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 *table, int write,
 255                                  void __user *buffer, size_t *lenp,
 256                                  loff_t *ppos)
 257 {
 258         struct net *net = container_of(table->data, struct net,
 259             ipv4.sysctl_tcp_fastopen);
 260         struct ctl_table tbl = { .maxlen = (TCP_FASTOPEN_KEY_LENGTH * 2 + 10) };
 261         struct tcp_fastopen_context *ctxt;
 262         int ret;
 263         u32  user_key[4]; /* 16 bytes, matching TCP_FASTOPEN_KEY_LENGTH */
 264
 265         tbl.data = kmalloc(tbl.maxlen, GFP_KERNEL);
 266         if (!tbl.data)
 267                 return -ENOMEM;
 268
 269         rcu_read_lock();
 270         ctxt = rcu_dereference(net->ipv4.tcp_fastopen_ctx);
 271         if (ctxt)
 272                 memcpy(user_key, ctxt->key, TCP_FASTOPEN_KEY_LENGTH);
 273         else
 274                 memset(user_key, 0, sizeof(user_key));
 275         rcu_read_unlock();
 276
 277         snprintf(tbl.data, tbl.maxlen, "%08x-%08x-%08x-%08x",
 278                 user_key[0], user_key[1], user_key[2], user_key[3]);
 279         ret = proc_dostring(&tbl, write, buffer, lenp, ppos);
 280
 281         if (write && ret == 0) {
 282                 if (sscanf(tbl.data, "%x-%x-%x-%x", user_key, user_key + 1,
 283                            user_key + 2, user_key + 3) != 4) {
 284                         ret = -EINVAL;
 285                         goto bad_key;
 286                 }
 287                 tcp_fastopen_reset_cipher(net, NULL, user_key,
 288                                           TCP_FASTOPEN_KEY_LENGTH);
 289         }
 290
 291 bad_key:
 292         pr_debug("proc FO key set 0x%x-%x-%x-%x <- 0x%s: %u\n",
 293                user_key[0], user_key[1], user_key[2], user_key[3],
 294                (char *)tbl.data, ret);
 295         kfree(tbl.data);
 296         return ret;
 297 }
 298
 299 static void proc_configure_early_demux(int enabled, int protocol)
 300 {
 301         struct net_protocol *ipprot;
 302 #if IS_ENABLED(CONFIG_IPV6)
 303         struct inet6_protocol *ip6prot;
 304 #endif
 305
 306         rcu_read_lock();
 307
 308         ipprot = rcu_dereference(inet_protos[protocol]);
 309         if (ipprot)
 310                 ipprot->early_demux = enabled ? ipprot->early_demux_handler :
 311                                                 NULL;
 312
 313 #if IS_ENABLED(CONFIG_IPV6)
 314         ip6prot = rcu_dereference(inet6_protos[protocol]);
 315         if (ip6prot)
 316                 ip6prot->early_demux = enabled ? ip6prot->early_demux_handler :
 317                                                  NULL;
 318 #endif
 319         rcu_read_unlock();
 320 }
 321
 322 static int proc_tcp_early_demux(struct ctl_table *table, int write,
 323                                 void __user *buffer, size_t *lenp, loff_t *ppos)
 324 {
 325         int ret = 0;
 326
 327         ret = proc_dointvec(table, write, buffer, lenp, ppos);
 328
 329         if (write && !ret) {
 330                 int enabled = init_net.ipv4.sysctl_tcp_early_demux;
 331
 332                 proc_configure_early_demux(enabled, IPPROTO_TCP);
 333         }
 334
 335         return ret;
 336 }
 337
 338 static int proc_udp_early_demux(struct ctl_table *table, int write,
 339                                 void __user *buffer, size_t *lenp, loff_t *ppos)
 340 {
 341         int ret = 0;
 342
 343         ret = proc_dointvec(table, write, buffer, lenp, ppos);
 344
 345         if (write && !ret) {
 346                 int enabled = init_net.ipv4.sysctl_udp_early_demux;
 347
 348                 proc_configure_early_demux(enabled, IPPROTO_UDP);
 349         }
 350
 351         return ret;
 352 }
 353
 354 static int proc_tfo_blackhole_detect_timeout(struct ctl_table *table,
 355                                              int write,
 356                                              void __user *buffer,
 357                                              size_t *lenp, loff_t *ppos)
 358 {
 359         struct net *net = container_of(table->data, struct net,
 360             ipv4.sysctl_tcp_fastopen_blackhole_timeout);
 361         int ret;
 362
 363         ret = proc_dointvec_minmax(table, write, buffer, lenp, ppos);
 364         if (write && ret == 0)
