1 /* 2 * ip6_flowlabel.c IPv6 flowlabel manager. 3 * 4 * This program is free software; you can redistribute it and/or 5 * modify it under the terms of the GNU General Public License 6 * as published by the Free Software Foundation; either version 7 * 2 of the License, or (at your option) any later version. 8 * 9 * Authors: Alexey Kuznetsov, <kuznet@ms2.inr.ac.ru> 10 */ 11 12 #include <linux/capability.h> 13 #include <linux/config.h> 14 #include <linux/errno.h> 15 #include <linux/types.h> 16 #include <linux/socket.h> 17 #include <linux/net.h> 18 #include <linux/netdevice.h> 19 #include <linux/if_arp.h> 20 #include <linux/in6.h> 21 #include <linux/route.h> 22 #include <linux/proc_fs.h> 23 #include <linux/seq_file.h> 24 25 #include <net/sock.h> 26 27 #include <net/ipv6.h> 28 #include <net/ndisc.h> 29 #include <net/protocol.h> 30 #include <net/ip6_route.h> 31 #include <net/addrconf.h> 32 #include <net/rawv6.h> 33 #include <net/icmp.h> 34 #include <net/transp_v6.h> 35 36 #include <asm/uaccess.h> 37 38 #define FL_MIN_LINGER 6 /* Minimal linger. It is set to 6sec specified 39 in old IPv6 RFC. Well, it was reasonable value. 40 */ 41 #define FL_MAX_LINGER 60 /* Maximal linger timeout */ 42 43 /* FL hash table */ 44 45 #define FL_MAX_PER_SOCK 32 46 #define FL_MAX_SIZE 4096 47 #define FL_HASH_MASK 255 48 #define FL_HASH(l) (ntohl(l)&FL_HASH_MASK) 49 50 static atomic_t fl_size = ATOMIC_INIT(0); 51 static struct ip6_flowlabel *fl_ht[FL_HASH_MASK+1]; 52 53 static void ip6_fl_gc(unsigned long dummy); 54 static DEFINE_TIMER(ip6_fl_gc_timer, ip6_fl_gc, 0, 0); 55 56 /* FL hash table lock: it protects only of GC */ 57 58 static DEFINE_RWLOCK(ip6_fl_lock); 59 60 /* Big socket sock */ 61 62 static DEFINE_RWLOCK(ip6_sk_fl_lock); 63 64 65 static __inline__ struct ip6_flowlabel * __fl_lookup(u32 label) 66 { 67 struct ip6_flowlabel *fl; 68 69 for (fl=fl_ht[FL_HASH(label)]; fl; fl = fl->next) { 70 if (fl->label == label) 71 return fl; 72 } 73 return NULL; 74 } 75 76 static struct ip6_flowlabel * fl_lookup(u32 label) 77 { 78 struct ip6_flowlabel *fl; 79 80 read_lock_bh(&ip6_fl_lock); 81 fl = __fl_lookup(label); 82 if (fl) 83 atomic_inc(&fl->users); 84 read_unlock_bh(&ip6_fl_lock); 85 return fl; 86 } 87 88 89 static void fl_free(struct ip6_flowlabel *fl) 90 { 91 if (fl) 92 kfree(fl->opt); 93 kfree(fl); 94 } 95 96 static void fl_release(struct ip6_flowlabel *fl) 97 { 98 write_lock_bh(&ip6_fl_lock); 99 100 fl->lastuse = jiffies; 101 if (atomic_dec_and_test(&fl->users)) { 102 unsigned long ttd = fl->lastuse + fl->linger; 103 if (time_after(ttd, fl->expires)) 104 fl->expires = ttd; 105 ttd = fl->expires; 106 if (fl->opt && fl->share == IPV6_FL_S_EXCL) { 107 struct ipv6_txoptions *opt = fl->opt; 108 fl->opt = NULL; 109 kfree(opt); 110 } 111 if (!timer_pending(&ip6_fl_gc_timer) || 112 time_after(ip6_fl_gc_timer.expires, ttd)) 113 mod_timer(&ip6_fl_gc_timer, ttd); 