1 /* $OpenBSD: uipc_mbuf.c,v 1.85 2007/07/20 09:59:19 claudio Exp $ */
2 /* $NetBSD: uipc_mbuf.c,v 1.15.4.1 1996/06/13 17:11:44 cgd Exp $ */
3
4 /*
5 * Copyright (c) 1982, 1986, 1988, 1991, 1993
6 * The Regents of the University of California. All rights reserved.
7 *
8 * Redistribution and use in source and binary forms, with or without
9 * modification, are permitted provided that the following conditions
10 * are met:
11 * 1. Redistributions of source code must retain the above copyright
12 * notice, this list of conditions and the following disclaimer.
13 * 2. Redistributions in binary form must reproduce the above copyright
14 * notice, this list of conditions and the following disclaimer in the
15 * documentation and/or other materials provided with the distribution.
16 * 3. Neither the name of the University nor the names of its contributors
17 * may be used to endorse or promote products derived from this software
18 * without specific prior written permission.
19 *
20 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
21 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
22 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
23 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
24 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
25 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
26 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
27 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
28 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
29 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
30 * SUCH DAMAGE.
31 *
32 * @(#)uipc_mbuf.c 8.2 (Berkeley) 1/4/94
33 */
34
35 /*
36 * @(#)COPYRIGHT 1.1 (NRL) 17 January 1995
37 *
38 * NRL grants permission for redistribution and use in source and binary
39 * forms, with or without modification, of the software and documentation
40 * created at NRL provided that the following conditions are met:
41 *
42 * 1. Redistributions of source code must retain the above copyright
43 * notice, this list of conditions and the following disclaimer.
44 * 2. Redistributions in binary form must reproduce the above copyright
45 * notice, this list of conditions and the following disclaimer in the
46 * documentation and/or other materials provided with the distribution.
47 * 3. All advertising materials mentioning features or use of this software
48 * must display the following acknowledgements:
49 * This product includes software developed by the University of
50 * California, Berkeley and its contributors.
51 * This product includes software developed at the Information
52 * Technology Division, US Naval Research Laboratory.
53 * 4. Neither the name of the NRL nor the names of its contributors
54 * may be used to endorse or promote products derived from this software
55 * without specific prior written permission.
56 *
57 * THE SOFTWARE PROVIDED BY NRL IS PROVIDED BY NRL AND CONTRIBUTORS ``AS
58 * IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
59 * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A
60 * PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL NRL OR
61 * CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL,
62 * EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO,
63 * PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR
64 * PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
65 * LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
66 * NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
67 * SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
68 *
69 * The views and conclusions contained in the software and documentation
70 * are those of the authors and should not be interpreted as representing
71 * official policies, either expressed or implied, of the US Naval
72 * Research Laboratory (NRL).
73 */
74
75 #include <sys/param.h>
76 #include <sys/systm.h>
77 #include <sys/proc.h>
78 #include <sys/malloc.h>
79 #define MBTYPES
80 #include <sys/mbuf.h>
81 #include <sys/kernel.h>
82 #include <sys/syslog.h>
83 #include <sys/domain.h>
84 #include <sys/protosw.h>
85 #include <sys/pool.h>
86
87 #include <machine/cpu.h>
88
89 #include <uvm/uvm_extern.h>
90
91 struct mbstat mbstat; /* mbuf stats */
92 struct pool mbpool; /* mbuf pool */
93 struct pool mclpool; /* mbuf cluster pool */
94
95 int max_linkhdr; /* largest link-level header */
96 int max_protohdr; /* largest protocol header */
97 int max_hdr; /* largest link+protocol header */
98 int max_datalen; /* MHLEN - max_hdr */
99
100 struct mbuf *m_copym0(struct mbuf *, int, int, int, int);
101 void nmbclust_update(void);
102
103
104 const char *mclpool_warnmsg =
105 "WARNING: mclpool limit reached; increase kern.maxclusters";
106
107 /*
108 * Initialize the mbuf allocator.
