xref: /freebsd/sys/sys/mbuf.h (revision daf1cffce2e07931f27c6c6998652e90df6ba87e)
1 /*
2  * Copyright (c) 1982, 1986, 1988, 1993
3  *	The Regents of the University of California.  All rights reserved.
4  *
5  * Redistribution and use in source and binary forms, with or without
6  * modification, are permitted provided that the following conditions
7  * are met:
8  * 1. Redistributions of source code must retain the above copyright
9  *    notice, this list of conditions and the following disclaimer.
10  * 2. Redistributions in binary form must reproduce the above copyright
11  *    notice, this list of conditions and the following disclaimer in the
12  *    documentation and/or other materials provided with the distribution.
13  * 3. All advertising materials mentioning features or use of this software
14  *    must display the following acknowledgement:
15  *	This product includes software developed by the University of
16  *	California, Berkeley and its contributors.
17  * 4. Neither the name of the University nor the names of its contributors
18  *    may be used to endorse or promote products derived from this software
19  *    without specific prior written permission.
20  *
21  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
22  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
23  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
24  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
25  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
26  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
27  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
28  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
29  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
30  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
31  * SUCH DAMAGE.
32  *
33  *	@(#)mbuf.h	8.5 (Berkeley) 2/19/95
34  * $FreeBSD$
35  */
36 
37 #ifndef _SYS_MBUF_H_
38 #define	_SYS_MBUF_H_
39 
40 /*
41  * Mbufs are of a single size, MSIZE (machine/param.h), which
42  * includes overhead.  An mbuf may add a single "mbuf cluster" of size
43  * MCLBYTES (also in machine/param.h), which has no additional overhead
44  * and is used instead of the internal data area; this is done when
45  * at least MINCLSIZE of data must be stored.
46  */
47 
48 #define	MLEN		(MSIZE - sizeof(struct m_hdr))	/* normal data len */
49 #define	MHLEN		(MLEN - sizeof(struct pkthdr))	/* data len w/pkthdr */
50 
51 #define	MINCLSIZE	(MHLEN + 1)	/* smallest amount to put in cluster */
52 #define	M_MAXCOMPRESS	(MHLEN / 2)	/* max amount to copy for compression */
53 
54 /*
55  * Macros for type conversion
56  * mtod(m, t) -	convert mbuf pointer to data pointer of correct type
57  * dtom(x) -	convert data pointer within mbuf to mbuf pointer (XXX)
58  * mtocl(x) -	convert pointer within cluster to cluster index #
59  * cltom(x) -	convert cluster # to ptr to beginning of cluster
60  */
61 #define	mtod(m, t)	((t)((m)->m_data))
62 #define	dtom(x)		((struct mbuf *)((intptr_t)(x) & ~(MSIZE-1)))
63 #define	mtocl(x)	(((uintptr_t)(x) - (uintptr_t)mbutl) >> MCLSHIFT)
64 #define	cltom(x)	((caddr_t)((uintptr_t)mbutl + \
65 			    ((uintptr_t)(x) << MCLSHIFT)))
66 
67 /* header at beginning of each mbuf: */
68 struct m_hdr {
69 	struct	mbuf *mh_next;		/* next buffer in chain */
70 	struct	mbuf *mh_nextpkt;	/* next chain in queue/record */
71 	caddr_t	mh_data;		/* location of data */
72 	int	mh_len;			/* amount of data in this mbuf */
73 	short	mh_type;		/* type of data in this mbuf */
74 	short	mh_flags;		/* flags; see below */
75 };
76 
77 /* record/packet header in first mbuf of chain; valid if M_PKTHDR set */
78 struct pkthdr {
79 	struct	ifnet *rcvif;		/* rcv interface */
80 	int	len;			/* total packet length */
81 	/* variables for ip and tcp reassembly */
82 	caddr_t	header;			/* pointer to packet header */
83 };
84 
85 /* description of external storage mapped into mbuf, valid if M_EXT set */