114 } 115 116 write_unlock_bh(&ip6_fl_lock); 117 } 118 119 static void ip6_fl_gc(unsigned long dummy) 120 { 121 int i; 122 unsigned long now = jiffies; 123 unsigned long sched = 0; 124 125 write_lock(&ip6_fl_lock); 126 127 for (i=0; i<=FL_HASH_MASK; i++) { 128 struct ip6_flowlabel *fl, **flp; 129 flp = &fl_ht[i]; 130 while ((fl=*flp) != NULL) { 131 if (atomic_read(&fl->users) == 0) { 132 unsigned long ttd = fl->lastuse + fl->linger; 133 if (time_after(ttd, fl->expires)) 134 fl->expires = ttd; 135 ttd = fl->expires; 136 if (time_after_eq(now, ttd)) { 137 *flp = fl->next; 138 fl_free(fl); 139 atomic_dec(&fl_size); 140 continue; 141 } 142 if (!sched || time_before(ttd, sched)) 143 sched = ttd; 144 } 145 flp = &fl->next; 146 } 147 } 148 if (!sched && atomic_read(&fl_size)) 149 sched = now + FL_MAX_LINGER; 150 if (sched) { 151 ip6_fl_gc_timer.expires = sched; 152 add_timer(&ip6_fl_gc_timer); 153 } 154 write_unlock(&ip6_fl_lock); 155 } 156 157 static int fl_intern(struct ip6_flowlabel *fl, __u32 label) 158 { 159 fl->label = label & IPV6_FLOWLABEL_MASK; 160 161 write_lock_bh(&ip6_fl_lock); 162 if (label == 0) { 163 for (;;) { 164 fl->label = htonl(net_random())&IPV6_FLOWLABEL_MASK; 165 if (fl->label) { 166 struct ip6_flowlabel *lfl; 167 lfl = __fl_lookup(fl->label); 168 if (lfl == NULL) 169 break; 170 } 171 } 172 } 173 174 fl->lastuse = jiffies; 175 fl->next = fl_ht[FL_HASH(fl->label)]; 176 fl_ht[FL_HASH(fl->label)] = fl; 177 atomic_inc(&fl_size); 178 write_unlock_bh(&ip6_fl_lock); 179 return 0; 180 } 181 182 183 184 /* Socket flowlabel lists */ 185 186 struct ip6_flowlabel * fl6_sock_lookup(struct sock *sk, u32 label) 187 { 188 struct ipv6_fl_socklist *sfl; 189 struct ipv6_pinfo *np = inet6_sk(sk); 190 191 label &= IPV6_FLOWLABEL_MASK; 192 193 for (sfl=np->ipv6_fl_list; sfl; sfl = sfl->next) { 194 struct ip6_flowlabel *fl = sfl->fl; 195 if (fl->label == label) { 196 fl->lastuse = jiffies; 197 atomic_inc(&fl->users); 198 return fl; 199 } 200 } 201 return NULL; 202 } 203 204 EXPORT_SYMBOL_GPL(fl6_sock_lookup); 205 206 void fl6_free_socklist(struct sock *sk) 207 { 208 struct ipv6_pinfo *np = inet6_sk(sk); 209 struct ipv6_fl_socklist *sfl; 210 211 while ((sfl = np->ipv6_fl_list) != NULL) { 212 np->ipv6_fl_list = sfl->next; 213 fl_release(sfl->fl); 214 kfree(sfl); 215 } 216 } 217 218 /* Service routines */ 219 220 221 /* 222 It is the only difficult place. flowlabel enforces equal headers 223 before and including routing header, however user may supply options 224 following rthdr. 