109 */
110 void
111 mbinit(void)
112 {
113 pool_init(&mbpool, MSIZE, 0, 0, 0, "mbpl", NULL);
114 pool_init(&mclpool, MCLBYTES, 0, 0, 0, "mclpl", NULL);
115
116 nmbclust_update();
117
118 /*
119 * Set a low water mark for both mbufs and clusters. This should
120 * help ensure that they can be allocated in a memory starvation
121 * situation. This is important for e.g. diskless systems which
122 * must allocate mbufs in order for the pagedaemon to clean pages.
123 */
124 pool_setlowat(&mbpool, mblowat);
125 pool_setlowat(&mclpool, mcllowat);
126 }
127
128 void
129 nmbclust_update(void)
130 {
131 /*
132 * Set the hard limit on the mclpool to the number of
133 * mbuf clusters the kernel is to support. Log the limit
134 * reached message max once a minute.
135 */
136 (void)pool_sethardlimit(&mclpool, nmbclust, mclpool_warnmsg, 60);
137 pool_sethiwat(&mbpool, nmbclust);
138 }
139
140 void
141 m_reclaim(void *arg, int flags)
142 {
143 struct domain *dp;
144 struct protosw *pr;
145 int s = splvm();
146
147 for (dp = domains; dp; dp = dp->dom_next)
148 for (pr = dp->dom_protosw; pr < dp->dom_protoswNPROTOSW; pr++)
149 if (pr->pr_drain)
150 (*pr->pr_drain)();
151 splx(s);
152 mbstat.m_drain++;
153 }
154
155 /*
156 * Space allocation routines.
157 */
158 struct mbuf *
159 m_get(int nowait, int type)
160 {
161 struct mbuf *m;
162 int s;
163
164 s = splvm();
165 m = pool_get(&mbpool, nowait == M_WAIT ? PR_WAITOK|PR_LIMITFAIL : 0);
166 if (m) {
167 m->m_type = type;
168 mbstat.m_mtypes[type]++;
169 m->m_next = (struct mbuf *)NULL;
170 m->m_nextpkt = (struct mbuf *)NULL;
171 m->m_data = m->m_dat;
172 m->m_flags = 0;
173 }
174 splx(s);
175 return (m);
176 }
177
178 struct mbuf *
179 m_gethdr(int nowait, int type)
180 {
181 struct mbuf *m;
182 int s;
183
184 s = splvm();
185 m = pool_get(&mbpool, nowait == M_WAIT ? PR_WAITOK|PR_LIMITFAIL : 0);
186 if (m) {
187 m->m_type = type;
188 mbstat.m_mtypes[type]++;
189 m->m_next = (struct mbuf *)NULL;
190 m->m_nextpkt = (struct mbuf *)NULL;
191 m->m_data = m->m_pktdat;
192 m->m_flags = M_PKTHDR;
193 m->m_pkthdr.rcvif = NULL;
194 SLIST_INIT(&m->m_pkthdr.tags);
195 m->m_pkthdr.csum_flags = 0;
196 m->m_pkthdr.pf.hdr = NULL;
197 m->m_pkthdr.pf.rtableid = 0;
198 m->m_pkthdr.pf.qid = 0;
199 m->m_pkthdr.pf.tag = 0;
200 m->m_pkthdr.pf.flags = 0;
201 m->m_pkthdr.pf.routed = 0;
202 }
203 splx(s);
204 return (m);
205 }
206
207 struct mbuf *
208 m_getclr(int nowait, int type)
209 {
210 struct mbuf *m;
211
212 MGET(m, nowait, type);
213 if (m == NULL)
214 return (NULL);
215 memset(mtod(m, caddr_t), 0, MLEN);
216 return (m);
217 }
218
219 void
220 m_clget(struct mbuf *m, int how)
221 {
222 int s;
223
224 s = splvm();
225 m->m_ext.ext_buf =
226 pool_get(&mclpool, how == M_WAIT ? (PR_WAITOK|PR_LIMITFAIL) : 0);
227 splx(s);
228 if (m->m_ext.ext_buf != NULL) {
229 m->m_data = m->m_ext.ext_buf;
230 m->m_flags |= M_EXT|M_CLUSTER;
231 m->m_ext.ext_size = MCLBYTES;
232 m->m_ext.ext_free = NULL;
233 m->m_ext.ext_arg = NULL;
234 MCLINITREFERENCE(m);
235 }
236 }
237
238 struct mbuf *
239 m_free(struct mbuf *m)
240 {
241 struct mbuf *n;
242 int s;
243
