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