xref: /freebsd/sys/kern/uipc_mbuf2.c (revision 2be1a816b9ff69588e55be0a84cbe2a31efc0f2f)
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  * 4. Neither the name of the University nor the names of its contributors
45  *    may be used to endorse or promote products derived from this software
46  *    without specific prior written permission.
47  *
48  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
49  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
50  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
51  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
52  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
53  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
54  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
55  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
56  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
57  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
58  * SUCH DAMAGE.
59  *
60  *	@(#)uipc_mbuf.c	8.4 (Berkeley) 2/14/95
61  */
62 
63 #include <sys/cdefs.h>
64 __FBSDID("$FreeBSD$");
65 
66 /*#define PULLDOWN_DEBUG*/
67 
68 #include "opt_mac.h"
69 
70 #include <sys/param.h>
71 #include <sys/systm.h>
72 #include <sys/kernel.h>
73 #include <sys/lock.h>
74 #include <sys/malloc.h>
75 #include <sys/mbuf.h>
76 #include <sys/mutex.h>
77 
78 #include <security/mac/mac_framework.h>
79 
80 static MALLOC_DEFINE(M_PACKET_TAGS, MBUF_TAG_MEM_NAME,
81     "packet-attached information");
82 
83 /* can't call it m_dup(), as freebsd[34] uses m_dup() with different arg */
84 static struct mbuf *m_dup1(struct mbuf *, int, int, int);
85 
86 /*
87  * ensure that [off, off + len) is contiguous on the mbuf chain "m".
88  * packet chain before "off" is kept untouched.
89  * if offp == NULL, the target will start at <retval, 0> on resulting chain.
90  * if offp != NULL, the target will start at <retval, *offp> on resulting chain.
91  *
92  * on error return (NULL return value), original "m" will be freed.
93  *
94  * XXX: M_TRAILINGSPACE/M_LEADINGSPACE only permitted on writable ext_buf.
95  */
96 struct mbuf *
97 m_pulldown(struct mbuf *m, int off, int len, int *offp)
98 {
99 	struct mbuf *n, *o;
100 	int hlen, tlen, olen;
101 	int writable;
102 
103 	/* check invalid arguments. */
104 	if (m == NULL)
105 		panic("m == NULL in m_pulldown()");
106 	if (len > MCLBYTES) {
107 		m_freem(m);
108 		return NULL;	/* impossible */
109 	}
110 
111 #ifdef PULLDOWN_DEBUG
112     {
113 	struct mbuf *t;
114 	printf("before:");
115 	for (t = m; t; t = t->m_next)
116 		printf(" %d", t->m_len);
117 	printf("\n");
118     }
119 #endif
120 	n = m;
121 	while (n != NULL && off > 0) {
122 		if (n->m_len > off)
123 			break;
124 		off -= n->m_len;
125 		n = n->m_next;
126 	}
127 	/* be sure to point non-empty mbuf */
128 	while (n != NULL && n->m_len == 0)
129 		n = n->m_next;
130 	if (!n) {
131 		m_freem(m);
132 		return NULL;	/* mbuf chain too short */
133 	}
134 
135 	/*
136 	 * XXX: This code is flawed because it considers a "writable" mbuf
137 	 *      data region to require all of the following:
138 	 *	  (i) mbuf _has_ to have M_EXT set; if it is just a regular
139 	 *	      mbuf, it is still not considered "writable."
140 	 *	  (ii) since mbuf has M_EXT, the ext_type _has_ to be
141 	 *	       EXT_CLUSTER. Anything else makes it non-writable.
142 	 *	  (iii) M_WRITABLE() must evaluate true.
143 	 *      Ideally, the requirement should only be (iii).
