1 /*- 2 * SPDX-License-Identifier: BSD-3-Clause 3 * 4 * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project. 5 * 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 project 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 PROJECT 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 PROJECT 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 * $KAME: frag6.c,v 1.33 2002/01/07 11:34:48 kjc Exp $ 32 */ 33 34 #include <sys/cdefs.h> 35 __FBSDID("$FreeBSD$"); 36 37 #include "opt_rss.h" 38 39 #include <sys/param.h> 40 #include <sys/systm.h> 41 #include <sys/malloc.h> 42 #include <sys/mbuf.h> 43 #include <sys/domain.h> 44 #include <sys/eventhandler.h> 45 #include <sys/protosw.h> 46 #include <sys/socket.h> 47 #include <sys/errno.h> 48 #include <sys/time.h> 49 #include <sys/kernel.h> 50 #include <sys/syslog.h> 51 52 #include <net/if.h> 53 #include <net/if_var.h> 54 #include <net/netisr.h> 55 #include <net/route.h> 56 #include <net/vnet.h> 57 58 #include <netinet/in.h> 59 #include <netinet/in_var.h> 60 #include <netinet/ip6.h> 61 #include <netinet6/ip6_var.h> 62 #include <netinet/icmp6.h> 63 #include <netinet/in_systm.h> /* for ECN definitions */ 64 #include <netinet/ip.h> /* for ECN definitions */ 65 66 #include <security/mac/mac_framework.h> 67 68 static void frag6_enq(struct ip6asfrag *, struct ip6asfrag *); 69 static void frag6_deq(struct ip6asfrag *); 70 static void frag6_insque(struct ip6q *, struct ip6q *); 71 static void frag6_remque(struct ip6q *); 72 static void frag6_freef(struct ip6q *); 73 74 static struct mtx ip6qlock; 75 /* 76 * These fields all protected by ip6qlock. 77 */ 78 static VNET_DEFINE(u_int, frag6_nfragpackets); 79 static VNET_DEFINE(u_int, frag6_nfrags); 80 static VNET_DEFINE(struct ip6q, ip6q); /* ip6 reassemble queue */ 81 82 #define V_frag6_nfragpackets VNET(frag6_nfragpackets) 83 #define V_frag6_nfrags VNET(frag6_nfrags) 84 #define V_ip6q VNET(ip6q) 85 86 #define IP6Q_LOCK_INIT() mtx_init(&ip6qlock, "ip6qlock", NULL, MTX_DEF); 87 #define IP6Q_LOCK() mtx_lock(&ip6qlock) 88 #define IP6Q_TRYLOCK() mtx_trylock(&ip6qlock) 89 #define IP6Q_LOCK_ASSERT() mtx_assert(&ip6qlock, MA_OWNED) 90 #define IP6Q_UNLOCK() mtx_unlock(&ip6qlock) 91 92 static MALLOC_DEFINE(M_FTABLE, "fragment", "fragment reassembly header"); 93 94 /* 95 * Initialise reassembly queue and fragment identifier. 96 */ 97 static void 98 frag6_change(void *tag) 99 { 100 101 V_ip6_maxfragpackets = nmbclusters / 4; 102 V_ip6_maxfrags = nmbclusters / 4; 103 } 104 105 void 106 frag6_init(void) 107 { 108 109 V_ip6_maxfragpackets = nmbclusters / 4; 110 V_ip6_maxfrags = nmbclusters / 4; 111 V_ip6q.ip6q_next = V_ip6q.ip6q_prev = &V_ip6q; 112 113 if (!IS_DEFAULT_VNET(curvnet)) 114 return; 115 116 EVENTHANDLER_REGISTER(nmbclusters_change, 117 frag6_change, NULL, EVENTHANDLER_PRI_ANY); 118 119 IP6Q_LOCK_INIT(); 120 } 121 122 /* 123 * In RFC2460, fragment and reassembly rule do not agree with each other, 124 * in terms of next header field handling in fragment header. 