net/ipv4/sysctl_net_ipv4.c

   1 /*
   2  * sysctl_net_ipv4.c: sysctl interface to net IPV4 subsystem.
   3  *
   4  * Begun April 1, 1996, Mike Shaver.
   5  * Added /proc/sys/net/ipv4 directory entry (empty =) ). [MS]
   6  */
   7
   8 #include <linux/mm.h>
   9 #include <linux/module.h>
  10 #include <linux/sysctl.h>
  11 #include <linux/igmp.h>
  12 #include <linux/inetdevice.h>
  13 #include <linux/seqlock.h>
  14 #include <linux/init.h>
  15 #include <linux/slab.h>
  16 #include <linux/nsproxy.h>
  17 #include <linux/swap.h>
  18 #include <net/snmp.h>
  19 #include <net/icmp.h>
  20 #include <net/ip.h>
  21 #include <net/route.h>
  22 #include <net/tcp.h>
  23 #include <net/udp.h>
  24 #include <net/cipso_ipv4.h>
  25 #include <net/inet_frag.h>
  26 #include <net/ping.h>
  27 #include <net/tcp_memcontrol.h>
  28
  29 static int zero;
  30 static int one = 1;
  31 static int four = 4;
  32 static int thousand = 1000;
  33 static int gso_max_segs = GSO_MAX_SEGS;
  34 static int tcp_retr1_max = 255;
  35 static int ip_local_port_range_min[] = { 1, 1 };
  36 static int ip_local_port_range_max[] = { 65535, 65535 };
  37 static int tcp_adv_win_scale_min = -31;
  38 static int tcp_adv_win_scale_max = 31;
  39 static int ip_ttl_min = 1;
  40 static int ip_ttl_max = 255;
  41 static int tcp_syn_retries_min = 1;
  42 static int tcp_syn_retries_max = MAX_TCP_SYNCNT;
  43 static int ip_ping_group_range_min[] = { 0, 0 };
  44 static int ip_ping_group_range_max[] = { GID_T_MAX, GID_T_MAX };
  45
  46 /* Update system visible IP port range */
  47 static void set_local_port_range(struct net *net, int range[2])
  48 {
  49         bool same_parity = !((range[0] ^ range[1]) & 1);
  50
  51         write_seqlock_bh(&net->ipv4.ip_local_ports.lock);
  52         if (same_parity && !net->ipv4.ip_local_ports.warned) {
  53                 net->ipv4.ip_local_ports.warned = true;
  54                 pr_err_ratelimited("ip_local_port_range: prefer different parity for start/end values.\n");
  55         }
  56         net->ipv4.ip_local_ports.range[0] = range[0];
  57         net->ipv4.ip_local_ports.range[1] = range[1];
  58         write_sequnlock_bh(&net->ipv4.ip_local_ports.lock);
  59 }
  60
  61 /* Validate changes from /proc interface. */
  62 static int ipv4_local_port_range(struct ctl_table *table, int write,
  63                                  void __user *buffer,
  64                                  size_t *lenp, loff_t *ppos)
  65 {
  66         struct net *net =
  67                 container_of(table->data, struct net, ipv4.ip_local_ports.range);
  68         int ret;
  69         int range[2];
  70         struct ctl_table tmp = {
  71                 .data = &range,
  72                 .maxlen = sizeof(range),
  73                 .mode = table->mode,
  74                 .extra1 = &ip_local_port_range_min,
  75                 .extra2 = &ip_local_port_range_max,
  76         };
  77
  78         inet_get_local_port_range(net, &range[0], &range[1]);
  79
  80         ret = proc_dointvec_minmax(&tmp, write, buffer, lenp, ppos);
  81
  82         if (write && ret == 0) {
  83                 if (range[1] < range[0])
  84                         ret = -EINVAL;
  85                 else
  86                         set_local_port_range(net, range);
  87         }
  88
  89         return ret;
  90 }
  91
  92
  93 static void inet_get_ping_group_range_table(struct ctl_table *table, kgid_t *low, kgid_t *high)
  94 {
  95         kgid_t *data = table->data;
  96         struct net *net =
  97                 container_of(table->data, struct net, ipv4.ping_group_range.range);
  98         unsigned int seq;
  99         do {
 100                 seq = read_seqbegin(&net->ipv4.ip_local_ports.lock);
 101
 102                 *low = data[0];
 103                 *high = data[1];
 104         } while (read_seqretry(&net->ipv4.ip_local_ports.lock, seq));
 105 }
 106
 107 /* Update system visible IP port range */
 108 static void set_ping_group_range(struct ctl_table *table, kgid_t low, kgid_t high)
 109 {
 110         kgid_t *data = table->data;
 111         struct net *net =
 112                 container_of(table->data, struct net, ipv4.ping_group_range.range);
 113         write_seqlock(&net->ipv4.ip_local_ports.lock);
 114         data[0] = low;
 115         data[1] = high;
 116         write_sequnlock(&net->ipv4.ip_local_ports.lock);
 117 }
 118
 119 /* Validate changes from /proc interface. */
 120 static int ipv4_ping_group_range(struct ctl_table *table, int write,
 121                                  void __user *buffer,
 122                                  size_t *lenp, loff_t *ppos)
 123 {
 124         struct user_namespace *user_ns = current_user_ns();
 125         int ret;
 126         gid_t urange[2];
 127         kgid_t low, high;
 128         struct ctl_table tmp = {
 129                 .data = &urange,
 130                 .maxlen = sizeof(urange),
 131                 .mode = table->mode,
 132                 .extra1 = &ip_ping_group_range_min,
 133                 .extra2 = &ip_ping_group_range_max,
 134         };
 135
 136         inet_get_ping_group_range_table(table, &low, &high);
 137         urange[0] = from_kgid_munged(user_ns, low);
 138         urange[1] = from_kgid_munged(user_ns, high);
 139         ret = proc_dointvec_minmax(&tmp, write, buffer, lenp, ppos);
 140
 141         if (write && ret == 0) {
 142                 low = make_kgid(user_ns, urange[0]);
 143                 high = make_kgid(user_ns, urange[1]);
