uipc_mbuf.c revision 100960
1/* 2 * Copyright (c) 1982, 1986, 1988, 1991, 1993 3 * The Regents of the University of California. All rights reserved. 4 * 5 * Redistribution and use in source and binary forms, with or without 6 * modification, are permitted provided that the following conditions 7 * are met: 8 * 1. Redistributions of source code must retain the above copyright 9 * notice, this list of conditions and the following disclaimer. 10 * 2. Redistributions in binary form must reproduce the above copyright 11 * notice, this list of conditions and the following disclaimer in the 12 * documentation and/or other materials provided with the distribution. 13 * 3. All advertising materials mentioning features or use of this software 14 * must display the following acknowledgement: 15 * This product includes software developed by the University of 16 * California, Berkeley and its contributors. 17 * 4. Neither the name of the University nor the names of its contributors 18 * may be used to endorse or promote products derived from this software 19 * without specific prior written permission. 20 * 21 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 22 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 23 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 24 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 25 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 26 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 27 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 28 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 29 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 30 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 31 * SUCH DAMAGE. 32 * 33 * @(#)uipc_mbuf.c 8.2 (Berkeley) 1/4/94 34 * $FreeBSD: head/sys/kern/uipc_mbuf.c 100960 2002-07-30 18:28:58Z rwatson $ 35 */ 36 37#include "opt_param.h" 38#include <sys/param.h> 39#include <sys/systm.h> 40#include <sys/kernel.h> 41#include <sys/lock.h> 42#include <sys/malloc.h> 43#include <sys/mbuf.h> 44#include <sys/sysctl.h> 45#include <sys/domain.h> 46#include <sys/protosw.h> 47 48int max_linkhdr; 49int max_protohdr; 50int max_hdr; 51int max_datalen; 52 53/* 54 * sysctl(8) exported objects 55 */ 56SYSCTL_DECL(_kern_ipc); 57SYSCTL_INT(_kern_ipc, KIPC_MAX_LINKHDR, max_linkhdr, CTLFLAG_RW, 58 &max_linkhdr, 0, ""); 59SYSCTL_INT(_kern_ipc, KIPC_MAX_PROTOHDR, max_protohdr, CTLFLAG_RW, 60 &max_protohdr, 0, ""); 61SYSCTL_INT(_kern_ipc, KIPC_MAX_HDR, max_hdr, CTLFLAG_RW, &max_hdr, 0, ""); 62SYSCTL_INT(_kern_ipc, KIPC_MAX_DATALEN, max_datalen, CTLFLAG_RW, 63 &max_datalen, 0, ""); 64 65/* 66 * Copy mbuf pkthdr from "from" to "to". 67 * "from" must have M_PKTHDR set, and "to" must be empty. 68 * aux pointer will be moved to "to". 69 */ 70void 71m_copy_pkthdr(struct mbuf *to, struct mbuf *from) 72{ 73 74#if 0 75 KASSERT(to->m_flags & M_PKTHDR, 76 ("m_copy_pkthdr() called on non-header")); 77#endif 78 to->m_data = to->m_pktdat; 79 to->m_flags = from->m_flags & M_COPYFLAGS; 80 to->m_pkthdr = from->m_pkthdr; 81 from->m_pkthdr.aux = NULL; 82} 83 84/* 85 * Lesser-used path for M_PREPEND: 86 * allocate new mbuf to prepend to chain, 87 * copy junk along. 