xref: /freebsd/sys/kern/uipc_mbuf.c (revision b2d48be1bc7df45ddd13b143a160d0acb5a383c5)
1 /*-
2  * Copyright (c) 1982, 1986, 1988, 1991, 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  * 4. Neither the name of the University nor the names of its contributors
14  *    may be used to endorse or promote products derived from this software
15  *    without specific prior written permission.
16  *
17  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
18  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
19  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
20  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
21  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
22  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
23  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
24  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
25  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
26  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
27  * SUCH DAMAGE.
28  *
29  *	@(#)uipc_mbuf.c	8.2 (Berkeley) 1/4/94
30  */
31 
32 #include <sys/cdefs.h>
33 __FBSDID("$FreeBSD$");
34 
35 #include "opt_param.h"
36 #include "opt_mbuf_stress_test.h"
37 #include "opt_mbuf_profiling.h"
38 
39 #include <sys/param.h>
40 #include <sys/systm.h>
41 #include <sys/kernel.h>
42 #include <sys/limits.h>
43 #include <sys/lock.h>
44 #include <sys/malloc.h>
45 #include <sys/mbuf.h>
46 #include <sys/sysctl.h>
47 #include <sys/domain.h>
48 #include <sys/protosw.h>
49 #include <sys/uio.h>
50 
51 int	max_linkhdr;
52 int	max_protohdr;
53 int	max_hdr;
54 int	max_datalen;
55 #ifdef MBUF_STRESS_TEST
56 int	m_defragpackets;
57 int	m_defragbytes;
58 int	m_defraguseless;
59 int	m_defragfailure;
60 int	m_defragrandomfailures;
61 #endif
62 
63 /*
64  * sysctl(8) exported objects
65  */
66 SYSCTL_INT(_kern_ipc, KIPC_MAX_LINKHDR, max_linkhdr, CTLFLAG_RD,
67 	   &max_linkhdr, 0, "Size of largest link layer header");
68 SYSCTL_INT(_kern_ipc, KIPC_MAX_PROTOHDR, max_protohdr, CTLFLAG_RD,
69 	   &max_protohdr, 0, "Size of largest protocol layer header");
70 SYSCTL_INT(_kern_ipc, KIPC_MAX_HDR, max_hdr, CTLFLAG_RD,
71 	   &max_hdr, 0, "Size of largest link plus protocol header");
72 SYSCTL_INT(_kern_ipc, KIPC_MAX_DATALEN, max_datalen, CTLFLAG_RD,
73 	   &max_datalen, 0, "Minimum space left in mbuf after max_hdr");
74 #ifdef MBUF_STRESS_TEST
75 SYSCTL_INT(_kern_ipc, OID_AUTO, m_defragpackets, CTLFLAG_RD,
76 	   &m_defragpackets, 0, "");
77 SYSCTL_INT(_kern_ipc, OID_AUTO, m_defragbytes, CTLFLAG_RD,
78 	   &m_defragbytes, 0, "");
79 SYSCTL_INT(_kern_ipc, OID_AUTO, m_defraguseless, CTLFLAG_RD,
80 	   &m_defraguseless, 0, "");
81 SYSCTL_INT(_kern_ipc, OID_AUTO, m_defragfailure, CTLFLAG_RD,
82 	   &m_defragfailure, 0, "");
83 SYSCTL_INT(_kern_ipc, OID_AUTO, m_defragrandomfailures, CTLFLAG_RW,
84 	   &m_defragrandomfailures, 0, "");
85 #endif
86 
87 /*
88  * Ensure the correct size of various mbuf parameters.  It could be off due
89  * to compiler-induced padding and alignment artifacts.
90  */
91 CTASSERT(MSIZE - offsetof(struct mbuf, m_dat) == MLEN);
92 CTASSERT(MSIZE - offsetof(struct mbuf, m_pktdat) == MHLEN);
93 
94 /*
95  * mbuf data storage should be 64-bit aligned regardless of architectural
96  * pointer size; check this is the case with and without a packet header.
97  */
98 CTASSERT(offsetof(struct mbuf, m_dat) % 8 == 0);
99 CTASSERT(offsetof(struct mbuf, m_pktdat) % 8 == 0);
100 
101 /*
102  * While the specific values here don't matter too much (i.e., +/- a few
103  * words), we do want to ensure that changes to these values are carefully
104  * reasoned about and properly documented.  This is especially the case as
105  * network-protocol and device-driver modules encode these layouts, and must
106  * be recompiled if the structures change.  Check these values at compile time
107  * against the ones documented in comments in mbuf.h.
108  *
109  * NB: Possibly they should be documented there via #define's and not just
110  * comments.
111  */
112 #if defined(__LP64__)
113 CTASSERT(offsetof(struct mbuf, m_dat) == 32);
114 CTASSERT(sizeof(struct pkthdr) == 56);
115 CTASSERT(sizeof(struct m_ext) == 48);
116 #else
117 CTASSERT(offsetof(struct mbuf, m_dat) == 24);
118 CTASSERT(sizeof(struct pkthdr) == 48);
119 CTASSERT(sizeof(struct m_ext) == 28);
120 #endif
121 
122 /*
123  * Assert that the queue(3) macros produce code of the same size as an old
124  * plain pointer does.
125  */
126 #ifdef INVARIANTS
127 static struct mbuf m_assertbuf;
128 CTASSERT(sizeof(m_assertbuf.m_slist) == sizeof(m_assertbuf.m_next));
129 CTASSERT(sizeof(m_assertbuf.m_stailq) == sizeof(m_assertbuf.m_next));
130 CTASSERT(sizeof(m_assertbuf.m_slistpkt) == sizeof(m_assertbuf.m_nextpkt));
131 CTASSERT(sizeof(m_assertbuf.m_stailqpkt) == sizeof(m_assertbuf.m_nextpkt));
132 #endif
133 
134 /*
135  * m_get2() allocates minimum mbuf that would fit "size" argument.
136  */
137 struct mbuf *
138 m_get2(int size, int how, short type, int flags)
139 {
140 	struct mb_args args;
141 	struct mbuf *m, *n;
142 
143 	args.flags = flags;
144 	args.type = type;
145 
146 	if (size <= MHLEN || (size <= MLEN && (flags & M_PKTHDR) == 0))
147 		return (uma_zalloc_arg(zone_mbuf, &args, how));
148 	if (size <= MCLBYTES)
149 		return (uma_zalloc_arg(zone_pack, &args, how));
150 
151 	if (size > MJUMPAGESIZE)
152 		return (NULL);
153 
154 	m = uma_zalloc_arg(zone_mbuf, &args, how);
155 	if (m == NULL)
156 		return (NULL);
157 
158 	n = uma_zalloc_arg(zone_jumbop, m, how);
159 	if (n == NULL) {
160 		uma_zfree(zone_mbuf, m);
161 		return (NULL);
162 	}
163 
164 	return (m);
165 }
166 
167 /*
168  * m_getjcl() returns an mbuf with a cluster of the specified size attached.
169  * For size it takes MCLBYTES, MJUMPAGESIZE, MJUM9BYTES, MJUM16BYTES.
170  */
171 struct mbuf *
172 m_getjcl(int how, short type, int flags, int size)
173 {
174 	struct mb_args args;
175 	struct mbuf *m, *n;
176 	uma_zone_t zone;
177 
178 	if (size == MCLBYTES)
179 		return m_getcl(how, type, flags);
180 
181 	args.flags = flags;
182 	args.type = type;
183 
184 	m = uma_zalloc_arg(zone_mbuf, &args, how);
185 	if (m == NULL)
186 		return (NULL);
187 
188 	zone = m_getzone(size);
189 	n = uma_zalloc_arg(zone, m, how);
190 	if (n == NULL) {
191 		uma_zfree(zone_mbuf, m);
192 		return (NULL);
193 	}
194 	return (m);
195 }
196 
197 /*
198  * Allocate a given length worth of mbufs and/or clusters (whatever fits
199  * best) and return a pointer to the top of the allocated chain.  If an
200  * existing mbuf chain is provided, then we will append the new chain
201  * to the existing one but still return the top of the newly allocated
202  * chain.
203  */
204 struct mbuf *
205 m_getm2(struct mbuf *m, int len, int how, short type, int flags)
206 {
207 	struct mbuf *mb, *nm = NULL, *mtail = NULL;
208 
209 	KASSERT(len >= 0, ("%s: len is < 0", __func__));
210 
211 	/* Validate flags. */
212 	flags &= (M_PKTHDR | M_EOR);
213 
214 	/* Packet header mbuf must be first in chain. */
215 	if ((flags & M_PKTHDR) && m != NULL)
216 		flags &= ~M_PKTHDR;
217 
218 	/* Loop and append maximum sized mbufs to the chain tail. */
219 	while (len > 0) {
220 		if (len > MCLBYTES)
221 			mb = m_getjcl(how, type, (flags & M_PKTHDR),
222 			    MJUMPAGESIZE);
223 		else if (len >= MINCLSIZE)
224 			mb = m_getcl(how, type, (flags & M_PKTHDR));
225 		else if (flags & M_PKTHDR)
226 			mb = m_gethdr(how, type);
227 		else
228 			mb = m_get(how, type);
229 
230 		/* Fail the whole operation if one mbuf can't be allocated. */
231 		if (mb == NULL) {
232 			if (nm != NULL)
233 				m_freem(nm);
234 			return (NULL);
235 		}
236 
237 		/* Book keeping. */
238 		len -= M_SIZE(mb);
239 		if (mtail != NULL)
240 			mtail->m_next = mb;
241 		else
242 			nm = mb;
243 		mtail = mb;
244 		flags &= ~M_PKTHDR;	/* Only valid on the first mbuf. */
245 	}
246 	if (flags & M_EOR)
247 		mtail->m_flags |= M_EOR;  /* Only valid on the last mbuf. */
248 
249 	/* If mbuf was supplied, append new chain to the end of it. */
250 	if (m != NULL) {
251 		for (mtail = m; mtail->m_next != NULL; mtail = mtail->m_next)
252 			;
253 		mtail->m_next = nm;
254 		mtail->m_flags &= ~M_EOR;
255 	} else
256 		m = nm;
257 
258 	return (m);
259 }
260 
261 /*
262  * Free an entire chain of mbufs and associated external buffers, if
263  * applicable.