144 	 *
145 	 * If we're writable, we're sure we're writable, because the ref. count
146 	 * cannot increase from 1, as that would require posession of mbuf
147 	 * n by someone else (which is impossible). However, if we're _not_
148 	 * writable, we may eventually become writable )if the ref. count drops
149 	 * to 1), but we'll fail to notice it unless we re-evaluate
150 	 * M_WRITABLE(). For now, we only evaluate once at the beginning and
151 	 * live with this.
152 	 */
153 	/*
154 	 * XXX: This is dumb. If we're just a regular mbuf with no M_EXT,
155 	 *      then we're not "writable," according to this code.
156 	 */
157 	writable = 0;
158 	if ((n->m_flags & M_EXT) == 0 ||
159 	    (n->m_ext.ext_type == EXT_CLUSTER && M_WRITABLE(n)))
160 		writable = 1;
161 
162 	/*
163 	 * the target data is on <n, off>.
164 	 * if we got enough data on the mbuf "n", we're done.
165 	 */
166 	if ((off == 0 || offp) && len <= n->m_len - off && writable)
167 		goto ok;
168 
169 	/*
170 	 * when len <= n->m_len - off and off != 0, it is a special case.
171 	 * len bytes from <n, off> sits in single mbuf, but the caller does
172 	 * not like the starting position (off).
173 	 * chop the current mbuf into two pieces, set off to 0.
174 	 */
175 	if (len <= n->m_len - off) {
176 		o = m_dup1(n, off, n->m_len - off, M_DONTWAIT);
177 		if (o == NULL) {
178 			m_freem(m);
179 			return NULL;	/* ENOBUFS */
180 		}
181 		n->m_len = off;
182 		o->m_next = n->m_next;
183 		n->m_next = o;
184 		n = n->m_next;
185 		off = 0;
186 		goto ok;
187 	}
188 
189 	/*
190 	 * we need to take hlen from <n, off> and tlen from <n->m_next, 0>,
191 	 * and construct contiguous mbuf with m_len == len.
192 	 * note that hlen + tlen == len, and tlen > 0.
193 	 */
194 	hlen = n->m_len - off;
195 	tlen = len - hlen;
196 
197 	/*
198 	 * ensure that we have enough trailing data on mbuf chain.
199 	 * if not, we can do nothing about the chain.
200 	 */
201 	olen = 0;
202 	for (o = n->m_next; o != NULL; o = o->m_next)
203 		olen += o->m_len;
204 	if (hlen + olen < len) {
205 		m_freem(m);
206 		return NULL;	/* mbuf chain too short */
207 	}
208 
209 	/*
210 	 * easy cases first.
211 	 * we need to use m_copydata() to get data from <n->m_next, 0>.
212 	 */
213 	if ((off == 0 || offp) && M_TRAILINGSPACE(n) >= tlen
214 	 && writable) {
215 		m_copydata(n->m_next, 0, tlen, mtod(n, caddr_t) + n->m_len);
216 		n->m_len += tlen;
217 		m_adj(n->m_next, tlen);
218 		goto ok;
219 	}
220 	if ((off == 0 || offp) && M_LEADINGSPACE(n->m_next) >= hlen
221 	 && writable) {
222 		n->m_next->m_data -= hlen;
223 		n->m_next->m_len += hlen;
224 		bcopy(mtod(n, caddr_t) + off, mtod(n->m_next, caddr_t), hlen);
225 		n->m_len -= hlen;
226 		n = n->m_next;
227 		off = 0;
228 		goto ok;
229 	}
230 
231 	/*
232 	 * now, we need to do the hard way.  don't m_copy as there's no room
233 	 * on both end.
