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