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/protocol.h>
  28
  29 static int zero;
  30 static int one = 1;
  31 static int four = 4;
  32 static int thousand = 1000;
  33 static int gso_max_segs = GSO_MAX_SEGS;
  34 static int tcp_retr1_max = 255;
  35 static int ip_local_port_range_min[] = { 1, 1 };
  36 static int ip_local_port_range_max[] = { 65535, 65535 };
  37 static int tcp_adv_win_scale_min = -31;
  38 static int tcp_adv_win_scale_max = 31;
  39 static int ip_privileged_port_min;
  40 static int ip_privileged_port_max = 65535;
  41 static int ip_ttl_min = 1;
  42 static int ip_ttl_max = 255;
  43 static int tcp_syn_retries_min = 1;
  44 static int tcp_syn_retries_max = MAX_TCP_SYNCNT;
  45 static int ip_ping_group_range_min[] = { 0, 0 };
  46 static int ip_ping_group_range_max[] = { GID_T_MAX, GID_T_MAX };
  47
  48 /* obsolete */
  49 static int sysctl_tcp_low_latency __read_mostly;
  50
  51 /* Update system visible IP port range */
  52 static void set_local_port_range(struct net *net, int range[2])
  53 {
  54         bool same_parity = !((range[0] ^ range[1]) & 1);
  55
  56         write_seqlock_bh(&net->ipv4.ip_local_ports.lock);
  57         if (same_parity && !net->ipv4.ip_local_ports.warned) {
  58                 net->ipv4.ip_local_ports.warned = true;
  59                 pr_err_ratelimited("ip_local_port_range: prefer different parity for start/end values.\n");
  60         }
  61         net->ipv4.ip_local_ports.range[0] = range[0];
  62         net->ipv4.ip_local_ports.range[1] = range[1];
  63         write_sequnlock_bh(&net->ipv4.ip_local_ports.lock);
  64 }
  65
  66 /* Validate changes from /proc interface. */
  67 static int ipv4_local_port_range(struct ctl_table *table, int write,
  68                                  void __user *buffer,
  69                                  size_t *lenp, loff_t *ppos)
  70 {
  71         struct net *net =
  72                 container_of(table->data, struct net, ipv4.ip_local_ports.range);
  73         int ret;
  74         int range[2];
  75         struct ctl_table tmp = {
  76                 .data = &range,
  77                 .maxlen = sizeof(range),
  78                 .mode = table->mode,
  79                 .extra1 = &ip_local_port_range_min,
  80                 .extra2 = &ip_local_port_range_max,
  81         };
  82
  83         inet_get_local_port_range(net, &range[0], &range[1]);
  84
  85         ret = proc_dointvec_minmax(&tmp, write, buffer, lenp, ppos);
  86
  87         if (write && ret == 0) {
  88                 /* Ensure that the upper limit is not smaller than the lower,
  89                  * and that the lower does not encroach upon the privileged
  90                  * port limit.
  91                  */
  92                 if ((range[1] < range[0]) ||
  93                     (range[0] < net->ipv4.sysctl_ip_prot_sock))
  94                         ret = -EINVAL;
  95                 else
  96                         set_local_port_range(net, range);
  97         }
  98
  99         return ret;
 100 }
 101
 102 /* Validate changes from /proc interface. */
 103 static int ipv4_privileged_ports(struct ctl_table *table, int write,
 104                                 void __user *buffer, size_t *lenp, loff_t *ppos)
 105 {
 106         struct net *net = container_of(table->data, struct net,
 107             ipv4.sysctl_ip_prot_sock);
 108         int ret;
 109         int pports;
 110         int range[2];
 111         struct ctl_table tmp = {
 112                 .data = &pports,
 113                 .maxlen = sizeof(pports),
 114                 .mode = table->mode,
 115                 .extra1 = &ip_privileged_port_min,
 116                 .extra2 = &ip_privileged_port_max,
 117         };
 118
 119         pports = net->ipv4.sysctl_ip_prot_sock;
 120
 121         ret = proc_dointvec_minmax(&tmp, write, buffer, lenp, ppos);
 122
 123         if (write && ret == 0) {
 124                 inet_get_local_port_range(net, &range[0], &range[1]);
 125                 /* Ensure that the local port range doesn't overlap with the
 126                  * privileged port range.
 127                  */
 128                 if (range[0] < pports)
 129                         ret = -EINVAL;
 130                 else
 131                         net->ipv4.sysctl_ip_prot_sock = pports;
 132         }
 133
 134         return ret;
 135 }
 136
 137 static void inet_get_ping_group_range_table(struct ctl_table *table, kgid_t *low, kgid_t *high)
 138 {
 139         kgid_t *data = table->data;
 140         struct net *net =
 141                 container_of(table->data, struct net, ipv4.ping_group_range.range);
 142         unsigned int seq;
 143         do {
 144                 seq = read_seqbegin(&net->ipv4.ping_group_range.lock);
 145
 146                 *low = data[0];
 147                 *high = data[1];
 148         } while (read_seqretry(&net->ipv4.ping_group_range.lock, seq));
 149 }
 150
 151 /* Update system visible IP port range */
 152 static void set_ping_group_range(struct ctl_table *table, kgid_t low, kgid_t high)
 153 {
 154         kgid_t *data = table->data;
 155         struct net *net =
 156                 container_of(table->data, struct net, ipv4.ping_group_range.range);
 157         write_seqlock(&net->ipv4.ping_group_range.lock);
 158         data[0] = low;
 159         data[1] = high;
 160         write_sequnlock(&net->ipv4.ping_group_range.lock);
 161 }
 162
 163 /* Validate changes from /proc interface. */
 164 static int ipv4_ping_group_range(struct ctl_table *table, int write,
 165                                  void __user *buffer,
 166                                  size_t *lenp, loff_t *ppos)
 167 {
 168         struct user_namespace *user_ns = current_user_ns();
 169         int ret;
 170         gid_t urange[2];
 171         kgid_t low, high;
 172         struct ctl_table tmp = {
 173                 .data = &urange,
 174                 .maxlen = sizeof(urange),
 175                 .mode = table->mode,
 176                 .extra1 = &ip_ping_group_range_min,
 177                 .extra2 = &ip_ping_group_range_max,
 178         };
 179
 180         inet_get_ping_group_range_table(table, &low, &high);
 181         urange[0] = from_kgid_munged(user_ns, low);
 182         urange[1] = from_kgid_munged(user_ns, high);
