1 /*- 2 * SPDX-License-Identifier: BSD-3-Clause 3 * 4 * Copyright (c) 1982, 1986, 1988, 1991, 1993 5 * The Regents of the University of California. All rights reserved. 6 * 7 * Redistribution and use in source and binary forms, with or without 8 * modification, are permitted provided that the following conditions 9 * are met: 10 * 1. Redistributions of source code must retain the above copyright 11 * notice, this list of conditions and the following disclaimer. 12 * 2. Redistributions in binary form must reproduce the above copyright 13 * notice, this list of conditions and the following disclaimer in the 14 * documentation and/or other materials provided with the distribution. 15 * 3. Neither the name of the University nor the names of its contributors 16 * may be used to endorse or promote products derived from this software 17 * without specific prior written permission. 18 * 19 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 20 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 21 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 22 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 23 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 24 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 25 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 26 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 27 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 28 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 29 * SUCH DAMAGE. 30 * 31 * @(#)uipc_mbuf.c 8.2 (Berkeley) 1/4/94 32 */ 33 34 #include <sys/cdefs.h> 35 __FBSDID("$FreeBSD$"); 36 37 #include "opt_param.h" 38 #include "opt_mbuf_stress_test.h" 39 #include "opt_mbuf_profiling.h" 40 41 #include <sys/param.h> 42 #include <sys/systm.h> 43 #include <sys/kernel.h> 44 #include <sys/limits.h> 45 #include <sys/lock.h> 46 #include <sys/malloc.h> 47 #include <sys/mbuf.h> 48 #include <sys/sysctl.h> 49 #include <sys/domain.h> 50 #include <sys/protosw.h> 51 #include <sys/uio.h> 52 #include <sys/vmmeter.h> 53 #include <sys/sbuf.h> 54 #include <sys/sdt.h> 55 #include <vm/vm.h> 56 #include <vm/vm_pageout.h> 57 #include <vm/vm_page.h> 58 59 SDT_PROBE_DEFINE5_XLATE(sdt, , , m__init, 60 "struct mbuf *", "mbufinfo_t *", 61 "uint32_t", "uint32_t", 62 "uint16_t", "uint16_t", 63 "uint32_t", "uint32_t", 64 "uint32_t", "uint32_t"); 65 66 SDT_PROBE_DEFINE3_XLATE(sdt, , , m__gethdr_raw, 67 "uint32_t", "uint32_t", 68 "uint16_t", "uint16_t", 69 "struct mbuf *", "mbufinfo_t *"); 70 71 SDT_PROBE_DEFINE3_XLATE(sdt, , , m__gethdr, 72 "uint32_t", "uint32_t", 73 "uint16_t", "uint16_t", 74 "struct mbuf *", "mbufinfo_t *"); 75 76 SDT_PROBE_DEFINE3_XLATE(sdt, , , m__get_raw, 77 "uint32_t", "uint32_t", 78 "uint16_t", "uint16_t", 79 "struct mbuf *", "mbufinfo_t *"); 80 81 SDT_PROBE_DEFINE3_XLATE(sdt, , , m__get, 82 "uint32_t", "uint32_t", 83 "uint16_t", "uint16_t", 84 "struct mbuf *", "mbufinfo_t *"); 85 86 SDT_PROBE_DEFINE4_XLATE(sdt, , , m__getcl, 87 "uint32_t", "uint32_t", 88 "uint16_t", "uint16_t", 89 "uint32_t", "uint32_t", 90 "struct mbuf *", "mbufinfo_t *"); 91 92 SDT_PROBE_DEFINE5_XLATE(sdt, , , m__getjcl, 93 "uint32_t", "uint32_t", 94 "uint16_t", "uint16_t", 95 "uint32_t", "uint32_t", 96 "uint32_t", "uint32_t", 97 "struct mbuf *", "mbufinfo_t *"); 98 99 SDT_PROBE_DEFINE3_XLATE(sdt, , , m__clget, 100 "struct mbuf *", "mbufinfo_t *", 101 "uint32_t", "uint32_t", 102 "uint32_t", "uint32_t"); 103 104 SDT_PROBE_DEFINE4_XLATE(sdt, , , m__cljget, 105 "struct mbuf *", "mbufinfo_t *", 106 "uint32_t", "uint32_t", 107 "uint32_t", "uint32_t", 108 "void*", "void*"); 109 110 SDT_PROBE_DEFINE(sdt, , , m__cljset); 111 112 SDT_PROBE_DEFINE1_XLATE(sdt, , , m__free, 113 "struct mbuf *", "mbufinfo_t *"); 114 115 SDT_PROBE_DEFINE1_XLATE(sdt, , , m__freem, 116 "struct mbuf *", "mbufinfo_t *"); 117 118 #include <security/mac/mac_framework.h> 119 120 int max_linkhdr; 121 int max_protohdr; 122 int max_hdr; 123 int max_datalen; 124 #ifdef MBUF_STRESS_TEST 125 int m_defragpackets; 126 int m_defragbytes; 127 int m_defraguseless; 128 int m_defragfailure; 129 int m_defragrandomfailures; 130 #endif 131 132 /* 133 * sysctl(8) exported objects 134 */ 135 SYSCTL_INT(_kern_ipc, KIPC_MAX_LINKHDR, max_linkhdr, CTLFLAG_RD, 136 &max_linkhdr, 0, "Size of largest link layer header"); 137 SYSCTL_INT(_kern_ipc, KIPC_MAX_PROTOHDR, max_protohdr, CTLFLAG_RD, 138 &max_protohdr, 0, "Size of largest protocol layer header"); 139 SYSCTL_INT(_kern_ipc, KIPC_MAX_HDR, max_hdr, CTLFLAG_RD, 140 &max_hdr, 0, "Size of largest link plus protocol header"); 141 SYSCTL_INT(_kern_ipc, KIPC_MAX_DATALEN, max_datalen, CTLFLAG_RD, 142 &max_datalen, 0, "Minimum space left in mbuf after max_hdr"); 143 #ifdef MBUF_STRESS_TEST 144 SYSCTL_INT(_kern_ipc, OID_AUTO, m_defragpackets, CTLFLAG_RD, 145 &m_defragpackets, 0, ""); 146 SYSCTL_INT(_kern_ipc, OID_AUTO, m_defragbytes, CTLFLAG_RD, 147 &m_defragbytes, 0, ""); 148 SYSCTL_INT(_kern_ipc, OID_AUTO, m_defraguseless, CTLFLAG_RD, 149 &m_defraguseless, 0, ""); 150 SYSCTL_INT(_kern_ipc, OID_AUTO, m_defragfailure, CTLFLAG_RD, 151 &m_defragfailure, 0, ""); 152 SYSCTL_INT(_kern_ipc, OID_AUTO, m_defragrandomfailures, CTLFLAG_RW, 153 &m_defragrandomfailures, 0, ""); 154 #endif 155 156 /* 157 * Ensure the correct size of various mbuf parameters. It could be off due 158 * to compiler-induced padding and alignment artifacts. 159 */ 160 CTASSERT(MSIZE - offsetof(struct mbuf, m_dat) == MLEN); 161 CTASSERT(MSIZE - offsetof(struct mbuf, m_pktdat) == MHLEN); 162 163 /* 164 * mbuf data storage should be 64-bit aligned regardless of architectural 165 * pointer size; check this is the case with and without a packet header. 166 */ 167 CTASSERT(offsetof(struct mbuf, m_dat) % 8 == 0); 168 CTASSERT(offsetof(struct mbuf, m_pktdat) % 8 == 0); 169 170 /* 171 * While the specific values here don't matter too much (i.e., +/- a few 172 * words), we do want to ensure that changes to these values are carefully 173 * reasoned about and properly documented. This is especially the case as 174 * network-protocol and device-driver modules encode these layouts, and must 175 * be recompiled if the structures change. Check these values at compile time 176 * against the ones documented in comments in mbuf.h. 177 * 178 * NB: Possibly they should be documented there via #define's and not just 179 * comments. 180 */ 181 #if defined(__LP64__) 182 CTASSERT(offsetof(struct mbuf, m_dat) == 32); 183 CTASSERT(sizeof(struct pkthdr) == 64); 184 CTASSERT(sizeof(struct m_ext) == 160); 185 #else 186 CTASSERT(offsetof(struct mbuf, m_dat) == 24); 187 CTASSERT(sizeof(struct pkthdr) == 56); 188 #if defined(__powerpc__) && defined(BOOKE) 189 /* PowerPC booke has 64-bit physical pointers. */ 190 CTASSERT(sizeof(struct m_ext) == 176); 191 #else 192 CTASSERT(sizeof(struct m_ext) == 172); 193 #endif 194 #endif 195 196 /* 197 * Assert that the queue(3) macros produce code of the same size as an old 198 * plain pointer does. 199 */ 200 #ifdef INVARIANTS 201 static struct mbuf __used m_assertbuf; 202 CTASSERT(sizeof(m_assertbuf.m_slist) == sizeof(m_assertbuf.m_next)); 203 CTASSERT(sizeof(m_assertbuf.m_stailq) == sizeof(m_assertbuf.m_next)); 204 CTASSERT(sizeof(m_assertbuf.m_slistpkt) == sizeof(m_assertbuf.m_nextpkt)); 205 CTASSERT(sizeof(m_assertbuf.m_stailqpkt) == sizeof(m_assertbuf.m_nextpkt)); 206 #endif 207 208 /* 209 * Attach the cluster from *m to *n, set up m_ext in *n 210 * and bump the refcount of the cluster. 211 */ 212 void 213 mb_dupcl(struct mbuf *n, struct mbuf *m) 214 { 215 volatile u_int *refcnt; 216 217 KASSERT(m->m_flags & (M_EXT|M_EXTPG), 218 ("%s: M_EXT|M_EXTPG not set on %p", __func__, m)); 219 KASSERT(!(n->m_flags & (M_EXT|M_EXTPG)), 220 ("%s: M_EXT|M_EXTPG set on %p", __func__, n)); 221 222 /* 223 * Cache access optimization. 