264  */
265 void
266 m_freem(struct mbuf *mb)
267 {
268 
269 	while (mb != NULL)
270 		mb = m_free(mb);
271 }
272 
273 /*-
274  * Configure a provided mbuf to refer to the provided external storage
275  * buffer and setup a reference count for said buffer.  If the setting
276  * up of the reference count fails, the M_EXT bit will not be set.  If
277  * successfull, the M_EXT bit is set in the mbuf's flags.
278  *
279  * Arguments:
280  *    mb     The existing mbuf to which to attach the provided buffer.
281  *    buf    The address of the provided external storage buffer.
282  *    size   The size of the provided buffer.
283  *    freef  A pointer to a routine that is responsible for freeing the
284  *           provided external storage buffer.
285  *    args   A pointer to an argument structure (of any type) to be passed
286  *           to the provided freef routine (may be NULL).
287  *    flags  Any other flags to be passed to the provided mbuf.
288  *    type   The type that the external storage buffer should be
289  *           labeled with.
290  *
291  * Returns:
292  *    Nothing.
293  */
294 int
295 m_extadd(struct mbuf *mb, caddr_t buf, u_int size,
296     void (*freef)(struct mbuf *, void *, void *), void *arg1, void *arg2,
297     int flags, int type, int wait)
298 {
299 	KASSERT(type != EXT_CLUSTER, ("%s: EXT_CLUSTER not allowed", __func__));
300 
301 	if (type != EXT_EXTREF)
302 		mb->m_ext.ext_cnt = uma_zalloc(zone_ext_refcnt, wait);
303 
304 	if (mb->m_ext.ext_cnt == NULL)
305 		return (ENOMEM);
306 
307 	*(mb->m_ext.ext_cnt) = 1;
308 	mb->m_flags |= (M_EXT | flags);
309 	mb->m_ext.ext_buf = buf;
310 	mb->m_data = mb->m_ext.ext_buf;
311 	mb->m_ext.ext_size = size;
312 	mb->m_ext.ext_free = freef;
313 	mb->m_ext.ext_arg1 = arg1;
314 	mb->m_ext.ext_arg2 = arg2;
315 	mb->m_ext.ext_type = type;
316 	mb->m_ext.ext_flags = 0;
317 
318 	return (0);
319 }
320 
321 /*
322  * Non-directly-exported function to clean up after mbufs with M_EXT
323  * storage attached to them if the reference count hits 1.
324  */
325 void
326 mb_free_ext(struct mbuf *m)
327 {
328 	int freembuf;
329 
330 	KASSERT(m->m_flags & M_EXT, ("%s: M_EXT not set on %p", __func__, m));
331 
332 	/*
333 	 * Check if the header is embedded in the cluster.
334 	 */
335 	freembuf = (m->m_flags & M_NOFREE) ? 0 : 1;
336 
337 	switch (m->m_ext.ext_type) {
338 	case EXT_SFBUF:
339 		sf_ext_free(m->m_ext.ext_arg1, m->m_ext.ext_arg2);
340 		break;
341 	default:
342 		KASSERT(m->m_ext.ext_cnt != NULL,
343 		    ("%s: no refcounting pointer on %p", __func__, m));
344 		/*
345 		 * Free attached storage if this mbuf is the only
346 		 * reference to it.
347 		 */
348 		if (*(m->m_ext.ext_cnt) != 1) {
349 			if (atomic_fetchadd_int(m->m_ext.ext_cnt, -1) != 1)
350 				break;
351 		}
352 
353 		switch (m->m_ext.ext_type) {
354 		case EXT_PACKET:	/* The packet zone is special. */
355 			if (*(m->m_ext.ext_cnt) == 0)
356 				*(m->m_ext.ext_cnt) = 1;
357 			uma_zfree(zone_pack, m);
358 			return;		/* Job done. */
359 		case EXT_CLUSTER:
360 			uma_zfree(zone_clust, m->m_ext.ext_buf);
361 			break;
362 		case EXT_JUMBOP:
363 			uma_zfree(zone_jumbop, m->m_ext.ext_buf);
364 			break;
365 		case EXT_JUMBO9:
366 			uma_zfree(zone_jumbo9, m->m_ext.ext_buf);
367 			break;
368 		case EXT_JUMBO16:
369 			uma_zfree(zone_jumbo16, m->m_ext.ext_buf);
370 			break;
371 		case EXT_NET_DRV:
372 		case EXT_MOD_TYPE:
373 		case EXT_DISPOSABLE:
374 			*(m->m_ext.ext_cnt) = 0;
375 			uma_zfree(zone_ext_refcnt, __DEVOLATILE(u_int *,
376 				m->m_ext.ext_cnt));
377 			/* FALLTHROUGH */
378 		case EXT_EXTREF:
379 			KASSERT(m->m_ext.ext_free != NULL,
380 				("%s: ext_free not set", __func__));
381 			(*(m->m_ext.ext_free))(m, m->m_ext.ext_arg1,
382 			    m->m_ext.ext_arg2);
383 			break;
384 		default:
385 			KASSERT(m->m_ext.ext_type == 0,
386 				("%s: unknown ext_type", __func__));
387 		}
388 	}
389 
390 	if (freembuf)
391 		uma_zfree(zone_mbuf, m);
392 }
393 
394 /*
395  * Attach the cluster from *m to *n, set up m_ext in *n
396  * and bump the refcount of the cluster.
397  */
398 static void
399 mb_dupcl(struct mbuf *n, const struct mbuf *m)
400 {
401 
402 	KASSERT(m->m_flags & M_EXT, ("%s: M_EXT not set on %p", __func__, m));
403 	KASSERT(!(n->m_flags & M_EXT), ("%s: M_EXT set on %p", __func__, n));
404 
405 	switch (m->m_ext.ext_type) {
406 	case EXT_SFBUF:
407 		sf_ext_ref(m->m_ext.ext_arg1, m->m_ext.ext_arg2);
408 		break;
409 	default:
410 		KASSERT(m->m_ext.ext_cnt != NULL,
411 		    ("%s: no refcounting pointer on %p", __func__, m));
412 		if (*(m->m_ext.ext_cnt) == 1)
413 			*(m->m_ext.ext_cnt) += 1;
414 		else
415 			atomic_add_int(m->m_ext.ext_cnt, 1);
416 	}
417 
418 	n->m_ext = m->m_ext;
419 	n->m_flags |= M_EXT;
420 	n->m_flags |= m->m_flags & M_RDONLY;
421 }
422 
423 void
424 m_demote_pkthdr(struct mbuf *m)
425 {
426 
427 	M_ASSERTPKTHDR(m);
428 
429 	m_tag_delete_chain(m, NULL);
430 	m->m_flags &= ~M_PKTHDR;
431 	bzero(&m->m_pkthdr, sizeof(struct pkthdr));
432 }
433 
434 /*
435  * Clean up mbuf (chain) from any tags and packet headers.
436  * If "all" is set then the first mbuf in the chain will be
437  * cleaned too.
438  */
439 void
440 m_demote(struct mbuf *m0, int all, int flags)
441 {
442 	struct mbuf *m;
443 
444 	for (m = all ? m0 : m0->m_next; m != NULL; m = m->m_next) {
445 		KASSERT(m->m_nextpkt == NULL, ("%s: m_nextpkt in m %p, m0 %p",
446 		    __func__, m, m0));
447 		if (m->m_flags & M_PKTHDR)
448 			m_demote_pkthdr(m);
449 		m->m_flags = m->m_flags & (M_EXT | M_RDONLY | M_NOFREE | flags);
450 	}
451 }
452 
453 /*
454  * Sanity checks on mbuf (chain) for use in KASSERT() and general
455  * debugging.
456  * Returns 0 or panics when bad and 1 on all tests passed.
457  * Sanitize, 0 to run M_SANITY_ACTION, 1 to garble things so they
458  * blow up later.
459  */
460 int
461 m_sanity(struct mbuf *m0, int sanitize)
462 {
463 	struct mbuf *m;
464 	caddr_t a, b;
465 	int pktlen = 0;
466 
467 #ifdef INVARIANTS
468 #define	M_SANITY_ACTION(s)	panic("mbuf %p: " s, m)
469 #else
470 #define	M_SANITY_ACTION(s)	printf("mbuf %p: " s, m)
471 #endif
472 
473 	for (m = m0; m != NULL; m = m->m_next) {
474 		/*
475 		 * Basic pointer checks.  If any of these fails then some
476 		 * unrelated kernel memory before or after us is trashed.
477 		 * No way to recover from that.