86 struct m_ext {
87 	caddr_t	ext_buf;		/* start of buffer */
88 	void	(*ext_free)		/* free routine if not the usual */
89 		__P((caddr_t, u_int));
90 	u_int	ext_size;		/* size of buffer, for ext_free */
91 	void	(*ext_ref)		/* add a reference to the ext object */
92 		__P((caddr_t, u_int));
93 };
94 
95 struct mbuf {
96 	struct	m_hdr m_hdr;
97 	union {
98 		struct {
99 			struct	pkthdr MH_pkthdr;	/* M_PKTHDR set */
100 			union {
101 				struct	m_ext MH_ext;	/* M_EXT set */
102 				char	MH_databuf[MHLEN];
103 			} MH_dat;
104 		} MH;
105 		char	M_databuf[MLEN];		/* !M_PKTHDR, !M_EXT */
106 	} M_dat;
107 };
108 #define	m_next		m_hdr.mh_next
109 #define	m_len		m_hdr.mh_len
110 #define	m_data		m_hdr.mh_data
111 #define	m_type		m_hdr.mh_type
112 #define	m_flags		m_hdr.mh_flags
113 #define	m_nextpkt	m_hdr.mh_nextpkt
114 #define	m_act		m_nextpkt
115 #define	m_pkthdr	M_dat.MH.MH_pkthdr
116 #define	m_ext		M_dat.MH.MH_dat.MH_ext
117 #define	m_pktdat	M_dat.MH.MH_dat.MH_databuf
118 #define	m_dat		M_dat.M_databuf
119 
120 /* mbuf flags */
121 #define	M_EXT		0x0001	/* has associated external storage */
122 #define	M_PKTHDR	0x0002	/* start of record */
123 #define	M_EOR		0x0004	/* end of record */
124 #define	M_PROTO1	0x0008	/* protocol-specific */
125 #define	M_PROTO2	0x0010	/* protocol-specific */
126 #define	M_PROTO3	0x0020	/* protocol-specific */
127 #define	M_PROTO4	0x0040	/* protocol-specific */
128 #define	M_PROTO5	0x0080	/* protocol-specific */
129 
130 /* mbuf pkthdr flags, also in m_flags */
131 #define	M_BCAST		0x0100	/* send/received as link-level broadcast */
132 #define	M_MCAST		0x0200	/* send/received as link-level multicast */
133 #define	M_FRAG		0x0400	/* packet is a fragment of a larger packet */
134 
135 /* flags copied when copying m_pkthdr */
136 #define	M_COPYFLAGS	(M_PKTHDR|M_EOR|M_PROTO1|M_PROTO1|M_PROTO2|M_PROTO3 | \
137 			    M_PROTO4|M_PROTO5|M_BCAST|M_MCAST|M_FRAG)
138 
139 /* mbuf types */
140 #define	MT_FREE		0	/* should be on free list */
141 #define	MT_DATA		1	/* dynamic (data) allocation */
142 #define	MT_HEADER	2	/* packet header */
143 #if 0
144 #define	MT_SOCKET	3	/* socket structure */
145 #define	MT_PCB		4	/* protocol control block */
146 #define	MT_RTABLE	5	/* routing tables */
147 #define	MT_HTABLE	6	/* IMP host tables */
148 #define	MT_ATABLE	7	/* address resolution tables */
149 #endif
150 #define	MT_SONAME	8	/* socket name */
151 #if 0
152 #define	MT_SOOPTS	10	/* socket options */
153 #endif
154 #define	MT_FTABLE	11	/* fragment reassembly header */
155 #if 0
156 #define	MT_RIGHTS	12	/* access rights */
157 #define	MT_IFADDR	13	/* interface address */
158 #endif
159 #define	MT_CONTROL	14	/* extra-data protocol message */
160 #define	MT_OOBDATA	15	/* expedited data  */
161 
162 /*
163  * mbuf statistics
164  */
165 struct mbstat {
166 	u_long	m_mbufs;	/* mbufs obtained from page pool */
167 	u_long	m_clusters;	/* clusters obtained from page pool */
168 	u_long	m_spare;	/* spare field */
169 	u_long	m_clfree;	/* free clusters */
170 	u_long	m_drops;	/* times failed to find space */
171 	u_long	m_wait;		/* times waited for space */
172 	u_long	m_drain;	/* times drained protocols for space */
173 	u_short	m_mtypes[256];	/* type specific mbuf allocations */
174 	u_long	m_mcfail;	/* times m_copym failed */
175 	u_long	m_mpfail;	/* times m_pullup failed */
176 	u_long	m_msize;	/* length of an mbuf */
177 	u_long	m_mclbytes;	/* length of an mbuf cluster */
178 	u_long	m_minclsize;	/* min length of data to allocate a cluster */
179 	u_long	m_mlen;		/* length of data in an mbuf */
180 	u_long	m_mhlen;	/* length of data in a header mbuf */
181 };
182 
183 /* flags to m_get/MGET */
184 #define	M_DONTWAIT	1
185 #define	M_WAIT		0
186 
187 /* Freelists:
188  *
189  * Normal mbuf clusters are normally treated as character arrays
190  * after allocation, but use the first word of the buffer as a free list
191  * pointer while on the free list.