 365                 atomic_set(&net->ipv4.tfo_active_disable_times, 0);
 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_max_orphans",
 391                 .data           = &sysctl_tcp_max_orphans,
 392                 .maxlen         = sizeof(int),
 393                 .mode           = 0644,
 394                 .proc_handler   = proc_dointvec
 395         },
 396         {
 397                 .procname       = "tcp_abort_on_overflow",
 398                 .data           = &sysctl_tcp_abort_on_overflow,
 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       = "inet_peer_threshold",
 412                 .data           = &inet_peer_threshold,
 413                 .maxlen         = sizeof(int),
 414                 .mode           = 0644,
 415                 .proc_handler   = proc_dointvec
 416         },
 417         {
 418                 .procname       = "inet_peer_minttl",
 419                 .data           = &inet_peer_minttl,
 420                 .maxlen         = sizeof(int),
 421                 .mode           = 0644,
 422                 .proc_handler   = proc_dointvec_jiffies,
 423         },
 424         {
 425                 .procname       = "inet_peer_maxttl",
 426                 .data           = &inet_peer_maxttl,
 427                 .maxlen         = sizeof(int),
 428                 .mode           = 0644,
 429                 .proc_handler   = proc_dointvec_jiffies,
 430         },
 431         {
 432                 .procname       = "tcp_fack",
 433                 .data           = &sysctl_tcp_fack,
 434                 .maxlen         = sizeof(int),
 435                 .mode           = 0644,
 436                 .proc_handler   = proc_dointvec
 437         },
 438         {
 439                 .procname       = "tcp_max_reordering",
 440                 .data           = &sysctl_tcp_max_reordering,
 441                 .maxlen         = sizeof(int),
 442                 .mode           = 0644,
 443                 .proc_handler   = proc_dointvec
 444         },
 445         {
 446                 .procname       = "tcp_dsack",
 447                 .data           = &sysctl_tcp_dsack,
 448                 .maxlen         = sizeof(int),
 449                 .mode           = 0644,
 450                 .proc_handler   = proc_dointvec
 451         },
 452         {
 453                 .procname       = "tcp_mem",
 454                 .maxlen         = sizeof(sysctl_tcp_mem),
 455                 .data           = &sysctl_tcp_mem,
 456                 .mode           = 0644,
 457                 .proc_handler   = proc_doulongvec_minmax,
 458         },
 459         {
 460                 .procname       = "tcp_wmem",
 461                 .data           = &sysctl_tcp_wmem,
 462                 .maxlen         = sizeof(sysctl_tcp_wmem),
 463                 .mode           = 0644,
 464                 .proc_handler   = proc_dointvec_minmax,
 465                 .extra1         = &one,
 466         },
 467         {
 468                 .procname       = "tcp_rmem",
 469                 .data           = &sysctl_tcp_rmem,
 470                 .maxlen         = sizeof(sysctl_tcp_rmem),
 471                 .mode           = 0644,
 472                 .proc_handler   = proc_dointvec_minmax,
 473                 .extra1         = &one,
 474         },
 475         {
 476                 .procname       = "tcp_app_win",
 477                 .data           = &sysctl_tcp_app_win,
 478                 .maxlen         = sizeof(int),
 479                 .mode           = 0644,
 480                 .proc_handler   = proc_dointvec
 481         },
 482         {
 483                 .procname       = "tcp_adv_win_scale",
 484                 .data           = &sysctl_tcp_adv_win_scale,
 485                 .maxlen         = sizeof(int),
 486                 .mode           = 0644,
 487                 .proc_handler   = proc_dointvec_minmax,
 488                 .extra1         = &tcp_adv_win_scale_min,
 489                 .extra2         = &tcp_adv_win_scale_max,
 490         },
 491         {
 492                 .procname       = "tcp_frto",
 493                 .data           = &sysctl_tcp_frto,
 494                 .maxlen         = sizeof(int),
 495                 .mode           = 0644,
 496                 .proc_handler   = proc_dointvec
