| 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/errno.h> |
| 14 | #include <linux/types.h> |
| 15 | #include <linux/socket.h> |
| 16 | #include <linux/net.h> |
| 17 | #include <linux/netdevice.h> |
| 18 | #include <linux/if_arp.h> |
| 19 | #include <linux/in6.h> |
| 20 | #include <linux/route.h> |
| 21 | #include <linux/proc_fs.h> |
| 22 | #include <linux/seq_file.h> |
| 23 | #include <linux/slab.h> |
| 24 | #include <linux/export.h> |
| 25 | #include <linux/pid_namespace.h> |
| 26 | |
| 27 | #include <net/net_namespace.h> |
| 28 | #include <net/sock.h> |
| 29 | |
| 30 | #include <net/ipv6.h> |
| 31 | #include <net/ndisc.h> |
| 32 | #include <net/protocol.h> |
| 33 | #include <net/ip6_route.h> |
| 34 | #include <net/addrconf.h> |
| 35 | #include <net/rawv6.h> |
| 36 | #include <net/icmp.h> |
| 37 | #include <net/transp_v6.h> |
| 38 | |
| 39 | #include <asm/uaccess.h> |
| 40 | |
| 41 | #define FL_MIN_LINGER 6 /* Minimal linger. It is set to 6sec specified |
| 42 | in old IPv6 RFC. Well, it was reasonable value. |
| 43 | */ |
| 44 | #define FL_MAX_LINGER 60 /* Maximal linger timeout */ |
| 45 | |
| 46 | /* FL hash table */ |
| 47 | |
| 48 | #define FL_MAX_PER_SOCK 32 |
| 49 | #define FL_MAX_SIZE 4096 |
| 50 | #define FL_HASH_MASK 255 |
| 51 | #define FL_HASH(l) (ntohl(l)&FL_HASH_MASK) |
| 52 | |
| 53 | static atomic_t fl_size = ATOMIC_INIT(0); |
| 54 | static struct ip6_flowlabel __rcu *fl_ht[FL_HASH_MASK+1]; |
| 55 | |
| 56 | static void ip6_fl_gc(unsigned long dummy); |
| 57 | static DEFINE_TIMER(ip6_fl_gc_timer, ip6_fl_gc, 0, 0); |
| 58 | |
| 59 | /* FL hash table lock: it protects only of GC */ |
| 60 | |
| 61 | static DEFINE_SPINLOCK(ip6_fl_lock); |
| 62 | |
| 63 | /* Big socket sock */ |
| 64 | |
| 65 | static DEFINE_SPINLOCK(ip6_sk_fl_lock); |
| 66 | |
| 67 | #define for_each_fl_rcu(hash, fl) \ |
| 68 | for (fl = rcu_dereference_bh(fl_ht[(hash)]); \ |
| 69 | fl != NULL; \ |
| 70 | fl = rcu_dereference_bh(fl->next)) |
| 71 | #define for_each_fl_continue_rcu(fl) \ |
| 72 | for (fl = rcu_dereference_bh(fl->next); \ |
| 73 | fl != NULL; \ |
| 74 | fl = rcu_dereference_bh(fl->next)) |
| 75 | |
| 76 | #define for_each_sk_fl_rcu(np, sfl) \ |
| 77 | for (sfl = rcu_dereference_bh(np->ipv6_fl_list); \ |
| 78 | sfl != NULL; \ |
| 79 | sfl = rcu_dereference_bh(sfl->next)) |
| 80 | |
| 81 | static inline struct ip6_flowlabel *__fl_lookup(struct net *net, __be32 label) |
| 82 | { |
| 83 | struct ip6_flowlabel *fl; |
| 84 | |
| 85 | for_each_fl_rcu(FL_HASH(label), fl) { |
| 86 | if (fl->label == label && net_eq(fl->fl_net, net)) |
| 87 | return fl; |
| 88 | } |
| 89 | return NULL; |
| 90 | } |
| 91 | |
| 92 | static struct ip6_flowlabel *fl_lookup(struct net *net, __be32 label) |
| 93 | { |
| 94 | struct ip6_flowlabel *fl; |
| 95 | |
| 96 | rcu_read_lock_bh(); |
| 97 | fl = __fl_lookup(net, label); |
| 98 | if (fl && !atomic_inc_not_zero(&fl->users)) |
| 99 | fl = NULL; |
| 100 | rcu_read_unlock_bh(); |
| 101 | return fl; |
| 102 | } |
| 103 | |
| 104 | |
| 105 | static void fl_free(struct ip6_flowlabel *fl) |
| 106 | { |
| 107 | if (fl) { |
| 108 | if (fl->share == IPV6_FL_S_PROCESS) |
| 109 | put_pid(fl->owner.pid); |
| 110 | release_net(fl->fl_net); |
| 111 | kfree(fl->opt); |
| 112 | kfree_rcu(fl, rcu); |
| 113 | } |
| 114 | } |
| 115 | |