192  */
193 union mcluster {
194 	union	mcluster *mcl_next;
195 	char	mcl_buf[MCLBYTES];
196 };
197 
198 
199 /*
200  * These are identifying numbers passed to the m_mballoc_wait function,
201  * allowing us to determine whether the call came from an MGETHDR or
202  * an MGET.
203  */
204 #define	MGETHDR_C      1
205 #define	MGET_C         2
206 
207 /*
208  * Wake up the next instance (if any) of m_mballoc_wait() which is
209  * waiting for an mbuf to be freed.  This should be called at splimp().
210  *
211  * XXX: If there is another free mbuf, this routine will be called [again]
212  * from the m_mballoc_wait routine in order to wake another sleep instance.
213  */
214 #define	MMBWAKEUP() do {						\
215 	if (m_mballoc_wid) {						\
216 		m_mballoc_wid--;					\
217 		wakeup_one(&m_mballoc_wid); 				\
218 	}								\
219 } while (0)
220 
221 /*
222  * Same as above, but for mbuf cluster(s).
223  */
224 #define	MCLWAKEUP() do {						\
225 	if (m_clalloc_wid) {						\
226 		m_clalloc_wid--;					\
227 		wakeup_one(&m_clalloc_wid);				\
228 	}								\
229 } while (0)
230 
231 /*
232  * mbuf utility macros:
233  *
234  *	MBUFLOCK(code)
235  * prevents a section of code from from being interrupted by network
236  * drivers.
237  */
238 #define	MBUFLOCK(code) do {						\
239 	int _ms = splimp();						\
240 									\
241 	{ code }							\
242 	splx(_ms);							\
243 } while (0)
244 
245 /*
246  * mbuf allocation/deallocation macros:
247  *
248  *	MGET(struct mbuf *m, int how, int type)
249  * allocates an mbuf and initializes it to contain internal data.
250  *
251  *	MGETHDR(struct mbuf *m, int how, int type)
252  * allocates an mbuf and initializes it to contain a packet header
253  * and internal data.
254  *
255  * Warning: MGETHDR() does *not* initialize m->m_pkthdr.rcvif.
