xref: /netbsd-src/sys/kern/uipc_mbuf.c (revision ce2c90c7c172d95d2402a5b3d96d8f8e6d138a21)
1 /*	$NetBSD: uipc_mbuf.c,v 1.115 2006/10/12 01:32:19 christos Exp $	*/
2 
3 /*-
4  * Copyright (c) 1999, 2001 The NetBSD Foundation, Inc.
5  * All rights reserved.
6  *
7  * This code is derived from software contributed to The NetBSD Foundation
8  * by Jason R. Thorpe of the Numerical Aerospace Simulation Facility,
9  * NASA Ames Research Center.
10  *
11  * Redistribution and use in source and binary forms, with or without
12  * modification, are permitted provided that the following conditions
13  * are met:
14  * 1. Redistributions of source code must retain the above copyright
15  *    notice, this list of conditions and the following disclaimer.
16  * 2. Redistributions in binary form must reproduce the above copyright
17  *    notice, this list of conditions and the following disclaimer in the
18  *    documentation and/or other materials provided with the distribution.
19  * 3. All advertising materials mentioning features or use of this software
20  *    must display the following acknowledgement:
21  *	This product includes software developed by the NetBSD
22  *	Foundation, Inc. and its contributors.
23  * 4. Neither the name of The NetBSD Foundation nor the names of its
24  *    contributors may be used to endorse or promote products derived
25  *    from this software without specific prior written permission.
26  *
27  * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
28  * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
29  * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
30  * PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
31  * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
32  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
33  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
34  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
35  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
36  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
37  * POSSIBILITY OF SUCH DAMAGE.
38  */
39 
40 /*
41  * Copyright (c) 1982, 1986, 1988, 1991, 1993
42  *	The Regents of the University of California.  All rights reserved.
43  *
44  * Redistribution and use in source and binary forms, with or without
45  * modification, are permitted provided that the following conditions
46  * are met:
47  * 1. Redistributions of source code must retain the above copyright
48  *    notice, this list of conditions and the following disclaimer.
49  * 2. Redistributions in binary form must reproduce the above copyright
50  *    notice, this list of conditions and the following disclaimer in the
51  *    documentation and/or other materials provided with the distribution.
52  * 3. Neither the name of the University nor the names of its contributors
53  *    may be used to endorse or promote products derived from this software
54  *    without specific prior written permission.
55  *
56  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
57  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
58  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
59  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
60  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
61  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
62  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
63  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
64  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
65  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
66  * SUCH DAMAGE.
67  *
68  *	@(#)uipc_mbuf.c	8.4 (Berkeley) 2/14/95
69  */
70 
71 #include <sys/cdefs.h>
72 __KERNEL_RCSID(0, "$NetBSD: uipc_mbuf.c,v 1.115 2006/10/12 01:32:19 christos Exp $");
73 
74 #include "opt_mbuftrace.h"
75 #include "opt_ddb.h"
76 
77 #include <sys/param.h>
78 #include <sys/systm.h>
79 #include <sys/proc.h>
80 #include <sys/malloc.h>
81 #define MBTYPES
82 #include <sys/mbuf.h>
83 #include <sys/kernel.h>
84 #include <sys/syslog.h>
85 #include <sys/domain.h>
86 #include <sys/protosw.h>
87 #include <sys/pool.h>
88 #include <sys/socket.h>
89 #include <sys/sysctl.h>
90 
91 #include <net/if.h>
92 
93 #include <uvm/uvm.h>
94 
95 
96 struct	pool mbpool;		/* mbuf pool */
97 struct	pool mclpool;		/* mbuf cluster pool */
98 
99 struct pool_cache mbpool_cache;
100 struct pool_cache mclpool_cache;
101 
102 struct mbstat mbstat;
103 int	max_linkhdr;
104 int	max_protohdr;
105 int	max_hdr;
106 int	max_datalen;
107 
108 static int mb_ctor(void *, void *, int);
109 
110 static void	*mclpool_alloc(struct pool *, int);
111 static void	mclpool_release(struct pool *, void *);
112 
113 static struct pool_allocator mclpool_allocator = {
114 	.pa_alloc = mclpool_alloc,
115 	.pa_free = mclpool_release,
116 };
117 
118 static struct mbuf *m_copym0(struct mbuf *, int, int, int, int);
119 static struct mbuf *m_split0(struct mbuf *, int, int, int);
120 static int m_copyback0(struct mbuf **, int, int, const void *, int, int);
121 
122 /* flags for m_copyback0 */
123 #define	M_COPYBACK0_COPYBACK	0x0001	/* copyback from cp */
124 #define	M_COPYBACK0_PRESERVE	0x0002	/* preserve original data */
125 #define	M_COPYBACK0_COW		0x0004	/* do copy-on-write */
126 #define	M_COPYBACK0_EXTEND	0x0008	/* extend chain */
127 
128 static const char mclpool_warnmsg[] =
129     "WARNING: mclpool limit reached; increase NMBCLUSTERS";
130 
131 MALLOC_DEFINE(M_MBUF, "mbuf", "mbuf");
132 
133 #ifdef MBUFTRACE
134 struct mownerhead mowners = LIST_HEAD_INITIALIZER(mowners);
135 struct mowner unknown_mowners[] = {
136 	MOWNER_INIT("unknown", "free"),
137 	MOWNER_INIT("unknown", "data"),
138 	MOWNER_INIT("unknown", "header"),
139 	MOWNER_INIT("unknown", "soname"),
140 	MOWNER_INIT("unknown", "soopts"),
141 	MOWNER_INIT("unknown", "ftable"),
142 	MOWNER_INIT("unknown", "control"),
143 	MOWNER_INIT("unknown", "oobdata"),
144 };
145 struct mowner revoked_mowner = MOWNER_INIT("revoked", "");
146 #endif
147 
148 /*
149  * Initialize the mbuf allocator.
150  */
151 void
152 mbinit(void)
153 {
154 
155 	KASSERT(sizeof(struct _m_ext) <= MHLEN);
156 	KASSERT(sizeof(struct mbuf) == MSIZE);
157 
158 	mclpool_allocator.pa_backingmap = mb_map;
159 	pool_init(&mbpool, msize, 0, 0, 0, "mbpl", NULL);
160 	pool_init(&mclpool, mclbytes, 0, 0, 0, "mclpl", &mclpool_allocator);
161 
162 	pool_set_drain_hook(&mbpool, m_reclaim, NULL);
163 	pool_set_drain_hook(&mclpool, m_reclaim, NULL);
164 
165 	pool_cache_init(&mbpool_cache, &mbpool, mb_ctor, NULL, NULL);
166 	pool_cache_init(&mclpool_cache, &mclpool, NULL, NULL, NULL);
167 
168 	/*
169 	 * Set the hard limit on the mclpool to the number of
170 	 * mbuf clusters the kernel is to support.  Log the limit
171 	 * reached message max once a minute.
