net/ipv6/ip6_flowlabel.c

   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/in6.h>
  19 #include <linux/proc_fs.h>
  20 #include <linux/seq_file.h>
  21 #include <linux/slab.h>
  22 #include <linux/export.h>
  23 #include <linux/pid_namespace.h>
  24
  25 #include <net/net_namespace.h>
  26 #include <net/sock.h>
  27
  28 #include <net/ipv6.h>
  29 #include <net/rawv6.h>
  30 #include <net/transp_v6.h>
  31
  32 #include <asm/uaccess.h>
  33
  34 #define FL_MIN_LINGER   6       /* Minimal linger. It is set to 6sec specified
  35                                    in old IPv6 RFC. Well, it was reasonable value.
  36                                  */
  37 #define FL_MAX_LINGER   150     /* Maximal linger timeout */
  38
  39 /* FL hash table */
  40
  41 #define FL_MAX_PER_SOCK 32
  42 #define FL_MAX_SIZE     4096
  43 #define FL_HASH_MASK    255
  44 #define FL_HASH(l)      (ntohl(l)&FL_HASH_MASK)
  45
  46 static atomic_t fl_size = ATOMIC_INIT(0);
  47 static struct ip6_flowlabel __rcu *fl_ht[FL_HASH_MASK+1];
  48
  49 static void ip6_fl_gc(unsigned long dummy);
  50 static DEFINE_TIMER(ip6_fl_gc_timer, ip6_fl_gc, 0, 0);
  51
  52 /* FL hash table lock: it protects only of GC */
  53
  54 static DEFINE_SPINLOCK(ip6_fl_lock);
  55
  56 /* Big socket sock */
  57
  58 static DEFINE_SPINLOCK(ip6_sk_fl_lock);
  59
  60 #define for_each_fl_rcu(hash, fl)                               \
  61         for (fl = rcu_dereference_bh(fl_ht[(hash)]);            \
  62              fl != NULL;                                        \
  63              fl = rcu_dereference_bh(fl->next))
  64 #define for_each_fl_continue_rcu(fl)                            \
  65         for (fl = rcu_dereference_bh(fl->next);                 \
  66              fl != NULL;                                        \
  67              fl = rcu_dereference_bh(fl->next))
  68
  69 #define for_each_sk_fl_rcu(np, sfl)                             \
  70         for (sfl = rcu_dereference_bh(np->ipv6_fl_list);        \
  71              sfl != NULL;                                       \
  72              sfl = rcu_dereference_bh(sfl->next))
  73
  74 static inline struct ip6_flowlabel *__fl_lookup(struct net *net, __be32 label)
  75 {
  76         struct ip6_flowlabel *fl;
  77
  78         for_each_fl_rcu(FL_HASH(label), fl) {
  79                 if (fl->label == label && net_eq(fl->fl_net, net))
  80                         return fl;
  81         }
  82         return NULL;
  83 }
  84
  85 static struct ip6_flowlabel *fl_lookup(struct net *net, __be32 label)
  86 {
  87         struct ip6_flowlabel *fl;
  88
  89         rcu_read_lock_bh();
  90         fl = __fl_lookup(net, label);
  91         if (fl && !atomic_inc_not_zero(&fl->users))
  92                 fl = NULL;
  93         rcu_read_unlock_bh();
  94         return fl;
  95 }
  96
  97
  98 static void fl_free(struct ip6_flowlabel *fl)
  99 {
 100         if (fl) {
 101                 if (fl->share == IPV6_FL_S_PROCESS)
 102                         put_pid(fl->owner.pid);
 103                 release_net(fl->fl_net);
 104                 kfree(fl->opt);
 105                 kfree_rcu(fl, rcu);
 106         }
 107 }
 108
 109 static void fl_release(struct ip6_flowlabel *fl)
 110 {
 111         spin_lock_bh(&ip6_fl_lock);
 112
 113         fl->lastuse = jiffies;
 114         if (atomic_dec_and_test(&fl->users)) {
 115                 unsigned long ttd = fl->lastuse + fl->linger;
 116                 if (time_after(ttd, fl->expires))
 117                         fl->expires = ttd;
 118                 ttd = fl->expires;
 119                 if (fl->opt && fl->share == IPV6_FL_S_EXCL) {
 120                         struct ipv6_txoptions *opt = fl->opt;
 121                         fl->opt = NULL;
 122                         kfree(opt);
 123                 }