256  */
257 #define	MGET(m, how, type) do {						\
258 	struct mbuf *_mm;						\
259 	int _mhow = (how);						\
260 	int _mtype = (type);						\
261 	int _ms = splimp();						\
262 									\
263 	if (mmbfree == NULL)						\
264 		(void)m_mballoc(1, _mhow);				\
265 	_mm = mmbfree;							\
266 	if (_mm != NULL) {						\
267 		mmbfree = _mm->m_next;					\
268 		mbstat.m_mtypes[MT_FREE]--;				\
269 		_mm->m_type = _mtype;					\
270 		mbstat.m_mtypes[_mtype]++;				\
271 		_mm->m_next = NULL;					\
272 		_mm->m_nextpkt = NULL;					\
273 		_mm->m_data = _mm->m_dat;				\
274 		_mm->m_flags = 0;					\
275 		(m) = _mm;						\
276 		splx(_ms);						\
277 	} else {							\
278 		splx(_ms);						\
279 		_mm = m_retry(_mhow, _mtype);				\
280 		if (_mm == NULL && _mhow == M_WAIT)			\
281 			(m) = m_mballoc_wait(MGET_C, _mtype);		\
282 		else							\
283 			(m) = _mm;					\
284 	}								\
285 } while (0)
286 
287 #define	MGETHDR(m, how, type) do {					\
288 	struct mbuf *_mm;						\
289 	int _mhow = (how);						\
290 	int _mtype = (type);						\
291 	int _ms = splimp();						\
292 									\
293 	if (mmbfree == NULL)						\
294 		(void)m_mballoc(1, _mhow);				\
295 	_mm = mmbfree;							\
296 	if (_mm != NULL) {						\
297 		mmbfree = _mm->m_next;					\
298 		mbstat.m_mtypes[MT_FREE]--;				\
299 		_mm->m_type = _mtype;					\
300 		mbstat.m_mtypes[_mtype]++;				\
301 		_mm->m_next = NULL;					\
302 		_mm->m_nextpkt = NULL;					\
303 		_mm->m_data = _mm->m_pktdat;				\
304 		_mm->m_flags = M_PKTHDR;				\
305 		(m) = _mm;						\
306 		splx(_ms);						\
307 	} else {							\
308 		splx(_ms);						\
309 		_mm = m_retryhdr(_mhow, _mtype);			\
310 		if (_mm == NULL && _mhow == M_WAIT)			\
311 			(m) = m_mballoc_wait(MGETHDR_C, _mtype);	\
312 		else							\
313 			(m) = _mm;					\
314 	}								\
315 } while (0)
316 
317 /*
318  * Mbuf cluster macros.
319  * MCLALLOC(caddr_t p, int how) allocates an mbuf cluster.
320  * MCLGET adds such clusters to a normal mbuf;
321  * the flag M_EXT is set upon success.
322  * MCLFREE releases a reference to a cluster allocated by MCLALLOC,
323  * freeing the cluster if the reference count has reached 0.
324  */
325 #define	MCLALLOC(p, how) do {						\
326 	caddr_t _mp;							\
327 	int _mhow = (how);						\
328 	int _ms = splimp();						\
329 									\
330 	if (mclfree == NULL)						\
331 		(void)m_clalloc(1, _mhow);				\
332 	_mp = (caddr_t)mclfree;						\
333 	if (_mp != NULL) {						\
334 		mclrefcnt[mtocl(_mp)]++;				\
335 		mbstat.m_clfree--;					\
336 		mclfree = ((union mcluster *)_mp)->mcl_next;		\
337 		(p) = _mp;						\
338 		splx(_ms);						\
339 	} else {							\
340 		splx(_ms);						\
341 		if (_mhow == M_WAIT)					\
342 			(p) = m_clalloc_wait();				\
343 		else							\
344 			(p) = NULL;					\
345 	}								\
346 } while (0)
347 
348 #define	MCLGET(m, how) do {						\
349 	struct mbuf *_mm = (m);						\
350 									\
351 	MCLALLOC(_mm->m_ext.ext_buf, (how));				\
352 	if (_mm->m_ext.ext_buf != NULL) {				\
353 		_mm->m_data = _mm->m_ext.ext_buf;			\
354 		_mm->m_flags |= M_EXT;					\
355 		_mm->m_ext.ext_free = NULL;				\
356 		_mm->m_ext.ext_ref = NULL;				\
357 		_mm->m_ext.ext_size = MCLBYTES;				\
358 	}								\
359 } while (0)
360 
361 #define	MCLFREE1(p) do {						\
362 	union mcluster *_mp = (union mcluster *)(p);			\
363 									\
364 	if (--mclrefcnt[mtocl(_mp)] == 0) {				\
365 		_mp->mcl_next = mclfree;				\
366 		mclfree = _mp;						\
367 		mbstat.m_clfree++;					\
368 		MCLWAKEUP();						\
369 	}								\
370 } while (0)
371 
372 #define	MCLFREE(p) MBUFLOCK(						\
373 	MCLFREE1(p);							\
374 )
375 
376 #define	MEXTFREE1(m) do {						\
377 		struct mbuf *_mm = (m);					\
378 									\
379 		if (_mm->m_ext.ext_free != NULL)			\
380 			(*_mm->m_ext.ext_free)(_mm->m_ext.ext_buf,	\
381 		    	    _mm->m_ext.ext_size);			\
382 		else							\
383 			MCLFREE1(_mm->m_ext.ext_buf);			\
384 } while (0)
385 
386 #define	MEXTFREE(m) MBUFLOCK(						\
387 	MEXTFREE1(m);							\
388 )
389 
390 /*
391  * MFREE(struct mbuf *m, struct mbuf *n)
392  * Free a single mbuf and associated external storage.
