xref: /netbsd-src/sys/kern/uipc_mbuf.c (revision 8b0f9554ff8762542c4defc4f70e1eb76fb508fa)
1 /*	$NetBSD: uipc_mbuf.c,v 1.123 2007/11/14 14:11:57 yamt 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 1.123 2007/11/14 14:11:57 yamt 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 pool_cache_t mb_cache;	/* mbuf cache */
96 pool_cache_t mcl_cache;	/* mbuf cluster cache */
97 
98 struct mbstat mbstat;
99 int	max_linkhdr;
100 int	max_protohdr;
101 int	max_hdr;
102 int	max_datalen;
103 
104 static int mb_ctor(void *, void *, int);
105 
106 static void	*mclpool_alloc(struct pool *, int);
107 static void	mclpool_release(struct pool *, void *);
108 
109 static struct pool_allocator mclpool_allocator = {
110 	.pa_alloc = mclpool_alloc,
111 	.pa_free = mclpool_release,
112 };
113 
114 static struct mbuf *m_copym0(struct mbuf *, int, int, int, int);
115 static struct mbuf *m_split0(struct mbuf *, int, int, int);
116 static int m_copyback0(struct mbuf **, int, int, const void *, int, int);
117 
118 /* flags for m_copyback0 */
119 #define	M_COPYBACK0_COPYBACK	0x0001	/* copyback from cp */
120 #define	M_COPYBACK0_PRESERVE	0x0002	/* preserve original data */
121 #define	M_COPYBACK0_COW		0x0004	/* do copy-on-write */
122 #define	M_COPYBACK0_EXTEND	0x0008	/* extend chain */
123 
124 static const char mclpool_warnmsg[] =
125     "WARNING: mclpool limit reached; increase NMBCLUSTERS";
126 
127 MALLOC_DEFINE(M_MBUF, "mbuf", "mbuf");
128 
129 #ifdef MBUFTRACE
130 struct mownerhead mowners = LIST_HEAD_INITIALIZER(mowners);
131 struct mowner unknown_mowners[] = {
132 	MOWNER_INIT("unknown", "free"),
133 	MOWNER_INIT("unknown", "data"),
134 	MOWNER_INIT("unknown", "header"),
135 	MOWNER_INIT("unknown", "soname"),
136 	MOWNER_INIT("unknown", "soopts"),
137 	MOWNER_INIT("unknown", "ftable"),
138 	MOWNER_INIT("unknown", "control"),
139 	MOWNER_INIT("unknown", "oobdata"),
140 };
141 struct mowner revoked_mowner = MOWNER_INIT("revoked", "");
142 #endif
143 
144 /*
145  * Initialize the mbuf allocator.
146  */
147 void
148 mbinit(void)
149 {
150 
151 	KASSERT(sizeof(struct _m_ext) <= MHLEN);
152 	KASSERT(sizeof(struct mbuf) == MSIZE);
153 
154 	mclpool_allocator.pa_backingmap = mb_map;
155 
156 	mb_cache = pool_cache_init(msize, 0, 0, 0, "mbpl",
157 	    NULL, IPL_VM, mb_ctor, NULL, NULL);
158 	KASSERT(mb_cache != NULL);
159 
160 	mcl_cache = pool_cache_init(mclbytes, 0, 0, 0, "mclpl",
161 	    &mclpool_allocator, IPL_VM, NULL, NULL, NULL);
162 	KASSERT(mcl_cache != NULL);
163 
164 	pool_cache_set_drain_hook(mb_cache, m_reclaim, NULL);
165 	pool_cache_set_drain_hook(mcl_cache, m_reclaim, NULL);
166 
167 	/*
168 	 * Set the hard limit on the mclpool to the number of
169 	 * mbuf clusters the kernel is to support.  Log the limit
170 	 * reached message max once a minute.
171 	 */
172 	pool_cache_sethardlimit(mcl_cache, nmbclusters, mclpool_warnmsg, 60);
173 
174 	/*
175 	 * Set a low water mark for both mbufs and clusters.  This should
176 	 * help ensure that they can be allocated in a memory starvation
177 	 * situation.  This is important for e.g. diskless systems which
178 	 * must allocate mbufs in order for the pagedaemon to clean pages.
179 	 */
180 	pool_cache_setlowat(mb_cache, mblowat);
181 	pool_cache_setlowat(mcl_cache, mcllowat);
182 
183 #ifdef MBUFTRACE
184 	{
185 		/*
186 		 * Attach the unknown mowners.
187 		 */
188 		int i;
189 		MOWNER_ATTACH(&revoked_mowner);
190 		for (i = sizeof(unknown_mowners)/sizeof(unknown_mowners[0]);
191 		     i-- > 0; )
192 			MOWNER_ATTACH(&unknown_mowners[i]);
193 	}
194 #endif
195 }
196 
197 /*
198  * sysctl helper routine for the kern.mbuf subtree.  nmbclusters may
199  * or may not be writable, and mblowat and mcllowat need range
200  * checking and pool tweaking after being reset.