 144                 if (!gid_valid(low) || !gid_valid(high) ||
 145                     (urange[1] < urange[0]) || gid_lt(high, low)) {
 146                         low = make_kgid(&init_user_ns, 1);
 147                         high = make_kgid(&init_user_ns, 0);
 148                 }
 149                 set_ping_group_range(table, low, high);
 150         }
 151
 152         return ret;
 153 }
 154
 155 /* Validate changes from /proc interface. */
 156 static int proc_tcp_default_init_rwnd(struct ctl_table *ctl, int write,
 157                                       void __user *buffer,
 158                                       size_t *lenp, loff_t *ppos)
 159 {
 160         int old_value = *(int *)ctl->data;
 161         int ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
 162         int new_value = *(int *)ctl->data;
 163
 164         if (write && ret == 0 && (new_value < 3 || new_value > 100))
 165                 *(int *)ctl->data = old_value;
 166
 167         return ret;
 168 }
 169
 170 static int proc_tcp_congestion_control(struct ctl_table *ctl, int write,
 171                                        void __user *buffer, size_t *lenp, loff_t *ppos)
 172 {
 173         char val[TCP_CA_NAME_MAX];
 174         struct ctl_table tbl = {
 175                 .data = val,
 176                 .maxlen = TCP_CA_NAME_MAX,
 177         };
 178         int ret;
 179
 180         tcp_get_default_congestion_control(val);
 181
 182         ret = proc_dostring(&tbl, write, buffer, lenp, ppos);
 183         if (write && ret == 0)
 184                 ret = tcp_set_default_congestion_control(val);
 185         return ret;
 186 }
 187
 188 static int proc_tcp_available_congestion_control(struct ctl_table *ctl,
 189                                                  int write,
 190                                                  void __user *buffer, size_t *lenp,
 191                                                  loff_t *ppos)
 192 {
 193         struct ctl_table tbl = { .maxlen = TCP_CA_BUF_MAX, };
 194         int ret;
 195
 196         tbl.data = kmalloc(tbl.maxlen, GFP_USER);
 197         if (!tbl.data)
 198                 return -ENOMEM;
 199         tcp_get_available_congestion_control(tbl.data, TCP_CA_BUF_MAX);
 200         ret = proc_dostring(&tbl, write, buffer, lenp, ppos);
 201         kfree(tbl.data);
 202         return ret;
 203 }
 204
 205 static int proc_allowed_congestion_control(struct ctl_table *ctl,
 206                                            int write,
 207                                            void __user *buffer, size_t *lenp,
 208                                            loff_t *ppos)
 209 {
 210         struct ctl_table tbl = { .maxlen = TCP_CA_BUF_MAX };
 211         int ret;
 212
 213         tbl.data = kmalloc(tbl.maxlen, GFP_USER);
 214         if (!tbl.data)
 215                 return -ENOMEM;
 216
 217         tcp_get_allowed_congestion_control(tbl.data, tbl.maxlen);
 218         ret = proc_dostring(&tbl, write, buffer, lenp, ppos);
 219         if (write && ret == 0)
 220                 ret = tcp_set_allowed_congestion_control(tbl.data);
 221         kfree(tbl.data);
 222         return ret;
 223 }
 224
 225 static int proc_tcp_fastopen_key(struct ctl_table *ctl, int write,
 226                                  void __user *buffer, size_t *lenp,
 227                                  loff_t *ppos)
 228 {
 229         struct ctl_table tbl = { .maxlen = (TCP_FASTOPEN_KEY_LENGTH * 2 + 10) };
 230         struct tcp_fastopen_context *ctxt;
 231         int ret;
 232         u32  user_key[4]; /* 16 bytes, matching TCP_FASTOPEN_KEY_LENGTH */
 233
 234         tbl.data = kmalloc(tbl.maxlen, GFP_KERNEL);
 235         if (!tbl.data)
 236                 return -ENOMEM;
 237
 238         rcu_read_lock();
 239         ctxt = rcu_dereference(tcp_fastopen_ctx);
 240         if (ctxt)
 241                 memcpy(user_key, ctxt->key, TCP_FASTOPEN_KEY_LENGTH);
 242         else
 243                 memset(user_key, 0, sizeof(user_key));
 244         rcu_read_unlock();
 245
 246         snprintf(tbl.data, tbl.maxlen, "%08x-%08x-%08x-%08x",
 247                 user_key[0], user_key[1], user_key[2], user_key[3]);
 248         ret = proc_dostring(&tbl, write, buffer, lenp, ppos);
 249
 250         if (write && ret == 0) {
 251                 if (sscanf(tbl.data, "%x-%x-%x-%x", user_key, user_key + 1,
 252                            user_key + 2, user_key + 3) != 4) {
 253                         ret = -EINVAL;
 254                         goto bad_key;
 255                 }
 256                 /* Generate a dummy secret but don't publish it. This
 257                  * is needed so we don't regenerate a new key on the
 258                  * first invocation of tcp_fastopen_cookie_gen
 259                  */
 260                 tcp_fastopen_init_key_once(false);
 261                 tcp_fastopen_reset_cipher(user_key, TCP_FASTOPEN_KEY_LENGTH);
 262         }
 263
 264 bad_key:
 265         pr_debug("proc FO key set 0x%x-%x-%x-%x <- 0x%s: %u\n",
 266                user_key[0], user_key[1], user_key[2], user_key[3],
 267                (char *)tbl.data, ret);
 268         kfree(tbl.data);
 269         return ret;
 270 }
 271
 272 static struct ctl_table ipv4_table[] = {
 273         {
 274                 .procname       = "tcp_timestamps",
 275                 .data           = &sysctl_tcp_timestamps,
 276                 .maxlen         = sizeof(int),
 277                 .mode           = 0644,