125 * While the sender will use the same value for all of the fragmented packets, 126 * receiver is suggested not to check the consistency. 127 * 128 * fragment rule (p20): 129 * (2) A Fragment header containing: 130 * The Next Header value that identifies the first header of 131 * the Fragmentable Part of the original packet. 132 * -> next header field is same for all fragments 133 * 134 * reassembly rule (p21): 135 * The Next Header field of the last header of the Unfragmentable 136 * Part is obtained from the Next Header field of the first 137 * fragment's Fragment header. 138 * -> should grab it from the first fragment only 139 * 140 * The following note also contradicts with fragment rule - no one is going to 141 * send different fragment with different next header field. 142 * 143 * additional note (p22): 144 * The Next Header values in the Fragment headers of different 145 * fragments of the same original packet may differ. Only the value 146 * from the Offset zero fragment packet is used for reassembly. 147 * -> should grab it from the first fragment only 148 * 149 * There is no explicit reason given in the RFC. Historical reason maybe? 150 */ 151 /* 152 * Fragment input 153 */ 154 int 155 frag6_input(struct mbuf **mp, int *offp, int proto) 156 { 157 struct mbuf *m = *mp, *t; 158 struct ip6_hdr *ip6; 159 struct ip6_frag *ip6f; 160 struct ip6q *q6; 161 struct ip6asfrag *af6, *ip6af, *af6dwn; 162 struct in6_ifaddr *ia; 163 int offset = *offp, nxt, i, next; 164 int first_frag = 0; 165 int fragoff, frgpartlen; /* must be larger than u_int16_t */ 166 struct ifnet *dstifp; 167 u_int8_t ecn, ecn0; 168 #ifdef RSS 169 struct m_tag *mtag; 170 struct ip6_direct_ctx *ip6dc; 171 #endif 172 173 #if 0 174 char ip6buf[INET6_ADDRSTRLEN]; 175 #endif 176 177 ip6 = mtod(m, struct ip6_hdr *); 178 #ifndef PULLDOWN_TEST 179 IP6_EXTHDR_CHECK(m, offset, sizeof(struct ip6_frag), IPPROTO_DONE); 180 ip6f = (struct ip6_frag *)((caddr_t)ip6 + offset); 181 #else 182 IP6_EXTHDR_GET(ip6f, struct ip6_frag *, m, offset, sizeof(*ip6f)); 183 if (ip6f == NULL) 184 return (IPPROTO_DONE); 185 #endif 186 187 dstifp = NULL; 188 /* find the destination interface of the packet. */ 189 ia = in6ifa_ifwithaddr(&ip6->ip6_dst, 0 /* XXX */); 190 if (ia != NULL) { 191 dstifp = ia->ia_ifp; 192 ifa_free(&ia->ia_ifa); 193 } 194 /* jumbo payload can't contain a fragment header */ 195 if (ip6->ip6_plen == 0) { 196 icmp6_error(m, ICMP6_PARAM_PROB, ICMP6_PARAMPROB_HEADER, offset); 197 in6_ifstat_inc(dstifp, ifs6_reass_fail); 198 return IPPROTO_DONE; 199 } 200 201 /* 202 * check whether fragment packet's fragment length is 203 * multiple of 8 octets. 