244 s = splvm();
245 mbstat.m_mtypes[m->m_type]--;
246 if (m->m_flags & M_PKTHDR)
247 m_tag_delete_chain(m);
248 if (m->m_flags & M_EXT) {
249 if (MCLISREFERENCED(m))
250 _MCLDEREFERENCE(m);
251 else if (m->m_flags & M_CLUSTER)
252 pool_put(&mclpool, m->m_ext.ext_buf);
253 else if (m->m_ext.ext_free)
254 (*(m->m_ext.ext_free))(m->m_ext.ext_buf,
255 m->m_ext.ext_size, m->m_ext.ext_arg);
256 else
257 free(m->m_ext.ext_buf,m->m_ext.ext_type);
258 m->m_flags &= ~(M_CLUSTER|M_EXT);
259 m->m_ext.ext_size = 0;
260 }
261 n = m->m_next;
262 pool_put(&mbpool, m);
263 splx(s);
264
265 return (n);
266 }
267
268 void
269 m_freem(struct mbuf *m)
270 {
271 struct mbuf *n;
272
273 if (m == NULL)
274 return;
275 do {
276 MFREE(m, n);
277 } while ((m = n) != NULL);
278 }
279
280 /*
281 * Mbuffer utility routines.
282 */
283
284 /*
285 * Lesser-used path for M_PREPEND:
286 * allocate new mbuf to prepend to chain,
287 * copy junk along.
288 */
289 struct mbuf *
290 m_prepend(struct mbuf *m, int len, int how)
291 {
292 struct mbuf *mn;
293
294 if (len > MHLEN)
295 panic("mbuf prepend length too big");
296
297 MGET(mn, how, m->m_type);
298 if (mn == NULL) {
299 m_freem(m);
300 return (NULL);
301 }
302 if (m->m_flags & M_PKTHDR)
303 M_MOVE_PKTHDR(mn, m);
304 mn->m_next = m;
305 m = mn;
306 MH_ALIGN(m, len);
307 m->m_len = len;
308 return (m);
309 }
310
311 /*
312 * Make a copy of an mbuf chain starting "off" bytes from the beginning,
313 * continuing for "len" bytes. If len is M_COPYALL, copy to end of mbuf.
314 * The wait parameter is a choice of M_WAIT/M_DONTWAIT from caller.
315 */
316 int MCFail;
317
318 struct mbuf *
319 m_copym(struct mbuf *m, int off, int len, int wait)
320 {
321 return m_copym0(m, off, len, wait, 0); /* shallow copy on M_EXT */
322 }
323
324 /*
325 * m_copym2() is like m_copym(), except it COPIES cluster mbufs, instead
326 * of merely bumping the reference count.
327 */
328 struct mbuf *
329 m_copym2(struct mbuf *m, int off, int len, int wait)
330 {
331 return m_copym0(m, off, len, wait, 1); /* deep copy */
332 }
333
334 struct mbuf *
335 m_copym0(struct mbuf *m, int off, int len, int wait, int deep)
336 {
337 struct mbuf *n, **np;
338 struct mbuf *top;
339 int copyhdr = 0;
340
341 if (off < 0 || len < 0)
342 panic("m_copym0: off %d, len %d", off, len);
343 if (off == 0 && m->m_flags & M_PKTHDR)
344 copyhdr = 1;
345 while (off > 0) {
346 if (m == NULL)
347 panic("m_copym0: null mbuf");
348 if (off < m->m_len)
349 break;
350 off -= m->m_len;
351 m = m->m_next;
352 }
353 np = ⊤
354 top = NULL;
355 while (len > 0) {
356 if (m == NULL) {
357 if (len != M_COPYALL)
358 panic("m_copym0: m == NULL and not COPYALL");
359 break;
360 }
361 MGET(n, wait, m->m_type);
362 *np = n;
363 if (n == NULL)
364 goto nospace;
365 if (copyhdr) {
366 M_DUP_PKTHDR(n, m);
367 if (len != M_COPYALL)
368 n->m_pkthdr.len = len;
369 copyhdr = 0;
370 }
371 n->m_len = min(len, m->m_len - off);
372 if (m->m_flags & M_EXT) {
373 if (!deep) {
374 n->m_data = m->m_data + off;
375 n->m_ext = m->m_ext;
376 MCLADDREFERENCE(m, n);
377 } else {
378 /*
379 * we are unsure about the way m was allocated.