 124                 if (!timer_pending(&ip6_fl_gc_timer) ||
 125                     time_after(ip6_fl_gc_timer.expires, ttd))
 126                         mod_timer(&ip6_fl_gc_timer, ttd);
 127         }
 128         spin_unlock_bh(&ip6_fl_lock);
 129 }
 130
 131 static void ip6_fl_gc(unsigned long dummy)
 132 {
 133         int i;
 134         unsigned long now = jiffies;
 135         unsigned long sched = 0;
 136
 137         spin_lock(&ip6_fl_lock);
 138
 139         for (i=0; i<=FL_HASH_MASK; i++) {
 140                 struct ip6_flowlabel *fl;
 141                 struct ip6_flowlabel __rcu **flp;
 142
 143                 flp = &fl_ht[i];
 144                 while ((fl = rcu_dereference_protected(*flp,
 145                                                        lockdep_is_held(&ip6_fl_lock))) != NULL) {
 146                         if (atomic_read(&fl->users) == 0) {
 147                                 unsigned long ttd = fl->lastuse + fl->linger;
 148                                 if (time_after(ttd, fl->expires))
 149                                         fl->expires = ttd;
 150                                 ttd = fl->expires;
 151                                 if (time_after_eq(now, ttd)) {
 152                                         *flp = fl->next;
 153                                         fl_free(fl);
 154                                         atomic_dec(&fl_size);
 155                                         continue;
 156                                 }
 157                                 if (!sched || time_before(ttd, sched))
 158                                         sched = ttd;
 159                         }
 160                         flp = &fl->next;
 161                 }
 162         }
 163         if (!sched && atomic_read(&fl_size))
 164                 sched = now + FL_MAX_LINGER;
 165         if (sched) {
 166                 mod_timer(&ip6_fl_gc_timer, sched);
 167         }
 168         spin_unlock(&ip6_fl_lock);
 169 }
 170
 171 static void __net_exit ip6_fl_purge(struct net *net)
 172 {
 173         int i;
 174
 175         spin_lock(&ip6_fl_lock);
 176         for (i = 0; i <= FL_HASH_MASK; i++) {
 177                 struct ip6_flowlabel *fl;
 178                 struct ip6_flowlabel __rcu **flp;
 179
 180                 flp = &fl_ht[i];
 181                 while ((fl = rcu_dereference_protected(*flp,
 182                                                        lockdep_is_held(&ip6_fl_lock))) != NULL) {
 183                         if (net_eq(fl->fl_net, net) &&
 184                             atomic_read(&fl->users) == 0) {
 185                                 *flp = fl->next;
 186                                 fl_free(fl);
 187                                 atomic_dec(&fl_size);
 188                                 continue;
 189                         }
 190                         flp = &fl->next;
 191                 }
 192         }
 193         spin_unlock(&ip6_fl_lock);
 194 }
 195
 196 static struct ip6_flowlabel *fl_intern(struct net *net,
 197                                        struct ip6_flowlabel *fl, __be32 label)
 198 {
 199         struct ip6_flowlabel *lfl;
 200
 201         fl->label = label & IPV6_FLOWLABEL_MASK;
 202
 203         spin_lock_bh(&ip6_fl_lock);
 204         if (label == 0) {
 205                 for (;;) {
 206                         fl->label = htonl(prandom_u32())&IPV6_FLOWLABEL_MASK;
 207                         if (fl->label) {
 208                                 lfl = __fl_lookup(net, fl->label);
 209                                 if (lfl == NULL)
 210                                         break;
 211                         }
 212                 }
 213         } else {
 214                 /*
 215                  * we dropper the ip6_fl_lock, so this entry could reappear
 216                  * and we need to recheck with it.