 278                 .proc_handler   = proc_dointvec
 279         },
 280         {
 281                 .procname       = "tcp_window_scaling",
 282                 .data           = &sysctl_tcp_window_scaling,
 283                 .maxlen         = sizeof(int),
 284                 .mode           = 0644,
 285                 .proc_handler   = proc_dointvec
 286         },
 287         {
 288                 .procname       = "tcp_sack",
 289                 .data           = &sysctl_tcp_sack,
 290                 .maxlen         = sizeof(int),
 291                 .mode           = 0644,
 292                 .proc_handler   = proc_dointvec
 293         },
 294         {
 295                 .procname       = "tcp_retrans_collapse",
 296                 .data           = &sysctl_tcp_retrans_collapse,
 297                 .maxlen         = sizeof(int),
 298                 .mode           = 0644,
 299                 .proc_handler   = proc_dointvec
 300         },
 301         {
 302                 .procname       = "ip_default_ttl",
 303                 .data           = &sysctl_ip_default_ttl,
 304                 .maxlen         = sizeof(int),
 305                 .mode           = 0644,
 306                 .proc_handler   = proc_dointvec_minmax,
 307                 .extra1         = &ip_ttl_min,
 308                 .extra2         = &ip_ttl_max,
 309         },
 310         {
 311                 .procname       = "tcp_syn_retries",
 312                 .data           = &sysctl_tcp_syn_retries,
 313                 .maxlen         = sizeof(int),
 314                 .mode           = 0644,
 315                 .proc_handler   = proc_dointvec_minmax,
 316                 .extra1         = &tcp_syn_retries_min,
 317                 .extra2         = &tcp_syn_retries_max
 318         },
 319         {
 320                 .procname       = "tcp_synack_retries",
 321                 .data           = &sysctl_tcp_synack_retries,
 322                 .maxlen         = sizeof(int),
 323                 .mode           = 0644,
 324                 .proc_handler   = proc_dointvec
 325         },
 326         {
 327                 .procname       = "tcp_max_orphans",
 328                 .data           = &sysctl_tcp_max_orphans,
 329                 .maxlen         = sizeof(int),
 330                 .mode           = 0644,
 331                 .proc_handler   = proc_dointvec
 332         },
 333         {
 334                 .procname       = "tcp_max_tw_buckets",
 335                 .data           = &tcp_death_row.sysctl_max_tw_buckets,
 336                 .maxlen         = sizeof(int),
 337                 .mode           = 0644,
 338                 .proc_handler   = proc_dointvec
 339         },
 340         {
 341                 .procname       = "ip_early_demux",
 342                 .data           = &sysctl_ip_early_demux,
 343                 .maxlen         = sizeof(int),
 344                 .mode           = 0644,
 345                 .proc_handler   = proc_dointvec
 346         },
 347         {
 348                 .procname       = "ip_dynaddr",
 349                 .data           = &sysctl_ip_dynaddr,
 350                 .maxlen         = sizeof(int),
 351                 .mode           = 0644,
 352                 .proc_handler   = proc_dointvec
 353         },
 354         {
 355                 .procname       = "tcp_keepalive_time",
 356                 .data           = &sysctl_tcp_keepalive_time,
 357                 .maxlen         = sizeof(int),
 358                 .mode           = 0644,
 359                 .proc_handler   = proc_dointvec_jiffies,
 360         },
 361         {
 362                 .procname       = "tcp_keepalive_probes",
 363                 .data           = &sysctl_tcp_keepalive_probes,
 364                 .maxlen         = sizeof(int),
 365                 .mode           = 0644,
 366                 .proc_handler   = proc_dointvec
 367         },
 368         {
 369                 .procname       = "tcp_keepalive_intvl",
 370                 .data           = &sysctl_tcp_keepalive_intvl,
 371                 .maxlen         = sizeof(int),
 372                 .mode           = 0644,
 373                 .proc_handler   = proc_dointvec_jiffies,
 374         },
 375         {
 376                 .procname       = "tcp_retries1",
 377                 .data           = &sysctl_tcp_retries1,
 378                 .maxlen         = sizeof(int),
 379                 .mode           = 0644,
 380                 .proc_handler   = proc_dointvec_minmax,
 381                 .extra2         = &tcp_retr1_max
 382         },
 383         {
 384                 .procname       = "tcp_retries2",
 385                 .data           = &sysctl_tcp_retries2,
 386                 .maxlen         = sizeof(int),
 387                 .mode           = 0644,
 388                 .proc_handler   = proc_dointvec
 389         },
 390         {
 391                 .procname       = "tcp_fin_timeout",
 392                 .data           = &sysctl_tcp_fin_timeout,
 393                 .maxlen         = sizeof(int),
 394                 .mode           = 0644,
 395                 .proc_handler   = proc_dointvec_jiffies,
 396         },
 397 #ifdef CONFIG_SYN_COOKIES
 398         {
 399                 .procname       = "tcp_syncookies",
 400                 .data           = &sysctl_tcp_syncookies,
 401                 .maxlen         = sizeof(int),
 402                 .mode           = 0644,
 403                 .proc_handler   = proc_dointvec