224 * 225 * o Regular M_EXT storage doesn't need full copy of m_ext, since 226 * the holder of the 'ext_count' is responsible to carry the free 227 * routine and its arguments. 228 * o M_EXTPG data is split between main part of mbuf and m_ext, the 229 * main part is copied in full, the m_ext part is similar to M_EXT. 230 * o EXT_EXTREF, where 'ext_cnt' doesn't point into mbuf at all, is 231 * special - it needs full copy of m_ext into each mbuf, since any 232 * copy could end up as the last to free. 233 */ 234 if (m->m_flags & M_EXTPG) { 235 bcopy(&m->m_epg_startcopy, &n->m_epg_startcopy, 236 __rangeof(struct mbuf, m_epg_startcopy, m_epg_endcopy)); 237 bcopy(&m->m_ext, &n->m_ext, m_epg_ext_copylen); 238 } else if (m->m_ext.ext_type == EXT_EXTREF) 239 bcopy(&m->m_ext, &n->m_ext, sizeof(struct m_ext)); 240 else 241 bcopy(&m->m_ext, &n->m_ext, m_ext_copylen); 242 243 n->m_flags |= m->m_flags & (M_RDONLY | M_EXT | M_EXTPG); 244 245 /* See if this is the mbuf that holds the embedded refcount. */ 246 if (m->m_ext.ext_flags & EXT_FLAG_EMBREF) { 247 refcnt = n->m_ext.ext_cnt = &m->m_ext.ext_count; 248 n->m_ext.ext_flags &= ~EXT_FLAG_EMBREF; 249 } else { 250 KASSERT(m->m_ext.ext_cnt != NULL, 251 ("%s: no refcounting pointer on %p", __func__, m)); 252 refcnt = m->m_ext.ext_cnt; 253 } 254 255 if (*refcnt == 1) 256 *refcnt += 1; 257 else 258 atomic_add_int(refcnt, 1); 259 } 260 261 void 262 m_demote_pkthdr(struct mbuf *m) 263 { 264 265 M_ASSERTPKTHDR(m); 266 M_ASSERT_NO_SND_TAG(m); 267 268 m_tag_delete_chain(m, NULL); 269 m->m_flags &= ~M_PKTHDR; 270 bzero(&m->m_pkthdr, sizeof(struct pkthdr)); 271 } 272 273 /* 274 * Clean up mbuf (chain) from any tags and packet headers. 275 * If "all" is set then the first mbuf in the chain will be 276 * cleaned too. 277 */ 278 void 279 m_demote(struct mbuf *m0, int all, int flags) 280 { 281 struct mbuf *m; 282 283 flags |= M_DEMOTEFLAGS; 284 285 for (m = all ? m0 : m0->m_next; m != NULL; m = m->m_next) { 286 KASSERT(m->m_nextpkt == NULL, ("%s: m_nextpkt in m %p, m0 %p", 287 __func__, m, m0)); 288 if (m->m_flags & M_PKTHDR) 289 m_demote_pkthdr(m); 290 m->m_flags &= flags; 291 } 292 } 293 294 /* 295 * Sanity checks on mbuf (chain) for use in KASSERT() and general 296 * debugging. 297 * Returns 0 or panics when bad and 1 on all tests passed. 298 * Sanitize, 0 to run M_SANITY_ACTION, 1 to garble things so they 299 * blow up later. 300 */ 301 int 302 m_sanity(struct mbuf *m0, int sanitize) 303 { 304 struct mbuf *m; 305 caddr_t a, b; 306 int pktlen = 0; 307 308 #ifdef INVARIANTS 309 #define M_SANITY_ACTION(s) panic("mbuf %p: " s, m) 310 #else 311 #define M_SANITY_ACTION(s) printf("mbuf %p: " s, m) 312 #endif 313 314 for (m = m0; m != NULL; m = m->m_next) { 315 /* 316 * Basic pointer checks. If any of these fails then some 317 * unrelated kernel memory before or after us is trashed. 318 * No way to recover from that. 319 */ 320 a = M_START(m); 321 b = a + M_SIZE(m); 322 if ((caddr_t)m->m_data < a) 323 M_SANITY_ACTION("m_data outside mbuf data range left"); 324 if ((caddr_t)m->m_data > b) 325 M_SANITY_ACTION("m_data outside mbuf data range right"); 326 if ((caddr_t)m->m_data + m->m_len > b) 327 M_SANITY_ACTION("m_data + m_len exeeds mbuf space"); 328 329 /* m->m_nextpkt may only be set on first mbuf in chain. */ 330 if (m != m0 && m->m_nextpkt != NULL) { 331 if (sanitize) { 332 m_freem(m->m_nextpkt); 333 m->m_nextpkt = (struct mbuf *)0xDEADC0DE; 334 } else 335 M_SANITY_ACTION("m->m_nextpkt on in-chain mbuf"); 336 } 337 338 /* packet length (not mbuf length!) calculation */ 339 if (m0->m_flags & M_PKTHDR) 340 pktlen += m->m_len; 341 342 /* m_tags may only be attached to first mbuf in chain. */ 343 if (m != m0 && m->m_flags & M_PKTHDR && 344 !SLIST_EMPTY(&m->m_pkthdr.tags)) { 345 if (sanitize) { 346 m_tag_delete_chain(m, NULL); 347 /* put in 0xDEADC0DE perhaps? */ 348 } else 349 M_SANITY_ACTION("m_tags on in-chain mbuf"); 350 } 351 352 /* M_PKTHDR may only be set on first mbuf in chain */ 353 if (m != m0 && m->m_flags & M_PKTHDR) { 354 if (sanitize) { 355 bzero(&m->m_pkthdr, sizeof(m->m_pkthdr)); 356 m->m_flags &= ~M_PKTHDR; 357 /* put in 0xDEADCODE and leave hdr flag in */ 358 } else 359 M_SANITY_ACTION("M_PKTHDR on in-chain mbuf"); 360 } 361 } 362 m = m0; 363 if (pktlen && pktlen != m->m_pkthdr.len) { 364 if (sanitize) 365 m->m_pkthdr.len = 0; 366 else 367 M_SANITY_ACTION("m_pkthdr.len != mbuf chain length"); 368 } 369 return 1; 370 371 #undef M_SANITY_ACTION 372 } 373 374 /* 375 * Non-inlined part of m_init(). 376 */ 377 int 378 m_pkthdr_init(struct mbuf *m, int how) 379 { 380 #ifdef MAC 381 int error; 382 #endif 383 m->m_data = m->m_pktdat; 384 bzero(&m->m_pkthdr, sizeof(m->m_pkthdr)); 385 #ifdef NUMA 386 m->m_pkthdr.numa_domain = M_NODOM; 387 #endif 388 #ifdef MAC 389 /* If the label init fails, fail the alloc */ 390 error = mac_mbuf_init(m, how); 391 if (error) 392 return (error); 393 #endif 394 395 return (0); 396 } 397 398 /* 399 * "Move" mbuf pkthdr from "from" to "to". 400 * "from" must have M_PKTHDR set, and "to" must be empty. 401 */ 402 void 403 m_move_pkthdr(struct mbuf *to, struct mbuf *from) 404 { 405 406 #if 0 407 /* see below for why these are not enabled */ 408 M_ASSERTPKTHDR(to); 409 /* Note: with MAC, this may not be a good assertion. */ 410 KASSERT(SLIST_EMPTY(&to->m_pkthdr.tags), 411 ("m_move_pkthdr: to has tags")); 412 #endif 413 #ifdef MAC 414 /* 415 * XXXMAC: It could be this should also occur for non-MAC? 416 */ 417 if (to->m_flags & M_PKTHDR) 418 m_tag_delete_chain(to, NULL); 419 #endif 420 to->m_flags = (from->m_flags & M_COPYFLAGS) | 421 (to->m_flags & (M_EXT | M_EXTPG)); 422 if ((to->m_flags & M_EXT) == 0) 423 to->m_data = to->m_pktdat; 424 to->m_pkthdr = from->m_pkthdr; /* especially tags */ 425 SLIST_INIT(&from->m_pkthdr.tags); /* purge tags from src */ 426 from->m_flags &= ~M_PKTHDR; 427 if (from->m_pkthdr.csum_flags & CSUM_SND_TAG) { 428 from->m_pkthdr.csum_flags &= ~CSUM_SND_TAG; 429 from->m_pkthdr.snd_tag = NULL; 430 } 431 } 432 433 /* 434 * Duplicate "from"'s mbuf pkthdr in "to". 435 * "from" must have M_PKTHDR set, and "to" must be empty. 436 * In particular, this does a deep copy of the packet tags. 437 */ 438 int 439 m_dup_pkthdr(struct mbuf *to, const struct mbuf *from, int how) 440 { 441 442 #if 0 443 /* 444 * The mbuf allocator only initializes the pkthdr 445 * when the mbuf is allocated with m_gethdr(). Many users 446 * (e.g. m_copy*, m_prepend) use m_get() and then 447 * smash the pkthdr as needed causing these 448 * assertions to trip. For now just disable them. 449 */ 450 M_ASSERTPKTHDR(to); 451 /* Note: with MAC, this may not be a good assertion. */ 452 KASSERT(SLIST_EMPTY(&to->m_pkthdr.tags), ("m_dup_pkthdr: to has tags")); 453 #endif 454 MBUF_CHECKSLEEP(how); 455 #ifdef MAC 456 if (to->m_flags & M_PKTHDR) 457 m_tag_delete_chain(to, NULL); 458 #endif 459 to->m_flags = (from->m_flags & M_COPYFLAGS) | 460 (to->m_flags & (M_EXT | M_EXTPG)); 461 if ((to->m_flags & M_EXT) == 0) 462 to->m_data = to->m_pktdat; 463 to->m_pkthdr = from->m_pkthdr; 464 if (from->m_pkthdr.csum_flags & CSUM_SND_TAG) 465 m_snd_tag_ref(from->m_pkthdr.snd_tag); 466 SLIST_INIT(&to->m_pkthdr.tags); 467 return (m_tag_copy_chain(to, from, how)); 468 } 469 470 /* 471 * Lesser-used path for M_PREPEND: 472 * allocate new mbuf to prepend to chain, 473 * copy junk along. 474 */ 475 struct mbuf * 476 m_prepend(struct mbuf *m, int len, int how) 477 { 478 struct mbuf *mn; 479 480 if (m->m_flags & M_PKTHDR) 481 mn = m_gethdr(how, m->m_type); 482 else 483 mn = m_get(how, m->m_type); 484 if (mn == NULL) { 485 m_freem(m); 486 return (NULL); 487 } 488 if (m->m_flags & M_PKTHDR) 489 m_move_pkthdr(mn, m); 490 mn->m_next = m; 491 m = mn; 492 if (len < M_SIZE(m)) 493 M_ALIGN(m, len); 494 m->m_len = len; 495 return (m); 496 } 497 498 /* 499 * Make a copy of an mbuf chain starting "off0" bytes from the beginning, 500 * continuing for "len" bytes. If len is M_COPYALL, copy to end of mbuf. 501 * The wait parameter is a choice of M_WAITOK/M_NOWAIT from caller. 