 217                  *
 218                  * OTOH no need to search the active socket first, like it is
 219                  * done in ipv6_flowlabel_opt - sock is locked, so new entry
 220                  * with the same label can only appear on another sock
 221                  */
 222                 lfl = __fl_lookup(net, fl->label);
 223                 if (lfl != NULL) {
 224                         atomic_inc(&lfl->users);
 225                         spin_unlock_bh(&ip6_fl_lock);
 226                         return lfl;
 227                 }
 228         }
 229
 230         fl->lastuse = jiffies;
 231         fl->next = fl_ht[FL_HASH(fl->label)];
 232         rcu_assign_pointer(fl_ht[FL_HASH(fl->label)], fl);
 233         atomic_inc(&fl_size);
 234         spin_unlock_bh(&ip6_fl_lock);
 235         return NULL;
 236 }
 237
 238
 239
 240 /* Socket flowlabel lists */
 241
 242 struct ip6_flowlabel * fl6_sock_lookup(struct sock *sk, __be32 label)
 243 {
 244         struct ipv6_fl_socklist *sfl;
 245         struct ipv6_pinfo *np = inet6_sk(sk);
 246
 247         label &= IPV6_FLOWLABEL_MASK;
 248
 249         rcu_read_lock_bh();
 250         for_each_sk_fl_rcu(np, sfl) {
 251                 struct ip6_flowlabel *fl = sfl->fl;
 252                 if (fl->label == label) {
 253                         fl->lastuse = jiffies;
 254                         atomic_inc(&fl->users);
 255                         rcu_read_unlock_bh();
 256                         return fl;
 257                 }
 258         }
 259         rcu_read_unlock_bh();
 260         return NULL;
 261 }
 262
 263 EXPORT_SYMBOL_GPL(fl6_sock_lookup);
 264
 265 void fl6_free_socklist(struct sock *sk)
 266 {
 267         struct ipv6_pinfo *np = inet6_sk(sk);
 268         struct ipv6_fl_socklist *sfl;
 269
 270         if (!rcu_access_pointer(np->ipv6_fl_list))
 271                 return;
 272
 273         spin_lock_bh(&ip6_sk_fl_lock);
 274         while ((sfl = rcu_dereference_protected(np->ipv6_fl_list,
 275                                                 lockdep_is_held(&ip6_sk_fl_lock))) != NULL) {
 276                 np->ipv6_fl_list = sfl->next;
 277                 spin_unlock_bh(&ip6_sk_fl_lock);
 278
 279                 fl_release(sfl->fl);
 280                 kfree_rcu(sfl, rcu);
 281
 282                 spin_lock_bh(&ip6_sk_fl_lock);
 283         }
 284         spin_unlock_bh(&ip6_sk_fl_lock);
 285 }
 286
 287 /* Service routines */
 288
 289
 290 /*
 291    It is the only difficult place. flowlabel enforces equal headers
 292    before and including routing header, however user may supply options
 293    following rthdr.
 294  */
 295
 296 struct ipv6_txoptions *fl6_merge_options(struct ipv6_txoptions * opt_space,
 297                                          struct ip6_flowlabel * fl,
 298                                          struct ipv6_txoptions * fopt)
 299 {
 300         struct ipv6_txoptions * fl_opt = fl->opt;
 301
 302         if (fopt == NULL || fopt->opt_flen == 0)
 303                 return fl_opt;
 304
 305         if (fl_opt != NULL) {
 306                 opt_space->hopopt = fl_opt->hopopt;
 307                 opt_space->dst0opt = fl_opt->dst0opt;
 308                 opt_space->srcrt = fl_opt->srcrt;
 309                 opt_space->opt_nflen = fl_opt->opt_nflen;
 310         } else {
 311                 if (fopt->opt_nflen == 0)
 312                         return fopt;
 313                 opt_space->hopopt = NULL;
 314                 opt_space->dst0opt = NULL;
 315                 opt_space->srcrt = NULL;
 316                 opt_space->opt_nflen = 0;
 317         }
 318         opt_space->dst1opt = fopt->dst1opt;
 319         opt_space->opt_flen = fopt->opt_flen;
 320         return opt_space;
 321 }
 322 EXPORT_SYMBOL_GPL(fl6_merge_options);
 323
 324 static unsigned long check_linger(unsigned long ttl)
 325 {
 326         if (ttl < FL_MIN_LINGER)
 327                 return FL_MIN_LINGER*HZ;
 328         if (ttl > FL_MAX_LINGER && !capable(CAP_NET_ADMIN))
 329                 return 0;
 330         return ttl*HZ;
 331 }
 332
 333 static int fl6_renew(struct ip6_flowlabel *fl, unsigned long linger, unsigned long expires)
 334 {
 335         linger = check_linger(linger);
 336         if (!linger)
 337                 return -EPERM;
