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
  28 static int zero;
  29 static int one = 1;
  30 static int four = 4;
  31 static int thousand = 1000;
  32 static int gso_max_segs = GSO_MAX_SEGS;
  33 static int tcp_retr1_max = 255;
  34 static int ip_local_port_range_min[] = { 1, 1 };
  35 static int ip_local_port_range_max[] = { 65535, 65535 };
  36 static int tcp_adv_win_scale_min = -31;
  37 static int tcp_adv_win_scale_max = 31;
  38 static int ip_ttl_min = 1;
  39 static int ip_ttl_max = 255;
  40 static int tcp_syn_retries_min = 1;
  41 static int tcp_syn_retries_max = MAX_TCP_SYNCNT;
  42 static int ip_ping_group_range_min[] = { 0, 0 };
  43 static int ip_ping_group_range_max[] = { GID_T_MAX, GID_T_MAX };
  44
  45 /* Update system visible IP port range */
  46 static void set_local_port_range(struct net *net, int range[2])
  47 {
  48         bool same_parity = !((range[0] ^ range[1]) & 1);
  49
  50         write_seqlock_bh(&net->ipv4.ip_local_ports.lock);
  51         if (same_parity && !net->ipv4.ip_local_ports.warned) {
  52                 net->ipv4.ip_local_ports.warned = true;
  53                 pr_err_ratelimited("ip_local_port_range: prefer different parity for start/end values.\n");
  54         }
  55         net->ipv4.ip_local_ports.range[0] = range[0];
  56         net->ipv4.ip_local_ports.range[1] = range[1];
  57         write_sequnlock_bh(&net->ipv4.ip_local_ports.lock);
  58 }
  59
  60 /* Validate changes from /proc interface. */
  61 static int ipv4_local_port_range(struct ctl_table *table, int write,
  62                                  void __user *buffer,
  63                                  size_t *lenp, loff_t *ppos)
  64 {
  65         struct net *net =
  66                 container_of(table->data, struct net, ipv4.ip_local_ports.range);
  67         int ret;
  68         int range[2];
  69         struct ctl_table tmp = {
  70                 .data = &range,
  71                 .maxlen = sizeof(range),
  72                 .mode = table->mode,
  73                 .extra1 = &ip_local_port_range_min,
  74                 .extra2 = &ip_local_port_range_max,
  75         };
  76
  77         inet_get_local_port_range(net, &range[0], &range[1]);
  78
  79         ret = proc_dointvec_minmax(&tmp, write, buffer, lenp, ppos);
  80
  81         if (write && ret == 0) {
  82                 if (range[1] < range[0])
  83                         ret = -EINVAL;
  84                 else
  85                         set_local_port_range(net, range);
  86         }
  87
  88         return ret;
  89 }
  90
  91
  92 static void inet_get_ping_group_range_table(struct ctl_table *table, kgid_t *low, kgid_t *high)
  93 {
  94         kgid_t *data = table->data;
  95         struct net *net =
  96                 container_of(table->data, struct net, ipv4.ping_group_range.range);
  97         unsigned int seq;
  98         do {
  99                 seq = read_seqbegin(&net->ipv4.ip_local_ports.lock);
 100
 101                 *low = data[0];
 102                 *high = data[1];
 103         } while (read_seqretry(&net->ipv4.ip_local_ports.lock, seq));
 104 }
 105
 106 /* Update system visible IP port range */
 107 static void set_ping_group_range(struct ctl_table *table, kgid_t low, kgid_t high)
 108 {
 109         kgid_t *data = table->data;
 110         struct net *net =
 111                 container_of(table->data, struct net, ipv4.ping_group_range.range);
 112         write_seqlock(&net->ipv4.ip_local_ports.lock);
 113         data[0] = low;
 114         data[1] = high;
 115         write_sequnlock(&net->ipv4.ip_local_ports.lock);
 116 }
 117
 118 /* Validate changes from /proc interface. */
 119 static int ipv4_ping_group_range(struct ctl_table *table, int write,
 120                                  void __user *buffer,
 121                                  size_t *lenp, loff_t *ppos)
 122 {
 123         struct user_namespace *user_ns = current_user_ns();
 124         int ret;
 125         gid_t urange[2];
 126         kgid_t low, high;
 127         struct ctl_table tmp = {
 128                 .data = &urange,
 129                 .maxlen = sizeof(urange),
 130                 .mode = table->mode,
 131                 .extra1 = &ip_ping_group_range_min,
 132                 .extra2 = &ip_ping_group_range_max,
 133         };
 134
 135         inet_get_ping_group_range_table(table, &low, &high);
 136         urange[0] = from_kgid_munged(user_ns, low);
 137         urange[1] = from_kgid_munged(user_ns, high);
 138         ret = proc_dointvec_minmax(&tmp, write, buffer, lenp, ppos);
 139
 140         if (write && ret == 0) {
 141                 low = make_kgid(user_ns, urange[0]);
 142                 high = make_kgid(user_ns, urange[1]);
 143                 if (!gid_valid(low) || !gid_valid(high) ||
 144                     (urange[1] < urange[0]) || gid_lt(high, low)) {
 145                         low = make_kgid(&init_user_ns, 1);
 146                         high = make_kgid(&init_user_ns, 0);
 147                 }
 148                 set_ping_group_range(table, low, high);
 149         }
 150
 151         return ret;
 152 }
 153
 154 static int proc_tcp_congestion_control(struct ctl_table *ctl, int write,
 155                                        void __user *buffer, size_t *lenp, loff_t *ppos)
 156 {
 157         char val[TCP_CA_NAME_MAX];
