xref: /freebsd/sys/kern/uipc_mbuf2.c (revision eaa797943eeac5614edfdc8f6309f332343c3dd2)
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. 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 <sys/param.h>
69 #include <sys/systm.h>
70 #include <sys/kernel.h>
71 #include <sys/lock.h>
72 #include <sys/malloc.h>
73 #include <sys/mbuf.h>
74 #include <sys/mutex.h>
75 
76 #include <security/mac/mac_framework.h>
77 
78 static MALLOC_DEFINE(M_PACKET_TAGS, MBUF_TAG_MEM_NAME,
79     "packet-attached information");
80 
81 /* can't call it m_dup(), as freebsd[34] uses m_dup() with different arg */
82 static struct mbuf *m_dup1(struct mbuf *, int, int, int);
83 
84 /*
85  * ensure that [off, off + len) is contiguous on the mbuf chain "m".
86  * packet chain before "off" is kept untouched.
87  * if offp == NULL, the target will start at <retval, 0> on resulting chain.
88  * if offp != NULL, the target will start at <retval, *offp> on resulting chain.
89  *
90  * on error return (NULL return value), original "m" will be freed.
91  *
92  * XXX: M_TRAILINGSPACE/M_LEADINGSPACE only permitted on writable ext_buf.
93  */
94 struct mbuf *
95 m_pulldown(struct mbuf *m, int off, int len, int *offp)
96 {
97 	struct mbuf *n, *o;
98 	int hlen, tlen, olen;
99 	int writable;
100 
101 	/* check invalid arguments. */
102 	if (m == NULL)
103 		panic("m == NULL in m_pulldown()");
104 	if (len > MCLBYTES) {
105 		m_freem(m);
106 		return NULL;	/* impossible */
107 	}
108 
109 #ifdef PULLDOWN_DEBUG
110     {
111 	struct mbuf *t;
112 	printf("before:");
113 	for (t = m; t; t = t->m_next)
114 		printf(" %d", t->m_len);
115 	printf("\n");
116     }
117 #endif
118 	n = m;
119 	while (n != NULL && off > 0) {
120 		if (n->m_len > off)
121 			break;
122 		off -= n->m_len;
123 		n = n->m_next;
124 	}
125 	/* be sure to point non-empty mbuf */
126 	while (n != NULL && n->m_len == 0)
127 		n = n->m_next;
128 	if (!n) {
129 		m_freem(m);
130 		return NULL;	/* mbuf chain too short */
131 	}
132 
133 	/*
134 	 * The following comment is dated but still partially applies:
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 possession 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 	writable = 0;
154 	if ((n->m_flags & M_EXT) == 0 ||
155 	    (n->m_ext.ext_type == EXT_CLUSTER && M_WRITABLE(n)))
156 		writable = 1;
157 
158 	/*
159 	 * the target data is on <n, off>.
160 	 * if we got enough data on the mbuf "n", we're done.
161 	 */
162 	if ((off == 0 || offp) && len <= n->m_len - off && writable)
163 		goto ok;
164 
165 	/*
166 	 * when len <= n->m_len - off and off != 0, it is a special case.
167 	 * len bytes from <n, off> sits in single mbuf, but the caller does
168 	 * not like the starting position (off).
169 	 * chop the current mbuf into two pieces, set off to 0.
170 	 */
171 	if (len <= n->m_len - off) {
172 		o = m_dup1(n, off, n->m_len - off, M_NOWAIT);
173 		if (o == NULL) {
174 			m_freem(m);
175 			return NULL;	/* ENOBUFS */
176 		}
177 		n->m_len = off;
178 		o->m_next = n->m_next;
179 		n->m_next = o;
180 		n = n->m_next;
181 		off = 0;
182 		goto ok;
183 	}
184 
185 	/*
186 	 * we need to take hlen from <n, off> and tlen from <n->m_next, 0>,
187 	 * and construct contiguous mbuf with m_len == len.
188 	 * note that hlen + tlen == len, and tlen > 0.
189 	 */
190 	hlen = n->m_len - off;
191 	tlen = len - hlen;
192 
193 	/*
194 	 * ensure that we have enough trailing data on mbuf chain.
195 	 * if not, we can do nothing about the chain.
196 	 */
197 	olen = 0;
198 	for (o = n->m_next; o != NULL; o = o->m_next)
199 		olen += o->m_len;
200 	if (hlen + olen < len) {
201 		m_freem(m);
202 		return NULL;	/* mbuf chain too short */
203 	}
204 
205 	/*
206 	 * easy cases first.
207 	 * we need to use m_copydata() to get data from <n->m_next, 0>.
