xref: /freebsd/sys/kern/uipc_mbuf2.c (revision 2e3f49888ec8851bafb22011533217487764fdb0)
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  * SPDX-License-Identifier: BSD-3-Clause
6  *
7  * Copyright (C) 1999 WIDE Project.
8  * All rights reserved.
9  *
10  * Redistribution and use in source and binary forms, with or without
11  * modification, are permitted provided that the following conditions
12  * are met:
13  * 1. Redistributions of source code must retain the above copyright
14  *    notice, this list of conditions and the following disclaimer.
15  * 2. Redistributions in binary form must reproduce the above copyright
16  *    notice, this list of conditions and the following disclaimer in the
17  *    documentation and/or other materials provided with the distribution.
18  * 3. Neither the name of the project nor the names of its contributors
19  *    may be used to endorse or promote products derived from this software
20  *    without specific prior written permission.
21  *
22  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
23  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
24  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
25  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
26  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
27  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
28  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
29  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
30  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
31  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
32  * SUCH DAMAGE.
33  */
34 /*-
35  * Copyright (c) 1982, 1986, 1988, 1991, 1993
36  *	The Regents of the University of California.  All rights reserved.
37  *
38  * Redistribution and use in source and binary forms, with or without
39  * modification, are permitted provided that the following conditions
40  * are met:
41  * 1. Redistributions of source code must retain the above copyright
42  *    notice, this list of conditions and the following disclaimer.
43  * 2. Redistributions in binary form must reproduce the above copyright
44  *    notice, this list of conditions and the following disclaimer in the
45  *    documentation and/or other materials provided with the distribution.
46  * 3. Neither the name of the University nor the names of its contributors
47  *    may be used to endorse or promote products derived from this software
48  *    without specific prior written permission.
49  *
50  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
51  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
52  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
53  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
54  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
55  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
56  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
57  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
58  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
59  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
60  * SUCH DAMAGE.
61  */
62 
63 #include <sys/cdefs.h>
64 /*#define PULLDOWN_DEBUG*/
65 
66 #include <sys/param.h>
67 #include <sys/systm.h>
68 #include <sys/kernel.h>
69 #include <sys/lock.h>
70 #include <sys/malloc.h>
71 #include <sys/mbuf.h>
72 #include <sys/mutex.h>
73 
74 #include <security/mac/mac_framework.h>
75 
76 static MALLOC_DEFINE(M_PACKET_TAGS, MBUF_TAG_MEM_NAME,
77     "packet-attached information");
78 
79 /* can't call it m_dup(), as freebsd[34] uses m_dup() with different arg */
80 static struct mbuf *m_dup1(struct mbuf *, int, int, int);
81 
82 /*
83  * ensure that [off, off + len) is contiguous on the mbuf chain "m".
84  * packet chain before "off" is kept untouched.
85  * if offp == NULL, the target will start at <retval, 0> on resulting chain.
86  * if offp != NULL, the target will start at <retval, *offp> on resulting chain.
87  *
88  * on error return (NULL return value), original "m" will be freed.
89  *
90  * XXX: M_TRAILINGSPACE/M_LEADINGSPACE only permitted on writable ext_buf.
91  */
92 struct mbuf *
93 m_pulldown(struct mbuf *m, int off, int len, int *offp)
94 {
95 	struct mbuf *n, *o;
96 	int hlen, tlen, olen;
97 	int writable;
98 
99 	/* check invalid arguments. */
100 	KASSERT(m != NULL, ("%s: fix caller: m is NULL off %d len %d offp %p\n",
101 	    __func__, off, len, offp));
102 	if (len > MCLBYTES) {
103 		m_freem(m);
104 		return NULL;	/* impossible */
105 	}
106 
107 #ifdef PULLDOWN_DEBUG
108     {
109 	struct mbuf *t;
110 	printf("before:");
111 	for (t = m; t; t = t->m_next)
112 		printf(" %d", t->m_len);
113 	printf("\n");
114     }
115 #endif
116 	n = m;
117 	while (n != NULL && off > 0) {
118 		if (n->m_len > off)
119 			break;
120 		off -= n->m_len;
121 		n = n->m_next;
122 	}
123 	/* be sure to point non-empty mbuf */
124 	while (n != NULL && n->m_len == 0)
125 		n = n->m_next;
126 	if (!n) {
127 		m_freem(m);
128 		return NULL;	/* mbuf chain too short */
129 	}
130 
131 	/*
132 	 * The following comment is dated but still partially applies:
133 	 *
134 	 * XXX: This code is flawed because it considers a "writable" mbuf
135 	 *      data region to require all of the following:
136 	 *	  (i) mbuf _has_ to have M_EXT set; if it is just a regular
137 	 *	      mbuf, it is still not considered "writable."