 338         expires = check_linger(expires);
 339         if (!expires)
 340                 return -EPERM;
 341
 342         spin_lock_bh(&ip6_fl_lock);
 343         fl->lastuse = jiffies;
 344         if (time_before(fl->linger, linger))
 345                 fl->linger = linger;
 346         if (time_before(expires, fl->linger))
 347                 expires = fl->linger;
 348         if (time_before(fl->expires, fl->lastuse + expires))
 349                 fl->expires = fl->lastuse + expires;
 350         spin_unlock_bh(&ip6_fl_lock);
 351
 352         return 0;
 353 }
 354
 355 static struct ip6_flowlabel *
 356 fl_create(struct net *net, struct sock *sk, struct in6_flowlabel_req *freq,
 357           char __user *optval, int optlen, int *err_p)
 358 {
 359         struct ip6_flowlabel *fl = NULL;
 360         int olen;
 361         int addr_type;
 362         int err;
 363
 364         olen = optlen - CMSG_ALIGN(sizeof(*freq));
 365         err = -EINVAL;
 366         if (olen > 64 * 1024)
 367                 goto done;
 368
 369         err = -ENOMEM;
 370         fl = kzalloc(sizeof(*fl), GFP_KERNEL);
 371         if (fl == NULL)
 372                 goto done;
 373
 374         if (olen > 0) {
 375                 struct msghdr msg;
 376                 struct flowi6 flowi6;
 377                 int junk;
 378
 379                 err = -ENOMEM;
 380                 fl->opt = kmalloc(sizeof(*fl->opt) + olen, GFP_KERNEL);
 381                 if (fl->opt == NULL)
 382                         goto done;
 383
 384                 memset(fl->opt, 0, sizeof(*fl->opt));
 385                 fl->opt->tot_len = sizeof(*fl->opt) + olen;
 386                 err = -EFAULT;
 387                 if (copy_from_user(fl->opt+1, optval+CMSG_ALIGN(sizeof(*freq)), olen))
 388                         goto done;
 389
 390                 msg.msg_controllen = olen;
 391                 msg.msg_control = (void*)(fl->opt+1);
 392                 memset(&flowi6, 0, sizeof(flowi6));
 393
 394                 err = ip6_datagram_send_ctl(net, sk, &msg, &flowi6, fl->opt,
 395                                             &junk, &junk, &junk);
 396                 if (err)
 397                         goto done;
 398                 err = -EINVAL;
 399                 if (fl->opt->opt_flen)
 400                         goto done;
 401                 if (fl->opt->opt_nflen == 0) {
 402                         kfree(fl->opt);
 403                         fl->opt = NULL;
 404                 }
 405         }
 406
 407         fl->fl_net = hold_net(net);
 408         fl->expires = jiffies;
 409         err = fl6_renew(fl, freq->flr_linger, freq->flr_expires);
 410         if (err)
 411                 goto done;
 412         fl->share = freq->flr_share;
 413         addr_type = ipv6_addr_type(&freq->flr_dst);
 414         if ((addr_type & IPV6_ADDR_MAPPED) ||
 415             addr_type == IPV6_ADDR_ANY) {
 416                 err = -EINVAL;
 417                 goto done;
 418         }
 419         fl->dst = freq->flr_dst;
 420         atomic_set(&fl->users, 1);
 421         switch (fl->share) {
 422         case IPV6_FL_S_EXCL:
 423         case IPV6_FL_S_ANY:
 424                 break;
 425         case IPV6_FL_S_PROCESS:
 426                 fl->owner.pid = get_task_pid(current, PIDTYPE_PID);
 427                 break;
 428         case IPV6_FL_S_USER:
 429                 fl->owner.uid = current_euid();
 430                 break;
 431         default:
 432                 err = -EINVAL;
 433                 goto done;
 434         }
 435         return fl;
 436
 437 done:
 438         fl_free(fl);
 439         *err_p = err;
 440         return NULL;
 441 }
 442
 443 static int mem_check(struct sock *sk)
 444 {
 445         struct ipv6_pinfo *np = inet6_sk(sk);
 446         struct ipv6_fl_socklist *sfl;
 447         int room = FL_MAX_SIZE - atomic_read(&fl_size);
 448         int count = 0;
 449
 450         if (room > FL_MAX_SIZE - FL_MAX_PER_SOCK)
 451                 return 0;
 452
 453         rcu_read_lock_bh();
 454         for_each_sk_fl_rcu(np, sfl)
 455                 count++;
 456         rcu_read_unlock_bh();
 457
 458         if (room <= 0 ||
 459             ((count >= FL_MAX_PER_SOCK ||
 460               (count > 0 && room < FL_MAX_SIZE/2) || room < FL_MAX_SIZE/4) &&