 158         struct ctl_table tbl = {
 159                 .data = val,
 160                 .maxlen = TCP_CA_NAME_MAX,
 161         };
 162         int ret;
 163
 164         tcp_get_default_congestion_control(val);
 165
 166         ret = proc_dostring(&tbl, write, buffer, lenp, ppos);
 167         if (write && ret == 0)
 168                 ret = tcp_set_default_congestion_control(val);
 169         return ret;
 170 }
 171
 172 static int proc_tcp_available_congestion_control(struct ctl_table *ctl,
 173                                                  int write,
 174                                                  void __user *buffer, size_t *lenp,
 175                                                  loff_t *ppos)
 176 {
 177         struct ctl_table tbl = { .maxlen = TCP_CA_BUF_MAX, };
 178         int ret;
 179
 180         tbl.data = kmalloc(tbl.maxlen, GFP_USER);
 181         if (!tbl.data)
 182                 return -ENOMEM;
 183         tcp_get_available_congestion_control(tbl.data, TCP_CA_BUF_MAX);
 184         ret = proc_dostring(&tbl, write, buffer, lenp, ppos);
 185         kfree(tbl.data);
 186         return ret;
 187 }
 188
 189 static int proc_allowed_congestion_control(struct ctl_table *ctl,
 190                                            int write,
 191                                            void __user *buffer, size_t *lenp,
 192                                            loff_t *ppos)
 193 {
 194         struct ctl_table tbl = { .maxlen = TCP_CA_BUF_MAX };
 195         int ret;
 196
 197         tbl.data = kmalloc(tbl.maxlen, GFP_USER);
 198         if (!tbl.data)
 199                 return -ENOMEM;
 200
 201         tcp_get_allowed_congestion_control(tbl.data, tbl.maxlen);
 202         ret = proc_dostring(&tbl, write, buffer, lenp, ppos);
 203         if (write && ret == 0)
 204                 ret = tcp_set_allowed_congestion_control(tbl.data);
 205         kfree(tbl.data);
 206         return ret;
 207 }
 208
 209 static int proc_tcp_fastopen_key(struct ctl_table *ctl, int write,
 210                                  void __user *buffer, size_t *lenp,
 211                                  loff_t *ppos)
 212 {
 213         struct ctl_table tbl = { .maxlen = (TCP_FASTOPEN_KEY_LENGTH * 2 + 10) };
 214         struct tcp_fastopen_context *ctxt;
 215         int ret;
 216         u32  user_key[4]; /* 16 bytes, matching TCP_FASTOPEN_KEY_LENGTH */
 217
 218         tbl.data = kmalloc(tbl.maxlen, GFP_KERNEL);
 219         if (!tbl.data)
 220                 return -ENOMEM;
 221
 222         rcu_read_lock();
 223         ctxt = rcu_dereference(tcp_fastopen_ctx);
 224         if (ctxt)
 225                 memcpy(user_key, ctxt->key, TCP_FASTOPEN_KEY_LENGTH);
 226         else
 227                 memset(user_key, 0, sizeof(user_key));
 228         rcu_read_unlock();
 229
 230         snprintf(tbl.data, tbl.maxlen, "%08x-%08x-%08x-%08x",
 231                 user_key[0], user_key[1], user_key[2], user_key[3]);
 232         ret = proc_dostring(&tbl, write, buffer, lenp, ppos);
 233
 234         if (write && ret == 0) {
 235                 if (sscanf(tbl.data, "%x-%x-%x-%x", user_key, user_key + 1,
 236                            user_key + 2, user_key + 3) != 4) {
 237                         ret = -EINVAL;
 238                         goto bad_key;
 239                 }
 240                 /* Generate a dummy secret but don't publish it. This
 241                  * is needed so we don't regenerate a new key on the
 242                  * first invocation of tcp_fastopen_cookie_gen
 243                  */
 244                 tcp_fastopen_init_key_once(false);
 245                 tcp_fastopen_reset_cipher(user_key, TCP_FASTOPEN_KEY_LENGTH);
 246         }
 247
 248 bad_key:
 249         pr_debug("proc FO key set 0x%x-%x-%x-%x <- 0x%s: %u\n",
 250                user_key[0], user_key[1], user_key[2], user_key[3],
 251                (char *)tbl.data, ret);
 252         kfree(tbl.data);
 253         return ret;
 254 }
 255
 256 static struct ctl_table ipv4_table[] = {
 257         {
 258                 .procname       = "tcp_timestamps",
 259                 .data           = &sysctl_tcp_timestamps,
 260                 .maxlen         = sizeof(int),
 261                 .mode           = 0644,
 262                 .proc_handler   = proc_dointvec
 263         },
 264         {
 265                 .procname       = "tcp_window_scaling",
 266                 .data           = &sysctl_tcp_window_scaling,
 267                 .maxlen         = sizeof(int),
 268                 .mode           = 0644,
 269                 .proc_handler   = proc_dointvec
 270         },
 271         {
 272                 .procname       = "tcp_sack",
 273                 .data           = &sysctl_tcp_sack,
 274                 .maxlen         = sizeof(int),
 275                 .mode           = 0644,
 276                 .proc_handler   = proc_dointvec
 277         },
 278         {
 279                 .procname       = "tcp_retrans_collapse",
 280                 .data           = &sysctl_tcp_retrans_collapse,
 281                 .maxlen         = sizeof(int),
 282                 .mode           = 0644,
 283                 .proc_handler   = proc_dointvec
 284         },
 285         {
 286                 .procname       = "ip_default_ttl",
 287                 .data           = &sysctl_ip_default_ttl,
