xref: /freebsd/sys/kern/uipc_mbuf2.c (revision c4f02a891fe62fe1277c89859922804ea2c27bcd)
1 /*	$KAME: uipc_mbuf2.c,v 1.31 2001/11/28 11:08:53 itojun Exp $	*/
2 /*	$NetBSD: uipc_mbuf.c,v 1.40 1999/04/01 00:23:25 thorpej Exp $	*/
3 
4 /*
5  * Copyright (C) 1999 WIDE Project.
6  * 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 project 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 PROJECT 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 PROJECT 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 /*
33  * Copyright (c) 1982, 1986, 1988, 1991, 1993
34  *	The Regents of the University of California.  All rights reserved.
35  *
36  * Redistribution and use in source and binary forms, with or without
37  * modification, are permitted provided that the following conditions
38  * are met:
39  * 1. Redistributions of source code must retain the above copyright
40  *    notice, this list of conditions and the following disclaimer.
41  * 2. Redistributions in binary form must reproduce the above copyright
42  *    notice, this list of conditions and the following disclaimer in the
43  *    documentation and/or other materials provided with the distribution.
44  * 3. All advertising materials mentioning features or use of this software
45  *    must display the following acknowledgement:
46  *	This product includes software developed by the University of
47  *	California, Berkeley and its contributors.
48  * 4. Neither the name of the University nor the names of its contributors
49  *    may be used to endorse or promote products derived from this software
50  *    without specific prior written permission.
51  *
52  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
53  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
54  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
55  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
56  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
57  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
58  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
59  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
60  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
61  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
62  * SUCH DAMAGE.
63  *
64  *	@(#)uipc_mbuf.c	8.4 (Berkeley) 2/14/95
65  */
66 
67 #include <sys/cdefs.h>
68 __FBSDID("$FreeBSD$");
69 
70 /*#define PULLDOWN_DEBUG*/
71 
72 #include "opt_mac.h"
73 
74 #include <sys/param.h>
75 #include <sys/systm.h>
76 #include <sys/kernel.h>
77 #include <sys/lock.h>
78 #include <sys/mac.h>
79 #include <sys/malloc.h>
80 #include <sys/mbuf.h>
81 #include <sys/mutex.h>
82 
83 MALLOC_DEFINE(M_PACKET_TAGS, "tag", "packet-attached information");
84 
85 /* can't call it m_dup(), as freebsd[34] uses m_dup() with different arg */
86 static struct mbuf *m_dup1(struct mbuf *, int, int, int);
87 
88 /*
89  * ensure that [off, off + len) is contiguous on the mbuf chain "m".
90  * packet chain before "off" is kept untouched.
91  * if offp == NULL, the target will start at <retval, 0> on resulting chain.
92  * if offp != NULL, the target will start at <retval, *offp> on resulting chain.
93  *
94  * on error return (NULL return value), original "m" will be freed.
95  *
96  * XXX: M_TRAILINGSPACE/M_LEADINGSPACE only permitted on writable ext_buf.
97  */
98 struct mbuf *
99 m_pulldown(struct mbuf *m, int off, int len, int *offp)
100 {
101 	struct mbuf *n, *o;
102 	int hlen, tlen, olen;
103 	int writable;
104 
105 	/* check invalid arguments. */
106 	if (m == NULL)
107 		panic("m == NULL in m_pulldown()");
108 	if (len > MCLBYTES) {
109 		m_freem(m);
110 		return NULL;	/* impossible */
111 	}
112 
113 #ifdef PULLDOWN_DEBUG
114     {
115 	struct mbuf *t;
116 	printf("before:");
117 	for (t = m; t; t = t->m_next)
118 		printf(" %d", t->m_len);
119 	printf("\n");
120     }
121 #endif
122 	n = m;
123 	while (n != NULL && off > 0) {
124 		if (n->m_len > off)
125 			break;
126 		off -= n->m_len;
127 		n = n->m_next;
128 	}
129 	/* be sure to point non-empty mbuf */
130 	while (n != NULL && n->m_len == 0)
131 		n = n->m_next;
132 	if (!n) {
133 		m_freem(m);
134 		return NULL;	/* mbuf chain too short */
135 	}
136 
137 	/*
138 	 * XXX: This code is flawed because it considers a "writable" mbuf
139 	 *      data region to require all of the following:
140 	 *	  (i) mbuf _has_ to have M_EXT set; if it is just a regular
141 	 *	      mbuf, it is still not considered "writable."