225 */ 226 227 struct ipv6_txoptions *fl6_merge_options(struct ipv6_txoptions * opt_space, 228 struct ip6_flowlabel * fl, 229 struct ipv6_txoptions * fopt) 230 { 231 struct ipv6_txoptions * fl_opt = fl->opt; 232 233 if (fopt == NULL || fopt->opt_flen == 0) 234 return fl_opt; 235 236 if (fl_opt != NULL) { 237 opt_space->hopopt = fl_opt->hopopt; 238 opt_space->dst0opt = fl_opt->dst0opt; 239 opt_space->srcrt = fl_opt->srcrt; 240 opt_space->opt_nflen = fl_opt->opt_nflen; 241 } else { 242 if (fopt->opt_nflen == 0) 243 return fopt; 244 opt_space->hopopt = NULL; 245 opt_space->dst0opt = NULL; 246 opt_space->srcrt = NULL; 247 opt_space->opt_nflen = 0; 248 } 249 opt_space->dst1opt = fopt->dst1opt; 250 opt_space->opt_flen = fopt->opt_flen; 251 return opt_space; 252 } 253 254 static unsigned long check_linger(unsigned long ttl) 255 { 256 if (ttl < FL_MIN_LINGER) 257 return FL_MIN_LINGER*HZ; 258 if (ttl > FL_MAX_LINGER && !capable(CAP_NET_ADMIN)) 259 return 0; 260 return ttl*HZ; 261 } 262 263 static int fl6_renew(struct ip6_flowlabel *fl, unsigned long linger, unsigned long expires) 264 { 265 linger = check_linger(linger); 266 if (!linger) 267 return -EPERM; 268 expires = check_linger(expires); 269 if (!expires) 270 return -EPERM; 271 fl->lastuse = jiffies; 272 if (time_before(fl->linger, linger)) 273 fl->linger = linger; 274 if (time_before(expires, fl->linger)) 275 expires = fl->linger; 276 if (time_before(fl->expires, fl->lastuse + expires)) 277 fl->expires = fl->lastuse + expires; 278 return 0; 279 } 280 281 static struct ip6_flowlabel * 282 fl_create(struct in6_flowlabel_req *freq, char __user *optval, int optlen, int *err_p) 283 { 284 struct ip6_flowlabel *fl; 285 int olen; 286 int addr_type; 287 int err; 288 289 err = -ENOMEM; 290 fl = kzalloc(sizeof(*fl), GFP_KERNEL); 291 if (fl == NULL) 292 goto done; 293 294 olen = optlen - CMSG_ALIGN(sizeof(*freq)); 295 if (olen > 0) { 296 struct msghdr msg; 297 struct flowi flowi; 298 int junk; 299 300 err = -ENOMEM; 301 fl->opt = kmalloc(sizeof(*fl->opt) + olen, GFP_KERNEL); 302 if (fl->opt == NULL) 303 goto done; 304 305 memset(fl->opt, 0, sizeof(*fl->opt)); 306 fl->opt->tot_len = sizeof(*fl->opt) + olen; 307 err = -EFAULT; 308 if (copy_from_user(fl->opt+1, optval+CMSG_ALIGN(sizeof(*freq)), olen)) 309 goto done; 310 311 msg.msg_controllen = olen; 312 msg.msg_control = (void*)(fl->opt+1); 313 flowi.oif = 0; 314 315 err = datagram_send_ctl(&msg, &flowi, fl->opt, &junk, &junk); 316 if (err) 317 goto done; 318 err = -EINVAL; 319 if (fl->opt->opt_flen) 320 goto done; 321 if (fl->opt->opt_nflen == 0) { 322 kfree(fl->opt); 323 fl->opt = NULL; 324 } 325 } 326 327 fl->expires = jiffies; 328 err = fl6_renew(fl, freq->flr_linger, freq->flr_expires); 329 if (err) 330 goto done; 331 fl->share = freq->flr_share; 332 addr_type = ipv6_addr_type(&freq->flr_dst); 333 if ((addr_type&IPV6_ADDR_MAPPED) 334 || addr_type == IPV6_ADDR_ANY) 335 goto done; 336 ipv6_addr_copy(&fl->dst, &freq->flr_dst); 337 atomic_set(&fl->users, 1); 338 switch (fl->share) { 339 case IPV6_FL_S_EXCL: 340 case IPV6_FL_S_ANY: 341 break; 342 case IPV6_FL_S_PROCESS: 343 fl->owner = current->pid; 344 break; 345 case IPV6_FL_S_USER: 346 fl->owner = current->euid; 347 break; 348 default: 349 err = -EINVAL; 