 288                 .maxlen         = sizeof(int),
 289                 .mode           = 0644,
 290                 .proc_handler   = proc_dointvec_minmax,
 291                 .extra1         = &ip_ttl_min,
 292                 .extra2         = &ip_ttl_max,
 293         },
 294         {
 295                 .procname       = "tcp_syn_retries",
 296                 .data           = &sysctl_tcp_syn_retries,
 297                 .maxlen         = sizeof(int),
 298                 .mode           = 0644,
 299                 .proc_handler   = proc_dointvec_minmax,
 300                 .extra1         = &tcp_syn_retries_min,
 301                 .extra2         = &tcp_syn_retries_max
 302         },
 303         {
 304                 .procname       = "tcp_synack_retries",
 305                 .data           = &sysctl_tcp_synack_retries,
 306                 .maxlen         = sizeof(int),
 307                 .mode           = 0644,
 308                 .proc_handler   = proc_dointvec
 309         },
 310         {
 311                 .procname       = "tcp_max_orphans",
 312                 .data           = &sysctl_tcp_max_orphans,
 313                 .maxlen         = sizeof(int),
 314                 .mode           = 0644,
 315                 .proc_handler   = proc_dointvec
 316         },
 317         {
 318                 .procname       = "tcp_max_tw_buckets",
 319                 .data           = &tcp_death_row.sysctl_max_tw_buckets,
 320                 .maxlen         = sizeof(int),
 321                 .mode           = 0644,
 322                 .proc_handler   = proc_dointvec
 323         },
 324         {
 325                 .procname       = "ip_early_demux",
 326                 .data           = &sysctl_ip_early_demux,
 327                 .maxlen         = sizeof(int),
 328                 .mode           = 0644,
 329                 .proc_handler   = proc_dointvec
 330         },
 331         {
 332                 .procname       = "ip_dynaddr",
 333                 .data           = &sysctl_ip_dynaddr,
 334                 .maxlen         = sizeof(int),
 335                 .mode           = 0644,
 336                 .proc_handler   = proc_dointvec
 337         },
 338         {
 339                 .procname       = "tcp_retries1",
 340                 .data           = &sysctl_tcp_retries1,
 341                 .maxlen         = sizeof(int),
 342                 .mode           = 0644,
 343                 .proc_handler   = proc_dointvec_minmax,
 344                 .extra2         = &tcp_retr1_max
 345         },
 346         {
 347                 .procname       = "tcp_retries2",
 348                 .data           = &sysctl_tcp_retries2,
 349                 .maxlen         = sizeof(int),
 350                 .mode           = 0644,
 351                 .proc_handler   = proc_dointvec
 352         },
 353         {
 354                 .procname       = "tcp_fin_timeout",
 355                 .data           = &sysctl_tcp_fin_timeout,
 356                 .maxlen         = sizeof(int),
 357                 .mode           = 0644,
 358                 .proc_handler   = proc_dointvec_jiffies,
 359         },
 360 #ifdef CONFIG_SYN_COOKIES
 361         {
 362                 .procname       = "tcp_syncookies",
 363                 .data           = &sysctl_tcp_syncookies,
 364                 .maxlen         = sizeof(int),
 365                 .mode           = 0644,
 366                 .proc_handler   = proc_dointvec
 367         },
 368 #endif
 369         {
 370                 .procname       = "tcp_fastopen",
 371                 .data           = &sysctl_tcp_fastopen,
 372                 .maxlen         = sizeof(int),
 373                 .mode           = 0644,
 374                 .proc_handler   = proc_dointvec,
 375         },
 376         {
 377                 .procname       = "tcp_fastopen_key",
 378                 .mode           = 0600,
 379                 .maxlen         = ((TCP_FASTOPEN_KEY_LENGTH * 2) + 10),
 380                 .proc_handler   = proc_tcp_fastopen_key,
 381         },
 382         {
 383                 .procname       = "tcp_tw_recycle",
 384                 .data           = &tcp_death_row.sysctl_tw_recycle,
 385                 .maxlen         = sizeof(int),
 386                 .mode           = 0644,
 387                 .proc_handler   = proc_dointvec
 388         },
 389         {
 390                 .procname       = "tcp_abort_on_overflow",
 391                 .data           = &sysctl_tcp_abort_on_overflow,
 392                 .maxlen         = sizeof(int),
 393                 .mode           = 0644,
 394                 .proc_handler   = proc_dointvec
 395         },
 396         {
 397                 .procname       = "tcp_stdurg",
 398                 .data           = &sysctl_tcp_stdurg,
 399                 .maxlen         = sizeof(int),
 400                 .mode           = 0644,
 401                 .proc_handler   = proc_dointvec
 402         },
 403         {
 404                 .procname       = "tcp_rfc1337",
 405                 .data           = &sysctl_tcp_rfc1337,
 406                 .maxlen         = sizeof(int),
 407                 .mode           = 0644,
 408                 .proc_handler   = proc_dointvec
 409         },
 410         {
 411                 .procname       = "tcp_max_syn_backlog",