| 116 | static void fl_release(struct ip6_flowlabel *fl) |
| 117 | { |
| 118 | spin_lock_bh(&ip6_fl_lock); |
| 119 | |
| 120 | fl->lastuse = jiffies; |
| 121 | if (atomic_dec_and_test(&fl->users)) { |
| 122 | unsigned long ttd = fl->lastuse + fl->linger; |
| 123 | if (time_after(ttd, fl->expires)) |
| 124 | fl->expires = ttd; |
| 125 | ttd = fl->expires; |
| 126 | if (fl->opt && fl->share == IPV6_FL_S_EXCL) { |
| 127 | struct ipv6_txoptions *opt = fl->opt; |
| 128 | fl->opt = NULL; |
| 129 | kfree(opt); |
| 130 | } |
| 131 | if (!timer_pending(&ip6_fl_gc_timer) || |
| 132 | time_after(ip6_fl_gc_timer.expires, ttd)) |
| 133 | mod_timer(&ip6_fl_gc_timer, ttd); |
| 134 | } |
| 135 | spin_unlock_bh(&ip6_fl_lock); |
| 136 | } |
| 137 | |
| 138 | static void ip6_fl_gc(unsigned long dummy) |
| 139 | { |
| 140 | int i; |
| 141 | unsigned long now = jiffies; |
| 142 | unsigned long sched = 0; |
| 143 | |
| 144 | spin_lock(&ip6_fl_lock); |
| 145 | |
| 146 | for (i=0; i<=FL_HASH_MASK; i++) { |
| 147 | struct ip6_flowlabel *fl; |
| 148 | struct ip6_flowlabel __rcu **flp; |
| 149 | |
| 150 | flp = &fl_ht[i]; |
| 151 | while ((fl = rcu_dereference_protected(*flp, |
| 152 | lockdep_is_held(&ip6_fl_lock))) != NULL) { |
| 153 | if (atomic_read(&fl->users) == 0) { |
| 154 | unsigned long ttd = fl->lastuse + fl->linger; |
| 155 | if (time_after(ttd, fl->expires)) |
| 156 | fl->expires = ttd; |
| 157 | ttd = fl->expires; |
| 158 | if (time_after_eq(now, ttd)) { |
| 159 | *flp = fl->next; |
| 160 | fl_free(fl); |
| 161 | atomic_dec(&fl_size); |
| 162 | continue; |
| 163 | } |
| 164 | if (!sched || time_before(ttd, sched)) |
| 165 | sched = ttd; |
| 166 | } |
| 167 | flp = &fl->next; |
| 168 | } |
| 169 | } |
| 170 | if (!sched && atomic_read(&fl_size)) |
| 171 | sched = now + FL_MAX_LINGER; |
| 172 | if (sched) { |
| 173 | mod_timer(&ip6_fl_gc_timer, sched); |
| 174 | } |
| 175 | spin_unlock(&ip6_fl_lock); |
| 176 | } |
| 177 | |
| 178 | static void __net_exit ip6_fl_purge(struct net *net) |
| 179 | { |
| 180 | int i; |
| 181 | |
| 182 | spin_lock(&ip6_fl_lock); |
| 183 | for (i = 0; i <= FL_HASH_MASK; i++) { |
| 184 | struct ip6_flowlabel *fl; |
| 185 | struct ip6_flowlabel __rcu **flp; |
| 186 | |
| 187 | flp = &fl_ht[i]; |
| 188 | while ((fl = rcu_dereference_protected(*flp, |
| 189 | lockdep_is_held(&ip6_fl_lock))) != NULL) { |
| 190 | if (net_eq(fl->fl_net, net) && |
| 191 | atomic_read(&fl->users) == 0) { |
| 192 | *flp = fl->next; |
| 193 | fl_free(fl); |
| 194 | atomic_dec(&fl_size); |
| 195 | continue; |
| 196 | } |
| 197 | flp = &fl->next; |
| 198 | } |
| 199 | } |
| 200 | spin_unlock(&ip6_fl_lock); |
| 201 | } |
| 202 | |
| 203 | static struct ip6_flowlabel *fl_intern(struct net *net, |
| 204 | struct ip6_flowlabel *fl, __be32 label) |
| 205 | { |
| 206 | struct ip6_flowlabel *lfl; |
| 207 | |
| 208 | fl->label = label & IPV6_FLOWLABEL_MASK; |
| 209 | |
| 210 | spin_lock_bh(&ip6_fl_lock); |
| 211 | if (label == 0) { |
| 212 | for (;;) { |
| 213 | fl->label = htonl(net_random())&IPV6_FLOWLABEL_MASK; |
| 214 | if (fl->label) { |
| 215 | lfl = __fl_lookup(net, fl->label); |
| 216 | if (lfl == NULL) |
| 217 | break; |
| 218 | } |
| 219 | } |
| 220 | } else { |
| 221 | /* |
| 222 | * we dropper the ip6_fl_lock, so this entry could reappear |