142 	 *	  (ii) since mbuf has M_EXT, the ext_type _has_ to be
143 	 *	       EXT_CLUSTER. Anything else makes it non-writable.
144 	 *	  (iii) M_WRITABLE() must evaluate true.
145 	 *      Ideally, the requirement should only be (iii).
146 	 *
147 	 * If we're writable, we're sure we're writable, because the ref. count
148 	 * cannot increase from 1, as that would require posession of mbuf
149 	 * n by someone else (which is impossible). However, if we're _not_
150 	 * writable, we may eventually become writable )if the ref. count drops
151 	 * to 1), but we'll fail to notice it unless we re-evaluate
152 	 * M_WRITABLE(). For now, we only evaluate once at the beginning and
153 	 * live with this.
154 	 */
155 	/*
156 	 * XXX: This is dumb. If we're just a regular mbuf with no M_EXT,
157 	 *      then we're not "writable," according to this code.
158 	 */
159 	writable = 0;
160 	if ((n->m_flags & M_EXT) == 0 ||
161 	    (n->m_ext.ext_type == EXT_CLUSTER && M_WRITABLE(n)))
162 		writable = 1;
163 
164 	/*
165 	 * the target data is on <n, off>.
166 	 * if we got enough data on the mbuf "n", we're done.
167 	 */
168 	if ((off == 0 || offp) && len <= n->m_len - off && writable)
169 		goto ok;
170 
171 	/*
172 	 * when len <= n->m_len - off and off != 0, it is a special case.
173 	 * len bytes from <n, off> sits in single mbuf, but the caller does
174 	 * not like the starting position (off).
175 	 * chop the current mbuf into two pieces, set off to 0.
176 	 */
177 	if (len <= n->m_len - off) {
178 		o = m_dup1(n, off, n->m_len - off, M_DONTWAIT);
179 		if (o == NULL) {
180 			m_freem(m);
181 			return NULL;	/* ENOBUFS */
182 		}
183 		n->m_len = off;
184 		o->m_next = n->m_next;
185 		n->m_next = o;
186 		n = n->m_next;
187 		off = 0;
188 		goto ok;
189 	}
190 
191 	/*
192 	 * we need to take hlen from <n, off> and tlen from <n->m_next, 0>,
193 	 * and construct contiguous mbuf with m_len == len.
194 	 * note that hlen + tlen == len, and tlen > 0.
195 	 */
196 	hlen = n->m_len - off;
197 	tlen = len - hlen;
198 
199 	/*
200 	 * ensure that we have enough trailing data on mbuf chain.
201 	 * if not, we can do nothing about the chain.
202 	 */
203 	olen = 0;
204 	for (o = n->m_next; o != NULL; o = o->m_next)
205 		olen += o->m_len;
206 	if (hlen + olen < len) {
207 		m_freem(m);
208 		return NULL;	/* mbuf chain too short */
209 	}
210 
211 	/*
212 	 * easy cases first.
213 	 * we need to use m_copydata() to get data from <n->m_next, 0>.
214 	 */
215 	if ((off == 0 || offp) && M_TRAILINGSPACE(n) >= tlen
216 	 && writable) {
217 		m_copydata(n->m_next, 0, tlen, mtod(n, caddr_t) + n->m_len);
218 		n->m_len += tlen;
219 		m_adj(n->m_next, tlen);
220 		goto ok;
221 	}
222 	if ((off == 0 || offp) && M_LEADINGSPACE(n->m_next) >= hlen
223 	 && writable) {
224 		n->m_next->m_data -= hlen;
225 		n->m_next->m_len += hlen;
226 		bcopy(mtod(n, caddr_t) + off, mtod(n->m_next, caddr_t), hlen);
227 		n->m_len -= hlen;
228 		n = n->m_next;
229 		off = 0;
230 		goto ok;
231 	}
232 
233 	/*
234 	 * now, we need to do the hard way.  don't m_copy as there's no room
235 	 * on both end.