201  */
202 static int
203 sysctl_kern_mbuf(SYSCTLFN_ARGS)
204 {
205 	int error, newval;
206 	struct sysctlnode node;
207 
208 	node = *rnode;
209 	node.sysctl_data = &newval;
210 	switch (rnode->sysctl_num) {
211 	case MBUF_NMBCLUSTERS:
212 		if (mb_map != NULL) {
213 			node.sysctl_flags &= ~CTLFLAG_READWRITE;
214 			node.sysctl_flags |= CTLFLAG_READONLY;
215 		}
216 		/* FALLTHROUGH */
217 	case MBUF_MBLOWAT:
218 	case MBUF_MCLLOWAT:
219 		newval = *(int*)rnode->sysctl_data;
220 		break;
221 	default:
222 		return (EOPNOTSUPP);
223 	}
224 
225 	error = sysctl_lookup(SYSCTLFN_CALL(&node));
226 	if (error || newp == NULL)
227 		return (error);
228 	if (newval < 0)
229 		return (EINVAL);
230 
231 	switch (node.sysctl_num) {
232 	case MBUF_NMBCLUSTERS:
233 		if (newval < nmbclusters)
234 			return (EINVAL);
235 		nmbclusters = newval;
236 		pool_cache_sethardlimit(mcl_cache, nmbclusters,
237 		    mclpool_warnmsg, 60);
238 		break;
239 	case MBUF_MBLOWAT:
240 		mblowat = newval;
241 		pool_cache_setlowat(mb_cache, mblowat);
242 		break;
243 	case MBUF_MCLLOWAT:
244 		mcllowat = newval;
245 		pool_cache_setlowat(mcl_cache, 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, (char *)oldp + len, sizeof(*mo));
272 			if (error)
273 				break;
274 		}
275 		len += sizeof(*mo);
276 	}
277 
278 	if (error == 0)
279 		*oldlenp = len;
280 
281 	return (error);
282 }
283 #endif /* MBUFTRACE */
284 
285 SYSCTL_SETUP(sysctl_kern_mbuf_setup, "sysctl kern.mbuf subtree setup")
286 {
287 
288 	sysctl_createv(clog, 0, NULL, NULL,
289 		       CTLFLAG_PERMANENT,
290 		       CTLTYPE_NODE, "kern", NULL,
291 		       NULL, 0, NULL, 0,
292 		       CTL_KERN, CTL_EOL);
293 	sysctl_createv(clog, 0, NULL, NULL,
294 		       CTLFLAG_PERMANENT,
295 		       CTLTYPE_NODE, "mbuf",
296 		       SYSCTL_DESCR("mbuf control variables"),
297 		       NULL, 0, NULL, 0,
298 		       CTL_KERN, KERN_MBUF, CTL_EOL);
299 
300 	sysctl_createv(clog, 0, NULL, NULL,
301 		       CTLFLAG_PERMANENT|CTLFLAG_IMMEDIATE,
302 		       CTLTYPE_INT, "msize",
303 		       SYSCTL_DESCR("mbuf base size"),
304 		       NULL, msize, NULL, 0,
305 		       CTL_KERN, KERN_MBUF, MBUF_MSIZE, CTL_EOL);
306 	sysctl_createv(clog, 0, NULL, NULL,
307 		       CTLFLAG_PERMANENT|CTLFLAG_IMMEDIATE,
308 		       CTLTYPE_INT, "mclbytes",
309 		       SYSCTL_DESCR("mbuf cluster size"),
310 		       NULL, mclbytes, NULL, 0,
311 		       CTL_KERN, KERN_MBUF, MBUF_MCLBYTES, CTL_EOL);
312 	sysctl_createv(clog, 0, NULL, NULL,
313 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
314 		       CTLTYPE_INT, "nmbclusters",
315 		       SYSCTL_DESCR("Limit on the number of mbuf clusters"),
316 		       sysctl_kern_mbuf, 0, &nmbclusters, 0,
317 		       CTL_KERN, KERN_MBUF, MBUF_NMBCLUSTERS, CTL_EOL);
318 	sysctl_createv(clog, 0, NULL, NULL,
319 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
320 		       CTLTYPE_INT, "mblowat",
321 		       SYSCTL_DESCR("mbuf low water mark"),
322 		       sysctl_kern_mbuf, 0, &mblowat, 0,
323 		       CTL_KERN, KERN_MBUF, MBUF_MBLOWAT, CTL_EOL);
324 	sysctl_createv(clog, 0, NULL, NULL,
325 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
326 		       CTLTYPE_INT, "mcllowat",
327 		       SYSCTL_DESCR("mbuf cluster low water mark"),
328 		       sysctl_kern_mbuf, 0, &mcllowat, 0,
329 		       CTL_KERN, KERN_MBUF, MBUF_MCLLOWAT, CTL_EOL);
330 	sysctl_createv(clog, 0, NULL, NULL,
331 		       CTLFLAG_PERMANENT,
332 		       CTLTYPE_STRUCT, "stats",
333 		       SYSCTL_DESCR("mbuf allocation statistics"),
334 		       NULL, 0, &mbstat, sizeof(mbstat),
335 		       CTL_KERN, KERN_MBUF, MBUF_STATS, CTL_EOL);
336 #ifdef MBUFTRACE
337 	sysctl_createv(clog, 0, NULL, NULL,
338 		       CTLFLAG_PERMANENT,
339 		       CTLTYPE_STRUCT, "mowners",
340 		       SYSCTL_DESCR("Information about mbuf owners"),
341 		       sysctl_kern_mbuf_mowners, 0, NULL, 0,
342 		       CTL_KERN, KERN_MBUF, MBUF_MOWNERS, CTL_EOL);
343 #endif /* MBUFTRACE */
344 }
345 
346 static void *
347 mclpool_alloc(struct pool *pp, int flags)
348 {
349 	bool waitok = (flags & PR_WAITOK) ? true : false;
350 
351 	return ((void *)uvm_km_alloc_poolpage(mb_map, waitok));
352 }
353 
354 static void
355 mclpool_release(struct pool *pp, void *v)
356 {
357 
358 	uvm_km_free_poolpage(mb_map, (vaddr_t)v);
359 }
360 
361 /*ARGSUSED*/
362 static int
363 mb_ctor(void *arg, void *object, int flags)
364 {
365 	struct mbuf *m = object;
366 
367 #ifdef POOL_VTOPHYS
368 	m->m_paddr = POOL_VTOPHYS(m);
369 #else
370 	m->m_paddr = M_PADDR_INVALID;
371 #endif
372 	return (0);
373 }
374 
375 void
376 m_reclaim(void *arg, int flags)
377 {
378 	struct domain *dp;
379 	const struct protosw *pr;
380 	struct ifnet *ifp;
381 	int s;
382 
383 	KERNEL_LOCK(1, NULL);
384 	s = splvm();
385 	DOMAIN_FOREACH(dp) {
386 		for (pr = dp->dom_protosw;
387 		     pr < dp->dom_protoswNPROTOSW; pr++)
388 			if (pr->pr_drain)
389 				(*pr->pr_drain)();
390 	}
391 	IFNET_FOREACH(ifp) {
392 		if (ifp->if_drain)
393 			(*ifp->if_drain)(ifp);
394 	}
395 	splx(s);
396 	mbstat.m_drain++;
397 	KERNEL_UNLOCK_ONE(NULL);
398 }
399 
400 /*
401  * Space allocation routines.