 404         },
 405 #endif
 406         {
 407                 .procname       = "tcp_fastopen",
 408                 .data           = &sysctl_tcp_fastopen,
 409                 .maxlen         = sizeof(int),
 410                 .mode           = 0644,
 411                 .proc_handler   = proc_dointvec,
 412         },
 413         {
 414                 .procname       = "tcp_fastopen_key",
 415                 .mode           = 0600,
 416                 .maxlen         = ((TCP_FASTOPEN_KEY_LENGTH * 2) + 10),
 417                 .proc_handler   = proc_tcp_fastopen_key,
 418         },
 419         {
 420                 .procname       = "tcp_tw_recycle",
 421                 .data           = &tcp_death_row.sysctl_tw_recycle,
 422                 .maxlen         = sizeof(int),
 423                 .mode           = 0644,
 424                 .proc_handler   = proc_dointvec
 425         },
 426         {
 427                 .procname       = "tcp_abort_on_overflow",
 428                 .data           = &sysctl_tcp_abort_on_overflow,
 429                 .maxlen         = sizeof(int),
 430                 .mode           = 0644,
 431                 .proc_handler   = proc_dointvec
 432         },
 433         {
 434                 .procname       = "tcp_stdurg",
 435                 .data           = &sysctl_tcp_stdurg,
 436                 .maxlen         = sizeof(int),
 437                 .mode           = 0644,
 438                 .proc_handler   = proc_dointvec
 439         },
 440         {
 441                 .procname       = "tcp_rfc1337",
 442                 .data           = &sysctl_tcp_rfc1337,
 443                 .maxlen         = sizeof(int),
 444                 .mode           = 0644,
 445                 .proc_handler   = proc_dointvec
 446         },
 447         {
 448                 .procname       = "tcp_max_syn_backlog",
 449                 .data           = &sysctl_max_syn_backlog,
 450                 .maxlen         = sizeof(int),
 451                 .mode           = 0644,
 452                 .proc_handler   = proc_dointvec
 453         },
 454         {
 455                 .procname       = "igmp_max_memberships",
 456                 .data           = &sysctl_igmp_max_memberships,
 457                 .maxlen         = sizeof(int),
 458                 .mode           = 0644,
 459                 .proc_handler   = proc_dointvec
 460         },
 461         {
 462                 .procname       = "igmp_max_msf",
 463                 .data           = &sysctl_igmp_max_msf,
 464                 .maxlen         = sizeof(int),
 465                 .mode           = 0644,
 466                 .proc_handler   = proc_dointvec
 467         },
 468 #ifdef CONFIG_IP_MULTICAST
 469         {
 470                 .procname       = "igmp_qrv",
 471                 .data           = &sysctl_igmp_qrv,
 472                 .maxlen         = sizeof(int),
 473                 .mode           = 0644,
 474                 .proc_handler   = proc_dointvec_minmax,
 475                 .extra1         = &one
 476         },
 477 #endif
 478         {
 479                 .procname       = "inet_peer_threshold",
 480                 .data           = &inet_peer_threshold,
 481                 .maxlen         = sizeof(int),
 482                 .mode           = 0644,
 483                 .proc_handler   = proc_dointvec
 484         },
 485         {
 486                 .procname       = "inet_peer_minttl",
 487                 .data           = &inet_peer_minttl,
 488                 .maxlen         = sizeof(int),
 489                 .mode           = 0644,
 490                 .proc_handler   = proc_dointvec_jiffies,
 491         },
 492         {
 493                 .procname       = "inet_peer_maxttl",
 494                 .data           = &inet_peer_maxttl,
 495                 .maxlen         = sizeof(int),
 496                 .mode           = 0644,
 497                 .proc_handler   = proc_dointvec_jiffies,
 498         },
 499         {
 500                 .procname       = "tcp_orphan_retries",
 501                 .data           = &sysctl_tcp_orphan_retries,
 502                 .maxlen         = sizeof(int),
 503                 .mode           = 0644,
 504                 .proc_handler   = proc_dointvec
 505         },
 506         {
 507                 .procname       = "tcp_fack",
 508                 .data           = &sysctl_tcp_fack,
 509                 .maxlen         = sizeof(int),
 510                 .mode           = 0644,
 511                 .proc_handler   = proc_dointvec
 512         },
 513         {
 514                 .procname       = "tcp_recovery",
 515                 .data           = &sysctl_tcp_recovery,
 516                 .maxlen         = sizeof(int),
 517                 .mode           = 0644,
 518                 .proc_handler   = proc_dointvec,
 519         },
 520         {
 521                 .procname       = "tcp_reordering",
 522                 .data           = &sysctl_tcp_reordering,
 523                 .maxlen         = sizeof(int),
 524                 .mode           = 0644,
 525                 .proc_handler   = proc_dointvec
 526         },
 527         {
 528                 .procname       = "tcp_max_reordering",
 529                 .data           = &sysctl_tcp_max_reordering,
 530                 .maxlen         = sizeof(int),
 531                 .mode           = 0644,
 532                 .proc_handler   = proc_dointvec
 533         },