204 * sizeof(struct ip6_frag) == 8 205 * sizeof(struct ip6_hdr) = 40 206 */ 207 if ((ip6f->ip6f_offlg & IP6F_MORE_FRAG) && 208 (((ntohs(ip6->ip6_plen) - offset) & 0x7) != 0)) { 209 icmp6_error(m, ICMP6_PARAM_PROB, ICMP6_PARAMPROB_HEADER, 210 offsetof(struct ip6_hdr, ip6_plen)); 211 in6_ifstat_inc(dstifp, ifs6_reass_fail); 212 return IPPROTO_DONE; 213 } 214 215 IP6STAT_INC(ip6s_fragments); 216 in6_ifstat_inc(dstifp, ifs6_reass_reqd); 217 218 /* offset now points to data portion */ 219 offset += sizeof(struct ip6_frag); 220 221 /* 222 * RFC 6946: Handle "atomic" fragments (offset and m bit set to 0) 223 * upfront, unrelated to any reassembly. Just skip the fragment header. 224 */ 225 if ((ip6f->ip6f_offlg & ~IP6F_RESERVED_MASK) == 0) { 226 /* XXX-BZ we want dedicated counters for this. */ 227 IP6STAT_INC(ip6s_reassembled); 228 in6_ifstat_inc(dstifp, ifs6_reass_ok); 229 *offp = offset; 230 return (ip6f->ip6f_nxt); 231 } 232 233 IP6Q_LOCK(); 234 235 /* 236 * Enforce upper bound on number of fragments. 237 * If maxfrag is 0, never accept fragments. 238 * If maxfrag is -1, accept all fragments without limitation. 239 */ 240 if (V_ip6_maxfrags < 0) 241 ; 242 else if (V_frag6_nfrags >= (u_int)V_ip6_maxfrags) 243 goto dropfrag; 244 245 for (q6 = V_ip6q.ip6q_next; q6 != &V_ip6q; q6 = q6->ip6q_next) 246 if (ip6f->ip6f_ident == q6->ip6q_ident && 247 IN6_ARE_ADDR_EQUAL(&ip6->ip6_src, &q6->ip6q_src) && 248 IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst, &q6->ip6q_dst) 249 #ifdef MAC 250 && mac_ip6q_match(m, q6) 251 #endif 252 ) 253 break; 254 255 if (q6 == &V_ip6q) { 256 /* 257 * the first fragment to arrive, create a reassembly queue. 258 */ 259 first_frag = 1; 260 261 /* 262 * Enforce upper bound on number of fragmented packets 263 * for which we attempt reassembly; 264 * If maxfragpackets is 0, never accept fragments. 265 * If maxfragpackets is -1, accept all fragments without 266 * limitation. 267 */ 268 if (V_ip6_maxfragpackets < 0) 269 ; 270 else if (V_frag6_nfragpackets >= (u_int)V_ip6_maxfragpackets) 271 goto dropfrag; 272 V_frag6_nfragpackets++; 273 q6 = (struct ip6q *)malloc(sizeof(struct ip6q), M_FTABLE, 274 M_NOWAIT); 275 if (q6 == NULL) 276 goto dropfrag; 277 bzero(q6, sizeof(*q6)); 278 #ifdef MAC 279 if (mac_ip6q_init(q6, M_NOWAIT) != 0) { 280 free(q6, M_FTABLE); 281 goto dropfrag; 282 } 283 mac_ip6q_create(m, q6); 284 #endif 285 frag6_insque(q6, &V_ip6q); 286 287 /* ip6q_nxt will be filled afterwards, from 1st fragment */ 288 q6->ip6q_down = q6->ip6q_up = (struct ip6asfrag *)q6; 289 #ifdef notyet 290 q6->ip6q_nxtp = (u_char *)nxtp; 291 #endif 292 q6->ip6q_ident = ip6f->ip6f_ident; 293 q6->ip6q_ttl = IPV6_FRAGTTL; 294 q6->ip6q_src = ip6->ip6_src; 295 q6->ip6q_dst = ip6->ip6_dst; 296 q6->ip6q_ecn = 297 (ntohl(ip6->ip6_flow) >> 20) & IPTOS_ECN_MASK; 298 q6->ip6q_unfrglen = -1; /* The 1st fragment has not arrived. */ 299 300 q6->ip6q_nfrag = 0; 301 } 302 303 /* 304 * If it's the 1st fragment, record the length of the 305 * unfragmentable part and the next header of the fragment header. 