| 223 | * and we need to recheck with it. |
| 224 | * |
| 225 | * OTOH no need to search the active socket first, like it is |
| 226 | * done in ipv6_flowlabel_opt - sock is locked, so new entry |
| 227 | * with the same label can only appear on another sock |
| 228 | */ |
| 229 | lfl = __fl_lookup(net, fl->label); |
| 230 | if (lfl != NULL) { |
| 231 | atomic_inc(&lfl->users); |
| 232 | spin_unlock_bh(&ip6_fl_lock); |
| 233 | return lfl; |
| 234 | } |
| 235 | } |
| 236 | |
| 237 | fl->lastuse = jiffies; |
| 238 | fl->next = fl_ht[FL_HASH(fl->label)]; |
| 239 | rcu_assign_pointer(fl_ht[FL_HASH(fl->label)], fl); |
| 240 | atomic_inc(&fl_size); |
| 241 | spin_unlock_bh(&ip6_fl_lock); |
| 242 | return NULL; |
| 243 | } |
| 244 | |
| 245 | |
| 246 | |
| 247 | /* Socket flowlabel lists */ |
| 248 | |
| 249 | struct ip6_flowlabel * fl6_sock_lookup(struct sock *sk, __be32 label) |
| 250 | { |
| 251 | struct ipv6_fl_socklist *sfl; |
| 252 | struct ipv6_pinfo *np = inet6_sk(sk); |
| 253 | |
| 254 | label &= IPV6_FLOWLABEL_MASK; |
| 255 | |
| 256 | rcu_read_lock_bh(); |
| 257 | for_each_sk_fl_rcu(np, sfl) { |
| 258 | struct ip6_flowlabel *fl = sfl->fl; |
| 259 | if (fl->label == label) { |
| 260 | fl->lastuse = jiffies; |
| 261 | atomic_inc(&fl->users); |
| 262 | rcu_read_unlock_bh(); |
| 263 | return fl; |
| 264 | } |
| 265 | } |
| 266 | rcu_read_unlock_bh(); |
| 267 | return NULL; |
| 268 | } |
| 269 | |
| 270 | EXPORT_SYMBOL_GPL(fl6_sock_lookup); |
| 271 | |
| 272 | void fl6_free_socklist(struct sock *sk) |
| 273 | { |
| 274 | struct ipv6_pinfo *np = inet6_sk(sk); |
| 275 | struct ipv6_fl_socklist *sfl; |
| 276 | |
| 277 | if (!rcu_access_pointer(np->ipv6_fl_list)) |
| 278 | return; |
| 279 | |
| 280 | spin_lock_bh(&ip6_sk_fl_lock); |
| 281 | while ((sfl = rcu_dereference_protected(np->ipv6_fl_list, |
| 282 | lockdep_is_held(&ip6_sk_fl_lock))) != NULL) { |
| 283 | np->ipv6_fl_list = sfl->next; |
| 284 | spin_unlock_bh(&ip6_sk_fl_lock); |
| 285 | |
| 286 | fl_release(sfl->fl); |
| 287 | kfree_rcu(sfl, rcu); |
| 288 | |
| 289 | spin_lock_bh(&ip6_sk_fl_lock); |
| 290 | } |
| 291 | spin_unlock_bh(&ip6_sk_fl_lock); |
| 292 | } |
| 293 | |
| 294 | /* Service routines */ |
| 295 | |
| 296 | |
| 297 | /* |
| 298 | It is the only difficult place. flowlabel enforces equal headers |
| 299 | before and including routing header, however user may supply options |
| 300 | following rthdr. |
| 301 | */ |
| 302 | |
| 303 | struct ipv6_txoptions *fl6_merge_options(struct ipv6_txoptions * opt_space, |
| 304 | struct ip6_flowlabel * fl, |
| 305 | struct ipv6_txoptions * fopt) |
| 306 | { |
| 307 | struct ipv6_txoptions * fl_opt = fl->opt; |
| 308 | |
| 309 | if (fopt == NULL || fopt->opt_flen == 0) |
| 310 | return fl_opt; |
| 311 | |
| 312 | if (fl_opt != NULL) { |
| 313 | opt_space->hopopt = fl_opt->hopopt; |
| 314 | opt_space->dst0opt = fl_opt->dst0opt; |
| 315 | opt_space->srcrt = fl_opt->srcrt; |
| 316 | opt_space->opt_nflen = fl_opt->opt_nflen; |
| 317 | } else { |
| 318 | if (fopt->opt_nflen == 0) |
| 319 | return fopt; |
| 320 | opt_space->hopopt = NULL; |
| 321 | opt_space->dst0opt = NULL; |
| 322 | opt_space->srcrt = NULL; |
| 323 | opt_space->opt_nflen = 0; |
| 324 | } |
| 325 | opt_space->dst1opt = fopt->dst1opt; |
| 326 | opt_space->opt_flen = fopt->opt_flen; |