502 * Note that the copy is read-only, because clusters are not copied, 503 * only their reference counts are incremented. 504 */ 505 struct mbuf * 506 m_copym(struct mbuf *m, int off0, int len, int wait) 507 { 508 struct mbuf *n, **np; 509 int off = off0; 510 struct mbuf *top; 511 int copyhdr = 0; 512 513 KASSERT(off >= 0, ("m_copym, negative off %d", off)); 514 KASSERT(len >= 0, ("m_copym, negative len %d", len)); 515 MBUF_CHECKSLEEP(wait); 516 if (off == 0 && m->m_flags & M_PKTHDR) 517 copyhdr = 1; 518 while (off > 0) { 519 KASSERT(m != NULL, ("m_copym, offset > size of mbuf chain")); 520 if (off < m->m_len) 521 break; 522 off -= m->m_len; 523 m = m->m_next; 524 } 525 np = ⊤ 526 top = NULL; 527 while (len > 0) { 528 if (m == NULL) { 529 KASSERT(len == M_COPYALL, 530 ("m_copym, length > size of mbuf chain")); 531 break; 532 } 533 if (copyhdr) 534 n = m_gethdr(wait, m->m_type); 535 else 536 n = m_get(wait, m->m_type); 537 *np = n; 538 if (n == NULL) 539 goto nospace; 540 if (copyhdr) { 541 if (!m_dup_pkthdr(n, m, wait)) 542 goto nospace; 543 if (len == M_COPYALL) 544 n->m_pkthdr.len -= off0; 545 else 546 n->m_pkthdr.len = len; 547 copyhdr = 0; 548 } 549 n->m_len = min(len, m->m_len - off); 550 if (m->m_flags & (M_EXT|M_EXTPG)) { 551 n->m_data = m->m_data + off; 552 mb_dupcl(n, m); 553 } else 554 bcopy(mtod(m, caddr_t)+off, mtod(n, caddr_t), 555 (u_int)n->m_len); 556 if (len != M_COPYALL) 557 len -= n->m_len; 558 off = 0; 559 m = m->m_next; 560 np = &n->m_next; 561 } 562 563 return (top); 564 nospace: 565 m_freem(top); 566 return (NULL); 567 } 568 569 /* 570 * Copy an entire packet, including header (which must be present). 571 * An optimization of the common case `m_copym(m, 0, M_COPYALL, how)'. 572 * Note that the copy is read-only, because clusters are not copied, 573 * only their reference counts are incremented. 574 * Preserve alignment of the first mbuf so if the creator has left 575 * some room at the beginning (e.g. for inserting protocol headers) 576 * the copies still have the room available. 577 */ 578 struct mbuf * 579 m_copypacket(struct mbuf *m, int how) 580 { 581 struct mbuf *top, *n, *o; 582 583 MBUF_CHECKSLEEP(how); 584 n = m_get(how, m->m_type); 585 top = n; 586 if (n == NULL) 587 goto nospace; 588 589 if (!m_dup_pkthdr(n, m, how)) 590 goto nospace; 591 n->m_len = m->m_len; 592 if (m->m_flags & (M_EXT|M_EXTPG)) { 593 n->m_data = m->m_data; 594 mb_dupcl(n, m); 595 } else { 596 n->m_data = n->m_pktdat + (m->m_data - m->m_pktdat ); 597 bcopy(mtod(m, char *), mtod(n, char *), n->m_len); 598 } 599 600 m = m->m_next; 601 while (m) { 602 o = m_get(how, m->m_type); 603 if (o == NULL) 604 goto nospace; 605 606 n->m_next = o; 607 n = n->m_next; 608 609 n->m_len = m->m_len; 610 if (m->m_flags & (M_EXT|M_EXTPG)) { 611 n->m_data = m->m_data; 612 mb_dupcl(n, m); 613 } else { 614 bcopy(mtod(m, char *), mtod(n, char *), n->m_len); 615 } 616 617 m = m->m_next; 618 } 619 return top; 620 nospace: 621 m_freem(top); 622 return (NULL); 623 } 624 625 static void 626 m_copyfromunmapped(const struct mbuf *m, int off, int len, caddr_t cp) 627 { 628 struct iovec iov; 629 struct uio uio; 630 int error __diagused; 631 632 KASSERT(off >= 0, ("m_copyfromunmapped: negative off %d", off)); 633 KASSERT(len >= 0, ("m_copyfromunmapped: negative len %d", len)); 634 KASSERT(off < m->m_len, 635 ("m_copyfromunmapped: len exceeds mbuf length")); 636 iov.iov_base = cp; 637 iov.iov_len = len; 638 uio.uio_resid = len; 639 uio.uio_iov = &iov; 640 uio.uio_segflg = UIO_SYSSPACE; 641 uio.uio_iovcnt = 1; 642 uio.uio_offset = 0; 643 uio.uio_rw = UIO_READ; 644 error = m_unmapped_uiomove(m, off, &uio, len); 645 KASSERT(error == 0, ("m_unmapped_uiomove failed: off %d, len %d", off, 646 len)); 647 } 648 649 /* 650 * Copy data from an mbuf chain starting "off" bytes from the beginning, 651 * continuing for "len" bytes, into the indicated buffer. 652 */ 653 void 654 m_copydata(const struct mbuf *m, int off, int len, caddr_t cp) 655 { 656 u_int count; 657 658 KASSERT(off >= 0, ("m_copydata, negative off %d", off)); 659 KASSERT(len >= 0, ("m_copydata, negative len %d", len)); 660 while (off > 0) { 661 KASSERT(m != NULL, ("m_copydata, offset > size of mbuf chain")); 662 if (off < m->m_len) 663 break; 664 off -= m->m_len; 665 m = m->m_next; 666 } 667 while (len > 0) { 668 KASSERT(m != NULL, ("m_copydata, length > size of mbuf chain")); 669 count = min(m->m_len - off, len); 670 if ((m->m_flags & M_EXTPG) != 0) 671 m_copyfromunmapped(m, off, count, cp); 672 else 673 bcopy(mtod(m, caddr_t) + off, cp, count); 674 len -= count; 675 cp += count; 676 off = 0; 677 m = m->m_next; 678 } 679 } 680 681 /* 682 * Copy a packet header mbuf chain into a completely new chain, including 683 * copying any mbuf clusters. Use this instead of m_copypacket() when 684 * you need a writable copy of an mbuf chain. 685 */ 686 struct mbuf * 687 m_dup(const struct mbuf *m, int how) 688 { 689 struct mbuf **p, *top = NULL; 690 int remain, moff, nsize; 691 692 MBUF_CHECKSLEEP(how); 693 /* Sanity check */ 694 if (m == NULL) 695 return (NULL); 696 M_ASSERTPKTHDR(m); 697 698 /* While there's more data, get a new mbuf, tack it on, and fill it */ 699 remain = m->m_pkthdr.len; 700 moff = 0; 701 p = ⊤ 702 while (remain > 0 || top == NULL) { /* allow m->m_pkthdr.len == 0 */ 703 struct mbuf *n; 704 705 /* Get the next new mbuf */ 706 if (remain >= MINCLSIZE) { 707 n = m_getcl(how, m->m_type, 0); 708 nsize = MCLBYTES; 709 } else { 710 n = m_get(how, m->m_type); 711 nsize = MLEN; 712 } 713 if (n == NULL) 714 goto nospace; 715 716 if (top == NULL) { /* First one, must be PKTHDR */ 717 if (!m_dup_pkthdr(n, m, how)) { 718 m_free(n); 719 goto nospace; 720 } 721 if ((n->m_flags & M_EXT) == 0) 722 nsize = MHLEN; 723 n->m_flags &= ~M_RDONLY; 724 } 725 n->m_len = 0; 726 727 /* Link it into the new chain */ 728 *p = n; 729 p = &n->m_next; 730 731 /* Copy data from original mbuf(s) into new mbuf */ 732 while (n->m_len < nsize && m != NULL) { 733 int chunk = min(nsize - n->m_len, m->m_len - moff); 734 735 m_copydata(m, moff, chunk, n->m_data + n->m_len); 736 moff += chunk; 737 n->m_len += chunk; 738 remain -= chunk; 739 if (moff == m->m_len) { 740 m = m->m_next; 741 moff = 0; 742 } 743 } 744 745 /* Check correct total mbuf length */ 746 KASSERT((remain > 0 && m != NULL) || (remain == 0 && m == NULL), 747 ("%s: bogus m_pkthdr.len", __func__)); 748 } 749 return (top); 750 751 nospace: 752 m_freem(top); 753 return (NULL); 754 } 755 756 /* 757 * Concatenate mbuf chain n to m. 758 * Both chains must be of the same type (e.g. MT_DATA). 759 * Any m_pkthdr is not updated. 760 */ 761 void 762 m_cat(struct mbuf *m, struct mbuf *n) 763 { 764 while (m->m_next) 765 m = m->m_next; 766 while (n) { 767 if (!M_WRITABLE(m) || 768 (n->m_flags & M_EXTPG) != 0 || 769 M_TRAILINGSPACE(m) < n->m_len) { 770 /* just join the two chains */ 771 m->m_next = n; 772 return; 773 } 774 /* splat the data from one into the other */ 775 bcopy(mtod(n, caddr_t), mtod(m, caddr_t) + m->m_len, 776 (u_int)n->m_len); 777 m->m_len += n->m_len; 778 n = m_free(n); 779 } 780 } 781 782 /* 783 * Concatenate two pkthdr mbuf chains. 