234 	 */
235 	if (len > MLEN)
236 		o = m_getcl(M_DONTWAIT, m->m_type, 0);
237 	else
238 		o = m_get(M_DONTWAIT, m->m_type);
239 	if (!o) {
240 		m_freem(m);
241 		return NULL;	/* ENOBUFS */
242 	}
243 	/* get hlen from <n, off> into <o, 0> */
244 	o->m_len = hlen;
245 	bcopy(mtod(n, caddr_t) + off, mtod(o, caddr_t), hlen);
246 	n->m_len -= hlen;
247 	/* get tlen from <n->m_next, 0> into <o, hlen> */
248 	m_copydata(n->m_next, 0, tlen, mtod(o, caddr_t) + o->m_len);
249 	o->m_len += tlen;
250 	m_adj(n->m_next, tlen);
251 	o->m_next = n->m_next;
252 	n->m_next = o;
253 	n = o;
254 	off = 0;
255 
256 ok:
257 #ifdef PULLDOWN_DEBUG
258     {
259 	struct mbuf *t;
260 	printf("after:");
261 	for (t = m; t; t = t->m_next)
262 		printf("%c%d", t == n ? '*' : ' ', t->m_len);
263 	printf(" (off=%d)\n", off);
264     }
265 #endif
266 	if (offp)
267 		*offp = off;
268 	return n;
269 }
270 
271 static struct mbuf *
272 m_dup1(struct mbuf *m, int off, int len, int wait)
273 {
274 	struct mbuf *n;
275 	int copyhdr;
276 
277 	if (len > MCLBYTES)
278 		return NULL;
279 	if (off == 0 && (m->m_flags & M_PKTHDR) != 0)
280 		copyhdr = 1;
281 	else
282 		copyhdr = 0;
283 	if (len >= MINCLSIZE) {
284 		if (copyhdr == 1)
285 			n = m_getcl(wait, m->m_type, M_PKTHDR);
286 		else
287 			n = m_getcl(wait, m->m_type, 0);
288 	} else {
289 		if (copyhdr == 1)
290 			n = m_gethdr(wait, m->m_type);
291 		else
292 			n = m_get(wait, m->m_type);
293 	}
294 	if (!n)
295 		return NULL; /* ENOBUFS */
296 
297 	if (copyhdr && !m_dup_pkthdr(n, m, wait)) {
298 		m_free(n);
299 		return NULL;
300 	}
301 	m_copydata(m, off, len, mtod(n, caddr_t));
302 	n->m_len = len;
303 	return n;
304 }
305 
306 /* Free a packet tag. */
307 void
308 m_tag_free_default(struct m_tag *t)
309 {
310 #ifdef MAC
311 	if (t->m_tag_id == PACKET_TAG_MACLABEL)
312 		mac_mbuf_tag_destroy(t);
313 #endif
314 	free(t, M_PACKET_TAGS);
315 }
316 
317 /* Get a packet tag structure along with specified data following. */
318 struct m_tag *
319 m_tag_alloc(u_int32_t cookie, int type, int len, int wait)
320 {
321 	struct m_tag *t;
322 
323 	MBUF_CHECKSLEEP(wait);
324 	if (len < 0)
325 		return NULL;
326 	t = malloc(len + sizeof(struct m_tag), M_PACKET_TAGS, wait);
327 	if (t == NULL)
328 		return NULL;
329 	m_tag_setup(t, cookie, type, len);
330 	t->m_tag_free = m_tag_free_default;
331 	return t;
332 }
333 
334 /* Unlink and free a packet tag. */
335 void
336 m_tag_delete(struct mbuf *m, struct m_tag *t)
337 {
338 
339 	KASSERT(m && t, ("m_tag_delete: null argument, m %p t %p", m, t));
340 	m_tag_unlink(m, t);
341 	m_tag_free(t);
342 }
343 
344 /* Unlink and free a packet tag chain, starting from given tag. */
345 void
346 m_tag_delete_chain(struct mbuf *m, struct m_tag *t)
347 {
348 	struct m_tag *p, *q;
349 
350 	KASSERT(m, ("m_tag_delete_chain: null mbuf"));
351 	if (t != NULL)
352 		p = t;
353 	else
354 		p = SLIST_FIRST(&m->m_pkthdr.tags);
355 	if (p == NULL)
356 		return;
357 	while ((q = SLIST_NEXT(p, m_tag_link)) != NULL)
358 		m_tag_delete(m, q);
359 	m_tag_delete(m, p);
360 }
361 
362 /*
363  * Strip off all tags that would normally vanish when
364  * passing through a network interface.  Only persistent
365  * tags will exist after this; these are expected to remain
366  * so long as the mbuf chain exists, regardless of the
367  * path the mbufs take.