172 	 */
173 	pool_sethardlimit(&mclpool, nmbclusters, mclpool_warnmsg, 60);
174 
175 	/*
176 	 * Set a low water mark for both mbufs and clusters.  This should
177 	 * help ensure that they can be allocated in a memory starvation
178 	 * situation.  This is important for e.g. diskless systems which
179 	 * must allocate mbufs in order for the pagedaemon to clean pages.
180 	 */
181 	pool_setlowat(&mbpool, mblowat);
182 	pool_setlowat(&mclpool, mcllowat);
183 
184 #ifdef MBUFTRACE
185 	{
186 		/*
187 		 * Attach the unknown mowners.
188 		 */
189 		int i;
190 		MOWNER_ATTACH(&revoked_mowner);
191 		for (i = sizeof(unknown_mowners)/sizeof(unknown_mowners[0]);
192 		     i-- > 0; )
193 			MOWNER_ATTACH(&unknown_mowners[i]);
194 	}
195 #endif
196 }
197 
198 /*
199  * sysctl helper routine for the kern.mbuf subtree.  nmbclusters may
200  * or may not be writable, and mblowat and mcllowat need range
201  * checking and pool tweaking after being reset.
202  */
203 static int
204 sysctl_kern_mbuf(SYSCTLFN_ARGS)
205 {
206 	int error, newval;
207 	struct sysctlnode node;
208 
209 	node = *rnode;
210 	node.sysctl_data = &newval;
211 	switch (rnode->sysctl_num) {
212 	case MBUF_NMBCLUSTERS:
213 		if (mb_map != NULL) {
214 			node.sysctl_flags &= ~CTLFLAG_READWRITE;
215 			node.sysctl_flags |= CTLFLAG_READONLY;
216 		}
217 		/* FALLTHROUGH */
218 	case MBUF_MBLOWAT:
219 	case MBUF_MCLLOWAT:
220 		newval = *(int*)rnode->sysctl_data;
221 		break;
222 	default:
223 		return (EOPNOTSUPP);
224 	}
225 
226 	error = sysctl_lookup(SYSCTLFN_CALL(&node));
227 	if (error || newp == NULL)
228 		return (error);
229 	if (newval < 0)
230 		return (EINVAL);
231 
232 	switch (node.sysctl_num) {
233 	case MBUF_NMBCLUSTERS:
234 		if (newval < nmbclusters)
235 			return (EINVAL);
236 		nmbclusters = newval;
237 		pool_sethardlimit(&mclpool, nmbclusters, mclpool_warnmsg, 60);
238 		break;
239 	case MBUF_MBLOWAT:
240 		mblowat = newval;
241 		pool_setlowat(&mbpool, mblowat);
242 		break;
243 	case MBUF_MCLLOWAT:
244 		mcllowat = newval;
245 		pool_setlowat(&mclpool, mcllowat);
246 		break;
247 	}
248 
249 	return (0);
250 }
251 
252 #ifdef MBUFTRACE
253 static int
254 sysctl_kern_mbuf_mowners(SYSCTLFN_ARGS)
255 {
256 	struct mowner *mo;
257 	size_t len = 0;
258 	int error = 0;
259 
260 	if (namelen != 0)
261 		return (EINVAL);
262 	if (newp != NULL)
263 		return (EPERM);
264 
265 	LIST_FOREACH(mo, &mowners, mo_link) {
266 		if (oldp != NULL) {
267 			if (*oldlenp - len < sizeof(*mo)) {
268 				error = ENOMEM;
269 				break;
270 			}
271 			error = copyout(mo, (caddr_t) oldp + len,
272 					sizeof(*mo));
273 			if (error)
274 				break;
275 		}
276 		len += sizeof(*mo);
277 	}
278 
279 	if (error == 0)
280 		*oldlenp = len;
281 
282 	return (error);
283 }
284 #endif /* MBUFTRACE */
285 
286 SYSCTL_SETUP(sysctl_kern_mbuf_setup, "sysctl kern.mbuf subtree setup")
287 {
288 
289 	sysctl_createv(clog, 0, NULL, NULL,
290 		       CTLFLAG_PERMANENT,
291 		       CTLTYPE_NODE, "kern", NULL,
292 		       NULL, 0, NULL, 0,
293 		       CTL_KERN, CTL_EOL);
294 	sysctl_createv(clog, 0, NULL, NULL,
295 		       CTLFLAG_PERMANENT,
296 		       CTLTYPE_NODE, "mbuf",
297 		       SYSCTL_DESCR("mbuf control variables"),
298 		       NULL, 0, NULL, 0,
299 		       CTL_KERN, KERN_MBUF, CTL_EOL);
300 
301 	sysctl_createv(clog, 0, NULL, NULL,
302 		       CTLFLAG_PERMANENT|CTLFLAG_IMMEDIATE,
303 		       CTLTYPE_INT, "msize",
304 		       SYSCTL_DESCR("mbuf base size"),
305 		       NULL, msize, NULL, 0,
306 		       CTL_KERN, KERN_MBUF, MBUF_MSIZE, CTL_EOL);
307 	sysctl_createv(clog, 0, NULL, NULL,
308 		       CTLFLAG_PERMANENT|CTLFLAG_IMMEDIATE,
309 		       CTLTYPE_INT, "mclbytes",
310 		       SYSCTL_DESCR("mbuf cluster size"),
311 		       NULL, mclbytes, NULL, 0,
312 		       CTL_KERN, KERN_MBUF, MBUF_MCLBYTES, CTL_EOL);
313 	sysctl_createv(clog, 0, NULL, NULL,
314 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
315 		       CTLTYPE_INT, "nmbclusters",
316 		       SYSCTL_DESCR("Limit on the number of mbuf clusters"),
317 		       sysctl_kern_mbuf, 0, &nmbclusters, 0,
318 		       CTL_KERN, KERN_MBUF, MBUF_NMBCLUSTERS, CTL_EOL);
319 	sysctl_createv(clog, 0, NULL, NULL,
320 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
321 		       CTLTYPE_INT, "mblowat",