 183         ret = proc_dointvec_minmax(&tmp, write, buffer, lenp, ppos);
 184
 185         if (write && ret == 0) {
 186                 low = make_kgid(user_ns, urange[0]);
 187                 high = make_kgid(user_ns, urange[1]);
 188                 if (!gid_valid(low) || !gid_valid(high) ||
 189                     (urange[1] < urange[0]) || gid_lt(high, low)) {
 190                         low = make_kgid(&init_user_ns, 1);
 191                         high = make_kgid(&init_user_ns, 0);
 192                 }
 193                 set_ping_group_range(table, low, high);
 194         }
 195
 196         return ret;
 197 }
 198
 199 static int proc_tcp_congestion_control(struct ctl_table *ctl, int write,
 200                                        void __user *buffer, size_t *lenp, loff_t *ppos)
 201 {
 202         char val[TCP_CA_NAME_MAX];
 203         struct ctl_table tbl = {
 204                 .data = val,
 205                 .maxlen = TCP_CA_NAME_MAX,
 206         };
 207         int ret;
 208
 209         tcp_get_default_congestion_control(val);
 210
 211         ret = proc_dostring(&tbl, write, buffer, lenp, ppos);
 212         if (write && ret == 0)
 213                 ret = tcp_set_default_congestion_control(val);
 214         return ret;
 215 }
 216
 217 static int proc_tcp_available_congestion_control(struct ctl_table *ctl,
 218                                                  int write,
 219                                                  void __user *buffer, size_t *lenp,
 220                                                  loff_t *ppos)
 221 {
 222         struct ctl_table tbl = { .maxlen = TCP_CA_BUF_MAX, };
 223         int ret;
 224
 225         tbl.data = kmalloc(tbl.maxlen, GFP_USER);
 226         if (!tbl.data)
 227                 return -ENOMEM;
 228         tcp_get_available_congestion_control(tbl.data, TCP_CA_BUF_MAX);
 229         ret = proc_dostring(&tbl, write, buffer, lenp, ppos);
 230         kfree(tbl.data);
 231         return ret;
 232 }
 233
 234 static int proc_allowed_congestion_control(struct ctl_table *ctl,
 235                                            int write,
 236                                            void __user *buffer, size_t *lenp,
 237                                            loff_t *ppos)
 238 {
 239         struct ctl_table tbl = { .maxlen = TCP_CA_BUF_MAX };
 240         int ret;
 241
 242         tbl.data = kmalloc(tbl.maxlen, GFP_USER);
 243         if (!tbl.data)
 244                 return -ENOMEM;
 245
 246         tcp_get_allowed_congestion_control(tbl.data, tbl.maxlen);
 247         ret = proc_dostring(&tbl, write, buffer, lenp, ppos);
 248         if (write && ret == 0)
 249                 ret = tcp_set_allowed_congestion_control(tbl.data);
 250         kfree(tbl.data);
 251         return ret;
 252 }
 253
 254 static int proc_tcp_fastopen_key(struct ctl_table *ctl, int write,
 255                                  void __user *buffer, size_t *lenp,
 256                                  loff_t *ppos)
 257 {
 258         struct ctl_table tbl = { .maxlen = (TCP_FASTOPEN_KEY_LENGTH * 2 + 10) };
 259         struct tcp_fastopen_context *ctxt;
 260         int ret;
 261         u32  user_key[4]; /* 16 bytes, matching TCP_FASTOPEN_KEY_LENGTH */
 262
 263         tbl.data = kmalloc(tbl.maxlen, GFP_KERNEL);
 264         if (!tbl.data)
 265                 return -ENOMEM;
 266
 267         rcu_read_lock();
 268         ctxt = rcu_dereference(tcp_fastopen_ctx);
 269         if (ctxt)
 270                 memcpy(user_key, ctxt->key, TCP_FASTOPEN_KEY_LENGTH);
 271         else
 272                 memset(user_key, 0, sizeof(user_key));
 273         rcu_read_unlock();
 274
 275         snprintf(tbl.data, tbl.maxlen, "%08x-%08x-%08x-%08x",
 276                 user_key[0], user_key[1], user_key[2], user_key[3]);
 277         ret = proc_dostring(&tbl, write, buffer, lenp, ppos);
 278
 279         if (write && ret == 0) {
 280                 if (sscanf(tbl.data, "%x-%x-%x-%x", user_key, user_key + 1,
 281                            user_key + 2, user_key + 3) != 4) {
 282                         ret = -EINVAL;
 283                         goto bad_key;
 284                 }
 285                 /* Generate a dummy secret but don't publish it. This
 286                  * is needed so we don't regenerate a new key on the
 287                  * first invocation of tcp_fastopen_cookie_gen
 288                  */
 289                 tcp_fastopen_init_key_once(false);
 290                 tcp_fastopen_reset_cipher(user_key, TCP_FASTOPEN_KEY_LENGTH);
 291         }
 292
 293 bad_key:
 294         pr_debug("proc FO key set 0x%x-%x-%x-%x <- 0x%s: %u\n",
 295                user_key[0], user_key[1], user_key[2], user_key[3],
 296                (char *)tbl.data, ret);
 297         kfree(tbl.data);
 298         return ret;
 299 }
 300
 301 static void proc_configure_early_demux(int enabled, int protocol)
 302 {
 303         struct net_protocol *ipprot;
 304 #if IS_ENABLED(CONFIG_IPV6)
 305         struct inet6_protocol *ip6prot;
 306 #endif
 307
 308         rcu_read_lock();
 309
 310         ipprot = rcu_dereference(inet_protos[protocol]);
 311         if (ipprot)
 312                 ipprot->early_demux = enabled ? ipprot->early_demux_handler :
 313                                                 NULL;
 314
 315 #if IS_ENABLED(CONFIG_IPV6)
 316         ip6prot = rcu_dereference(inet6_protos[protocol]);
 317         if (ip6prot)
 318                 ip6prot->early_demux = enabled ? ip6prot->early_demux_handler :
 319                                                  NULL;
 320 #endif
 321         rcu_read_unlock();
 322 }
 323
 324 static int proc_tcp_early_demux(struct ctl_table *table, int write,
 325                                 void __user *buffer, size_t *lenp, loff_t *ppos)
 326 {
 327         int ret = 0;
 328
 329         ret = proc_dointvec(table, write, buffer, lenp, ppos);
 330
 331         if (write && !ret) {
 332                 int enabled = init_net.ipv4.sysctl_tcp_early_demux;
 333
 334                 proc_configure_early_demux(enabled, IPPROTO_TCP);
 335         }
 336
 337         return ret;
 338 }
 339
 340 static int proc_udp_early_demux(struct ctl_table *table, int write,
 341                                 void __user *buffer, size_t *lenp, loff_t *ppos)