350 goto done; 351 } 352 return fl; 353 354 done: 355 fl_free(fl); 356 *err_p = err; 357 return NULL; 358 } 359 360 static int mem_check(struct sock *sk) 361 { 362 struct ipv6_pinfo *np = inet6_sk(sk); 363 struct ipv6_fl_socklist *sfl; 364 int room = FL_MAX_SIZE - atomic_read(&fl_size); 365 int count = 0; 366 367 if (room > FL_MAX_SIZE - FL_MAX_PER_SOCK) 368 return 0; 369 370 for (sfl = np->ipv6_fl_list; sfl; sfl = sfl->next) 371 count++; 372 373 if (room <= 0 || 374 ((count >= FL_MAX_PER_SOCK || 375 (count > 0 && room < FL_MAX_SIZE/2) || room < FL_MAX_SIZE/4) 376 && !capable(CAP_NET_ADMIN))) 377 return -ENOBUFS; 378 379 return 0; 380 } 381 382 static int ipv6_hdr_cmp(struct ipv6_opt_hdr *h1, struct ipv6_opt_hdr *h2) 383 { 384 if (h1 == h2) 385 return 0; 386 if (h1 == NULL || h2 == NULL) 387 return 1; 388 if (h1->hdrlen != h2->hdrlen) 389 return 1; 390 return memcmp(h1+1, h2+1, ((h1->hdrlen+1)<<3) - sizeof(*h1)); 391 } 392 393 static int ipv6_opt_cmp(struct ipv6_txoptions *o1, struct ipv6_txoptions *o2) 394 { 395 if (o1 == o2) 396 return 0; 397 if (o1 == NULL || o2 == NULL) 398 return 1; 399 if (o1->opt_nflen != o2->opt_nflen) 400 return 1; 401 if (ipv6_hdr_cmp(o1->hopopt, o2->hopopt)) 402 return 1; 403 if (ipv6_hdr_cmp(o1->dst0opt, o2->dst0opt)) 404 return 1; 405 if (ipv6_hdr_cmp((struct ipv6_opt_hdr *)o1->srcrt, (struct ipv6_opt_hdr *)o2->srcrt)) 406 return 1; 407 return 0; 408 } 409 410 int ipv6_flowlabel_opt(struct sock *sk, char __user *optval, int optlen) 411 { 412 int err; 413 struct ipv6_pinfo *np = inet6_sk(sk); 414 struct in6_flowlabel_req freq; 415 struct ipv6_fl_socklist *sfl1=NULL; 416 struct ipv6_fl_socklist *sfl, **sflp; 417 struct ip6_flowlabel *fl; 418 419 if (optlen < sizeof(freq)) 420 return -EINVAL; 421 422 if (copy_from_user(&freq, optval, sizeof(freq))) 423 return -EFAULT; 424 425 switch (freq.flr_action) { 426 case IPV6_FL_A_PUT: 427 write_lock_bh(&ip6_sk_fl_lock); 428 for (sflp = &np->ipv6_fl_list; (sfl=*sflp)!=NULL; sflp = &sfl->next) { 429 if (sfl->fl->label == freq.flr_label) { 430 if (freq.flr_label == (np->flow_label&IPV6_FLOWLABEL_MASK)) 431 np->flow_label &= ~IPV6_FLOWLABEL_MASK; 432 *sflp = sfl->next; 433 write_unlock_bh(&ip6_sk_fl_lock); 434 fl_release(sfl->fl); 435 kfree(sfl); 436 return 0; 437 } 438 } 439 write_unlock_bh(&ip6_sk_fl_lock); 440 return -ESRCH; 441 442 case IPV6_FL_A_RENEW: 443 read_lock_bh(&ip6_sk_fl_lock); 444 for (sfl = np->ipv6_fl_list; sfl; sfl = sfl->next) { 445 if (sfl->fl->label == freq.flr_label) { 446 err = fl6_renew(sfl->fl, freq.flr_linger, freq.flr_expires); 447 read_unlock_bh(&ip6_sk_fl_lock); 448 return err; 449 } 450 } 451 read_unlock_bh(&ip6_sk_fl_lock); 452 453 if (freq.flr_share == IPV6_FL_S_NONE && capable(CAP_NET_ADMIN)) { 454 fl = fl_lookup(freq.flr_label); 455 if (fl) { 456 err = fl6_renew(fl, freq.flr_linger, freq.flr_expires); 457 fl_release(fl); 458 return