 497         },
 498         {
 499                 .procname       = "tcp_min_rtt_wlen",
 500                 .data           = &sysctl_tcp_min_rtt_wlen,
 501                 .maxlen         = sizeof(int),
 502                 .mode           = 0644,
 503                 .proc_handler   = proc_dointvec
 504         },
 505         {
 506                 .procname       = "tcp_low_latency",
 507                 .data           = &sysctl_tcp_low_latency,
 508                 .maxlen         = sizeof(int),
 509                 .mode           = 0644,
 510                 .proc_handler   = proc_dointvec
 511         },
 512         {
 513                 .procname       = "tcp_no_metrics_save",
 514                 .data           = &sysctl_tcp_nometrics_save,
 515                 .maxlen         = sizeof(int),
 516                 .mode           = 0644,
 517                 .proc_handler   = proc_dointvec,
 518         },
 519         {
 520                 .procname       = "tcp_moderate_rcvbuf",
 521                 .data           = &sysctl_tcp_moderate_rcvbuf,
 522                 .maxlen         = sizeof(int),
 523                 .mode           = 0644,
 524                 .proc_handler   = proc_dointvec,
 525         },
 526         {
 527                 .procname       = "tcp_tso_win_divisor",
 528                 .data           = &sysctl_tcp_tso_win_divisor,
 529                 .maxlen         = sizeof(int),
 530                 .mode           = 0644,
 531                 .proc_handler   = proc_dointvec,
 532         },
 533         {
 534                 .procname       = "tcp_congestion_control",
 535                 .mode           = 0644,
 536                 .maxlen         = TCP_CA_NAME_MAX,
 537                 .proc_handler   = proc_tcp_congestion_control,
 538         },
 539         {
 540                 .procname       = "tcp_workaround_signed_windows",
 541                 .data           = &sysctl_tcp_workaround_signed_windows,
 542                 .maxlen         = sizeof(int),
 543                 .mode           = 0644,
 544                 .proc_handler   = proc_dointvec
 545         },
 546         {
 547                 .procname       = "tcp_limit_output_bytes",
 548                 .data           = &sysctl_tcp_limit_output_bytes,
 549                 .maxlen         = sizeof(int),
 550                 .mode           = 0644,
 551                 .proc_handler   = proc_dointvec
 552         },
 553         {
 554                 .procname       = "tcp_challenge_ack_limit",
 555                 .data           = &sysctl_tcp_challenge_ack_limit,
 556                 .maxlen         = sizeof(int),
 557                 .mode           = 0644,
 558                 .proc_handler   = proc_dointvec
 559         },
 560 #ifdef CONFIG_NETLABEL
 561         {
 562                 .procname       = "cipso_cache_enable",
 563                 .data           = &cipso_v4_cache_enabled,
 564                 .maxlen         = sizeof(int),
 565                 .mode           = 0644,
 566                 .proc_handler   = proc_dointvec,
 567         },
 568         {
 569                 .procname       = "cipso_cache_bucket_size",
 570                 .data           = &cipso_v4_cache_bucketsize,
 571                 .maxlen         = sizeof(int),
 572                 .mode           = 0644,
 573                 .proc_handler   = proc_dointvec,
 574         },
 575         {
 576                 .procname       = "cipso_rbm_optfmt",
 577                 .data           = &cipso_v4_rbm_optfmt,
 578                 .maxlen         = sizeof(int),
 579                 .mode           = 0644,
 580                 .proc_handler   = proc_dointvec,
 581         },
 582         {
 583                 .procname       = "cipso_rbm_strictvalid",
 584                 .data           = &cipso_v4_rbm_strictvalid,
 585                 .maxlen         = sizeof(int),
 586                 .mode           = 0644,
 587                 .proc_handler   = proc_dointvec,
 588         },
 589 #endif /* CONFIG_NETLABEL */
 590         {
 591                 .procname       = "tcp_available_congestion_control",
 592                 .maxlen         = TCP_CA_BUF_MAX,
 593                 .mode           = 0444,
 594                 .proc_handler   = proc_tcp_available_congestion_control,