322 		       SYSCTL_DESCR("mbuf low water mark"),
323 		       sysctl_kern_mbuf, 0, &mblowat, 0,
324 		       CTL_KERN, KERN_MBUF, MBUF_MBLOWAT, CTL_EOL);
325 	sysctl_createv(clog, 0, NULL, NULL,
326 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
327 		       CTLTYPE_INT, "mcllowat",
328 		       SYSCTL_DESCR("mbuf cluster low water mark"),
329 		       sysctl_kern_mbuf, 0, &mcllowat, 0,
330 		       CTL_KERN, KERN_MBUF, MBUF_MCLLOWAT, CTL_EOL);
331 	sysctl_createv(clog, 0, NULL, NULL,
332 		       CTLFLAG_PERMANENT,
333 		       CTLTYPE_STRUCT, "stats",
334 		       SYSCTL_DESCR("mbuf allocation statistics"),
335 		       NULL, 0, &mbstat, sizeof(mbstat),
336 		       CTL_KERN, KERN_MBUF, MBUF_STATS, CTL_EOL);
337 #ifdef MBUFTRACE
338 	sysctl_createv(clog, 0, NULL, NULL,
339 		       CTLFLAG_PERMANENT,
340 		       CTLTYPE_STRUCT, "mowners",
341 		       SYSCTL_DESCR("Information about mbuf owners"),
342 		       sysctl_kern_mbuf_mowners, 0, NULL, 0,
343 		       CTL_KERN, KERN_MBUF, MBUF_MOWNERS, CTL_EOL);
344 #endif /* MBUFTRACE */
345 }
346 
347 static void *
348 mclpool_alloc(struct pool *pp __unused, int flags)
349 {
350 	boolean_t waitok = (flags & PR_WAITOK) ? TRUE : FALSE;
351 
352 	return ((void *)uvm_km_alloc_poolpage(mb_map, waitok));
353 }
354 
355 static void
356 mclpool_release(struct pool *pp __unused, void *v)
357 {
358 
359 	uvm_km_free_poolpage(mb_map, (vaddr_t)v);
360 }
361 
362 /*ARGSUSED*/
363 static int
364 mb_ctor(void *arg __unused, void *object, int flags __unused)
365 {
366 	struct mbuf *m = object;
367 
368 #ifdef POOL_VTOPHYS
369 	m->m_paddr = POOL_VTOPHYS(m);
370 #else
371 	m->m_paddr = M_PADDR_INVALID;
372 #endif
373 	return (0);
374 }
375 
376 void
377 m_reclaim(void *arg __unused, int flags __unused)
378 {
379 	struct domain *dp;
380 	const struct protosw *pr;
381 	struct ifnet *ifp;
382 	int s = splvm();
383 
384 	DOMAIN_FOREACH(dp) {
385 		for (pr = dp->dom_protosw;
386 		     pr < dp->dom_protoswNPROTOSW; pr++)
387 			if (pr->pr_drain)
388 				(*pr->pr_drain)();
389 	}
390 	IFNET_FOREACH(ifp) {
391 		if (ifp->if_drain)
392 			(*ifp->if_drain)(ifp);
393 	}
394 	splx(s);
395 	mbstat.m_drain++;
396 }
397 
398 /*
399  * Space allocation routines.
400  * These are also available as macros
401  * for critical paths.
402  */
403 struct mbuf *
404 m_get(int nowait, int type)
405 {
406 	struct mbuf *m;
407 
408 	MGET(m, nowait, type);
409 	return (m);
410 }
411 
412 struct mbuf *
413 m_gethdr(int nowait, int type)
414 {
415 	struct mbuf *m;
416 
417 	MGETHDR(m, nowait, type);
418 	return (m);
419 }
420 
421 struct mbuf *
422 m_getclr(int nowait, int type)
423 {
424 	struct mbuf *m;
425 
426 	MGET(m, nowait, type);
427 	if (m == 0)
428 		return (NULL);
429 	memset(mtod(m, caddr_t), 0, MLEN);
430 	return (m);
431 }
432 
433 void
434 m_clget(struct mbuf *m, int nowait)
435 {
436 
437 	MCLGET(m, nowait);
438 }
439 
440 struct mbuf *
441 m_free(struct mbuf *m)
442 {
443 	struct mbuf *n;
444 
445 	MFREE(m, n);
446 	return (n);
447 }
448 
449 void
450 m_freem(struct mbuf *m)
451 {
452 	struct mbuf *n;
453 
454 	if (m == NULL)
455 		return;
456 	do {
457 		MFREE(m, n);
458 		m = n;
459 	} while (m);
460 }
461 
462 #ifdef MBUFTRACE
463 /*
464  * Walk a chain of mbufs, claiming ownership of each mbuf in the chain.
465  */
466 void
467 m_claimm(struct mbuf *m, struct mowner *mo)
468 {
469 
470 	for (; m != NULL; m = m->m_next)
471 		MCLAIM(m, mo);
472 }
473 #endif
474 
475 /*
476  * Mbuffer utility routines.
477  */
478 
479 /*
480  * Lesser-used path for M_PREPEND:
481  * allocate new mbuf to prepend to chain,
482  * copy junk along.
483  */
484 struct mbuf *
485 m_prepend(struct mbuf *m, int len, int how)
486 {
487 	struct mbuf *mn;
488 
489 	MGET(mn, how, m->m_type);
490 	if (mn == (struct mbuf *)NULL) {
491 		m_freem(m);
492 		return ((struct mbuf *)NULL);
493 	}
494 	if (m->m_flags & M_PKTHDR) {
495 		M_MOVE_PKTHDR(mn, m);
496 	} else {
497 		MCLAIM(mn, m->m_owner);
498 	}
499 	mn->m_next = m;
500 	m = mn;
501 	if (len < MHLEN)
502 		MH_ALIGN(m, len);
503 	m->m_len = len;
504 	return (m);
505 }
506 
507 /*
508  * Make a copy of an mbuf chain starting "off0" bytes from the beginning,
509  * continuing for "len" bytes.  If len is M_COPYALL, copy to end of mbuf.
510  * The wait parameter is a choice of M_WAIT/M_DONTWAIT from caller.