 412                 .data           = &sysctl_max_syn_backlog,
 413                 .maxlen         = sizeof(int),
 414                 .mode           = 0644,
 415                 .proc_handler   = proc_dointvec
 416         },
 417         {
 418                 .procname       = "igmp_max_memberships",
 419                 .data           = &sysctl_igmp_max_memberships,
 420                 .maxlen         = sizeof(int),
 421                 .mode           = 0644,
 422                 .proc_handler   = proc_dointvec
 423         },
 424         {
 425                 .procname       = "igmp_max_msf",
 426                 .data           = &sysctl_igmp_max_msf,
 427                 .maxlen         = sizeof(int),
 428                 .mode           = 0644,
 429                 .proc_handler   = proc_dointvec
 430         },
 431 #ifdef CONFIG_IP_MULTICAST
 432         {
 433                 .procname       = "igmp_qrv",
 434                 .data           = &sysctl_igmp_qrv,
 435                 .maxlen         = sizeof(int),
 436                 .mode           = 0644,
 437                 .proc_handler   = proc_dointvec_minmax,
 438                 .extra1         = &one
 439         },
 440 #endif
 441         {
 442                 .procname       = "inet_peer_threshold",
 443                 .data           = &inet_peer_threshold,
 444                 .maxlen         = sizeof(int),
 445                 .mode           = 0644,
 446                 .proc_handler   = proc_dointvec
 447         },
 448         {
 449                 .procname       = "inet_peer_minttl",
 450                 .data           = &inet_peer_minttl,
 451                 .maxlen         = sizeof(int),
 452                 .mode           = 0644,
 453                 .proc_handler   = proc_dointvec_jiffies,
 454         },
 455         {
 456                 .procname       = "inet_peer_maxttl",
 457                 .data           = &inet_peer_maxttl,
 458                 .maxlen         = sizeof(int),
 459                 .mode           = 0644,
 460                 .proc_handler   = proc_dointvec_jiffies,
 461         },
 462         {
 463                 .procname       = "tcp_orphan_retries",
 464                 .data           = &sysctl_tcp_orphan_retries,
 465                 .maxlen         = sizeof(int),
 466                 .mode           = 0644,
 467                 .proc_handler   = proc_dointvec
 468         },
 469         {
 470                 .procname       = "tcp_fack",
 471                 .data           = &sysctl_tcp_fack,
 472                 .maxlen         = sizeof(int),
 473                 .mode           = 0644,
 474                 .proc_handler   = proc_dointvec
 475         },
 476         {
 477                 .procname       = "tcp_recovery",
 478                 .data           = &sysctl_tcp_recovery,
 479                 .maxlen         = sizeof(int),
 480                 .mode           = 0644,
 481                 .proc_handler   = proc_dointvec,
 482         },
 483         {
 484                 .procname       = "tcp_reordering",
 485                 .data           = &sysctl_tcp_reordering,
 486                 .maxlen         = sizeof(int),
 487                 .mode           = 0644,
 488                 .proc_handler   = proc_dointvec
 489         },
 490         {
 491                 .procname       = "tcp_max_reordering",
 492                 .data           = &sysctl_tcp_max_reordering,
 493                 .maxlen         = sizeof(int),
 494                 .mode           = 0644,
 495                 .proc_handler   = proc_dointvec
 496         },
 497         {
 498                 .procname       = "tcp_dsack",
 499                 .data           = &sysctl_tcp_dsack,
 500                 .maxlen         = sizeof(int),
 501                 .mode           = 0644,
 502                 .proc_handler   = proc_dointvec
 503         },
 504         {
 505                 .procname       = "tcp_mem",
 506                 .maxlen         = sizeof(sysctl_tcp_mem),
 507                 .data           = &sysctl_tcp_mem,
 508                 .mode           = 0644,
 509                 .proc_handler   = proc_doulongvec_minmax,
 510         },
 511         {
 512                 .procname       = "tcp_wmem",
 513                 .data           = &sysctl_tcp_wmem,
 514                 .maxlen         = sizeof(sysctl_tcp_wmem),
 515                 .mode           = 0644,
 516                 .proc_handler   = proc_dointvec_minmax,
 517                 .extra1         = &one,
 518         },
 519         {
 520                 .procname       = "tcp_notsent_lowat",
 521                 .data           = &sysctl_tcp_notsent_lowat,
 522                 .maxlen         = sizeof(sysctl_tcp_notsent_lowat),
 523                 .mode           = 0644,
 524                 .proc_handler   = proc_dointvec,
 525         },
 526         {
 527                 .procname       = "tcp_rmem",
 528                 .data           = &sysctl_tcp_rmem,
 529                 .maxlen         = sizeof(sysctl_tcp_rmem),
 530                 .mode           = 0644,
 531                 .proc_handler   = proc_dointvec_minmax,
 532                 .extra1         = &one,
 533         },
 534         {
 535                 .procname       = "tcp_app_win",