 534         {
 535                 .procname       = "tcp_dsack",
 536                 .data           = &sysctl_tcp_dsack,
 537                 .maxlen         = sizeof(int),
 538                 .mode           = 0644,
 539                 .proc_handler   = proc_dointvec
 540         },
 541         {
 542                 .procname       = "tcp_mem",
 543                 .maxlen         = sizeof(sysctl_tcp_mem),
 544                 .data           = &sysctl_tcp_mem,
 545                 .mode           = 0644,
 546                 .proc_handler   = proc_doulongvec_minmax,
 547         },
 548         {
 549                 .procname       = "tcp_wmem",
 550                 .data           = &sysctl_tcp_wmem,
 551                 .maxlen         = sizeof(sysctl_tcp_wmem),
 552                 .mode           = 0644,
 553                 .proc_handler   = proc_dointvec_minmax,
 554                 .extra1         = &one,
 555         },
 556         {
 557                 .procname       = "tcp_notsent_lowat",
 558                 .data           = &sysctl_tcp_notsent_lowat,
 559                 .maxlen         = sizeof(sysctl_tcp_notsent_lowat),
 560                 .mode           = 0644,
 561                 .proc_handler   = proc_dointvec,
 562         },
 563         {
 564                 .procname       = "tcp_rmem",
 565                 .data           = &sysctl_tcp_rmem,
 566                 .maxlen         = sizeof(sysctl_tcp_rmem),
 567                 .mode           = 0644,
 568                 .proc_handler   = proc_dointvec_minmax,
 569                 .extra1         = &one,
 570         },
 571         {
 572                 .procname       = "tcp_app_win",
 573                 .data           = &sysctl_tcp_app_win,
 574                 .maxlen         = sizeof(int),
 575                 .mode           = 0644,
 576                 .proc_handler   = proc_dointvec
 577         },
 578         {
 579                 .procname       = "tcp_adv_win_scale",
 580                 .data           = &sysctl_tcp_adv_win_scale,
 581                 .maxlen         = sizeof(int),
 582                 .mode           = 0644,
 583                 .proc_handler   = proc_dointvec_minmax,
 584                 .extra1         = &tcp_adv_win_scale_min,
 585                 .extra2         = &tcp_adv_win_scale_max,
 586         },
 587         {
 588                 .procname       = "tcp_tw_reuse",
 589                 .data           = &sysctl_tcp_tw_reuse,
 590                 .maxlen         = sizeof(int),
 591                 .mode           = 0644,
 592                 .proc_handler   = proc_dointvec
 593         },
 594         {
 595                 .procname       = "tcp_frto",
 596                 .data           = &sysctl_tcp_frto,
 597                 .maxlen         = sizeof(int),
 598                 .mode           = 0644,
 599                 .proc_handler   = proc_dointvec
 600         },
 601         {
 602                 .procname       = "tcp_min_rtt_wlen",
 603                 .data           = &sysctl_tcp_min_rtt_wlen,
 604                 .maxlen         = sizeof(int),
 605                 .mode           = 0644,
 606                 .proc_handler   = proc_dointvec
 607         },
 608         {
 609                 .procname       = "tcp_low_latency",
 610                 .data           = &sysctl_tcp_low_latency,
 611                 .maxlen         = sizeof(int),
 612                 .mode           = 0644,
 613                 .proc_handler   = proc_dointvec
 614         },
 615         {
 616                 .procname       = "tcp_no_metrics_save",
 617                 .data           = &sysctl_tcp_nometrics_save,
 618                 .maxlen         = sizeof(int),
 619                 .mode           = 0644,
 620                 .proc_handler   = proc_dointvec,
 621         },
 622         {
 623                 .procname       = "tcp_moderate_rcvbuf",
 624                 .data           = &sysctl_tcp_moderate_rcvbuf,
 625                 .maxlen         = sizeof(int),
 626                 .mode           = 0644,
 627                 .proc_handler   = proc_dointvec,
 628         },
 629         {
 630                 .procname       = "tcp_tso_win_divisor",
 631                 .data           = &sysctl_tcp_tso_win_divisor,
 632                 .maxlen         = sizeof(int),
 633                 .mode           = 0644,
 634                 .proc_handler   = proc_dointvec,
 635         },
 636         {
 637                 .procname       = "tcp_congestion_control",
 638                 .mode           = 0644,
 639                 .maxlen         = TCP_CA_NAME_MAX,
 640                 .proc_handler   = proc_tcp_congestion_control,
 641         },
 642         {
 643                 .procname       = "tcp_workaround_signed_windows",
 644                 .data           = &sysctl_tcp_workaround_signed_windows,
 645                 .maxlen         = sizeof(int),
 646                 .mode           = 0644,
 647                 .proc_handler   = proc_dointvec
 648         },
 649         {
 650                 .procname       = "tcp_limit_output_bytes",
 651                 .data           = &sysctl_tcp_limit_output_bytes,
 652                 .maxlen         = sizeof(int),
 653                 .mode           = 0644,
 654                 .proc_handler   = proc_dointvec
 655         },
 656         {
 657                 .procname       = "tcp_challenge_ack_limit",
 658                 .data           = &sysctl_tcp_challenge_ack_limit,