236 	 */
237 	MGET(o, M_DONTWAIT, m->m_type);
238 	if (o && len > MLEN) {
239 		MCLGET(o, M_DONTWAIT);
240 		if ((o->m_flags & M_EXT) == 0) {
241 			m_free(o);
242 			o = NULL;
243 		}
244 	}
245 	if (!o) {
246 		m_freem(m);
247 		return NULL;	/* ENOBUFS */
248 	}
249 	/* get hlen from <n, off> into <o, 0> */
250 	o->m_len = hlen;
251 	bcopy(mtod(n, caddr_t) + off, mtod(o, caddr_t), hlen);
252 	n->m_len -= hlen;
253 	/* get tlen from <n->m_next, 0> into <o, hlen> */
254 	m_copydata(n->m_next, 0, tlen, mtod(o, caddr_t) + o->m_len);
255 	o->m_len += tlen;
256 	m_adj(n->m_next, tlen);
257 	o->m_next = n->m_next;
258 	n->m_next = o;
259 	n = o;
260 	off = 0;
261 
262 ok:
263 #ifdef PULLDOWN_DEBUG
264     {
265 	struct mbuf *t;
266 	printf("after:");
267 	for (t = m; t; t = t->m_next)
268 		printf("%c%d", t == n ? '*' : ' ', t->m_len);
269 	printf(" (off=%d)\n", off);
270     }
271 #endif
272 	if (offp)
273 		*offp = off;
274 	return n;
275 }
276 
277 static struct mbuf *
278 m_dup1(struct mbuf *m, int off, int len, int wait)
279 {
280 	struct mbuf *n;
281 	int l;
282 	int copyhdr;
283 
284 	if (len > MCLBYTES)
285 		return NULL;
286 	if (off == 0 && (m->m_flags & M_PKTHDR) != 0) {
287 		copyhdr = 1;
288 		MGETHDR(n, wait, m->m_type);
289 		l = MHLEN;
290 	} else {
291 		copyhdr = 0;
292 		MGET(n, wait, m->m_type);
293 		l = MLEN;
294 	}
295 	if (n && len > l) {
296 		MCLGET(n, wait);
297 		if ((n->m_flags & M_EXT) == 0) {
298 			m_free(n);
299 			n = NULL;
300 		}
301 	}
302 	if (!n)
303 		return NULL;
304 
305 	if (copyhdr && !m_dup_pkthdr(n, m, wait)) {
306 		m_free(n);
307 		return NULL;
308 	}
309 	m_copydata(m, off, len, mtod(n, caddr_t));
310 	return n;
311 }
312 
313 /* Get a packet tag structure along with specified data following. */
314 struct m_tag *
315 m_tag_alloc(u_int32_t cookie, int type, int len, int wait)
316 {
317 	struct m_tag *t;
318 
319 	if (len < 0)
320 		return NULL;
321 	t = malloc(len + sizeof(struct m_tag), M_PACKET_TAGS, wait);
322 	if (t == NULL)
323 		return NULL;
324 	t->m_tag_id = type;
325 	t->m_tag_len = len;
326 	t->m_tag_cookie = cookie;
327 	return t;
328 }
329 
330 /* Free a packet tag. */
331 void
332 m_tag_free(struct m_tag *t)
333 {
334 #ifdef MAC
335 	if (t->m_tag_id == PACKET_TAG_MACLABEL)
336 		mac_destroy_mbuf_tag(t);
337 #endif
338 	free(t, M_PACKET_TAGS);
339 }
340 
341 /* Prepend a packet tag. */
342 void
343 m_tag_prepend(struct mbuf *m, struct m_tag *t)
344 {
345 	KASSERT(m && t, ("m_tag_prepend: null argument, m %p t %p", m, t));
346 	SLIST_INSERT_HEAD(&m->m_pkthdr.tags, t, m_tag_link);
347 }
348 
349 /* Unlink a packet tag. */
350 void
351 m_tag_unlink(struct mbuf *m, struct m_tag *t)
352 {
353 	KASSERT(m && t, ("m_tag_unlink: null argument, m %p t %p", m, t));
354 	SLIST_REMOVE(&m->m_pkthdr.tags, t, m_tag, m_tag_link);
355 }
356 
357 /* Unlink and free a packet tag. */
358 void
359 m_tag_delete(struct mbuf *m, struct m_tag *t)
360 {
361 	KASSERT(m && t, ("m_tag_delete: null argument, m %p t %p", m, t));
362 	m_tag_unlink(m, t);
363 	m_tag_free(t);