402  * These are also available as macros
403  * for critical paths.
404  */
405 struct mbuf *
406 m_get(int nowait, int type)
407 {
408 	struct mbuf *m;
409 
410 	MGET(m, nowait, type);
411 	return (m);
412 }
413 
414 struct mbuf *
415 m_gethdr(int nowait, int type)
416 {
417 	struct mbuf *m;
418 
419 	MGETHDR(m, nowait, type);
420 	return (m);
421 }
422 
423 struct mbuf *
424 m_getclr(int nowait, int type)
425 {
426 	struct mbuf *m;
427 
428 	MGET(m, nowait, type);
429 	if (m == 0)
430 		return (NULL);
431 	memset(mtod(m, void *), 0, MLEN);
432 	return (m);
433 }
434 
435 void
436 m_clget(struct mbuf *m, int nowait)
437 {
438 
439 	MCLGET(m, nowait);
440 }
441 
442 struct mbuf *
443 m_free(struct mbuf *m)
444 {
445 	struct mbuf *n;
446 
447 	MFREE(m, n);
448 	return (n);
449 }
450 
451 void
452 m_freem(struct mbuf *m)
453 {
454 	struct mbuf *n;
455 
456 	if (m == NULL)
457 		return;
458 	do {
459 		MFREE(m, n);
460 		m = n;
461 	} while (m);
462 }
463 
464 #ifdef MBUFTRACE
465 /*
466  * Walk a chain of mbufs, claiming ownership of each mbuf in the chain.
467  */
468 void
469 m_claimm(struct mbuf *m, struct mowner *mo)
470 {
471 
472 	for (; m != NULL; m = m->m_next)
473 		MCLAIM(m, mo);
474 }
475 #endif
476 
477 /*
478  * Mbuffer utility routines.
479  */
480 
481 /*
482  * Lesser-used path for M_PREPEND:
483  * allocate new mbuf to prepend to chain,
484  * copy junk along.
485  */
486 struct mbuf *
487 m_prepend(struct mbuf *m, int len, int how)
488 {
489 	struct mbuf *mn;
490 
491 	MGET(mn, how, m->m_type);
492 	if (mn == (struct mbuf *)NULL) {
493 		m_freem(m);
494 		return ((struct mbuf *)NULL);
495 	}
496 	if (m->m_flags & M_PKTHDR) {
497 		M_MOVE_PKTHDR(mn, m);
498 	} else {
499 		MCLAIM(mn, m->m_owner);
500 	}
501 	mn->m_next = m;
502 	m = mn;
503 	if (len < MHLEN)
504 		MH_ALIGN(m, len);
505 	m->m_len = len;
506 	return (m);
507 }
508 
509 /*
510  * Make a copy of an mbuf chain starting "off0" bytes from the beginning,
511  * continuing for "len" bytes.  If len is M_COPYALL, copy to end of mbuf.
512  * The wait parameter is a choice of M_WAIT/M_DONTWAIT from caller.
513  */
514 int MCFail;
515 
516 struct mbuf *
517 m_copym(struct mbuf *m, int off0, int len, int wait)
518 {
519 
520 	return m_copym0(m, off0, len, wait, 0);	/* shallow copy on M_EXT */
521 }
522 
523 struct mbuf *
524 m_dup(struct mbuf *m, int off0, int len, int wait)
525 {
526 
527 	return m_copym0(m, off0, len, wait, 1);	/* deep copy */
528 }
529 
530 static struct mbuf *
531 m_copym0(struct mbuf *m, int off0, int len, int wait, int deep)
532 {
533 	struct mbuf *n, **np;
534 	int off = off0;
535 	struct mbuf *top;
536 	int copyhdr = 0;
537 
538 	if (off < 0 || len < 0)
539 		panic("m_copym: off %d, len %d", off, len);
540 	if (off == 0 && m->m_flags & M_PKTHDR)
541 		copyhdr = 1;
542 	while (off > 0) {
543 		if (m == 0)
544 			panic("m_copym: m == 0, off %d", off);
545 		if (off < m->m_len)
546 			break;
547 		off -= m->m_len;
548 		m = m->m_next;
549 	}
550 	np = &top;
551 	top = 0;
552 	while (len > 0) {
553 		if (m == 0) {
554 			if (len != M_COPYALL)
555 				panic("m_copym: m == 0, len %d [!COPYALL]",
556 				    len);
557 			break;
558 		}
559 		MGET(n, wait, m->m_type);
560 		*np = n;
561 		if (n == 0)
562 			goto nospace;
563 		MCLAIM(n, m->m_owner);
564 		if (copyhdr) {
565 			M_COPY_PKTHDR(n, m);
566 			if (len == M_COPYALL)
567 				n->m_pkthdr.len -= off0;
568 			else
569 				n->m_pkthdr.len = len;
570 			copyhdr = 0;
571 		}
572 		n->m_len = min(len, m->m_len - off);
573 		if (m->m_flags & M_EXT) {
574 			if (!deep) {
575 				n->m_data = m->m_data + off;
576 				n->m_ext = m->m_ext;
577 				MCLADDREFERENCE(m, n);
578 			} else {
579 				/*
580 				 * we are unsure about the way m was allocated.
581 				 * copy into multiple MCLBYTES cluster mbufs.
582 				 */
583 				MCLGET(n, wait);
584 				n->m_len = 0;
585 				n->m_len = M_TRAILINGSPACE(n);
586 				n->m_len = min(n->m_len, len);
587 				n->m_len = min(n->m_len, m->m_len - off);
588 				memcpy(mtod(n, void *), mtod(m, char *) + off,
589 				    (unsigned)n->m_len);
590 			}
591 		} else
592 			memcpy(mtod(n, void *), mtod(m, char *) + off,
593 			    (unsigned)n->m_len);
594 		if (len != M_COPYALL)
595 			len -= n->m_len;
596 		off += n->m_len;
597 #ifdef DIAGNOSTIC
598 		if (off > m->m_len)
599 			panic("m_copym0 overrun");
600 #endif
601 		if (off == m->m_len) {
602 			m = m->m_next;
603 			off = 0;
604 		}
605 		np = &n->m_next;
606 	}
607 	if (top == 0)
608 		MCFail++;
609 	return (top);
610 nospace:
611 	m_freem(top);
612 	MCFail++;
613 	return (NULL);
614 }
615 
616 /*
617  * Copy an entire packet, including header (which must be present).