 536                 .data           = &sysctl_tcp_app_win,
 537                 .maxlen         = sizeof(int),
 538                 .mode           = 0644,
 539                 .proc_handler   = proc_dointvec
 540         },
 541         {
 542                 .procname       = "tcp_adv_win_scale",
 543                 .data           = &sysctl_tcp_adv_win_scale,
 544                 .maxlen         = sizeof(int),
 545                 .mode           = 0644,
 546                 .proc_handler   = proc_dointvec_minmax,
 547                 .extra1         = &tcp_adv_win_scale_min,
 548                 .extra2         = &tcp_adv_win_scale_max,
 549         },
 550         {
 551                 .procname       = "tcp_tw_reuse",
 552                 .data           = &sysctl_tcp_tw_reuse,
 553                 .maxlen         = sizeof(int),
 554                 .mode           = 0644,
 555                 .proc_handler   = proc_dointvec
 556         },
 557         {
 558                 .procname       = "tcp_frto",
 559                 .data           = &sysctl_tcp_frto,
 560                 .maxlen         = sizeof(int),
 561                 .mode           = 0644,
 562                 .proc_handler   = proc_dointvec
 563         },
 564         {
 565                 .procname       = "tcp_min_rtt_wlen",
 566                 .data           = &sysctl_tcp_min_rtt_wlen,
 567                 .maxlen         = sizeof(int),
 568                 .mode           = 0644,
 569                 .proc_handler   = proc_dointvec
 570         },
 571         {
 572                 .procname       = "tcp_low_latency",
 573                 .data           = &sysctl_tcp_low_latency,
 574                 .maxlen         = sizeof(int),
 575                 .mode           = 0644,
 576                 .proc_handler   = proc_dointvec
 577         },
 578         {
 579                 .procname       = "tcp_no_metrics_save",
 580                 .data           = &sysctl_tcp_nometrics_save,
 581                 .maxlen         = sizeof(int),
 582                 .mode           = 0644,
 583                 .proc_handler   = proc_dointvec,
 584         },
 585         {
 586                 .procname       = "tcp_moderate_rcvbuf",
 587                 .data           = &sysctl_tcp_moderate_rcvbuf,
 588                 .maxlen         = sizeof(int),
 589                 .mode           = 0644,
 590                 .proc_handler   = proc_dointvec,
 591         },
 592         {
 593                 .procname       = "tcp_tso_win_divisor",
 594                 .data           = &sysctl_tcp_tso_win_divisor,
 595                 .maxlen         = sizeof(int),
 596                 .mode           = 0644,
 597                 .proc_handler   = proc_dointvec,
 598         },
 599         {
 600                 .procname       = "tcp_congestion_control",
 601                 .mode           = 0644,
 602                 .maxlen         = TCP_CA_NAME_MAX,
 603                 .proc_handler   = proc_tcp_congestion_control,
 604         },
 605         {
 606                 .procname       = "tcp_workaround_signed_windows",
 607                 .data           = &sysctl_tcp_workaround_signed_windows,
 608                 .maxlen         = sizeof(int),
 609                 .mode           = 0644,
 610                 .proc_handler   = proc_dointvec
 611         },
 612         {
 613                 .procname       = "tcp_limit_output_bytes",
 614                 .data           = &sysctl_tcp_limit_output_bytes,
 615                 .maxlen         = sizeof(int),
 616                 .mode           = 0644,
 617                 .proc_handler   = proc_dointvec
 618         },
 619         {
 620                 .procname       = "tcp_challenge_ack_limit",
 621                 .data           = &sysctl_tcp_challenge_ack_limit,
 622                 .maxlen         = sizeof(int),
 623                 .mode           = 0644,
 624                 .proc_handler   = proc_dointvec
 625         },
 626         {
 627                 .procname       = "tcp_slow_start_after_idle",
 628                 .data           = &sysctl_tcp_slow_start_after_idle,
 629                 .maxlen         = sizeof(int),
 630                 .mode           = 0644,
 631                 .proc_handler   = proc_dointvec
 632         },
 633 #ifdef CONFIG_NETLABEL
 634         {
 635                 .procname       = "cipso_cache_enable",
 636                 .data           = &cipso_v4_cache_enabled,
 637                 .maxlen         = sizeof(int),
 638                 .mode           = 0644,
 639                 .proc_handler   = proc_dointvec,
 640         },
 641         {
 642                 .procname       = "cipso_cache_bucket_size",
 643                 .data           = &cipso_v4_cache_bucketsize,
 644                 .maxlen         = sizeof(int),
 645                 .mode           = 0644,
 646                 .proc_handler   = proc_dointvec,
 647         },
 648         {
 649                 .procname       = "cipso_rbm_optfmt",
 650                 .data           = &cipso_v4_rbm_optfmt,
 651                 .maxlen         = sizeof(int),
 652                 .mode           = 0644,
 653                 .proc_handler   = proc_dointvec,
 654         },
 655         {
 656                 .procname       = "cipso_rbm_strictvalid",
 657                 .data           = &cipso_v4_rbm_strictvalid,