88 */ 89struct mbuf * 90m_prepend(struct mbuf *m, int len, int how) 91{ 92 struct mbuf *mn; 93 94 MGET(mn, how, m->m_type); 95 if (mn == NULL) { 96 m_freem(m); 97 return (NULL); 98 } 99 if (m->m_flags & M_PKTHDR) { 100 M_COPY_PKTHDR(mn, m); 101 m->m_flags &= ~M_PKTHDR; 102 } 103 mn->m_next = m; 104 m = mn; 105 if (len < MHLEN) 106 MH_ALIGN(m, len); 107 m->m_len = len; 108 return (m); 109} 110 111/* 112 * Make a copy of an mbuf chain starting "off0" bytes from the beginning, 113 * continuing for "len" bytes. If len is M_COPYALL, copy to end of mbuf. 114 * The wait parameter is a choice of M_TRYWAIT/M_DONTWAIT from caller. 115 * Note that the copy is read-only, because clusters are not copied, 116 * only their reference counts are incremented. 117 */ 118struct mbuf * 119m_copym(struct mbuf *m, int off0, int len, int wait) 120{ 121 struct mbuf *n, **np; 122 int off = off0; 123 struct mbuf *top; 124 int copyhdr = 0; 125 126 KASSERT(off >= 0, ("m_copym, negative off %d", off)); 127 KASSERT(len >= 0, ("m_copym, negative len %d", len)); 128 if (off == 0 && m->m_flags & M_PKTHDR) 129 copyhdr = 1; 130 while (off > 0) { 131 KASSERT(m != NULL, ("m_copym, offset > size of mbuf chain")); 132 if (off < m->m_len) 133 break; 134 off -= m->m_len; 135 m = m->m_next; 136 } 137 np = ⊤ 138 top = 0; 139 while (len > 0) { 140 if (m == NULL) { 141 KASSERT(len == M_COPYALL, 142 ("m_copym, length > size of mbuf chain")); 143 break; 144 } 145 MGET(n, wait, m->m_type); 146 *np = n; 147 if (n == NULL) 148 goto nospace; 149 if (copyhdr) { 150 M_COPY_PKTHDR(n, m); 151 if (len == M_COPYALL) 152 n->m_pkthdr.len -= off0; 153 else 154 n->m_pkthdr.len = len; 155 copyhdr = 0; 156 } 157 n->m_len = min(len, m->m_len - off); 158 if (m->m_flags & M_EXT) { 159 n->m_data = m->m_data + off; 160 n->m_ext = m->m_ext; 161 n->m_flags |= M_EXT; 162 MEXT_ADD_REF(m); 163 } else 164 bcopy(mtod(m, caddr_t)+off, mtod(n, caddr_t), 165 (unsigned)n->m_len); 166 if (len != M_COPYALL) 167 len -= n->m_len; 168 off = 0; 169 m = m->m_next; 170 np = &n->m_next; 171 } 172 if (top == NULL) 173 mbstat.m_mcfail++; /* XXX: No consistency. */ 174 175 return (top); 176nospace: 177 m_freem(top); 178 mbstat.m_mcfail++; /* XXX: No consistency. */ 179 return (NULL); 180} 181 182/* 183 * Copy an entire packet, including header (which must be present). 184 * An optimization of the common case `m_copym(m, 0, M_COPYALL, how)'. 185 * Note that the copy is read-only, because clusters are not copied, 186 * only their reference counts are incremented. 187 * Preserve alignment of the first mbuf so if the creator has left 188 * some room at the beginning (e.g. for inserting protocol headers) 189 * the copies still have the room available. 