618  * An optimization of the common case `m_copym(m, 0, M_COPYALL, how)'.
619  */
620 struct mbuf *
621 m_copypacket(struct mbuf *m, int how)
622 {
623 	struct mbuf *top, *n, *o;
624 
625 	MGET(n, how, m->m_type);
626 	top = n;
627 	if (!n)
628 		goto nospace;
629 
630 	MCLAIM(n, m->m_owner);
631 	M_COPY_PKTHDR(n, m);
632 	n->m_len = m->m_len;
633 	if (m->m_flags & M_EXT) {
634 		n->m_data = m->m_data;
635 		n->m_ext = m->m_ext;
636 		MCLADDREFERENCE(m, n);
637 	} else {
638 		memcpy(mtod(n, char *), mtod(m, char *), n->m_len);
639 	}
640 
641 	m = m->m_next;
642 	while (m) {
643 		MGET(o, how, m->m_type);
644 		if (!o)
645 			goto nospace;
646 
647 		MCLAIM(o, m->m_owner);
648 		n->m_next = o;
649 		n = n->m_next;
650 
651 		n->m_len = m->m_len;
652 		if (m->m_flags & M_EXT) {
653 			n->m_data = m->m_data;
654 			n->m_ext = m->m_ext;
655 			MCLADDREFERENCE(m, n);
656 		} else {
657 			memcpy(mtod(n, char *), mtod(m, char *), n->m_len);
658 		}
659 
660 		m = m->m_next;
661 	}
662 	return top;
663 nospace:
664 	m_freem(top);
665 	MCFail++;
666 	return NULL;
667 }
668 
669 /*
670  * Copy data from an mbuf chain starting "off" bytes from the beginning,
671  * continuing for "len" bytes, into the indicated buffer.
672  */
673 void
674 m_copydata(struct mbuf *m, int off, int len, void *vp)
675 {
676 	unsigned	count;
677 	void *		cp = vp;
678 
679 	if (off < 0 || len < 0)
680 		panic("m_copydata: off %d, len %d", off, len);
681 	while (off > 0) {
682 		if (m == NULL)
683 			panic("m_copydata: m == NULL, off %d", off);
684 		if (off < m->m_len)
685 			break;
686 		off -= m->m_len;
687 		m = m->m_next;
688 	}
689 	while (len > 0) {
690 		if (m == NULL)
691 			panic("m_copydata: m == NULL, len %d", len);
692 		count = min(m->m_len - off, len);
693 		memcpy(cp, mtod(m, char *) + off, count);
694 		len -= count;
695 		cp = (char *)cp + count;
696 		off = 0;
697 		m = m->m_next;
698 	}
699 }
700 
701 /*
702  * Concatenate mbuf chain n to m.
703  * n might be copied into m (when n->m_len is small), therefore data portion of
704  * n could be copied into an mbuf of different mbuf type.
705  * Any m_pkthdr is not updated.
706  */
707 void
708 m_cat(struct mbuf *m, struct mbuf *n)
709 {
710 
711 	while (m->m_next)
712 		m = m->m_next;
713 	while (n) {
714 		if (M_READONLY(m) || n->m_len > M_TRAILINGSPACE(m)) {
715 			/* just join the two chains */
716 			m->m_next = n;
717 			return;
718 		}
719 		/* splat the data from one into the other */
720 		memcpy(mtod(m, char *) + m->m_len, mtod(n, void *),
721 		    (u_int)n->m_len);
722 		m->m_len += n->m_len;
723 		n = m_free(n);
724 	}
725 }
726 
727 void
728 m_adj(struct mbuf *mp, int req_len)
729 {
730 	int len = req_len;
731 	struct mbuf *m;
732 	int count;
733 
734 	if ((m = mp) == NULL)
735 		return;
736 	if (len >= 0) {
737 		/*
738 		 * Trim from head.
739 		 */
740 		while (m != NULL && len > 0) {
741 			if (m->m_len <= len) {
742 				len -= m->m_len;
743 				m->m_len = 0;
744 				m = m->m_next;
745 			} else {
746 				m->m_len -= len;
747 				m->m_data += len;
748 				len = 0;
749 			}
750 		}
751 		m = mp;
752 		if (mp->m_flags & M_PKTHDR)
753 			m->m_pkthdr.len -= (req_len - len);
754 	} else {
755 		/*
756 		 * Trim from tail.  Scan the mbuf chain,
757 		 * calculating its length and finding the last mbuf.
758 		 * If the adjustment only affects this mbuf, then just
759 		 * adjust and return.  Otherwise, rescan and truncate
760 		 * after the remaining size.
761 		 */
762 		len = -len;
763 		count = 0;
764 		for (;;) {
765 			count += m->m_len;
766 			if (m->m_next == (struct mbuf *)0)
767 				break;
768 			m = m->m_next;
769 		}
770 		if (m->m_len >= len) {
771 			m->m_len -= len;
772 			if (mp->m_flags & M_PKTHDR)
773 				mp->m_pkthdr.len -= len;
774 			return;
775 		}
776 		count -= len;
777 		if (count < 0)
778 			count = 0;
779 		/*
780 		 * Correct length for chain is "count".
781 		 * Find the mbuf with last data, adjust its length,
782 		 * and toss data from remaining mbufs on chain.
783 		 */
784 		m = mp;
785 		if (m->m_flags & M_PKTHDR)
786 			m->m_pkthdr.len = count;
787 		for (; m; m = m->m_next) {
788 			if (m->m_len >= count) {
789 				m->m_len = count;
790 				break;
791 			}
792 			count -= m->m_len;
793 		}
794 		if (m)
795 			while (m->m_next)
796 				(m = m->m_next)->m_len = 0;
797 	}
798 }
799 
800 /*
801  * Rearrange an mbuf chain so that len bytes are contiguous
802  * and in the data area of an mbuf (so that mtod and dtom
803  * will work for a structure of size len).  Returns the resulting
804  * mbuf chain on success, frees it and returns null on failure.