478 		 */
479 		a = M_START(m);
480 		b = a + M_SIZE(m);
481 		if ((caddr_t)m->m_data < a)
482 			M_SANITY_ACTION("m_data outside mbuf data range left");
483 		if ((caddr_t)m->m_data > b)
484 			M_SANITY_ACTION("m_data outside mbuf data range right");
485 		if ((caddr_t)m->m_data + m->m_len > b)
486 			M_SANITY_ACTION("m_data + m_len exeeds mbuf space");
487 
488 		/* m->m_nextpkt may only be set on first mbuf in chain. */
489 		if (m != m0 && m->m_nextpkt != NULL) {
490 			if (sanitize) {
491 				m_freem(m->m_nextpkt);
492 				m->m_nextpkt = (struct mbuf *)0xDEADC0DE;
493 			} else
494 				M_SANITY_ACTION("m->m_nextpkt on in-chain mbuf");
495 		}
496 
497 		/* packet length (not mbuf length!) calculation */
498 		if (m0->m_flags & M_PKTHDR)
499 			pktlen += m->m_len;
500 
501 		/* m_tags may only be attached to first mbuf in chain. */
502 		if (m != m0 && m->m_flags & M_PKTHDR &&
503 		    !SLIST_EMPTY(&m->m_pkthdr.tags)) {
504 			if (sanitize) {
505 				m_tag_delete_chain(m, NULL);
506 				/* put in 0xDEADC0DE perhaps? */
507 			} else
508 				M_SANITY_ACTION("m_tags on in-chain mbuf");
509 		}
510 
511 		/* M_PKTHDR may only be set on first mbuf in chain */
512 		if (m != m0 && m->m_flags & M_PKTHDR) {
513 			if (sanitize) {
514 				bzero(&m->m_pkthdr, sizeof(m->m_pkthdr));
515 				m->m_flags &= ~M_PKTHDR;
516 				/* put in 0xDEADCODE and leave hdr flag in */
517 			} else
518 				M_SANITY_ACTION("M_PKTHDR on in-chain mbuf");
519 		}
520 	}
521 	m = m0;
522 	if (pktlen && pktlen != m->m_pkthdr.len) {
523 		if (sanitize)
524 			m->m_pkthdr.len = 0;
525 		else
526 			M_SANITY_ACTION("m_pkthdr.len != mbuf chain length");
527 	}
528 	return 1;
529 
530 #undef	M_SANITY_ACTION
531 }
532 
533 
534 /*
535  * "Move" mbuf pkthdr from "from" to "to".
536  * "from" must have M_PKTHDR set, and "to" must be empty.
537  */
538 void
539 m_move_pkthdr(struct mbuf *to, struct mbuf *from)
540 {
541 
542 #if 0
543 	/* see below for why these are not enabled */
544 	M_ASSERTPKTHDR(to);
545 	/* Note: with MAC, this may not be a good assertion. */
546 	KASSERT(SLIST_EMPTY(&to->m_pkthdr.tags),
547 	    ("m_move_pkthdr: to has tags"));
548 #endif
549 #ifdef MAC
550 	/*
551 	 * XXXMAC: It could be this should also occur for non-MAC?
552 	 */
553 	if (to->m_flags & M_PKTHDR)
554 		m_tag_delete_chain(to, NULL);
555 #endif
556 	to->m_flags = (from->m_flags & M_COPYFLAGS) | (to->m_flags & M_EXT);
557 	if ((to->m_flags & M_EXT) == 0)
558 		to->m_data = to->m_pktdat;
559 	to->m_pkthdr = from->m_pkthdr;		/* especially tags */
560 	SLIST_INIT(&from->m_pkthdr.tags);	/* purge tags from src */
561 	from->m_flags &= ~M_PKTHDR;
562 }
563 
564 /*
565  * Duplicate "from"'s mbuf pkthdr in "to".
566  * "from" must have M_PKTHDR set, and "to" must be empty.
567  * In particular, this does a deep copy of the packet tags.
568  */
569 int
570 m_dup_pkthdr(struct mbuf *to, const struct mbuf *from, int how)
571 {
572 
573 #if 0
574 	/*
575 	 * The mbuf allocator only initializes the pkthdr
576 	 * when the mbuf is allocated with m_gethdr(). Many users
577 	 * (e.g. m_copy*, m_prepend) use m_get() and then
578 	 * smash the pkthdr as needed causing these
579 	 * assertions to trip.  For now just disable them.
580 	 */
581 	M_ASSERTPKTHDR(to);
582 	/* Note: with MAC, this may not be a good assertion. */
583 	KASSERT(SLIST_EMPTY(&to->m_pkthdr.tags), ("m_dup_pkthdr: to has tags"));
584 #endif
585 	MBUF_CHECKSLEEP(how);
586 #ifdef MAC
587 	if (to->m_flags & M_PKTHDR)
588 		m_tag_delete_chain(to, NULL);
589 #endif
590 	to->m_flags = (from->m_flags & M_COPYFLAGS) | (to->m_flags & M_EXT);
591 	if ((to->m_flags & M_EXT) == 0)
592 		to->m_data = to->m_pktdat;
593 	to->m_pkthdr = from->m_pkthdr;
594 	SLIST_INIT(&to->m_pkthdr.tags);
595 	return (m_tag_copy_chain(to, from, how));
596 }
597 
598 /*
599  * Lesser-used path for M_PREPEND:
600  * allocate new mbuf to prepend to chain,
601  * copy junk along.
602  */
603 struct mbuf *
604 m_prepend(struct mbuf *m, int len, int how)
605 {
606 	struct mbuf *mn;
607 
608 	if (m->m_flags & M_PKTHDR)
609 		mn = m_gethdr(how, m->m_type);
610 	else
611 		mn = m_get(how, m->m_type);
612 	if (mn == NULL) {
613 		m_freem(m);
614 		return (NULL);
615 	}
616 	if (m->m_flags & M_PKTHDR)
617 		m_move_pkthdr(mn, m);
618 	mn->m_next = m;
619 	m = mn;
620 	if (len < M_SIZE(m))
621 		M_ALIGN(m, len);
622 	m->m_len = len;
623 	return (m);
624 }
625 
626 /*
627  * Make a copy of an mbuf chain starting "off0" bytes from the beginning,
628  * continuing for "len" bytes.  If len is M_COPYALL, copy to end of mbuf.
629  * The wait parameter is a choice of M_WAITOK/M_NOWAIT from caller.
630  * Note that the copy is read-only, because clusters are not copied,
631  * only their reference counts are incremented.
632  */
633 struct mbuf *
634 m_copym(const struct mbuf *m, int off0, int len, int wait)
635 {
636 	struct mbuf *n, **np;
637 	int off = off0;
638 	struct mbuf *top;
639 	int copyhdr = 0;
640 
641 	KASSERT(off >= 0, ("m_copym, negative off %d", off));
642 	KASSERT(len >= 0, ("m_copym, negative len %d", len));
643 	MBUF_CHECKSLEEP(wait);
644 	if (off == 0 && m->m_flags & M_PKTHDR)
645 		copyhdr = 1;
646 	while (off > 0) {
647 		KASSERT(m != NULL, ("m_copym, offset > size of mbuf chain"));
648 		if (off < m->m_len)
649 			break;
650 		off -= m->m_len;
651 		m = m->m_next;
652 	}
653 	np = &top;
654 	top = 0;
655 	while (len > 0) {
656 		if (m == NULL) {
657 			KASSERT(len == M_COPYALL,
658 			    ("m_copym, length > size of mbuf chain"));
659 			break;
660 		}
661 		if (copyhdr)
662 			n = m_gethdr(wait, m->m_type);
663 		else
664 			n = m_get(wait, m->m_type);
665 		*np = n;
666 		if (n == NULL)
667 			goto nospace;
668 		if (copyhdr) {
669 			if (!m_dup_pkthdr(n, m, wait))
670 				goto nospace;
671 			if (len == M_COPYALL)
672 				n->m_pkthdr.len -= off0;
673 			else
674 				n->m_pkthdr.len = len;
675 			copyhdr = 0;
676 		}
677 		n->m_len = min(len, m->m_len - off);
678 		if (m->m_flags & M_EXT) {
679 			n->m_data = m->m_data + off;
680 			mb_dupcl(n, m);
681 		} else
682 			bcopy(mtod(m, caddr_t)+off, mtod(n, caddr_t),
683 			    (u_int)n->m_len);
684 		if (len != M_COPYALL)
685 			len -= n->m_len;
686 		off = 0;
687 		m = m->m_next;
688 		np = &n->m_next;
689 	}
690 
691 	return (top);
692 nospace:
693 	m_freem(top);
694 	return (NULL);
695 }
696 
697 /*
698  * Copy an entire packet, including header (which must be present).
699  * An optimization of the common case `m_copym(m, 0, M_COPYALL, how)'.
700  * Note that the copy is read-only, because clusters are not copied,
701  * only their reference counts are incremented.
702  * Preserve alignment of the first mbuf so if the creator has left
703  * some room at the beginning (e.g. for inserting protocol headers)
704  * the copies still have the room available.
705  */
706 struct mbuf *
707 m_copypacket(struct mbuf *m, int how)
708 {
709 	struct mbuf *top, *n, *o;
710 
711 	MBUF_CHECKSLEEP(how);
712 	n = m_get(how, m->m_type);
713 	top = n;
714 	if (n == NULL)
715 		goto nospace;
716 
717 	if (!m_dup_pkthdr(n, m, how))
718 		goto nospace;
719 	n->m_len = m->m_len;
720 	if (m->m_flags & M_EXT) {
721 		n->m_data = m->m_data;
722 		mb_dupcl(n, m);
723 	} else {
724 		n->m_data = n->m_pktdat + (m->m_data - m->m_pktdat );
725 		bcopy(mtod(m, char *), mtod(n, char *), n->m_len);
726 	}
727 
728 	m = m->m_next;
729 	while (m) {
730 		o = m_get(how, m->m_type);
731 		if (o == NULL)
732 			goto nospace;
733 
734 		n->m_next = o;
735 		n = n->m_next;
736 
737 		n->m_len = m->m_len;
738 		if (m->m_flags & M_EXT) {
739 			n->m_data = m->m_data;
740 			mb_dupcl(n, m);
741 		} else {
742 			bcopy(mtod(m, char *), mtod(n, char *), n->m_len);
743 		}
744 
745 		m = m->m_next;
746 	}
747 	return top;
748 nospace:
749 	m_freem(top);
750 	return (NULL);
751 }
752 
753 /*
754  * Copy data from an mbuf chain starting "off" bytes from the beginning,
755  * continuing for "len" bytes, into the indicated buffer.