511  */
512 int MCFail;
513 
514 struct mbuf *
515 m_copym(struct mbuf *m, int off0, int len, int wait)
516 {
517 
518 	return m_copym0(m, off0, len, wait, 0);	/* shallow copy on M_EXT */
519 }
520 
521 struct mbuf *
522 m_dup(struct mbuf *m, int off0, int len, int wait)
523 {
524 
525 	return m_copym0(m, off0, len, wait, 1);	/* deep copy */
526 }
527 
528 static struct mbuf *
529 m_copym0(struct mbuf *m, int off0, int len, int wait, int deep)
530 {
531 	struct mbuf *n, **np;
532 	int off = off0;
533 	struct mbuf *top;
534 	int copyhdr = 0;
535 
536 	if (off < 0 || len < 0)
537 		panic("m_copym: off %d, len %d", off, len);
538 	if (off == 0 && m->m_flags & M_PKTHDR)
539 		copyhdr = 1;
540 	while (off > 0) {
541 		if (m == 0)
542 			panic("m_copym: m == 0, off %d", off);
543 		if (off < m->m_len)
544 			break;
545 		off -= m->m_len;
546 		m = m->m_next;
547 	}
548 	np = &top;
549 	top = 0;
550 	while (len > 0) {
551 		if (m == 0) {
552 			if (len != M_COPYALL)
553 				panic("m_copym: m == 0, len %d [!COPYALL]",
554 				    len);
555 			break;
556 		}
557 		MGET(n, wait, m->m_type);
558 		*np = n;
559 		if (n == 0)
560 			goto nospace;
561 		MCLAIM(n, m->m_owner);
562 		if (copyhdr) {
563 			M_COPY_PKTHDR(n, m);
564 			if (len == M_COPYALL)
565 				n->m_pkthdr.len -= off0;
566 			else
567 				n->m_pkthdr.len = len;
568 			copyhdr = 0;
569 		}
570 		n->m_len = min(len, m->m_len - off);
571 		if (m->m_flags & M_EXT) {
572 			if (!deep) {
573 				n->m_data = m->m_data + off;
574 				n->m_ext = m->m_ext;
575 				MCLADDREFERENCE(m, n);
576 			} else {
577 				/*
578 				 * we are unsure about the way m was allocated.
579 				 * copy into multiple MCLBYTES cluster mbufs.
580 				 */
581 				MCLGET(n, wait);
582 				n->m_len = 0;
583 				n->m_len = M_TRAILINGSPACE(n);
584 				n->m_len = min(n->m_len, len);
585 				n->m_len = min(n->m_len, m->m_len - off);
586 				memcpy(mtod(n, caddr_t), mtod(m, caddr_t) + off,
587 				    (unsigned)n->m_len);
588 			}
589 		} else
590 			memcpy(mtod(n, caddr_t), mtod(m, caddr_t)+off,
591 			    (unsigned)n->m_len);
592 		if (len != M_COPYALL)
593 			len -= n->m_len;
594 		off += n->m_len;
595 #ifdef DIAGNOSTIC
596 		if (off > m->m_len)
597 			panic("m_copym0 overrun");
598 #endif
599 		if (off == m->m_len) {
600 			m = m->m_next;
601 			off = 0;
602 		}
603 		np = &n->m_next;
604 	}
605 	if (top == 0)
606 		MCFail++;
607 	return (top);
608 nospace:
609 	m_freem(top);
610 	MCFail++;
611 	return (NULL);
612 }
613 
614 /*
615  * Copy an entire packet, including header (which must be present).
616  * An optimization of the common case `m_copym(m, 0, M_COPYALL, how)'.
617  */
618 struct mbuf *
619 m_copypacket(struct mbuf *m, int how)
620 {
621 	struct mbuf *top, *n, *o;
622 
623 	MGET(n, how, m->m_type);
624 	top = n;
625 	if (!n)
626 		goto nospace;
627 
628 	MCLAIM(n, m->m_owner);
629 	M_COPY_PKTHDR(n, m);
630 	n->m_len = m->m_len;
631 	if (m->m_flags & M_EXT) {
632 		n->m_data = m->m_data;
633 		n->m_ext = m->m_ext;
634 		MCLADDREFERENCE(m, n);
635 	} else {
636 		memcpy(mtod(n, char *), mtod(m, char *), n->m_len);
637 	}
638 
639 	m = m->m_next;
640 	while (m) {
641 		MGET(o, how, m->m_type);
642 		if (!o)
643 			goto nospace;
644 
645 		MCLAIM(o, m->m_owner);
646 		n->m_next = o;
647 		n = n->m_next;
648 
649 		n->m_len = m->m_len;
650 		if (m->m_flags & M_EXT) {
651 			n->m_data = m->m_data;
652 			n->m_ext = m->m_ext;
653 			MCLADDREFERENCE(m, n);
654 		} else {
655 			memcpy(mtod(n, char *), mtod(m, char *), n->m_len);
656 		}
657 
658 		m = m->m_next;
659 	}
660 	return top;
661 nospace:
662 	m_freem(top);
663 	MCFail++;
664 	return NULL;
665 }
666 
667 /*
668  * Copy data from an mbuf chain starting "off" bytes from the beginning,
669  * continuing for "len" bytes, into the indicated buffer.
670  */
671 void
672 m_copydata(struct mbuf *m, int off, int len, void *vp)
673 {
674 	unsigned	count;
675 	caddr_t		cp = vp;
676 
677 	if (off < 0 || len < 0)
678 		panic("m_copydata: off %d, len %d", off, len);
679 	while (off > 0) {
680 		if (m == NULL)
681 			panic("m_copydata: m == NULL, off %d", off);
682 		if (off < m->m_len)
683 			break;
684 		off -= m->m_len;
685 		m = m->m_next;
686 	}
687 	while (len > 0) {
688 		if (m == NULL)
689 			panic("m_copydata: m == NULL, len %d", len);
690 		count = min(m->m_len - off, len);
691 		memcpy(cp, mtod(m, caddr_t) + off, count);
692 		len -= count;
693 		cp += count;
694 		off = 0;
695 		m = m->m_next;
696 	}
697 }
698 
699 /*
700  * Concatenate mbuf chain n to m.
701  * n might be copied into m (when n->m_len is small), therefore data portion of
702  * n could be copied into an mbuf of different mbuf type.
703  * Any m_pkthdr is not updated.
704  */
705 void
706 m_cat(struct mbuf *m, struct mbuf *n)
707 {
708 
709 	while (m->m_next)
710 		m = m->m_next;
711 	while (n) {
712 		if (M_READONLY(m) || n->m_len > M_TRAILINGSPACE(m)) {
713 			/* just join the two chains */
714 			m->m_next = n;
715 			return;
716 		}
717 		/* splat the data from one into the other */
718 		memcpy(mtod(m, caddr_t) + m->m_len, mtod(n, caddr_t),
719 		    (u_int)n->m_len);
720 		m->m_len += n->m_len;
721 		n = m_free(n);
722 	}
723 }
724 
725 void
726 m_adj(struct mbuf *mp, int req_len)
727 {
728 	int len = req_len;
729 	struct mbuf *m;
730 	int count;
731 
732 	if ((m = mp) == NULL)
733 		return;
734 	if (len >= 0) {
735 		/*
736 		 * Trim from head.
737 		 */
738 		while (m != NULL && len > 0) {
739 			if (m->m_len <= len) {
740 				len -= m->m_len;
741 				m->m_len = 0;
742 				m = m->m_next;
743 			} else {
744 				m->m_len -= len;
745 				m->m_data += len;
746 				len = 0;
747 			}
748 		}
749 		m = mp;
750 		if (mp->m_flags & M_PKTHDR)
751 			m->m_pkthdr.len -= (req_len - len);
752 	} else {
753 		/*
754 		 * Trim from tail.  Scan the mbuf chain,
755 		 * calculating its length and finding the last mbuf.