805  * If there is room, it will add up to max_protohdr-len extra bytes to the
806  * contiguous region in an attempt to avoid being called next time.
807  */
808 int MPFail;
809 
810 struct mbuf *
811 m_pullup(struct mbuf *n, int len)
812 {
813 	struct mbuf *m;
814 	int count;
815 	int space;
816 
817 	/*
818 	 * If first mbuf has no cluster, and has room for len bytes
819 	 * without shifting current data, pullup into it,
820 	 * otherwise allocate a new mbuf to prepend to the chain.
821 	 */
822 	if ((n->m_flags & M_EXT) == 0 &&
823 	    n->m_data + len < &n->m_dat[MLEN] && n->m_next) {
824 		if (n->m_len >= len)
825 			return (n);
826 		m = n;
827 		n = n->m_next;
828 		len -= m->m_len;
829 	} else {
830 		if (len > MHLEN)
831 			goto bad;
832 		MGET(m, M_DONTWAIT, n->m_type);
833 		if (m == 0)
834 			goto bad;
835 		MCLAIM(m, n->m_owner);
836 		m->m_len = 0;
837 		if (n->m_flags & M_PKTHDR) {
838 			M_MOVE_PKTHDR(m, n);
839 		}
840 	}
841 	space = &m->m_dat[MLEN] - (m->m_data + m->m_len);
842 	do {
843 		count = min(min(max(len, max_protohdr), space), n->m_len);
844 		memcpy(mtod(m, char *) + m->m_len, mtod(n, void *),
845 		  (unsigned)count);
846 		len -= count;
847 		m->m_len += count;
848 		n->m_len -= count;
849 		space -= count;
850 		if (n->m_len)
851 			n->m_data += count;
852 		else
853 			n = m_free(n);
854 	} while (len > 0 && n);
855 	if (len > 0) {
856 		(void) m_free(m);
857 		goto bad;
858 	}
859 	m->m_next = n;
860 	return (m);
861 bad:
862 	m_freem(n);
863 	MPFail++;
864 	return (NULL);
865 }
866 
867 /*
868  * Like m_pullup(), except a new mbuf is always allocated, and we allow
869  * the amount of empty space before the data in the new mbuf to be specified
870  * (in the event that the caller expects to prepend later).
871  */
872 int MSFail;
873 
874 struct mbuf *
875 m_copyup(struct mbuf *n, int len, int dstoff)
876 {
877 	struct mbuf *m;
878 	int count, space;
879 
880 	if (len > (MHLEN - dstoff))
881 		goto bad;
882 	MGET(m, M_DONTWAIT, n->m_type);
883 	if (m == NULL)
884 		goto bad;
885 	MCLAIM(m, n->m_owner);
886 	m->m_len = 0;
887 	if (n->m_flags & M_PKTHDR) {
888 		M_MOVE_PKTHDR(m, n);
889 	}
890 	m->m_data += dstoff;
891 	space = &m->m_dat[MLEN] - (m->m_data + m->m_len);
892 	do {
893 		count = min(min(max(len, max_protohdr), space), n->m_len);
894 		memcpy(mtod(m, char *) + m->m_len, mtod(n, void *),
895 		    (unsigned)count);
896 		len -= count;
897 		m->m_len += count;
898 		n->m_len -= count;
899 		space -= count;
900 		if (n->m_len)
901 			n->m_data += count;
902 		else
903 			n = m_free(n);
904 	} while (len > 0 && n);
905 	if (len > 0) {
906 		(void) m_free(m);
907 		goto bad;
908 	}
909 	m->m_next = n;
910 	return (m);
911  bad:
912 	m_freem(n);
913 	MSFail++;
914 	return (NULL);
915 }
916 
917 /*
918  * Partition an mbuf chain in two pieces, returning the tail --
919  * all but the first len0 bytes.  In case of failure, it returns NULL and
920  * attempts to restore the chain to its original state.
921  */
922 struct mbuf *
923 m_split(struct mbuf *m0, int len0, int wait)
924 {
925 
926 	return m_split0(m0, len0, wait, 1);
927 }
928 
929 static struct mbuf *
930 m_split0(struct mbuf *m0, int len0, int wait, int copyhdr)
931 {
932 	struct mbuf *m, *n;
933 	unsigned len = len0, remain, len_save;
934 
935 	for (m = m0; m && len > m->m_len; m = m->m_next)
936 		len -= m->m_len;
937 	if (m == 0)
938 		return (NULL);
939 	remain = m->m_len - len;
940 	if (copyhdr && (m0->m_flags & M_PKTHDR)) {
941 		MGETHDR(n, wait, m0->m_type);
942 		if (n == 0)
943 			return (NULL);
944 		MCLAIM(n, m0->m_owner);
945 		n->m_pkthdr.rcvif = m0->m_pkthdr.rcvif;
946 		n->m_pkthdr.len = m0->m_pkthdr.len - len0;
947 		len_save = m0->m_pkthdr.len;
948 		m0->m_pkthdr.len = len0;
949 		if (m->m_flags & M_EXT)
950 			goto extpacket;
951 		if (remain > MHLEN) {
952 			/* m can't be the lead packet */
953 			MH_ALIGN(n, 0);
954 			n->m_next = m_split(m, len, wait);
955 			if (n->m_next == 0) {
956 				(void) m_free(n);
957 				m0->m_pkthdr.len = len_save;
958 				return (NULL);
959 			} else
960 				return (n);
961 		} else
962 			MH_ALIGN(n, remain);
963 	} else if (remain == 0) {
964 		n = m->m_next;
965 		m->m_next = 0;
966 		return (n);
967 	} else {
968 		MGET(n, wait, m->m_type);
969 		if (n == 0)
970 			return (NULL);
971 		MCLAIM(n, m->m_owner);
972 		M_ALIGN(n, remain);
973 	}
974 extpacket:
975 	if (m->m_flags & M_EXT) {
976 		n->m_ext = m->m_ext;
977 		MCLADDREFERENCE(m, n);
978 		n->m_data = m->m_data + len;
979 	} else {
980 		memcpy(mtod(n, void *), mtod(m, char *) + len, remain);
981 	}
982 	n->m_len = remain;
983 	m->m_len = len;
984 	n->m_next = m->m_next;
985 	m->m_next = 0;
986 	return (n);
987 }
988 /*
989  * Routine to copy from device local memory into mbufs.