 342 {
 343         int ret = 0;
 344
 345         ret = proc_dointvec(table, write, buffer, lenp, ppos);
 346
 347         if (write && !ret) {
 348                 int enabled = init_net.ipv4.sysctl_udp_early_demux;
 349
 350                 proc_configure_early_demux(enabled, IPPROTO_UDP);
 351         }
 352
 353         return ret;
 354 }
 355
 356 static int proc_tfo_blackhole_detect_timeout(struct ctl_table *table,
 357                                              int write,
 358                                              void __user *buffer,
 359                                              size_t *lenp, loff_t *ppos)
 360 {
 361         int ret;
 362
 363         ret = proc_dointvec_minmax(table, write, buffer, lenp, ppos);
 364         if (write && ret == 0)
 365                 tcp_fastopen_active_timeout_reset();
 366
 367         return ret;
 368 }
 369
 370 static int proc_tcp_available_ulp(struct ctl_table *ctl,
 371                                   int write,
 372                                   void __user *buffer, size_t *lenp,
 373                                   loff_t *ppos)
 374 {
 375         struct ctl_table tbl = { .maxlen = TCP_ULP_BUF_MAX, };
 376         int ret;
 377
 378         tbl.data = kmalloc(tbl.maxlen, GFP_USER);
 379         if (!tbl.data)
 380                 return -ENOMEM;
 381         tcp_get_available_ulp(tbl.data, TCP_ULP_BUF_MAX);
 382         ret = proc_dostring(&tbl, write, buffer, lenp, ppos);
 383         kfree(tbl.data);
 384
 385         return ret;
 386 }
 387
 388 static struct ctl_table ipv4_table[] = {
 389         {
 390                 .procname       = "tcp_retrans_collapse",
 391                 .data           = &sysctl_tcp_retrans_collapse,
 392                 .maxlen         = sizeof(int),
 393                 .mode           = 0644,
 394                 .proc_handler   = proc_dointvec
 395         },
 396         {
 397                 .procname       = "tcp_max_orphans",
 398                 .data           = &sysctl_tcp_max_orphans,
 399                 .maxlen         = sizeof(int),
 400                 .mode           = 0644,
 401                 .proc_handler   = proc_dointvec
 402         },
 403         {
 404                 .procname       = "tcp_fastopen_key",
 405                 .mode           = 0600,
 406                 .maxlen         = ((TCP_FASTOPEN_KEY_LENGTH * 2) + 10),
 407                 .proc_handler   = proc_tcp_fastopen_key,
 408         },
 409         {
 410                 .procname       = "tcp_fastopen_blackhole_timeout_sec",
 411                 .data           = &sysctl_tcp_fastopen_blackhole_timeout,
 412                 .maxlen         = sizeof(int),
 413                 .mode           = 0644,
 414                 .proc_handler   = proc_tfo_blackhole_detect_timeout,
 415                 .extra1         = &zero,
 416         },
 417         {
 418                 .procname       = "tcp_abort_on_overflow",
 419                 .data           = &sysctl_tcp_abort_on_overflow,
 420                 .maxlen         = sizeof(int),
 421                 .mode           = 0644,
 422                 .proc_handler   = proc_dointvec
 423         },
 424         {
 425                 .procname       = "tcp_stdurg",
 426                 .data           = &sysctl_tcp_stdurg,
 427                 .maxlen         = sizeof(int),
 428                 .mode           = 0644,
 429                 .proc_handler   = proc_dointvec
 430         },
 431         {
 432                 .procname       = "tcp_rfc1337",
 433                 .data           = &sysctl_tcp_rfc1337,
 434                 .maxlen         = sizeof(int),
 435                 .mode           = 0644,
 436                 .proc_handler   = proc_dointvec
 437         },
 438         {
 439                 .procname       = "inet_peer_threshold",
 440                 .data           = &inet_peer_threshold,
 441                 .maxlen         = sizeof(int),
 442                 .mode           = 0644,
 443                 .proc_handler   = proc_dointvec
 444         },
 445         {
 446                 .procname       = "inet_peer_minttl",
 447                 .data           = &inet_peer_minttl,
 448                 .maxlen         = sizeof(int),
 449                 .mode           = 0644,
 450                 .proc_handler   = proc_dointvec_jiffies,
 451         },
 452         {
 453                 .procname       = "inet_peer_maxttl",
 454                 .data           = &inet_peer_maxttl,
 455                 .maxlen         = sizeof(int),
 456                 .mode           = 0644,
 457                 .proc_handler   = proc_dointvec_jiffies,
 458         },
 459         {
 460                 .procname       = "tcp_fack",
 461                 .data           = &sysctl_tcp_fack,
 462                 .maxlen         = sizeof(int),
 463                 .mode           = 0644,
 464                 .proc_handler   = proc_dointvec
 465         },
 466         {
 467                 .procname       = "tcp_recovery",
 468                 .data           = &sysctl_tcp_recovery,
 469                 .maxlen         = sizeof(int),
 470                 .mode           = 0644,
 471                 .proc_handler   = proc_dointvec,
 472         },
 473         {
 474                 .procname       = "tcp_max_reordering",
 475                 .data           = &sysctl_tcp_max_reordering,
 476                 .maxlen         = sizeof(int),
 477                 .mode           = 0644,
 478                 .proc_handler   = proc_dointvec
 479         },
 480         {
 481                 .procname       = "tcp_dsack",
 482                 .data           = &sysctl_tcp_dsack,
 483                 .maxlen         = sizeof(int),
 484                 .mode           = 0644,
 485                 .proc_handler   = proc_dointvec
 486         },
 487         {
 488                 .procname       = "tcp_mem",
 489                 .maxlen         = sizeof(sysctl_tcp_mem),
 490                 .data           = &sysctl_tcp_mem,
 491                 .mode           = 0644,
 492                 .proc_handler   = proc_doulongvec_minmax,
 493         },
 494         {
 495                 .procname       = "tcp_wmem",
 496                 .data           = &sysctl_tcp_wmem,