208 	 */
209 	if ((off == 0 || offp) && M_TRAILINGSPACE(n) >= tlen
210 	 && writable) {
211 		m_copydata(n->m_next, 0, tlen, mtod(n, caddr_t) + n->m_len);
212 		n->m_len += tlen;
213 		m_adj(n->m_next, tlen);
214 		goto ok;
215 	}
216 	if ((off == 0 || offp) && M_LEADINGSPACE(n->m_next) >= hlen
217 	 && writable) {
218 		n->m_next->m_data -= hlen;
219 		n->m_next->m_len += hlen;
220 		bcopy(mtod(n, caddr_t) + off, mtod(n->m_next, caddr_t), hlen);
221 		n->m_len -= hlen;
222 		n = n->m_next;
223 		off = 0;
224 		goto ok;
225 	}
226 
227 	/*
228 	 * now, we need to do the hard way.  don't m_copy as there's no room
229 	 * on both end.
230 	 */
231 	if (len > MLEN)
232 		o = m_getcl(M_NOWAIT, m->m_type, 0);
233 	else
234 		o = m_get(M_NOWAIT, m->m_type);
235 	if (!o) {
236 		m_freem(m);
237 		return NULL;	/* ENOBUFS */
238 	}
239 	/* get hlen from <n, off> into <o, 0> */
240 	o->m_len = hlen;
241 	bcopy(mtod(n, caddr_t) + off, mtod(o, caddr_t), hlen);
242 	n->m_len -= hlen;
243 	/* get tlen from <n->m_next, 0> into <o, hlen> */
244 	m_copydata(n->m_next, 0, tlen, mtod(o, caddr_t) + o->m_len);
245 	o->m_len += tlen;
246 	m_adj(n->m_next, tlen);
247 	o->m_next = n->m_next;
248 	n->m_next = o;
249 	n = o;
250 	off = 0;
251 
252 ok:
253 #ifdef PULLDOWN_DEBUG
254     {
255 	struct mbuf *t;
256 	printf("after:");
257 	for (t = m; t; t = t->m_next)
258 		printf("%c%d", t == n ? '*' : ' ', t->m_len);
259 	printf(" (off=%d)\n", off);
260     }
261 #endif
262 	if (offp)
263 		*offp = off;
264 	return n;
265 }
266 
267 static struct mbuf *
268 m_dup1(struct mbuf *m, int off, int len, int wait)
269 {
270 	struct mbuf *n;
271 	int copyhdr;
272 
273 	if (len > MCLBYTES)
274 		return NULL;
275 	if (off == 0 && (m->m_flags & M_PKTHDR) != 0)
276 		copyhdr = 1;
277 	else
278 		copyhdr = 0;
279 	if (len >= MINCLSIZE) {
280 		if (copyhdr == 1)
281 			n = m_getcl(wait, m->m_type, M_PKTHDR);
282 		else
283 			n = m_getcl(wait, m->m_type, 0);
284 	} else {
285 		if (copyhdr == 1)
286 			n = m_gethdr(wait, m->m_type);
287 		else
288 			n = m_get(wait, m->m_type);
289 	}
290 	if (!n)
291 		return NULL; /* ENOBUFS */
292 
293 	if (copyhdr && !m_dup_pkthdr(n, m, wait)) {
294 		m_free(n);
295 		return NULL;
296 	}
297 	m_copydata(m, off, len, mtod(n, caddr_t));
298 	n->m_len = len;
299 	return n;
300 }
301 
302 /* Free a packet tag. */
303 void
304 m_tag_free_default(struct m_tag *t)
305 {
306 #ifdef MAC
307 	if (t->m_tag_id == PACKET_TAG_MACLABEL)
308 		mac_mbuf_tag_destroy(t);
309 #endif
310 	free(t, M_PACKET_TAGS);
311 }
312 
313 /* Get a packet tag structure along with specified data following. */
314 struct m_tag *
315 m_tag_alloc(uint32_t cookie, int type, int len, int wait)
316 {
317 	struct m_tag *t;
318 
319 	MBUF_CHECKSLEEP(wait);
320 	if (len < 0)
321 		return NULL;
322 	t = malloc(len + sizeof(struct m_tag), M_PACKET_TAGS, wait);
323 	if (t == NULL)
324 		return NULL;
325 	m_tag_setup(t, cookie, type, len);
326 	t->m_tag_free = m_tag_free_default;
327 	return t;
328 }
329 
330 /* Unlink and free a packet tag. */
331 void
332 m_tag_delete(struct mbuf *m, struct m_tag *t)
333 {
334 
335 	KASSERT(m && t, ("m_tag_delete: null argument, m %p t %p", m, t));
336 	m_tag_unlink(m, t);
337 	m_tag_free(t);
338 }
339 
340 /* Unlink and free a packet tag chain, starting from given tag. */
341 void
342 m_tag_delete_chain(struct mbuf *m, struct m_tag *t)
343 {
344 	struct m_tag *p, *q;
345 
346 	KASSERT(m, ("m_tag_delete_chain: null mbuf"));
347 	if (t != NULL)
348 		p = t;
349 	else
350 		p = SLIST_FIRST(&m->m_pkthdr.tags);
351 	if (p == NULL)
352 		return;
353 	while ((q = SLIST_NEXT(p, m_tag_link)) != NULL)
354 		m_tag_delete(m, q);
355 	m_tag_delete(m, p);
356 }
357 
358 /*
359  * Strip off all tags that would normally vanish when
360  * passing through a network interface.  Only persistent
361  * tags will exist after this; these are expected to remain
362  * so long as the mbuf chain exists, regardless of the
363  * path the mbufs take.