756 		 * If the adjustment only affects this mbuf, then just
757 		 * adjust and return.  Otherwise, rescan and truncate
758 		 * after the remaining size.
759 		 */
760 		len = -len;
761 		count = 0;
762 		for (;;) {
763 			count += m->m_len;
764 			if (m->m_next == (struct mbuf *)0)
765 				break;
766 			m = m->m_next;
767 		}
768 		if (m->m_len >= len) {
769 			m->m_len -= len;
770 			if (mp->m_flags & M_PKTHDR)
771 				mp->m_pkthdr.len -= len;
772 			return;
773 		}
774 		count -= len;
775 		if (count < 0)
776 			count = 0;
777 		/*
778 		 * Correct length for chain is "count".
779 		 * Find the mbuf with last data, adjust its length,
780 		 * and toss data from remaining mbufs on chain.
781 		 */
782 		m = mp;
783 		if (m->m_flags & M_PKTHDR)
784 			m->m_pkthdr.len = count;
785 		for (; m; m = m->m_next) {
786 			if (m->m_len >= count) {
787 				m->m_len = count;
788 				break;
789 			}
790 			count -= m->m_len;
791 		}
792 		if (m)
793 			while (m->m_next)
794 				(m = m->m_next)->m_len = 0;
795 	}
796 }
797 
798 /*
799  * Rearrange an mbuf chain so that len bytes are contiguous
800  * and in the data area of an mbuf (so that mtod and dtom
801  * will work for a structure of size len).  Returns the resulting
802  * mbuf chain on success, frees it and returns null on failure.
803  * If there is room, it will add up to max_protohdr-len extra bytes to the
804  * contiguous region in an attempt to avoid being called next time.
805  */
806 int MPFail;
807 
808 struct mbuf *
809 m_pullup(struct mbuf *n, int len)
810 {
811 	struct mbuf *m;
812 	int count;
813 	int space;
814 
815 	/*
816 	 * If first mbuf has no cluster, and has room for len bytes
817 	 * without shifting current data, pullup into it,
818 	 * otherwise allocate a new mbuf to prepend to the chain.
819 	 */
820 	if ((n->m_flags & M_EXT) == 0 &&
821 	    n->m_data + len < &n->m_dat[MLEN] && n->m_next) {
822 		if (n->m_len >= len)
823 			return (n);
824 		m = n;
825 		n = n->m_next;
826 		len -= m->m_len;
827 	} else {
828 		if (len > MHLEN)
829 			goto bad;
830 		MGET(m, M_DONTWAIT, n->m_type);
831 		if (m == 0)
832 			goto bad;
833 		MCLAIM(m, n->m_owner);
834 		m->m_len = 0;
835 		if (n->m_flags & M_PKTHDR) {
836 			M_MOVE_PKTHDR(m, n);
837 		}
838 	}
839 	space = &m->m_dat[MLEN] - (m->m_data + m->m_len);
840 	do {
841 		count = min(min(max(len, max_protohdr), space), n->m_len);
842 		memcpy(mtod(m, caddr_t) + m->m_len, mtod(n, caddr_t),
843 		  (unsigned)count);
844 		len -= count;
845 		m->m_len += count;
846 		n->m_len -= count;
847 		space -= count;
848 		if (n->m_len)
849 			n->m_data += count;
850 		else
851 			n = m_free(n);
852 	} while (len > 0 && n);
853 	if (len > 0) {
854 		(void) m_free(m);
855 		goto bad;
856 	}
857 	m->m_next = n;
858 	return (m);
859 bad:
860 	m_freem(n);
861 	MPFail++;
862 	return (NULL);
863 }
864 
865 /*
866  * Like m_pullup(), except a new mbuf is always allocated, and we allow
867  * the amount of empty space before the data in the new mbuf to be specified
868  * (in the event that the caller expects to prepend later).
869  */
870 int MSFail;
871 
872 struct mbuf *
873 m_copyup(struct mbuf *n, int len, int dstoff)
874 {
875 	struct mbuf *m;
876 	int count, space;
877 
878 	if (len > (MHLEN - dstoff))
879 		goto bad;
880 	MGET(m, M_DONTWAIT, n->m_type);
881 	if (m == NULL)
882 		goto bad;
883 	MCLAIM(m, n->m_owner);
884 	m->m_len = 0;
885 	if (n->m_flags & M_PKTHDR) {
886 		M_MOVE_PKTHDR(m, n);
887 	}
888 	m->m_data += dstoff;
889 	space = &m->m_dat[MLEN] - (m->m_data + m->m_len);
890 	do {
891 		count = min(min(max(len, max_protohdr), space), n->m_len);
892 		memcpy(mtod(m, caddr_t) + m->m_len, mtod(n, caddr_t),
893 		    (unsigned)count);
894 		len -= count;
895 		m->m_len += count;
896 		n->m_len -= count;
897 		space -= count;
898 		if (n->m_len)
899 			n->m_data += count;
900 		else
901 			n = m_free(n);
902 	} while (len > 0 && n);
903 	if (len > 0) {
904 		(void) m_free(m);
905 		goto bad;
906 	}
907 	m->m_next = n;
908 	return (m);
909  bad:
910 	m_freem(n);
911 	MSFail++;
912 	return (NULL);
913 }
914 
915 /*
916  * Partition an mbuf chain in two pieces, returning the tail --
917  * all but the first len0 bytes.  In case of failure, it returns NULL and
918  * attempts to restore the chain to its original state.