138 	 *	  (ii) since mbuf has M_EXT, the ext_type _has_ to be
139 	 *	       EXT_CLUSTER. Anything else makes it non-writable.
140 	 *	  (iii) M_WRITABLE() must evaluate true.
141 	 *      Ideally, the requirement should only be (iii).
142 	 *
143 	 * If we're writable, we're sure we're writable, because the ref. count
144 	 * cannot increase from 1, as that would require possession of mbuf
145 	 * n by someone else (which is impossible). However, if we're _not_
146 	 * writable, we may eventually become writable )if the ref. count drops
147 	 * to 1), but we'll fail to notice it unless we re-evaluate
148 	 * M_WRITABLE(). For now, we only evaluate once at the beginning and
149 	 * live with this.
150 	 */
151 	writable = 0;
152 	if ((n->m_flags & M_EXT) == 0 ||
153 	    (n->m_ext.ext_type == EXT_CLUSTER && M_WRITABLE(n)))
154 		writable = 1;
155 
156 	/*
157 	 * the target data is on <n, off>.
158 	 * if we got enough data on the mbuf "n", we're done.
159 	 */
160 	if ((off == 0 || offp) && len <= n->m_len - off)
161 		goto ok;
162 
163 	/*
164 	 * when len <= n->m_len - off and off != 0, it is a special case.
165 	 * len bytes from <n, off> sits in single mbuf, but the caller does
166 	 * not like the starting position (off).
167 	 * chop the current mbuf into two pieces, set off to 0.
168 	 */
169 	if (len <= n->m_len - off) {
170 		o = m_dup1(n, off, n->m_len - off, M_NOWAIT);
171 		if (o == NULL) {
172 			m_freem(m);
173 			return NULL;	/* ENOBUFS */
174 		}
175 		n->m_len = off;
176 		o->m_next = n->m_next;
177 		n->m_next = o;
178 		n = n->m_next;
179 		off = 0;
180 		goto ok;
181 	}
182 
183 	/*
184 	 * we need to take hlen from <n, off> and tlen from <n->m_next, 0>,
185 	 * and construct contiguous mbuf with m_len == len.
186 	 * note that hlen + tlen == len, and tlen > 0.
187 	 */
188 	hlen = n->m_len - off;
189 	tlen = len - hlen;
190 
191 	/*
192 	 * ensure that we have enough trailing data on mbuf chain.
193 	 * if not, we can do nothing about the chain.
194 	 */
195 	olen = 0;
196 	for (o = n->m_next; o != NULL; o = o->m_next)
197 		olen += o->m_len;
198 	if (hlen + olen < len) {
199 		m_freem(m);
200 		return NULL;	/* mbuf chain too short */
201 	}
202 
203 	/*
204 	 * easy cases first.
205 	 * we need to use m_copydata() to get data from <n->m_next, 0>.
206 	 */
207 	if ((off == 0 || offp) && M_TRAILINGSPACE(n) >= tlen
208 	 && writable) {
209 		m_copydata(n->m_next, 0, tlen, mtod(n, caddr_t) + n->m_len);
210 		n->m_len += tlen;
211 		m_adj(n->m_next, tlen);
212 		goto ok;
213 	}
214 	if ((off == 0 || offp) && M_LEADINGSPACE(n->m_next) >= hlen
215 	 && writable && n->m_next->m_len >= tlen) {
216 		n->m_next->m_data -= hlen;
217 		n->m_next->m_len += hlen;
218 		bcopy(mtod(n, caddr_t) + off, mtod(n->m_next, caddr_t), hlen);
219 		n->m_len -= hlen;
220 		n = n->m_next;
221 		off = 0;
222 		goto ok;
223 	}
224 
225 	/*
226 	 * now, we need to do the hard way.  don't m_copy as there's no room
227 	 * on both end.