784 */ 785 void 786 m_catpkt(struct mbuf *m, struct mbuf *n) 787 { 788 789 M_ASSERTPKTHDR(m); 790 M_ASSERTPKTHDR(n); 791 792 m->m_pkthdr.len += n->m_pkthdr.len; 793 m_demote(n, 1, 0); 794 795 m_cat(m, n); 796 } 797 798 void 799 m_adj(struct mbuf *mp, int req_len) 800 { 801 int len = req_len; 802 struct mbuf *m; 803 int count; 804 805 if ((m = mp) == NULL) 806 return; 807 if (len >= 0) { 808 /* 809 * Trim from head. 810 */ 811 while (m != NULL && len > 0) { 812 if (m->m_len <= len) { 813 len -= m->m_len; 814 m->m_len = 0; 815 m = m->m_next; 816 } else { 817 m->m_len -= len; 818 m->m_data += len; 819 len = 0; 820 } 821 } 822 if (mp->m_flags & M_PKTHDR) 823 mp->m_pkthdr.len -= (req_len - len); 824 } else { 825 /* 826 * Trim from tail. Scan the mbuf chain, 827 * calculating its length and finding the last mbuf. 828 * If the adjustment only affects this mbuf, then just 829 * adjust and return. Otherwise, rescan and truncate 830 * after the remaining size. 831 */ 832 len = -len; 833 count = 0; 834 for (;;) { 835 count += m->m_len; 836 if (m->m_next == (struct mbuf *)0) 837 break; 838 m = m->m_next; 839 } 840 if (m->m_len >= len) { 841 m->m_len -= len; 842 if (mp->m_flags & M_PKTHDR) 843 mp->m_pkthdr.len -= len; 844 return; 845 } 846 count -= len; 847 if (count < 0) 848 count = 0; 849 /* 850 * Correct length for chain is "count". 851 * Find the mbuf with last data, adjust its length, 852 * and toss data from remaining mbufs on chain. 853 */ 854 m = mp; 855 if (m->m_flags & M_PKTHDR) 856 m->m_pkthdr.len = count; 857 for (; m; m = m->m_next) { 858 if (m->m_len >= count) { 859 m->m_len = count; 860 if (m->m_next != NULL) { 861 m_freem(m->m_next); 862 m->m_next = NULL; 863 } 864 break; 865 } 866 count -= m->m_len; 867 } 868 } 869 } 870 871 void 872 m_adj_decap(struct mbuf *mp, int len) 873 { 874 uint8_t rsstype; 875 876 m_adj(mp, len); 877 if ((mp->m_flags & M_PKTHDR) != 0) { 878 /* 879 * If flowid was calculated by card from the inner 880 * headers, move flowid to the decapsulated mbuf 881 * chain, otherwise clear. This depends on the 882 * internals of m_adj, which keeps pkthdr as is, in 883 * particular not changing rsstype and flowid. 884 */ 885 rsstype = mp->m_pkthdr.rsstype; 886 if ((rsstype & M_HASHTYPE_INNER) != 0) { 887 M_HASHTYPE_SET(mp, rsstype & ~M_HASHTYPE_INNER); 888 } else { 889 M_HASHTYPE_CLEAR(mp); 890 } 891 } 892 } 893 894 /* 895 * Rearange an mbuf chain so that len bytes are contiguous 896 * and in the data area of an mbuf (so that mtod will work 897 * for a structure of size len). Returns the resulting 898 * mbuf chain on success, frees it and returns null on failure. 899 * If there is room, it will add up to max_protohdr-len extra bytes to the 900 * contiguous region in an attempt to avoid being called next time. 901 */ 902 struct mbuf * 903 m_pullup(struct mbuf *n, int len) 904 { 905 struct mbuf *m; 906 int count; 907 int space; 908 909 KASSERT((n->m_flags & M_EXTPG) == 0, 910 ("%s: unmapped mbuf %p", __func__, n)); 911 912 /* 913 * If first mbuf has no cluster, and has room for len bytes 914 * without shifting current data, pullup into it, 915 * otherwise allocate a new mbuf to prepend to the chain. 916 */ 917 if ((n->m_flags & M_EXT) == 0 && 918 n->m_data + len < &n->m_dat[MLEN] && n->m_next) { 919 if (n->m_len >= len) 920 return (n); 921 m = n; 922 n = n->m_next; 923 len -= m->m_len; 924 } else { 925 if (len > MHLEN) 926 goto bad; 927 m = m_get(M_NOWAIT, n->m_type); 928 if (m == NULL) 929 goto bad; 930 if (n->m_flags & M_PKTHDR) 931 m_move_pkthdr(m, n); 932 } 933 space = &m->m_dat[MLEN] - (m->m_data + m->m_len); 934 do { 935 count = min(min(max(len, max_protohdr), space), n->m_len); 936 bcopy(mtod(n, caddr_t), mtod(m, caddr_t) + m->m_len, 937 (u_int)count); 938 len -= count; 939 m->m_len += count; 940 n->m_len -= count; 941 space -= count; 942 if (n->m_len) 943 n->m_data += count; 944 else 945 n = m_free(n); 946 } while (len > 0 && n); 947 if (len > 0) { 948 (void) m_free(m); 949 goto bad; 950 } 951 m->m_next = n; 952 return (m); 953 bad: 954 m_freem(n); 955 return (NULL); 956 } 957 958 /* 959 * Like m_pullup(), except a new mbuf is always allocated, and we allow 960 * the amount of empty space before the data in the new mbuf to be specified 961 * (in the event that the caller expects to prepend later). 962 */ 963 struct mbuf * 964 m_copyup(struct mbuf *n, int len, int dstoff) 965 { 966 struct mbuf *m; 967 int count, space; 968 969 if (len > (MHLEN - dstoff)) 970 goto bad; 971 m = m_get(M_NOWAIT, n->m_type); 972 if (m == NULL) 973 goto bad; 974 if (n->m_flags & M_PKTHDR) 975 m_move_pkthdr(m, n); 976 m->m_data += dstoff; 977 space = &m->m_dat[MLEN] - (m->m_data + m->m_len); 978 do { 979 count = min(min(max(len, max_protohdr), space), n->m_len); 980 memcpy(mtod(m, caddr_t) + m->m_len, mtod(n, caddr_t), 981 (unsigned)count); 982 len -= count; 983 m->m_len += count; 984 n->m_len -= count; 985 space -= count; 986 if (n->m_len) 987 n->m_data += count; 988 else 989 n = m_free(n); 990 } while (len > 0 && n); 991 if (len > 0) { 992 (void) m_free(m); 993 goto bad; 994 } 995 m->m_next = n; 996 return (m); 997 bad: 998 m_freem(n); 999 return (NULL); 1000 } 1001 1002 /* 1003 * Partition an mbuf chain in two pieces, returning the tail -- 1004 * all but the first len0 bytes. In case of failure, it returns NULL and 1005 * attempts to restore the chain to its original state. 1006 * 1007 * Note that the resulting mbufs might be read-only, because the new 1008 * mbuf can end up sharing an mbuf cluster with the original mbuf if 1009 * the "breaking point" happens to lie within a cluster mbuf. Use the 1010 * M_WRITABLE() macro to check for this case. 1011 */ 1012 struct mbuf * 1013 m_split(struct mbuf *m0, int len0, int wait) 1014 { 1015 struct mbuf *m, *n; 1016 u_int len = len0, remain; 1017 1018 MBUF_CHECKSLEEP(wait); 1019 for (m = m0; m && len > m->m_len; m = m->m_next) 1020 len -= m->m_len; 1021 if (m == NULL) 1022 return (NULL); 1023 remain = m->m_len - len; 1024 if (m0->m_flags & M_PKTHDR && remain == 0) { 1025 n = m_gethdr(wait, m0->m_type); 1026 if (n == NULL) 1027 return (NULL); 1028 n->m_next = m->m_next; 1029 m->m_next = NULL; 1030 if (m0->m_pkthdr.csum_flags & CSUM_SND_TAG) { 1031 n->m_pkthdr.snd_tag = 1032 m_snd_tag_ref(m0->m_pkthdr.snd_tag); 1033 n->m_pkthdr.csum_flags |= CSUM_SND_TAG; 1034 } else 1035 n->m_pkthdr.rcvif = m0->m_pkthdr.rcvif; 1036 n->m_pkthdr.len = m0->m_pkthdr.len - len0; 1037 m0->m_pkthdr.len = len0; 1038 return (n); 1039 } else if (m0->m_flags & M_PKTHDR) { 1040 n = m_gethdr(wait, m0->m_type); 1041 if (n == NULL) 1042 return (NULL); 1043 if (m0->m_pkthdr.csum_flags & CSUM_SND_TAG) { 1044 n->m_pkthdr.snd_tag = 1045 m_snd_tag_ref(m0->m_pkthdr.snd_tag); 1046 n->m_pkthdr.csum_flags |= CSUM_SND_TAG; 1047 } else 1048 n->m_pkthdr.rcvif = m0->m_pkthdr.rcvif; 1049 n->m_pkthdr.len = m0->m_pkthdr.len - len0; 1050 m0->m_pkthdr.len = len0; 1051 if (m->m_flags & (M_EXT|M_EXTPG)) 1052 goto extpacket; 1053 if (remain > MHLEN) { 1054 /* m can't be the lead packet */ 1055 M_ALIGN(n, 0); 1056 n->m_next = m_split(m, len, wait); 1057 if (n->m_next == NULL) { 1058 (void) m_free(n); 1059 return (NULL); 1060 } else { 1061 n->m_len = 0; 1062 return (n); 1063 } 1064 } else 1065 M_ALIGN(n, remain); 1066 } else if (remain == 0) { 1067 n = m->m_next; 1068 m->m_next = NULL; 1069 return (n); 1070 } else { 1071 n = m_get(wait, m->m_type); 1072 if (n == NULL) 1073 return (NULL); 1074 M_ALIGN(n, remain); 1075 } 1076 extpacket: 1077 if (m->m_flags & (M_EXT|M_EXTPG)) { 1078 n->m_data = m->m_data + len; 1079 mb_dupcl(n, m); 1080 } else { 1081 bcopy(mtod(m, caddr_t) + len, mtod(n, caddr_t), remain); 1082 } 1083 n->m_len = remain; 1084 m->m_len = len; 1085 n->m_next = m->m_next; 1086 m->m_next = NULL; 1087 return (n); 1088 } 1089 /* 1090 * Routine to copy from device local memory into mbufs. 