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 tcp_retr1_max = 255;
  33 static int ip_local_port_range_min[] = { 1, 1 };
  34 static int ip_local_port_range_max[] = { 65535, 65535 };
  35 static int tcp_adv_win_scale_min = -31;
  36 static int tcp_adv_win_scale_max = 31;
  37 static int ip_ttl_min = 1;
  38 static int ip_ttl_max = 255;
  39 static int tcp_syn_retries_min = 1;
  40 static int tcp_syn_retries_max = MAX_TCP_SYNCNT;
  41 static int ip_ping_group_range_min[] = { 0, 0 };
  42 static int ip_ping_group_range_max[] = { GID_T_MAX, GID_T_MAX };
  43
  44 /* Update system visible IP port range */
  45 static void set_local_port_range(int range[2])
  46 {
  47         write_seqlock(&sysctl_local_ports.lock);
  48         sysctl_local_ports.range[0] = range[0];
  49         sysctl_local_ports.range[1] = range[1];
  50         write_sequnlock(&sysctl_local_ports.lock);
  51 }
  52
  53 /* Validate changes from /proc interface. */
  54 static int ipv4_local_port_range(ctl_table *table, int write,
  55                                  void __user *buffer,
  56                                  size_t *lenp, loff_t *ppos)
  57 {
  58         int ret;
  59         int range[2];
  60         ctl_table tmp = {
  61                 .data = &range,
  62                 .maxlen = sizeof(range),
  63                 .mode = table->mode,
  64                 .extra1 = &ip_local_port_range_min,
  65                 .extra2 = &ip_local_port_range_max,
  66         };
  67
  68         inet_get_local_port_range(range, range + 1);
  69         ret = proc_dointvec_minmax(&tmp, write, buffer, lenp, ppos);
  70
  71         if (write && ret == 0) {
  72                 if (range[1] < range[0])
  73                         ret = -EINVAL;
  74                 else
  75                         set_local_port_range(range);
  76         }
  77
  78         return ret;
  79 }
  80
  81
  82 static void inet_get_ping_group_range_table(struct ctl_table *table, kgid_t *low, kgid_t *high)
  83 {
  84         kgid_t *data = table->data;
  85         unsigned int seq;
  86         do {
  87                 seq = read_seqbegin(&sysctl_local_ports.lock);
  88
  89                 *low = data[0];
  90                 *high = data[1];
  91         } while (read_seqretry(&sysctl_local_ports.lock, seq));
  92 }
  93
  94 /* Update system visible IP port range */
  95 static void set_ping_group_range(struct ctl_table *table, kgid_t low, kgid_t high)
  96 {
  97         kgid_t *data = table->data;
  98         write_seqlock(&sysctl_local_ports.lock);
  99         data[0] = low;
 100         data[1] = high;
 101         write_sequnlock(&sysctl_local_ports.lock);
 102 }
 103
 104 /* Validate changes from /proc interface. */
 105 static int ipv4_ping_group_range(ctl_table *table, int write,
 106                                  void __user *buffer,
 107                                  size_t *lenp, loff_t *ppos)
 108 {
 109         struct user_namespace *user_ns = current_user_ns();
 110         int ret;
 111         gid_t urange[2];
 112         kgid_t low, high;
 113         ctl_table tmp = {
 114                 .data = &urange,
 115                 .maxlen = sizeof(urange),
 116                 .mode = table->mode,
 117                 .extra1 = &ip_ping_group_range_min,
 118                 .extra2 = &ip_ping_group_range_max,
 119         };
 120
 121         inet_get_ping_group_range_table(table, &low, &high);
 122         urange[0] = from_kgid_munged(user_ns, low);
 123         urange[1] = from_kgid_munged(user_ns, high);
 124         ret = proc_dointvec_minmax(&tmp, write, buffer, lenp, ppos);
 125
 126         if (write && ret == 0) {
 127                 low = make_kgid(user_ns, urange[0]);
 128                 high = make_kgid(user_ns, urange[1]);
 129                 if (!gid_valid(low) || !gid_valid(high) ||
 130                     (urange[1] < urange[0]) || gid_lt(high, low)) {
 131                         low = make_kgid(&init_user_ns, 1);
 132                         high = make_kgid(&init_user_ns, 0);
 133                 }
 134                 set_ping_group_range(table, low, high);
 135         }
 136
 137         return ret;
 138 }
 139
 140 static int proc_tcp_congestion_control(ctl_table *ctl, int write,
 141                                        void __user *buffer, size_t *lenp, loff_t *ppos)
 142 {
 143         char val[TCP_CA_NAME_MAX];
 144         ctl_table tbl = {
 145                 .data = val,
 146                 .maxlen = TCP_CA_NAME_MAX,
 147         };
 148         int ret;
 149
 150         tcp_get_default_congestion_control(val);
 151
 152         ret = proc_dostring(&tbl, write, buffer, lenp, ppos);
 153         if (write && ret == 0)