228 	 */
229 	if (len > MLEN)
230 		o = m_getcl(M_NOWAIT, m->m_type, 0);
231 	else
232 		o = m_get(M_NOWAIT, m->m_type);
233 	if (!o) {
234 		m_freem(m);
235 		return NULL;	/* ENOBUFS */
236 	}
237 	/* get hlen from <n, off> into <o, 0> */
238 	o->m_len = hlen;
239 	bcopy(mtod(n, caddr_t) + off, mtod(o, caddr_t), hlen);
240 	n->m_len -= hlen;
241 	/* get tlen from <n->m_next, 0> into <o, hlen> */
242 	m_copydata(n->m_next, 0, tlen, mtod(o, caddr_t) + o->m_len);
243 	o->m_len += tlen;
244 	m_adj(n->m_next, tlen);
245 	o->m_next = n->m_next;
246 	n->m_next = o;
247 	n = o;
248 	off = 0;
249 
250 ok:
251 #ifdef PULLDOWN_DEBUG
252     {
253 	struct mbuf *t;
254 	printf("after:");
255 	for (t = m; t; t = t->m_next)
256 		printf("%c%d", t == n ? '*' : ' ', t->m_len);
257 	printf(" (off=%d)\n", off);
258     }
259 #endif
260 	if (offp)
261 		*offp = off;
262 	return n;
263 }
264 
265 static struct mbuf *
266 m_dup1(struct mbuf *m, int off, int len, int wait)
267 {
268 	struct mbuf *n;
269 	int copyhdr;
270 
271 	if (len > MCLBYTES)
272 		return NULL;
273 	if (off == 0 && (m->m_flags & M_PKTHDR) != 0)
274 		copyhdr = 1;
275 	else
276 		copyhdr = 0;
277 	if (len >= MINCLSIZE) {
278 		if (copyhdr == 1)
279 			n = m_getcl(wait, m->m_type, M_PKTHDR);
280 		else
281 			n = m_getcl(wait, m->m_type, 0);
282 	} else {
283 		if (copyhdr == 1)
284 			n = m_gethdr(wait, m->m_type);
285 		else
286 			n = m_get(wait, m->m_type);
287 	}
288 	if (!n)
289 		return NULL; /* ENOBUFS */
290 
291 	if (copyhdr && !m_dup_pkthdr(n, m, wait)) {
292 		m_free(n);
293 		return NULL;
294 	}
295 	m_copydata(m, off, len, mtod(n, caddr_t));
296 	n->m_len = len;
297 	return n;
298 }
299 
300 /* Free a packet tag. */
301 void
302 m_tag_free_default(struct m_tag *t)
303 {
304 #ifdef MAC
305 	if (t->m_tag_id == PACKET_TAG_MACLABEL)
306 		mac_mbuf_tag_destroy(t);
307 #endif
308 	free(t, M_PACKET_TAGS);
309 }
310 
311 /* Get a packet tag structure along with specified data following. */
312 struct m_tag *
313 m_tag_alloc(uint32_t cookie, uint16_t type, int len, int wait)
314 {
315 	struct m_tag *t;
316 
317 	MBUF_CHECKSLEEP(wait);
318 	if (len < 0)
319 		return NULL;
320 	t = malloc(len + sizeof(struct m_tag), M_PACKET_TAGS, wait);
321 	if (t == NULL)
322 		return NULL;
323 	m_tag_setup(t, cookie, type, len);
324 	t->m_tag_free = m_tag_free_default;
325 	return t;
326 }
327 
328 /* Unlink and free a packet tag. */
329 void
330 m_tag_delete(struct mbuf *m, struct m_tag *t)
331 {
332 
333 	KASSERT(m && t, ("m_tag_delete: null argument, m %p t %p", m, t));
334 	m_tag_unlink(m, t);
335 	m_tag_free(t);
336 }
337 
338 /* Unlink and free a packet tag chain, starting from given tag. */
339 void
340 m_tag_delete_chain(struct mbuf *m, struct m_tag *t)
341 {
342 	struct m_tag *p, *q;
343 
344 	KASSERT(m, ("m_tag_delete_chain: null mbuf"));
345 	if (t != NULL)
346 		p = t;
347 	else
348 		p = SLIST_FIRST(&m->m_pkthdr.tags);
349 	if (p == NULL)
350 		return;
351 	while ((q = SLIST_NEXT(p, m_tag_link)) != NULL)
352 		m_tag_delete(m, q);
353 	m_tag_delete(m, p);
354 }
355 
356 /*
357  * Strip off all tags that would normally vanish when
358  * passing through a network interface.  Only persistent
359  * tags will exist after this; these are expected to remain
360  * so long as the mbuf chain exists, regardless of the
361  * path the mbufs take.