| 327 | return opt_space; |
| 328 | } |
| 329 | EXPORT_SYMBOL_GPL(fl6_merge_options); |
| 330 | |
| 331 | static unsigned long check_linger(unsigned long ttl) |
| 332 | { |
| 333 | if (ttl < FL_MIN_LINGER) |
| 334 | return FL_MIN_LINGER*HZ; |
| 335 | if (ttl > FL_MAX_LINGER && !capable(CAP_NET_ADMIN)) |
| 336 | return 0; |
| 337 | return ttl*HZ; |
| 338 | } |
| 339 | |
| 340 | static int fl6_renew(struct ip6_flowlabel *fl, unsigned long linger, unsigned long expires) |
| 341 | { |
| 342 | linger = check_linger(linger); |
| 343 | if (!linger) |
| 344 | return -EPERM; |
| 345 | expires = check_linger(expires); |
| 346 | if (!expires) |
| 347 | return -EPERM; |
| 348 | fl->lastuse = jiffies; |
| 349 | if (time_before(fl->linger, linger)) |
| 350 | fl->linger = linger; |
| 351 | if (time_before(expires, fl->linger)) |
| 352 | expires = fl->linger; |
| 353 | if (time_before(fl->expires, fl->lastuse + expires)) |
| 354 | fl->expires = fl->lastuse + expires; |
| 355 | return 0; |
| 356 | } |
| 357 | |
| 358 | static struct ip6_flowlabel * |
| 359 | fl_create(struct net *net, struct sock *sk, struct in6_flowlabel_req *freq, |
| 360 | char __user *optval, int optlen, int *err_p) |
| 361 | { |
| 362 | struct ip6_flowlabel *fl = NULL; |
| 363 | int olen; |
| 364 | int addr_type; |
| 365 | int err; |
| 366 | |
| 367 | olen = optlen - CMSG_ALIGN(sizeof(*freq)); |
| 368 | err = -EINVAL; |
| 369 | if (olen > 64 * 1024) |
| 370 | goto done; |
| 371 | |
| 372 | err = -ENOMEM; |
| 373 | fl = kzalloc(sizeof(*fl), GFP_KERNEL); |
| 374 | if (fl == NULL) |
| 375 | goto done; |
| 376 | |
| 377 | if (olen > 0) { |
| 378 | struct msghdr msg; |
| 379 | struct flowi6 flowi6; |
| 380 | int junk; |
| 381 | |
| 382 | err = -ENOMEM; |
| 383 | fl->opt = kmalloc(sizeof(*fl->opt) + olen, GFP_KERNEL); |
| 384 | if (fl->opt == NULL) |
| 385 | goto done; |
| 386 | |
| 387 | memset(fl->opt, 0, sizeof(*fl->opt)); |
| 388 | fl->opt->tot_len = sizeof(*fl->opt) + olen; |
| 389 | err = -EFAULT; |
| 390 | if (copy_from_user(fl->opt+1, optval+CMSG_ALIGN(sizeof(*freq)), olen)) |
| 391 | goto done; |
| 392 | |
| 393 | msg.msg_controllen = olen; |
| 394 | msg.msg_control = (void*)(fl->opt+1); |
| 395 | memset(&flowi6, 0, sizeof(flowi6)); |
| 396 | |
| 397 | err = ip6_datagram_send_ctl(net, sk, &msg, &flowi6, fl->opt, |
| 398 | &junk, &junk, &junk); |
| 399 | if (err) |
| 400 | goto done; |
| 401 | err = -EINVAL; |
| 402 | if (fl->opt->opt_flen) |
| 403 | goto done; |
| 404 | if (fl->opt->opt_nflen == 0) { |
| 405 | kfree(fl->opt); |
| 406 | fl->opt = NULL; |
| 407 | } |
| 408 | } |
| 409 | |
| 410 | fl->fl_net = hold_net(net); |
| 411 | fl->expires = jiffies; |
| 412 | err = fl6_renew(fl, freq->flr_linger, freq->flr_expires); |
| 413 | if (err) |
| 414 | goto done; |
| 415 | fl->share = freq->flr_share; |
| 416 | addr_type = ipv6_addr_type(&freq->flr_dst); |
| 417 | if ((addr_type & IPV6_ADDR_MAPPED) || |
| 418 | addr_type == IPV6_ADDR_ANY) { |
| 419 | err = -EINVAL; |
| 420 | goto done; |
| 421 | } |
| 422 | fl->dst = freq->flr_dst; |
| 423 | atomic_set(&fl->users, 1); |
| 424 | switch (fl->share) { |
| 425 | case IPV6_FL_S_EXCL: |
| 426 | case IPV6_FL_S_ANY: |
| 427 | break; |
| 428 | case IPV6_FL_S_PROCESS: |
| 429 | fl->owner.pid = get_task_pid(current, PIDTYPE_PID); |
| 430 | break; |