190 */ 191struct mbuf * 192m_copypacket(struct mbuf *m, int how) 193{ 194 struct mbuf *top, *n, *o; 195 196 MGET(n, how, m->m_type); 197 top = n; 198 if (n == NULL) 199 goto nospace; 200 201 M_COPY_PKTHDR(n, m); 202 n->m_len = m->m_len; 203 if (m->m_flags & M_EXT) { 204 n->m_data = m->m_data; 205 n->m_ext = m->m_ext; 206 n->m_flags |= M_EXT; 207 MEXT_ADD_REF(m); 208 } else { 209 n->m_data = n->m_pktdat + (m->m_data - m->m_pktdat ); 210 bcopy(mtod(m, char *), mtod(n, char *), n->m_len); 211 } 212 213 m = m->m_next; 214 while (m) { 215 MGET(o, how, m->m_type); 216 if (o == NULL) 217 goto nospace; 218 219 n->m_next = o; 220 n = n->m_next; 221 222 n->m_len = m->m_len; 223 if (m->m_flags & M_EXT) { 224 n->m_data = m->m_data; 225 n->m_ext = m->m_ext; 226 n->m_flags |= M_EXT; 227 MEXT_ADD_REF(m); 228 } else { 229 bcopy(mtod(m, char *), mtod(n, char *), n->m_len); 230 } 231 232 m = m->m_next; 233 } 234 return top; 235nospace: 236 m_freem(top); 237 mbstat.m_mcfail++; /* XXX: No consistency. */ 238 return (NULL); 239} 240 241/* 242 * Copy data from an mbuf chain starting "off" bytes from the beginning, 243 * continuing for "len" bytes, into the indicated buffer. 244 */ 245void 246m_copydata(const struct mbuf *m, int off, int len, caddr_t cp) 247{ 248 unsigned count; 249 250 KASSERT(off >= 0, ("m_copydata, negative off %d", off)); 251 KASSERT(len >= 0, ("m_copydata, negative len %d", len)); 252 while (off > 0) { 253 KASSERT(m != NULL, ("m_copydata, offset > size of mbuf chain")); 254 if (off < m->m_len) 255 break; 256 off -= m->m_len; 257 m = m->m_next; 258 } 259 while (len > 0) { 260 KASSERT(m != NULL, ("m_copydata, length > size of mbuf chain")); 261 count = min(m->m_len - off, len); 262 bcopy(mtod(m, caddr_t) + off, cp, count); 263 len -= count; 264 cp += count; 265 off = 0; 266 m = m->m_next; 267 } 268} 269 270/* 271 * Copy a packet header mbuf chain into a completely new chain, including 272 * copying any mbuf clusters. Use this instead of m_copypacket() when 273 * you need a writable copy of an mbuf chain. 274 */ 275struct mbuf * 276m_dup(struct mbuf *m, int how) 277{ 278 struct mbuf **p, *top = NULL; 279 int remain, moff, nsize; 280 281 /* Sanity check */ 282 if (m == NULL) 283 return (NULL); 284 KASSERT((m->m_flags & M_PKTHDR) != 0, ("%s: !PKTHDR", __func__)); 285 286 /* While there's more data, get a new mbuf, tack it on, and fill it */ 287 remain = m->m_pkthdr.len; 288 moff = 0; 289 p = ⊤ 290 while (remain > 0 || top == NULL) { /* allow m->m_pkthdr.len == 0 */ 291 struct mbuf *n; 292 293 /* Get the next new mbuf */ 294 MGET(n, how, m->m_type); 295 if (n == NULL) 296 goto nospace; 297 if (top == NULL) { /* first one, must be PKTHDR */ 298 M_COPY_PKTHDR(n, m); 299 nsize = MHLEN; 300 } else /* not the first one */ 301 nsize = MLEN; 302 if (remain >= MINCLSIZE) { 303 MCLGET(n, how); 304 if ((n->m_flags & M_EXT) == 0) { 305 (void)m_free(n); 306 goto nospace; 307 } 308 nsize = MCLBYTES; 309 } 310 n->m_len = 0; 311 312 /* Link it into the new chain */ 313 *p = n; 314 p = &n->m_next; 315 316 /* Copy data from original mbuf(s) into new mbuf */ 317 while (n->m_len < nsize && m != NULL) { 318 int chunk = min(nsize - n->m_len, m->m_len - moff); 319 320 bcopy(m->m_data + moff, n->m_data + n->m_len, chunk); 321 moff += chunk; 322 n->m_len += chunk; 323 remain -= chunk; 324 if (moff == m->m_len) { 325 m = m->m_next; 326 moff = 0; 327 } 328 } 329 330 /* Check correct total mbuf length */ 331 KASSERT((remain > 0 && m != NULL) || (remain == 0 && m == NULL), 332 ("%s: bogus m_pkthdr.len", __func__)); 333 } 334 return (top); 335 336nospace: 337 m_freem(top); 338 mbstat.m_mcfail++; /* XXX: No consistency. */ 339 return (NULL); 340} 341 342/* 343 * Concatenate mbuf chain n to m. 344 * Both chains must be of the same type (e.g. MT_DATA). 345 * Any m_pkthdr is not updated. 346 */ 347void 348m_cat(struct mbuf *m, struct mbuf *n) 349{ 350 while (m->m_next) 351 m = m->m_next; 352 while (n) { 353 if (m->m_flags & M_EXT || 354 m->m_data + m->m_len + n->m_len >= &m->m_dat[MLEN]) { 355 /* just join the two chains */ 356 m->m_next = n; 357 return; 358 } 359 /* splat the data from one into the other */ 360 bcopy(mtod(n, caddr_t), mtod(m, caddr_t) + m->m_len, 361 (u_int)n->m_len); 362 m->m_len += n->m_len; 363 n = m_free(n); 364 } 365} 366 367void 368m_adj(struct mbuf *mp, int req_len) 369{ 370 int len = req_len; 371 struct mbuf *m; 372 int count; 373 374 if ((m = mp) == NULL) 375 return; 376 if (len >= 0) { 377 /* 378 * Trim from head. 379 */ 380 while (m != NULL && len > 0) { 381 if (m->m_len <= len) { 382 len -= m->m_len; 383 m->m_len = 0; 384 m = m->m_next; 385 } else { 386 m->m_len -= len; 387 m->m_data += len; 388 len = 0; 389 } 390 } 391 m = mp; 392 if (mp->m_flags & M_PKTHDR) 393 m->m_pkthdr.len -= (req_len - len); 394 } else { 395 /* 396 * Trim from tail. Scan the mbuf chain, 397 * calculating its length and finding the last mbuf. 398 * If the adjustment only affects this mbuf, then just 399 * adjust and return. Otherwise, rescan and truncate 400 * after the remaining size. 401 */ 402 len = -len; 403 count = 0; 404 for (;;) { 405 count += m->m_len; 406 if (m->m_next == (struct mbuf *)0) 407 break; 408 m = m->m_next; 409 } 410 if (m->m_len >= len) { 411 m->m_len -= len; 412 if (mp->m_flags & M_PKTHDR) 413 mp->m_pkthdr.len -= len; 414 return; 415 } 416 count -= len; 417 if (count < 0) 418 count = 0; 419 /* 420 * Correct length for chain is "count". 421 * Find the mbuf with last data, adjust its length, 422 * and toss data from remaining mbufs on chain. 423 */ 424 m = mp; 425 if (m->m_flags & M_PKTHDR) 426 m->m_pkthdr.len = count; 427 for (; m; m = m->m_next) { 428 if (m->m_len >= count) { 429 m->m_len = count; 430 break; 431 } 432 count -= m->m_len; 433 } 434 while (m->m_next) 435 (m = m->m_next) ->m_len = 0; 436 } 437} 438 439/* 440 * Rearange an mbuf chain so that len bytes are contiguous 441 * and in the data area of an mbuf (so that mtod and dtom 442 * will work for a structure of size len). Returns the resulting 443 * mbuf chain on success, frees it and returns null on failure. 