380 * copy into multiple MCLBYTES cluster mbufs.
381 */
382 MCLGET(n, wait);
383 n->m_len = 0;
384 n->m_len = M_TRAILINGSPACE(n);
385 n->m_len = min(n->m_len, len);
386 n->m_len = min(n->m_len, m->m_len - off);
387 memcpy(mtod(n, caddr_t), mtod(m, caddr_t) + off,
388 (unsigned)n->m_len);
389 }
390 } else
391 memcpy(mtod(n, caddr_t), mtod(m, caddr_t) + off,
392 (unsigned)n->m_len);
393 if (len != M_COPYALL)
394 len -= n->m_len;
395 off += n->m_len;
396 #ifdef DIAGNOSTIC
397 if (off > m->m_len)
398 panic("m_copym0 overrun");
399 #endif
400 if (off == m->m_len) {
401 m = m->m_next;
402 off = 0;
403 }
404 np = &n->m_next;
405 }
406 if (top == NULL)
407 MCFail++;
408 return (top);
409 nospace:
410 m_freem(top);
411 MCFail++;
412 return (NULL);
413 }
414
415 /*
416 * Copy data from an mbuf chain starting "off" bytes from the beginning,
417 * continuing for "len" bytes, into the indicated buffer.
418 */
419 void
420 m_copydata(struct mbuf *m, int off, int len, caddr_t cp)
421 {
422 unsigned count;
423
424 if (off < 0)
425 panic("m_copydata: off %d < 0", off);
426 if (len < 0)
427 panic("m_copydata: len %d < 0", len);
428 while (off > 0) {
429 if (m == NULL)
430 panic("m_copydata: null mbuf in skip");
431 if (off < m->m_len)
432 break;
433 off -= m->m_len;
434 m = m->m_next;
435 }
436 while (len > 0) {
437 if (m == NULL)
438 panic("m_copydata: null mbuf");
439 count = min(m->m_len - off, len);
440 bcopy(mtod(m, caddr_t) + off, cp, count);
441 len -= count;
442 cp += count;
443 off = 0;
444 m = m->m_next;
445 }
446 }
447
448 /*
449 * Copy data from a buffer back into the indicated mbuf chain,
450 * starting "off" bytes from the beginning, extending the mbuf
451 * chain if necessary. The mbuf needs to be properly initialized
452 * including the setting of m_len.
453 */
454 void
455 m_copyback(struct mbuf *m0, int off, int len, const void *_cp)
456 {
457 int mlen;
458 struct mbuf *m = m0, *n;
459 int totlen = 0;
460 caddr_t cp = (caddr_t)_cp;
461
462 if (m0 == NULL)
463 return;
464 while (off > (mlen = m->m_len)) {
465 off -= mlen;
466 totlen += mlen;
467 if (m->m_next == NULL) {
468 n = m_getclr(M_DONTWAIT, m->m_type);
469 if (n == NULL)
470 goto out;
471 n->m_len = min(MLEN, len + off);
472 m->m_next = n;
473 }
474 m = m->m_next;
475 }
476 while (len > 0) {
477 mlen = min (m->m_len - off, len);
478 bcopy(cp, off + mtod(m, caddr_t), (unsigned)mlen);
479 cp += mlen;
480 len -= mlen;
481 mlen += off;
482 off = 0;
483 totlen += mlen;
484 if (len == 0)
485 break;
486 if (m->m_next == NULL) {
487 n = m_get(M_DONTWAIT, m->m_type);
488 if (n == NULL)
489 break;
490 n->m_len = min(MLEN, len);
491 m->m_next = n;
492 }
493 m = m->m_next;
494 }
495 out: if (((m = m0)->m_flags & M_PKTHDR) && (m->m_pkthdr.len < totlen))
496 m->m_pkthdr.len = totlen;
497 }
498
499 /*
500 * Concatenate mbuf chain n to m.