306 */ 307 fragoff = ntohs(ip6f->ip6f_offlg & IP6F_OFF_MASK); 308 if (fragoff == 0) { 309 q6->ip6q_unfrglen = offset - sizeof(struct ip6_hdr) - 310 sizeof(struct ip6_frag); 311 q6->ip6q_nxt = ip6f->ip6f_nxt; 312 } 313 314 /* 315 * Check that the reassembled packet would not exceed 65535 bytes 316 * in size. 317 * If it would exceed, discard the fragment and return an ICMP error. 318 */ 319 frgpartlen = sizeof(struct ip6_hdr) + ntohs(ip6->ip6_plen) - offset; 320 if (q6->ip6q_unfrglen >= 0) { 321 /* The 1st fragment has already arrived. */ 322 if (q6->ip6q_unfrglen + fragoff + frgpartlen > IPV6_MAXPACKET) { 323 icmp6_error(m, ICMP6_PARAM_PROB, ICMP6_PARAMPROB_HEADER, 324 offset - sizeof(struct ip6_frag) + 325 offsetof(struct ip6_frag, ip6f_offlg)); 326 IP6Q_UNLOCK(); 327 return (IPPROTO_DONE); 328 } 329 } else if (fragoff + frgpartlen > IPV6_MAXPACKET) { 330 icmp6_error(m, ICMP6_PARAM_PROB, ICMP6_PARAMPROB_HEADER, 331 offset - sizeof(struct ip6_frag) + 332 offsetof(struct ip6_frag, ip6f_offlg)); 333 IP6Q_UNLOCK(); 334 return (IPPROTO_DONE); 335 } 336 /* 337 * If it's the first fragment, do the above check for each 338 * fragment already stored in the reassembly queue. 339 */ 340 if (fragoff == 0) { 341 for (af6 = q6->ip6q_down; af6 != (struct ip6asfrag *)q6; 342 af6 = af6dwn) { 343 af6dwn = af6->ip6af_down; 344 345 if (q6->ip6q_unfrglen + af6->ip6af_off + af6->ip6af_frglen > 346 IPV6_MAXPACKET) { 347 struct mbuf *merr = IP6_REASS_MBUF(af6); 348 struct ip6_hdr *ip6err; 349 int erroff = af6->ip6af_offset; 350 351 /* dequeue the fragment. */ 352 frag6_deq(af6); 353 free(af6, M_FTABLE); 354 355 /* adjust pointer. */ 356 ip6err = mtod(merr, struct ip6_hdr *); 357 358 /* 359 * Restore source and destination addresses 360 * in the erroneous IPv6 header. 361 */ 362 ip6err->ip6_src = q6->ip6q_src; 363 ip6err->ip6_dst = q6->ip6q_dst; 364 365 icmp6_error(merr, ICMP6_PARAM_PROB, 366 ICMP6_PARAMPROB_HEADER, 367 erroff - sizeof(struct ip6_frag) + 368 offsetof(struct ip6_frag, ip6f_offlg)); 369 } 370 } 371 } 372 373 ip6af = (struct ip6asfrag *)malloc(sizeof(struct ip6asfrag), M_FTABLE, 374 M_NOWAIT); 375 if (ip6af == NULL) 376 goto dropfrag; 377 bzero(ip6af, sizeof(*ip6af)); 378 ip6af->ip6af_mff = ip6f->ip6f_offlg & IP6F_MORE_FRAG; 379 ip6af->ip6af_off = fragoff; 380 ip6af->ip6af_frglen = frgpartlen; 381 ip6af->ip6af_offset = offset; 382 IP6_REASS_MBUF(ip6af) = m; 383 384 if (first_frag) { 385 af6 = (struct ip6asfrag *)q6; 386 goto insert; 387 } 388 389 /* 390 * Handle ECN by comparing this segment with the first one; 391 * if CE is set, do not lose CE. 392 * drop if CE and not-ECT are mixed for the same packet. 