 595         },
 596         {
 597                 .procname       = "tcp_allowed_congestion_control",
 598                 .maxlen         = TCP_CA_BUF_MAX,
 599                 .mode           = 0644,
 600                 .proc_handler   = proc_allowed_congestion_control,
 601         },
 602         {
 603                 .procname       = "tcp_min_tso_segs",
 604                 .data           = &sysctl_tcp_min_tso_segs,
 605                 .maxlen         = sizeof(int),
 606                 .mode           = 0644,
 607                 .proc_handler   = proc_dointvec_minmax,
 608                 .extra1         = &one,
 609                 .extra2         = &gso_max_segs,
 610         },
 611         {
 612                 .procname       = "tcp_pacing_ss_ratio",
 613                 .data           = &sysctl_tcp_pacing_ss_ratio,
 614                 .maxlen         = sizeof(int),
 615                 .mode           = 0644,
 616                 .proc_handler   = proc_dointvec_minmax,
 617                 .extra1         = &zero,
 618                 .extra2         = &thousand,
 619         },
 620         {
 621                 .procname       = "tcp_pacing_ca_ratio",
 622                 .data           = &sysctl_tcp_pacing_ca_ratio,
 623                 .maxlen         = sizeof(int),
 624                 .mode           = 0644,
 625                 .proc_handler   = proc_dointvec_minmax,
 626                 .extra1         = &zero,
 627                 .extra2         = &thousand,
 628         },
 629         {
 630                 .procname       = "tcp_autocorking",
 631                 .data           = &sysctl_tcp_autocorking,
 632                 .maxlen         = sizeof(int),
 633                 .mode           = 0644,
 634                 .proc_handler   = proc_dointvec_minmax,
 635                 .extra1         = &zero,
 636                 .extra2         = &one,
 637         },
 638         {
 639                 .procname       = "tcp_invalid_ratelimit",
 640                 .data           = &sysctl_tcp_invalid_ratelimit,
 641                 .maxlen         = sizeof(int),
 642                 .mode           = 0644,
 643                 .proc_handler   = proc_dointvec_ms_jiffies,
 644         },
 645         {
 646                 .procname       = "tcp_available_ulp",
 647                 .maxlen         = TCP_ULP_BUF_MAX,
 648                 .mode           = 0444,
 649                 .proc_handler   = proc_tcp_available_ulp,
 650         },
 651         {
 652                 .procname       = "icmp_msgs_per_sec",
 653                 .data           = &sysctl_icmp_msgs_per_sec,
 654                 .maxlen         = sizeof(int),
 655                 .mode           = 0644,
 656                 .proc_handler   = proc_dointvec_minmax,
 657                 .extra1         = &zero,
 658         },
 659         {
 660                 .procname       = "icmp_msgs_burst",
 661                 .data           = &sysctl_icmp_msgs_burst,
 662                 .maxlen         = sizeof(int),
 663                 .mode           = 0644,
 664                 .proc_handler   = proc_dointvec_minmax,
 665                 .extra1         = &zero,
 666         },
 667         {
 668                 .procname       = "udp_mem",
 669                 .data           = &sysctl_udp_mem,
 670                 .maxlen         = sizeof(sysctl_udp_mem),
 671                 .mode           = 0644,
 672                 .proc_handler   = proc_doulongvec_minmax,
 673         },
 674         {
 675                 .procname       = "udp_rmem_min",
 676                 .data           = &sysctl_udp_rmem_min,
 677                 .maxlen         = sizeof(sysctl_udp_rmem_min),
 678                 .mode           = 0644,
 679                 .proc_handler   = proc_dointvec_minmax,
 680                 .extra1         = &one
 681         },
 682         {
 683                 .procname       = "udp_wmem_min",
 684                 .data           = &sysctl_udp_wmem_min,
 685                 .maxlen         = sizeof(sysctl_udp_wmem_min),
 686                 .mode           = 0644,
 687                 .proc_handler   = proc_dointvec_minmax,
 688                 .extra1         = &one
 689         },
 690         { }
 691 };
 692