990  */
991 struct mbuf *
992 m_devget(char *buf, int totlen, int off0, struct ifnet *ifp,
993     void (*copy)(const void *from, void *to, size_t len))
994 {
995 	struct mbuf *m;
996 	struct mbuf *top = 0, **mp = &top;
997 	int off = off0, len;
998 	char *cp;
999 	char *epkt;
1000 
1001 	cp = buf;
1002 	epkt = cp + totlen;
1003 	if (off) {
1004 		/*
1005 		 * If 'off' is non-zero, packet is trailer-encapsulated,
1006 		 * so we have to skip the type and length fields.
1007 		 */
1008 		cp += off + 2 * sizeof(uint16_t);
1009 		totlen -= 2 * sizeof(uint16_t);
1010 	}
1011 	MGETHDR(m, M_DONTWAIT, MT_DATA);
1012 	if (m == 0)
1013 		return (NULL);
1014 	m->m_pkthdr.rcvif = ifp;
1015 	m->m_pkthdr.len = totlen;
1016 	m->m_len = MHLEN;
1017 
1018 	while (totlen > 0) {
1019 		if (top) {
1020 			MGET(m, M_DONTWAIT, MT_DATA);
1021 			if (m == 0) {
1022 				m_freem(top);
1023 				return (NULL);
1024 			}
1025 			m->m_len = MLEN;
1026 		}
1027 		len = min(totlen, epkt - cp);
1028 		if (len >= MINCLSIZE) {
1029 			MCLGET(m, M_DONTWAIT);
1030 			if ((m->m_flags & M_EXT) == 0) {
1031 				m_free(m);
1032 				m_freem(top);
1033 				return (NULL);
1034 			}
1035 			m->m_len = len = min(len, MCLBYTES);
1036 		} else {
1037 			/*
1038 			 * Place initial small packet/header at end of mbuf.
1039 			 */
1040 			if (len < m->m_len) {
1041 				if (top == 0 && len + max_linkhdr <= m->m_len)
1042 					m->m_data += max_linkhdr;
1043 				m->m_len = len;
1044 			} else
1045 				len = m->m_len;
1046 		}
1047 		if (copy)
1048 			copy(cp, mtod(m, void *), (size_t)len);
1049 		else
1050 			memcpy(mtod(m, void *), cp, (size_t)len);
1051 		cp += len;
1052 		*mp = m;
1053 		mp = &m->m_next;
1054 		totlen -= len;
1055 		if (cp == epkt)
1056 			cp = buf;
1057 	}
1058 	return (top);
1059 }
1060 
1061 /*
1062  * Copy data from a buffer back into the indicated mbuf chain,
1063  * starting "off" bytes from the beginning, extending the mbuf
1064  * chain if necessary.
1065  */
1066 void
1067 m_copyback(struct mbuf *m0, int off, int len, const void *cp)
1068 {
1069 #if defined(DEBUG)
1070 	struct mbuf *origm = m0;
1071 	int error;
1072 #endif /* defined(DEBUG) */
1073 
1074 	if (m0 == NULL)
1075 		return;
1076 
1077 #if defined(DEBUG)
1078 	error =
1079 #endif /* defined(DEBUG) */
1080 	m_copyback0(&m0, off, len, cp,
1081 	    M_COPYBACK0_COPYBACK|M_COPYBACK0_EXTEND, M_DONTWAIT);
1082 
1083 #if defined(DEBUG)
1084 	if (error != 0 || (m0 != NULL && origm != m0))
1085 		panic("m_copyback");
1086 #endif /* defined(DEBUG) */
1087 }
1088 
1089 struct mbuf *
1090 m_copyback_cow(struct mbuf *m0, int off, int len, const void *cp, int how)
1091 {
1092 	int error;
1093 
1094 	/* don't support chain expansion */
1095 	KDASSERT(off + len <= m_length(m0));
1096 
1097 	error = m_copyback0(&m0, off, len, cp,
1098 	    M_COPYBACK0_COPYBACK|M_COPYBACK0_COW, how);
1099 	if (error) {
1100 		/*
1101 		 * no way to recover from partial success.
1102 		 * just free the chain.
1103 		 */
1104 		m_freem(m0);
1105 		return NULL;
1106 	}
1107 	return m0;
1108 }
1109 
1110 /*
1111  * m_makewritable: ensure the specified range writable.
1112  */
1113 int
1114 m_makewritable(struct mbuf **mp, int off, int len, int how)
1115 {
1116 	int error;
1117 #if defined(DEBUG)
1118 	struct mbuf *n;
1119 	int origlen, reslen;
1120 
1121 	origlen = m_length(*mp);
1122 #endif /* defined(DEBUG) */
1123 
1124 #if 0 /* M_COPYALL is large enough */
1125 	if (len == M_COPYALL)
1126 		len = m_length(*mp) - off; /* XXX */
1127 #endif
1128 
1129 	error = m_copyback0(mp, off, len, NULL,
1130 	    M_COPYBACK0_PRESERVE|M_COPYBACK0_COW, how);
1131 
1132 #if defined(DEBUG)
1133 	reslen = 0;
1134 	for (n = *mp; n; n = n->m_next)
1135 		reslen += n->m_len;
1136 	if (origlen != reslen)
1137 		panic("m_makewritable: length changed");
1138 	if (((*mp)->m_flags & M_PKTHDR) != 0 && reslen != (*mp)->m_pkthdr.len)
1139 		panic("m_makewritable: inconsist");
1140 #endif /* defined(DEBUG) */
1141 
1142 	return error;
1143 }
1144 
1145 int
1146 m_copyback0(struct mbuf **mp0, int off, int len, const void *vp, int flags,
1147     int how)
1148 {
1149 	int mlen;
1150 	struct mbuf *m, *n;
1151 	struct mbuf **mp;
1152 	int totlen = 0;
1153 	const char *cp = vp;
1154 
1155 	KASSERT(mp0 != NULL);
1156 	KASSERT(*mp0 != NULL);
1157 	KASSERT((flags & M_COPYBACK0_PRESERVE) == 0 || cp == NULL);
1158 	KASSERT((flags & M_COPYBACK0_COPYBACK) == 0 || cp != NULL);
1159 
1160 	/*
1161 	 * we don't bother to update "totlen" in the case of M_COPYBACK0_COW,
1162 	 * assuming that M_COPYBACK0_EXTEND and M_COPYBACK0_COW are exclusive.