 658                 .maxlen         = sizeof(int),
 659                 .mode           = 0644,
 660                 .proc_handler   = proc_dointvec,
 661         },
 662 #endif /* CONFIG_NETLABEL */
 663         {
 664                 .procname       = "tcp_available_congestion_control",
 665                 .maxlen         = TCP_CA_BUF_MAX,
 666                 .mode           = 0444,
 667                 .proc_handler   = proc_tcp_available_congestion_control,
 668         },
 669         {
 670                 .procname       = "tcp_allowed_congestion_control",
 671                 .maxlen         = TCP_CA_BUF_MAX,
 672                 .mode           = 0644,
 673                 .proc_handler   = proc_allowed_congestion_control,
 674         },
 675         {
 676                 .procname       = "tcp_thin_linear_timeouts",
 677                 .data           = &sysctl_tcp_thin_linear_timeouts,
 678                 .maxlen         = sizeof(int),
 679                 .mode           = 0644,
 680                 .proc_handler   = proc_dointvec
 681         },
 682         {
 683                 .procname       = "tcp_thin_dupack",
 684                 .data           = &sysctl_tcp_thin_dupack,
 685                 .maxlen         = sizeof(int),
 686                 .mode           = 0644,
 687                 .proc_handler   = proc_dointvec
 688         },
 689         {
 690                 .procname       = "tcp_early_retrans",
 691                 .data           = &sysctl_tcp_early_retrans,
 692                 .maxlen         = sizeof(int),
 693                 .mode           = 0644,
 694                 .proc_handler   = proc_dointvec_minmax,
 695                 .extra1         = &zero,
 696                 .extra2         = &four,
 697         },
 698         {
 699                 .procname       = "tcp_min_tso_segs",
 700                 .data           = &sysctl_tcp_min_tso_segs,
 701                 .maxlen         = sizeof(int),
 702                 .mode           = 0644,
 703                 .proc_handler   = proc_dointvec_minmax,
 704                 .extra1         = &one,
 705                 .extra2         = &gso_max_segs,
 706         },
 707         {
 708                 .procname       = "tcp_pacing_ss_ratio",
 709                 .data           = &sysctl_tcp_pacing_ss_ratio,
 710                 .maxlen         = sizeof(int),
 711                 .mode           = 0644,
 712                 .proc_handler   = proc_dointvec_minmax,
 713                 .extra1         = &zero,
 714                 .extra2         = &thousand,
 715         },
 716         {
 717                 .procname       = "tcp_pacing_ca_ratio",
 718                 .data           = &sysctl_tcp_pacing_ca_ratio,
 719                 .maxlen         = sizeof(int),
 720                 .mode           = 0644,
 721                 .proc_handler   = proc_dointvec_minmax,
 722                 .extra1         = &zero,
 723                 .extra2         = &thousand,
 724         },
 725         {
 726                 .procname       = "tcp_autocorking",
 727                 .data           = &sysctl_tcp_autocorking,
 728                 .maxlen         = sizeof(int),
 729                 .mode           = 0644,
 730                 .proc_handler   = proc_dointvec_minmax,
 731                 .extra1         = &zero,
 732                 .extra2         = &one,
 733         },
 734         {
 735                 .procname       = "tcp_invalid_ratelimit",
 736                 .data           = &sysctl_tcp_invalid_ratelimit,
 737                 .maxlen         = sizeof(int),
 738                 .mode           = 0644,
 739                 .proc_handler   = proc_dointvec_ms_jiffies,
 740         },
 741         {
 742                 .procname       = "icmp_msgs_per_sec",
 743                 .data           = &sysctl_icmp_msgs_per_sec,
 744                 .maxlen         = sizeof(int),
 745                 .mode           = 0644,
 746                 .proc_handler   = proc_dointvec_minmax,
 747                 .extra1         = &zero,
 748         },
 749         {
 750                 .procname       = "icmp_msgs_burst",
 751                 .data           = &sysctl_icmp_msgs_burst,
 752                 .maxlen         = sizeof(int),
 753                 .mode           = 0644,
 754                 .proc_handler   = proc_dointvec_minmax,
 755                 .extra1         = &zero,
 756         },
 757         {
 758                 .procname       = "udp_mem",
 759                 .data           = &sysctl_udp_mem,
 760                 .maxlen         = sizeof(sysctl_udp_mem),
 761                 .mode           = 0644,
 762                 .proc_handler   = proc_doulongvec_minmax,
 763         },
 764         {
 765                 .procname       = "udp_rmem_min",
 766                 .data           = &sysctl_udp_rmem_min,
 767                 .maxlen         = sizeof(sysctl_udp_rmem_min),
 768                 .mode           = 0644,
 769                 .proc_handler   = proc_dointvec_minmax,
 770                 .extra1         = &one
 771         },
 772         {
 773                 .procname       = "udp_wmem_min",
 774                 .data           = &sysctl_udp_wmem_min,
 775                 .maxlen         = sizeof(sysctl_udp_wmem_min),
