| 1 | /* $NetBSD: ip_reass.c,v 1.10 2016/04/26 08:44:44 ozaki-r Exp $ */ |
| 2 | |
| 3 | /* |
| 4 | * Copyright (c) 1982, 1986, 1988, 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 | * @(#)ip_input.c 8.2 (Berkeley) 1/4/94 |
| 32 | */ |
| 33 | |
| 34 | /* |
| 35 | * IP reassembly. |
| 36 | * |
| 37 | * Additive-Increase/Multiplicative-Decrease (AIMD) strategy for IP |
| 38 | * reassembly queue buffer managment. |
| 39 | * |
| 40 | * We keep a count of total IP fragments (NB: not fragmented packets), |
| 41 | * awaiting reassembly (ip_nfrags) and a limit (ip_maxfrags) on fragments. |
| 42 | * If ip_nfrags exceeds ip_maxfrags the limit, we drop half the total |
| 43 | * fragments in reassembly queues. This AIMD policy avoids repeatedly |
| 44 | * deleting single packets under heavy fragmentation load (e.g., from lossy |
| 45 | * NFS peers). |
| 46 | */ |
| 47 | |
| 48 | #include <sys/cdefs.h> |
| 49 | __KERNEL_RCSID(0, "$NetBSD: ip_reass.c,v 1.10 2016/04/26 08:44:44 ozaki-r Exp $" ); |
| 50 | |
| 51 | #include <sys/param.h> |
| 52 | #include <sys/types.h> |
| 53 | |
| 54 | #include <sys/malloc.h> |
| 55 | #include <sys/mbuf.h> |
| 56 | #include <sys/mutex.h> |
| 57 | #include <sys/domain.h> |
| 58 | #include <sys/protosw.h> |
| 59 | #include <sys/pool.h> |
| 60 | #include <sys/queue.h> |
| 61 | #include <sys/sysctl.h> |
| 62 | #include <sys/systm.h> |
| 63 | |
| 64 | #include <net/if.h> |
| 65 | |
| 66 | #include <netinet/in.h> |
| 67 | #include <netinet/in_systm.h> |
| 68 | #include <netinet/ip.h> |
| 69 | #include <netinet/in_pcb.h> |
| 70 | #include <netinet/ip_var.h> |
| 71 | #include <netinet/in_proto.h> |
| 72 | #include <netinet/ip_private.h> |
| 73 | #include <netinet/in_var.h> |
| 74 | |
| 75 | /* |
| 76 | * IP reassembly queue structures. Each fragment being reassembled is |
| 77 | * attached to one of these structures. They are timed out after TTL |
| 78 | * drops to 0, and may also be reclaimed if memory becomes tight. |
| 79 | */ |
| 80 | |
| 81 | typedef struct ipfr_qent { |
| 82 | TAILQ_ENTRY(ipfr_qent) ipqe_q; |
| 83 | struct ip * ipqe_ip; |
| 84 | struct mbuf * ipqe_m; |
| 85 | bool ipqe_mff; |
| 86 | } ipfr_qent_t; |
| 87 | |
| 88 | TAILQ_HEAD(ipfr_qent_head, ipfr_qent); |
| 89 | |
| 90 | typedef struct ipfr_queue { |
| 91 | LIST_ENTRY(ipfr_queue) ipq_q; /* to other reass headers */ |
| 92 | struct ipfr_qent_head ipq_fragq; /* queue of fragment entries */ |
| 93 | uint8_t ipq_ttl; /* time for reass q to live */ |
| 94 | uint8_t ipq_p; /* protocol of this fragment */ |
| 95 | uint16_t ipq_id; /* sequence id for reassembly */ |
| 96 | struct in_addr ipq_src; |
| 97 | struct in_addr ipq_dst; |
| 98 | uint16_t ipq_nfrags; /* frags in this queue entry */ |