err; 459 } 460 } 461 return -ESRCH; 462 463 case IPV6_FL_A_GET: 464 if (freq.flr_label & ~IPV6_FLOWLABEL_MASK) 465 return -EINVAL; 466 467 fl = fl_create(&freq, optval, optlen, &err); 468 if (fl == NULL) 469 return err; 470 sfl1 = kmalloc(sizeof(*sfl1), GFP_KERNEL); 471 472 if (freq.flr_label) { 473 struct ip6_flowlabel *fl1 = NULL; 474 475 err = -EEXIST; 476 read_lock_bh(&ip6_sk_fl_lock); 477 for (sfl = np->ipv6_fl_list; sfl; sfl = sfl->next) { 478 if (sfl->fl->label == freq.flr_label) { 479 if (freq.flr_flags&IPV6_FL_F_EXCL) { 480 read_unlock_bh(&ip6_sk_fl_lock); 481 goto done; 482 } 483 fl1 = sfl->fl; 484 atomic_inc(&fl1->users); 485 break; 486 } 487 } 488 read_unlock_bh(&ip6_sk_fl_lock); 489 490 if (fl1 == NULL) 491 fl1 = fl_lookup(freq.flr_label); 492 if (fl1) { 493 err = -EEXIST; 494 if (freq.flr_flags&IPV6_FL_F_EXCL) 495 goto release; 496 err = -EPERM; 497 if (fl1->share == IPV6_FL_S_EXCL || 498 fl1->share != fl->share || 499 fl1->owner != fl->owner) 500 goto release; 501 502 err = -EINVAL; 503 if (!ipv6_addr_equal(&fl1->dst, &fl->dst) || 504 ipv6_opt_cmp(fl1->opt, fl->opt)) 505 goto release; 506 507 err = -ENOMEM; 508 if (sfl1 == NULL) 509 goto release; 510 if (fl->linger > fl1->linger) 511 fl1->linger = fl->linger; 512 if ((long)(fl->expires - fl1->expires) > 0) 513 fl1->expires = fl->expires; 514 write_lock_bh(&ip6_sk_fl_lock); 515 sfl1->fl = fl1; 516 sfl1->next = np->ipv6_fl_list; 517 np->ipv6_fl_list = sfl1; 518 write_unlock_bh(&ip6_sk_fl_lock); 519 fl_free(fl); 520 return 0; 521 522 release: 523 fl_release(fl1); 524 goto done; 525 } 526 } 527 err = -ENOENT; 528 if (!(freq.flr_flags&IPV6_FL_F_CREATE)) 529 goto done; 530 531 err = -ENOMEM; 532 if (sfl1 == NULL || (err = mem_check(sk)) != 0) 533 goto done; 534 535 err = fl_intern(fl, freq.flr_label); 536 if (err) 537 goto done; 538 539 if (!freq.flr_label) { 540 if (copy_to_user(&((struct in6_flowlabel_req __user *) optval)->flr_label, 541 &fl->label, sizeof(fl->label))) { 542 /* Intentionally ignore fault. */ 543 } 544 } 545 546 sfl1->fl = fl; 547 sfl1->next = np->ipv6_fl_list; 548 np->ipv6_fl_list = sfl1; 549 return 0; 550 551 default: 552 return -EINVAL; 553 } 554 555 done: 556 fl_free(fl); 557 kfree(sfl1); 558 return err; 559 } 560 561 #ifdef CONFIG_PROC_FS 562 563 struct ip6fl_iter_state { 564 int bucket; 565 }; 566 567 #define ip6fl_seq_private(seq) ((struct ip6fl_iter_state *)(seq)->private) 568 569 static struct ip6_flowlabel *ip6fl_get_first(struct seq_file *seq) 570 { 571 struct ip6_flowlabel *fl = NULL; 572 struct ip6fl_iter_state *state = ip6fl_seq_private(seq); 573 574 for (state->bucket = 0; state->bucket <= FL_HASH_MASK; ++state->bucket) { 575 if (fl_ht[state->bucket]) { 576 fl = fl_ht[state->bucket]; 577 break; 578 } 579 } 580 return fl; 581 } 582 583 static struct ip6_flowlabel *ip6fl_get_next(struct seq_file *seq, struct ip6_flowlabel *fl) 584 { 585 