501 * n might be copied into m (when n->m_len is small), therefore data portion of
502 * n could be copied into an mbuf of different mbuf type.
503 * Therefore both chains should be of the same type (e.g. MT_DATA).
504 * Any m_pkthdr is not updated.
505 */
506 void
507 m_cat(struct mbuf *m, struct mbuf *n)
508 {
509 while (m->m_next)
510 m = m->m_next;
511 while (n) {
512 if (M_READONLY(m) || n->m_len > M_TRAILINGSPACE(m)) {
513 /* just join the two chains */
514 m->m_next = n;
515 return;
516 }
517 /* splat the data from one into the other */
518 bcopy(mtod(n, caddr_t), mtod(m, caddr_t) + m->m_len,
519 (u_int)n->m_len);
520 m->m_len += n->m_len;
521 n = m_free(n);
522 }
523 }
524
525 void
526 m_adj(struct mbuf *mp, int req_len)
527 {
528 int len = req_len;
529 struct mbuf *m;
530 int count;
531
532 if ((m = mp) == NULL)
533 return;
534 if (len >= 0) {
535 /*
536 * Trim from head.
537 */
538 while (m != NULL && len > 0) {
539 if (m->m_len <= len) {
540 len -= m->m_len;
541 m->m_len = 0;
542 m = m->m_next;
543 } else {
544 m->m_len -= len;
545 m->m_data += len;
546 len = 0;
547 }
548 }
549 m = mp;
550 if (mp->m_flags & M_PKTHDR)
551 m->m_pkthdr.len -= (req_len - len);
552 } else {
553 /*
554 * Trim from tail. Scan the mbuf chain,
555 * calculating its length and finding the last mbuf.
556 * If the adjustment only affects this mbuf, then just
557 * adjust and return. Otherwise, rescan and truncate
558 * after the remaining size.
559 */
560 len = -len;
561 count = 0;
562 for (;;) {
563 count += m->m_len;
564 if (m->m_next == NULL)
565 break;
566 m = m->m_next;
567 }
568 if (m->m_len >= len) {
569 m->m_len -= len;
570 if (mp->m_flags & M_PKTHDR)
571 mp->m_pkthdr.len -= len;
572 return;
573 }
574 count -= len;
575 if (count < 0)
576 count = 0;
577 /*
578 * Correct length for chain is "count".
579 * Find the mbuf with last data, adjust its length,
580 * and toss data from remaining mbufs on chain.
581 */
582 m = mp;
583 if (m->m_flags & M_PKTHDR)
584 m->m_pkthdr.len = count;
585 for (; m; m = m->m_next) {
586 if (m->m_len >= count) {
587 m->m_len = count;
588 break;
589 }
590 count -= m->m_len;
591 }
592 while ((m = m->m_next) != NULL)
593 m->m_len = 0;
594 }
595 }
596
597 /*
598 * Rearange an mbuf chain so that len bytes are contiguous
599 * and in the data area of an mbuf (so that mtod and dtom
600 * will work for a structure of size len). Returns the resulting
601 * mbuf chain on success, frees it and returns null on failure.
602 * If there is room, it will add up to max_protohdr-len extra bytes to the
603 * contiguous region in an attempt to avoid being called next time.
604 */
605 int MPFail;
606
607 struct mbuf *
608 m_pullup(struct mbuf *n, int len)
609 {
610 struct mbuf *m;
611 int count;
612 int space;
613
614 /*
615 * If first mbuf has no cluster, and has room for len bytes
616 * without shifting current data, pullup into it,
617 * otherwise allocate a new mbuf to prepend to the chain.