 154                 ret = tcp_set_default_congestion_control(val);
 155         return ret;
 156 }
 157
 158 static int proc_tcp_available_congestion_control(ctl_table *ctl,
 159                                                  int write,
 160                                                  void __user *buffer, size_t *lenp,
 161                                                  loff_t *ppos)
 162 {
 163         ctl_table tbl = { .maxlen = TCP_CA_BUF_MAX, };
 164         int ret;
 165
 166         tbl.data = kmalloc(tbl.maxlen, GFP_USER);
 167         if (!tbl.data)
 168                 return -ENOMEM;
 169         tcp_get_available_congestion_control(tbl.data, TCP_CA_BUF_MAX);
 170         ret = proc_dostring(&tbl, write, buffer, lenp, ppos);
 171         kfree(tbl.data);
 172         return ret;
 173 }
 174
 175 static int proc_allowed_congestion_control(ctl_table *ctl,
 176                                            int write,
 177                                            void __user *buffer, size_t *lenp,
 178                                            loff_t *ppos)
 179 {
 180         ctl_table tbl = { .maxlen = TCP_CA_BUF_MAX };
 181         int ret;
 182
 183         tbl.data = kmalloc(tbl.maxlen, GFP_USER);
 184         if (!tbl.data)
 185                 return -ENOMEM;
 186
 187         tcp_get_allowed_congestion_control(tbl.data, tbl.maxlen);
 188         ret = proc_dostring(&tbl, write, buffer, lenp, ppos);
 189         if (write && ret == 0)
 190                 ret = tcp_set_allowed_congestion_control(tbl.data);
 191         kfree(tbl.data);
 192         return ret;
 193 }
 194
 195 static int ipv4_tcp_mem(ctl_table *ctl, int write,
 196                            void __user *buffer, size_t *lenp,
 197                            loff_t *ppos)
 198 {
 199         int ret;
 200         unsigned long vec[3];
 201         struct net *net = current->nsproxy->net_ns;
 202 #ifdef CONFIG_MEMCG_KMEM
 203         struct mem_cgroup *memcg;
 204 #endif
 205
 206         ctl_table tmp = {
 207                 .data = &vec,
 208                 .maxlen = sizeof(vec),
 209                 .mode = ctl->mode,
 210         };
 211
 212         if (!write) {
 213                 ctl->data = &net->ipv4.sysctl_tcp_mem;
 214                 return proc_doulongvec_minmax(ctl, write, buffer, lenp, ppos);
 215         }
 216
 217         ret = proc_doulongvec_minmax(&tmp, write, buffer, lenp, ppos);
 218         if (ret)
 219                 return ret;
 220
 221 #ifdef CONFIG_MEMCG_KMEM
 222         rcu_read_lock();
 223         memcg = mem_cgroup_from_task(current);
 224
 225         tcp_prot_mem(memcg, vec[0], 0);
 226         tcp_prot_mem(memcg, vec[1], 1);
 227         tcp_prot_mem(memcg, vec[2], 2);
 228         rcu_read_unlock();
 229 #endif
 230
 231         net->ipv4.sysctl_tcp_mem[0] = vec[0];
 232         net->ipv4.sysctl_tcp_mem[1] = vec[1];
 233         net->ipv4.sysctl_tcp_mem[2] = vec[2];
 234
 235         return 0;
 236 }
 237
 238 static int proc_tcp_fastopen_key(ctl_table *ctl, int write, void __user *buffer,
 239                                  size_t *lenp, loff_t *ppos)
 240 {
 241         ctl_table tbl = { .maxlen = (TCP_FASTOPEN_KEY_LENGTH * 2 + 10) };
 242         struct tcp_fastopen_context *ctxt;
 243         int ret;
 244         u32  user_key[4]; /* 16 bytes, matching TCP_FASTOPEN_KEY_LENGTH */
 245
 246         tbl.data = kmalloc(tbl.maxlen, GFP_KERNEL);
 247         if (!tbl.data)
 248                 return -ENOMEM;
 249
 250         rcu_read_lock();
 251         ctxt = rcu_dereference(tcp_fastopen_ctx);
 252         if (ctxt)
 253                 memcpy(user_key, ctxt->key, TCP_FASTOPEN_KEY_LENGTH);
 254         else
 255                 memset(user_key, 0, sizeof(user_key));
 256         rcu_read_unlock();
 257
 258         snprintf(tbl.data, tbl.maxlen, "%08x-%08x-%08x-%08x",
 259                 user_key[0], user_key[1], user_key[2], user_key[3]);
 260         ret = proc_dostring(&tbl, write, buffer, lenp, ppos);
 261
 262         if (write && ret == 0) {
 263                 if (sscanf(tbl.data, "%x-%x-%x-%x", user_key, user_key + 1,
 264                            user_key + 2, user_key + 3) != 4) {
 265                         ret = -EINVAL;
 266                         goto bad_key;
 267                 }
 268                 tcp_fastopen_reset_cipher(user_key, TCP_FASTOPEN_KEY_LENGTH);
 269         }
 270
 271 bad_key:
 272         pr_debug("proc FO key set 0x%x-%x-%x-%x <- 0x%s: %u\n",
 273                user_key[0], user_key[1], user_key[2], user_key[3],
 274                (char *)tbl.data, ret);
 275         kfree(tbl.data);
 276         return ret;
 277 }
 278
 279 static struct ctl_table ipv4_table[] = {