struct ip6fl_iter_state *state = ip6fl_seq_private(seq); 586 587 fl = fl->next; 588 while (!fl) { 589 if (++state->bucket <= FL_HASH_MASK) 590 fl = fl_ht[state->bucket]; 591 } 592 return fl; 593 } 594 595 static struct ip6_flowlabel *ip6fl_get_idx(struct seq_file *seq, loff_t pos) 596 { 597 struct ip6_flowlabel *fl = ip6fl_get_first(seq); 598 if (fl) 599 while (pos && (fl = ip6fl_get_next(seq, fl)) != NULL) 600 --pos; 601 return pos ? NULL : fl; 602 } 603 604 static void *ip6fl_seq_start(struct seq_file *seq, loff_t *pos) 605 { 606 read_lock_bh(&ip6_fl_lock); 607 return *pos ? ip6fl_get_idx(seq, *pos - 1) : SEQ_START_TOKEN; 608 } 609 610 static void *ip6fl_seq_next(struct seq_file *seq, void *v, loff_t *pos) 611 { 612 struct ip6_flowlabel *fl; 613 614 if (v == SEQ_START_TOKEN) 615 fl = ip6fl_get_first(seq); 616 else 617 fl = ip6fl_get_next(seq, v); 618 ++*pos; 619 return fl; 620 } 621 622 static void ip6fl_seq_stop(struct seq_file *seq, void *v) 623 { 624 read_unlock_bh(&ip6_fl_lock); 625 } 626 627 static void ip6fl_fl_seq_show(struct seq_file *seq, struct ip6_flowlabel *fl) 628 { 629 while(fl) { 630 seq_printf(seq, 631 "%05X %-1d %-6d %-6d %-6ld %-8ld " NIP6_SEQFMT " %-4d\n", 632 (unsigned)ntohl(fl->label), 633 fl->share, 634 (unsigned)fl->owner, 635 atomic_read(&fl->users), 636 fl->linger/HZ, 637 (long)(fl->expires - jiffies)/HZ, 638 NIP6(fl->dst), 639 fl->opt ? fl->opt->opt_nflen : 0); 640 fl = fl->next; 641 } 642 } 643 644 static int ip6fl_seq_show(struct seq_file *seq, void *v) 645 { 646 if (v == SEQ_START_TOKEN) 647 seq_printf(seq, "%-5s %-1s %-6s %-6s %-6s %-8s %-32s %s\n", 648 "Label", "S", "Owner", "Users", "Linger", "Expires", "Dst", "Opt"); 649 else 650 ip6fl_fl_seq_show(seq, v); 651 return 0; 652 } 653 654 static struct seq_operations ip6fl_seq_ops = { 655 .start = ip6fl_seq_start, 656 .next = ip6fl_seq_next, 657 .stop = ip6fl_seq_stop, 658 .show = ip6fl_seq_show, 659 }; 660 661 static int ip6fl_seq_open(struct inode *inode, struct file *file) 662 { 663 struct seq_file *seq; 664 int rc = -ENOMEM; 665 struct ip6fl_iter_state *s = kzalloc(sizeof(*s), GFP_KERNEL); 666 667 if (!s) 668 goto out; 669 670 rc = seq_open(file, &ip6fl_seq_ops); 671 if (rc) 672 goto out_kfree; 673 674 seq = file->private_data; 675 seq->private = s; 676 out: 677 return rc; 678 out_kfree: 679 kfree(s); 680 goto out; 681 } 682 683 static struct file_operations ip6fl_seq_fops = { 684 .owner = THIS_MODULE, 685 .open = ip6fl_seq_open, 686 .read = seq_read, 687 .llseek = seq_lseek, 688 .release = seq_release_private, 689 }; 690 #endif 691 692 693 void ip6_flowlabel_init(void) 694 { 695 #ifdef CONFIG_PROC_FS 696 proc_net_fops_create("ip6_flowlabel", S_IRUGO, &ip6fl_seq_fops); 697 #endif 698 } 699 700 void ip6_flowlabel_cleanup(void) 701 { 702 del_timer(&ip6_fl_gc_timer); 703 #ifdef CONFIG_PROC_FS 704 proc_net_remove("ip6_flowlabel"); 705 #endif 706 } 707