 497                 .maxlen         = sizeof(sysctl_tcp_wmem),
 498                 .mode           = 0644,
 499                 .proc_handler   = proc_dointvec_minmax,
 500                 .extra1         = &one,
 501         },
 502         {
 503                 .procname       = "tcp_rmem",
 504                 .data           = &sysctl_tcp_rmem,
 505                 .maxlen         = sizeof(sysctl_tcp_rmem),
 506                 .mode           = 0644,
 507                 .proc_handler   = proc_dointvec_minmax,
 508                 .extra1         = &one,
 509         },
 510         {
 511                 .procname       = "tcp_app_win",
 512                 .data           = &sysctl_tcp_app_win,
 513                 .maxlen         = sizeof(int),
 514                 .mode           = 0644,
 515                 .proc_handler   = proc_dointvec
 516         },
 517         {
 518                 .procname       = "tcp_adv_win_scale",
 519                 .data           = &sysctl_tcp_adv_win_scale,
 520                 .maxlen         = sizeof(int),
 521                 .mode           = 0644,
 522                 .proc_handler   = proc_dointvec_minmax,
 523                 .extra1         = &tcp_adv_win_scale_min,
 524                 .extra2         = &tcp_adv_win_scale_max,
 525         },
 526         {
 527                 .procname       = "tcp_frto",
 528                 .data           = &sysctl_tcp_frto,
 529                 .maxlen         = sizeof(int),
 530                 .mode           = 0644,
 531                 .proc_handler   = proc_dointvec
 532         },
 533         {
 534                 .procname       = "tcp_min_rtt_wlen",
 535                 .data           = &sysctl_tcp_min_rtt_wlen,
 536                 .maxlen         = sizeof(int),
 537                 .mode           = 0644,
 538                 .proc_handler   = proc_dointvec
 539         },
 540         {
 541                 .procname       = "tcp_low_latency",
 542                 .data           = &sysctl_tcp_low_latency,
 543                 .maxlen         = sizeof(int),
 544                 .mode           = 0644,
 545                 .proc_handler   = proc_dointvec
 546         },
 547         {
 548                 .procname       = "tcp_no_metrics_save",
 549                 .data           = &sysctl_tcp_nometrics_save,
 550                 .maxlen         = sizeof(int),
 551                 .mode           = 0644,
 552                 .proc_handler   = proc_dointvec,
 553         },
 554         {
 555                 .procname       = "tcp_moderate_rcvbuf",
 556                 .data           = &sysctl_tcp_moderate_rcvbuf,
 557                 .maxlen         = sizeof(int),
 558                 .mode           = 0644,
 559                 .proc_handler   = proc_dointvec,
 560         },
 561         {
 562                 .procname       = "tcp_tso_win_divisor",
 563                 .data           = &sysctl_tcp_tso_win_divisor,
 564                 .maxlen         = sizeof(int),
 565                 .mode           = 0644,
 566                 .proc_handler   = proc_dointvec,
 567         },
 568         {
 569                 .procname       = "tcp_congestion_control",
 570                 .mode           = 0644,
 571                 .maxlen         = TCP_CA_NAME_MAX,
 572                 .proc_handler   = proc_tcp_congestion_control,
 573         },
 574         {
 575                 .procname       = "tcp_workaround_signed_windows",
 576                 .data           = &sysctl_tcp_workaround_signed_windows,
 577                 .maxlen         = sizeof(int),
 578                 .mode           = 0644,
 579                 .proc_handler   = proc_dointvec
 580         },
 581         {
 582                 .procname       = "tcp_limit_output_bytes",
 583                 .data           = &sysctl_tcp_limit_output_bytes,
 584                 .maxlen         = sizeof(int),
 585                 .mode           = 0644,
 586                 .proc_handler   = proc_dointvec
 587         },
 588         {
 589                 .procname       = "tcp_challenge_ack_limit",
 590                 .data           = &sysctl_tcp_challenge_ack_limit,
 591                 .maxlen         = sizeof(int),
 592                 .mode           = 0644,
 593                 .proc_handler   = proc_dointvec
 594         },
 595         {
 596                 .procname       = "tcp_slow_start_after_idle",
 597                 .data           = &sysctl_tcp_slow_start_after_idle,
 598                 .maxlen         = sizeof(int),
 599                 .mode           = 0644,
 600                 .proc_handler   = proc_dointvec
 601         },
 602 #ifdef CONFIG_NETLABEL
 603         {
 604                 .procname       = "cipso_cache_enable",
 605                 .data           = &cipso_v4_cache_enabled,
 606                 .maxlen         = sizeof(int),
 607                 .mode           = 0644,
 608                 .proc_handler   = proc_dointvec,
 609         },
 610         {
 611                 .procname       = "cipso_cache_bucket_size",
 612                 .data           = &cipso_v4_cache_bucketsize,
 613                 .maxlen         = sizeof(int),
 614                 .mode           = 0644,
 615                 .proc_handler   = proc_dointvec,
 616         },
 617         {
 618                 .procname       = "cipso_rbm_optfmt",
 619                 .data           = &cipso_v4_rbm_optfmt,
 620                 .maxlen         = sizeof(int),
 621                 .mode           = 0644,
 622                 .proc_handler   = proc_dointvec,
 623         },
 624         {
 625                 .procname       = "cipso_rbm_strictvalid",
 626                 .data           = &cipso_v4_rbm_strictvalid,
 627                 .maxlen         = sizeof(int),
 628                 .mode           = 0644,
 629                 .proc_handler   = proc_dointvec,
 630         },
 631 #endif /* CONFIG_NETLABEL */
 632         {
 633                 .procname       = "tcp_available_congestion_control",
 634                 .maxlen         = TCP_CA_BUF_MAX,
 635                 .mode           = 0444,
 636                 .proc_handler   = proc_tcp_available_congestion_control,
 637         },
 638         {