 461              !capable(CAP_NET_ADMIN)))
 462                 return -ENOBUFS;
 463
 464         return 0;
 465 }
 466
 467 static inline void fl_link(struct ipv6_pinfo *np, struct ipv6_fl_socklist *sfl,
 468                 struct ip6_flowlabel *fl)
 469 {
 470         spin_lock_bh(&ip6_sk_fl_lock);
 471         sfl->fl = fl;
 472         sfl->next = np->ipv6_fl_list;
 473         rcu_assign_pointer(np->ipv6_fl_list, sfl);
 474         spin_unlock_bh(&ip6_sk_fl_lock);
 475 }
 476
 477 int ipv6_flowlabel_opt_get(struct sock *sk, struct in6_flowlabel_req *freq,
 478                            int flags)
 479 {
 480         struct ipv6_pinfo *np = inet6_sk(sk);
 481         struct ipv6_fl_socklist *sfl;
 482
 483         if (flags & IPV6_FL_F_REMOTE) {
 484                 freq->flr_label = np->rcv_flowinfo & IPV6_FLOWLABEL_MASK;
 485                 return 0;
 486         }
 487
 488         if (np->repflow) {
 489                 freq->flr_label = np->flow_label;
 490                 return 0;
 491         }
 492
 493         rcu_read_lock_bh();
 494
 495         for_each_sk_fl_rcu(np, sfl) {
 496                 if (sfl->fl->label == (np->flow_label & IPV6_FLOWLABEL_MASK)) {
 497                         spin_lock_bh(&ip6_fl_lock);
 498                         freq->flr_label = sfl->fl->label;
 499                         freq->flr_dst = sfl->fl->dst;
 500                         freq->flr_share = sfl->fl->share;
 501                         freq->flr_expires = (sfl->fl->expires - jiffies) / HZ;
 502                         freq->flr_linger = sfl->fl->linger / HZ;
 503
 504                         spin_unlock_bh(&ip6_fl_lock);
 505                         rcu_read_unlock_bh();
 506                         return 0;
 507                 }
 508         }
 509         rcu_read_unlock_bh();
 510
 511         return -ENOENT;
 512 }
 513
 514 int ipv6_flowlabel_opt(struct sock *sk, char __user *optval, int optlen)
 515 {
 516         int uninitialized_var(err);
 517         struct net *net = sock_net(sk);
 518         struct ipv6_pinfo *np = inet6_sk(sk);
 519         struct in6_flowlabel_req freq;
 520         struct ipv6_fl_socklist *sfl1=NULL;
 521         struct ipv6_fl_socklist *sfl;
 522         struct ipv6_fl_socklist __rcu **sflp;
 523         struct ip6_flowlabel *fl, *fl1 = NULL;
 524
 525
 526         if (optlen < sizeof(freq))
 527                 return -EINVAL;
 528
 529         if (copy_from_user(&freq, optval, sizeof(freq)))
 530                 return -EFAULT;
 531
 532         switch (freq.flr_action) {
 533         case IPV6_FL_A_PUT:
 534                 if (freq.flr_flags & IPV6_FL_F_REFLECT) {
 535                         if (sk->sk_protocol != IPPROTO_TCP)
 536                                 return -ENOPROTOOPT;
 537                         if (!np->repflow)
 538                                 return -ESRCH;
 539                         np->flow_label = 0;
 540                         np->repflow = 0;
 541                         return 0;
 542                 }
 543                 spin_lock_bh(&ip6_sk_fl_lock);
 544                 for (sflp = &np->ipv6_fl_list;
 545                      (sfl = rcu_dereference(*sflp))!=NULL;
 546                      sflp = &sfl->next) {
 547                         if (sfl->fl->label == freq.flr_label) {
 548                                 if (freq.flr_label == (np->flow_label&IPV6_FLOWLABEL_MASK))
 549                                         np->flow_label &= ~IPV6_FLOWLABEL_MASK;
 550                                 *sflp = rcu_dereference(sfl->next);
 551                                 spin_unlock_bh(&ip6_sk_fl_lock);
 552                                 fl_release(sfl->fl);
 553                                 kfree_rcu(sfl, rcu);
 554                                 return 0;
 555                         }
 556                 }
 557                 spin_unlock_bh(&ip6_sk_fl_lock);
 558                 return -ESRCH;
 559
 560         case IPV6_FL_A_RENEW:
 561                 rcu_read_lock_bh();
 562                 for_each_sk_fl_rcu(np, sfl) {
 563                         if (sfl->fl->label == freq.flr_label) {
 564                                 err = fl6_renew(sfl->fl, freq.flr_linger, freq.flr_expires);
 565                                 rcu_read_unlock_bh();
 566                                 return err;