| 99 | uint8_t ipq_tos; /* TOS of this fragment */ |
| 100 | } ipfr_queue_t; |
| 101 | |
| 102 | /* |
| 103 | * Hash table of IP reassembly queues. |
| 104 | */ |
| 105 | #define IPREASS_HASH_SHIFT 6 |
| 106 | #define IPREASS_HASH_SIZE (1 << IPREASS_HASH_SHIFT) |
| 107 | #define IPREASS_HASH_MASK (IPREASS_HASH_SIZE - 1) |
| 108 | #define IPREASS_HASH(x, y) \ |
| 109 | (((((x) & 0xf) | ((((x) >> 8) & 0xf) << 4)) ^ (y)) & IPREASS_HASH_MASK) |
| 110 | |
| 111 | static LIST_HEAD(, ipfr_queue) ip_frags[IPREASS_HASH_SIZE]; |
| 112 | static pool_cache_t ipfren_cache; |
| 113 | static kmutex_t ipfr_lock; |
| 114 | |
| 115 | /* Number of packets in reassembly queue and total number of fragments. */ |
| 116 | static int ip_nfragpackets; |
| 117 | static int ip_nfrags; |
| 118 | |
| 119 | /* Limits on packet and fragments. */ |
| 120 | static int ip_maxfragpackets; |
| 121 | static int ip_maxfrags; |
| 122 | |
| 123 | /* |
| 124 | * Cached copy of nmbclusters. If nbclusters is different, recalculate |
| 125 | * IP parameters derived from nmbclusters. |
| 126 | */ |
| 127 | static int ip_nmbclusters; |
| 128 | |
| 129 | /* |
| 130 | * IP reassembly TTL machinery for multiplicative drop. |
| 131 | */ |
| 132 | static u_int fragttl_histo[IPFRAGTTL + 1]; |
| 133 | |
| 134 | static struct sysctllog *ip_reass_sysctllog; |
| 135 | |
| 136 | void sysctl_ip_reass_setup(void); |
| 137 | static void ip_nmbclusters_changed(void); |
| 138 | |
| 139 | static struct mbuf * ip_reass(ipfr_qent_t *, ipfr_queue_t *, u_int); |
| 140 | static u_int ip_reass_ttl_decr(u_int ticks); |
| 141 | static void ip_reass_drophalf(void); |
| 142 | static void ip_freef(ipfr_queue_t *); |
| 143 | |
| 144 | /* |
| 145 | * ip_reass_init: |
| 146 | * |
| 147 | * Initialization of IP reassembly mechanism. |
| 148 | */ |
| 149 | void |
| 150 | ip_reass_init(void) |
| 151 | { |
| 152 | int i; |
| 153 | |
| 154 | ipfren_cache = pool_cache_init(sizeof(ipfr_qent_t), coherency_unit, |
| 155 | 0, 0, "ipfrenpl" , NULL, IPL_NET, NULL, NULL, NULL); |
| 156 | mutex_init(&ipfr_lock, MUTEX_DEFAULT, IPL_VM); |
| 157 | |
| 158 | for (i = 0; i < IPREASS_HASH_SIZE; i++) { |
| 159 | LIST_INIT(&ip_frags[i]); |
| 160 | } |
| 161 | ip_maxfragpackets = 200; |
| 162 | ip_maxfrags = 0; |
| 163 | ip_nmbclusters_changed(); |
| 164 | |
| 165 | sysctl_ip_reass_setup(); |
| 166 | } |
| 167 | |
| 168 | void |
| 169 | sysctl_ip_reass_setup(void) |
| 170 | { |
| 171 | |
| 172 | sysctl_createv(&ip_reass_sysctllog, 0, NULL, NULL, |
| 173 | CTLFLAG_PERMANENT, |
| 174 | CTLTYPE_NODE, "inet" , |
| 175 | SYSCTL_DESCR("PF_INET related settings" ), |
| 176 | NULL, 0, NULL, 0, |
| 177 | CTL_NET, PF_INET, CTL_EOL); |
| 178 | sysctl_createv(&ip_reass_sysctllog, 0, NULL, NULL, |
| 179 | CTLFLAG_PERMANENT, |