393  * Place the successor, if any, in n.
394  */
395 #define	MFREE(m, n) MBUFLOCK(						\
396 	struct mbuf *_mm = (m);						\
397 									\
398 	mbstat.m_mtypes[_mm->m_type]--;					\
399 	if (_mm->m_flags & M_EXT)					\
400 		MEXTFREE1(m);						\
401 	(n) = _mm->m_next;						\
402 	_mm->m_type = MT_FREE;						\
403 	mbstat.m_mtypes[MT_FREE]++;					\
404 	_mm->m_next = mmbfree;						\
405 	mmbfree = _mm;							\
406 	MMBWAKEUP();							\
407 )
408 
409 /*
410  * Copy mbuf pkthdr from "from" to "to".
411  * from must have M_PKTHDR set, and to must be empty.
412  */
413 #define	M_COPY_PKTHDR(to, from) do {					\
414 	struct mbuf *_mfrom = (from);					\
415 	struct mbuf *_mto = (to);					\
416 									\
417 	_mto->m_data = _mto->m_pktdat;					\
418 	_mto->m_flags = _mfrom->m_flags & M_COPYFLAGS;			\
419 	_mto->m_pkthdr = _mfrom->m_pkthdr;				\
420 } while (0)
421 
422 /*
423  * Set the m_data pointer of a newly-allocated mbuf (m_get/MGET) to place
424  * an object of the specified size at the end of the mbuf, longword aligned.
425  */
426 #define	M_ALIGN(m, len) do {						\
427 	(m)->m_data += (MLEN - (len)) & ~(sizeof(long) - 1);		\
428 } while (0)
429 
430 /*
431  * As above, for mbufs allocated with m_gethdr/MGETHDR
432  * or initialized by M_COPY_PKTHDR.
433  */
434 #define	MH_ALIGN(m, len) do {						\
435 	(m)->m_data += (MHLEN - (len)) & ~(sizeof(long) - 1);		\
436 } while (0)
437 
438 /*
439  * Compute the amount of space available
440  * before the current start of data in an mbuf.
441  */
442 #define	M_LEADINGSPACE(m)						\
443 	((m)->m_flags & M_EXT ?						\
444 	    /* (m)->m_data - (m)->m_ext.ext_buf */ 0 :			\
445 	    (m)->m_flags & M_PKTHDR ? (m)->m_data - (m)->m_pktdat :	\
446 	    (m)->m_data - (m)->m_dat)
447 
448 /*
449  * Compute the amount of space available
450  * after the end of data in an mbuf.
451  */
452 #define	M_TRAILINGSPACE(m)						\
453 	((m)->m_flags & M_EXT ? (m)->m_ext.ext_buf +			\
454 	    (m)->m_ext.ext_size - ((m)->m_data + (m)->m_len) :		\
455 	    &(m)->m_dat[MLEN] - ((m)->m_data + (m)->m_len))
456 
457 /*
458  * Arrange to prepend space of size plen to mbuf m.
459  * If a new mbuf must be allocated, how specifies whether to wait.
460  * If how is M_DONTWAIT and allocation fails, the original mbuf chain
461  * is freed and m is set to NULL.