 567                         }
 568                 }
 569                 rcu_read_unlock_bh();
 570
 571                 if (freq.flr_share == IPV6_FL_S_NONE &&
 572                     ns_capable(net->user_ns, CAP_NET_ADMIN)) {
 573                         fl = fl_lookup(net, freq.flr_label);
 574                         if (fl) {
 575                                 err = fl6_renew(fl, freq.flr_linger, freq.flr_expires);
 576                                 fl_release(fl);
 577                                 return err;
 578                         }
 579                 }
 580                 return -ESRCH;
 581
 582         case IPV6_FL_A_GET:
 583                 if (freq.flr_flags & IPV6_FL_F_REFLECT) {
 584                         struct net *net = sock_net(sk);
 585                         if (net->ipv6.sysctl.flowlabel_consistency) {
 586                                 net_info_ratelimited("Can not set IPV6_FL_F_REFLECT if flowlabel_consistency sysctl is enable\n");
 587                                 return -EPERM;
 588                         }
 589
 590                         if (sk->sk_protocol != IPPROTO_TCP)
 591                                 return -ENOPROTOOPT;
 592
 593                         np->repflow = 1;
 594                         return 0;
 595                 }
 596
 597                 if (freq.flr_label & ~IPV6_FLOWLABEL_MASK)
 598                         return -EINVAL;
 599
 600                 fl = fl_create(net, sk, &freq, optval, optlen, &err);
 601                 if (fl == NULL)
 602                         return err;
 603                 sfl1 = kmalloc(sizeof(*sfl1), GFP_KERNEL);
 604
 605                 if (freq.flr_label) {
 606                         err = -EEXIST;
 607                         rcu_read_lock_bh();
 608                         for_each_sk_fl_rcu(np, sfl) {
 609                                 if (sfl->fl->label == freq.flr_label) {
 610                                         if (freq.flr_flags&IPV6_FL_F_EXCL) {
 611                                                 rcu_read_unlock_bh();
 612                                                 goto done;
 613                                         }
 614                                         fl1 = sfl->fl;
 615                                         atomic_inc(&fl1->users);
 616                                         break;
 617                                 }
 618                         }
 619                         rcu_read_unlock_bh();
 620
 621                         if (fl1 == NULL)
 622                                 fl1 = fl_lookup(net, freq.flr_label);
 623                         if (fl1) {
 624 recheck:
 625                                 err = -EEXIST;
 626                                 if (freq.flr_flags&IPV6_FL_F_EXCL)
 627                                         goto release;
 628                                 err = -EPERM;
 629                                 if (fl1->share == IPV6_FL_S_EXCL ||
 630                                     fl1->share != fl->share ||
 631                                     ((fl1->share == IPV6_FL_S_PROCESS) &&
 632                                      (fl1->owner.pid == fl->owner.pid)) ||
 633                                     ((fl1->share == IPV6_FL_S_USER) &&
 634                                      uid_eq(fl1->owner.uid, fl->owner.uid)))
 635                                         goto release;
 636
 637                                 err = -ENOMEM;
 638                                 if (sfl1 == NULL)
 639                                         goto release;
 640                                 if (fl->linger > fl1->linger)
 641                                         fl1->linger = fl->linger;
 642                                 if ((long)(fl->expires - fl1->expires) > 0)
 643                                         fl1->expires = fl->expires;
 644                                 fl_link(np, sfl1, fl1);
 645                                 fl_free(fl);
 646                                 return 0;
 647
 648 release:
 649                                 fl_release(fl1);
 650                                 goto done;
 651                         }
 652                 }
 653                 err = -ENOENT;
 654                 if (!(freq.flr_flags&IPV6_FL_F_CREATE))