 693 static struct ctl_table ipv4_net_table[] = {
 694         {
 695                 .procname       = "icmp_echo_ignore_all",
 696                 .data           = &init_net.ipv4.sysctl_icmp_echo_ignore_all,
 697                 .maxlen         = sizeof(int),
 698                 .mode           = 0644,
 699                 .proc_handler   = proc_dointvec
 700         },
 701         {
 702                 .procname       = "icmp_echo_ignore_broadcasts",
 703                 .data           = &init_net.ipv4.sysctl_icmp_echo_ignore_broadcasts,
 704                 .maxlen         = sizeof(int),
 705                 .mode           = 0644,
 706                 .proc_handler   = proc_dointvec
 707         },
 708         {
 709                 .procname       = "icmp_ignore_bogus_error_responses",
 710                 .data           = &init_net.ipv4.sysctl_icmp_ignore_bogus_error_responses,
 711                 .maxlen         = sizeof(int),
 712                 .mode           = 0644,
 713                 .proc_handler   = proc_dointvec
 714         },
 715         {
 716                 .procname       = "icmp_errors_use_inbound_ifaddr",
 717                 .data           = &init_net.ipv4.sysctl_icmp_errors_use_inbound_ifaddr,
 718                 .maxlen         = sizeof(int),
 719                 .mode           = 0644,
 720                 .proc_handler   = proc_dointvec
 721         },
 722         {
 723                 .procname       = "icmp_ratelimit",
 724                 .data           = &init_net.ipv4.sysctl_icmp_ratelimit,
 725                 .maxlen         = sizeof(int),
 726                 .mode           = 0644,
 727                 .proc_handler   = proc_dointvec_ms_jiffies,
 728         },
 729         {
 730                 .procname       = "icmp_ratemask",
 731                 .data           = &init_net.ipv4.sysctl_icmp_ratemask,
 732                 .maxlen         = sizeof(int),
 733                 .mode           = 0644,
 734                 .proc_handler   = proc_dointvec
 735         },
 736         {
 737                 .procname       = "ping_group_range",
 738                 .data           = &init_net.ipv4.ping_group_range.range,
 739                 .maxlen         = sizeof(gid_t)*2,
 740                 .mode           = 0644,
 741                 .proc_handler   = ipv4_ping_group_range,
 742         },
 743         {
 744                 .procname       = "tcp_ecn",
 745                 .data           = &init_net.ipv4.sysctl_tcp_ecn,
 746                 .maxlen         = sizeof(int),
 747                 .mode           = 0644,
 748                 .proc_handler   = proc_dointvec
 749         },
 750         {
 751                 .procname       = "tcp_ecn_fallback",
 752                 .data           = &init_net.ipv4.sysctl_tcp_ecn_fallback,
 753                 .maxlen         = sizeof(int),
 754                 .mode           = 0644,
 755                 .proc_handler   = proc_dointvec
 756         },
 757         {
 758                 .procname       = "ip_dynaddr",
 759                 .data           = &init_net.ipv4.sysctl_ip_dynaddr,
 760                 .maxlen         = sizeof(int),
 761                 .mode           = 0644,
 762                 .proc_handler   = proc_dointvec
 763         },
 764         {
 765                 .procname       = "ip_early_demux",
 766                 .data           = &init_net.ipv4.sysctl_ip_early_demux,
 767                 .maxlen         = sizeof(int),
 768                 .mode           = 0644,
 769                 .proc_handler   = proc_dointvec
 770         },
 771         {
 772                 .procname       = "udp_early_demux",
 773                 .data           = &init_net.ipv4.sysctl_udp_early_demux,
 774                 .maxlen         = sizeof(int),
 775                 .mode           = 0644,
 776                 .proc_handler   = proc_udp_early_demux
 777         },
 778         {
 779                 .procname       = "tcp_early_demux",
 780                 .data           = &init_net.ipv4.sysctl_tcp_early_demux,
 781                 .maxlen         = sizeof(int),
 782                 .mode           = 0644,
 783                 .proc_handler   = proc_tcp_early_demux
 784         },
 785         {
 786                 .procname       = "ip_default_ttl",
