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