1091 * Note that `off' argument is offset into first mbuf of target chain from 1092 * which to begin copying the data to. 1093 */ 1094 struct mbuf * 1095 m_devget(char *buf, int totlen, int off, struct ifnet *ifp, 1096 void (*copy)(char *from, caddr_t to, u_int len)) 1097 { 1098 struct mbuf *m; 1099 struct mbuf *top = NULL, **mp = ⊤ 1100 int len; 1101 1102 if (off < 0 || off > MHLEN) 1103 return (NULL); 1104 1105 while (totlen > 0) { 1106 if (top == NULL) { /* First one, must be PKTHDR */ 1107 if (totlen + off >= MINCLSIZE) { 1108 m = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR); 1109 len = MCLBYTES; 1110 } else { 1111 m = m_gethdr(M_NOWAIT, MT_DATA); 1112 len = MHLEN; 1113 1114 /* Place initial small packet/header at end of mbuf */ 1115 if (m && totlen + off + max_linkhdr <= MHLEN) { 1116 m->m_data += max_linkhdr; 1117 len -= max_linkhdr; 1118 } 1119 } 1120 if (m == NULL) 1121 return NULL; 1122 m->m_pkthdr.rcvif = ifp; 1123 m->m_pkthdr.len = totlen; 1124 } else { 1125 if (totlen + off >= MINCLSIZE) { 1126 m = m_getcl(M_NOWAIT, MT_DATA, 0); 1127 len = MCLBYTES; 1128 } else { 1129 m = m_get(M_NOWAIT, MT_DATA); 1130 len = MLEN; 1131 } 1132 if (m == NULL) { 1133 m_freem(top); 1134 return NULL; 1135 } 1136 } 1137 if (off) { 1138 m->m_data += off; 1139 len -= off; 1140 off = 0; 1141 } 1142 m->m_len = len = min(totlen, len); 1143 if (copy) 1144 copy(buf, mtod(m, caddr_t), (u_int)len); 1145 else 1146 bcopy(buf, mtod(m, caddr_t), (u_int)len); 1147 buf += len; 1148 *mp = m; 1149 mp = &m->m_next; 1150 totlen -= len; 1151 } 1152 return (top); 1153 } 1154 1155 static void 1156 m_copytounmapped(const struct mbuf *m, int off, int len, c_caddr_t cp) 1157 { 1158 struct iovec iov; 1159 struct uio uio; 1160 int error __diagused; 1161 1162 KASSERT(off >= 0, ("m_copytounmapped: negative off %d", off)); 1163 KASSERT(len >= 0, ("m_copytounmapped: negative len %d", len)); 1164 KASSERT(off < m->m_len, ("m_copytounmapped: len exceeds mbuf length")); 1165 iov.iov_base = __DECONST(caddr_t, cp); 1166 iov.iov_len = len; 1167 uio.uio_resid = len; 1168 uio.uio_iov = &iov; 1169 uio.uio_segflg = UIO_SYSSPACE; 1170 uio.uio_iovcnt = 1; 1171 uio.uio_offset = 0; 1172 uio.uio_rw = UIO_WRITE; 1173 error = m_unmapped_uiomove(m, off, &uio, len); 1174 KASSERT(error == 0, ("m_unmapped_uiomove failed: off %d, len %d", off, 1175 len)); 1176 } 1177 1178 /* 1179 * Copy data from a buffer back into the indicated mbuf chain, 1180 * starting "off" bytes from the beginning, extending the mbuf 1181 * chain if necessary. 1182 */ 1183 void 1184 m_copyback(struct mbuf *m0, int off, int len, c_caddr_t cp) 1185 { 1186 int mlen; 1187 struct mbuf *m = m0, *n; 1188 int totlen = 0; 1189 1190 if (m0 == NULL) 1191 return; 1192 while (off > (mlen = m->m_len)) { 1193 off -= mlen; 1194 totlen += mlen; 1195 if (m->m_next == NULL) { 1196 n = m_get(M_NOWAIT, m->m_type); 1197 if (n == NULL) 1198 goto out; 1199 bzero(mtod(n, caddr_t), MLEN); 1200 n->m_len = min(MLEN, len + off); 1201 m->m_next = n; 1202 } 1203 m = m->m_next; 1204 } 1205 while (len > 0) { 1206 if (m->m_next == NULL && (len > m->m_len - off)) { 1207 m->m_len += min(len - (m->m_len - off), 1208 M_TRAILINGSPACE(m)); 1209 } 1210 mlen = min (m->m_len - off, len); 1211 if ((m->m_flags & M_EXTPG) != 0) 1212 m_copytounmapped(m, off, mlen, cp); 1213 else 1214 bcopy(cp, off + mtod(m, caddr_t), (u_int)mlen); 1215 cp += mlen; 1216 len -= mlen; 1217 mlen += off; 1218 off = 0; 1219 totlen += mlen; 1220 if (len == 0) 1221 break; 1222 if (m->m_next == NULL) { 1223 n = m_get(M_NOWAIT, m->m_type); 1224 if (n == NULL) 1225 break; 1226 n->m_len = min(MLEN, len); 1227 m->m_next = n; 1228 } 1229 m = m->m_next; 1230 } 1231 out: if (((m = m0)->m_flags & M_PKTHDR) && (m->m_pkthdr.len < totlen)) 1232 m->m_pkthdr.len = totlen; 1233 } 1234 1235 /* 1236 * Append the specified data to the indicated mbuf chain, 1237 * Extend the mbuf chain if the new data does not fit in 1238 * existing space. 1239 * 1240 * Return 1 if able to complete the job; otherwise 0. 1241 */ 1242 int 1243 m_append(struct mbuf *m0, int len, c_caddr_t cp) 1244 { 1245 struct mbuf *m, *n; 1246 int remainder, space; 1247 1248 for (m = m0; m->m_next != NULL; m = m->m_next) 1249 ; 1250 remainder = len; 1251 space = M_TRAILINGSPACE(m); 1252 if (space > 0) { 1253 /* 1254 * Copy into available space. 1255 */ 1256 if (space > remainder) 1257 space = remainder; 1258 bcopy(cp, mtod(m, caddr_t) + m->m_len, space); 1259 m->m_len += space; 1260 cp += space, remainder -= space; 1261 } 1262 while (remainder > 0) { 1263 /* 1264 * Allocate a new mbuf; could check space 1265 * and allocate a cluster instead. 1266 */ 1267 n = m_get(M_NOWAIT, m->m_type); 1268 if (n == NULL) 1269 break; 1270 n->m_len = min(MLEN, remainder); 1271 bcopy(cp, mtod(n, caddr_t), n->m_len); 1272 cp += n->m_len, remainder -= n->m_len; 1273 m->m_next = n; 1274 m = n; 1275 } 1276 if (m0->m_flags & M_PKTHDR) 1277 m0->m_pkthdr.len += len - remainder; 1278 return (remainder == 0); 1279 } 1280 1281 static int 1282 m_apply_extpg_one(struct mbuf *m, int off, int len, 1283 int (*f)(void *, void *, u_int), void *arg) 1284 { 1285 void *p; 1286 u_int i, count, pgoff, pglen; 1287 int rval; 1288 1289 KASSERT(PMAP_HAS_DMAP, 1290 ("m_apply_extpg_one does not support unmapped mbufs")); 1291 off += mtod(m, vm_offset_t); 1292 if (off < m->m_epg_hdrlen) { 1293 count = min(m->m_epg_hdrlen - off, len); 1294 rval = f(arg, m->m_epg_hdr + off, count); 1295 if (rval) 1296 return (rval); 1297 len -= count; 1298 off = 0; 1299 } else 1300 off -= m->m_epg_hdrlen; 1301 pgoff = m->m_epg_1st_off; 1302 for (i = 0; i < m->m_epg_npgs && len > 0; i++) { 1303 pglen = m_epg_pagelen(m, i, pgoff); 1304 if (off < pglen) { 1305 count = min(pglen - off, len); 1306 p = (void *)PHYS_TO_DMAP(m->m_epg_pa[i] + pgoff + off); 1307 rval = f(arg, p, count); 1308 if (rval) 1309 return (rval); 1310 len -= count; 1311 off = 0; 1312 } else 1313 off -= pglen; 1314 pgoff = 0; 1315 } 1316 if (len > 0) { 1317 KASSERT(off < m->m_epg_trllen, 1318 ("m_apply_extpg_one: offset beyond trailer")); 1319 KASSERT(len <= m->m_epg_trllen - off, 1320 ("m_apply_extpg_one: length beyond trailer")); 1321 return (f(arg, m->m_epg_trail + off, len)); 1322 } 1323 return (0); 1324 } 1325 1326 /* Apply function f to the data in a single mbuf. */ 1327 static int 1328 m_apply_one(struct mbuf *m, int off, int len, 1329 int (*f)(void *, void *, u_int), void *arg) 1330 { 1331 if ((m->m_flags & M_EXTPG) != 0) 1332 return (m_apply_extpg_one(m, off, len, f, arg)); 1333 else 1334 return (f(arg, mtod(m, caddr_t) + off, len)); 1335 } 1336 1337 /* 1338 * Apply function f to the data in an mbuf chain starting "off" bytes from 1339 * the beginning, continuing for "len" bytes. 1340 */ 1341 int 1342 m_apply(struct mbuf *m, int off, int len, 1343 int (*f)(void *, void *, u_int), void *arg) 1344 { 1345 u_int count; 1346 int rval; 1347 1348 KASSERT(off >= 0, ("m_apply, negative off %d", off)); 1349 KASSERT(len >= 0, ("m_apply, negative len %d", len)); 1350 while (off > 0) { 1351 KASSERT(m != NULL, ("m_apply, offset > size of mbuf chain")); 1352 if (off < m->m_len) 1353 break; 1354 off -= m->m_len; 1355 m = m->m_next; 1356 } 1357 while (len > 0) { 1358 KASSERT(m != NULL, ("m_apply, offset > size of mbuf chain")); 1359 count = min(m->m_len - off, len); 1360 rval = m_apply_one(m, off, count, f, arg); 1361 if (rval) 1362 return (rval); 1363 len -= count; 1364 off = 0; 1365 m = m->m_next; 1366 } 1367 return (0); 1368 } 1369 1370 /* 1371 * Return a pointer to mbuf/offset of location in mbuf chain. 