444 * If there is room, it will add up to max_protohdr-len extra bytes to the 445 * contiguous region in an attempt to avoid being called next time. 446 */ 447struct mbuf * 448m_pullup(struct mbuf *n, int len) 449{ 450 struct mbuf *m; 451 int count; 452 int space; 453 454 /* 455 * If first mbuf has no cluster, and has room for len bytes 456 * without shifting current data, pullup into it, 457 * otherwise allocate a new mbuf to prepend to the chain. 458 */ 459 if ((n->m_flags & M_EXT) == 0 && 460 n->m_data + len < &n->m_dat[MLEN] && n->m_next) { 461 if (n->m_len >= len) 462 return (n); 463 m = n; 464 n = n->m_next; 465 len -= m->m_len; 466 } else { 467 if (len > MHLEN) 468 goto bad; 469 MGET(m, M_DONTWAIT, n->m_type); 470 if (m == NULL) 471 goto bad; 472 m->m_len = 0; 473 if (n->m_flags & M_PKTHDR) { 474 M_COPY_PKTHDR(m, n); 475 n->m_flags &= ~M_PKTHDR; 476 } 477 } 478 space = &m->m_dat[MLEN] - (m->m_data + m->m_len); 479 do { 480 count = min(min(max(len, max_protohdr), space), n->m_len); 481 bcopy(mtod(n, caddr_t), mtod(m, caddr_t) + m->m_len, 482 (unsigned)count); 483 len -= count; 484 m->m_len += count; 485 n->m_len -= count; 486 space -= count; 487 if (n->m_len) 488 n->m_data += count; 489 else 490 n = m_free(n); 491 } while (len > 0 && n); 492 if (len > 0) { 493 (void) m_free(m); 494 goto bad; 495 } 496 m->m_next = n; 497 return (m); 498bad: 499 m_freem(n); 500 mbstat.m_mpfail++; /* XXX: No consistency. */ 501 return (NULL); 502} 503 504/* 505 * Partition an mbuf chain in two pieces, returning the tail -- 506 * all but the first len0 bytes. In case of failure, it returns NULL and 507 * attempts to restore the chain to its original state. 508 * 509 * Note that the resulting mbufs might be read-only, because the new 510 * mbuf can end up sharing an mbuf cluster with the original mbuf if 511 * the "breaking point" happens to lie within a cluster mbuf. Use the 512 * M_WRITABLE() macro to check for this case. 513 */ 514struct mbuf * 515m_split(struct mbuf *m0, int len0, int wait) 516{ 517 struct mbuf *m, *n; 518 unsigned len = len0, remain; 519 520 for (m = m0; m && len > m->m_len; m = m->m_next) 521 len -= m->m_len; 522 if (m == NULL) 523 return (NULL); 524 remain = m->m_len - len; 525 if (m0->m_flags & M_PKTHDR) { 526 MGETHDR(n, wait, m0->m_type); 527 if (n == NULL) 528 return (NULL); 529 n->m_pkthdr.rcvif = m0->m_pkthdr.rcvif; 530 n->m_pkthdr.len = m0->m_pkthdr.len - len0; 531 m0->m_pkthdr.len = len0; 532 if (m->m_flags & M_EXT) 533 goto extpacket; 534 if (remain > MHLEN) { 535 /* m can't be the lead packet */ 536 MH_ALIGN(n, 0); 537 n->m_next = m_split(m, len, wait); 538 if (n->m_next == NULL) { 539 (void) m_free(n); 540 return (NULL); 541 } else { 542 n->m_len = 0; 543 return (n); 544 } 545 } else 546 MH_ALIGN(n, remain); 547 } else if (remain == 0) { 548 n = m->m_next; 549 m->m_next = NULL; 550 return (n); 551 } else { 552 MGET(n, wait, m->m_type); 553 if (n == NULL) 554 return (NULL); 555 M_ALIGN(n, remain); 556 } 557extpacket: 558 if (m->m_flags & M_EXT) { 559 n->m_flags |= M_EXT; 560 n->m_ext = m->m_ext; 561 MEXT_ADD_REF(m); 562 n->m_data = m->m_data + len; 563 } else { 564 bcopy(mtod(m, caddr_t) + len, mtod(n, caddr_t), remain); 565 } 566 n->m_len = remain; 567 m->m_len = len; 568 n->m_next = m->m_next; 569 m->m_next = NULL; 570 return (n); 571} 572/* 573 * Routine to copy from device local memory into mbufs. 