919  */
920 struct mbuf *
921 m_split(struct mbuf *m0, int len0, int wait)
922 {
923 
924 	return m_split0(m0, len0, wait, 1);
925 }
926 
927 static struct mbuf *
928 m_split0(struct mbuf *m0, int len0, int wait, int copyhdr)
929 {
930 	struct mbuf *m, *n;
931 	unsigned len = len0, remain, len_save;
932 
933 	for (m = m0; m && len > m->m_len; m = m->m_next)
934 		len -= m->m_len;
935 	if (m == 0)
936 		return (NULL);
937 	remain = m->m_len - len;
938 	if (copyhdr && (m0->m_flags & M_PKTHDR)) {
939 		MGETHDR(n, wait, m0->m_type);
940 		if (n == 0)
941 			return (NULL);
942 		MCLAIM(n, m0->m_owner);
943 		n->m_pkthdr.rcvif = m0->m_pkthdr.rcvif;
944 		n->m_pkthdr.len = m0->m_pkthdr.len - len0;
945 		len_save = m0->m_pkthdr.len;
946 		m0->m_pkthdr.len = len0;
947 		if (m->m_flags & M_EXT)
948 			goto extpacket;
949 		if (remain > MHLEN) {
950 			/* m can't be the lead packet */
951 			MH_ALIGN(n, 0);
952 			n->m_next = m_split(m, len, wait);
953 			if (n->m_next == 0) {
954 				(void) m_free(n);
955 				m0->m_pkthdr.len = len_save;
956 				return (NULL);
957 			} else
958 				return (n);
959 		} else
960 			MH_ALIGN(n, remain);
961 	} else if (remain == 0) {
962 		n = m->m_next;
963 		m->m_next = 0;
964 		return (n);
965 	} else {
966 		MGET(n, wait, m->m_type);
967 		if (n == 0)
968 			return (NULL);
969 		MCLAIM(n, m->m_owner);
970 		M_ALIGN(n, remain);
971 	}
972 extpacket:
973 	if (m->m_flags & M_EXT) {
974 		n->m_ext = m->m_ext;
975 		MCLADDREFERENCE(m, n);
976 		n->m_data = m->m_data + len;
977 	} else {
978 		memcpy(mtod(n, caddr_t), mtod(m, caddr_t) + len, remain);
979 	}
980 	n->m_len = remain;
981 	m->m_len = len;
982 	n->m_next = m->m_next;
983 	m->m_next = 0;
984 	return (n);
985 }
986 /*
987  * Routine to copy from device local memory into mbufs.
988  */
989 struct mbuf *
990 m_devget(char *buf, int totlen, int off0, struct ifnet *ifp,
991     void (*copy)(const void *from, void *to, size_t len))
992 {
993 	struct mbuf *m;
994 	struct mbuf *top = 0, **mp = &top;
995 	int off = off0, len;
996 	char *cp;
997 	char *epkt;
998 
999 	cp = buf;
1000 	epkt = cp + totlen;
1001 	if (off) {
1002 		/*
1003 		 * If 'off' is non-zero, packet is trailer-encapsulated,
1004 		 * so we have to skip the type and length fields.
1005 		 */
1006 		cp += off + 2 * sizeof(uint16_t);
1007 		totlen -= 2 * sizeof(uint16_t);
1008 	}
1009 	MGETHDR(m, M_DONTWAIT, MT_DATA);
1010 	if (m == 0)
1011 		return (NULL);
1012 	m->m_pkthdr.rcvif = ifp;
1013 	m->m_pkthdr.len = totlen;
1014 	m->m_len = MHLEN;
1015 
1016 	while (totlen > 0) {
1017 		if (top) {
1018 			MGET(m, M_DONTWAIT, MT_DATA);
1019 			if (m == 0) {
1020 				m_freem(top);
1021 				return (NULL);
1022 			}
1023 			m->m_len = MLEN;
1024 		}
1025 		len = min(totlen, epkt - cp);
1026 		if (len >= MINCLSIZE) {
1027 			MCLGET(m, M_DONTWAIT);
1028 			if ((m->m_flags & M_EXT) == 0) {
1029 				m_free(m);
1030 				m_freem(top);
1031 				return (NULL);
1032 			}
1033 			m->m_len = len = min(len, MCLBYTES);
1034 		} else {
1035 			/*
1036 			 * Place initial small packet/header at end of mbuf.
1037 			 */
1038 			if (len < m->m_len) {
1039 				if (top == 0 && len + max_linkhdr <= m->m_len)
1040 					m->m_data += max_linkhdr;
1041 				m->m_len = len;
1042 			} else
1043 				len = m->m_len;
1044 		}
1045 		if (copy)
1046 			copy(cp, mtod(m, caddr_t), (size_t)len);
1047 		else
1048 			memcpy(mtod(m, caddr_t), cp, (size_t)len);
1049 		cp += len;
1050 		*mp = m;
1051 		mp = &m->m_next;
1052 		totlen -= len;
1053 		if (cp == epkt)
1054 			cp = buf;
1055 	}
1056 	return (top);
1057 }
1058 
1059 /*
1060  * Copy data from a buffer back into the indicated mbuf chain,
1061  * starting "off" bytes from the beginning, extending the mbuf
1062  * chain if necessary.
1063  */
1064 void
1065 m_copyback(struct mbuf *m0, int off, int len, const void *cp)
1066 {
1067 #if defined(DEBUG)
1068 	struct mbuf *origm = m0;
1069 	int error;
1070 #endif /* defined(DEBUG) */
1071 
1072 	if (m0 == NULL)
1073 		return;
1074 
1075 #if defined(DEBUG)
1076 	error =
1077 #endif /* defined(DEBUG) */
1078 	m_copyback0(&m0, off, len, cp,
1079 	    M_COPYBACK0_COPYBACK|M_COPYBACK0_EXTEND, M_DONTWAIT);
1080 
1081 #if defined(DEBUG)
1082 	if (error != 0 || (m0 != NULL && origm != m0))
1083 		panic("m_copyback");
1084 #endif /* defined(DEBUG) */
1085 }
1086 
1087 struct mbuf *
1088 m_copyback_cow(struct mbuf *m0, int off, int len, const void *cp, int how)
1089 {
1090 	int error;
1091 
1092 	/* don't support chain expansion */
1093 	KDASSERT(off + len <= m_length(m0));
1094 
1095 	error = m_copyback0(&m0, off, len, cp,
1096 	    M_COPYBACK0_COPYBACK|M_COPYBACK0_COW, how);
1097 	if (error) {
1098 		/*
1099 		 * no way to recover from partial success.
1100 		 * just free the chain.
1101 		 */
1102 		m_freem(m0);
1103 		return NULL;
1104 	}
1105 	return m0;
1106 }
1107 
1108 /*
1109  * m_makewritable: ensure the specified range writable.