1372 */ 1373 struct mbuf * 1374 m_getptr(struct mbuf *m, int loc, int *off) 1375 { 1376 1377 while (loc >= 0) { 1378 /* Normal end of search. */ 1379 if (m->m_len > loc) { 1380 *off = loc; 1381 return (m); 1382 } else { 1383 loc -= m->m_len; 1384 if (m->m_next == NULL) { 1385 if (loc == 0) { 1386 /* Point at the end of valid data. */ 1387 *off = m->m_len; 1388 return (m); 1389 } 1390 return (NULL); 1391 } 1392 m = m->m_next; 1393 } 1394 } 1395 return (NULL); 1396 } 1397 1398 void 1399 m_print(const struct mbuf *m, int maxlen) 1400 { 1401 int len; 1402 int pdata; 1403 const struct mbuf *m2; 1404 1405 if (m == NULL) { 1406 printf("mbuf: %p\n", m); 1407 return; 1408 } 1409 1410 if (m->m_flags & M_PKTHDR) 1411 len = m->m_pkthdr.len; 1412 else 1413 len = -1; 1414 m2 = m; 1415 while (m2 != NULL && (len == -1 || len)) { 1416 pdata = m2->m_len; 1417 if (maxlen != -1 && pdata > maxlen) 1418 pdata = maxlen; 1419 printf("mbuf: %p len: %d, next: %p, %b%s", m2, m2->m_len, 1420 m2->m_next, m2->m_flags, "\20\20freelist\17skipfw" 1421 "\11proto5\10proto4\7proto3\6proto2\5proto1\4rdonly" 1422 "\3eor\2pkthdr\1ext", pdata ? "" : "\n"); 1423 if (pdata) 1424 printf(", %*D\n", pdata, (u_char *)m2->m_data, "-"); 1425 if (len != -1) 1426 len -= m2->m_len; 1427 m2 = m2->m_next; 1428 } 1429 if (len > 0) 1430 printf("%d bytes unaccounted for.\n", len); 1431 return; 1432 } 1433 1434 u_int 1435 m_fixhdr(struct mbuf *m0) 1436 { 1437 u_int len; 1438 1439 len = m_length(m0, NULL); 1440 m0->m_pkthdr.len = len; 1441 return (len); 1442 } 1443 1444 u_int 1445 m_length(struct mbuf *m0, struct mbuf **last) 1446 { 1447 struct mbuf *m; 1448 u_int len; 1449 1450 len = 0; 1451 for (m = m0; m != NULL; m = m->m_next) { 1452 len += m->m_len; 1453 if (m->m_next == NULL) 1454 break; 1455 } 1456 if (last != NULL) 1457 *last = m; 1458 return (len); 1459 } 1460 1461 /* 1462 * Defragment a mbuf chain, returning the shortest possible 1463 * chain of mbufs and clusters. If allocation fails and 1464 * this cannot be completed, NULL will be returned, but 1465 * the passed in chain will be unchanged. Upon success, 1466 * the original chain will be freed, and the new chain 1467 * will be returned. 1468 * 1469 * If a non-packet header is passed in, the original 1470 * mbuf (chain?) will be returned unharmed. 1471 */ 1472 struct mbuf * 1473 m_defrag(struct mbuf *m0, int how) 1474 { 1475 struct mbuf *m_new = NULL, *m_final = NULL; 1476 int progress = 0, length; 1477 1478 MBUF_CHECKSLEEP(how); 1479 if (!(m0->m_flags & M_PKTHDR)) 1480 return (m0); 1481 1482 m_fixhdr(m0); /* Needed sanity check */ 1483 1484 #ifdef MBUF_STRESS_TEST 1485 if (m_defragrandomfailures) { 1486 int temp = arc4random() & 0xff; 1487 if (temp == 0xba) 1488 goto nospace; 1489 } 1490 #endif 1491 1492 if (m0->m_pkthdr.len > MHLEN) 1493 m_final = m_getcl(how, MT_DATA, M_PKTHDR); 1494 else 1495 m_final = m_gethdr(how, MT_DATA); 1496 1497 if (m_final == NULL) 1498 goto nospace; 1499 1500 if (m_dup_pkthdr(m_final, m0, how) == 0) 1501 goto nospace; 1502 1503 m_new = m_final; 1504 1505 while (progress < m0->m_pkthdr.len) { 1506 length = m0->m_pkthdr.len - progress; 1507 if (length > MCLBYTES) 1508 length = MCLBYTES; 1509 1510 if (m_new == NULL) { 1511 if (length > MLEN) 1512 m_new = m_getcl(how, MT_DATA, 0); 1513 else 1514 m_new = m_get(how, MT_DATA); 1515 if (m_new == NULL) 1516 goto nospace; 1517 } 1518 1519 m_copydata(m0, progress, length, mtod(m_new, caddr_t)); 1520 progress += length; 1521 m_new->m_len = length; 1522 if (m_new != m_final) 1523 m_cat(m_final, m_new); 1524 m_new = NULL; 1525 } 1526 #ifdef MBUF_STRESS_TEST 1527 if (m0->m_next == NULL) 1528 m_defraguseless++; 1529 #endif 1530 m_freem(m0); 1531 m0 = m_final; 1532 #ifdef MBUF_STRESS_TEST 1533 m_defragpackets++; 1534 m_defragbytes += m0->m_pkthdr.len; 1535 #endif 1536 return (m0); 1537 nospace: 1538 #ifdef MBUF_STRESS_TEST 1539 m_defragfailure++; 1540 #endif 1541 if (m_final) 1542 m_freem(m_final); 1543 return (NULL); 1544 } 1545 1546 /* 1547 * Return the number of fragments an mbuf will use. This is usually 1548 * used as a proxy for the number of scatter/gather elements needed by 1549 * a DMA engine to access an mbuf. In general mapped mbufs are 1550 * assumed to be backed by physically contiguous buffers that only 1551 * need a single fragment. Unmapped mbufs, on the other hand, can 1552 * span disjoint physical pages. 1553 */ 1554 static int 1555 frags_per_mbuf(struct mbuf *m) 1556 { 1557 int frags; 1558 1559 if ((m->m_flags & M_EXTPG) == 0) 1560 return (1); 1561 1562 /* 1563 * The header and trailer are counted as a single fragment 1564 * each when present. 1565 * 1566 * XXX: This overestimates the number of fragments by assuming 1567 * all the backing physical pages are disjoint. 1568 */ 1569 frags = 0; 1570 if (m->m_epg_hdrlen != 0) 1571 frags++; 1572 frags += m->m_epg_npgs; 1573 if (m->m_epg_trllen != 0) 1574 frags++; 1575 1576 return (frags); 1577 } 1578 1579 /* 1580 * Defragment an mbuf chain, returning at most maxfrags separate 1581 * mbufs+clusters. If this is not possible NULL is returned and 1582 * the original mbuf chain is left in its present (potentially 1583 * modified) state. We use two techniques: collapsing consecutive 1584 * mbufs and replacing consecutive mbufs by a cluster. 1585 * 1586 * NB: this should really be named m_defrag but that name is taken 1587 */ 1588 struct mbuf * 1589 m_collapse(struct mbuf *m0, int how, int maxfrags) 1590 { 1591 struct mbuf *m, *n, *n2, **prev; 1592 u_int curfrags; 1593 1594 /* 1595 * Calculate the current number of frags. 1596 */ 1597 curfrags = 0; 1598 for (m = m0; m != NULL; m = m->m_next) 1599 curfrags += frags_per_mbuf(m); 1600 /* 1601 * First, try to collapse mbufs. Note that we always collapse 1602 * towards the front so we don't need to deal with moving the 1603 * pkthdr. This may be suboptimal if the first mbuf has much 1604 * less data than the following. 1605 */ 1606 m = m0; 1607 again: 1608 for (;;) { 1609 n = m->m_next; 1610 if (n == NULL) 1611 break; 1612 if (M_WRITABLE(m) && 1613 n->m_len < M_TRAILINGSPACE(m)) { 1614 m_copydata(n, 0, n->m_len, 1615 mtod(m, char *) + m->m_len); 1616 m->m_len += n->m_len; 1617 m->m_next = n->m_next; 1618 curfrags -= frags_per_mbuf(n); 1619 m_free(n); 1620 if (curfrags <= maxfrags) 1621 return m0; 1622 } else 1623 m = n; 1624 } 1625 KASSERT(maxfrags > 1, 1626 ("maxfrags %u, but normal collapse failed", maxfrags)); 1627 /* 1628 * Collapse consecutive mbufs to a cluster. 1629 */ 1630 prev = &m0->m_next; /* NB: not the first mbuf */ 1631 while ((n = *prev) != NULL) { 1632 if ((n2 = n->m_next) != NULL && 1633 n->m_len + n2->m_len < MCLBYTES) { 1634 m = m_getcl(how, MT_DATA, 0); 1635 if (m == NULL) 1636 goto bad; 1637 m_copydata(n, 0, n->m_len, mtod(m, char *)); 1638 m_copydata(n2, 0, n2->m_len, 1639 mtod(m, char *) + n->m_len); 1640 m->m_len = n->m_len + n2->m_len; 1641 m->m_next = n2->m_next; 1642 *prev = m; 1643 curfrags += 1; /* For the new cluster */ 1644 curfrags -= frags_per_mbuf(n); 1645 curfrags -= frags_per_mbuf(n2); 1646 m_free(n); 1647 m_free(n2); 1648 if (curfrags <= maxfrags) 1649 return m0; 1650 /* 1651 * Still not there, try the normal collapse 1652 * again before we allocate another cluster. 1653 */ 1654 goto again; 1655 } 1656 prev = &n->m_next; 1657 } 1658 /* 1659 * No place where we can collapse to a cluster; punt. 1660 * This can occur if, for example, you request 2 frags 1661 * but the packet requires that both be clusters (we 1662 * never reallocate the first mbuf to avoid moving the 1663 * packet header). 1664 */ 1665 bad: 1666 return NULL; 1667 } 1668 1669 #ifdef MBUF_STRESS_TEST 1670 1671 /* 1672 * Fragment an mbuf chain. There's no reason you'd ever want to do 1673 * this in normal usage, but it's great for stress testing various 1674 * mbuf consumers. 1675 * 1676 * If fragmentation is not possible, the original chain will be 1677 * returned. 1678 * 1679 * Possible length values: 1680 * 0 no fragmentation will occur 1681 * > 0 each fragment will be of the specified length 1682 * -1 each fragment will be the same random value in length 1683 * -2 each fragment's length will be entirely random 1684 * (Random values range from 1 to 256) 1685 */ 1686 struct mbuf * 1687 m_fragment(struct mbuf *m0, int how, int length) 1688 { 1689 struct mbuf *m_first, *m_last; 1690 int divisor = 255, progress = 0, fraglen; 1691 1692 if (!