| 180 | CTLTYPE_NODE, "ip" , |
| 181 | SYSCTL_DESCR("IPv4 related settings" ), |
| 182 | NULL, 0, NULL, 0, |
| 183 | CTL_NET, PF_INET, IPPROTO_IP, CTL_EOL); |
| 184 | |
| 185 | sysctl_createv(&ip_reass_sysctllog, 0, NULL, NULL, |
| 186 | CTLFLAG_PERMANENT|CTLFLAG_READWRITE, |
| 187 | CTLTYPE_INT, "maxfragpackets" , |
| 188 | SYSCTL_DESCR("Maximum number of fragments to retain for " |
| 189 | "possible reassembly" ), |
| 190 | NULL, 0, &ip_maxfragpackets, 0, |
| 191 | CTL_NET, PF_INET, IPPROTO_IP, IPCTL_MAXFRAGPACKETS, CTL_EOL); |
| 192 | } |
| 193 | |
| 194 | #define CHECK_NMBCLUSTER_PARAMS() \ |
| 195 | do { \ |
| 196 | if (__predict_false(ip_nmbclusters != nmbclusters)) \ |
| 197 | ip_nmbclusters_changed(); \ |
| 198 | } while (/*CONSTCOND*/0) |
| 199 | |
| 200 | /* |
| 201 | * Compute IP limits derived from the value of nmbclusters. |
| 202 | */ |
| 203 | static void |
| 204 | ip_nmbclusters_changed(void) |
| 205 | { |
| 206 | ip_maxfrags = nmbclusters / 4; |
| 207 | ip_nmbclusters = nmbclusters; |
| 208 | } |
| 209 | |
| 210 | /* |
| 211 | * ip_reass: |
| 212 | * |
| 213 | * Take incoming datagram fragment and try to reassemble it into whole |
| 214 | * datagram. If a chain for reassembly of this datagram already exists, |
| 215 | * then it is given as 'fp'; otherwise have to make a chain. |
| 216 | */ |
| 217 | struct mbuf * |
| 218 | ip_reass(ipfr_qent_t *ipqe, ipfr_queue_t *fp, const u_int hash) |
| 219 | { |
| 220 | struct ip *ip = ipqe->ipqe_ip, *qip; |
| 221 | const int hlen = ip->ip_hl << 2; |
| 222 | struct mbuf *m = ipqe->ipqe_m, *t; |
| 223 | ipfr_qent_t *nq, *p, *q; |
| 224 | int i, next; |
| 225 | |
| 226 | KASSERT(mutex_owned(&ipfr_lock)); |
| 227 | |
| 228 | /* |
| 229 | * Presence of header sizes in mbufs would confuse code below. |
| 230 | */ |
| 231 | m->m_data += hlen; |
| 232 | m->m_len -= hlen; |
| 233 | |
| 234 | #ifdef notyet |
| 235 | /* Make sure fragment limit is up-to-date. */ |
| 236 | CHECK_NMBCLUSTER_PARAMS(); |
| 237 | |
| 238 | /* If we have too many fragments, drop the older half. */ |
| 239 | if (ip_nfrags >= ip_maxfrags) { |
| 240 | ip_reass_drophalf(void); |
| 241 | } |
| 242 | #endif |
| 243 | |
| 244 | /* |
| 245 | * We are about to add a fragment; increment frag count. |
| 246 | */ |
| 247 | ip_nfrags++; |
| 248 | |
| 249 | /* |
| 250 | * If first fragment to arrive, create a reassembly queue. |
| 251 | */ |
| 252 | if (fp == NULL) { |
| 253 | /* |
| 254 | * Enforce upper bound on number of fragmented packets |
| 255 | * for which we attempt reassembly: a) if maxfrag is 0, |
| 256 | * never accept fragments b) if maxfrag is -1, accept |
| 257 | * all fragments without limitation. |
| 258 | */ |
| 259 | if (ip_maxfragpackets < 0) |
| 260 | ; |