756  */
757 void
758 m_copydata(const struct mbuf *m, int off, int len, caddr_t cp)
759 {
760 	u_int count;
761 
762 	KASSERT(off >= 0, ("m_copydata, negative off %d", off));
763 	KASSERT(len >= 0, ("m_copydata, negative len %d", len));
764 	while (off > 0) {
765 		KASSERT(m != NULL, ("m_copydata, offset > size of mbuf chain"));
766 		if (off < m->m_len)
767 			break;
768 		off -= m->m_len;
769 		m = m->m_next;
770 	}
771 	while (len > 0) {
772 		KASSERT(m != NULL, ("m_copydata, length > size of mbuf chain"));
773 		count = min(m->m_len - off, len);
774 		bcopy(mtod(m, caddr_t) + off, cp, count);
775 		len -= count;
776 		cp += count;
777 		off = 0;
778 		m = m->m_next;
779 	}
780 }
781 
782 /*
783  * Copy a packet header mbuf chain into a completely new chain, including
784  * copying any mbuf clusters.  Use this instead of m_copypacket() when
785  * you need a writable copy of an mbuf chain.
786  */
787 struct mbuf *
788 m_dup(const struct mbuf *m, int how)
789 {
790 	struct mbuf **p, *top = NULL;
791 	int remain, moff, nsize;
792 
793 	MBUF_CHECKSLEEP(how);
794 	/* Sanity check */
795 	if (m == NULL)
796 		return (NULL);
797 	M_ASSERTPKTHDR(m);
798 
799 	/* While there's more data, get a new mbuf, tack it on, and fill it */
800 	remain = m->m_pkthdr.len;
801 	moff = 0;
802 	p = &top;
803 	while (remain > 0 || top == NULL) {	/* allow m->m_pkthdr.len == 0 */
804 		struct mbuf *n;
805 
806 		/* Get the next new mbuf */
807 		if (remain >= MINCLSIZE) {
808 			n = m_getcl(how, m->m_type, 0);
809 			nsize = MCLBYTES;
810 		} else {
811 			n = m_get(how, m->m_type);
812 			nsize = MLEN;
813 		}
814 		if (n == NULL)
815 			goto nospace;
816 
817 		if (top == NULL) {		/* First one, must be PKTHDR */
818 			if (!m_dup_pkthdr(n, m, how)) {
819 				m_free(n);
820 				goto nospace;
821 			}
822 			if ((n->m_flags & M_EXT) == 0)
823 				nsize = MHLEN;
824 			n->m_flags &= ~M_RDONLY;
825 		}
826 		n->m_len = 0;
827 
828 		/* Link it into the new chain */
829 		*p = n;
830 		p = &n->m_next;
831 
832 		/* Copy data from original mbuf(s) into new mbuf */
833 		while (n->m_len < nsize && m != NULL) {
834 			int chunk = min(nsize - n->m_len, m->m_len - moff);
835 
836 			bcopy(m->m_data + moff, n->m_data + n->m_len, chunk);
837 			moff += chunk;
838 			n->m_len += chunk;
839 			remain -= chunk;
840 			if (moff == m->m_len) {
841 				m = m->m_next;
842 				moff = 0;
843 			}
844 		}
845 
846 		/* Check correct total mbuf length */
847 		KASSERT((remain > 0 && m != NULL) || (remain == 0 && m == NULL),
848 		    	("%s: bogus m_pkthdr.len", __func__));
849 	}
850 	return (top);
851 
852 nospace:
853 	m_freem(top);
854 	return (NULL);
855 }
856 
857 /*
858  * Concatenate mbuf chain n to m.
859  * Both chains must be of the same type (e.g. MT_DATA).
860  * Any m_pkthdr is not updated.
861  */
862 void
863 m_cat(struct mbuf *m, struct mbuf *n)
864 {
865 	while (m->m_next)
866 		m = m->m_next;
867 	while (n) {
868 		if (!M_WRITABLE(m) ||
869 		    M_TRAILINGSPACE(m) < n->m_len) {
870 			/* just join the two chains */
871 			m->m_next = n;
872 			return;
873 		}
874 		/* splat the data from one into the other */
875 		bcopy(mtod(n, caddr_t), mtod(m, caddr_t) + m->m_len,
876 		    (u_int)n->m_len);
877 		m->m_len += n->m_len;
878 		n = m_free(n);
879 	}
880 }
881 
882 /*
883  * Concatenate two pkthdr mbuf chains.
884  */
885 void
886 m_catpkt(struct mbuf *m, struct mbuf *n)
887 {
888 
889 	M_ASSERTPKTHDR(m);
890 	M_ASSERTPKTHDR(n);
891 
892 	m->m_pkthdr.len += n->m_pkthdr.len;
893 	m_demote(n, 1, 0);
894 
895 	m_cat(m, n);
896 }
897 
898 void
899 m_adj(struct mbuf *mp, int req_len)
900 {
901 	int len = req_len;
902 	struct mbuf *m;
903 	int count;
904 
905 	if ((m = mp) == NULL)
906 		return;
907 	if (len >= 0) {
908 		/*
909 		 * Trim from head.
910 		 */
911 		while (m != NULL && len > 0) {
912 			if (m->m_len <= len) {
913 				len -= m->m_len;
914 				m->m_len = 0;
915 				m = m->m_next;
916 			} else {
917 				m->m_len -= len;
918 				m->m_data += len;
919 				len = 0;
920 			}
921 		}
922 		if (mp->m_flags & M_PKTHDR)
923 			mp->m_pkthdr.len -= (req_len - len);
924 	} else {
925 		/*
926 		 * Trim from tail.  Scan the mbuf chain,
927 		 * calculating its length and finding the last mbuf.
928 		 * If the adjustment only affects this mbuf, then just
929 		 * adjust and return.  Otherwise, rescan and truncate
930 		 * after the remaining size.
931 		 */
932 		len = -len;
933 		count = 0;
934 		for (;;) {
935 			count += m->m_len;
936 			if (m->m_next == (struct mbuf *)0)
937 				break;
938 			m = m->m_next;
939 		}
940 		if (m->m_len >= len) {
941 			m->m_len -= len;
942 			if (mp->m_flags & M_PKTHDR)
943 				mp->m_pkthdr.len -= len;
944 			return;
945 		}
946 		count -= len;
947 		if (count < 0)
948 			count = 0;
949 		/*
950 		 * Correct length for chain is "count".
951 		 * Find the mbuf with last data, adjust its length,
952 		 * and toss data from remaining mbufs on chain.
953 		 */
954 		m = mp;
955 		if (m->m_flags & M_PKTHDR)
956 			m->m_pkthdr.len = count;
957 		for (; m; m = m->m_next) {
958 			if (m->m_len >= count) {
959 				m->m_len = count;
960 				if (m->m_next != NULL) {
961 					m_freem(m->m_next);
962 					m->m_next = NULL;
963 				}
964 				break;
965 			}
966 			count -= m->m_len;
967 		}
968 	}
969 }
970 
971 /*
972  * Rearange an mbuf chain so that len bytes are contiguous
973  * and in the data area of an mbuf (so that mtod will work
974  * for a structure of size len).  Returns the resulting
975  * mbuf chain on success, frees it and returns null on failure.
976  * If there is room, it will add up to max_protohdr-len extra bytes to the
977  * contiguous region in an attempt to avoid being called next time.
978  */
979 struct mbuf *
980 m_pullup(struct mbuf *n, int len)
981 {
982 	struct mbuf *m;
983 	int count;
984 	int space;
985 
986 	/*
987 	 * If first mbuf has no cluster, and has room for len bytes
988 	 * without shifting current data, pullup into it,
989 	 * otherwise allocate a new mbuf to prepend to the chain.
990 	 */
991 	if ((n->m_flags & M_EXT) == 0 &&
992 	    n->m_data + len < &n->m_dat[MLEN] && n->m_next) {
993 		if (n->m_len >= len)
994 			return (n);
995 		m = n;
996 		n = n->m_next;
997 		len -= m->m_len;
998 	} else {
999 		if (len > MHLEN)
1000 			goto bad;
1001 		m = m_get(M_NOWAIT, n->m_type);
1002 		if (m == NULL)
1003 			goto bad;
1004 		if (n->m_flags & M_PKTHDR)
1005 			m_move_pkthdr(m, n);
1006 	}
1007 	space = &m->m_dat[MLEN] - (m->m_data + m->m_len);
1008 	do {
1009 		count = min(min(max(len, max_protohdr), space), n->m_len);
1010 		bcopy(mtod(n, caddr_t), mtod(m, caddr_t) + m->m_len,
1011 		  (u_int)count);
1012 		len -= count;
1013 		m->m_len += count;
1014 		n->m_len -= count;
1015 		space -= count;
1016 		if (n->m_len)
1017 			n->m_data += count;
1018 		else
1019 			n = m_free(n);
1020 	} while (len > 0 && n);
1021 	if (len > 0) {
1022 		(void) m_free(m);
1023 		goto bad;
1024 	}
1025 	m->m_next = n;
1026 	return (m);
1027 bad:
1028 	m_freem(n);
1029 	return (NULL);
1030 }
1031 
1032 /*
1033  * Like m_pullup(), except a new mbuf is always allocated, and we allow
1034  * the amount of empty space before the data in the new mbuf to be specified
1035  * (in the event that the caller expects to prepend later).