 655                         goto done;
 656
 657                 err = -ENOMEM;
 658                 if (sfl1 == NULL || (err = mem_check(sk)) != 0)
 659                         goto done;
 660
 661                 fl1 = fl_intern(net, fl, freq.flr_label);
 662                 if (fl1 != NULL)
 663                         goto recheck;
 664
 665                 if (!freq.flr_label) {
 666                         if (copy_to_user(&((struct in6_flowlabel_req __user *) optval)->flr_label,
 667                                          &fl->label, sizeof(fl->label))) {
 668                                 /* Intentionally ignore fault. */
 669                         }
 670                 }
 671
 672                 fl_link(np, sfl1, fl);
 673                 return 0;
 674
 675         default:
 676                 return -EINVAL;
 677         }
 678
 679 done:
 680         fl_free(fl);
 681         kfree(sfl1);
 682         return err;
 683 }
 684
 685 #ifdef CONFIG_PROC_FS
 686
 687 struct ip6fl_iter_state {
 688         struct seq_net_private p;
 689         struct pid_namespace *pid_ns;
 690         int bucket;
 691 };
 692
 693 #define ip6fl_seq_private(seq)  ((struct ip6fl_iter_state *)(seq)->private)
 694
 695 static struct ip6_flowlabel *ip6fl_get_first(struct seq_file *seq)
 696 {
 697         struct ip6_flowlabel *fl = NULL;
 698         struct ip6fl_iter_state *state = ip6fl_seq_private(seq);
 699         struct net *net = seq_file_net(seq);
 700
 701         for (state->bucket = 0; state->bucket <= FL_HASH_MASK; ++state->bucket) {
 702                 for_each_fl_rcu(state->bucket, fl) {
 703                         if (net_eq(fl->fl_net, net))
 704                                 goto out;
 705                 }
 706         }
 707         fl = NULL;
 708 out:
 709         return fl;
 710 }
 711
 712 static struct ip6_flowlabel *ip6fl_get_next(struct seq_file *seq, struct ip6_flowlabel *fl)
 713 {
 714         struct ip6fl_iter_state *state = ip6fl_seq_private(seq);
 715         struct net *net = seq_file_net(seq);
 716
 717         for_each_fl_continue_rcu(fl) {
 718                 if (net_eq(fl->fl_net, net))
 719                         goto out;
 720         }
 721
 722 try_again:
 723         if (++state->bucket <= FL_HASH_MASK) {
 724                 for_each_fl_rcu(state->bucket, fl) {
 725                         if (net_eq(fl->fl_net, net))
 726                                 goto out;
 727                 }
 728                 goto try_again;
 729         }
 730         fl = NULL;
 731
 732 out:
 733         return fl;
 734 }
 735
 736 static struct ip6_flowlabel *ip6fl_get_idx(struct seq_file *seq, loff_t pos)
 737 {
 738         struct ip6_flowlabel *fl = ip6fl_get_first(seq);
 739         if (fl)
 740                 while (pos && (fl = ip6fl_get_next(seq, fl)) != NULL)
 741                         --pos;
 742         return pos ? NULL : fl;
 743 }
 744
 745 static void *ip6fl_seq_start(struct seq_file *seq, loff_t *pos)
 746         __acquires(RCU)
 747 {
 748         rcu_read_lock_bh();
 749         return *pos ? ip6fl_get_idx(seq, *pos - 1) : SEQ_START_TOKEN;
 750 }
 751
 752 static void *ip6fl_seq_next(struct seq_file *seq, void *v, loff_t *pos)
 753 {
 754         struct ip6_flowlabel *fl;
 755
 756         if (v == SEQ_START_TOKEN)
 757                 fl = ip6fl_get_first(seq);
 758         else
 759                 fl = ip6fl_get_next(seq, v);
 760         ++*pos;
 761         return fl;
 762 }
 763
 764 static void ip6fl_seq_stop(struct seq_file *seq, void *v)
 765         __releases(RCU)
 766 {
 767         rcu_read_unlock_bh();
 768 }
 769
 770 static int ip6fl_seq_show(struct seq_file *seq, void *v)
 771 {
 772         struct ip6fl_iter_state *state = ip6fl_seq_private(seq);
 773         if (v == SEQ_START_TOKEN)
 774                 seq_printf(seq, "%-5s %-1s %-6s %-6s %-6s %-8s %-32s %s\n",
 775                            "Label", "S", "Owner", "Users", "Linger", "Expires", "Dst", "Opt");
 776         else {
 777                 struct ip6_flowlabel *fl = v;
 778                 seq_printf(seq,
 779                            "%05X %-1d %-6d %-6d %-6ld %-8ld %pi6 %-4d\n",
 780                            (unsigned int)ntohl(fl->label),
 781                            fl->share,
 782                            ((fl->share == IPV6_FL_S_PROCESS) ?