 280         {
 281                 .procname       = "tcp_timestamps",
 282                 .data           = &sysctl_tcp_timestamps,
 283                 .maxlen         = sizeof(int),
 284                 .mode           = 0644,
 285                 .proc_handler   = proc_dointvec
 286         },
 287         {
 288                 .procname       = "tcp_window_scaling",
 289                 .data           = &sysctl_tcp_window_scaling,
 290                 .maxlen         = sizeof(int),
 291                 .mode           = 0644,
 292                 .proc_handler   = proc_dointvec
 293         },
 294         {
 295                 .procname       = "tcp_sack",
 296                 .data           = &sysctl_tcp_sack,
 297                 .maxlen         = sizeof(int),
 298                 .mode           = 0644,
 299                 .proc_handler   = proc_dointvec
 300         },
 301         {
 302                 .procname       = "tcp_retrans_collapse",
 303                 .data           = &sysctl_tcp_retrans_collapse,
 304                 .maxlen         = sizeof(int),
 305                 .mode           = 0644,
 306                 .proc_handler   = proc_dointvec
 307         },
 308         {
 309                 .procname       = "ip_default_ttl",
 310                 .data           = &sysctl_ip_default_ttl,
 311                 .maxlen         = sizeof(int),
 312                 .mode           = 0644,
 313                 .proc_handler   = proc_dointvec_minmax,
 314                 .extra1         = &ip_ttl_min,
 315                 .extra2         = &ip_ttl_max,
 316         },
 317         {
 318                 .procname       = "ip_no_pmtu_disc",
 319                 .data           = &ipv4_config.no_pmtu_disc,
 320                 .maxlen         = sizeof(int),
 321                 .mode           = 0644,
 322                 .proc_handler   = proc_dointvec
 323         },
 324         {
 325                 .procname       = "ip_nonlocal_bind",
 326                 .data           = &sysctl_ip_nonlocal_bind,
 327                 .maxlen         = sizeof(int),
 328                 .mode           = 0644,
 329                 .proc_handler   = proc_dointvec
 330         },
 331         {
 332                 .procname       = "tcp_syn_retries",
 333                 .data           = &sysctl_tcp_syn_retries,
 334                 .maxlen         = sizeof(int),
 335                 .mode           = 0644,
 336                 .proc_handler   = proc_dointvec_minmax,
 337                 .extra1         = &tcp_syn_retries_min,
 338                 .extra2         = &tcp_syn_retries_max
 339         },
 340         {
 341                 .procname       = "tcp_synack_retries",
 342                 .data           = &sysctl_tcp_synack_retries,
 343                 .maxlen         = sizeof(int),
 344                 .mode           = 0644,
 345                 .proc_handler   = proc_dointvec
 346         },
 347         {
 348                 .procname       = "tcp_max_orphans",
 349                 .data           = &sysctl_tcp_max_orphans,
 350                 .maxlen         = sizeof(int),
 351                 .mode           = 0644,
 352                 .proc_handler   = proc_dointvec
 353         },
 354         {
 355                 .procname       = "tcp_max_tw_buckets",
 356                 .data           = &tcp_death_row.sysctl_max_tw_buckets,
 357                 .maxlen         = sizeof(int),
 358                 .mode           = 0644,
 359                 .proc_handler   = proc_dointvec
 360         },
 361         {
 362                 .procname       = "ip_early_demux",
 363                 .data           = &sysctl_ip_early_demux,
 364                 .maxlen         = sizeof(int),
 365                 .mode           = 0644,
 366                 .proc_handler   = proc_dointvec
 367         },
 368         {
 369                 .procname       = "ip_dynaddr",
 370                 .data           = &sysctl_ip_dynaddr,
 371                 .maxlen         = sizeof(int),
 372                 .mode           = 0644,
 373                 .proc_handler   = proc_dointvec
 374         },
 375         {
 376                 .procname       = "tcp_keepalive_time",
 377                 .data           = &sysctl_tcp_keepalive_time,
 378                 .maxlen         = sizeof(int),
 379                 .mode           = 0644,
 380                 .proc_handler   = proc_dointvec_jiffies,
 381         },
 382         {
 383                 .procname       = "tcp_keepalive_probes",
 384                 .data           = &sysctl_tcp_keepalive_probes,
 385                 .maxlen         = sizeof(int),
 386                 .mode           = 0644,