364  */
365 void
366 m_tag_delete_nonpersistent(struct mbuf *m)
367 {
368 	struct m_tag *p, *q;
369 
370 	SLIST_FOREACH_SAFE(p, &m->m_pkthdr.tags, m_tag_link, q)
371 		if ((p->m_tag_id & MTAG_PERSISTENT) == 0)
372 			m_tag_delete(m, p);
373 }
374 
375 /* Find a tag, starting from a given position. */
376 struct m_tag *
377 m_tag_locate(struct mbuf *m, uint32_t cookie, int type, struct m_tag *t)
378 {
379 	struct m_tag *p;
380 
381 	KASSERT(m, ("m_tag_locate: null mbuf"));
382 	if (t == NULL)
383 		p = SLIST_FIRST(&m->m_pkthdr.tags);
384 	else
385 		p = SLIST_NEXT(t, m_tag_link);
386 	while (p != NULL) {
387 		if (p->m_tag_cookie == cookie && p->m_tag_id == type)
388 			return p;
389 		p = SLIST_NEXT(p, m_tag_link);
390 	}
391 	return NULL;
392 }
393 
394 /* Copy a single tag. */
395 struct m_tag *
396 m_tag_copy(struct m_tag *t, int how)
397 {
398 	struct m_tag *p;
399 
400 	MBUF_CHECKSLEEP(how);
401 	KASSERT(t, ("m_tag_copy: null tag"));
402 	p = m_tag_alloc(t->m_tag_cookie, t->m_tag_id, t->m_tag_len, how);
403 	if (p == NULL)
404 		return (NULL);
405 #ifdef MAC
406 	/*
407 	 * XXXMAC: we should probably pass off the initialization, and
408 	 * copying here?  can we hide that PACKET_TAG_MACLABEL is
409 	 * special from the mbuf code?
410 	 */
411 	if (t->m_tag_id == PACKET_TAG_MACLABEL) {
412 		if (mac_mbuf_tag_init(p, how) != 0) {
413 			m_tag_free(p);
414 			return (NULL);
415 		}
416 		mac_mbuf_tag_copy(t, p);
417 	} else
418 #endif
419 		bcopy(t + 1, p + 1, t->m_tag_len); /* Copy the data */
420 	return p;
421 }
422 
423 /*
424  * Copy two tag chains. The destination mbuf (to) loses any attached
425  * tags even if the operation fails. This should not be a problem, as
426  * m_tag_copy_chain() is typically called with a newly-allocated
427  * destination mbuf.
428  */
429 int
430 m_tag_copy_chain(struct mbuf *to, const struct mbuf *from, int how)
431 {
432 	struct m_tag *p, *t, *tprev = NULL;
433 
434 	MBUF_CHECKSLEEP(how);
435 	KASSERT(to && from,
436 		("m_tag_copy_chain: null argument, to %p from %p", to, from));
437 	m_tag_delete_chain(to, NULL);
438 	SLIST_FOREACH(p, &from->m_pkthdr.tags, m_tag_link) {
439 		t = m_tag_copy(p, how);
440 		if (t == NULL) {
441 			m_tag_delete_chain(to, NULL);
442 			return 0;
443 		}
444 		if (tprev == NULL)
445 			SLIST_INSERT_HEAD(&to->m_pkthdr.tags, t, m_tag_link);
446 		else
447 			SLIST_INSERT_AFTER(tprev, t, m_tag_link);
448 		tprev = t;
449 	}
450 	return 1;
451 }
452