 776                 .mode           = 0644,
 777                 .proc_handler   = proc_dointvec_minmax,
 778                 .extra1         = &one
 779         },
 780         { }
 781 };
 782
 783 static struct ctl_table ipv4_net_table[] = {
 784         {
 785                 .procname       = "icmp_echo_ignore_all",
 786                 .data           = &init_net.ipv4.sysctl_icmp_echo_ignore_all,
 787                 .maxlen         = sizeof(int),
 788                 .mode           = 0644,
 789                 .proc_handler   = proc_dointvec
 790         },
 791         {
 792                 .procname       = "icmp_echo_ignore_broadcasts",
 793                 .data           = &init_net.ipv4.sysctl_icmp_echo_ignore_broadcasts,
 794                 .maxlen         = sizeof(int),
 795                 .mode           = 0644,
 796                 .proc_handler   = proc_dointvec
 797         },
 798         {
 799                 .procname       = "icmp_ignore_bogus_error_responses",
 800                 .data           = &init_net.ipv4.sysctl_icmp_ignore_bogus_error_responses,
 801                 .maxlen         = sizeof(int),
 802                 .mode           = 0644,
 803                 .proc_handler   = proc_dointvec
 804         },
 805         {
 806                 .procname       = "icmp_errors_use_inbound_ifaddr",
 807                 .data           = &init_net.ipv4.sysctl_icmp_errors_use_inbound_ifaddr,
 808                 .maxlen         = sizeof(int),
 809                 .mode           = 0644,
 810                 .proc_handler   = proc_dointvec
 811         },
 812         {
 813                 .procname       = "icmp_ratelimit",
 814                 .data           = &init_net.ipv4.sysctl_icmp_ratelimit,
 815                 .maxlen         = sizeof(int),
 816                 .mode           = 0644,
 817                 .proc_handler   = proc_dointvec_ms_jiffies,
 818         },
 819         {
 820                 .procname       = "icmp_ratemask",
 821                 .data           = &init_net.ipv4.sysctl_icmp_ratemask,
 822                 .maxlen         = sizeof(int),
 823                 .mode           = 0644,
 824                 .proc_handler   = proc_dointvec
 825         },
 826         {
 827                 .procname       = "ping_group_range",
 828                 .data           = &init_net.ipv4.ping_group_range.range,
 829                 .maxlen         = sizeof(gid_t)*2,
 830                 .mode           = 0644,
 831                 .proc_handler   = ipv4_ping_group_range,
 832         },
 833         {
 834                 .procname       = "tcp_ecn",
 835                 .data           = &init_net.ipv4.sysctl_tcp_ecn,
 836                 .maxlen         = sizeof(int),
 837                 .mode           = 0644,
 838                 .proc_handler   = proc_dointvec
 839         },
 840         {
 841                 .procname       = "tcp_ecn_fallback",
 842                 .data           = &init_net.ipv4.sysctl_tcp_ecn_fallback,
 843                 .maxlen         = sizeof(int),
 844                 .mode           = 0644,
 845                 .proc_handler   = proc_dointvec
 846         },
 847         {
 848                 .procname       = "ip_local_port_range",
 849                 .maxlen         = sizeof(init_net.ipv4.ip_local_ports.range),
 850                 .data           = &init_net.ipv4.ip_local_ports.range,
 851                 .mode           = 0644,
 852                 .proc_handler   = ipv4_local_port_range,
 853         },
 854         {
 855                 .procname       = "ip_local_reserved_ports",
 856                 .data           = &init_net.ipv4.sysctl_local_reserved_ports,
 857                 .maxlen         = 65536,
 858                 .mode           = 0644,
 859                 .proc_handler   = proc_do_large_bitmap,
 860         },
 861         {
 862                 .procname       = "ip_no_pmtu_disc",
 863                 .data           = &init_net.ipv4.sysctl_ip_no_pmtu_disc,
 864                 .maxlen         = sizeof(int),
 865                 .mode           = 0644,
 866                 .proc_handler   = proc_dointvec
 867         },
 868         {
 869                 .procname       = "ip_forward_use_pmtu",
 870                 .data           = &init_net.ipv4.sysctl_ip_fwd_use_pmtu,
 871                 .maxlen         = sizeof(int),
 872                 .mode           = 0644,
 873                 .proc_handler   = proc_dointvec,
 874         },
 875         {
 876                 .procname       = "ip_nonlocal_bind",
 877                 .data           = &init_net.ipv4.sysctl_ip_nonlocal_bind,
 878                 .maxlen         = sizeof(int),
 879                 .mode           = 0644,
 880                 .proc_handler   = proc_dointvec
 881         },
 882         {
 883                 .procname       = "fwmark_reflect",
 884                 .data           = &init_net.ipv4.sysctl_fwmark_reflect,
 885                 .maxlen         = sizeof(int),
 886                 .mode           = 0644,
 887                 .proc_handler   = proc_dointvec,
 888         },
 889         {
 890                 .procname       = "tcp_fwmark_accept",
 891                 .data           = &init_net.ipv4.sysctl_tcp_fwmark_accept,
 892                 .maxlen         = sizeof(int),
 893                 .mode           = 0644,