 387                 .proc_handler   = proc_dointvec
 388         },
 389         {
 390                 .procname       = "tcp_keepalive_intvl",
 391                 .data           = &sysctl_tcp_keepalive_intvl,
 392                 .maxlen         = sizeof(int),
 393                 .mode           = 0644,
 394                 .proc_handler   = proc_dointvec_jiffies,
 395         },
 396         {
 397                 .procname       = "tcp_retries1",
 398                 .data           = &sysctl_tcp_retries1,
 399                 .maxlen         = sizeof(int),
 400                 .mode           = 0644,
 401                 .proc_handler   = proc_dointvec_minmax,
 402                 .extra2         = &tcp_retr1_max
 403         },
 404         {
 405                 .procname       = "tcp_retries2",
 406                 .data           = &sysctl_tcp_retries2,
 407                 .maxlen         = sizeof(int),
 408                 .mode           = 0644,
 409                 .proc_handler   = proc_dointvec
 410         },
 411         {
 412                 .procname       = "tcp_fin_timeout",
 413                 .data           = &sysctl_tcp_fin_timeout,
 414                 .maxlen         = sizeof(int),
 415                 .mode           = 0644,
 416                 .proc_handler   = proc_dointvec_jiffies,
 417         },
 418 #ifdef CONFIG_SYN_COOKIES
 419         {
 420                 .procname       = "tcp_syncookies",
 421                 .data           = &sysctl_tcp_syncookies,
 422                 .maxlen         = sizeof(int),
 423                 .mode           = 0644,
 424                 .proc_handler   = proc_dointvec
 425         },
 426 #endif
 427         {
 428                 .procname       = "tcp_fastopen",
 429                 .data           = &sysctl_tcp_fastopen,
 430                 .maxlen         = sizeof(int),
 431                 .mode           = 0644,
 432                 .proc_handler   = proc_dointvec,
 433         },
 434         {
 435                 .procname       = "tcp_fastopen_key",
 436                 .mode           = 0600,
 437                 .maxlen         = ((TCP_FASTOPEN_KEY_LENGTH * 2) + 10),
 438                 .proc_handler   = proc_tcp_fastopen_key,
 439         },
 440         {
 441                 .procname       = "tcp_tw_recycle",
 442                 .data           = &tcp_death_row.sysctl_tw_recycle,
 443                 .maxlen         = sizeof(int),
 444                 .mode           = 0644,
 445                 .proc_handler   = proc_dointvec
 446         },
 447         {
 448                 .procname       = "tcp_abort_on_overflow",
 449                 .data           = &sysctl_tcp_abort_on_overflow,
 450                 .maxlen         = sizeof(int),
 451                 .mode           = 0644,
 452                 .proc_handler   = proc_dointvec
 453         },
 454         {
 455                 .procname       = "tcp_stdurg",
 456                 .data           = &sysctl_tcp_stdurg,
 457                 .maxlen         = sizeof(int),
 458                 .mode           = 0644,
 459                 .proc_handler   = proc_dointvec
 460         },
 461         {
 462                 .procname       = "tcp_rfc1337",
 463                 .data           = &sysctl_tcp_rfc1337,
 464                 .maxlen         = sizeof(int),
 465                 .mode           = 0644,
 466                 .proc_handler   = proc_dointvec
 467         },
 468         {
 469                 .procname       = "tcp_max_syn_backlog",
 470                 .data           = &sysctl_max_syn_backlog,
 471                 .maxlen         = sizeof(int),
 472                 .mode           = 0644,
 473                 .proc_handler   = proc_dointvec
 474         },
 475         {
 476                 .procname       = "ip_local_port_range",
 477                 .data           = &sysctl_local_ports.range,
 478                 .maxlen         = sizeof(sysctl_local_ports.range),
 479                 .mode           = 0644,
 480                 .proc_handler   = ipv4_local_port_range,
 481         },
 482         {
 483                 .procname       = "ip_local_reserved_ports",
 484                 .data           = NULL, /* initialized in sysctl_ipv4_init */
 485                 .maxlen         = 65536,
 486                 .mode           = 0644,
 487                 .proc_handler   = proc_do_large_bitmap,
 488         },
 489         {
 490                 .procname       = "igmp_max_memberships",
 491                 .data           = &sysctl_igmp_max_memberships,
 492                 .maxlen         = sizeof(int),
 493                 .mode           = 0644,