 639                 .procname       = "tcp_allowed_congestion_control",
 640                 .maxlen         = TCP_CA_BUF_MAX,
 641                 .mode           = 0644,
 642                 .proc_handler   = proc_allowed_congestion_control,
 643         },
 644         {
 645                 .procname       = "tcp_thin_linear_timeouts",
 646                 .data           = &sysctl_tcp_thin_linear_timeouts,
 647                 .maxlen         = sizeof(int),
 648                 .mode           = 0644,
 649                 .proc_handler   = proc_dointvec
 650         },
 651         {
 652                 .procname       = "tcp_early_retrans",
 653                 .data           = &sysctl_tcp_early_retrans,
 654                 .maxlen         = sizeof(int),
 655                 .mode           = 0644,
 656                 .proc_handler   = proc_dointvec_minmax,
 657                 .extra1         = &zero,
 658                 .extra2         = &four,
 659         },
 660         {
 661                 .procname       = "tcp_min_tso_segs",
 662                 .data           = &sysctl_tcp_min_tso_segs,
 663                 .maxlen         = sizeof(int),
 664                 .mode           = 0644,
 665                 .proc_handler   = proc_dointvec_minmax,
 666                 .extra1         = &one,
 667                 .extra2         = &gso_max_segs,
 668         },
 669         {
 670                 .procname       = "tcp_pacing_ss_ratio",
 671                 .data           = &sysctl_tcp_pacing_ss_ratio,
 672                 .maxlen         = sizeof(int),
 673                 .mode           = 0644,
 674                 .proc_handler   = proc_dointvec_minmax,
 675                 .extra1         = &zero,
 676                 .extra2         = &thousand,
 677         },
 678         {
 679                 .procname       = "tcp_pacing_ca_ratio",
 680                 .data           = &sysctl_tcp_pacing_ca_ratio,
 681                 .maxlen         = sizeof(int),
 682                 .mode           = 0644,
 683                 .proc_handler   = proc_dointvec_minmax,
 684                 .extra1         = &zero,
 685                 .extra2         = &thousand,
 686         },
 687         {
 688                 .procname       = "tcp_autocorking",
 689                 .data           = &sysctl_tcp_autocorking,
 690                 .maxlen         = sizeof(int),
 691                 .mode           = 0644,
 692                 .proc_handler   = proc_dointvec_minmax,
 693                 .extra1         = &zero,
 694                 .extra2         = &one,
 695         },
 696         {
 697                 .procname       = "tcp_invalid_ratelimit",
 698                 .data           = &sysctl_tcp_invalid_ratelimit,
 699                 .maxlen         = sizeof(int),
 700                 .mode           = 0644,
 701                 .proc_handler   = proc_dointvec_ms_jiffies,
 702         },
 703         {
 704                 .procname       = "tcp_available_ulp",
 705                 .maxlen         = TCP_ULP_BUF_MAX,
 706                 .mode           = 0444,
 707                 .proc_handler   = proc_tcp_available_ulp,
 708         },
 709         {
 710                 .procname       = "icmp_msgs_per_sec",
 711                 .data           = &sysctl_icmp_msgs_per_sec,
 712                 .maxlen         = sizeof(int),
 713                 .mode           = 0644,
 714                 .proc_handler   = proc_dointvec_minmax,
 715                 .extra1         = &zero,
 716         },
 717         {
 718                 .procname       = "icmp_msgs_burst",
 719                 .data           = &sysctl_icmp_msgs_burst,
 720                 .maxlen         = sizeof(int),
 721                 .mode           = 0644,
 722                 .proc_handler   = proc_dointvec_minmax,
 723                 .extra1         = &zero,
 724         },
 725         {
 726                 .procname       = "udp_mem",
 727                 .data           = &sysctl_udp_mem,
 728                 .maxlen         = sizeof(sysctl_udp_mem),
 729                 .mode           = 0644,
 730                 .proc_handler   = proc_doulongvec_minmax,
 731         },
 732         {
 733                 .procname       = "udp_rmem_min",
 734                 .data           = &sysctl_udp_rmem_min,
 735                 .maxlen         = sizeof(sysctl_udp_rmem_min),
 736                 .mode           = 0644,
 737                 .proc_handler   = proc_dointvec_minmax,
 738                 .extra1         = &one
 739         },
 740         {
 741                 .procname       = "udp_wmem_min",
 742                 .data           = &sysctl_udp_wmem_min,
 743                 .maxlen         = sizeof(sysctl_udp_wmem_min),
 744                 .mode           = 0644,
 745                 .proc_handler   = proc_dointvec_minmax,
 746                 .extra1         = &one
 747         },
 748         { }
 749 };
 750
 751 static struct ctl_table ipv4_net_table[] = {
 752         {
 753                 .procname       = "icmp_echo_ignore_all",
 754                 .data           = &init_net.ipv4.sysctl_icmp_echo_ignore_all,
 755                 .maxlen         = sizeof(int),
 756                 .mode           = 0644,
 757                 .proc_handler   = proc_dointvec
 758         },
 759         {
 760                 .procname       = "icmp_echo_ignore_broadcasts",
 761                 .data           = &init_net.ipv4.sysctl_icmp_echo_ignore_broadcasts,
 762                 .maxlen         = sizeof(int),
 763                 .mode           = 0644,
 764                 .proc_handler   = proc_dointvec
 765         },
 766         {
 767                 .procname       = "icmp_ignore_bogus_error_responses",
 768                 .data           = &init_net.ipv4.sysctl_icmp_ignore_bogus_error_responses,
 769                 .maxlen         = sizeof(int),
 770                 .mode           = 0644,
 771                 .proc_handler   = proc_dointvec
 772         },
 773         {
 774                 .procname       = "icmp_errors_use_inbound_ifaddr",
 775                 .data           = &init_net.ipv4.sysctl_icmp_errors_use_inbound_ifaddr,
 776                 .maxlen         = sizeof(int),
 777                 .mode           = 0644,
 778                 .proc_handler   = proc_dointvec