 783                             pid_nr_ns(fl->owner.pid, state->pid_ns) :
 784                             ((fl->share == IPV6_FL_S_USER) ?
 785                              from_kuid_munged(seq_user_ns(seq), fl->owner.uid) :
 786                              0)),
 787                            atomic_read(&fl->users),
 788                            fl->linger/HZ,
 789                            (long)(fl->expires - jiffies)/HZ,
 790                            &fl->dst,
 791                            fl->opt ? fl->opt->opt_nflen : 0);
 792         }
 793         return 0;
 794 }
 795
 796 static const struct seq_operations ip6fl_seq_ops = {
 797         .start  =       ip6fl_seq_start,
 798         .next   =       ip6fl_seq_next,
 799         .stop   =       ip6fl_seq_stop,
 800         .show   =       ip6fl_seq_show,
 801 };
 802
 803 static int ip6fl_seq_open(struct inode *inode, struct file *file)
 804 {
 805         struct seq_file *seq;
 806         struct ip6fl_iter_state *state;
 807         int err;
 808
 809         err = seq_open_net(inode, file, &ip6fl_seq_ops,
 810                            sizeof(struct ip6fl_iter_state));
 811
 812         if (!err) {
 813                 seq = file->private_data;
 814                 state = ip6fl_seq_private(seq);
 815                 rcu_read_lock();
 816                 state->pid_ns = get_pid_ns(task_active_pid_ns(current));
 817                 rcu_read_unlock();
 818         }
 819         return err;
 820 }
 821
 822 static int ip6fl_seq_release(struct inode *inode, struct file *file)
 823 {
 824         struct seq_file *seq = file->private_data;
 825         struct ip6fl_iter_state *state = ip6fl_seq_private(seq);
 826         put_pid_ns(state->pid_ns);
 827         return seq_release_net(inode, file);
 828 }
 829
 830 static const struct file_operations ip6fl_seq_fops = {
 831         .owner          =       THIS_MODULE,
 832         .open           =       ip6fl_seq_open,
 833         .read           =       seq_read,
 834         .llseek         =       seq_lseek,
 835         .release        =       ip6fl_seq_release,
 836 };
 837
 838 static int __net_init ip6_flowlabel_proc_init(struct net *net)
 839 {
 840         if (!proc_create("ip6_flowlabel", S_IRUGO, net->proc_net,
 841                          &ip6fl_seq_fops))
 842                 return -ENOMEM;
 843         return 0;
 844 }
 845
 846 static void __net_exit ip6_flowlabel_proc_fini(struct net *net)
 847 {
 848         remove_proc_entry("ip6_flowlabel", net->proc_net);
 849 }
 850 #else
 851 static inline int ip6_flowlabel_proc_init(struct net *net)
 852 {
 853         return 0;
 854 }
 855 static inline void ip6_flowlabel_proc_fini(struct net *net)
 856 {
 857 }
 858 #endif
 859
 860 static void __net_exit ip6_flowlabel_net_exit(struct net *net)
 861 {
 862         ip6_fl_purge(net);
 863         ip6_flowlabel_proc_fini(net);
 864 }
 865
 866 static struct pernet_operations ip6_flowlabel_net_ops = {
 867         .init = ip6_flowlabel_proc_init,
 868         .exit = ip6_flowlabel_net_exit,
 869 };
 870
 871 int ip6_flowlabel_init(void)
 872 {
 873         return register_pernet_subsys(&ip6_flowlabel_net_ops);
 874 }
 875
 876 void ip6_flowlabel_cleanup(void)
 877 {
 878         del_timer(&ip6_fl_gc_timer);
 879         unregister_pernet_subsys(&ip6_flowlabel_net_ops);
 880 }