| 261 | else if (ip_nfragpackets >= ip_maxfragpackets) { |
| 262 | goto dropfrag; |
| 263 | } |
| 264 | fp = malloc(sizeof(ipfr_queue_t), M_FTABLE, M_NOWAIT); |
| 265 | if (fp == NULL) { |
| 266 | goto dropfrag; |
| 267 | } |
| 268 | ip_nfragpackets++; |
| 269 | TAILQ_INIT(&fp->ipq_fragq); |
| 270 | fp->ipq_nfrags = 1; |
| 271 | fp->ipq_ttl = IPFRAGTTL; |
| 272 | fp->ipq_p = ip->ip_p; |
| 273 | fp->ipq_id = ip->ip_id; |
| 274 | fp->ipq_tos = ip->ip_tos; |
| 275 | fp->ipq_src = ip->ip_src; |
| 276 | fp->ipq_dst = ip->ip_dst; |
| 277 | LIST_INSERT_HEAD(&ip_frags[hash], fp, ipq_q); |
| 278 | p = NULL; |
| 279 | goto insert; |
| 280 | } else { |
| 281 | fp->ipq_nfrags++; |
| 282 | } |
| 283 | |
| 284 | /* |
| 285 | * Find a segment which begins after this one does. |
| 286 | */ |
| 287 | TAILQ_FOREACH(q, &fp->ipq_fragq, ipqe_q) { |
| 288 | if (ntohs(q->ipqe_ip->ip_off) > ntohs(ip->ip_off)) |
| 289 | break; |
| 290 | } |
| 291 | if (q != NULL) { |
| 292 | p = TAILQ_PREV(q, ipfr_qent_head, ipqe_q); |
| 293 | } else { |
| 294 | p = TAILQ_LAST(&fp->ipq_fragq, ipfr_qent_head); |
| 295 | } |
| 296 | |
| 297 | /* |
| 298 | * If there is a preceding segment, it may provide some of our |
| 299 | * data already. If so, drop the data from the incoming segment. |
| 300 | * If it provides all of our data, drop us. |
| 301 | */ |
| 302 | if (p != NULL) { |
| 303 | i = ntohs(p->ipqe_ip->ip_off) + ntohs(p->ipqe_ip->ip_len) - |
| 304 | ntohs(ip->ip_off); |
| 305 | if (i > 0) { |
| 306 | if (i >= ntohs(ip->ip_len)) { |
| 307 | goto dropfrag; |
| 308 | } |
| 309 | m_adj(ipqe->ipqe_m, i); |
| 310 | ip->ip_off = htons(ntohs(ip->ip_off) + i); |
| 311 | ip->ip_len = htons(ntohs(ip->ip_len) - i); |
| 312 | } |
| 313 | } |
| 314 | |
| 315 | /* |
| 316 | * While we overlap succeeding segments trim them or, if they are |
| 317 | * completely covered, dequeue them. |
| 318 | */ |
| 319 | while (q != NULL) { |
| 320 | size_t end; |
| 321 | |
| 322 | qip = q->ipqe_ip; |
| 323 | end = ntohs(ip->ip_off) + ntohs(ip->ip_len); |
| 324 | if (end <= ntohs(qip->ip_off)) { |
| 325 | break; |
| 326 | } |
| 327 | i = end - ntohs(qip->ip_off); |
| 328 | if (i < ntohs(qip->ip_len)) { |
| 329 | qip->ip_len = htons(ntohs(qip->ip_len) - i); |
| 330 | qip->ip_off = htons(ntohs(qip->ip_off) + i); |
| 331 | m_adj(q->ipqe_m, i); |
| 332 | break; |
| 333 | } |
| 334 | nq = TAILQ_NEXT(q, ipqe_q); |
| 335 | m_freem(q->ipqe_m); |
| 336 | TAILQ_REMOVE(&fp->ipq_fragq, q, ipqe_q); |
| 337 | pool_cache_put(ipfren_cache, q); |
| 338 | fp->ipq_nfrags--; |
| 339 | ip_nfrags--; |
| 340 | q = nq; |
| 341 | } |
| 342 | |
| 343 | insert: |
| 344 | /* |
| 345 | * Stick new segment in its place; check for complete reassembly. |
| 346 | */ |
| 347 | if (p == NULL) { |