 779         },
 780         {
 781                 .procname       = "icmp_ratelimit",
 782                 .data           = &init_net.ipv4.sysctl_icmp_ratelimit,
 783                 .maxlen         = sizeof(int),
 784                 .mode           = 0644,
 785                 .proc_handler   = proc_dointvec_ms_jiffies,
 786         },
 787         {
 788                 .procname       = "icmp_ratemask",
 789                 .data           = &init_net.ipv4.sysctl_icmp_ratemask,
 790                 .maxlen         = sizeof(int),
 791                 .mode           = 0644,
 792                 .proc_handler   = proc_dointvec
 793         },
 794         {
 795                 .procname       = "ping_group_range",
 796                 .data           = &init_net.ipv4.ping_group_range.range,
 797                 .maxlen         = sizeof(gid_t)*2,
 798                 .mode           = 0644,
 799                 .proc_handler   = ipv4_ping_group_range,
 800         },
 801         {
 802                 .procname       = "tcp_ecn",
 803                 .data           = &init_net.ipv4.sysctl_tcp_ecn,
 804                 .maxlen         = sizeof(int),
 805                 .mode           = 0644,
 806                 .proc_handler   = proc_dointvec
 807         },
 808         {
 809                 .procname       = "tcp_ecn_fallback",
 810                 .data           = &init_net.ipv4.sysctl_tcp_ecn_fallback,
 811                 .maxlen         = sizeof(int),
 812                 .mode           = 0644,
 813                 .proc_handler   = proc_dointvec
 814         },
 815         {
 816                 .procname       = "ip_dynaddr",
 817                 .data           = &init_net.ipv4.sysctl_ip_dynaddr,
 818                 .maxlen         = sizeof(int),
 819                 .mode           = 0644,
 820                 .proc_handler   = proc_dointvec
 821         },
 822         {
 823                 .procname       = "ip_early_demux",
 824                 .data           = &init_net.ipv4.sysctl_ip_early_demux,
 825                 .maxlen         = sizeof(int),
 826                 .mode           = 0644,
 827                 .proc_handler   = proc_dointvec
 828         },
 829         {
 830                 .procname       = "udp_early_demux",
 831                 .data           = &init_net.ipv4.sysctl_udp_early_demux,
 832                 .maxlen         = sizeof(int),
 833                 .mode           = 0644,
 834                 .proc_handler   = proc_udp_early_demux
 835         },
 836         {
 837                 .procname       = "tcp_early_demux",
 838                 .data           = &init_net.ipv4.sysctl_tcp_early_demux,
 839                 .maxlen         = sizeof(int),
 840                 .mode           = 0644,
 841                 .proc_handler   = proc_tcp_early_demux
 842         },
 843         {
 844                 .procname       = "ip_default_ttl",
 845                 .data           = &init_net.ipv4.sysctl_ip_default_ttl,
 846                 .maxlen         = sizeof(int),
 847                 .mode           = 0644,
 848                 .proc_handler   = proc_dointvec_minmax,
 849                 .extra1         = &ip_ttl_min,
 850                 .extra2         = &ip_ttl_max,
 851         },
 852         {
 853                 .procname       = "ip_local_port_range",
 854                 .maxlen         = sizeof(init_net.ipv4.ip_local_ports.range),
 855                 .data           = &init_net.ipv4.ip_local_ports.range,
 856                 .mode           = 0644,
 857                 .proc_handler   = ipv4_local_port_range,
 858         },
 859         {
 860                 .procname       = "ip_local_reserved_ports",
 861                 .data           = &init_net.ipv4.sysctl_local_reserved_ports,
 862                 .maxlen         = 65536,
 863                 .mode           = 0644,
 864                 .proc_handler   = proc_do_large_bitmap,
 865         },
 866         {
 867                 .procname       = "ip_no_pmtu_disc",
 868                 .data           = &init_net.ipv4.sysctl_ip_no_pmtu_disc,
 869                 .maxlen         = sizeof(int),
 870                 .mode           = 0644,
 871                 .proc_handler   = proc_dointvec
 872         },
 873         {
 874                 .procname       = "ip_forward_use_pmtu",
 875                 .data           = &init_net.ipv4.sysctl_ip_fwd_use_pmtu,
 876                 .maxlen         = sizeof(int),
 877                 .mode           = 0644,
 878                 .proc_handler   = proc_dointvec,
 879         },
 880         {
 881                 .procname       = "ip_nonlocal_bind",
 882                 .data           = &init_net.ipv4.sysctl_ip_nonlocal_bind,
 883                 .maxlen         = sizeof(int),
 884                 .mode           = 0644,
 885                 .proc_handler   = proc_dointvec
 886         },
 887         {
 888                 .procname       = "fwmark_reflect",
 889                 .data           = &init_net.ipv4.sysctl_fwmark_reflect,
 890                 .maxlen         = sizeof(int),
 891                 .mode           = 0644,
 892                 .proc_handler   = proc_dointvec,
 893         },
 894         {
 895                 .procname       = "tcp_fwmark_accept",
 896                 .data           = &init_net.ipv4.sysctl_tcp_fwmark_accept,
 897                 .maxlen         = sizeof(int),
 898                 .mode           = 0644,
 899                 .proc_handler   = proc_dointvec,
 900         },
 901 #ifdef CONFIG_NET_L3_MASTER_DEV
 902         {
 903                 .procname       = "tcp_l3mdev_accept",
 904                 .data           = &init_net.ipv4.sysctl_tcp_l3mdev_accept,
 905                 .maxlen         = sizeof(int),
 906                 .mode           = 0644,
 907                 .proc_handler   = proc_dointvec_minmax,
 908                 .extra1         = &zero,
 909                 .extra2         = &one,
 910         },
 911 #endif
 912         {
 913                 .procname       = "tcp_mtu_probing",
 914                 .data           = &init_net.ipv4.sysctl_tcp_mtu_probing,
 915                 .maxlen         = sizeof(int),
 916                 .mode           = 0644,