393 */ 394 ecn = (ntohl(ip6->ip6_flow) >> 20) & IPTOS_ECN_MASK; 395 ecn0 = q6->ip6q_ecn; 396 if (ecn == IPTOS_ECN_CE) { 397 if (ecn0 == IPTOS_ECN_NOTECT) { 398 free(ip6af, M_FTABLE); 399 goto dropfrag; 400 } 401 if (ecn0 != IPTOS_ECN_CE) 402 q6->ip6q_ecn = IPTOS_ECN_CE; 403 } 404 if (ecn == IPTOS_ECN_NOTECT && ecn0 != IPTOS_ECN_NOTECT) { 405 free(ip6af, M_FTABLE); 406 goto dropfrag; 407 } 408 409 /* 410 * Find a segment which begins after this one does. 411 */ 412 for (af6 = q6->ip6q_down; af6 != (struct ip6asfrag *)q6; 413 af6 = af6->ip6af_down) 414 if (af6->ip6af_off > ip6af->ip6af_off) 415 break; 416 417 #if 0 418 /* 419 * If there is a preceding segment, it may provide some of 420 * our data already. If so, drop the data from the incoming 421 * segment. If it provides all of our data, drop us. 422 */ 423 if (af6->ip6af_up != (struct ip6asfrag *)q6) { 424 i = af6->ip6af_up->ip6af_off + af6->ip6af_up->ip6af_frglen 425 - ip6af->ip6af_off; 426 if (i > 0) { 427 if (i >= ip6af->ip6af_frglen) 428 goto dropfrag; 429 m_adj(IP6_REASS_MBUF(ip6af), i); 430 ip6af->ip6af_off += i; 431 ip6af->ip6af_frglen -= i; 432 } 433 } 434 435 /* 436 * While we overlap succeeding segments trim them or, 437 * if they are completely covered, dequeue them. 438 */ 439 while (af6 != (struct ip6asfrag *)q6 && 440 ip6af->ip6af_off + ip6af->ip6af_frglen > af6->ip6af_off) { 441 i = (ip6af->ip6af_off + ip6af->ip6af_frglen) - af6->ip6af_off; 442 if (i < af6->ip6af_frglen) { 443 af6->ip6af_frglen -= i; 444 af6->ip6af_off += i; 445 m_adj(IP6_REASS_MBUF(af6), i); 446 break; 447 } 448 af6 = af6->ip6af_down; 449 m_freem(IP6_REASS_MBUF(af6->ip6af_up)); 450 frag6_deq(af6->ip6af_up); 451 } 452 #else 453 /* 454 * If the incoming framgent overlaps some existing fragments in 455 * the reassembly queue, drop it, since it is dangerous to override 456 * existing fragments from a security point of view. 457 * We don't know which fragment is the bad guy - here we trust 458 * fragment that came in earlier, with no real reason. 459 * 460 * Note: due to changes after disabling this part, mbuf passed to 461 * m_adj() below now does not meet the requirement. 462 */ 463 if (af6->ip6af_up != (struct ip6asfrag *)q6) { 464 i = af6->ip6af_up->ip6af_off + af6->ip6af_up->ip6af_frglen 465 - ip6af->ip6af_off; 466 if (i > 0) { 467 #if 0 /* suppress the noisy log */ 468 log(LOG_ERR, "%d bytes of a fragment from %s " 469 "overlaps the previous fragment\n", 470 i, ip6_sprintf(ip6buf, &q6->ip6q_src)); 471 #endif 472 free(ip6af, M_FTABLE); 473 goto dropfrag; 474 } 475 } 476 if (af6 != (struct ip6asfrag *)q6) { 477 