 894                 .proc_handler   = proc_dointvec,
 895         },
 896 #ifdef CONFIG_NET_L3_MASTER_DEV
 897         {
 898                 .procname       = "tcp_l3mdev_accept",
 899                 .data           = &init_net.ipv4.sysctl_tcp_l3mdev_accept,
 900                 .maxlen         = sizeof(int),
 901                 .mode           = 0644,
 902                 .proc_handler   = proc_dointvec_minmax,
 903                 .extra1         = &zero,
 904                 .extra2         = &one,
 905         },
 906 #endif
 907         {
 908                 .procname       = "tcp_mtu_probing",
 909                 .data           = &init_net.ipv4.sysctl_tcp_mtu_probing,
 910                 .maxlen         = sizeof(int),
 911                 .mode           = 0644,
 912                 .proc_handler   = proc_dointvec,
 913         },
 914         {
 915                 .procname       = "tcp_base_mss",
 916                 .data           = &init_net.ipv4.sysctl_tcp_base_mss,
 917                 .maxlen         = sizeof(int),
 918                 .mode           = 0644,
 919                 .proc_handler   = proc_dointvec,
 920         },
 921         {
 922                 .procname       = "tcp_probe_threshold",
 923                 .data           = &init_net.ipv4.sysctl_tcp_probe_threshold,
 924                 .maxlen         = sizeof(int),
 925                 .mode           = 0644,
 926                 .proc_handler   = proc_dointvec,
 927         },
 928         {
 929                 .procname       = "tcp_probe_interval",
 930                 .data           = &init_net.ipv4.sysctl_tcp_probe_interval,
 931                 .maxlen         = sizeof(int),
 932                 .mode           = 0644,
 933                 .proc_handler   = proc_dointvec,
 934         },
 935         {
 936                 .procname       = "igmp_link_local_mcast_reports",
 937                 .data           = &sysctl_igmp_llm_reports,
 938                 .maxlen         = sizeof(int),
 939                 .mode           = 0644,
 940                 .proc_handler   = proc_dointvec
 941         },
 942         {
 943                 .procname       = "tcp_keepalive_time",
 944                 .data           = &init_net.ipv4.sysctl_tcp_keepalive_time,
 945                 .maxlen         = sizeof(int),
 946                 .mode           = 0644,
 947                 .proc_handler   = proc_dointvec_jiffies,
 948         },
 949         {
 950                 .procname       = "tcp_keepalive_probes",
 951                 .data           = &init_net.ipv4.sysctl_tcp_keepalive_probes,
 952                 .maxlen         = sizeof(int),
 953                 .mode           = 0644,
 954                 .proc_handler   = proc_dointvec
 955         },
 956         {
 957                 .procname       = "tcp_keepalive_intvl",
 958                 .data           = &init_net.ipv4.sysctl_tcp_keepalive_intvl,
 959                 .maxlen         = sizeof(int),
 960                 .mode           = 0644,
 961                 .proc_handler   = proc_dointvec_jiffies,
 962         },
 963         { }
 964 };
 965
 966 static __net_init int ipv4_sysctl_init_net(struct net *net)
 967 {
 968         struct ctl_table *table;
 969
 970         table = ipv4_net_table;
 971         if (!net_eq(net, &init_net)) {
 972                 int i;
 973
 974                 table = kmemdup(table, sizeof(ipv4_net_table), GFP_KERNEL);
 975                 if (!table)
 976                         goto err_alloc;
 977
 978                 /* Update the variables to point into the current struct net */
 979                 for (i = 0; i < ARRAY_SIZE(ipv4_net_table) - 1; i++)
 980                         table[i].data += (void *)net - (void *)&init_net;
 981         }
 982
 983         net->ipv4.ipv4_hdr = register_net_sysctl(net, "net/ipv4", table);
 984         if (!net->ipv4.ipv4_hdr)
 985                 goto err_reg;
 986
 987         net->ipv4.sysctl_local_reserved_ports = kzalloc(65536 / 8, GFP_KERNEL);
 988         if (!net->ipv4.sysctl_local_reserved_ports)
 989                 goto err_ports;
 990
 991         return 0;
 992
 993 err_ports:
 994         unregister_net_sysctl_table(net->ipv4.ipv4_hdr);
 995 err_reg:
 996         if (!net_eq(net, &init_net))
 997                 kfree(table);
 998 err_alloc:
 999         return -ENOMEM;
1000 }
1001
1002 static __net_exit void ipv4_sysctl_exit_net(struct net *net)
1003 {
1004         struct ctl_table *table;
1005
1006         kfree(net->ipv4.sysctl_local_reserved_ports);
1007         table = net->ipv4.ipv4_hdr->ctl_table_arg;
1008         unregister_net_sysctl_table(net->ipv4.ipv4_hdr);
1009         kfree(table);
1010 }
1011
1012 static __net_initdata struct pernet_operations ipv4_sysctl_ops = {
1013         .init = ipv4_sysctl_init_net,
1014         .exit = ipv4_sysctl_exit_net,
1015 };
1016
1017 static __init int sysctl_ipv4_init(void)
1018 {
1019         struct ctl_table_header *hdr;
1020
1021         hdr = register_net_sysctl(&init_net, "net/ipv4", ipv4_table);
1022         if (!hdr)
1023                 return -ENOMEM;
1024
1025         if (register_pernet_subsys(&ipv4_sysctl_ops)) {
1026                 unregister_net_sysctl_table(hdr);
1027                 return -ENOMEM;
1028         }
1029
1030         return 0;
1031 }
1032
1033 __initcall(sysctl_ipv4_init);