| 348 | TAILQ_INSERT_HEAD(&fp->ipq_fragq, ipqe, ipqe_q); |
| 349 | } else { |
| 350 | TAILQ_INSERT_AFTER(&fp->ipq_fragq, p, ipqe, ipqe_q); |
| 351 | } |
| 352 | next = 0; |
| 353 | TAILQ_FOREACH(q, &fp->ipq_fragq, ipqe_q) { |
| 354 | qip = q->ipqe_ip; |
| 355 | if (ntohs(qip->ip_off) != next) { |
| 356 | mutex_exit(&ipfr_lock); |
| 357 | return NULL; |
| 358 | } |
| 359 | next += ntohs(qip->ip_len); |
| 360 | } |
| 361 | p = TAILQ_LAST(&fp->ipq_fragq, ipfr_qent_head); |
| 362 | if (p->ipqe_mff) { |
| 363 | mutex_exit(&ipfr_lock); |
| 364 | return NULL; |
| 365 | } |
| 366 | |
| 367 | /* |
| 368 | * Reassembly is complete. Check for a bogus message size. |
| 369 | */ |
| 370 | q = TAILQ_FIRST(&fp->ipq_fragq); |
| 371 | ip = q->ipqe_ip; |
| 372 | if ((next + (ip->ip_hl << 2)) > IP_MAXPACKET) { |
| 373 | IP_STATINC(IP_STAT_TOOLONG); |
| 374 | ip_freef(fp); |
| 375 | mutex_exit(&ipfr_lock); |
| 376 | return NULL; |
| 377 | } |
| 378 | LIST_REMOVE(fp, ipq_q); |
| 379 | ip_nfrags -= fp->ipq_nfrags; |
| 380 | ip_nfragpackets--; |
| 381 | mutex_exit(&ipfr_lock); |
| 382 | |
| 383 | /* Concatenate all fragments. */ |
| 384 | m = q->ipqe_m; |
| 385 | t = m->m_next; |
| 386 | m->m_next = NULL; |
| 387 | m_cat(m, t); |
| 388 | nq = TAILQ_NEXT(q, ipqe_q); |
| 389 | pool_cache_put(ipfren_cache, q); |
| 390 | |
| 391 | for (q = nq; q != NULL; q = nq) { |
| 392 | t = q->ipqe_m; |
| 393 | nq = TAILQ_NEXT(q, ipqe_q); |
| 394 | pool_cache_put(ipfren_cache, q); |
| 395 | m_cat(m, t); |
| 396 | } |
| 397 | |
| 398 | /* |
| 399 | * Create header for new packet by modifying header of first |
| 400 | * packet. Dequeue and discard fragment reassembly header. Make |
| 401 | * header visible. |
| 402 | */ |
| 403 | ip->ip_len = htons((ip->ip_hl << 2) + next); |
| 404 | ip->ip_src = fp->ipq_src; |
| 405 | ip->ip_dst = fp->ipq_dst; |
| 406 | free(fp, M_FTABLE); |
| 407 | |
| 408 | m->m_len += (ip->ip_hl << 2); |
| 409 | m->m_data -= (ip->ip_hl << 2); |
| 410 | |
| 411 | /* Fix up mbuf. XXX This should be done elsewhere. */ |
| 412 | if (m->m_flags & M_PKTHDR) { |
| 413 | int plen = 0; |
| 414 | for (t = m; t; t = t->m_next) { |
| 415 | plen += t->m_len; |
| 416 | } |
| 417 | m->m_pkthdr.len = plen; |
| 418 | m->m_pkthdr.csum_flags = 0; |
| 419 | } |
| 420 | return m; |
| 421 | |
| 422 | dropfrag: |
| 423 | if (fp != NULL) { |
| 424 | fp->ipq_nfrags--; |
| 425 | } |
| 426 | ip_nfrags--; |
| 427 | IP_STATINC(IP_STAT_FRAGDROPPED); |
| 428 | mutex_exit(&ipfr_lock); |
| 429 | |
| 430 | pool_cache_put(ipfren_cache, ipqe); |
| 431 | m_freem(m); |
| 432 | return NULL; |
| 433 | } |
| 434 | |
| 435 | /* |
| 436 | * ip_freef: |
| 437 | * |
| 438 | * Free a fragment reassembly header and all associated datagrams. |
| 439 | */ |
| 440 | static void |