362  */
363 void
364 m_tag_delete_nonpersistent(struct mbuf *m)
365 {
366 	struct m_tag *p, *q;
367 
368 	SLIST_FOREACH_SAFE(p, &m->m_pkthdr.tags, m_tag_link, q)
369 		if ((p->m_tag_id & MTAG_PERSISTENT) == 0)
370 			m_tag_delete(m, p);
371 }
372 
373 /* Find a tag, starting from a given position. */
374 struct m_tag *
375 m_tag_locate(struct mbuf *m, uint32_t cookie, uint16_t type, struct m_tag *t)
376 {
377 	struct m_tag *p;
378 
379 	KASSERT(m, ("m_tag_locate: null mbuf"));
380 	if (t == NULL)
381 		p = SLIST_FIRST(&m->m_pkthdr.tags);
382 	else
383 		p = SLIST_NEXT(t, m_tag_link);
384 	while (p != NULL) {
385 		if (p->m_tag_cookie == cookie && p->m_tag_id == type)
386 			return p;
387 		p = SLIST_NEXT(p, m_tag_link);
388 	}
389 	return NULL;
390 }
391 
392 /* Copy a single tag. */
393 struct m_tag *
394 m_tag_copy(struct m_tag *t, int how)
395 {
396 	struct m_tag *p;
397 
398 	MBUF_CHECKSLEEP(how);
399 	KASSERT(t, ("m_tag_copy: null tag"));
400 	p = m_tag_alloc(t->m_tag_cookie, t->m_tag_id, t->m_tag_len, how);
401 	if (p == NULL)
402 		return (NULL);
403 #ifdef MAC
404 	/*
405 	 * XXXMAC: we should probably pass off the initialization, and
406 	 * copying here?  can we hide that PACKET_TAG_MACLABEL is
407 	 * special from the mbuf code?
408 	 */
409 	if (t->m_tag_id == PACKET_TAG_MACLABEL) {
410 		if (mac_mbuf_tag_init(p, how) != 0) {
411 			m_tag_free(p);
412 			return (NULL);
413 		}
414 		mac_mbuf_tag_copy(t, p);
415 	} else
416 #endif
417 		bcopy(t + 1, p + 1, t->m_tag_len); /* Copy the data */
418 	return p;
419 }
420 
421 /*
422  * Copy two tag chains. The destination mbuf (to) loses any attached
423  * tags even if the operation fails. This should not be a problem, as
424  * m_tag_copy_chain() is typically called with a newly-allocated
425  * destination mbuf.
426  */
427 int
428 m_tag_copy_chain(struct mbuf *to, const struct mbuf *from, int how)
429 {
430 	struct m_tag *p, *t, *tprev = NULL;
431 
432 	MBUF_CHECKSLEEP(how);
433 	KASSERT(to && from,
434 		("m_tag_copy_chain: null argument, to %p from %p", to, from));
435 	m_tag_delete_chain(to, NULL);
436 	SLIST_FOREACH(p, &from->m_pkthdr.tags, m_tag_link) {
437 		t = m_tag_copy(p, how);
438 		if (t == NULL) {
439 			m_tag_delete_chain(to, NULL);
440 			return 0;
441 		}
442 		if (tprev == NULL)
443 			SLIST_INSERT_HEAD(&to->m_pkthdr.tags, t, m_tag_link);
444 		else
445 			SLIST_INSERT_AFTER(tprev, t, m_tag_link);
446 		tprev = t;
447 	}
448 	return 1;
449 }
450