 494                 .proc_handler   = proc_dointvec
 495         },
 496         {
 497                 .procname       = "igmp_max_msf",
 498                 .data           = &sysctl_igmp_max_msf,
 499                 .maxlen         = sizeof(int),
 500                 .mode           = 0644,
 501                 .proc_handler   = proc_dointvec
 502         },
 503         {
 504                 .procname       = "inet_peer_threshold",
 505                 .data           = &inet_peer_threshold,
 506                 .maxlen         = sizeof(int),
 507                 .mode           = 0644,
 508                 .proc_handler   = proc_dointvec
 509         },
 510         {
 511                 .procname       = "inet_peer_minttl",
 512                 .data           = &inet_peer_minttl,
 513                 .maxlen         = sizeof(int),
 514                 .mode           = 0644,
 515                 .proc_handler   = proc_dointvec_jiffies,
 516         },
 517         {
 518                 .procname       = "inet_peer_maxttl",
 519                 .data           = &inet_peer_maxttl,
 520                 .maxlen         = sizeof(int),
 521                 .mode           = 0644,
 522                 .proc_handler   = proc_dointvec_jiffies,
 523         },
 524         {
 525                 .procname       = "tcp_orphan_retries",
 526                 .data           = &sysctl_tcp_orphan_retries,
 527                 .maxlen         = sizeof(int),
 528                 .mode           = 0644,
 529                 .proc_handler   = proc_dointvec
 530         },
 531         {
 532                 .procname       = "tcp_fack",
 533                 .data           = &sysctl_tcp_fack,
 534                 .maxlen         = sizeof(int),
 535                 .mode           = 0644,
 536                 .proc_handler   = proc_dointvec
 537         },
 538         {
 539                 .procname       = "tcp_reordering",
 540                 .data           = &sysctl_tcp_reordering,
 541                 .maxlen         = sizeof(int),
 542                 .mode           = 0644,
 543                 .proc_handler   = proc_dointvec
 544         },
 545         {
 546                 .procname       = "tcp_dsack",
 547                 .data           = &sysctl_tcp_dsack,
 548                 .maxlen         = sizeof(int),
 549                 .mode           = 0644,
 550                 .proc_handler   = proc_dointvec
 551         },
 552         {
 553                 .procname       = "tcp_wmem",
 554                 .data           = &sysctl_tcp_wmem,
 555                 .maxlen         = sizeof(sysctl_tcp_wmem),
 556                 .mode           = 0644,
 557                 .proc_handler   = proc_dointvec_minmax,
 558                 .extra1         = &one,
 559         },
 560         {
 561                 .procname       = "tcp_rmem",
 562                 .data           = &sysctl_tcp_rmem,
 563                 .maxlen         = sizeof(sysctl_tcp_rmem),
 564                 .mode           = 0644,
 565                 .proc_handler   = proc_dointvec_minmax,
 566                 .extra1         = &one,
 567         },
 568         {
 569                 .procname       = "tcp_app_win",
 570                 .data           = &sysctl_tcp_app_win,
 571                 .maxlen         = sizeof(int),
 572                 .mode           = 0644,
 573                 .proc_handler   = proc_dointvec
 574         },
 575         {
 576                 .procname       = "tcp_adv_win_scale",
 577                 .data           = &sysctl_tcp_adv_win_scale,
 578                 .maxlen         = sizeof(int),
 579                 .mode           = 0644,
 580                 .proc_handler   = proc_dointvec_minmax,
 581                 .extra1         = &tcp_adv_win_scale_min,
 582                 .extra2         = &tcp_adv_win_scale_max,
 583         },
 584         {
 585                 .procname       = "tcp_tw_reuse",
 586                 .data           = &sysctl_tcp_tw_reuse,
 587                 .maxlen         = sizeof(int),
 588                 .mode           = 0644,
 589                 .proc_handler   = proc_dointvec
 590         },
 591         {
 592                 .procname       = "tcp_frto",
 593                 .data           = &sysctl_tcp_frto,
 594                 .maxlen         = sizeof(int),
 595                 .mode           = 0644,
 596                 .proc_handler   = proc_dointvec
 597         },
 598         {
 599                 .procname       = "tcp_low_latency",
 600                 .data           = &sysctl_tcp_low_latency,
 601                 .maxlen         = sizeof(int),
 602                 .mode           = 0644,