574 * Note that `off' argument is offset into first mbuf of target chain from 575 * which to begin copying the data to. 576 */ 577struct mbuf * 578m_devget(char *buf, int totlen, int off, struct ifnet *ifp, 579 void (*copy)(char *from, caddr_t to, u_int len)) 580{ 581 struct mbuf *m; 582 struct mbuf *top = 0, **mp = ⊤ 583 int len; 584 585 if (off < 0 || off > MHLEN) 586 return (NULL); 587 588 MGETHDR(m, M_DONTWAIT, MT_DATA); 589 if (m == NULL) 590 return (NULL); 591 m->m_pkthdr.rcvif = ifp; 592 m->m_pkthdr.len = totlen; 593 len = MHLEN; 594 595 while (totlen > 0) { 596 if (top) { 597 MGET(m, M_DONTWAIT, MT_DATA); 598 if (m == NULL) { 599 m_freem(top); 600 return (NULL); 601 } 602 len = MLEN; 603 } 604 if (totlen + off >= MINCLSIZE) { 605 MCLGET(m, M_DONTWAIT); 606 if (m->m_flags & M_EXT) 607 len = MCLBYTES; 608 } else { 609 /* 610 * Place initial small packet/header at end of mbuf. 611 */ 612 if (top == NULL && totlen + off + max_linkhdr <= len) { 613 m->m_data += max_linkhdr; 614 len -= max_linkhdr; 615 } 616 } 617 if (off) { 618 m->m_data += off; 619 len -= off; 620 off = 0; 621 } 622 m->m_len = len = min(totlen, len); 623 if (copy) 624 copy(buf, mtod(m, caddr_t), (unsigned)len); 625 else 626 bcopy(buf, mtod(m, caddr_t), (unsigned)len); 627 buf += len; 628 *mp = m; 629 mp = &m->m_next; 630 totlen -= len; 631 } 632 return (top); 633} 634 635/* 636 * Copy data from a buffer back into the indicated mbuf chain, 637 * starting "off" bytes from the beginning, extending the mbuf 638 * chain if necessary. 639 */ 640void 641m_copyback(struct mbuf *m0, int off, int len, caddr_t cp) 642{ 643 int mlen; 644 struct mbuf *m = m0, *n; 645 int totlen = 0; 646 647 if (m0 == NULL) 648 return; 649 while (off > (mlen = m->m_len)) { 650 off -= mlen; 651 totlen += mlen; 652 if (m->m_next == NULL) { 653 n = m_get_clrd(M_DONTWAIT, m->m_type); 654 if (n == NULL) 655 goto out; 656 n->m_len = min(MLEN, len + off); 657 m->m_next = n; 658 } 659 m = m->m_next; 660 } 661 while (len > 0) { 662 mlen = min (m->m_len - off, len); 663 bcopy(cp, off + mtod(m, caddr_t), (unsigned)mlen); 664 cp += mlen; 665 len -= mlen; 666 mlen += off; 667 off = 0; 668 totlen += mlen; 669 if (len == 0) 670 break; 671 if (m->m_next == NULL) { 672 n = m_get(M_DONTWAIT, m->m_type); 673 if (n == NULL) 674 break; 675 n->m_len = min(MLEN, len); 676 m->m_next = n; 677 } 678 m = m->m_next; 679 } 680out: if (((m = m0)->m_flags & M_PKTHDR) && (m->m_pkthdr.len < totlen)) 681 m->m_pkthdr.len = totlen; 682} 683 684void 685m_print(const struct mbuf *m) 686{ 687 int len; 688 const struct mbuf *m2; 689 690 len = m->m_pkthdr.len; 691 m2 = m; 692 while (len) { 693 printf("%p %*D\n", m2, m2->m_len, (u_char *)m2->m_data, "-"); 694 len -= m2->m_len; 695 m2 = m2->m_next; 696 } 697 return; 698} 699