(m0->m_flags & M_PKTHDR)) 1693 return (m0); 1694 1695 if (length == 0 || length < -2) 1696 return (m0); 1697 if (length > MCLBYTES) 1698 length = MCLBYTES; 1699 if (length < 0 && divisor > MCLBYTES) 1700 divisor = MCLBYTES; 1701 if (length == -1) 1702 length = 1 + (arc4random() % divisor); 1703 if (length > 0) 1704 fraglen = length; 1705 1706 m_fixhdr(m0); /* Needed sanity check */ 1707 1708 m_first = m_getcl(how, MT_DATA, M_PKTHDR); 1709 if (m_first == NULL) 1710 goto nospace; 1711 1712 if (m_dup_pkthdr(m_first, m0, how) == 0) 1713 goto nospace; 1714 1715 m_last = m_first; 1716 1717 while (progress < m0->m_pkthdr.len) { 1718 if (length == -2) 1719 fraglen = 1 + (arc4random() % divisor); 1720 if (fraglen > m0->m_pkthdr.len - progress) 1721 fraglen = m0->m_pkthdr.len - progress; 1722 1723 if (progress != 0) { 1724 struct mbuf *m_new = m_getcl(how, MT_DATA, 0); 1725 if (m_new == NULL) 1726 goto nospace; 1727 1728 m_last->m_next = m_new; 1729 m_last = m_new; 1730 } 1731 1732 m_copydata(m0, progress, fraglen, mtod(m_last, caddr_t)); 1733 progress += fraglen; 1734 m_last->m_len = fraglen; 1735 } 1736 m_freem(m0); 1737 m0 = m_first; 1738 return (m0); 1739 nospace: 1740 if (m_first) 1741 m_freem(m_first); 1742 /* Return the original chain on failure */ 1743 return (m0); 1744 } 1745 1746 #endif 1747 1748 /* 1749 * Free pages from mbuf_ext_pgs, assuming they were allocated via 1750 * vm_page_alloc() and aren't associated with any object. Complement 1751 * to allocator from m_uiotombuf_nomap(). 1752 */ 1753 void 1754 mb_free_mext_pgs(struct mbuf *m) 1755 { 1756 vm_page_t pg; 1757 1758 M_ASSERTEXTPG(m); 1759 for (int i = 0; i < m->m_epg_npgs; i++) { 1760 pg = PHYS_TO_VM_PAGE(m->m_epg_pa[i]); 1761 vm_page_unwire_noq(pg); 1762 vm_page_free(pg); 1763 } 1764 } 1765 1766 static struct mbuf * 1767 m_uiotombuf_nomap(struct uio *uio, int how, int len, int maxseg, int flags) 1768 { 1769 struct mbuf *m, *mb, *prev; 1770 vm_page_t pg_array[MBUF_PEXT_MAX_PGS]; 1771 int error, length, i, needed; 1772 ssize_t total; 1773 int pflags = malloc2vm_flags(how) | VM_ALLOC_NODUMP | VM_ALLOC_WIRED; 1774 1775 MPASS((flags & M_PKTHDR) == 0); 1776 MPASS((how & M_ZERO) == 0); 1777 1778 /* 1779 * len can be zero or an arbitrary large value bound by 1780 * the total data supplied by the uio. 1781 */ 1782 if (len > 0) 1783 total = MIN(uio->uio_resid, len); 1784 else 1785 total = uio->uio_resid; 1786 1787 if (maxseg == 0) 1788 maxseg = MBUF_PEXT_MAX_PGS * PAGE_SIZE; 1789 1790 /* 1791 * If total is zero, return an empty mbuf. This can occur 1792 * for TLS 1.0 connections which send empty fragments as 1793 * a countermeasure against the known-IV weakness in CBC 1794 * ciphersuites. 1795 */ 1796 if (__predict_false(total == 0)) { 1797 mb = mb_alloc_ext_pgs(how, mb_free_mext_pgs); 1798 if (mb == NULL) 1799 return (NULL); 1800 mb->m_epg_flags = EPG_FLAG_ANON; 1801 return (mb); 1802 } 1803 1804 /* 1805 * Allocate the pages 1806 */ 1807 m = NULL; 1808 while (total > 0) { 1809 mb = mb_alloc_ext_pgs(how, mb_free_mext_pgs); 1810 if (mb == NULL) 1811 goto failed; 1812 if (m == NULL) 1813 m = mb; 1814 else 1815 prev->m_next = mb; 1816 prev = mb; 1817 mb->m_epg_flags = EPG_FLAG_ANON; 1818 needed = length = MIN(maxseg, total); 1819 for (i = 0; needed > 0; i++, needed -= PAGE_SIZE) { 1820 retry_page: 1821 pg_array[i] = vm_page_alloc_noobj(pflags); 1822 if (pg_array[i] == NULL) { 1823 if (how & M_NOWAIT) { 1824 goto failed; 1825 } else { 1826 vm_wait(NULL); 1827 goto retry_page; 1828 } 1829 } 1830 mb->m_epg_pa[i] = VM_PAGE_TO_PHYS(pg_array[i]); 1831 mb->m_epg_npgs++; 1832 } 1833 mb->m_epg_last_len = length - PAGE_SIZE * (mb->m_epg_npgs - 1); 1834 MBUF_EXT_PGS_ASSERT_SANITY(mb); 1835 total -= length; 1836 error = uiomove_fromphys(pg_array, 0, length, uio); 1837 if (error != 0) 1838 goto failed; 1839 mb->m_len = length; 1840 mb->m_ext.ext_size += PAGE_SIZE * mb->m_epg_npgs; 1841 if (flags & M_PKTHDR) 1842 m->m_pkthdr.len += length; 1843 } 1844 return (m); 1845 1846 failed: 1847 m_freem(m); 1848 return (NULL); 1849 } 1850 1851 /* 1852 * Copy the contents of uio into a properly sized mbuf chain. 1853 */ 1854 struct mbuf * 1855 m_uiotombuf(struct uio *uio, int how, int len, int align, int flags) 1856 { 1857 struct mbuf *m, *mb; 1858 int error, length; 1859 ssize_t total; 1860 int progress = 0; 1861 1862 if (flags & M_EXTPG) 1863 return (m_uiotombuf_nomap(uio, how, len, align, flags)); 1864 1865 /* 1866 * len can be zero or an arbitrary large value bound by 1867 * the total data supplied by the uio. 1868 */ 1869 if (len > 0) 1870 total = (uio->uio_resid < len) ? uio->uio_resid : len; 1871 else 1872 total = uio->uio_resid; 1873 1874 /* 1875 * The smallest unit returned by m_getm2() is a single mbuf 1876 * with pkthdr. We can't align past it. 1877 */ 1878 if (align >= MHLEN) 1879 return (NULL); 1880 1881 /* 1882 * Give us the full allocation or nothing. 1883 * If len is zero return the smallest empty mbuf. 1884 */ 1885 m = m_getm2(NULL, max(total + align, 1), how, MT_DATA, flags); 1886 if (m == NULL) 1887 return (NULL); 1888 m->m_data += align; 1889 1890 /* Fill all mbufs with uio data and update header information. */ 1891 for (mb = m; mb != NULL; mb = mb->m_next) { 1892 length = min(M_TRAILINGSPACE(mb), total - progress); 1893 1894 error = uiomove(mtod(mb, void *), length, uio); 1895 if (error) { 1896 m_freem(m); 1897 return (NULL); 1898 } 1899 1900 mb->m_len = length; 1901 progress += length; 1902 if (flags & M_PKTHDR) 1903 m->m_pkthdr.len += length; 1904 } 1905 KASSERT(progress == total, ("%s: progress != total", __func__)); 1906 1907 return (m); 1908 } 1909 1910 /* 1911 * Copy data to/from an unmapped mbuf into a uio limited by len if set. 1912 */ 1913 int 1914 m_unmapped_uiomove(const struct mbuf *m, int m_off, struct uio *uio, int len) 1915 { 1916 vm_page_t pg; 1917 int error, i, off, pglen, pgoff, seglen, segoff; 1918 1919 M_ASSERTEXTPG(m); 1920 error = 0; 1921 1922 /* Skip over any data removed from the front. */ 1923 off = mtod(m, vm_offset_t); 1924 1925 off += m_off; 1926 if (m->m_epg_hdrlen != 0) { 1927 if (off >= m->m_epg_hdrlen) { 1928 off -= m->m_epg_hdrlen; 1929 } else { 1930 seglen = m->m_epg_hdrlen - off; 1931 segoff = off; 1932 seglen = min(seglen, len); 1933 off = 0; 1934 len -= seglen; 1935 error = uiomove(__DECONST(void *, 1936 &m->m_epg_hdr[segoff]), seglen, uio); 1937 } 1938 } 1939 pgoff = m->m_epg_1st_off; 1940 for (i = 0; i < m->m_epg_npgs && error == 0 && len > 0; i++) { 1941 pglen = m_epg_pagelen(m, i, pgoff); 1942 if (off >= pglen) { 1943 off -= pglen; 1944 pgoff = 0; 1945 continue; 1946 } 1947 seglen = pglen - off; 1948 segoff = pgoff + off; 1949 off = 0; 1950 seglen = min(seglen, len); 1951 len -= seglen; 1952 pg = PHYS_TO_VM_PAGE(m->m_epg_pa[i]); 1953 error = uiomove_fromphys(&pg, segoff, seglen, uio); 1954 pgoff = 0; 1955 }; 1956 if (len != 0 && error == 0) { 1957 KASSERT((off + len) <= m->m_epg_trllen, 1958 ("off + len > trail (%d + %d > %d, m_off = %d)", off, len, 1959 m->m_epg_trllen, m_off)); 1960 error = uiomove(__DECONST(void *, &m->m_epg_trail[off]), 1961 len, uio); 1962 } 1963 return (error); 1964 } 1965 1966 /* 1967 * Copy an mbuf chain into a uio limited by len if set. 1968 */ 1969 int 1970 m_mbuftouio(struct uio *uio, const struct mbuf *m, int len) 1971 { 1972 int error, length, total; 1973 int progress = 0; 1974 1975 if (len > 0) 1976 total = min(uio->uio_resid, len); 1977 else 1978 total = uio->uio_resid; 1979 1980 /* Fill the uio with data from the mbufs. */ 1981 for (; m != NULL; m = m->m_next) { 1982 length = min(m->m_len, total - progress); 1983 1984 if ((m->m_flags & M_EXTPG) != 0) 1985 error = m_unmapped_uiomove(m, 0, uio, length); 1986 else 1987 error = uiomove(mtod(m, void *), length, uio); 1988 if (error) 1989 return (error); 1990 1991 progress += length; 1992 } 1993 1994 return (0); 1995 } 1996 1997 /* 1998 * Create a writable copy of the mbuf chain. While doing this 1999 * we compact the chain with a goal of producing a chain with 2000 * at most two mbufs. The second mbuf in this chain is likely 2001 * to be a cluster. The primary purpose of this work is to create 2002 * a writable packet for encryption, compression, etc. The 2003 * secondary goal is to linearize the data so the data can be 2004 * passed to crypto hardware in the most efficient manner possible. 