1036  */
1037 struct mbuf *
1038 m_copyup(struct mbuf *n, int len, int dstoff)
1039 {
1040 	struct mbuf *m;
1041 	int count, space;
1042 
1043 	if (len > (MHLEN - dstoff))
1044 		goto bad;
1045 	m = m_get(M_NOWAIT, n->m_type);
1046 	if (m == NULL)
1047 		goto bad;
1048 	if (n->m_flags & M_PKTHDR)
1049 		m_move_pkthdr(m, n);
1050 	m->m_data += dstoff;
1051 	space = &m->m_dat[MLEN] - (m->m_data + m->m_len);
1052 	do {
1053 		count = min(min(max(len, max_protohdr), space), n->m_len);
1054 		memcpy(mtod(m, caddr_t) + m->m_len, mtod(n, caddr_t),
1055 		    (unsigned)count);
1056 		len -= count;
1057 		m->m_len += count;
1058 		n->m_len -= count;
1059 		space -= count;
1060 		if (n->m_len)
1061 			n->m_data += count;
1062 		else
1063 			n = m_free(n);
1064 	} while (len > 0 && n);
1065 	if (len > 0) {
1066 		(void) m_free(m);
1067 		goto bad;
1068 	}
1069 	m->m_next = n;
1070 	return (m);
1071  bad:
1072 	m_freem(n);
1073 	return (NULL);
1074 }
1075 
1076 /*
1077  * Partition an mbuf chain in two pieces, returning the tail --
1078  * all but the first len0 bytes.  In case of failure, it returns NULL and
1079  * attempts to restore the chain to its original state.
1080  *
1081  * Note that the resulting mbufs might be read-only, because the new
1082  * mbuf can end up sharing an mbuf cluster with the original mbuf if
1083  * the "breaking point" happens to lie within a cluster mbuf. Use the
1084  * M_WRITABLE() macro to check for this case.
1085  */
1086 struct mbuf *
1087 m_split(struct mbuf *m0, int len0, int wait)
1088 {
1089 	struct mbuf *m, *n;
1090 	u_int len = len0, remain;
1091 
1092 	MBUF_CHECKSLEEP(wait);
1093 	for (m = m0; m && len > m->m_len; m = m->m_next)
1094 		len -= m->m_len;
1095 	if (m == NULL)
1096 		return (NULL);
1097 	remain = m->m_len - len;
1098 	if (m0->m_flags & M_PKTHDR && remain == 0) {
1099 		n = m_gethdr(wait, m0->m_type);
1100 		if (n == NULL)
1101 			return (NULL);
1102 		n->m_next = m->m_next;
1103 		m->m_next = NULL;
1104 		n->m_pkthdr.rcvif = m0->m_pkthdr.rcvif;
1105 		n->m_pkthdr.len = m0->m_pkthdr.len - len0;
1106 		m0->m_pkthdr.len = len0;
1107 		return (n);
1108 	} else if (m0->m_flags & M_PKTHDR) {
1109 		n = m_gethdr(wait, m0->m_type);
1110 		if (n == NULL)
1111 			return (NULL);
1112 		n->m_pkthdr.rcvif = m0->m_pkthdr.rcvif;
1113 		n->m_pkthdr.len = m0->m_pkthdr.len - len0;
1114 		m0->m_pkthdr.len = len0;
1115 		if (m->m_flags & M_EXT)
1116 			goto extpacket;
1117 		if (remain > MHLEN) {
1118 			/* m can't be the lead packet */
1119 			M_ALIGN(n, 0);
1120 			n->m_next = m_split(m, len, wait);
1121 			if (n->m_next == NULL) {
1122 				(void) m_free(n);
1123 				return (NULL);
1124 			} else {
1125 				n->m_len = 0;
1126 				return (n);
1127 			}
1128 		} else
1129 			M_ALIGN(n, remain);
1130 	} else if (remain == 0) {
1131 		n = m->m_next;
1132 		m->m_next = NULL;
1133 		return (n);
1134 	} else {
1135 		n = m_get(wait, m->m_type);
1136 		if (n == NULL)
1137 			return (NULL);
1138 		M_ALIGN(n, remain);
1139 	}
1140 extpacket:
1141 	if (m->m_flags & M_EXT) {
1142 		n->m_data = m->m_data + len;
1143 		mb_dupcl(n, m);
1144 	} else {
1145 		bcopy(mtod(m, caddr_t) + len, mtod(n, caddr_t), remain);
1146 	}
1147 	n->m_len = remain;
1148 	m->m_len = len;
1149 	n->m_next = m->m_next;
1150 	m->m_next = NULL;
1151 	return (n);
1152 }
1153 /*
1154  * Routine to copy from device local memory into mbufs.
1155  * Note that `off' argument is offset into first mbuf of target chain from
1156  * which to begin copying the data to.
1157  */
1158 struct mbuf *
1159 m_devget(char *buf, int totlen, int off, struct ifnet *ifp,
1160     void (*copy)(char *from, caddr_t to, u_int len))
1161 {
1162 	struct mbuf *m;
1163 	struct mbuf *top = NULL, **mp = &top;
1164 	int len;
1165 
1166 	if (off < 0 || off > MHLEN)
1167 		return (NULL);
1168 
1169 	while (totlen > 0) {
1170 		if (top == NULL) {	/* First one, must be PKTHDR */
1171 			if (totlen + off >= MINCLSIZE) {
1172 				m = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR);
1173 				len = MCLBYTES;
1174 			} else {
1175 				m = m_gethdr(M_NOWAIT, MT_DATA);
1176 				len = MHLEN;
1177 
1178 				/* Place initial small packet/header at end of mbuf */
1179 				if (m && totlen + off + max_linkhdr <= MLEN) {
1180 					m->m_data += max_linkhdr;
1181 					len -= max_linkhdr;
1182 				}
1183 			}
1184 			if (m == NULL)
1185 				return NULL;
1186 			m->m_pkthdr.rcvif = ifp;
1187 			m->m_pkthdr.len = totlen;
1188 		} else {
1189 			if (totlen + off >= MINCLSIZE) {
1190 				m = m_getcl(M_NOWAIT, MT_DATA, 0);
1191 				len = MCLBYTES;
1192 			} else {
1193 				m = m_get(M_NOWAIT, MT_DATA);
1194 				len = MLEN;
1195 			}
1196 			if (m == NULL) {
1197 				m_freem(top);
1198 				return NULL;
1199 			}
1200 		}
1201 		if (off) {
1202 			m->m_data += off;
1203 			len -= off;
1204 			off = 0;
1205 		}
1206 		m->m_len = len = min(totlen, len);
1207 		if (copy)
1208 			copy(buf, mtod(m, caddr_t), (u_int)len);
1209 		else
1210 			bcopy(buf, mtod(m, caddr_t), (u_int)len);
1211 		buf += len;
1212 		*mp = m;
1213 		mp = &m->m_next;
1214 		totlen -= len;
1215 	}
1216 	return (top);
1217 }
1218 
1219 /*
1220  * Copy data from a buffer back into the indicated mbuf chain,
1221  * starting "off" bytes from the beginning, extending the mbuf
1222  * chain if necessary.
1223  */
1224 void
1225 m_copyback(struct mbuf *m0, int off, int len, c_caddr_t cp)
1226 {
1227 	int mlen;
1228 	struct mbuf *m = m0, *n;
1229 	int totlen = 0;
1230 
1231 	if (m0 == NULL)
1232 		return;
1233 	while (off > (mlen = m->m_len)) {
1234 		off -= mlen;
1235 		totlen += mlen;
1236 		if (m->m_next == NULL) {
1237 			n = m_get(M_NOWAIT, m->m_type);
1238 			if (n == NULL)
1239 				goto out;
1240 			bzero(mtod(n, caddr_t), MLEN);
1241 			n->m_len = min(MLEN, len + off);
1242 			m->m_next = n;
1243 		}
1244 		m = m->m_next;
1245 	}
1246 	while (len > 0) {
1247 		if (m->m_next == NULL && (len > m->m_len - off)) {
1248 			m->m_len += min(len - (m->m_len - off),
1249 			    M_TRAILINGSPACE(m));
1250 		}
1251 		mlen = min (m->m_len - off, len);
1252 		bcopy(cp, off + mtod(m, caddr_t), (u_int)mlen);
1253 		cp += mlen;
1254 		len -= mlen;
1255 		mlen += off;
1256 		off = 0;
1257 		totlen += mlen;
1258 		if (len == 0)
1259 			break;
1260 		if (m->m_next == NULL) {
1261 			n = m_get(M_NOWAIT, m->m_type);
1262 			if (n == NULL)
1263 				break;
1264 			n->m_len = min(MLEN, len);
1265 			m->m_next = n;
1266 		}
1267 		m = m->m_next;
1268 	}
1269 out:	if (((m = m0)->m_flags & M_PKTHDR) && (m->m_pkthdr.len < totlen))
1270 		m->m_pkthdr.len = totlen;
1271 }
1272 
1273 /*
1274  * Append the specified data to the indicated mbuf chain,
1275  * Extend the mbuf chain if the new data does not fit in
1276  * existing space.
1277  *
1278  * Return 1 if able to complete the job; otherwise 0.
1279  */
1280 int
1281 m_append(struct mbuf *m0, int len, c_caddr_t cp)
1282 {
1283 	struct mbuf *m, *n;
1284 	int remainder, space;
1285 
1286 	for (m = m0; m->m_next != NULL; m = m->m_next)
1287 		;
1288 	remainder = len;
1289 	space = M_TRAILINGSPACE(m);
1290 	if (space > 0) {
1291 		/*
1292 		 * Copy into available space.
1293 		 */
1294 		if (space > remainder)
1295 			space = remainder;
1296 		bcopy(cp, mtod(m, caddr_t) + m->m_len, space);
1297 		m->m_len += space;
1298 		cp += space, remainder -= space;
1299 	}
1300 	while (remainder > 0) {
1301 		/*
1302 		 * Allocate a new mbuf; could check space
1303 		 * and allocate a cluster instead.