 917                 .proc_handler   = proc_dointvec,
 918         },
 919         {
 920                 .procname       = "tcp_base_mss",
 921                 .data           = &init_net.ipv4.sysctl_tcp_base_mss,
 922                 .maxlen         = sizeof(int),
 923                 .mode           = 0644,
 924                 .proc_handler   = proc_dointvec,
 925         },
 926         {
 927                 .procname       = "tcp_probe_threshold",
 928                 .data           = &init_net.ipv4.sysctl_tcp_probe_threshold,
 929                 .maxlen         = sizeof(int),
 930                 .mode           = 0644,
 931                 .proc_handler   = proc_dointvec,
 932         },
 933         {
 934                 .procname       = "tcp_probe_interval",
 935                 .data           = &init_net.ipv4.sysctl_tcp_probe_interval,
 936                 .maxlen         = sizeof(int),
 937                 .mode           = 0644,
 938                 .proc_handler   = proc_dointvec,
 939         },
 940         {
 941                 .procname       = "igmp_link_local_mcast_reports",
 942                 .data           = &init_net.ipv4.sysctl_igmp_llm_reports,
 943                 .maxlen         = sizeof(int),
 944                 .mode           = 0644,
 945                 .proc_handler   = proc_dointvec
 946         },
 947         {
 948                 .procname       = "igmp_max_memberships",
 949                 .data           = &init_net.ipv4.sysctl_igmp_max_memberships,
 950                 .maxlen         = sizeof(int),
 951                 .mode           = 0644,
 952                 .proc_handler   = proc_dointvec
 953         },
 954         {
 955                 .procname       = "igmp_max_msf",
 956                 .data           = &init_net.ipv4.sysctl_igmp_max_msf,
 957                 .maxlen         = sizeof(int),
 958                 .mode           = 0644,
 959                 .proc_handler   = proc_dointvec
 960         },
 961 #ifdef CONFIG_IP_MULTICAST
 962         {
 963                 .procname       = "igmp_qrv",
 964                 .data           = &init_net.ipv4.sysctl_igmp_qrv,
 965                 .maxlen         = sizeof(int),
 966                 .mode           = 0644,
 967                 .proc_handler   = proc_dointvec_minmax,
 968                 .extra1         = &one
 969         },
 970 #endif
 971         {
 972                 .procname       = "tcp_keepalive_time",
 973                 .data           = &init_net.ipv4.sysctl_tcp_keepalive_time,
 974                 .maxlen         = sizeof(int),
 975                 .mode           = 0644,
 976                 .proc_handler   = proc_dointvec_jiffies,
 977         },
 978         {
 979                 .procname       = "tcp_keepalive_probes",
 980                 .data           = &init_net.ipv4.sysctl_tcp_keepalive_probes,
 981                 .maxlen         = sizeof(int),
 982                 .mode           = 0644,
 983                 .proc_handler   = proc_dointvec
 984         },
 985         {
 986                 .procname       = "tcp_keepalive_intvl",
 987                 .data           = &init_net.ipv4.sysctl_tcp_keepalive_intvl,
 988                 .maxlen         = sizeof(int),
 989                 .mode           = 0644,
 990                 .proc_handler   = proc_dointvec_jiffies,
 991         },
 992         {
 993                 .procname       = "tcp_syn_retries",
 994                 .data           = &init_net.ipv4.sysctl_tcp_syn_retries,
 995                 .maxlen         = sizeof(int),
 996                 .mode           = 0644,
 997                 .proc_handler   = proc_dointvec_minmax,
 998                 .extra1         = &tcp_syn_retries_min,
 999                 .extra2         = &tcp_syn_retries_max
1000         },
1001         {
1002                 .procname       = "tcp_synack_retries",
1003                 .data           = &init_net.ipv4.sysctl_tcp_synack_retries,
1004                 .maxlen         = sizeof(int),
1005                 .mode           = 0644,
1006                 .proc_handler   = proc_dointvec
1007         },
1008 #ifdef CONFIG_SYN_COOKIES
1009         {
1010                 .procname       = "tcp_syncookies",
1011                 .data           = &init_net.ipv4.sysctl_tcp_syncookies,
1012                 .maxlen         = sizeof(int),
1013                 .mode           = 0644,
1014                 .proc_handler   = proc_dointvec
1015         },
1016 #endif
1017         {
1018                 .procname       = "tcp_reordering",
1019                 .data           = &init_net.ipv4.sysctl_tcp_reordering,
1020                 .maxlen         = sizeof(int),
1021                 .mode           = 0644,
1022                 .proc_handler   = proc_dointvec
1023         },
1024         {
1025                 .procname       = "tcp_retries1",
1026                 .data           = &init_net.ipv4.sysctl_tcp_retries1,
1027                 .maxlen         = sizeof(int),
1028                 .mode           = 0644,
1029                 .proc_handler   = proc_dointvec_minmax,
1030                 .extra2         = &tcp_retr1_max
1031         },
1032         {
1033                 .procname       = "tcp_retries2",
1034                 .data           = &init_net.ipv4.sysctl_tcp_retries2,
1035                 .maxlen         = sizeof(int),
1036                 .mode           = 0644,
1037                 .proc_handler   = proc_dointvec
1038         },
1039         {
1040                 .procname       = "tcp_orphan_retries",
1041                 .data           = &init_net.ipv4.sysctl_tcp_orphan_retries,
1042                 .maxlen         = sizeof(int),
1043                 .mode           = 0644,
1044                 .proc_handler   = proc_dointvec
1045         },
1046         {
1047                 .procname       = "tcp_fin_timeout",
1048                 .data           = &init_net.ipv4.sysctl_tcp_fin_timeout,
1049                 .maxlen         = sizeof(int),
1050                 .mode           = 0644,
1051                 .proc_handler   = proc_dointvec_jiffies,
1052         },
1053         {
1054                 .procname       = "tcp_notsent_lowat",
1055                 .data           = &init_net.ipv4.sysctl_tcp_notsent_lowat,
1056                 .maxlen         = sizeof(unsigned int),