1163 	 */
1164 
1165 	KASSERT((~flags & (M_COPYBACK0_EXTEND|M_COPYBACK0_COW)) != 0);
1166 
1167 	mp = mp0;
1168 	m = *mp;
1169 	while (off > (mlen = m->m_len)) {
1170 		off -= mlen;
1171 		totlen += mlen;
1172 		if (m->m_next == NULL) {
1173 			int tspace;
1174 extend:
1175 			if ((flags & M_COPYBACK0_EXTEND) == 0)
1176 				goto out;
1177 
1178 			/*
1179 			 * try to make some space at the end of "m".
1180 			 */
1181 
1182 			mlen = m->m_len;
1183 			if (off + len >= MINCLSIZE &&
1184 			    (m->m_flags & M_EXT) == 0 && m->m_len == 0) {
1185 				MCLGET(m, how);
1186 			}
1187 			tspace = M_TRAILINGSPACE(m);
1188 			if (tspace > 0) {
1189 				tspace = min(tspace, off + len);
1190 				KASSERT(tspace > 0);
1191 				memset(mtod(m, char *) + m->m_len, 0,
1192 				    min(off, tspace));
1193 				m->m_len += tspace;
1194 				off += mlen;
1195 				totlen -= mlen;
1196 				continue;
1197 			}
1198 
1199 			/*
1200 			 * need to allocate an mbuf.
1201 			 */
1202 
1203 			if (off + len >= MINCLSIZE) {
1204 				n = m_getcl(how, m->m_type, 0);
1205 			} else {
1206 				n = m_get(how, m->m_type);
1207 			}
1208 			if (n == NULL) {
1209 				goto out;
1210 			}
1211 			n->m_len = 0;
1212 			n->m_len = min(M_TRAILINGSPACE(n), off + len);
1213 			memset(mtod(n, char *), 0, min(n->m_len, off));
1214 			m->m_next = n;
1215 		}
1216 		mp = &m->m_next;
1217 		m = m->m_next;
1218 	}
1219 	while (len > 0) {
1220 		mlen = m->m_len - off;
1221 		if (mlen != 0 && M_READONLY(m)) {
1222 			char *datap;
1223 			int eatlen;
1224 
1225 			/*
1226 			 * this mbuf is read-only.
1227 			 * allocate a new writable mbuf and try again.
1228 			 */
1229 
1230 #if defined(DIAGNOSTIC)
1231 			if ((flags & M_COPYBACK0_COW) == 0)
1232 				panic("m_copyback0: read-only");
1233 #endif /* defined(DIAGNOSTIC) */
1234 
1235 			/*
1236 			 * if we're going to write into the middle of
1237 			 * a mbuf, split it first.
1238 			 */
1239 			if (off > 0 && len < mlen) {
1240 				n = m_split0(m, off, how, 0);
1241 				if (n == NULL)
1242 					goto enobufs;
1243 				m->m_next = n;
1244 				mp = &m->m_next;
1245 				m = n;
1246 				off = 0;
1247 				continue;
1248 			}
1249 
1250 			/*
1251 			 * XXX TODO coalesce into the trailingspace of
1252 			 * the previous mbuf when possible.
1253 			 */
1254 
1255 			/*
1256 			 * allocate a new mbuf.  copy packet header if needed.
1257 			 */
1258 			MGET(n, how, m->m_type);
1259 			if (n == NULL)
1260 				goto enobufs;
1261 			MCLAIM(n, m->m_owner);
1262 			if (off == 0 && (m->m_flags & M_PKTHDR) != 0) {
1263 				M_MOVE_PKTHDR(n, m);
1264 				n->m_len = MHLEN;
1265 			} else {
1266 				if (len >= MINCLSIZE)
1267 					MCLGET(n, M_DONTWAIT);
1268 				n->m_len =
1269 				    (n->m_flags & M_EXT) ? MCLBYTES : MLEN;
1270 			}
1271 			if (n->m_len > len)
1272 				n->m_len = len;
1273 
1274 			/*
1275 			 * free the region which has been overwritten.
1276 			 * copying data from old mbufs if requested.
1277 			 */
1278 			if (flags & M_COPYBACK0_PRESERVE)
1279 				datap = mtod(n, char *);
1280 			else
1281 				datap = NULL;
1282 			eatlen = n->m_len;
1283 			KDASSERT(off == 0 || eatlen >= mlen);
1284 			if (off > 0) {
1285 				KDASSERT(len >= mlen);
1286 				m->m_len = off;
1287 				m->m_next = n;
1288 				if (datap) {
1289 					m_copydata(m, off, mlen, datap);
1290 					datap += mlen;
1291 				}
1292 				eatlen -= mlen;
1293 				mp = &m->m_next;
1294 				m = m->m_next;
1295 			}
1296 			while (m != NULL && M_READONLY(m) &&
1297 			    n->m_type == m->m_type && eatlen > 0) {
1298 				mlen = min(eatlen, m->m_len);
1299 				if (datap) {
1300 					m_copydata(m, 0, mlen, datap);
1301 					datap += mlen;
1302 				}
1303 				m->m_data += mlen;
1304 				m->m_len -= mlen;
1305 				eatlen -= mlen;
1306 				if (m->m_len == 0)
1307 					*mp = m = m_free(m);
1308 			}
1309 			if (eatlen > 0)
1310 				n->m_len -= eatlen;
1311 			n->m_next = m;
1312 			*mp = m = n;
1313 			continue;
1314 		}
1315 		mlen = min(mlen, len);
1316 		if (flags & M_COPYBACK0_COPYBACK) {
1317 			memcpy(mtod(m, char *) + off, cp, (unsigned)mlen);
1318 			cp += mlen;
1319 		}
1320 		len -= mlen;
1321 		mlen += off;
1322 		off = 0;
1323 		totlen += mlen;
1324 		if (len == 0)
1325 			break;
1326 		if (m->m_next == NULL) {
1327 			goto extend;
1328 		}
1329 		mp = &m->m_next;
1330 		m = m->m_next;
1331 	}
1332 out:	if (((m = *mp0)->m_flags & M_PKTHDR) && (m->m_pkthdr.len < totlen)) {
1333 		KASSERT((flags & M_COPYBACK0_EXTEND) != 0);
1334 		m->m_pkthdr.len = totlen;
1335 	}
1336 
1337 	return 0;
1338 
1339 enobufs:
1340 	return ENOBUFS;
1341 }
1342 
1343 void
1344 m_move_pkthdr(struct mbuf *to, struct mbuf *from)
1345 {
1346 
1347 	KASSERT((to->m_flags & M_EXT) == 0);
1348 	KASSERT((to->m_flags & M_PKTHDR) == 0 || m_tag_first(to) == NULL);
1349 	KASSERT((from->m_flags & M_PKTHDR) != 0);
1350 
1351 	to->m_pkthdr = from->m_pkthdr;
1352 	to->m_flags = from->m_flags & M_COPYFLAGS;
1353 	to->m_data = to->m_pktdat;
1354 
1355 	from->m_flags &= ~M_PKTHDR;
1356 }
1357 
1358 /*
1359  * Apply function f to the data in an mbuf chain starting "off" bytes from the
1360  * beginning, continuing for "len" bytes.