 603                 .proc_handler   = proc_dointvec
 604         },
 605         {
 606                 .procname       = "tcp_no_metrics_save",
 607                 .data           = &sysctl_tcp_nometrics_save,
 608                 .maxlen         = sizeof(int),
 609                 .mode           = 0644,
 610                 .proc_handler   = proc_dointvec,
 611         },
 612         {
 613                 .procname       = "tcp_moderate_rcvbuf",
 614                 .data           = &sysctl_tcp_moderate_rcvbuf,
 615                 .maxlen         = sizeof(int),
 616                 .mode           = 0644,
 617                 .proc_handler   = proc_dointvec,
 618         },
 619         {
 620                 .procname       = "tcp_tso_win_divisor",
 621                 .data           = &sysctl_tcp_tso_win_divisor,
 622                 .maxlen         = sizeof(int),
 623                 .mode           = 0644,
 624                 .proc_handler   = proc_dointvec,
 625         },
 626         {
 627                 .procname       = "tcp_congestion_control",
 628                 .mode           = 0644,
 629                 .maxlen         = TCP_CA_NAME_MAX,
 630                 .proc_handler   = proc_tcp_congestion_control,
 631         },
 632         {
 633                 .procname       = "tcp_mtu_probing",
 634                 .data           = &sysctl_tcp_mtu_probing,
 635                 .maxlen         = sizeof(int),
 636                 .mode           = 0644,
 637                 .proc_handler   = proc_dointvec,
 638         },
 639         {
 640                 .procname       = "tcp_base_mss",
 641                 .data           = &sysctl_tcp_base_mss,
 642                 .maxlen         = sizeof(int),
 643                 .mode           = 0644,
 644                 .proc_handler   = proc_dointvec,
 645         },
 646         {
 647                 .procname       = "tcp_workaround_signed_windows",
 648                 .data           = &sysctl_tcp_workaround_signed_windows,
 649                 .maxlen         = sizeof(int),
 650                 .mode           = 0644,
 651                 .proc_handler   = proc_dointvec
 652         },
 653         {
 654                 .procname       = "tcp_limit_output_bytes",
 655                 .data           = &sysctl_tcp_limit_output_bytes,
 656                 .maxlen         = sizeof(int),
 657                 .mode           = 0644,
 658                 .proc_handler   = proc_dointvec
 659         },
 660         {
 661                 .procname       = "tcp_challenge_ack_limit",
 662                 .data           = &sysctl_tcp_challenge_ack_limit,
 663                 .maxlen         = sizeof(int),
 664                 .mode           = 0644,
 665                 .proc_handler   = proc_dointvec
 666         },
 667 #ifdef CONFIG_NET_DMA
 668         {
 669                 .procname       = "tcp_dma_copybreak",
 670                 .data           = &sysctl_tcp_dma_copybreak,
 671                 .maxlen         = sizeof(int),
 672                 .mode           = 0644,
 673                 .proc_handler   = proc_dointvec
 674         },
 675 #endif
 676         {
 677                 .procname       = "tcp_slow_start_after_idle",
 678                 .data           = &sysctl_tcp_slow_start_after_idle,
 679                 .maxlen         = sizeof(int),
 680                 .mode           = 0644,
 681                 .proc_handler   = proc_dointvec
 682         },
 683 #ifdef CONFIG_NETLABEL
 684         {
 685                 .procname       = "cipso_cache_enable",
 686                 .data           = &cipso_v4_cache_enabled,
 687                 .maxlen         = sizeof(int),
 688                 .mode           = 0644,
 689                 .proc_handler   = proc_dointvec,
 690         },
 691         {
 692                 .procname       = "cipso_cache_bucket_size",
 693                 .data           = &cipso_v4_cache_bucketsize,
 694                 .maxlen         = sizeof(int),
 695                 .mode           = 0644,
 696                 .proc_handler   = proc_dointvec,
 697         },
 698         {
 699                 .procname       = "cipso_rbm_optfmt",
 700                 .data           = &cipso_v4_rbm_optfmt,
 701                 .maxlen         = sizeof(int),
 702                 .mode           = 0644,
 703                 .proc_handler   = proc_dointvec,
 704         },
 705         {
 706                 .procname       = "cipso_rbm_strictvalid",
 707                 .data           = &cipso_v4_rbm_strictvalid,
 708                 .maxlen         = sizeof(int),
 709                 .mode           = 0644,
 710                 .proc_handler   = proc_dointvec,