 787                 .data           = &init_net.ipv4.sysctl_ip_default_ttl,
 788                 .maxlen         = sizeof(int),
 789                 .mode           = 0644,
 790                 .proc_handler   = proc_dointvec_minmax,
 791                 .extra1         = &ip_ttl_min,
 792                 .extra2         = &ip_ttl_max,
 793         },
 794         {
 795                 .procname       = "ip_local_port_range",
 796                 .maxlen         = sizeof(init_net.ipv4.ip_local_ports.range),
 797                 .data           = &init_net.ipv4.ip_local_ports.range,
 798                 .mode           = 0644,
 799                 .proc_handler   = ipv4_local_port_range,
 800         },
 801         {
 802                 .procname       = "ip_local_reserved_ports",
 803                 .data           = &init_net.ipv4.sysctl_local_reserved_ports,
 804                 .maxlen         = 65536,
 805                 .mode           = 0644,
 806                 .proc_handler   = proc_do_large_bitmap,
 807         },
 808         {
 809                 .procname       = "ip_no_pmtu_disc",
 810                 .data           = &init_net.ipv4.sysctl_ip_no_pmtu_disc,
 811                 .maxlen         = sizeof(int),
 812                 .mode           = 0644,
 813                 .proc_handler   = proc_dointvec
 814         },
 815         {
 816                 .procname       = "ip_forward_use_pmtu",
 817                 .data           = &init_net.ipv4.sysctl_ip_fwd_use_pmtu,
 818                 .maxlen         = sizeof(int),
 819                 .mode           = 0644,
 820                 .proc_handler   = proc_dointvec,
 821         },
 822         {
 823                 .procname       = "ip_nonlocal_bind",
 824                 .data           = &init_net.ipv4.sysctl_ip_nonlocal_bind,
 825                 .maxlen         = sizeof(int),
 826                 .mode           = 0644,
 827                 .proc_handler   = proc_dointvec
 828         },
 829         {
 830                 .procname       = "fwmark_reflect",
 831                 .data           = &init_net.ipv4.sysctl_fwmark_reflect,
 832                 .maxlen         = sizeof(int),
 833                 .mode           = 0644,
 834                 .proc_handler   = proc_dointvec,
 835         },
 836         {
 837                 .procname       = "tcp_fwmark_accept",
 838                 .data           = &init_net.ipv4.sysctl_tcp_fwmark_accept,
 839                 .maxlen         = sizeof(int),
 840                 .mode           = 0644,
 841                 .proc_handler   = proc_dointvec,
 842         },
 843 #ifdef CONFIG_NET_L3_MASTER_DEV
 844         {
 845                 .procname       = "tcp_l3mdev_accept",
 846                 .data           = &init_net.ipv4.sysctl_tcp_l3mdev_accept,
 847                 .maxlen         = sizeof(int),
 848                 .mode           = 0644,
 849                 .proc_handler   = proc_dointvec_minmax,
 850                 .extra1         = &zero,
 851                 .extra2         = &one,
 852         },
 853 #endif
 854         {
 855                 .procname       = "tcp_mtu_probing",
 856                 .data           = &init_net.ipv4.sysctl_tcp_mtu_probing,
 857                 .maxlen         = sizeof(int),
 858                 .mode           = 0644,
 859                 .proc_handler   = proc_dointvec,
 860         },
 861         {
 862                 .procname       = "tcp_base_mss",
 863                 .data           = &init_net.ipv4.sysctl_tcp_base_mss,
 864                 .maxlen         = sizeof(int),
 865                 .mode           = 0644,
 866                 .proc_handler   = proc_dointvec,
 867         },
 868         {
 869                 .procname       = "tcp_probe_threshold",
 870                 .data           = &init_net.ipv4.sysctl_tcp_probe_threshold,
 871                 .maxlen         = sizeof(int),
 872                 .mode           = 0644,
 873                 .proc_handler   = proc_dointvec,
 874         },
 875         {
 876                 .procname       = "tcp_probe_interval",
 877                 .data           = &init_net.ipv4.sysctl_tcp_probe_interval,
 878                 .maxlen         = sizeof(int),
 879                 .mode           = 0644,