1057                 .mode           = 0644,
1058                 .proc_handler   = proc_douintvec,
1059         },
1060         {
1061                 .procname       = "tcp_tw_reuse",
1062                 .data           = &init_net.ipv4.sysctl_tcp_tw_reuse,
1063                 .maxlen         = sizeof(int),
1064                 .mode           = 0644,
1065                 .proc_handler   = proc_dointvec
1066         },
1067         {
1068                 .procname       = "tcp_max_tw_buckets",
1069                 .data           = &init_net.ipv4.tcp_death_row.sysctl_max_tw_buckets,
1070                 .maxlen         = sizeof(int),
1071                 .mode           = 0644,
1072                 .proc_handler   = proc_dointvec
1073         },
1074         {
1075                 .procname       = "tcp_max_syn_backlog",
1076                 .data           = &init_net.ipv4.sysctl_max_syn_backlog,
1077                 .maxlen         = sizeof(int),
1078                 .mode           = 0644,
1079                 .proc_handler   = proc_dointvec
1080         },
1081         {
1082                 .procname       = "tcp_fastopen",
1083                 .data           = &init_net.ipv4.sysctl_tcp_fastopen,
1084                 .maxlen         = sizeof(int),
1085                 .mode           = 0644,
1086                 .proc_handler   = proc_dointvec,
1087         },
1088 #ifdef CONFIG_IP_ROUTE_MULTIPATH
1089         {
1090                 .procname       = "fib_multipath_use_neigh",
1091                 .data           = &init_net.ipv4.sysctl_fib_multipath_use_neigh,
1092                 .maxlen         = sizeof(int),
1093                 .mode           = 0644,
1094                 .proc_handler   = proc_dointvec_minmax,
1095                 .extra1         = &zero,
1096                 .extra2         = &one,
1097         },
1098         {
1099                 .procname       = "fib_multipath_hash_policy",
1100                 .data           = &init_net.ipv4.sysctl_fib_multipath_hash_policy,
1101                 .maxlen         = sizeof(int),
1102                 .mode           = 0644,
1103                 .proc_handler   = proc_dointvec_minmax,
1104                 .extra1         = &zero,
1105                 .extra2         = &one,
1106         },
1107 #endif
1108         {
1109                 .procname       = "ip_unprivileged_port_start",
1110                 .maxlen         = sizeof(int),
1111                 .data           = &init_net.ipv4.sysctl_ip_prot_sock,
1112                 .mode           = 0644,
1113                 .proc_handler   = ipv4_privileged_ports,
1114         },
1115 #ifdef CONFIG_NET_L3_MASTER_DEV
1116         {
1117                 .procname       = "udp_l3mdev_accept",
1118                 .data           = &init_net.ipv4.sysctl_udp_l3mdev_accept,
1119                 .maxlen         = sizeof(int),
1120                 .mode           = 0644,
1121                 .proc_handler   = proc_dointvec_minmax,
1122                 .extra1         = &zero,
1123                 .extra2         = &one,
1124         },
1125 #endif
1126         {
1127                 .procname       = "tcp_sack",
1128                 .data           = &init_net.ipv4.sysctl_tcp_sack,
1129                 .maxlen         = sizeof(int),
1130                 .mode           = 0644,
1131                 .proc_handler   = proc_dointvec
1132         },
1133         {
1134                 .procname       = "tcp_window_scaling",
1135                 .data           = &init_net.ipv4.sysctl_tcp_window_scaling,
1136                 .maxlen         = sizeof(int),
1137                 .mode           = 0644,
1138                 .proc_handler   = proc_dointvec
1139         },
1140         {
1141                 .procname       = "tcp_timestamps",
1142                 .data           = &init_net.ipv4.sysctl_tcp_timestamps,
1143                 .maxlen         = sizeof(int),
1144                 .mode           = 0644,
1145                 .proc_handler   = proc_dointvec
1146         },
1147         { }
1148 };
1149
1150 static __net_init int ipv4_sysctl_init_net(struct net *net)
1151 {
1152         struct ctl_table *table;
1153
1154         table = ipv4_net_table;
1155         if (!net_eq(net, &init_net)) {
1156                 int i;
1157
1158                 table = kmemdup(table, sizeof(ipv4_net_table), GFP_KERNEL);
1159                 if (!table)
1160                         goto err_alloc;
1161
1162                 /* Update the variables to point into the current struct net */
1163                 for (i = 0; i < ARRAY_SIZE(ipv4_net_table) - 1; i++)
1164                         table[i].data += (void *)net - (void *)&init_net;
1165         }
1166
1167         net->ipv4.ipv4_hdr = register_net_sysctl(net, "net/ipv4", table);
1168         if (!net->ipv4.ipv4_hdr)
1169                 goto err_reg;
1170
1171         net->ipv4.sysctl_local_reserved_ports = kzalloc(65536 / 8, GFP_KERNEL);
1172         if (!net->ipv4.sysctl_local_reserved_ports)
1173                 goto err_ports;
1174
1175         return 0;
1176
1177 err_ports:
1178         unregister_net_sysctl_table(net->ipv4.ipv4_hdr);
1179 err_reg:
1180         if (!net_eq(net, &init_net))
1181                 kfree(table);
1182 err_alloc:
1183         return -ENOMEM;
1184 }
1185
1186 static __net_exit void ipv4_sysctl_exit_net(struct net *net)
1187 {
1188         struct ctl_table *table;
1189
1190         kfree(net->ipv4.sysctl_local_reserved_ports);
1191         table = net->ipv4.ipv4_hdr->ctl_table_arg;
1192         unregister_net_sysctl_table(net->ipv4.ipv4_hdr);
1193         kfree(table);
1194 }
1195
1196 static __net_initdata struct pernet_operations ipv4_sysctl_ops = {
1197         .init = ipv4_sysctl_init_net,
1198         .exit = ipv4_sysctl_exit_net,
1199 };
1200
1201 static __init int sysctl_ipv4_init(void)
1202 {
1203         struct ctl_table_header *hdr;
1204
1205         hdr = register_net_sysctl(&init_net, "net/ipv4", ipv4_table);
1206         if (!hdr)
1207                 return -ENOMEM;
1208
1209         if (register_pernet_subsys(&ipv4_sysctl_ops)) {
1210                 unregister_net_sysctl_table(hdr);
1211                 return -ENOMEM;
1212         }
1213
1214         return 0;
1215 }
1216
1217 __initcall(sysctl_ipv4_init);