 711         },
 712 #endif /* CONFIG_NETLABEL */
 713         {
 714                 .procname       = "tcp_available_congestion_control",
 715                 .maxlen         = TCP_CA_BUF_MAX,
 716                 .mode           = 0444,
 717                 .proc_handler   = proc_tcp_available_congestion_control,
 718         },
 719         {
 720                 .procname       = "tcp_allowed_congestion_control",
 721                 .maxlen         = TCP_CA_BUF_MAX,
 722                 .mode           = 0644,
 723                 .proc_handler   = proc_allowed_congestion_control,
 724         },
 725         {
 726                 .procname       = "tcp_max_ssthresh",
 727                 .data           = &sysctl_tcp_max_ssthresh,
 728                 .maxlen         = sizeof(int),
 729                 .mode           = 0644,
 730                 .proc_handler   = proc_dointvec,
 731         },
 732         {
 733                 .procname       = "tcp_thin_linear_timeouts",
 734                 .data           = &sysctl_tcp_thin_linear_timeouts,
 735                 .maxlen         = sizeof(int),
 736                 .mode           = 0644,
 737                 .proc_handler   = proc_dointvec
 738         },
 739         {
 740                 .procname       = "tcp_thin_dupack",
 741                 .data           = &sysctl_tcp_thin_dupack,
 742                 .maxlen         = sizeof(int),
 743                 .mode           = 0644,
 744                 .proc_handler   = proc_dointvec
 745         },
 746         {
 747                 .procname       = "tcp_early_retrans",
 748                 .data           = &sysctl_tcp_early_retrans,
 749                 .maxlen         = sizeof(int),
 750                 .mode           = 0644,
 751                 .proc_handler   = proc_dointvec_minmax,
 752                 .extra1         = &zero,
 753                 .extra2         = &four,
 754         },
 755         {
 756                 .procname       = "udp_mem",
 757                 .data           = &sysctl_udp_mem,
 758                 .maxlen         = sizeof(sysctl_udp_mem),
 759                 .mode           = 0644,
 760                 .proc_handler   = proc_doulongvec_minmax,
 761         },
 762         {
 763                 .procname       = "udp_rmem_min",
 764                 .data           = &sysctl_udp_rmem_min,
 765                 .maxlen         = sizeof(sysctl_udp_rmem_min),
 766                 .mode           = 0644,
 767                 .proc_handler   = proc_dointvec_minmax,
 768                 .extra1         = &one
 769         },
 770         {
 771                 .procname       = "udp_wmem_min",
 772                 .data           = &sysctl_udp_wmem_min,
 773                 .maxlen         = sizeof(sysctl_udp_wmem_min),
 774                 .mode           = 0644,
 775                 .proc_handler   = proc_dointvec_minmax,
 776                 .extra1         = &one
 777         },
 778         { }
 779 };
 780
 781 static struct ctl_table ipv4_net_table[] = {
 782         {
 783                 .procname       = "icmp_echo_ignore_all",
 784                 .data           = &init_net.ipv4.sysctl_icmp_echo_ignore_all,
 785                 .maxlen         = sizeof(int),
 786                 .mode           = 0644,
 787                 .proc_handler   = proc_dointvec
 788         },
 789         {
 790                 .procname       = "icmp_echo_ignore_broadcasts",
 791                 .data           = &init_net.ipv4.sysctl_icmp_echo_ignore_broadcasts,
 792                 .maxlen         = sizeof(int),
 793                 .mode           = 0644,
 794                 .proc_handler   = proc_dointvec
 795         },
 796         {
 797                 .procname       = "icmp_ignore_bogus_error_responses",
 798                 .data           = &init_net.ipv4.sysctl_icmp_ignore_bogus_error_responses,
 799                 .maxlen         = sizeof(int),
 800                 .mode           = 0644,
 801                 .proc_handler   = proc_dointvec
 802         },
 803         {
 804                 .procname       = "icmp_errors_use_inbound_ifaddr",
 805                 .data           = &init_net.ipv4.sysctl_icmp_errors_use_inbound_ifaddr,
 806                 .maxlen         = sizeof(int),
 807                 .mode           = 0644,
 808                 .proc_handler   = proc_dointvec
 809         },
 810         {
 811                 .procname       = "icmp_ratelimit",
 812                 .data           = &init_net.ipv4.sysctl_icmp_ratelimit,
 813                 .maxlen         = sizeof(int),
 814                 .mode           = 0644,