368  */
369 void
370 m_tag_delete_nonpersistent(struct mbuf *m)
371 {
372 	struct m_tag *p, *q;
373 
374 	SLIST_FOREACH_SAFE(p, &m->m_pkthdr.tags, m_tag_link, q)
375 		if ((p->m_tag_id & MTAG_PERSISTENT) == 0)
376 			m_tag_delete(m, p);
377 }
378 
379 /* Find a tag, starting from a given position. */
380 struct m_tag *
381 m_tag_locate(struct mbuf *m, u_int32_t cookie, int type, struct m_tag *t)
382 {
383 	struct m_tag *p;
384 
385 	KASSERT(m, ("m_tag_locate: null mbuf"));
386 	if (t == NULL)
387 		p = SLIST_FIRST(&m->m_pkthdr.tags);
388 	else
389 		p = SLIST_NEXT(t, m_tag_link);
390 	while (p != NULL) {
391 		if (p->m_tag_cookie == cookie && p->m_tag_id == type)
392 			return p;
393 		p = SLIST_NEXT(p, m_tag_link);
394 	}
395 	return NULL;
396 }
397 
398 /* Copy a single tag. */
399 struct m_tag *
400 m_tag_copy(struct m_tag *t, int how)
401 {
402 	struct m_tag *p;
403 
404 	MBUF_CHECKSLEEP(how);
405 	KASSERT(t, ("m_tag_copy: null tag"));
406 	p = m_tag_alloc(t->m_tag_cookie, t->m_tag_id, t->m_tag_len, how);
407 	if (p == NULL)
408 		return (NULL);
409 #ifdef MAC
410 	/*
411 	 * XXXMAC: we should probably pass off the initialization, and
412 	 * copying here?  can we hide that PACKET_TAG_MACLABEL is
413 	 * special from the mbuf code?
414 	 */
415 	if (t->m_tag_id == PACKET_TAG_MACLABEL) {
416 		if (mac_mbuf_tag_init(p, how) != 0) {
417 			m_tag_free(p);
418 			return (NULL);
419 		}
420 		mac_mbuf_tag_copy(t, p);
421 	} else
422 #endif
423 		bcopy(t + 1, p + 1, t->m_tag_len); /* Copy the data */
424 	return p;
425 }
426 
427 /*
428  * Copy two tag chains. The destination mbuf (to) loses any attached
429  * tags even if the operation fails. This should not be a problem, as
430  * m_tag_copy_chain() is typically called with a newly-allocated
431  * destination mbuf.
432  */
433 int
434 m_tag_copy_chain(struct mbuf *to, struct mbuf *from, int how)
435 {
436 	struct m_tag *p, *t, *tprev = NULL;
437 
438 	MBUF_CHECKSLEEP(how);
439 	KASSERT(to && from,
440 		("m_tag_copy_chain: null argument, to %p from %p", to, from));
441 	m_tag_delete_chain(to, NULL);
442 	SLIST_FOREACH(p, &from->m_pkthdr.tags, m_tag_link) {
443 		t = m_tag_copy(p, how);
444 		if (t == NULL) {
445 			m_tag_delete_chain(to, NULL);
446 			return 0;
447 		}
448 		if (tprev == NULL)
449 			SLIST_INSERT_HEAD(&to->m_pkthdr.tags, t, m_tag_link);
450 		else
451 			SLIST_INSERT_AFTER(tprev, t, m_tag_link);
452 		tprev = t;
453 	}
454 	return 1;
455 }
456