 880                 .proc_handler   = proc_dointvec,
 881         },
 882         {
 883                 .procname       = "igmp_link_local_mcast_reports",
 884                 .data           = &init_net.ipv4.sysctl_igmp_llm_reports,
 885                 .maxlen         = sizeof(int),
 886                 .mode           = 0644,
 887                 .proc_handler   = proc_dointvec
 888         },
 889         {
 890                 .procname       = "igmp_max_memberships",
 891                 .data           = &init_net.ipv4.sysctl_igmp_max_memberships,
 892                 .maxlen         = sizeof(int),
 893                 .mode           = 0644,
 894                 .proc_handler   = proc_dointvec
 895         },
 896         {
 897                 .procname       = "igmp_max_msf",
 898                 .data           = &init_net.ipv4.sysctl_igmp_max_msf,
 899                 .maxlen         = sizeof(int),
 900                 .mode           = 0644,
 901                 .proc_handler   = proc_dointvec
 902         },
 903 #ifdef CONFIG_IP_MULTICAST
 904         {
 905                 .procname       = "igmp_qrv",
 906                 .data           = &init_net.ipv4.sysctl_igmp_qrv,
 907                 .maxlen         = sizeof(int),
 908                 .mode           = 0644,
 909                 .proc_handler   = proc_dointvec_minmax,
 910                 .extra1         = &one
 911         },
 912 #endif
 913         {
 914                 .procname       = "tcp_keepalive_time",
 915                 .data           = &init_net.ipv4.sysctl_tcp_keepalive_time,
 916                 .maxlen         = sizeof(int),
 917                 .mode           = 0644,
 918                 .proc_handler   = proc_dointvec_jiffies,
 919         },
 920         {
 921                 .procname       = "tcp_keepalive_probes",
 922                 .data           = &init_net.ipv4.sysctl_tcp_keepalive_probes,
 923                 .maxlen         = sizeof(int),
 924                 .mode           = 0644,
 925                 .proc_handler   = proc_dointvec
 926         },
 927         {
 928                 .procname       = "tcp_keepalive_intvl",
 929                 .data           = &init_net.ipv4.sysctl_tcp_keepalive_intvl,
 930                 .maxlen         = sizeof(int),
 931                 .mode           = 0644,
 932                 .proc_handler   = proc_dointvec_jiffies,
 933         },
 934         {
 935                 .procname       = "tcp_syn_retries",
 936                 .data           = &init_net.ipv4.sysctl_tcp_syn_retries,
 937                 .maxlen         = sizeof(int),
 938                 .mode           = 0644,
 939                 .proc_handler   = proc_dointvec_minmax,
 940                 .extra1         = &tcp_syn_retries_min,
 941                 .extra2         = &tcp_syn_retries_max
 942         },
 943         {
 944                 .procname       = "tcp_synack_retries",
 945                 .data           = &init_net.ipv4.sysctl_tcp_synack_retries,
 946                 .maxlen         = sizeof(int),
 947                 .mode           = 0644,
 948                 .proc_handler   = proc_dointvec
 949         },
 950 #ifdef CONFIG_SYN_COOKIES
 951         {
 952                 .procname       = "tcp_syncookies",
 953                 .data           = &init_net.ipv4.sysctl_tcp_syncookies,
 954                 .maxlen         = sizeof(int),
 955                 .mode           = 0644,
 956                 .proc_handler   = proc_dointvec
 957         },
 958 #endif
 959         {
 960                 .procname       = "tcp_reordering",
 961                 .data           = &init_net.ipv4.sysctl_tcp_reordering,
 962                 .maxlen         = sizeof(int),
 963                 .mode           = 0644,
 964                 .proc_handler   = proc_dointvec
 965         },
 966         {
 967                 .procname       = "tcp_retries1",
 968                 .data           = &init_net.ipv4.sysctl_tcp_retries1,
 969                 .maxlen         = sizeof(int),
 970                 .mode           = 0644,
 971                 .proc_handler   = proc_dointvec_minmax,
 972                 .extra2         = &tcp_retr1_max
 973         },
 974         {
 975                 .procname       = "tcp_retries2",
 976                 .data           = &init_net.ipv4.sysctl_tcp_retries2,