| 431 | case IPV6_FL_S_USER: |
| 432 | fl->owner.uid = current_euid(); |
| 433 | break; |
| 434 | default: |
| 435 | err = -EINVAL; |
| 436 | goto done; |
| 437 | } |
| 438 | return fl; |
| 439 | |
| 440 | done: |
| 441 | fl_free(fl); |
| 442 | *err_p = err; |
| 443 | return NULL; |
| 444 | } |
| 445 | |
| 446 | static int mem_check(struct sock *sk) |
| 447 | { |
| 448 | struct ipv6_pinfo *np = inet6_sk(sk); |
| 449 | struct ipv6_fl_socklist *sfl; |
| 450 | int room = FL_MAX_SIZE - atomic_read(&fl_size); |
| 451 | int count = 0; |
| 452 | |
| 453 | if (room > FL_MAX_SIZE - FL_MAX_PER_SOCK) |
| 454 | return 0; |
| 455 | |
| 456 | rcu_read_lock_bh(); |
| 457 | for_each_sk_fl_rcu(np, sfl) |
| 458 | count++; |
| 459 | rcu_read_unlock_bh(); |
| 460 | |
| 461 | if (room <= 0 || |
| 462 | ((count >= FL_MAX_PER_SOCK || |
| 463 | (count > 0 && room < FL_MAX_SIZE/2) || room < FL_MAX_SIZE/4) && |
| 464 | !capable(CAP_NET_ADMIN))) |
| 465 | return -ENOBUFS; |
| 466 | |
| 467 | return 0; |
| 468 | } |
| 469 | |
| 470 | static bool ipv6_hdr_cmp(struct ipv6_opt_hdr *h1, struct ipv6_opt_hdr *h2) |
| 471 | { |
| 472 | if (h1 == h2) |
| 473 | return false; |
| 474 | if (h1 == NULL || h2 == NULL) |
| 475 | return true; |
| 476 | if (h1->hdrlen != h2->hdrlen) |
| 477 | return true; |
| 478 | return memcmp(h1+1, h2+1, ((h1->hdrlen+1)<<3) - sizeof(*h1)); |
| 479 | } |
| 480 | |
| 481 | static bool ipv6_opt_cmp(struct ipv6_txoptions *o1, struct ipv6_txoptions *o2) |
| 482 | { |
| 483 | if (o1 == o2) |
| 484 | return false; |
| 485 | if (o1 == NULL || o2 == NULL) |
| 486 | return true; |
| 487 | if (o1->opt_nflen != o2->opt_nflen) |
| 488 | return true; |
| 489 | if (ipv6_hdr_cmp(o1->hopopt, o2->hopopt)) |
| 490 | return true; |
| 491 | if (ipv6_hdr_cmp(o1->dst0opt, o2->dst0opt)) |
| 492 | return true; |
| 493 | if (ipv6_hdr_cmp((struct ipv6_opt_hdr *)o1->srcrt, (struct ipv6_opt_hdr *)o2->srcrt)) |
| 494 | return true; |
| 495 | return false; |
| 496 | } |
| 497 | |
| 498 | static inline void fl_link(struct ipv6_pinfo *np, struct ipv6_fl_socklist *sfl, |
| 499 | struct ip6_flowlabel *fl) |
| 500 | { |
| 501 | spin_lock_bh(&ip6_sk_fl_lock); |
| 502 | sfl->fl = fl; |
| 503 | sfl->next = np->ipv6_fl_list; |
| 504 | rcu_assign_pointer(np->ipv6_fl_list, sfl); |
| 505 | spin_unlock_bh(&ip6_sk_fl_lock); |
| 506 | } |
| 507 | |
| 508 | int ipv6_flowlabel_opt(struct sock *sk, char __user *optval, int optlen) |
| 509 | { |
| 510 | int uninitialized_var(err); |
| 511 | struct net *net = sock_net(sk); |
| 512 | struct ipv6_pinfo *np = inet6_sk(sk); |
| 513 | struct in6_flowlabel_req freq; |
| 514 | struct ipv6_fl_socklist *sfl1=NULL; |
| 515 | struct ipv6_fl_socklist *sfl; |
| 516 | struct ipv6_fl_socklist __rcu **sflp; |
| 517 | struct ip6_flowlabel *fl, *fl1 = NULL; |
| 518 | |
| 519 | |
| 520 | if (optlen < sizeof(freq)) |
| 521 | return -EINVAL; |
| 522 | |
| 523 | if (copy_from_user(&freq, optval, sizeof(freq))) |
| 524 | return -EFAULT; |
| 525 | |
| 526 | switch (freq.flr_action) { |
| 527 | case IPV6_FL_A_PUT: |
| 528 | spin_lock_bh(&ip6_sk_fl_lock); |
| 529 | for (sflp = &np->ipv6_fl_list; |
| 530 | (sfl = rcu_dereference(*sflp))!=NULL; |
| 531 | sflp = &sfl->next) { |
| 532 | if (sfl->fl->label == freq.flr_label) { |
| 533 | if (freq.flr_label == (np->flow_label&IPV6_FLOWLABEL_MASK)) |