| 441 | ip_freef(ipfr_queue_t *fp) |
| 442 | { |
| 443 | ipfr_qent_t *q; |
| 444 | |
| 445 | KASSERT(mutex_owned(&ipfr_lock)); |
| 446 | |
| 447 | LIST_REMOVE(fp, ipq_q); |
| 448 | ip_nfrags -= fp->ipq_nfrags; |
| 449 | ip_nfragpackets--; |
| 450 | |
| 451 | while ((q = TAILQ_FIRST(&fp->ipq_fragq)) != NULL) { |
| 452 | TAILQ_REMOVE(&fp->ipq_fragq, q, ipqe_q); |
| 453 | m_freem(q->ipqe_m); |
| 454 | pool_cache_put(ipfren_cache, q); |
| 455 | } |
| 456 | free(fp, M_FTABLE); |
| 457 | } |
| 458 | |
| 459 | /* |
| 460 | * ip_reass_ttl_decr: |
| 461 | * |
| 462 | * Decrement TTL of all reasembly queue entries by `ticks'. Count |
| 463 | * number of distinct fragments (as opposed to partial, fragmented |
| 464 | * datagrams) inthe reassembly queue. While we traverse the entire |
| 465 | * reassembly queue, compute and return the median TTL over all |
| 466 | * fragments. |
| 467 | */ |
| 468 | static u_int |
| 469 | ip_reass_ttl_decr(u_int ticks) |
| 470 | { |
| 471 | u_int nfrags, median, dropfraction, keepfraction; |
| 472 | ipfr_queue_t *fp, *nfp; |
| 473 | int i; |
| 474 | |
| 475 | nfrags = 0; |
| 476 | memset(fragttl_histo, 0, sizeof(fragttl_histo)); |
| 477 | |
| 478 | for (i = 0; i < IPREASS_HASH_SIZE; i++) { |
| 479 | for (fp = LIST_FIRST(&ip_frags[i]); fp != NULL; fp = nfp) { |
| 480 | fp->ipq_ttl = ((fp->ipq_ttl <= ticks) ? |
| 481 | 0 : fp->ipq_ttl - ticks); |
| 482 | nfp = LIST_NEXT(fp, ipq_q); |
| 483 | if (fp->ipq_ttl == 0) { |
| 484 | IP_STATINC(IP_STAT_FRAGTIMEOUT); |
| 485 | ip_freef(fp); |
| 486 | } else { |
| 487 | nfrags += fp->ipq_nfrags; |
| 488 | fragttl_histo[fp->ipq_ttl] += fp->ipq_nfrags; |
| 489 | } |
| 490 | } |
| 491 | } |
| 492 | |
| 493 | KASSERT(ip_nfrags == nfrags); |
| 494 | |
| 495 | /* Find median (or other drop fraction) in histogram. */ |
| 496 | dropfraction = (ip_nfrags / 2); |
| 497 | keepfraction = ip_nfrags - dropfraction; |
| 498 | for (i = IPFRAGTTL, median = 0; i >= 0; i--) { |
| 499 | median += fragttl_histo[i]; |
| 500 | if (median >= keepfraction) |
| 501 | break; |
| 502 | } |
| 503 | |
| 504 | /* Return TTL of median (or other fraction). */ |
| 505 | return (u_int)i; |
| 506 | } |
| 507 | |
| 508 | static void |
| 509 | ip_reass_drophalf(void) |
| 510 | { |
| 511 | u_int median_ticks; |
| 512 | |
| 513 | KASSERT(mutex_owned(&ipfr_lock)); |
| 514 | |
| 515 | /* |
| 516 | * Compute median TTL of all fragments, and count frags |
| 517 | * with that TTL or lower (roughly half of all fragments). |
| 518 | */ |
| 519 | median_ticks = ip_reass_ttl_decr(0); |
| 520 | |
| 521 | /* Drop half. */ |
| 522 | median_ticks = ip_reass_ttl_decr(median_ticks); |
| 523 | } |
| 524 | |
| 525 | /* |
| 526 | * ip_reass_drain: drain off all datagram fragments. Do not acquire |