1304 		 */
1305 		n = m_get(M_NOWAIT, m->m_type);
1306 		if (n == NULL)
1307 			break;
1308 		n->m_len = min(MLEN, remainder);
1309 		bcopy(cp, mtod(n, caddr_t), n->m_len);
1310 		cp += n->m_len, remainder -= n->m_len;
1311 		m->m_next = n;
1312 		m = n;
1313 	}
1314 	if (m0->m_flags & M_PKTHDR)
1315 		m0->m_pkthdr.len += len - remainder;
1316 	return (remainder == 0);
1317 }
1318 
1319 /*
1320  * Apply function f to the data in an mbuf chain starting "off" bytes from
1321  * the beginning, continuing for "len" bytes.
1322  */
1323 int
1324 m_apply(struct mbuf *m, int off, int len,
1325     int (*f)(void *, void *, u_int), void *arg)
1326 {
1327 	u_int count;
1328 	int rval;
1329 
1330 	KASSERT(off >= 0, ("m_apply, negative off %d", off));
1331 	KASSERT(len >= 0, ("m_apply, negative len %d", len));
1332 	while (off > 0) {
1333 		KASSERT(m != NULL, ("m_apply, offset > size of mbuf chain"));
1334 		if (off < m->m_len)
1335 			break;
1336 		off -= m->m_len;
1337 		m = m->m_next;
1338 	}
1339 	while (len > 0) {
1340 		KASSERT(m != NULL, ("m_apply, offset > size of mbuf chain"));
1341 		count = min(m->m_len - off, len);
1342 		rval = (*f)(arg, mtod(m, caddr_t) + off, count);
1343 		if (rval)
1344 			return (rval);
1345 		len -= count;
1346 		off = 0;
1347 		m = m->m_next;
1348 	}
1349 	return (0);
1350 }
1351 
1352 /*
1353  * Return a pointer to mbuf/offset of location in mbuf chain.
1354  */
1355 struct mbuf *
1356 m_getptr(struct mbuf *m, int loc, int *off)
1357 {
1358 
1359 	while (loc >= 0) {
1360 		/* Normal end of search. */
1361 		if (m->m_len > loc) {
1362 			*off = loc;
1363 			return (m);
1364 		} else {
1365 			loc -= m->m_len;
1366 			if (m->m_next == NULL) {
1367 				if (loc == 0) {
1368 					/* Point at the end of valid data. */
1369 					*off = m->m_len;
1370 					return (m);
1371 				}
1372 				return (NULL);
1373 			}
1374 			m = m->m_next;
1375 		}
1376 	}
1377 	return (NULL);
1378 }
1379 
1380 void
1381 m_print(const struct mbuf *m, int maxlen)
1382 {
1383 	int len;
1384 	int pdata;
1385 	const struct mbuf *m2;
1386 
1387 	if (m == NULL) {
1388 		printf("mbuf: %p\n", m);
1389 		return;
1390 	}
1391 
1392 	if (m->m_flags & M_PKTHDR)
1393 		len = m->m_pkthdr.len;
1394 	else
1395 		len = -1;
1396 	m2 = m;
1397 	while (m2 != NULL && (len == -1 || len)) {
1398 		pdata = m2->m_len;
1399 		if (maxlen != -1 && pdata > maxlen)
1400 			pdata = maxlen;
1401 		printf("mbuf: %p len: %d, next: %p, %b%s", m2, m2->m_len,
1402 		    m2->m_next, m2->m_flags, "\20\20freelist\17skipfw"
1403 		    "\11proto5\10proto4\7proto3\6proto2\5proto1\4rdonly"
1404 		    "\3eor\2pkthdr\1ext", pdata ? "" : "\n");
1405 		if (pdata)
1406 			printf(", %*D\n", pdata, (u_char *)m2->m_data, "-");
1407 		if (len != -1)
1408 			len -= m2->m_len;
1409 		m2 = m2->m_next;
1410 	}
1411 	if (len > 0)
1412 		printf("%d bytes unaccounted for.\n", len);
1413 	return;
1414 }
1415 
1416 u_int
1417 m_fixhdr(struct mbuf *m0)
1418 {
1419 	u_int len;
1420 
1421 	len = m_length(m0, NULL);
1422 	m0->m_pkthdr.len = len;
1423 	return (len);
1424 }
1425 
1426 u_int
1427 m_length(struct mbuf *m0, struct mbuf **last)
1428 {
1429 	struct mbuf *m;
1430 	u_int len;
1431 
1432 	len = 0;
1433 	for (m = m0; m != NULL; m = m->m_next) {
1434 		len += m->m_len;
1435 		if (m->m_next == NULL)
1436 			break;
1437 	}
1438 	if (last != NULL)
1439 		*last = m;
1440 	return (len);
1441 }
1442 
1443 /*
1444  * Defragment a mbuf chain, returning the shortest possible
1445  * chain of mbufs and clusters.  If allocation fails and
1446  * this cannot be completed, NULL will be returned, but
1447  * the passed in chain will be unchanged.  Upon success,
1448  * the original chain will be freed, and the new chain
1449  * will be returned.
1450  *
1451  * If a non-packet header is passed in, the original
1452  * mbuf (chain?) will be returned unharmed.
1453  */
1454 struct mbuf *
1455 m_defrag(struct mbuf *m0, int how)
1456 {
1457 	struct mbuf *m_new = NULL, *m_final = NULL;
1458 	int progress = 0, length;
1459 
1460 	MBUF_CHECKSLEEP(how);
1461 	if (!(m0->m_flags & M_PKTHDR))
1462 		return (m0);
1463 
1464 	m_fixhdr(m0); /* Needed sanity check */
1465 
1466 #ifdef MBUF_STRESS_TEST
1467 	if (m_defragrandomfailures) {
1468 		int temp = arc4random() & 0xff;
1469 		if (temp == 0xba)
1470 			goto nospace;
1471 	}
1472 #endif
1473 
1474 	if (m0->m_pkthdr.len > MHLEN)
1475 		m_final = m_getcl(how, MT_DATA, M_PKTHDR);
1476 	else
1477 		m_final = m_gethdr(how, MT_DATA);
1478 
1479 	if (m_final == NULL)
1480 		goto nospace;
1481 
1482 	if (m_dup_pkthdr(m_final, m0, how) == 0)
1483 		goto nospace;
1484 
1485 	m_new = m_final;
1486 
1487 	while (progress < m0->m_pkthdr.len) {
1488 		length = m0->m_pkthdr.len - progress;
1489 		if (length > MCLBYTES)
1490 			length = MCLBYTES;
1491 
1492 		if (m_new == NULL) {
1493 			if (length > MLEN)
1494 				m_new = m_getcl(how, MT_DATA, 0);
1495 			else
1496 				m_new = m_get(how, MT_DATA);
1497 			if (m_new == NULL)
1498 				goto nospace;
1499 		}
1500 
1501 		m_copydata(m0, progress, length, mtod(m_new, caddr_t));
1502 		progress += length;
1503 		m_new->m_len = length;
1504 		if (m_new != m_final)
1505 			m_cat(m_final, m_new);
1506 		m_new = NULL;
1507 	}
1508 #ifdef MBUF_STRESS_TEST
1509 	if (m0->m_next == NULL)
1510 		m_defraguseless++;
1511 #endif
1512 	m_freem(m0);
1513 	m0 = m_final;
1514 #ifdef MBUF_STRESS_TEST
1515 	m_defragpackets++;
1516 	m_defragbytes += m0->m_pkthdr.len;
1517 #endif
1518 	return (m0);
1519 nospace:
1520 #ifdef MBUF_STRESS_TEST
1521 	m_defragfailure++;
1522 #endif
1523 	if (m_final)
1524 		m_freem(m_final);
1525 	return (NULL);
1526 }
1527 
1528 /*
1529  * Defragment an mbuf chain, returning at most maxfrags separate
1530  * mbufs+clusters.  If this is not possible NULL is returned and
1531  * the original mbuf chain is left in it's present (potentially
1532  * modified) state.  We use two techniques: collapsing consecutive
1533  * mbufs and replacing consecutive mbufs by a cluster.
1534  *
1535  * NB: this should really be named m_defrag but that name is taken
1536  */
1537 struct mbuf *
1538 m_collapse(struct mbuf *m0, int how, int maxfrags)
1539 {
1540 	struct mbuf *m, *n, *n2, **prev;
1541 	u_int curfrags;
1542 
1543 	/*
1544 	 * Calculate the current number of frags.
1545 	 */
1546 	curfrags = 0;
1547 	for (m = m0; m != NULL; m = m->m_next)
1548 		curfrags++;
1549 	/*
1550 	 * First, try to collapse mbufs.  Note that we always collapse
1551 	 * towards the front so we don't need to deal with moving the
1552 	 * pkthdr.  This may be suboptimal if the first mbuf has much
1553 	 * less data than the following.
1554 	 */
1555 	m = m0;
1556 again:
1557 	for (;;) {
1558 		n = m->m_next;
1559 		if (n == NULL)
1560 			break;
1561 		if (M_WRITABLE(m) &&
1562 		    n->m_len < M_TRAILINGSPACE(m)) {
1563 			bcopy(mtod(n, void *), mtod(m, char *) + m->m_len,
1564 				n->m_len);
1565 			m->m_len += n->m_len;
1566 			m->m_next = n->m_next;
1567 			m_free(n);
1568 			if (--curfrags <= maxfrags)
1569 				return m0;
1570 		} else
1571 			m = n;
1572 	}
1573 	KASSERT(maxfrags > 1,
1574 		("maxfrags %u, but normal collapse failed", maxfrags));
1575 	/*
1576 	 * Collapse consecutive mbufs to a cluster.