364 }
365 
366 /* Unlink and free a packet tag chain, starting from given tag. */
367 void
368 m_tag_delete_chain(struct mbuf *m, struct m_tag *t)
369 {
370 	struct m_tag *p, *q;
371 
372 	KASSERT(m, ("m_tag_delete_chain: null mbuf"));
373 	if (t != NULL)
374 		p = t;
375 	else
376 		p = SLIST_FIRST(&m->m_pkthdr.tags);
377 	if (p == NULL)
378 		return;
379 	while ((q = SLIST_NEXT(p, m_tag_link)) != NULL)
380 		m_tag_delete(m, q);
381 	m_tag_delete(m, p);
382 }
383 
384 /* Find a tag, starting from a given position. */
385 struct m_tag *
386 m_tag_locate(struct mbuf *m, u_int32_t cookie, int type, struct m_tag *t)
387 {
388 	struct m_tag *p;
389 
390 	KASSERT(m, ("m_tag_locate: null mbuf"));
391 	if (t == NULL)
392 		p = SLIST_FIRST(&m->m_pkthdr.tags);
393 	else
394 		p = SLIST_NEXT(t, m_tag_link);
395 	while (p != NULL) {
396 		if (p->m_tag_cookie == cookie && p->m_tag_id == type)
397 			return p;
398 		p = SLIST_NEXT(p, m_tag_link);
399 	}
400 	return NULL;
401 }
402 
403 /* Copy a single tag. */
404 struct m_tag *
405 m_tag_copy(struct m_tag *t, int how)
406 {
407 	struct m_tag *p;
408 
409 	KASSERT(t, ("m_tag_copy: null tag"));
410 	p = m_tag_alloc(t->m_tag_cookie, t->m_tag_id, t->m_tag_len, how);
411 	if (p == NULL)
412 		return (NULL);
413 #ifdef MAC
414 	/*
415 	 * XXXMAC: we should probably pass off the initialization, and
416 	 * copying here?  can we hide that PACKET_TAG_MACLABEL is
417 	 * special from the mbuf code?
418 	 */
419 	if (t->m_tag_id == PACKET_TAG_MACLABEL) {
420 		if (mac_init_mbuf_tag(p, how) != 0) {
421 			m_tag_free(p);
422 			return (NULL);
423 		}
424 		mac_copy_mbuf_tag(t, p);
425 	} else
426 #endif
427 		bcopy(t + 1, p + 1, t->m_tag_len); /* Copy the data */
428 	return p;
429 }
430 
431 /*
432  * Copy two tag chains. The destination mbuf (to) loses any attached
433  * tags even if the operation fails. This should not be a problem, as
434  * m_tag_copy_chain() is typically called with a newly-allocated
435  * destination mbuf.
436  */
437 int
438 m_tag_copy_chain(struct mbuf *to, struct mbuf *from, int how)
439 {
440 	struct m_tag *p, *t, *tprev = NULL;
441 
442 	KASSERT(to && from,
443 		("m_tag_copy_chain: null argument, to %p from %p", to, from));
444 	m_tag_delete_chain(to, NULL);
445 	SLIST_FOREACH(p, &from->m_pkthdr.tags, m_tag_link) {
446 		t = m_tag_copy(p, how);
447 		if (t == NULL) {
448 			m_tag_delete_chain(to, NULL);
449 			return 0;
450 		}
451 		if (tprev == NULL)
452 			SLIST_INSERT_HEAD(&to->m_pkthdr.tags, t, m_tag_link);
453 		else
454 			SLIST_INSERT_AFTER(tprev, t, m_tag_link);
455 		tprev = t;
456 	}
457 	return 1;
458 }
459 
460 /* Initialize tags on an mbuf. */
461 void
462 m_tag_init(struct mbuf *m)
463 {
464 	SLIST_INIT(&m->m_pkthdr.tags);
465 }
466 
467 /* Get first tag in chain. */
468 struct m_tag *
469 m_tag_first(struct mbuf *m)
470 {
471 	return SLIST_FIRST(&m->m_pkthdr.tags);
472 }
473 
474 /* Get next tag in chain. */
475 struct m_tag *
476 m_tag_next(struct mbuf *m, struct m_tag *t)
477 {
478 	return SLIST_NEXT(t, m_tag_link);
479 }
480