618 */
619 if ((n->m_flags & M_EXT) == 0 &&
620 n->m_data + len < &n->m_dat[MLEN] && n->m_next) {
621 if (n->m_len >= len)
622 return (n);
623 m = n;
624 n = n->m_next;
625 len -= m->m_len;
626 } else {
627 if (len > MHLEN)
628 goto bad;
629 MGET(m, M_DONTWAIT, n->m_type);
630 if (m == NULL)
631 goto bad;
632 m->m_len = 0;
633 if (n->m_flags & M_PKTHDR)
634 M_MOVE_PKTHDR(m, n);
635 }
636 space = &m->m_dat[MLEN] - (m->m_data + m->m_len);
637 do {
638 count = min(min(max(len, max_protohdr), space), n->m_len);
639 bcopy(mtod(n, caddr_t), mtod(m, caddr_t) + m->m_len,
640 (unsigned)count);
641 len -= count;
642 m->m_len += count;
643 n->m_len -= count;
644 space -= count;
645 if (n->m_len)
646 n->m_data += count;
647 else
648 n = m_free(n);
649 } while (len > 0 && n);
650 if (len > 0) {
651 (void)m_free(m);
652 goto bad;
653 }
654 m->m_next = n;
655 return (m);
656 bad:
657 m_freem(n);
658 MPFail++;
659 return (NULL);
660 }
661
662 /*
663 * m_pullup2() works like m_pullup, save that len can be <= MCLBYTES.
664 * m_pullup2() only works on values of len such that MHLEN < len <= MCLBYTES,
665 * it calls m_pullup() for values <= MHLEN. It also only coagulates the
666 * reqested number of bytes. (For those of us who expect unwieldly option
667 * headers.
668 *
669 * KEBE SAYS: Remember that dtom() calls with data in clusters does not work!
670 */
671 struct mbuf *
672 m_pullup2(struct mbuf *n, int len)
673 {
674 struct mbuf *m;
675 int count;
676
677 if (len <= MHLEN)
678 return m_pullup(n, len);
679 if ((n->m_flags & M_EXT) != 0 &&
680 n->m_data + len < &n->m_data[MCLBYTES] && n->m_next) {
681 if (n->m_len >= len)
682 return (n);
683 m = n;
684 n = n->m_next;
685 len -= m->m_len;
686 } else {
687 if (len > MCLBYTES)
688 goto bad;
689 MGET(m, M_DONTWAIT, n->m_type);
690 if (m == NULL)
691 goto bad;
692 MCLGET(m, M_DONTWAIT);
693 if ((m->m_flags & M_EXT) == 0)
694 goto bad;
695 m->m_len = 0;
696 if (n->m_flags & M_PKTHDR) {
697 /* Too many adverse side effects. */
698 /* M_MOVE_PKTHDR(m, n); */
699 m->m_flags = (n->m_flags & M_COPYFLAGS) |
700 M_EXT | M_CLUSTER;
701 M_MOVE_HDR(m, n);
702 /* n->m_data is cool. */
703 }
704 }
705
706 do {
707 count = min(len, n->m_len);
708 bcopy(mtod(n, caddr_t), mtod(m, caddr_t) + m->m_len,
709 (unsigned)count);
710 len -= count;
711 m->m_len += count;
712 n->m_len -= count;
713 if (n->m_len)
714 n->m_data += count;
715 else
716 n = m_free(n);
717 } while (len > 0 && n);
718 if (len > 0) {
719 (void)m_free(m);
720 goto bad;
721 }
722 m->m_next = n;
723
724 return (m);
725 bad:
726 m_freem(n);
727 MPFail++;
728 return (NULL);
729 }
730
731 /*
732 * Return a pointer to mbuf/offset of location in mbuf chain.
733 */
734 struct mbuf *
735 m_getptr(struct mbuf *m, int loc, int *off)
736 {
737 while (loc >= 0) {
738 /* Normal end of search */
739 if (m->m_len > loc) {
740 *off = loc;
741 return (m);
742 }
743 else {
744 loc -= m->m_len;
745
746 if (m->m_next == NULL) {
747 if (loc == 0) {
748 /* Point at the end of valid data */
749 *off = m->m_len;
750 return (m);
751 }
752 else
753 return (NULL);
754 } else
755 m = m->m_next;
756 }
757 }
758
759 return (NULL);
760 }
761
762 /*
763 * Inject a new mbuf chain of length siz in mbuf chain m0 at
764 * position len0. Returns a pointer to the first injected mbuf, or
765 * NULL on failure (m0 is left undisturbed). Note that if there is
766 * enough space for an object of size siz in the appropriate position,
767 * no memory will be allocated. Also, there will be no data movement in
768 * the first len0 bytes (pointers to that will remain valid).