462  */
463 #define	M_PREPEND(m, plen, how) do {					\
464 	struct mbuf **_mmp = &(m);					\
465 	struct mbuf *_mm = *_mmp;					\
466 	int _mplen = (plen);						\
467 	int __mhow = (how);						\
468 									\
469 	if (_mm == NULL) {						\
470 		MGET(_mm, __mhow, MT_DATA);				\
471 		if (_mm == NULL)					\
472 			break;						\
473 	}								\
474 	if (M_LEADINGSPACE(_mm) >= _mplen) {				\
475 		_mm->m_data -= _mplen;					\
476 		_mm->m_len += _mplen;					\
477 	} else								\
478 		_mm = m_prepend(_mm, _mplen, __mhow);			\
479 	if (_mm->m_flags & M_PKTHDR)					\
480 		_mm->m_pkthdr.len += _mplen;				\
481 	*_mmp = _mm;							\
482 } while (0)
483 
484 /* change mbuf to new type */
485 #define	MCHTYPE(m, t) do {						\
486 	struct mbuf *_mm = (m);						\
487 	int _mt = (t);							\
488 	int _ms = splimp();						\
489 									\
490 	mbstat.m_mtypes[_mm->m_type]--;					\
491 	mbstat.m_mtypes[_mt]++;						\
492 	splx(_ms);							\
493 	_mm->m_type = (_mt);						\
494 } while (0)
495 
496 /* length to m_copy to copy all */
497 #define	M_COPYALL	1000000000
498 
499 /* compatibility with 4.3 */
500 #define	m_copy(m, o, l)	m_copym((m), (o), (l), M_DONTWAIT)
501 
502 #ifdef _KERNEL
503 extern	u_int		 m_clalloc_wid;	/* mbuf cluster wait count */
504 extern	u_int		 m_mballoc_wid;	/* mbuf wait count */
505 extern	int		 max_linkhdr;	/* largest link-level header */
506 extern	int		 max_protohdr;	/* largest protocol header */
507 extern	int		 max_hdr;	/* largest link+protocol header */
508 extern	int		 max_datalen;	/* MHLEN - max_hdr */
509 extern	struct mbstat	 mbstat;
510 extern	int		 mbuf_wait;	/* mbuf sleep time */
511 extern	struct mbuf	*mbutl;		/* virtual address of mclusters */
512 extern	char		*mclrefcnt;	/* cluster reference counts */
513 extern	union mcluster	*mclfree;
514 extern	struct mbuf	*mmbfree;
515 extern	int		 nmbclusters;
516 extern	int		 nmbufs;
517 extern	int		 nsfbufs;
518 
519 void	m_adj __P((struct mbuf *, int));
520 void	m_cat __P((struct mbuf *,struct mbuf *));
521 int	m_clalloc __P((int, int));
522 caddr_t	m_clalloc_wait __P((void));
523 void	m_copyback __P((struct mbuf *, int, int, caddr_t));
524 void	m_copydata __P((struct mbuf *,int,int,caddr_t));
525 struct	mbuf *m_copym __P((struct mbuf *, int, int, int));
526 struct	mbuf *m_copypacket __P((struct mbuf *, int));
527 struct	mbuf *m_devget __P((char *, int, int, struct ifnet *,
528     void (*copy)(char *, caddr_t, u_int)));
529 struct	mbuf *m_dup __P((struct mbuf *, int));
530 struct	mbuf *m_free __P((struct mbuf *));
531 void	m_freem __P((struct mbuf *));
532 struct	mbuf *m_get __P((int, int));
533 struct	mbuf *m_getclr __P((int, int));
534 struct	mbuf *m_gethdr __P((int, int));
535 int	m_mballoc __P((int, int));
536 struct	mbuf *m_mballoc_wait __P((int, int));
537 struct	mbuf *m_prepend __P((struct mbuf *,int,int));
538 void	m_print __P((const struct mbuf *m));
539 struct	mbuf *m_pullup __P((struct mbuf *, int));
540 struct	mbuf *m_retry __P((int, int));
541 struct	mbuf *m_retryhdr __P((int, int));
542 struct	mbuf *m_split __P((struct mbuf *,int,int));
543 #endif /* _KERNEL */
544 
545 #endif /* !_SYS_MBUF_H_ */
546