 659                 .maxlen         = sizeof(int),
 660                 .mode           = 0644,
 661                 .proc_handler   = proc_dointvec
 662         },
 663         {
 664                 .procname       = "tcp_slow_start_after_idle",
 665                 .data           = &sysctl_tcp_slow_start_after_idle,
 666                 .maxlen         = sizeof(int),
 667                 .mode           = 0644,
 668                 .proc_handler   = proc_dointvec
 669         },
 670 #ifdef CONFIG_NETLABEL
 671         {
 672                 .procname       = "cipso_cache_enable",
 673                 .data           = &cipso_v4_cache_enabled,
 674                 .maxlen         = sizeof(int),
 675                 .mode           = 0644,
 676                 .proc_handler   = proc_dointvec,
 677         },
 678         {
 679                 .procname       = "cipso_cache_bucket_size",
 680                 .data           = &cipso_v4_cache_bucketsize,
 681                 .maxlen         = sizeof(int),
 682                 .mode           = 0644,
 683                 .proc_handler   = proc_dointvec,
 684         },
 685         {
 686                 .procname       = "cipso_rbm_optfmt",
 687                 .data           = &cipso_v4_rbm_optfmt,
 688                 .maxlen         = sizeof(int),
 689                 .mode           = 0644,
 690                 .proc_handler   = proc_dointvec,
 691         },
 692         {
 693                 .procname       = "cipso_rbm_strictvalid",
 694                 .data           = &cipso_v4_rbm_strictvalid,
 695                 .maxlen         = sizeof(int),
 696                 .mode           = 0644,
 697                 .proc_handler   = proc_dointvec,
 698         },
 699 #endif /* CONFIG_NETLABEL */
 700         {
 701                 .procname       = "tcp_available_congestion_control",
 702                 .maxlen         = TCP_CA_BUF_MAX,
 703                 .mode           = 0444,
 704                 .proc_handler   = proc_tcp_available_congestion_control,
 705         },
 706         {
 707                 .procname       = "tcp_allowed_congestion_control",
 708                 .maxlen         = TCP_CA_BUF_MAX,
 709                 .mode           = 0644,
 710                 .proc_handler   = proc_allowed_congestion_control,
 711         },
 712         {
 713                 .procname       = "tcp_thin_linear_timeouts",
 714                 .data           = &sysctl_tcp_thin_linear_timeouts,
 715                 .maxlen         = sizeof(int),
 716                 .mode           = 0644,
 717                 .proc_handler   = proc_dointvec
 718         },
 719         {
 720                 .procname       = "tcp_thin_dupack",
 721                 .data           = &sysctl_tcp_thin_dupack,
 722                 .maxlen         = sizeof(int),
 723                 .mode           = 0644,
 724                 .proc_handler   = proc_dointvec
 725         },
 726         {
 727                 .procname       = "tcp_early_retrans",
 728                 .data           = &sysctl_tcp_early_retrans,
 729                 .maxlen         = sizeof(int),
 730                 .mode           = 0644,
 731                 .proc_handler   = proc_dointvec_minmax,
 732                 .extra1         = &zero,
 733                 .extra2         = &four,
 734         },
 735         {
 736                 .procname       = "tcp_min_tso_segs",
 737                 .data           = &sysctl_tcp_min_tso_segs,
 738                 .maxlen         = sizeof(int),
 739                 .mode           = 0644,
 740                 .proc_handler   = proc_dointvec_minmax,
 741                 .extra1         = &one,
 742                 .extra2         = &gso_max_segs,
 743         },
 744         {
 745                 .procname       = "tcp_pacing_ss_ratio",
 746                 .data           = &sysctl_tcp_pacing_ss_ratio,
 747                 .maxlen         = sizeof(int),
 748                 .mode           = 0644,
 749                 .proc_handler   = proc_dointvec_minmax,
 750                 .extra1         = &zero,
 751                 .extra2         = &thousand,
 752         },
 753         {
 754                 .procname       = "tcp_pacing_ca_ratio",
 755                 .data           = &sysctl_tcp_pacing_ca_ratio,
 756                 .maxlen         = sizeof(int),
 757                 .mode           = 0644,
 758                 .proc_handler   = proc_dointvec_minmax,
 759                 .extra1         = &zero,
 760                 .extra2         = &thousand,
 761         },
 762         {
 763                 .procname       = "tcp_autocorking",
 764                 .data           = &sysctl_tcp_autocorking,
 765                 .maxlen         = sizeof(int),
 766                 .mode           = 0644,
 767                 .proc_handler   = proc_dointvec_minmax,
 768                 .extra1         = &zero,
 769                 .extra2         = &one,
 770         },
 771         {
 772                 .procname       = "tcp_invalid_ratelimit",
 773                 .data           = &sysctl_tcp_invalid_ratelimit,
 774                 .maxlen         = sizeof(int),
 775                 .mode           = 0644,
 776                 .proc_handler   = proc_dointvec_ms_jiffies,
 777         },
 778         {
 779                 .procname       = "tcp_default_init_rwnd",
 780                 .data           = &sysctl_tcp_default_init_rwnd,