i = (ip6af->ip6af_off + ip6af->ip6af_frglen) - af6->ip6af_off; 478 if (i > 0) { 479 #if 0 /* suppress the noisy log */ 480 log(LOG_ERR, "%d bytes of a fragment from %s " 481 "overlaps the succeeding fragment", 482 i, ip6_sprintf(ip6buf, &q6->ip6q_src)); 483 #endif 484 free(ip6af, M_FTABLE); 485 goto dropfrag; 486 } 487 } 488 #endif 489 490 insert: 491 #ifdef MAC 492 if (!first_frag) 493 mac_ip6q_update(m, q6); 494 #endif 495 496 /* 497 * Stick new segment in its place; 498 * check for complete reassembly. 499 * Move to front of packet queue, as we are 500 * the most recently active fragmented packet. 501 */ 502 frag6_enq(ip6af, af6->ip6af_up); 503 V_frag6_nfrags++; 504 q6->ip6q_nfrag++; 505 #if 0 /* xxx */ 506 if (q6 != V_ip6q.ip6q_next) { 507 frag6_remque(q6); 508 frag6_insque(q6, &V_ip6q); 509 } 510 #endif 511 next = 0; 512 for (af6 = q6->ip6q_down; af6 != (struct ip6asfrag *)q6; 513 af6 = af6->ip6af_down) { 514 if (af6->ip6af_off != next) { 515 IP6Q_UNLOCK(); 516 return IPPROTO_DONE; 517 } 518 next += af6->ip6af_frglen; 519 } 520 if (af6->ip6af_up->ip6af_mff) { 521 IP6Q_UNLOCK(); 522 return IPPROTO_DONE; 523 } 524 525 /* 526 * Reassembly is complete; concatenate fragments. 527 */ 528 ip6af = q6->ip6q_down; 529 t = m = IP6_REASS_MBUF(ip6af); 530 af6 = ip6af->ip6af_down; 531 frag6_deq(ip6af); 532 while (af6 != (struct ip6asfrag *)q6) { 533 m->m_pkthdr.csum_flags &= 534 IP6_REASS_MBUF(af6)->m_pkthdr.csum_flags; 535 m->m_pkthdr.csum_data += 536 IP6_REASS_MBUF(af6)->m_pkthdr.csum_data; 537 538 af6dwn = af6->ip6af_down; 539 frag6_deq(af6); 540 while (t->m_next) 541 t = t->m_next; 542 m_adj(IP6_REASS_MBUF(af6), af6->ip6af_offset); 543 m_cat(t, IP6_REASS_MBUF(af6)); 544 free(af6, M_FTABLE); 545 af6 = af6dwn; 546 } 547 548 while (m->m_pkthdr.csum_data & 0xffff0000) 549 m->m_pkthdr.csum_data = (m->m_pkthdr.csum_data & 0xffff) + 550 (m->m_pkthdr.csum_data >> 16); 551 552 /* adjust offset to point where the original next header starts */ 553 offset = ip6af->ip6af_offset - sizeof(struct ip6_frag); 554 free(ip6af, M_FTABLE); 555 ip6 = mtod(m, struct ip6_hdr *); 556 ip6->ip6_plen = htons((u_short)next + offset - sizeof(struct ip6_hdr)); 557 if (q6->ip6q_ecn == IPTOS_ECN_CE) 558 ip6->ip6_flow |= htonl(IPTOS_ECN_CE << 20); 559 nxt = q6->ip6q_nxt; 560 #ifdef notyet 561 *q6->ip6q_nxtp = (u_char)(nxt & 0xff); 562 #endif 563 564 if (ip6_deletefraghdr(m, offset, M_NOWAIT) != 0) { 565 frag6_remque(q6); 566 V_frag6_nfrags -= q6->ip6q_nfrag; 567 #ifdef MAC 568 mac_ip6q_destroy(q6); 569 #endif 570 free(q6, M_FTABLE); 571 V_frag6_nfragpackets--; 572 573 goto dropfrag; 574 } 575 576 /* 577 * Store NXT to the original. 