769 *
770 * XXX It is assumed that siz is less than the size of an mbuf at the moment.
771 */
772 struct mbuf *
773 m_inject(struct mbuf *m0, int len0, int siz, int wait)
774 {
775 struct mbuf *m, *n, *n2 = NULL, *n3;
776 unsigned len = len0, remain;
777
778 if ((siz >= MHLEN) || (len0 <= 0))
779 return (NULL);
780 for (m = m0; m && len > m->m_len; m = m->m_next)
781 len -= m->m_len;
782 if (m == NULL)
783 return (NULL);
784 remain = m->m_len - len;
785 if (remain == 0) {
786 if ((m->m_next) && (M_LEADINGSPACE(m->m_next) >= siz)) {
787 m->m_next->m_len += siz;
788 if (m0->m_flags & M_PKTHDR)
789 m0->m_pkthdr.len += siz;
790 m->m_next->m_data -= siz;
791 return m->m_next;
792 }
793 } else {
794 n2 = m_copym2(m, len, remain, wait);
795 if (n2 == NULL)
796 return (NULL);
797 }
798
799 MGET(n, wait, MT_DATA);
800 if (n == NULL) {
801 if (n2)
802 m_freem(n2);
803 return (NULL);
804 }
805
806 n->m_len = siz;
807 if (m0->m_flags & M_PKTHDR)
808 m0->m_pkthdr.len += siz;
809 m->m_len -= remain; /* Trim */
810 if (n2) {
811 for (n3 = n; n3->m_next != NULL; n3 = n3->m_next)
812 ;
813 n3->m_next = n2;
814 } else
815 n3 = n;
816 for (; n3->m_next != NULL; n3 = n3->m_next)
817 ;
818 n3->m_next = m->m_next;
819 m->m_next = n;
820 return n;
821 }
822
823 /*
824 * Partition an mbuf chain in two pieces, returning the tail --
825 * all but the first len0 bytes. In case of failure, it returns NULL and
826 * attempts to restore the chain to its original state.
827 */
828 struct mbuf *
829 m_split(struct mbuf *m0, int len0, int wait)
830 {
831 struct mbuf *m, *n;
832 unsigned len = len0, remain, olen;
833
834 for (m = m0; m && len > m->m_len; m = m->m_next)
835 len -= m->m_len;
836 if (m == NULL)
837 return (NULL);
838 remain = m->m_len - len;
839 if (m0->m_flags & M_PKTHDR) {
840 MGETHDR(n, wait, m0->m_type);
841 if (n == NULL)
842 return (NULL);
843 M_DUP_PKTHDR(n, m0);
844 n->m_pkthdr.len -= len0;
845 olen = m0->m_pkthdr.len;
846 m0->m_pkthdr.len = len0;
847 if (m->m_flags & M_EXT)
848 goto extpacket;
849 if (remain > MHLEN) {
850 /* m can't be the lead packet */
851 MH_ALIGN(n, 0);
852 n->m_next = m_split(m, len, wait);
853 if (n->m_next == NULL) {
854 (void) m_free(n);
855 m0->m_pkthdr.len = olen;
856 return (NULL);
857 } else
858 return (n);
859 } else
860 MH_ALIGN(n, remain);
861 } else if (remain == 0) {
862 n = m->m_next;
863 m->m_next = NULL;
864 return (n);
865 } else {
866 MGET(n, wait, m->m_type);
867 if (n == NULL)
868 return (NULL);
869 M_ALIGN(n, remain);
870 }
871 extpacket:
872 if (m->m_flags & M_EXT) {
873 n->m_ext = m->m_ext;
874 MCLADDREFERENCE(m, n);
875 n->m_data = m->m_data + len;
876 } else {
877 bcopy(mtod(m, caddr_t) + len, mtod(n, caddr_t), remain);
878 }
879 n->m_len = remain;
880 m->m_len = len;
881 n->m_next = m->m_next;
882 m->m_next = NULL;
883 return (n);
884 }
885
886 /*
887 * Routine to copy from device local memory into mbufs.