1110  */
1111 int
1112 m_makewritable(struct mbuf **mp, int off, int len, int how)
1113 {
1114 	int error;
1115 #if defined(DEBUG)
1116 	struct mbuf *n;
1117 	int origlen, reslen;
1118 
1119 	origlen = m_length(*mp);
1120 #endif /* defined(DEBUG) */
1121 
1122 #if 0 /* M_COPYALL is large enough */
1123 	if (len == M_COPYALL)
1124 		len = m_length(*mp) - off; /* XXX */
1125 #endif
1126 
1127 	error = m_copyback0(mp, off, len, NULL,
1128 	    M_COPYBACK0_PRESERVE|M_COPYBACK0_COW, how);
1129 
1130 #if defined(DEBUG)
1131 	reslen = 0;
1132 	for (n = *mp; n; n = n->m_next)
1133 		reslen += n->m_len;
1134 	if (origlen != reslen)
1135 		panic("m_makewritable: length changed");
1136 	if (((*mp)->m_flags & M_PKTHDR) != 0 && reslen != (*mp)->m_pkthdr.len)
1137 		panic("m_makewritable: inconsist");
1138 #endif /* defined(DEBUG) */
1139 
1140 	return error;
1141 }
1142 
1143 int
1144 m_copyback0(struct mbuf **mp0, int off, int len, const void *vp, int flags,
1145     int how)
1146 {
1147 	int mlen;
1148 	struct mbuf *m, *n;
1149 	struct mbuf **mp;
1150 	int totlen = 0;
1151 	const char *cp = vp;
1152 
1153 	KASSERT(mp0 != NULL);
1154 	KASSERT(*mp0 != NULL);
1155 	KASSERT((flags & M_COPYBACK0_PRESERVE) == 0 || cp == NULL);
1156 	KASSERT((flags & M_COPYBACK0_COPYBACK) == 0 || cp != NULL);
1157 
1158 	/*
1159 	 * we don't bother to update "totlen" in the case of M_COPYBACK0_COW,
1160 	 * assuming that M_COPYBACK0_EXTEND and M_COPYBACK0_COW are exclusive.
1161 	 */
1162 
1163 	KASSERT((~flags & (M_COPYBACK0_EXTEND|M_COPYBACK0_COW)) != 0);
1164 
1165 	mp = mp0;
1166 	m = *mp;
1167 	while (off > (mlen = m->m_len)) {
1168 		off -= mlen;
1169 		totlen += mlen;
1170 		if (m->m_next == NULL) {
1171 			int tspace;
1172 extend:
1173 			if ((flags & M_COPYBACK0_EXTEND) == 0)
1174 				goto out;
1175 
1176 			/*
1177 			 * try to make some space at the end of "m".
1178 			 */
1179 
1180 			mlen = m->m_len;
1181 			if (off + len >= MINCLSIZE &&
1182 			    (m->m_flags & M_EXT) == 0 && m->m_len == 0) {
1183 				MCLGET(m, how);
1184 			}
1185 			tspace = M_TRAILINGSPACE(m);
1186 			if (tspace > 0) {
1187 				tspace = min(tspace, off + len);
1188 				KASSERT(tspace > 0);
1189 				memset(mtod(m, char *) + m->m_len, 0,
1190 				    min(off, tspace));
1191 				m->m_len += tspace;
1192 				off += mlen;
1193 				totlen -= mlen;
1194 				continue;
1195 			}
1196 
1197 			/*
1198 			 * need to allocate an mbuf.
1199 			 */
1200 
1201 			if (off + len >= MINCLSIZE) {
1202 				n = m_getcl(how, m->m_type, 0);
1203 			} else {
1204 				n = m_get(how, m->m_type);
1205 			}
1206 			if (n == NULL) {
1207 				goto out;
1208 			}
1209 			n->m_len = 0;
1210 			n->m_len = min(M_TRAILINGSPACE(n), off + len);
1211 			memset(mtod(n, char *), 0, min(n->m_len, off));
1212 			m->m_next = n;
1213 		}
1214 		mp = &m->m_next;
1215 		m = m->m_next;
1216 	}
1217 	while (len > 0) {
1218 		mlen = m->m_len - off;
1219 		if (mlen != 0 && M_READONLY(m)) {
1220 			char *datap;
1221 			int eatlen;
1222 
1223 			/*
1224 			 * this mbuf is read-only.
1225 			 * allocate a new writable mbuf and try again.
1226 			 */
1227 
1228 #if defined(DIAGNOSTIC)
1229 			if ((flags & M_COPYBACK0_COW) == 0)
1230 				panic("m_copyback0: read-only");
1231 #endif /* defined(DIAGNOSTIC) */
1232 
1233 			/*
1234 			 * if we're going to write into the middle of
1235 			 * a mbuf, split it first.
1236 			 */
1237 			if (off > 0 && len < mlen) {
1238 				n = m_split0(m, off, how, 0);
1239 				if (n == NULL)
1240 					goto enobufs;
1241 				m->m_next = n;
1242 				mp = &m->m_next;
1243 				m = n;
1244 				off = 0;
1245 				continue;
1246 			}
1247 
1248 			/*
1249 			 * XXX TODO coalesce into the trailingspace of
1250 			 * the previous mbuf when possible.
1251 			 */
1252 
1253 			/*
1254 			 * allocate a new mbuf.  copy packet header if needed.
1255 			 */
1256 			MGET(n, how, m->m_type);
1257 			if (n == NULL)
1258 				goto enobufs;
1259 			MCLAIM(n, m->m_owner);
1260 			if (off == 0 && (m->m_flags & M_PKTHDR) != 0) {
1261 				M_MOVE_PKTHDR(n, m);
1262 				n->m_len = MHLEN;
1263 			} else {
1264 				if (len >= MINCLSIZE)
1265 					MCLGET(n, M_DONTWAIT);
1266 				n->m_len =
1267 				    (n->m_flags & M_EXT) ? MCLBYTES : MLEN;
1268 			}
1269 			if (n->m_len > len)
1270 				n->m_len = len;
1271 
1272 			/*
1273 			 * free the region which has been overwritten.
1274 			 * copying data from old mbufs if requested.