| 534 | np->flow_label &= ~IPV6_FLOWLABEL_MASK; |
| 535 | *sflp = rcu_dereference(sfl->next); |
| 536 | spin_unlock_bh(&ip6_sk_fl_lock); |
| 537 | fl_release(sfl->fl); |
| 538 | kfree_rcu(sfl, rcu); |
| 539 | return 0; |
| 540 | } |
| 541 | } |
| 542 | spin_unlock_bh(&ip6_sk_fl_lock); |
| 543 | return -ESRCH; |
| 544 | |
| 545 | case IPV6_FL_A_RENEW: |
| 546 | rcu_read_lock_bh(); |
| 547 | for_each_sk_fl_rcu(np, sfl) { |
| 548 | if (sfl->fl->label == freq.flr_label) { |
| 549 | err = fl6_renew(sfl->fl, freq.flr_linger, freq.flr_expires); |
| 550 | rcu_read_unlock_bh(); |
| 551 | return err; |
| 552 | } |
| 553 | } |
| 554 | rcu_read_unlock_bh(); |
| 555 | |
| 556 | if (freq.flr_share == IPV6_FL_S_NONE && |
| 557 | ns_capable(net->user_ns, CAP_NET_ADMIN)) { |
| 558 | fl = fl_lookup(net, freq.flr_label); |
| 559 | if (fl) { |
| 560 | err = fl6_renew(fl, freq.flr_linger, freq.flr_expires); |
| 561 | fl_release(fl); |
| 562 | return err; |
| 563 | } |
| 564 | } |
| 565 | return -ESRCH; |
| 566 | |
| 567 | case IPV6_FL_A_GET: |
| 568 | if (freq.flr_label & ~IPV6_FLOWLABEL_MASK) |
| 569 | return -EINVAL; |
| 570 | |
| 571 | fl = fl_create(net, sk, &freq, optval, optlen, &err); |
| 572 | if (fl == NULL) |
| 573 | return err; |
| 574 | sfl1 = kmalloc(sizeof(*sfl1), GFP_KERNEL); |
| 575 | |
| 576 | if (freq.flr_label) { |
| 577 | err = -EEXIST; |
| 578 | rcu_read_lock_bh(); |
| 579 | for_each_sk_fl_rcu(np, sfl) { |
| 580 | if (sfl->fl->label == freq.flr_label) { |
| 581 | if (freq.flr_flags&IPV6_FL_F_EXCL) { |
| 582 | rcu_read_unlock_bh(); |
| 583 | goto done; |
| 584 | } |
| 585 | fl1 = sfl->fl; |
| 586 | atomic_inc(&fl1->users); |
| 587 | break; |
| 588 | } |
| 589 | } |
| 590 | rcu_read_unlock_bh(); |
| 591 | |
| 592 | if (fl1 == NULL) |
| 593 | fl1 = fl_lookup(net, freq.flr_label); |
| 594 | if (fl1) { |
| 595 | recheck: |
| 596 | err = -EEXIST; |
| 597 | if (freq.flr_flags&IPV6_FL_F_EXCL) |
| 598 | goto release; |
| 599 | err = -EPERM; |
| 600 | if (fl1->share == IPV6_FL_S_EXCL || |
| 601 | fl1->share != fl->share || |
| 602 | ((fl1->share == IPV6_FL_S_PROCESS) && |
| 603 | (fl1->owner.pid == fl->owner.pid)) || |
| 604 | ((fl1->share == IPV6_FL_S_USER) && |
| 605 | uid_eq(fl1->owner.uid, fl->owner.uid))) |
| 606 | goto release; |
| 607 | |
| 608 | err = -EINVAL; |
| 609 | if (!ipv6_addr_equal(&fl1->dst, &fl->dst) || |
| 610 | ipv6_opt_cmp(fl1->opt, fl->opt)) |
| 611 | goto release; |
| 612 | |
| 613 | err = -ENOMEM; |
| 614 | if (sfl1 == NULL) |
| 615 | goto release; |
| 616 | if (fl->linger > fl1->linger) |
| 617 | fl1->linger = fl->linger; |
| 618 | if ((long)(fl->expires - fl1->expires) > 0) |
| 619 | fl1->expires = fl->expires; |
| 620 | fl_link(np, sfl1, fl1); |
| 621 | fl_free(fl); |
| 622 | return 0; |
| 623 | |
| 624 | release: |
| 625 | fl_release(fl1); |
| 626 | goto done; |
| 627 | } |
| 628 | } |
| 629 | err = -ENOENT; |
| 630 | if (!(freq.flr_flags&IPV6_FL_F_CREATE)) |
| 631 | goto done; |
| 632 | |
| 633 | err = -ENOMEM; |
| 634 | if (sfl1 == NULL || (err = mem_check(sk)) != 0) |
| 635 | goto done; |
| 636 | |
| 637 | fl1 = fl_intern(net, fl, freq.flr_label); |
| 638 | if (fl1 != NULL) |
| 639 | goto recheck; |
| 640 | |
| 641 | if (!freq.flr_label) { |
| 642 | if (copy_to_user(&((struct in6_flowlabel_req __user *) optval)->flr_label, |