 815                 .proc_handler   = proc_dointvec_ms_jiffies,
 816         },
 817         {
 818                 .procname       = "icmp_ratemask",
 819                 .data           = &init_net.ipv4.sysctl_icmp_ratemask,
 820                 .maxlen         = sizeof(int),
 821                 .mode           = 0644,
 822                 .proc_handler   = proc_dointvec
 823         },
 824         {
 825                 .procname       = "ping_group_range",
 826                 .data           = &init_net.ipv4.sysctl_ping_group_range,
 827                 .maxlen         = sizeof(gid_t)*2,
 828                 .mode           = 0644,
 829                 .proc_handler   = ipv4_ping_group_range,
 830         },
 831         {
 832                 .procname       = "tcp_ecn",
 833                 .data           = &init_net.ipv4.sysctl_tcp_ecn,
 834                 .maxlen         = sizeof(int),
 835                 .mode           = 0644,
 836                 .proc_handler   = proc_dointvec
 837         },
 838         {
 839                 .procname       = "tcp_mem",
 840                 .maxlen         = sizeof(init_net.ipv4.sysctl_tcp_mem),
 841                 .mode           = 0644,
 842                 .proc_handler   = ipv4_tcp_mem,
 843         },
 844         { }
 845 };
 846
 847 static __net_init int ipv4_sysctl_init_net(struct net *net)
 848 {
 849         struct ctl_table *table;
 850
 851         table = ipv4_net_table;
 852         if (!net_eq(net, &init_net)) {
 853                 table = kmemdup(table, sizeof(ipv4_net_table), GFP_KERNEL);
 854                 if (table == NULL)
 855                         goto err_alloc;
 856
 857                 table[0].data =
 858                         &net->ipv4.sysctl_icmp_echo_ignore_all;
 859                 table[1].data =
 860                         &net->ipv4.sysctl_icmp_echo_ignore_broadcasts;
 861                 table[2].data =
 862                         &net->ipv4.sysctl_icmp_ignore_bogus_error_responses;
 863                 table[3].data =
 864                         &net->ipv4.sysctl_icmp_errors_use_inbound_ifaddr;
 865                 table[4].data =
 866                         &net->ipv4.sysctl_icmp_ratelimit;
 867                 table[5].data =
 868                         &net->ipv4.sysctl_icmp_ratemask;
 869                 table[6].data =
 870                         &net->ipv4.sysctl_ping_group_range;
 871                 table[7].data =
 872                         &net->ipv4.sysctl_tcp_ecn;
 873
 874                 /* Don't export sysctls to unprivileged users */
 875                 if (net->user_ns != &init_user_ns)
 876                         table[0].procname = NULL;
 877         }
 878
 879         /*
 880          * Sane defaults - nobody may create ping sockets.
 881          * Boot scripts should set this to distro-specific group.
 882          */
 883         net->ipv4.sysctl_ping_group_range[0] = make_kgid(&init_user_ns, 1);
 884         net->ipv4.sysctl_ping_group_range[1] = make_kgid(&init_user_ns, 0);
 885
 886         tcp_init_mem(net);
 887
 888         net->ipv4.ipv4_hdr = register_net_sysctl(net, "net/ipv4", table);
 889         if (net->ipv4.ipv4_hdr == NULL)
 890                 goto err_reg;
 891
 892         return 0;
 893
 894 err_reg:
 895         if (!net_eq(net, &init_net))
 896                 kfree(table);
 897 err_alloc:
 898         return -ENOMEM;
 899 }
 900
 901 static __net_exit void ipv4_sysctl_exit_net(struct net *net)
 902 {
 903         struct ctl_table *table;
 904
 905         table = net->ipv4.ipv4_hdr->ctl_table_arg;
 906         unregister_net_sysctl_table(net->ipv4.ipv4_hdr);
 907         kfree(table);
 908 }
 909
 910 static __net_initdata struct pernet_operations ipv4_sysctl_ops = {
 911         .init = ipv4_sysctl_init_net,
 912         .exit = ipv4_sysctl_exit_net,
 913 };
 914
 915 static __init int sysctl_ipv4_init(void)
 916 {
 917         struct ctl_table_header *hdr;
 918         struct ctl_table *i;
 919
 920         for (i = ipv4_table; i->procname; i++) {
 921                 if (strcmp(i->procname, "ip_local_reserved_ports") == 0) {
 922                         i->data = sysctl_local_reserved_ports;
 923                         break;
 924                 }
 925         }
 926         if (!i->procname)
 927                 return -EINVAL;
 928
 929         hdr = register_net_sysctl(&init_net, "net/ipv4", ipv4_table);
 930         if (hdr == NULL)
 931                 return -ENOMEM;
 932
 933         if (register_pernet_subsys(&ipv4_sysctl_ops)) {
 934                 unregister_net_sysctl_table(hdr);
 935                 return -ENOMEM;
 936         }
 937
 938         return 0;
 939 }
 940
 941 __initcall(sysctl_ipv4_init);