 781                 .maxlen         = sizeof(int),
 782                 .mode           = 0644,
 783                 .proc_handler   = proc_tcp_default_init_rwnd
 784         },
 785         {
 786                 .procname       = "icmp_msgs_per_sec",
 787                 .data           = &sysctl_icmp_msgs_per_sec,
 788                 .maxlen         = sizeof(int),
 789                 .mode           = 0644,
 790                 .proc_handler   = proc_dointvec_minmax,
 791                 .extra1         = &zero,
 792         },
 793         {
 794                 .procname       = "icmp_msgs_burst",
 795                 .data           = &sysctl_icmp_msgs_burst,
 796                 .maxlen         = sizeof(int),
 797                 .mode           = 0644,
 798                 .proc_handler   = proc_dointvec_minmax,
 799                 .extra1         = &zero,
 800         },
 801         {
 802                 .procname       = "udp_mem",
 803                 .data           = &sysctl_udp_mem,
 804                 .maxlen         = sizeof(sysctl_udp_mem),
 805                 .mode           = 0644,
 806                 .proc_handler   = proc_doulongvec_minmax,
 807         },
 808         {
 809                 .procname       = "udp_rmem_min",
 810                 .data           = &sysctl_udp_rmem_min,
 811                 .maxlen         = sizeof(sysctl_udp_rmem_min),
 812                 .mode           = 0644,
 813                 .proc_handler   = proc_dointvec_minmax,
 814                 .extra1         = &one
 815         },
 816         {
 817                 .procname       = "udp_wmem_min",
 818                 .data           = &sysctl_udp_wmem_min,
 819                 .maxlen         = sizeof(sysctl_udp_wmem_min),
 820                 .mode           = 0644,
 821                 .proc_handler   = proc_dointvec_minmax,
 822                 .extra1         = &one
 823         },
 824         { }
 825 };
 826
 827 static struct ctl_table ipv4_net_table[] = {
 828         {
 829                 .procname       = "icmp_echo_ignore_all",
 830                 .data           = &init_net.ipv4.sysctl_icmp_echo_ignore_all,
 831                 .maxlen         = sizeof(int),
 832                 .mode           = 0644,
 833                 .proc_handler   = proc_dointvec
 834         },
 835         {
 836                 .procname       = "icmp_echo_ignore_broadcasts",
 837                 .data           = &init_net.ipv4.sysctl_icmp_echo_ignore_broadcasts,
 838                 .maxlen         = sizeof(int),
 839                 .mode           = 0644,
 840                 .proc_handler   = proc_dointvec
 841         },
 842         {
 843                 .procname       = "icmp_ignore_bogus_error_responses",
 844                 .data           = &init_net.ipv4.sysctl_icmp_ignore_bogus_error_responses,
 845                 .maxlen         = sizeof(int),
 846                 .mode           = 0644,
 847                 .proc_handler   = proc_dointvec
 848         },
 849         {
 850                 .procname       = "icmp_errors_use_inbound_ifaddr",
 851                 .data           = &init_net.ipv4.sysctl_icmp_errors_use_inbound_ifaddr,
 852                 .maxlen         = sizeof(int),
 853                 .mode           = 0644,
 854                 .proc_handler   = proc_dointvec
 855         },
 856         {
 857                 .procname       = "icmp_ratelimit",
 858                 .data           = &init_net.ipv4.sysctl_icmp_ratelimit,
 859                 .maxlen         = sizeof(int),
 860                 .mode           = 0644,
 861                 .proc_handler   = proc_dointvec_ms_jiffies,
 862         },
 863         {
 864                 .procname       = "icmp_ratemask",
 865                 .data           = &init_net.ipv4.sysctl_icmp_ratemask,
 866                 .maxlen         = sizeof(int),
 867                 .mode           = 0644,
 868                 .proc_handler   = proc_dointvec
 869         },
 870         {
 871                 .procname       = "ping_group_range",
 872                 .data           = &init_net.ipv4.ping_group_range.range,
 873                 .maxlen         = sizeof(gid_t)*2,
 874                 .mode           = 0644,
 875                 .proc_handler   = ipv4_ping_group_range,
 876         },
 877         {
 878                 .procname       = "tcp_ecn",
 879                 .data           = &init_net.ipv4.sysctl_tcp_ecn,
 880                 .maxlen         = sizeof(int),
 881                 .mode           = 0644,
 882                 .proc_handler   = proc_dointvec
 883         },
 884         {
 885                 .procname       = "tcp_ecn_fallback",
 886                 .data           = &init_net.ipv4.sysctl_tcp_ecn_fallback,
 887                 .maxlen         = sizeof(int),
 888                 .mode           = 0644,
 889                 .proc_handler   = proc_dointvec
 890         },
 891         {
 892                 .procname       = "ip_local_port_range",
 893                 .maxlen         = sizeof(init_net.ipv4.ip_local_ports.range),
 894                 .data           = &init_net.ipv4.ip_local_ports.range,
 895                 .mode           = 0644,
 896                 .proc_handler   = ipv4_local_port_range,
 897         },
 898         {
 899                 .procname       = "ip_local_reserved_ports",
 900                 .data           = &init_net.ipv4.sysctl_local_reserved_ports,
 901                 .maxlen         = 65536,
 902                 .mode           = 0644,