1275 			 */
1276 			if (flags & M_COPYBACK0_PRESERVE)
1277 				datap = mtod(n, char *);
1278 			else
1279 				datap = NULL;
1280 			eatlen = n->m_len;
1281 			KDASSERT(off == 0 || eatlen >= mlen);
1282 			if (off > 0) {
1283 				KDASSERT(len >= mlen);
1284 				m->m_len = off;
1285 				m->m_next = n;
1286 				if (datap) {
1287 					m_copydata(m, off, mlen, datap);
1288 					datap += mlen;
1289 				}
1290 				eatlen -= mlen;
1291 				mp = &m->m_next;
1292 				m = m->m_next;
1293 			}
1294 			while (m != NULL && M_READONLY(m) &&
1295 			    n->m_type == m->m_type && eatlen > 0) {
1296 				mlen = min(eatlen, m->m_len);
1297 				if (datap) {
1298 					m_copydata(m, 0, mlen, datap);
1299 					datap += mlen;
1300 				}
1301 				m->m_data += mlen;
1302 				m->m_len -= mlen;
1303 				eatlen -= mlen;
1304 				if (m->m_len == 0)
1305 					*mp = m = m_free(m);
1306 			}
1307 			if (eatlen > 0)
1308 				n->m_len -= eatlen;
1309 			n->m_next = m;
1310 			*mp = m = n;
1311 			continue;
1312 		}
1313 		mlen = min(mlen, len);
1314 		if (flags & M_COPYBACK0_COPYBACK) {
1315 			memcpy(mtod(m, caddr_t) + off, cp, (unsigned)mlen);
1316 			cp += mlen;
1317 		}
1318 		len -= mlen;
1319 		mlen += off;
1320 		off = 0;
1321 		totlen += mlen;
1322 		if (len == 0)
1323 			break;
1324 		if (m->m_next == NULL) {
1325 			goto extend;
1326 		}
1327 		mp = &m->m_next;
1328 		m = m->m_next;
1329 	}
1330 out:	if (((m = *mp0)->m_flags & M_PKTHDR) && (m->m_pkthdr.len < totlen)) {
1331 		KASSERT((flags & M_COPYBACK0_EXTEND) != 0);
1332 		m->m_pkthdr.len = totlen;
1333 	}
1334 
1335 	return 0;
1336 
1337 enobufs:
1338 	return ENOBUFS;
1339 }
1340 
1341 void
1342 m_move_pkthdr(struct mbuf *to, struct mbuf *from)
1343 {
1344 
1345 	KASSERT((to->m_flags & M_EXT) == 0);
1346 	KASSERT((to->m_flags & M_PKTHDR) == 0 || m_tag_first(to) == NULL);
1347 	KASSERT((from->m_flags & M_PKTHDR) != 0);
1348 
1349 	to->m_pkthdr = from->m_pkthdr;
1350 	to->m_flags = from->m_flags & M_COPYFLAGS;
1351 	to->m_data = to->m_pktdat;
1352 
1353 	from->m_flags &= ~M_PKTHDR;
1354 }
1355 
1356 /*
1357  * Apply function f to the data in an mbuf chain starting "off" bytes from the
1358  * beginning, continuing for "len" bytes.
1359  */
1360 int
1361 m_apply(struct mbuf *m, int off, int len,
1362     int (*f)(void *, caddr_t, unsigned int), void *arg)
1363 {
1364 	unsigned int count;
1365 	int rval;
1366 
1367 	KASSERT(len >= 0);
1368 	KASSERT(off >= 0);
1369 
1370 	while (off > 0) {
1371 		KASSERT(m != NULL);
1372 		if (off < m->m_len)
1373 			break;
1374 		off -= m->m_len;
1375 		m = m->m_next;
1376 	}
1377 	while (len > 0) {
1378 		KASSERT(m != NULL);
1379 		count = min(m->m_len - off, len);
1380 
1381 		rval = (*f)(arg, mtod(m, caddr_t) + off, count);
1382 		if (rval)
1383 			return (rval);
1384 
1385 		len -= count;
1386 		off = 0;
1387 		m = m->m_next;
1388 	}
1389 
1390 	return (0);
1391 }
1392 
1393 /*
1394  * Return a pointer to mbuf/offset of location in mbuf chain.
1395  */
1396 struct mbuf *
1397 m_getptr(struct mbuf *m, int loc, int *off)
1398 {
1399 
1400 	while (loc >= 0) {
1401 		/* Normal end of search */
1402 		if (m->m_len > loc) {
1403 	    		*off = loc;
1404 	    		return (m);
1405 		} else {
1406 	    		loc -= m->m_len;
1407 
1408 	    		if (m->m_next == NULL) {
1409 				if (loc == 0) {
1410  					/* Point at the end of valid data */
1411 		    			*off = m->m_len;
1412 		    			return (m);
1413 				} else
1414 		  			return (NULL);
1415 	    		} else
1416 	      			m = m->m_next;
1417 		}
1418     	}
1419 
1420 	return (NULL);
1421 }
1422 
1423 #if defined(DDB)
1424 void
1425 m_print(const struct mbuf *m, const char *modif, void (*pr)(const char *, ...))
1426 {
1427 	char ch;
1428 	boolean_t opt_c = FALSE;
1429 	char buf[512];
1430 
1431 	while ((ch = *(modif++)) != '\0') {
1432 		switch (ch) {
1433 		case 'c':
1434 			opt_c = TRUE;
1435 			break;
1436 		}
1437 	}
1438 
1439 nextchain:
1440 	(*pr)("MBUF %p\n", m);
1441 	bitmask_snprintf(m->m_flags, M_FLAGS_BITS, buf, sizeof(buf));
1442 	(*pr)("  data=%p, len=%d, type=%d, flags=0x%s\n",
1443 	    m->m_data, m->m_len, m->m_type, buf);
1444 	(*pr)("  owner=%p, next=%p, nextpkt=%p\n", m->m_owner, m->m_next,
1445 	    m->m_nextpkt);
1446 	(*pr)("  leadingspace=%u, trailingspace=%u, readonly=%u\n",
1447 	    (int)M_LEADINGSPACE(m), (int)M_TRAILINGSPACE(m),
1448 	    (int)M_READONLY(m));
1449 	if ((m->m_flags & M_PKTHDR) != 0) {
1450 		bitmask_snprintf(m->m_pkthdr.csum_flags, M_CSUM_BITS, buf,
1451 		    sizeof(buf));
1452 		(*pr)("  pktlen=%d, rcvif=%p, csum_flags=0x%s, csum_data=0x%"
1453 		    PRIx32 ", segsz=%u\n",
1454 		    m->m_pkthdr.len, m->m_pkthdr.rcvif,
1455 		    buf, m->m_pkthdr.csum_data, m->m_pkthdr.segsz);
1456 	}
1457 	if ((m->m_flags & M_EXT)) {
1458 		(*pr)("  shared=%u, ext_buf=%p, ext_size=%zd, "
1459 		    "ext_free=%p, ext_arg=%p\n",
1460 		    (int)MCLISREFERENCED(m),
1461 		    m->m_ext.ext_buf, m->m_ext.ext_size,
1462 		    m->m_ext.ext_free, m->m_ext.ext_arg);
1463 	}
1464 	if ((~m->m_flags & (M_EXT|M_EXT_PAGES)) == 0) {
1465 		vaddr_t sva = (vaddr_t)m->m_ext.ext_buf;
1466 		vaddr_t eva = sva + m->m_ext.ext_size;
1467 		int n = (round_page(eva) - trunc_page(sva)) >> PAGE_SHIFT;
1468 		int i;
1469 
1470 		(*pr)("  pages:");
1471 		for (i = 0; i < n; i ++) {
1472 			(*pr)(" %p", m->m_ext.ext_pgs[i]);
1473 		}
1474 		(*pr)("\n");
1475 	}
1476 
1477 	if (opt_c) {
1478 		m = m->m_next;
1479 		if (m != NULL) {
1480 			goto nextchain;
1481 		}
1482 	}
1483 }
1484 #endif /* defined(DDB) */
1485