1577 	 */
1578 	prev = &m0->m_next;		/* NB: not the first mbuf */
1579 	while ((n = *prev) != NULL) {
1580 		if ((n2 = n->m_next) != NULL &&
1581 		    n->m_len + n2->m_len < MCLBYTES) {
1582 			m = m_getcl(how, MT_DATA, 0);
1583 			if (m == NULL)
1584 				goto bad;
1585 			bcopy(mtod(n, void *), mtod(m, void *), n->m_len);
1586 			bcopy(mtod(n2, void *), mtod(m, char *) + n->m_len,
1587 				n2->m_len);
1588 			m->m_len = n->m_len + n2->m_len;
1589 			m->m_next = n2->m_next;
1590 			*prev = m;
1591 			m_free(n);
1592 			m_free(n2);
1593 			if (--curfrags <= maxfrags)	/* +1 cl -2 mbufs */
1594 				return m0;
1595 			/*
1596 			 * Still not there, try the normal collapse
1597 			 * again before we allocate another cluster.
1598 			 */
1599 			goto again;
1600 		}
1601 		prev = &n->m_next;
1602 	}
1603 	/*
1604 	 * No place where we can collapse to a cluster; punt.
1605 	 * This can occur if, for example, you request 2 frags
1606 	 * but the packet requires that both be clusters (we
1607 	 * never reallocate the first mbuf to avoid moving the
1608 	 * packet header).
1609 	 */
1610 bad:
1611 	return NULL;
1612 }
1613 
1614 #ifdef MBUF_STRESS_TEST
1615 
1616 /*
1617  * Fragment an mbuf chain.  There's no reason you'd ever want to do
1618  * this in normal usage, but it's great for stress testing various
1619  * mbuf consumers.
1620  *
1621  * If fragmentation is not possible, the original chain will be
1622  * returned.
1623  *
1624  * Possible length values:
1625  * 0	 no fragmentation will occur
1626  * > 0	each fragment will be of the specified length
1627  * -1	each fragment will be the same random value in length
1628  * -2	each fragment's length will be entirely random
1629  * (Random values range from 1 to 256)
1630  */
1631 struct mbuf *
1632 m_fragment(struct mbuf *m0, int how, int length)
1633 {
1634 	struct mbuf *m_new = NULL, *m_final = NULL;
1635 	int progress = 0;
1636 
1637 	if (!(m0->m_flags & M_PKTHDR))
1638 		return (m0);
1639 
1640 	if ((length == 0) || (length < -2))
1641 		return (m0);
1642 
1643 	m_fixhdr(m0); /* Needed sanity check */
1644 
1645 	m_final = m_getcl(how, MT_DATA, M_PKTHDR);
1646 
1647 	if (m_final == NULL)
1648 		goto nospace;
1649 
1650 	if (m_dup_pkthdr(m_final, m0, how) == 0)
1651 		goto nospace;
1652 
1653 	m_new = m_final;
1654 
1655 	if (length == -1)
1656 		length = 1 + (arc4random() & 255);
1657 
1658 	while (progress < m0->m_pkthdr.len) {
1659 		int fraglen;
1660 
1661 		if (length > 0)
1662 			fraglen = length;
1663 		else
1664 			fraglen = 1 + (arc4random() & 255);
1665 		if (fraglen > m0->m_pkthdr.len - progress)
1666 			fraglen = m0->m_pkthdr.len - progress;
1667 
1668 		if (fraglen > MCLBYTES)
1669 			fraglen = MCLBYTES;
1670 
1671 		if (m_new == NULL) {
1672 			m_new = m_getcl(how, MT_DATA, 0);
1673 			if (m_new == NULL)
1674 				goto nospace;
1675 		}
1676 
1677 		m_copydata(m0, progress, fraglen, mtod(m_new, caddr_t));
1678 		progress += fraglen;
1679 		m_new->m_len = fraglen;
1680 		if (m_new != m_final)
1681 			m_cat(m_final, m_new);
1682 		m_new = NULL;
1683 	}
1684 	m_freem(m0);
1685 	m0 = m_final;
1686 	return (m0);
1687 nospace:
1688 	if (m_final)
1689 		m_freem(m_final);
1690 	/* Return the original chain on failure */
1691 	return (m0);
1692 }
1693 
1694 #endif
1695 
1696 /*
1697  * Copy the contents of uio into a properly sized mbuf chain.
1698  */
1699 struct mbuf *
1700 m_uiotombuf(struct uio *uio, int how, int len, int align, int flags)
1701 {
1702 	struct mbuf *m, *mb;
1703 	int error, length;
1704 	ssize_t total;
1705 	int progress = 0;
1706 
1707 	/*
1708 	 * len can be zero or an arbitrary large value bound by
1709 	 * the total data supplied by the uio.
1710 	 */
1711 	if (len > 0)
1712 		total = min(uio->uio_resid, len);
1713 	else
1714 		total = uio->uio_resid;
1715 
1716 	/*
1717 	 * The smallest unit returned by m_getm2() is a single mbuf
1718 	 * with pkthdr.  We can't align past it.
1719 	 */
1720 	if (align >= MHLEN)
1721 		return (NULL);
1722 
1723 	/*
1724 	 * Give us the full allocation or nothing.
1725 	 * If len is zero return the smallest empty mbuf.
1726 	 */
1727 	m = m_getm2(NULL, max(total + align, 1), how, MT_DATA, flags);
1728 	if (m == NULL)
1729 		return (NULL);
1730 	m->m_data += align;
1731 
1732 	/* Fill all mbufs with uio data and update header information. */
1733 	for (mb = m; mb != NULL; mb = mb->m_next) {
1734 		length = min(M_TRAILINGSPACE(mb), total - progress);
1735 
1736 		error = uiomove(mtod(mb, void *), length, uio);
1737 		if (error) {
1738 			m_freem(m);
1739 			return (NULL);
1740 		}
1741 
1742 		mb->m_len = length;
1743 		progress += length;
1744 		if (flags & M_PKTHDR)
1745 			m->m_pkthdr.len += length;
1746 	}
1747 	KASSERT(progress == total, ("%s: progress != total", __func__));
1748 
1749 	return (m);
1750 }
1751 
1752 /*
1753  * Copy an mbuf chain into a uio limited by len if set.
1754  */
1755 int
1756 m_mbuftouio(struct uio *uio, struct mbuf *m, int len)
1757 {
1758 	int error, length, total;
1759 	int progress = 0;
1760 
1761 	if (len > 0)
1762 		total = min(uio->uio_resid, len);
1763 	else
1764 		total = uio->uio_resid;
1765 
1766 	/* Fill the uio with data from the mbufs. */
1767 	for (; m != NULL; m = m->m_next) {
1768 		length = min(m->m_len, total - progress);
1769 
1770 		error = uiomove(mtod(m, void *), length, uio);
1771 		if (error)
1772 			return (error);
1773 
1774 		progress += length;
1775 	}
1776 
1777 	return (0);
1778 }
1779 
1780 /*
1781  * Create a writable copy of the mbuf chain.  While doing this
1782  * we compact the chain with a goal of producing a chain with
1783  * at most two mbufs.  The second mbuf in this chain is likely
1784  * to be a cluster.  The primary purpose of this work is to create
1785  * a writable packet for encryption, compression, etc.  The
1786  * secondary goal is to linearize the data so the data can be
1787  * passed to crypto hardware in the most efficient manner possible.
1788  */
1789 struct mbuf *
1790 m_unshare(struct mbuf *m0, int how)
1791 {
1792 	struct mbuf *m, *mprev;
1793 	struct mbuf *n, *mfirst, *mlast;
1794 	int len, off;
1795 
1796 	mprev = NULL;
1797 	for (m = m0; m != NULL; m = mprev->m_next) {
1798 		/*
1799 		 * Regular mbufs are ignored unless there's a cluster
1800 		 * in front of it that we can use to coalesce.  We do
1801 		 * the latter mainly so later clusters can be coalesced
1802 		 * also w/o having to handle them specially (i.e. convert
1803 		 * mbuf+cluster -> cluster).  This optimization is heavily
1804 		 * influenced by the assumption that we're running over
1805 		 * Ethernet where MCLBYTES is large enough that the max
1806 		 * packet size will permit lots of coalescing into a
1807 		 * single cluster.  This in turn permits efficient
1808 		 * crypto operations, especially when using hardware.
1809 		 */
1810 		if ((m->m_flags & M_EXT) == 0) {
1811 			if (mprev && (mprev->m_flags & M_EXT) &&
1812 			    m->m_len <= M_TRAILINGSPACE(mprev)) {
1813 				/* XXX: this ignores mbuf types */
1814 				memcpy(mtod(mprev, caddr_t) + mprev->m_len,
1815 				    mtod(m, caddr_t), m->m_len);
1816 				mprev->m_len += m->m_len;
1817 				mprev->m_next = m->m_next;	/* unlink from chain */
1818 				m_free(m);			/* reclaim mbuf */
1819 #if 0
1820 				newipsecstat.ips_mbcoalesced++;
1821 #endif
1822 			} else {
1823 				mprev = m;
1824 			}
1825 			continue;
1826 		}
1827 		/*
1828 		 * Writable mbufs are left alone (for now).
1829 		 */
1830 		if (M_WRITABLE(m)) {
1831 			mprev = m;
1832 			continue;
1833 		}
1834 
1835 		/*
1836 		 * Not writable, replace with a copy or coalesce with
1837 		 * the previous mbuf if possible (since we have to copy
1838 		 * it anyway, we try to reduce the number of mbufs and
1839 		 * clusters so that future work is easier).