 903                 .proc_handler   = proc_do_large_bitmap,
 904         },
 905         {
 906                 .procname       = "ip_no_pmtu_disc",
 907                 .data           = &init_net.ipv4.sysctl_ip_no_pmtu_disc,
 908                 .maxlen         = sizeof(int),
 909                 .mode           = 0644,
 910                 .proc_handler   = proc_dointvec
 911         },
 912         {
 913                 .procname       = "ip_forward_use_pmtu",
 914                 .data           = &init_net.ipv4.sysctl_ip_fwd_use_pmtu,
 915                 .maxlen         = sizeof(int),
 916                 .mode           = 0644,
 917                 .proc_handler   = proc_dointvec,
 918         },
 919         {
 920                 .procname       = "ip_nonlocal_bind",
 921                 .data           = &init_net.ipv4.sysctl_ip_nonlocal_bind,
 922                 .maxlen         = sizeof(int),
 923                 .mode           = 0644,
 924                 .proc_handler   = proc_dointvec
 925         },
 926         {
 927                 .procname       = "fwmark_reflect",
 928                 .data           = &init_net.ipv4.sysctl_fwmark_reflect,
 929                 .maxlen         = sizeof(int),
 930                 .mode           = 0644,
 931                 .proc_handler   = proc_dointvec,
 932         },
 933         {
 934                 .procname       = "tcp_fwmark_accept",
 935                 .data           = &init_net.ipv4.sysctl_tcp_fwmark_accept,
 936                 .maxlen         = sizeof(int),
 937                 .mode           = 0644,
 938                 .proc_handler   = proc_dointvec,
 939         },
 940         {
 941                 .procname       = "tcp_mtu_probing",
 942                 .data           = &init_net.ipv4.sysctl_tcp_mtu_probing,
 943                 .maxlen         = sizeof(int),
 944                 .mode           = 0644,
 945                 .proc_handler   = proc_dointvec,
 946         },
 947         {
 948                 .procname       = "tcp_base_mss",
 949                 .data           = &init_net.ipv4.sysctl_tcp_base_mss,
 950                 .maxlen         = sizeof(int),
 951                 .mode           = 0644,
 952                 .proc_handler   = proc_dointvec,
 953         },
 954         {
 955                 .procname       = "tcp_probe_threshold",
 956                 .data           = &init_net.ipv4.sysctl_tcp_probe_threshold,
 957                 .maxlen         = sizeof(int),
 958                 .mode           = 0644,
 959                 .proc_handler   = proc_dointvec,
 960         },
 961         {
 962                 .procname       = "tcp_probe_interval",
 963                 .data           = &init_net.ipv4.sysctl_tcp_probe_interval,
 964                 .maxlen         = sizeof(int),
 965                 .mode           = 0644,
 966                 .proc_handler   = proc_dointvec,
 967         },
 968         {
 969                 .procname       = "igmp_link_local_mcast_reports",
 970                 .data           = &sysctl_igmp_llm_reports,
 971                 .maxlen         = sizeof(int),
 972                 .mode           = 0644,
 973                 .proc_handler   = proc_dointvec
 974         },
 975         { }
 976 };
 977
 978 static __net_init int ipv4_sysctl_init_net(struct net *net)
 979 {
 980         struct ctl_table *table;
 981
 982         table = ipv4_net_table;
 983         if (!net_eq(net, &init_net)) {
 984                 int i;
 985
 986                 table = kmemdup(table, sizeof(ipv4_net_table), GFP_KERNEL);
 987                 if (!table)
 988                         goto err_alloc;
 989
 990                 /* Update the variables to point into the current struct net */
 991                 for (i = 0; i < ARRAY_SIZE(ipv4_net_table) - 1; i++)
 992                         table[i].data += (void *)net - (void *)&init_net;
 993         }
 994
 995         net->ipv4.ipv4_hdr = register_net_sysctl(net, "net/ipv4", table);
 996         if (!net->ipv4.ipv4_hdr)
 997                 goto err_reg;
 998
 999         net->ipv4.sysctl_local_reserved_ports = kzalloc(65536 / 8, GFP_KERNEL);
1000         if (!net->ipv4.sysctl_local_reserved_ports)
1001                 goto err_ports;
1002
1003         return 0;
1004
1005 err_ports:
1006         unregister_net_sysctl_table(net->ipv4.ipv4_hdr);
1007 err_reg:
1008         if (!net_eq(net, &init_net))
1009                 kfree(table);
1010 err_alloc:
1011         return -ENOMEM;
1012 }
1013
1014 static __net_exit void ipv4_sysctl_exit_net(struct net *net)
1015 {
1016         struct ctl_table *table;
1017
1018         kfree(net->ipv4.sysctl_local_reserved_ports);
1019         table = net->ipv4.ipv4_hdr->ctl_table_arg;
1020         unregister_net_sysctl_table(net->ipv4.ipv4_hdr);
1021         kfree(table);
1022 }
1023
1024 static __net_initdata struct pernet_operations ipv4_sysctl_ops = {
1025         .init = ipv4_sysctl_init_net,
1026         .exit = ipv4_sysctl_exit_net,
1027 };
1028
1029 static __init int sysctl_ipv4_init(void)
1030 {
1031         struct ctl_table_header *hdr;
1032
1033         hdr = register_net_sysctl(&init_net, "net/ipv4", ipv4_table);
1034         if (!hdr)
1035                 return -ENOMEM;
1036
1037         if (register_pernet_subsys(&ipv4_sysctl_ops)) {
1038                 unregister_net_sysctl_table(hdr);
1039                 return -ENOMEM;
1040         }
1041
1042         return 0;
1043 }
1044
1045 __initcall(sysctl_ipv4_init);