2005 */ 2006 struct mbuf * 2007 m_unshare(struct mbuf *m0, int how) 2008 { 2009 struct mbuf *m, *mprev; 2010 struct mbuf *n, *mfirst, *mlast; 2011 int len, off; 2012 2013 mprev = NULL; 2014 for (m = m0; m != NULL; m = mprev->m_next) { 2015 /* 2016 * Regular mbufs are ignored unless there's a cluster 2017 * in front of it that we can use to coalesce. We do 2018 * the latter mainly so later clusters can be coalesced 2019 * also w/o having to handle them specially (i.e. convert 2020 * mbuf+cluster -> cluster). This optimization is heavily 2021 * influenced by the assumption that we're running over 2022 * Ethernet where MCLBYTES is large enough that the max 2023 * packet size will permit lots of coalescing into a 2024 * single cluster. This in turn permits efficient 2025 * crypto operations, especially when using hardware. 2026 */ 2027 if ((m->m_flags & M_EXT) == 0) { 2028 if (mprev && (mprev->m_flags & M_EXT) && 2029 m->m_len <= M_TRAILINGSPACE(mprev)) { 2030 /* XXX: this ignores mbuf types */ 2031 memcpy(mtod(mprev, caddr_t) + mprev->m_len, 2032 mtod(m, caddr_t), m->m_len); 2033 mprev->m_len += m->m_len; 2034 mprev->m_next = m->m_next; /* unlink from chain */ 2035 m_free(m); /* reclaim mbuf */ 2036 } else { 2037 mprev = m; 2038 } 2039 continue; 2040 } 2041 /* 2042 * Writable mbufs are left alone (for now). 2043 */ 2044 if (M_WRITABLE(m)) { 2045 mprev = m; 2046 continue; 2047 } 2048 2049 /* 2050 * Not writable, replace with a copy or coalesce with 2051 * the previous mbuf if possible (since we have to copy 2052 * it anyway, we try to reduce the number of mbufs and 2053 * clusters so that future work is easier). 2054 */ 2055 KASSERT(m->m_flags & M_EXT, ("m_flags 0x%x", m->m_flags)); 2056 /* NB: we only coalesce into a cluster or larger */ 2057 if (mprev != NULL && (mprev->m_flags & M_EXT) && 2058 m->m_len <= M_TRAILINGSPACE(mprev)) { 2059 /* XXX: this ignores mbuf types */ 2060 memcpy(mtod(mprev, caddr_t) + mprev->m_len, 2061 mtod(m, caddr_t), m->m_len); 2062 mprev->m_len += m->m_len; 2063 mprev->m_next = m->m_next; /* unlink from chain */ 2064 m_free(m); /* reclaim mbuf */ 2065 continue; 2066 } 2067 2068 /* 2069 * Allocate new space to hold the copy and copy the data. 2070 * We deal with jumbo mbufs (i.e. m_len > MCLBYTES) by 2071 * splitting them into clusters. We could just malloc a 2072 * buffer and make it external but too many device drivers 2073 * don't know how to break up the non-contiguous memory when 2074 * doing DMA. 2075 */ 2076 n = m_getcl(how, m->m_type, m->m_flags & M_COPYFLAGS); 2077 if (n == NULL) { 2078 m_freem(m0); 2079 return (NULL); 2080 } 2081 if (m->m_flags & M_PKTHDR) { 2082 KASSERT(mprev == NULL, ("%s: m0 %p, m %p has M_PKTHDR", 2083 __func__, m0, m)); 2084 m_move_pkthdr(n, m); 2085 } 2086 len = m->m_len; 2087 off = 0; 2088 mfirst = n; 2089 mlast = NULL; 2090 for (;;) { 2091 int cc = min(len, MCLBYTES); 2092 memcpy(mtod(n, caddr_t), mtod(m, caddr_t) + off, cc); 2093 n->m_len = cc; 2094 if (mlast != NULL) 2095 mlast->m_next = n; 2096 mlast = n; 2097 #if 0 2098 newipsecstat.ips_clcopied++; 2099 #endif 2100 2101 len -= cc; 2102 if (len <= 0) 2103 break; 2104 off += cc; 2105 2106 n = m_getcl(how, m->m_type, m->m_flags & M_COPYFLAGS); 2107 if (n == NULL) { 2108 m_freem(mfirst); 2109 m_freem(m0); 2110 return (NULL); 2111 } 2112 } 2113 n->m_next = m->m_next; 2114 if (mprev == NULL) 2115 m0 = mfirst; /* new head of chain */ 2116 else 2117 mprev->m_next = mfirst; /* replace old mbuf */ 2118 m_free(m); /* release old mbuf */ 2119 mprev = mfirst; 2120 } 2121 return (m0); 2122 } 2123 2124 #ifdef MBUF_PROFILING 2125 2126 #define MP_BUCKETS 32 /* don't just change this as things may overflow.*/ 2127 struct mbufprofile { 2128 uintmax_t wasted[MP_BUCKETS]; 2129 uintmax_t used[MP_BUCKETS]; 2130 uintmax_t segments[MP_BUCKETS]; 2131 } mbprof; 2132 2133 void 2134 m_profile(struct mbuf *m) 2135 { 2136 int segments = 0; 2137 int used = 0; 2138 int wasted = 0; 2139 2140 while (m) { 2141 segments++; 2142 used += m->m_len; 2143 if (m->m_flags & M_EXT) { 2144 wasted += MHLEN - sizeof(m->m_ext) + 2145 m->m_ext.ext_size - m->m_len; 2146 } else { 2147 if (m->m_flags & M_PKTHDR) 2148 wasted += MHLEN - m->m_len; 2149 else 2150 wasted += MLEN - m->m_len; 2151 } 2152 m = m->m_next; 2153 } 2154 /* be paranoid.. it helps */ 2155 if (segments > MP_BUCKETS - 1) 2156 segments = MP_BUCKETS - 1; 2157 if (used > 100000) 2158 used = 100000; 2159 if (wasted > 100000) 2160 wasted = 100000; 2161 /* store in the appropriate bucket */ 2162 /* don't bother locking. if it's slightly off, so what? */ 2163 mbprof.segments[segments]++; 2164 mbprof.used[fls(used)]++; 2165 mbprof.wasted[fls(wasted)]++; 2166 } 2167 2168 static int 2169 mbprof_handler(SYSCTL_HANDLER_ARGS) 2170 { 2171 char buf[256]; 2172 struct sbuf sb; 2173 int error; 2174 uint64_t *p; 2175 2176 sbuf_new_for_sysctl(&sb, buf, sizeof(buf), req); 2177 2178 p = &mbprof.wasted[0]; 2179 sbuf_printf(&sb, 2180 "wasted:\n" 2181 "%ju %ju %ju %ju %ju %ju %ju %ju " 2182 "%ju %ju %ju %ju %ju %ju %ju %ju\n", 2183 p[0], p[1], p[2], p[3], p[4], p[5], p[6], p[7], 2184 p[8], p[9], p[10], p[11], p[12], p[13], p[14], p[15]); 2185 #ifdef BIG_ARRAY 2186 p = &mbprof.wasted[16]; 2187 sbuf_printf(&sb, 2188 "%ju %ju %ju %ju %ju %ju %ju %ju " 2189 "%ju %ju %ju %ju %ju %ju %ju %ju\n", 2190 p[0], p[1], p[2], p[3], p[4], p[5], p[6], p[7], 2191 p[8], p[9], p[10], p[11], p[12], p[13], p[14], p[15]); 2192 #endif 2193 p = &mbprof.used[0]; 2194 sbuf_printf(&sb, 2195 "used:\n" 2196 "%ju %ju %ju %ju %ju %ju %ju %ju " 2197 "%ju %ju %ju %ju %ju %ju %ju %ju\n", 2198 p[0], p[1], p[2], p[3], p[4], p[5], p[6], p[7], 2199 p[8], p[9], p[10], p[11], p[12], p[13], p[14], p[15]); 2200 #ifdef BIG_ARRAY 2201 p = &mbprof.used[16]; 2202 sbuf_printf(&sb, 2203 "%ju %ju %ju %ju %ju %ju %ju %ju " 2204 "%ju %ju %ju %ju %ju %ju %ju %ju\n", 2205 p[0], p[1], p[2], p[3], p[4], p[5], p[6], p[7], 2206 p[8], p[9], p[10], p[11], p[12], p[13], p[14], p[15]); 2207 #endif 2208 p = &mbprof.segments[0]; 2209 sbuf_printf(&sb, 2210 "segments:\n" 2211 "%ju %ju %ju %ju %ju %ju %ju %ju " 2212 "%ju %ju %ju %ju %ju %ju %ju %ju\n", 2213 p[0], p[1], p[2], p[3], p[4], p[5], p[6], p[7], 2214 p[8], p[9], p[10], p[11], p[12], p[13], p[14], p[15]); 2215 #ifdef BIG_ARRAY 2216 p = &mbprof.segments[16]; 2217 sbuf_printf(&sb, 2218 "%ju %ju %ju %ju %ju %ju %ju %ju " 2219 "%ju %ju %ju %ju %ju %ju %ju %jju", 2220 p[0], p[1], p[2], p[3], p[4], p[5], p[6], p[7], 2221 p[8], p[9], p[10], p[11], p[12], p[13], p[14], p[15]); 2222 #endif 2223 2224 error = sbuf_finish(&sb); 2225 sbuf_delete(&sb); 2226 return (error); 2227 } 2228 2229 static int 2230 mbprof_clr_handler(SYSCTL_HANDLER_ARGS) 2231 { 2232 int clear, error; 2233 2234 clear = 0; 2235 error = sysctl_handle_int(oidp, &clear, 0, req); 2236 if (error || !req->newptr) 2237 return (error); 2238 2239 if (clear) { 2240 bzero(&mbprof, sizeof(mbprof)); 2241 } 2242 2243 return (error); 2244 } 2245 2246 SYSCTL_PROC(_kern_ipc, OID_AUTO, mbufprofile, 2247 CTLTYPE_STRING | CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, 0, 2248 mbprof_handler, "A", 2249 "mbuf profiling statistics"); 2250 2251 SYSCTL_PROC(_kern_ipc, OID_AUTO, mbufprofileclr, 2252 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE, NULL, 0, 2253 mbprof_clr_handler, "I", 2254 "clear mbuf profiling statistics"); 2255 #endif 2256