1840 		 */
1841 		KASSERT(m->m_flags & M_EXT, ("m_flags 0x%x", m->m_flags));
1842 		/* NB: we only coalesce into a cluster or larger */
1843 		if (mprev != NULL && (mprev->m_flags & M_EXT) &&
1844 		    m->m_len <= M_TRAILINGSPACE(mprev)) {
1845 			/* XXX: this ignores mbuf types */
1846 			memcpy(mtod(mprev, caddr_t) + mprev->m_len,
1847 			    mtod(m, caddr_t), m->m_len);
1848 			mprev->m_len += m->m_len;
1849 			mprev->m_next = m->m_next;	/* unlink from chain */
1850 			m_free(m);			/* reclaim mbuf */
1851 #if 0
1852 			newipsecstat.ips_clcoalesced++;
1853 #endif
1854 			continue;
1855 		}
1856 
1857 		/*
1858 		 * Allocate new space to hold the copy and copy the data.
1859 		 * We deal with jumbo mbufs (i.e. m_len > MCLBYTES) by
1860 		 * splitting them into clusters.  We could just malloc a
1861 		 * buffer and make it external but too many device drivers
1862 		 * don't know how to break up the non-contiguous memory when
1863 		 * doing DMA.
1864 		 */
1865 		n = m_getcl(how, m->m_type, m->m_flags);
1866 		if (n == NULL) {
1867 			m_freem(m0);
1868 			return (NULL);
1869 		}
1870 		if (m->m_flags & M_PKTHDR) {
1871 			KASSERT(mprev == NULL, ("%s: m0 %p, m %p has M_PKTHDR",
1872 			    __func__, m0, m));
1873 			m_move_pkthdr(n, m);
1874 		}
1875 		len = m->m_len;
1876 		off = 0;
1877 		mfirst = n;
1878 		mlast = NULL;
1879 		for (;;) {
1880 			int cc = min(len, MCLBYTES);
1881 			memcpy(mtod(n, caddr_t), mtod(m, caddr_t) + off, cc);
1882 			n->m_len = cc;
1883 			if (mlast != NULL)
1884 				mlast->m_next = n;
1885 			mlast = n;
1886 #if 0
1887 			newipsecstat.ips_clcopied++;
1888 #endif
1889 
1890 			len -= cc;
1891 			if (len <= 0)
1892 				break;
1893 			off += cc;
1894 
1895 			n = m_getcl(how, m->m_type, m->m_flags);
1896 			if (n == NULL) {
1897 				m_freem(mfirst);
1898 				m_freem(m0);
1899 				return (NULL);
1900 			}
1901 		}
1902 		n->m_next = m->m_next;
1903 		if (mprev == NULL)
1904 			m0 = mfirst;		/* new head of chain */
1905 		else
1906 			mprev->m_next = mfirst;	/* replace old mbuf */
1907 		m_free(m);			/* release old mbuf */
1908 		mprev = mfirst;
1909 	}
1910 	return (m0);
1911 }
1912 
1913 #ifdef MBUF_PROFILING
1914 
1915 #define MP_BUCKETS 32 /* don't just change this as things may overflow.*/
1916 struct mbufprofile {
1917 	uintmax_t wasted[MP_BUCKETS];
1918 	uintmax_t used[MP_BUCKETS];
1919 	uintmax_t segments[MP_BUCKETS];
1920 } mbprof;
1921 
1922 #define MP_MAXDIGITS 21	/* strlen("16,000,000,000,000,000,000") == 21 */
1923 #define MP_NUMLINES 6
1924 #define MP_NUMSPERLINE 16
1925 #define MP_EXTRABYTES 64	/* > strlen("used:\nwasted:\nsegments:\n") */
1926 /* work out max space needed and add a bit of spare space too */
1927 #define MP_MAXLINE ((MP_MAXDIGITS+1) * MP_NUMSPERLINE)
1928 #define MP_BUFSIZE ((MP_MAXLINE * MP_NUMLINES) + 1 + MP_EXTRABYTES)
1929 
1930 char mbprofbuf[MP_BUFSIZE];
1931 
1932 void
1933 m_profile(struct mbuf *m)
1934 {
1935 	int segments = 0;
1936 	int used = 0;
1937 	int wasted = 0;
1938 
1939 	while (m) {
1940 		segments++;
1941 		used += m->m_len;
1942 		if (m->m_flags & M_EXT) {
1943 			wasted += MHLEN - sizeof(m->m_ext) +
1944 			    m->m_ext.ext_size - m->m_len;
1945 		} else {
1946 			if (m->m_flags & M_PKTHDR)
1947 				wasted += MHLEN - m->m_len;
1948 			else
1949 				wasted += MLEN - m->m_len;
1950 		}
1951 		m = m->m_next;
1952 	}
1953 	/* be paranoid.. it helps */
1954 	if (segments > MP_BUCKETS - 1)
1955 		segments = MP_BUCKETS - 1;
1956 	if (used > 100000)
1957 		used = 100000;
1958 	if (wasted > 100000)
1959 		wasted = 100000;
1960 	/* store in the appropriate bucket */
1961 	/* don't bother locking. if it's slightly off, so what? */
1962 	mbprof.segments[segments]++;
1963 	mbprof.used[fls(used)]++;
1964 	mbprof.wasted[fls(wasted)]++;
1965 }
1966 
1967 static void
1968 mbprof_textify(void)
1969 {
1970 	int offset;
1971 	char *c;
1972 	uint64_t *p;
1973 
1974 	p = &mbprof.wasted[0];
1975 	c = mbprofbuf;
1976 	offset = snprintf(c, MP_MAXLINE + 10,
1977 	    "wasted:\n"
1978 	    "%ju %ju %ju %ju %ju %ju %ju %ju "
1979 	    "%ju %ju %ju %ju %ju %ju %ju %ju\n",
1980 	    p[0], p[1], p[2], p[3], p[4], p[5], p[6], p[7],
1981 	    p[8], p[9], p[10], p[11], p[12], p[13], p[14], p[15]);
1982 #ifdef BIG_ARRAY
1983 	p = &mbprof.wasted[16];
1984 	c += offset;
1985 	offset = snprintf(c, MP_MAXLINE,
1986 	    "%ju %ju %ju %ju %ju %ju %ju %ju "
1987 	    "%ju %ju %ju %ju %ju %ju %ju %ju\n",
1988 	    p[0], p[1], p[2], p[3], p[4], p[5], p[6], p[7],
1989 	    p[8], p[9], p[10], p[11], p[12], p[13], p[14], p[15]);
1990 #endif
1991 	p = &mbprof.used[0];
1992 	c += offset;
1993 	offset = snprintf(c, MP_MAXLINE + 10,
1994 	    "used:\n"
1995 	    "%ju %ju %ju %ju %ju %ju %ju %ju "
1996 	    "%ju %ju %ju %ju %ju %ju %ju %ju\n",
1997 	    p[0], p[1], p[2], p[3], p[4], p[5], p[6], p[7],
1998 	    p[8], p[9], p[10], p[11], p[12], p[13], p[14], p[15]);
1999 #ifdef BIG_ARRAY
2000 	p = &mbprof.used[16];
2001 	c += offset;
2002 	offset = snprintf(c, MP_MAXLINE,
2003 	    "%ju %ju %ju %ju %ju %ju %ju %ju "
2004 	    "%ju %ju %ju %ju %ju %ju %ju %ju\n",
2005 	    p[0], p[1], p[2], p[3], p[4], p[5], p[6], p[7],
2006 	    p[8], p[9], p[10], p[11], p[12], p[13], p[14], p[15]);
2007 #endif
2008 	p = &mbprof.segments[0];
2009 	c += offset;
2010 	offset = snprintf(c, MP_MAXLINE + 10,
2011 	    "segments:\n"
2012 	    "%ju %ju %ju %ju %ju %ju %ju %ju "
2013 	    "%ju %ju %ju %ju %ju %ju %ju %ju\n",
2014 	    p[0], p[1], p[2], p[3], p[4], p[5], p[6], p[7],
2015 	    p[8], p[9], p[10], p[11], p[12], p[13], p[14], p[15]);
2016 #ifdef BIG_ARRAY
2017 	p = &mbprof.segments[16];
2018 	c += offset;
2019 	offset = snprintf(c, MP_MAXLINE,
2020 	    "%ju %ju %ju %ju %ju %ju %ju %ju "
2021 	    "%ju %ju %ju %ju %ju %ju %ju %jju",
2022 	    p[0], p[1], p[2], p[3], p[4], p[5], p[6], p[7],
2023 	    p[8], p[9], p[10], p[11], p[12], p[13], p[14], p[15]);
2024 #endif
2025 }
2026 
2027 static int
2028 mbprof_handler(SYSCTL_HANDLER_ARGS)
2029 {
2030 	int error;
2031 
2032 	mbprof_textify();
2033 	error = SYSCTL_OUT(req, mbprofbuf, strlen(mbprofbuf) + 1);
2034 	return (error);
2035 }
2036 
2037 static int
2038 mbprof_clr_handler(SYSCTL_HANDLER_ARGS)
2039 {
2040 	int clear, error;
2041 
2042 	clear = 0;
2043 	error = sysctl_handle_int(oidp, &clear, 0, req);
2044 	if (error || !req->newptr)
2045 		return (error);
2046 
2047 	if (clear) {
2048 		bzero(&mbprof, sizeof(mbprof));
2049 	}
2050 
2051 	return (error);
2052 }
2053 
2054 
2055 SYSCTL_PROC(_kern_ipc, OID_AUTO, mbufprofile, CTLTYPE_STRING|CTLFLAG_RD,
2056 	    NULL, 0, mbprof_handler, "A", "mbuf profiling statistics");
2057 
2058 SYSCTL_PROC(_kern_ipc, OID_AUTO, mbufprofileclr, CTLTYPE_INT|CTLFLAG_RW,
2059 	    NULL, 0, mbprof_clr_handler, "I", "clear mbuf profiling statistics");
2060 #endif
2061 
2062