| 527 | * softnet_lock as can be called from hardware interrupt context. |
| 528 | */ |
| 529 | void |
| 530 | ip_reass_drain(void) |
| 531 | { |
| 532 | |
| 533 | /* |
| 534 | * We may be called from a device's interrupt context. If |
| 535 | * the ipq is already busy, just bail out now. |
| 536 | */ |
| 537 | if (mutex_tryenter(&ipfr_lock)) { |
| 538 | /* |
| 539 | * Drop half the total fragments now. If more mbufs are |
| 540 | * needed, we will be called again soon. |
| 541 | */ |
| 542 | ip_reass_drophalf(); |
| 543 | mutex_exit(&ipfr_lock); |
| 544 | } |
| 545 | } |
| 546 | |
| 547 | /* |
| 548 | * ip_reass_slowtimo: |
| 549 | * |
| 550 | * If a timer expires on a reassembly queue, discard it. |
| 551 | */ |
| 552 | void |
| 553 | ip_reass_slowtimo(void) |
| 554 | { |
| 555 | static u_int dropscanidx = 0; |
| 556 | u_int i, median_ttl; |
| 557 | |
| 558 | mutex_enter(&ipfr_lock); |
| 559 | |
| 560 | /* Age TTL of all fragments by 1 tick .*/ |
| 561 | median_ttl = ip_reass_ttl_decr(1); |
| 562 | |
| 563 | /* Make sure fragment limit is up-to-date. */ |
| 564 | CHECK_NMBCLUSTER_PARAMS(); |
| 565 | |
| 566 | /* If we have too many fragments, drop the older half. */ |
| 567 | if (ip_nfrags > ip_maxfrags) { |
| 568 | ip_reass_ttl_decr(median_ttl); |
| 569 | } |
| 570 | |
| 571 | /* |
| 572 | * If we are over the maximum number of fragmented packets (due to |
| 573 | * the limit being lowered), drain off enough to get down to the |
| 574 | * new limit. Start draining from the reassembly hashqueue most |
| 575 | * recently drained. |
| 576 | */ |
| 577 | if (ip_maxfragpackets < 0) |
| 578 | ; |
| 579 | else { |
| 580 | int wrapped = 0; |
| 581 | |
| 582 | i = dropscanidx; |
| 583 | while (ip_nfragpackets > ip_maxfragpackets && wrapped == 0) { |
| 584 | while (LIST_FIRST(&ip_frags[i]) != NULL) { |
| 585 | ip_freef(LIST_FIRST(&ip_frags[i])); |
| 586 | } |
| 587 | if (++i >= IPREASS_HASH_SIZE) { |
| 588 | i = 0; |
| 589 | } |
| 590 | /* |
| 591 | * Do not scan forever even if fragment counters are |
| 592 | * wrong: stop after scanning entire reassembly queue. |
| 593 | */ |
| 594 | if (i == dropscanidx) { |
| 595 | wrapped = 1; |
| 596 | } |
| 597 | } |
| 598 | dropscanidx = i; |
| 599 | } |
| 600 | mutex_exit(&ipfr_lock); |
| 601 | } |
| 602 | |
| 603 | /* |
| 604 | * ip_reass_packet: generic routine to perform IP reassembly. |
| 605 | * |
| 606 | * => Passed fragment should have IP_MF flag and/or offset set. |
| 607 | * => Fragment should not have other than IP_MF flags set. |
| 608 | * |
| 609 | * => Returns 0 on success or error otherwise. |
| 610 | * => On complete, m0 represents a constructed final packet. |
| 611 | */ |
| 612 | int |
| 613 | ip_reass_packet(struct mbuf **m0, struct ip *ip) |