888 */
889 struct mbuf *
890 m_devget(char *buf, int totlen, int off, struct ifnet *ifp,
891 void (*copy)(const void *, void *, size_t))
892 {
893 struct mbuf *m;
894 struct mbuf *top = NULL, **mp = ⊤
895 int len;
896 char *cp;
897 char *epkt;
898
899 cp = buf;
900 epkt = cp + totlen;
901 if (off) {
902 /*
903 * If 'off' is non-zero, packet is trailer-encapsulated,
904 * so we have to skip the type and length fields.
905 */
906 cp += off + 2 * sizeof(u_int16_t);
907 totlen -= 2 * sizeof(u_int16_t);
908 }
909 MGETHDR(m, M_DONTWAIT, MT_DATA);
910 if (m == NULL)
911 return (NULL);
912 m->m_pkthdr.rcvif = ifp;
913 m->m_pkthdr.len = totlen;
914 m->m_len = MHLEN;
915
916 while (totlen > 0) {
917 if (top != NULL) {
918 MGET(m, M_DONTWAIT, MT_DATA);
919 if (m == NULL) {
920 m_freem(top);
921 return (NULL);
922 }
923 m->m_len = MLEN;
924 }
925 len = min(totlen, epkt - cp);
926 if (len >= MINCLSIZE) {
927 MCLGET(m, M_DONTWAIT);
928 if (m->m_flags & M_EXT)
929 m->m_len = len = min(len, MCLBYTES);
930 else
931 len = m->m_len;
932 } else {
933 /*
934 * Place initial small packet/header at end of mbuf.
935 */
936 if (len < m->m_len) {
937 if (top == NULL &&
938 len + max_linkhdr <= m->m_len)
939 m->m_data += max_linkhdr;
940 m->m_len = len;
941 } else
942 len = m->m_len;
943 }
944 if (copy)
945 copy(cp, mtod(m, caddr_t), (size_t)len);
946 else
947 bcopy(cp, mtod(m, caddr_t), (size_t)len);
948 cp += len;
949 *mp = m;
950 mp = &m->m_next;
951 totlen -= len;
952 if (cp == epkt)
953 cp = buf;
954 }
955 return (top);
956 }
957
958 void
959 m_zero(struct mbuf *m)
960 {
961 while (m) {
962 #ifdef DIAGNOSTIC
963 if (M_READONLY(m))
964 panic("m_zero: M_READONLY");
965 #endif /* DIAGNOSTIC */
966 if (m->m_flags & M_EXT)
967 memset(m->m_ext.ext_buf, 0, m->m_ext.ext_size);
968 else {
969 if (m->m_flags & M_PKTHDR)
970 memset(m->m_pktdat, 0, MHLEN);
971 else
972 memset(m->m_dat, 0, MLEN);
973 }
974 m = m->m_next;
975 }
976 }
977
978 /*
979 * Apply function f to the data in an mbuf chain starting "off" bytes from the
980 * beginning, continuing for "len" bytes.
981 */
982 int
983 m_apply(struct mbuf *m, int off, int len,
984 int (*f)(caddr_t, caddr_t, unsigned int), caddr_t fstate)
985 {
986 int rval;
987 unsigned int count;
988
989 if (len < 0)
990 panic("m_apply: len %d < 0", len);
991 if (off < 0)
992 panic("m_apply: off %d < 0", off);
993 while (off > 0) {
994 if (m == NULL)
995 panic("m_apply: null mbuf in skip");
996 if (off < m->m_len)
997 break;
998 off -= m->m_len;
999 m = m->m_next;
1000 }
1001 while (len > 0) {
1002 if (m == NULL)
1003 panic("m_apply: null mbuf");
1004 count = min(m->m_len - off, len);
1005
1006 rval = f(fstate, mtod(m, caddr_t) + off, count);
1007 if (rval)
1008 return (rval);
1009
1010 len -= count;
1011 off = 0;
1012 m = m->m_next;
1013 }
1014
1015 return (0);
1016 }