578 */ 579 { 580 char *prvnxtp = ip6_get_prevhdr(m, offset); /* XXX */ 581 *prvnxtp = nxt; 582 } 583 584 frag6_remque(q6); 585 V_frag6_nfrags -= q6->ip6q_nfrag; 586 #ifdef MAC 587 mac_ip6q_reassemble(q6, m); 588 mac_ip6q_destroy(q6); 589 #endif 590 free(q6, M_FTABLE); 591 V_frag6_nfragpackets--; 592 593 if (m->m_flags & M_PKTHDR) { /* Isn't it always true? */ 594 int plen = 0; 595 for (t = m; t; t = t->m_next) 596 plen += t->m_len; 597 m->m_pkthdr.len = plen; 598 } 599 600 #ifdef RSS 601 mtag = m_tag_alloc(MTAG_ABI_IPV6, IPV6_TAG_DIRECT, sizeof(*ip6dc), 602 M_NOWAIT); 603 if (mtag == NULL) 604 goto dropfrag; 605 606 ip6dc = (struct ip6_direct_ctx *)(mtag + 1); 607 ip6dc->ip6dc_nxt = nxt; 608 ip6dc->ip6dc_off = offset; 609 610 m_tag_prepend(m, mtag); 611 #endif 612 613 IP6Q_UNLOCK(); 614 IP6STAT_INC(ip6s_reassembled); 615 in6_ifstat_inc(dstifp, ifs6_reass_ok); 616 617 #ifdef RSS 618 /* 619 * Queue/dispatch for reprocessing. 620 */ 621 netisr_dispatch(NETISR_IPV6_DIRECT, m); 622 return IPPROTO_DONE; 623 #endif 624 625 /* 626 * Tell launch routine the next header 627 */ 628 629 *mp = m; 630 *offp = offset; 631 632 return nxt; 633 634 dropfrag: 635 IP6Q_UNLOCK(); 636 in6_ifstat_inc(dstifp, ifs6_reass_fail); 637 IP6STAT_INC(ip6s_fragdropped); 638 m_freem(m); 639 return IPPROTO_DONE; 640 } 641 642 /* 643 * Free a fragment reassembly header and all 644 * associated datagrams. 645 */ 646 void 647 frag6_freef(struct ip6q *q6) 648 { 649 struct ip6asfrag *af6, *down6; 650 651 IP6Q_LOCK_ASSERT(); 652 653 for (af6 = q6->ip6q_down; af6 != (struct ip6asfrag *)q6; 654 af6 = down6) { 655 struct mbuf *m = IP6_REASS_MBUF(af6); 656 657 down6 = af6->ip6af_down; 658 frag6_deq(af6); 659 660 /* 661 * Return ICMP time exceeded error for the 1st fragment. 662 * Just free other fragments. 663 */ 664 if (af6->ip6af_off == 0) { 665 struct ip6_hdr *ip6; 666 667 /* adjust pointer */ 668 ip6 = mtod(m, struct ip6_hdr *); 669 670 /* restore source and destination addresses */ 671 ip6->ip6_src = q6->ip6q_src; 672 ip6->ip6_dst = q6->ip6q_dst; 673 674 icmp6_error(m, ICMP6_TIME_EXCEEDED, 675 ICMP6_TIME_EXCEED_REASSEMBLY, 0); 676 } else 677 m_freem(m); 678 free(af6, M_FTABLE); 679 } 680 frag6_remque(q6); 681 V_frag6_nfrags -= q6->ip6q_nfrag; 682 #ifdef MAC 683 mac_ip6q_destroy(q6); 684 #endif 685 free(q6, M_FTABLE); 686 V_frag6_nfragpackets--; 687 } 688 689 /* 690 * Put an ip fragment on a reassembly chain. 691 * Like insque, but pointers in middle of structure. 692 */ 693 void 694 frag6_enq(struct ip6asfrag *af6, struct ip6asfrag *up6) 695 { 696 697 IP6Q_LOCK_ASSERT(); 698 699 af6->ip6af_up = up6; 700 af6->ip6af_down = up6->ip6af_down; 701 up6->ip6af_down->ip6af_up = af6; 702 up6->ip6af_down = af6; 703 } 704 705 /* 706 * To frag6_enq as remque is to insque. 