| 643 | &fl->label, sizeof(fl->label))) { |
| 644 | /* Intentionally ignore fault. */ |
| 645 | } |
| 646 | } |
| 647 | |
| 648 | fl_link(np, sfl1, fl); |
| 649 | return 0; |
| 650 | |
| 651 | default: |
| 652 | return -EINVAL; |
| 653 | } |
| 654 | |
| 655 | done: |
| 656 | fl_free(fl); |
| 657 | kfree(sfl1); |
| 658 | return err; |
| 659 | } |
| 660 | |
| 661 | #ifdef CONFIG_PROC_FS |
| 662 | |
| 663 | struct ip6fl_iter_state { |
| 664 | struct seq_net_private p; |
| 665 | struct pid_namespace *pid_ns; |
| 666 | int bucket; |
| 667 | }; |
| 668 | |
| 669 | #define ip6fl_seq_private(seq) ((struct ip6fl_iter_state *)(seq)->private) |
| 670 | |
| 671 | static struct ip6_flowlabel *ip6fl_get_first(struct seq_file *seq) |
| 672 | { |
| 673 | struct ip6_flowlabel *fl = NULL; |
| 674 | struct ip6fl_iter_state *state = ip6fl_seq_private(seq); |
| 675 | struct net *net = seq_file_net(seq); |
| 676 | |
| 677 | for (state->bucket = 0; state->bucket <= FL_HASH_MASK; ++state->bucket) { |
| 678 | for_each_fl_rcu(state->bucket, fl) { |
| 679 | if (net_eq(fl->fl_net, net)) |
| 680 | goto out; |
| 681 | } |
| 682 | } |
| 683 | fl = NULL; |
| 684 | out: |
| 685 | return fl; |
| 686 | } |
| 687 | |
| 688 | static struct ip6_flowlabel *ip6fl_get_next(struct seq_file *seq, struct ip6_flowlabel *fl) |
| 689 | { |
| 690 | struct ip6fl_iter_state *state = ip6fl_seq_private(seq); |
| 691 | struct net *net = seq_file_net(seq); |
| 692 | |
| 693 | for_each_fl_continue_rcu(fl) { |
| 694 | if (net_eq(fl->fl_net, net)) |
| 695 | goto out; |
| 696 | } |
| 697 | |
| 698 | try_again: |
| 699 | if (++state->bucket <= FL_HASH_MASK) { |
| 700 | for_each_fl_rcu(state->bucket, fl) { |
| 701 | if (net_eq(fl->fl_net, net)) |
| 702 | goto out; |
| 703 | } |
| 704 | goto try_again; |
| 705 | } |
| 706 | fl = NULL; |
| 707 | |
| 708 | out: |
| 709 | return fl; |
| 710 | } |
| 711 | |
| 712 | static struct ip6_flowlabel *ip6fl_get_idx(struct seq_file *seq, loff_t pos) |
| 713 | { |
| 714 | struct ip6_flowlabel *fl = ip6fl_get_first(seq); |
| 715 | if (fl) |
| 716 | while (pos && (fl = ip6fl_get_next(seq, fl)) != NULL) |
| 717 | --pos; |
| 718 | return pos ? NULL : fl; |
| 719 | } |
| 720 | |
| 721 | static void *ip6fl_seq_start(struct seq_file *seq, loff_t *pos) |
| 722 | __acquires(RCU) |
| 723 | { |
| 724 | rcu_read_lock_bh(); |
| 725 | return *pos ? ip6fl_get_idx(seq, *pos - 1) : SEQ_START_TOKEN; |
| 726 | } |
| 727 | |
| 728 | static void *ip6fl_seq_next(struct seq_file *seq, void *v, loff_t *pos) |
| 729 | { |
| 730 | struct ip6_flowlabel *fl; |
| 731 | |
| 732 | if (v == SEQ_START_TOKEN) |
| 733 | fl = ip6fl_get_first(seq); |
| 734 | else |
| 735 | fl = ip6fl_get_next(seq, v); |
| 736 | ++*pos; |
| 737 | return fl; |
| 738 | } |
| 739 | |
| 740 | static void ip6fl_seq_stop(struct seq_file *seq, void *v) |
| 741 | __releases(RCU) |
| 742 | { |
| 743 | rcu_read_unlock_bh(); |
| 744 | } |
| 745 | |
| 746 | static int ip6fl_seq_show(struct seq_file *seq, void *v) |
| 747 | { |
| 748 | struct ip6fl_iter_state *state = ip6fl_seq_private(seq); |
| 749 | if (v == SEQ_START_TOKEN) |
| 750 | seq_printf(seq, "%-5s %-1s %-6s %-6s %-6s %-8s %-32s %s\n", |
| 751 | "Label", "S", "Owner", "Users", "Linger", "Expires", "Dst", "Opt"); |
| 752 | else { |