| 614 | { |
| 615 | const int hlen = ip->ip_hl << 2; |
| 616 | const int len = ntohs(ip->ip_len); |
| 617 | struct mbuf *m = *m0; |
| 618 | ipfr_queue_t *fp; |
| 619 | ipfr_qent_t *ipqe; |
| 620 | u_int hash, off, flen; |
| 621 | bool mff; |
| 622 | |
| 623 | /* |
| 624 | * Prevent TCP blind data attacks by not allowing non-initial |
| 625 | * fragments to start at less than 68 bytes (minimal fragment |
| 626 | * size) and making sure the first fragment is at least 68 |
| 627 | * bytes. |
| 628 | */ |
| 629 | off = (ntohs(ip->ip_off) & IP_OFFMASK) << 3; |
| 630 | if ((off > 0 ? off + hlen : len) < IP_MINFRAGSIZE - 1) { |
| 631 | IP_STATINC(IP_STAT_BADFRAGS); |
| 632 | return EINVAL; |
| 633 | } |
| 634 | |
| 635 | /* |
| 636 | * Fragment length and MF flag. Make sure that fragments have |
| 637 | * a data length which is non-zero and multiple of 8 bytes. |
| 638 | */ |
| 639 | flen = ntohs(ip->ip_len) - hlen; |
| 640 | mff = (ip->ip_off & htons(IP_MF)) != 0; |
| 641 | if (mff && (flen == 0 || (flen & 0x7) != 0)) { |
| 642 | IP_STATINC(IP_STAT_BADFRAGS); |
| 643 | return EINVAL; |
| 644 | } |
| 645 | |
| 646 | /* |
| 647 | * Adjust total IP length to not reflect header and convert |
| 648 | * offset of this to bytes. XXX: clobbers struct ip. |
| 649 | */ |
| 650 | ip->ip_len = htons(flen); |
| 651 | ip->ip_off = htons(off); |
| 652 | |
| 653 | /* Look for queue of fragments of this datagram. */ |
| 654 | mutex_enter(&ipfr_lock); |
| 655 | hash = IPREASS_HASH(ip->ip_src.s_addr, ip->ip_id); |
| 656 | LIST_FOREACH(fp, &ip_frags[hash], ipq_q) { |
| 657 | if (ip->ip_id != fp->ipq_id) |
| 658 | continue; |
| 659 | if (!in_hosteq(ip->ip_src, fp->ipq_src)) |
| 660 | continue; |
| 661 | if (!in_hosteq(ip->ip_dst, fp->ipq_dst)) |
| 662 | continue; |
| 663 | if (ip->ip_p != fp->ipq_p) |
| 664 | continue; |
| 665 | break; |
| 666 | } |
| 667 | |
| 668 | /* Make sure that TOS matches previous fragments. */ |
| 669 | if (fp && fp->ipq_tos != ip->ip_tos) { |
| 670 | IP_STATINC(IP_STAT_BADFRAGS); |
| 671 | mutex_exit(&ipfr_lock); |
| 672 | return EINVAL; |
| 673 | } |
| 674 | |
| 675 | /* |
| 676 | * Create new entry and attempt to reassembly. |
| 677 | */ |
| 678 | IP_STATINC(IP_STAT_FRAGMENTS); |
| 679 | ipqe = pool_cache_get(ipfren_cache, PR_NOWAIT); |
| 680 | if (ipqe == NULL) { |
| 681 | IP_STATINC(IP_STAT_RCVMEMDROP); |
| 682 | mutex_exit(&ipfr_lock); |
| 683 | return ENOMEM; |
| 684 | } |
| 685 | ipqe->ipqe_mff = mff; |
| 686 | ipqe->ipqe_m = m; |
| 687 | ipqe->ipqe_ip = ip; |
| 688 | |
| 689 | *m0 = ip_reass(ipqe, fp, hash); |
| 690 | if (*m0) { |
| 691 | /* Note that finally reassembled. */ |
| 692 | IP_STATINC(IP_STAT_REASSEMBLED); |
| 693 | } |
| 694 | return 0; |
| 695 | } |
| 696 | |