707 */ 708 void 709 frag6_deq(struct ip6asfrag *af6) 710 { 711 712 IP6Q_LOCK_ASSERT(); 713 714 af6->ip6af_up->ip6af_down = af6->ip6af_down; 715 af6->ip6af_down->ip6af_up = af6->ip6af_up; 716 } 717 718 void 719 frag6_insque(struct ip6q *new, struct ip6q *old) 720 { 721 722 IP6Q_LOCK_ASSERT(); 723 724 new->ip6q_prev = old; 725 new->ip6q_next = old->ip6q_next; 726 old->ip6q_next->ip6q_prev= new; 727 old->ip6q_next = new; 728 } 729 730 void 731 frag6_remque(struct ip6q *p6) 732 { 733 734 IP6Q_LOCK_ASSERT(); 735 736 p6->ip6q_prev->ip6q_next = p6->ip6q_next; 737 p6->ip6q_next->ip6q_prev = p6->ip6q_prev; 738 } 739 740 /* 741 * IPv6 reassembling timer processing; 742 * if a timer expires on a reassembly 743 * queue, discard it. 744 */ 745 void 746 frag6_slowtimo(void) 747 { 748 VNET_ITERATOR_DECL(vnet_iter); 749 struct ip6q *q6; 750 751 VNET_LIST_RLOCK_NOSLEEP(); 752 IP6Q_LOCK(); 753 VNET_FOREACH(vnet_iter) { 754 CURVNET_SET(vnet_iter); 755 q6 = V_ip6q.ip6q_next; 756 if (q6) 757 while (q6 != &V_ip6q) { 758 --q6->ip6q_ttl; 759 q6 = q6->ip6q_next; 760 if (q6->ip6q_prev->ip6q_ttl == 0) { 761 IP6STAT_INC(ip6s_fragtimeout); 762 /* XXX in6_ifstat_inc(ifp, ifs6_reass_fail) */ 763 frag6_freef(q6->ip6q_prev); 764 } 765 } 766 /* 767 * If we are over the maximum number of fragments 768 * (due to the limit being lowered), drain off 769 * enough to get down to the new limit. 770 */ 771 while (V_frag6_nfragpackets > (u_int)V_ip6_maxfragpackets && 772 V_ip6q.ip6q_prev) { 773 IP6STAT_INC(ip6s_fragoverflow); 774 /* XXX in6_ifstat_inc(ifp, ifs6_reass_fail) */ 775 frag6_freef(V_ip6q.ip6q_prev); 776 } 777 CURVNET_RESTORE(); 778 } 779 IP6Q_UNLOCK(); 780 VNET_LIST_RUNLOCK_NOSLEEP(); 781 } 782 783 /* 784 * Drain off all datagram fragments. 785 */ 786 void 787 frag6_drain(void) 788 { 789 VNET_ITERATOR_DECL(vnet_iter); 790 791 VNET_LIST_RLOCK_NOSLEEP(); 792 if (IP6Q_TRYLOCK() == 0) { 793 VNET_LIST_RUNLOCK_NOSLEEP(); 794 return; 795 } 796 VNET_FOREACH(vnet_iter) { 797 CURVNET_SET(vnet_iter); 798 while (V_ip6q.ip6q_next != &V_ip6q) { 799 IP6STAT_INC(ip6s_fragdropped); 800 /* XXX in6_ifstat_inc(ifp, ifs6_reass_fail) */ 801 frag6_freef(V_ip6q.ip6q_next); 802 } 803 CURVNET_RESTORE(); 804 } 805 IP6Q_UNLOCK(); 806 VNET_LIST_RUNLOCK_NOSLEEP(); 807 } 808 809 int 810 ip6_deletefraghdr(struct mbuf *m, int offset, int wait) 811 { 812 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *); 813 struct mbuf *t; 814 815 /* Delete frag6 header. */ 816 if (m->m_len >= offset + sizeof(struct ip6_frag)) { 817 /* This is the only possible case with !PULLDOWN_TEST. */ 818 bcopy(ip6, (char *)ip6 + sizeof(struct ip6_frag), 819 offset); 820 m->m_data += sizeof(struct ip6_frag); 821 m->m_len -= sizeof(struct ip6_frag); 822 } else { 823 /* This comes with no copy if the boundary is on cluster. */ 824 if ((t = m_split(m, offset, wait)) == NULL) 825 return (ENOMEM); 826 m_adj(t, sizeof(struct ip6_frag)); 827 m_cat(m, t); 828 } 829 830 return (0); 831 } 832