1361  */
1362 int
1363 m_apply(struct mbuf *m, int off, int len,
1364     int (*f)(void *, void *, unsigned int), void *arg)
1365 {
1366 	unsigned int count;
1367 	int rval;
1368 
1369 	KASSERT(len >= 0);
1370 	KASSERT(off >= 0);
1371 
1372 	while (off > 0) {
1373 		KASSERT(m != NULL);
1374 		if (off < m->m_len)
1375 			break;
1376 		off -= m->m_len;
1377 		m = m->m_next;
1378 	}
1379 	while (len > 0) {
1380 		KASSERT(m != NULL);
1381 		count = min(m->m_len - off, len);
1382 
1383 		rval = (*f)(arg, mtod(m, char *) + off, count);
1384 		if (rval)
1385 			return (rval);
1386 
1387 		len -= count;
1388 		off = 0;
1389 		m = m->m_next;
1390 	}
1391 
1392 	return (0);
1393 }
1394 
1395 /*
1396  * Return a pointer to mbuf/offset of location in mbuf chain.
1397  */
1398 struct mbuf *
1399 m_getptr(struct mbuf *m, int loc, int *off)
1400 {
1401 
1402 	while (loc >= 0) {
1403 		/* Normal end of search */
1404 		if (m->m_len > loc) {
1405 	    		*off = loc;
1406 	    		return (m);
1407 		} else {
1408 	    		loc -= m->m_len;
1409 
1410 	    		if (m->m_next == NULL) {
1411 				if (loc == 0) {
1412  					/* Point at the end of valid data */
1413 		    			*off = m->m_len;
1414 		    			return (m);
1415 				} else
1416 		  			return (NULL);
1417 	    		} else
1418 	      			m = m->m_next;
1419 		}
1420     	}
1421 
1422 	return (NULL);
1423 }
1424 
1425 #if defined(DDB)
1426 void
1427 m_print(const struct mbuf *m, const char *modif, void (*pr)(const char *, ...))
1428 {
1429 	char ch;
1430 	bool opt_c = false;
1431 	char buf[512];
1432 
1433 	while ((ch = *(modif++)) != '\0') {
1434 		switch (ch) {
1435 		case 'c':
1436 			opt_c = true;
1437 			break;
1438 		}
1439 	}
1440 
1441 nextchain:
1442 	(*pr)("MBUF %p\n", m);
1443 	bitmask_snprintf((u_int)m->m_flags, M_FLAGS_BITS, buf, sizeof(buf));
1444 	(*pr)("  data=%p, len=%d, type=%d, flags=0x%s\n",
1445 	    m->m_data, m->m_len, m->m_type, buf);
1446 	(*pr)("  owner=%p, next=%p, nextpkt=%p\n", m->m_owner, m->m_next,
1447 	    m->m_nextpkt);
1448 	(*pr)("  leadingspace=%u, trailingspace=%u, readonly=%u\n",
1449 	    (int)M_LEADINGSPACE(m), (int)M_TRAILINGSPACE(m),
1450 	    (int)M_READONLY(m));
1451 	if ((m->m_flags & M_PKTHDR) != 0) {
1452 		bitmask_snprintf(m->m_pkthdr.csum_flags, M_CSUM_BITS, buf,
1453 		    sizeof(buf));
1454 		(*pr)("  pktlen=%d, rcvif=%p, csum_flags=0x%s, csum_data=0x%"
1455 		    PRIx32 ", segsz=%u\n",
1456 		    m->m_pkthdr.len, m->m_pkthdr.rcvif,
1457 		    buf, m->m_pkthdr.csum_data, m->m_pkthdr.segsz);
1458 	}
1459 	if ((m->m_flags & M_EXT)) {
1460 		(*pr)("  shared=%u, ext_buf=%p, ext_size=%zd, "
1461 		    "ext_free=%p, ext_arg=%p\n",
1462 		    (int)MCLISREFERENCED(m),
1463 		    m->m_ext.ext_buf, m->m_ext.ext_size,
1464 		    m->m_ext.ext_free, m->m_ext.ext_arg);
1465 	}
1466 	if ((~m->m_flags & (M_EXT|M_EXT_PAGES)) == 0) {
1467 		vaddr_t sva = (vaddr_t)m->m_ext.ext_buf;
1468 		vaddr_t eva = sva + m->m_ext.ext_size;
1469 		int n = (round_page(eva) - trunc_page(sva)) >> PAGE_SHIFT;
1470 		int i;
1471 
1472 		(*pr)("  pages:");
1473 		for (i = 0; i < n; i ++) {
1474 			(*pr)(" %p", m->m_ext.ext_pgs[i]);
1475 		}
1476 		(*pr)("\n");
1477 	}
1478 
1479 	if (opt_c) {
1480 		m = m->m_next;
1481 		if (m != NULL) {
1482 			goto nextchain;
1483 		}
1484 	}
1485 }
1486 #endif /* defined(DDB) */
1487