| 753 | struct ip6_flowlabel *fl = v; |
| 754 | seq_printf(seq, |
| 755 | "%05X %-1d %-6d %-6d %-6ld %-8ld %pi6 %-4d\n", |
| 756 | (unsigned int)ntohl(fl->label), |
| 757 | fl->share, |
| 758 | ((fl->share == IPV6_FL_S_PROCESS) ? |
| 759 | pid_nr_ns(fl->owner.pid, state->pid_ns) : |
| 760 | ((fl->share == IPV6_FL_S_USER) ? |
| 761 | from_kuid_munged(seq_user_ns(seq), fl->owner.uid) : |
| 762 | 0)), |
| 763 | atomic_read(&fl->users), |
| 764 | fl->linger/HZ, |
| 765 | (long)(fl->expires - jiffies)/HZ, |
| 766 | &fl->dst, |
| 767 | fl->opt ? fl->opt->opt_nflen : 0); |
| 768 | } |
| 769 | return 0; |
| 770 | } |
| 771 | |
| 772 | static const struct seq_operations ip6fl_seq_ops = { |
| 773 | .start = ip6fl_seq_start, |
| 774 | .next = ip6fl_seq_next, |
| 775 | .stop = ip6fl_seq_stop, |
| 776 | .show = ip6fl_seq_show, |
| 777 | }; |
| 778 | |
| 779 | static int ip6fl_seq_open(struct inode *inode, struct file *file) |
| 780 | { |
| 781 | struct seq_file *seq; |
| 782 | struct ip6fl_iter_state *state; |
| 783 | int err; |
| 784 | |
| 785 | err = seq_open_net(inode, file, &ip6fl_seq_ops, |
| 786 | sizeof(struct ip6fl_iter_state)); |
| 787 | |
| 788 | if (!err) { |
| 789 | seq = file->private_data; |
| 790 | state = ip6fl_seq_private(seq); |
| 791 | rcu_read_lock(); |
| 792 | state->pid_ns = get_pid_ns(task_active_pid_ns(current)); |
| 793 | rcu_read_unlock(); |
| 794 | } |
| 795 | return err; |
| 796 | } |
| 797 | |
| 798 | static int ip6fl_seq_release(struct inode *inode, struct file *file) |
| 799 | { |
| 800 | struct seq_file *seq = file->private_data; |
| 801 | struct ip6fl_iter_state *state = ip6fl_seq_private(seq); |
| 802 | put_pid_ns(state->pid_ns); |
| 803 | return seq_release_net(inode, file); |
| 804 | } |
| 805 | |
| 806 | static const struct file_operations ip6fl_seq_fops = { |
| 807 | .owner = THIS_MODULE, |
| 808 | .open = ip6fl_seq_open, |
| 809 | .read = seq_read, |
| 810 | .llseek = seq_lseek, |
| 811 | .release = ip6fl_seq_release, |
| 812 | }; |
| 813 | |
| 814 | static int __net_init ip6_flowlabel_proc_init(struct net *net) |
| 815 | { |
| 816 | if (!proc_create("ip6_flowlabel", S_IRUGO, net->proc_net, |
| 817 | &ip6fl_seq_fops)) |
| 818 | return -ENOMEM; |
| 819 | return 0; |
| 820 | } |
| 821 | |
| 822 | static void __net_exit ip6_flowlabel_proc_fini(struct net *net) |
| 823 | { |
| 824 | remove_proc_entry("ip6_flowlabel", net->proc_net); |
| 825 | } |
| 826 | #else |
| 827 | static inline int ip6_flowlabel_proc_init(struct net *net) |
| 828 | { |
| 829 | return 0; |
| 830 | } |
| 831 | static inline void ip6_flowlabel_proc_fini(struct net *net) |
| 832 | { |
| 833 | } |
| 834 | #endif |
| 835 | |
| 836 | static void __net_exit ip6_flowlabel_net_exit(struct net *net) |
| 837 | { |
| 838 | ip6_fl_purge(net); |
| 839 | ip6_flowlabel_proc_fini(net); |
| 840 | } |
| 841 | |
| 842 | static struct pernet_operations ip6_flowlabel_net_ops = { |
| 843 | .init = ip6_flowlabel_proc_init, |
| 844 | .exit = ip6_flowlabel_net_exit, |
| 845 | }; |
| 846 | |
| 847 | int ip6_flowlabel_init(void) |
| 848 | { |
| 849 | return register_pernet_subsys(&ip6_flowlabel_net_ops); |
| 850 | } |
| 851 | |
| 852 | void ip6_flowlabel_cleanup(void) |
| 853 | { |
| 854 | del_timer(&ip6_fl_gc_timer); |
| 855 | unregister_pernet_subsys(&ip6_flowlabel_net_ops); |
| 856 | } |