ_FIRST(&mq->mq_head);
1279	STAILQ_INIT(&mq->mq_head);
1280	mq->mq_len = 0;
1281	return (m);
1282}
1283
1284static inline void
1285mbufq_drain(struct mbufq *mq)
1286{
1287	struct mbuf *m, *n;
1288
1289	n = mbufq_flush(mq);
1290	while ((m = n) != NULL) {
1291		n = STAILQ_NEXT(m, m_stailqpkt);
1292		m_freem(m);
1293	}
1294}
1295
1296static inline struct mbuf *
1297mbufq_first(const struct mbufq *mq)
1298{
1299
1300	return (STAILQ_FIRST(&mq->mq_head));
1301}
1302
1303static inline struct mbuf *
1304mbufq_last(const struct mbufq *mq)
1305{
1306
1307	return (STAILQ_LAST(&mq->mq_head, mbuf, m_stailqpkt));
1308}
1309
1310static inline int
1311mbufq_full(const struct mbufq *mq)
1312{
1313
1314	return (mq->mq_len >= mq->mq_maxlen);
1315}
1316
1317static inline int
1318mbufq_len(const struct mbufq *mq)
1319{
1320
1321	return (mq->mq_len);
1322}
1323
1324static inline int
1325mbufq_enqueue(struct mbufq *mq, struct mbuf *m)
1326{
1327
1328	if (mbufq_full(mq))
1329		return (ENOBUFS);
1330	STAILQ_INSERT_TAIL(&mq->mq_head, m, m_stailqpkt);
1331	mq->mq_len++;
1332	return (0);
1333}
1334
1335static inline struct mbuf *
1336mbufq_dequeue(struct mbufq *mq)
1337{
1338	struct mbuf *m;
1339
1340	m = STAILQ_FIRST(&mq->mq_head);
1341	if (m) {
1342		STAILQ_REMOVE_HEAD(&mq->mq_head, m_stailqpkt);
1343		m->m_nextpkt = NULL;
1344		mq->mq_len--;
1345	}
1346	return (m);
1347}
1348
1349static inline void
1350mbufq_prepend(struct mbufq *mq, struct mbuf *m)
1351{
1352
1353	STAILQ_INSERT_HEAD(&mq->mq_head, m, m_stailqpkt);
1354	mq->mq_len++;
1355}
1356
1357/*
1358 * Note: this doesn't enforce the maximum list size for dst.
1359 */
1360static inline void
1361mbufq_concat(struct mbufq *mq_dst, struct mbufq *mq_src)
1362{
1363
1364	mq_dst->mq_len += mq_src->mq_len;
1365	STAILQ_CONCAT(&mq_dst->mq_head, &mq_src->mq_head);
1366	mq_src->mq_len = 0;
1367}
1368
1369#ifdef _SYS_TIMESPEC_H_
1370static inline void
1371mbuf_tstmp2timespec(struct mbuf *m, struct timespec *ts)
1372{
1373
1374	KASSERT((m->m_flags & M_PKTHDR) != 0, ("mbuf %p no M_PKTHDR", m));
1375	KASSERT((m->m_flags & M_TSTMP) != 0, ("mbuf %p no M_TSTMP", m));
1376	ts->tv_sec = m->m_pkthdr.rcv_tstmp / 1000000000;
1377	ts->tv_nsec = m->m_pkthdr.rcv_tstmp % 1000000000;
1378}
1379#endif
1380
1381#ifdef NETDUMP
1382/* Invoked from the netdump client code. */
1383void	netdump_mbuf_drain(void);
1384void	netdump_mbuf_dump(void);
1385void	netdump_mbuf_reinit(int nmbuf, int nclust, int clsize);
1386#endif
1387
1388#endif /* _KERNEL */
1389#endif /* !_SYS_MBUF_H_ */
1390