 977                 .maxlen         = sizeof(int),
 978                 .mode           = 0644,
 979                 .proc_handler   = proc_dointvec
 980         },
 981         {
 982                 .procname       = "tcp_orphan_retries",
 983                 .data           = &init_net.ipv4.sysctl_tcp_orphan_retries,
 984                 .maxlen         = sizeof(int),
 985                 .mode           = 0644,
 986                 .proc_handler   = proc_dointvec
 987         },
 988         {
 989                 .procname       = "tcp_fin_timeout",
 990                 .data           = &init_net.ipv4.sysctl_tcp_fin_timeout,
 991                 .maxlen         = sizeof(int),
 992                 .mode           = 0644,
 993                 .proc_handler   = proc_dointvec_jiffies,
 994         },
 995         {
 996                 .procname       = "tcp_notsent_lowat",
 997                 .data           = &init_net.ipv4.sysctl_tcp_notsent_lowat,
 998                 .maxlen         = sizeof(unsigned int),
 999                 .mode           = 0644,
1000                 .proc_handler   = proc_douintvec,
1001         },
1002         {
1003                 .procname       = "tcp_tw_reuse",
1004                 .data           = &init_net.ipv4.sysctl_tcp_tw_reuse,
1005                 .maxlen         = sizeof(int),
1006                 .mode           = 0644,
1007                 .proc_handler   = proc_dointvec
1008         },
1009         {
1010                 .procname       = "tcp_max_tw_buckets",
1011                 .data           = &init_net.ipv4.tcp_death_row.sysctl_max_tw_buckets,
1012                 .maxlen         = sizeof(int),
1013                 .mode           = 0644,
1014                 .proc_handler   = proc_dointvec
1015         },
1016         {
1017                 .procname       = "tcp_max_syn_backlog",
1018                 .data           = &init_net.ipv4.sysctl_max_syn_backlog,
1019                 .maxlen         = sizeof(int),
1020                 .mode           = 0644,
1021                 .proc_handler   = proc_dointvec
1022         },
1023         {
1024                 .procname       = "tcp_fastopen",
1025                 .data           = &init_net.ipv4.sysctl_tcp_fastopen,
1026                 .maxlen         = sizeof(int),
1027                 .mode           = 0644,
1028                 .proc_handler   = proc_dointvec,
1029         },
1030         {
1031                 .procname       = "tcp_fastopen_key",
1032                 .mode           = 0600,
1033                 .data           = &init_net.ipv4.sysctl_tcp_fastopen,
1034                 .maxlen         = ((TCP_FASTOPEN_KEY_LENGTH * 2) + 10),
1035                 .proc_handler   = proc_tcp_fastopen_key,
1036         },
1037         {
1038                 .procname       = "tcp_fastopen_blackhole_timeout_sec",
1039                 .data           = &init_net.ipv4.sysctl_tcp_fastopen_blackhole_timeout,
1040                 .maxlen         = sizeof(int),
1041                 .mode           = 0644,
1042                 .proc_handler   = proc_tfo_blackhole_detect_timeout,
1043                 .extra1         = &zero,
1044         },
1045 #ifdef CONFIG_IP_ROUTE_MULTIPATH
1046         {
1047                 .procname       = "fib_multipath_use_neigh",
1048                 .data           = &init_net.ipv4.sysctl_fib_multipath_use_neigh,
1049                 .maxlen         = sizeof(int),
1050                 .mode           = 0644,
1051                 .proc_handler   = proc_dointvec_minmax,
1052                 .extra1         = &zero,
1053                 .extra2         = &one,
1054         },
1055         {
1056                 .procname       = "fib_multipath_hash_policy",
1057                 .data           = &init_net.ipv4.sysctl_fib_multipath_hash_policy,
1058                 .maxlen         = sizeof(int),
1059                 .mode           = 0644,
1060                 .proc_handler   = proc_dointvec_minmax,
1061                 .extra1         = &zero,
1062                 .extra2         = &one,
1063         },
1064 #endif
1065         {
1066                 .procname       = "ip_unprivileged_port_start",
1067                 .maxlen         = sizeof(int),
1068                 .data           = &init_net.ipv4.sysctl_ip_prot_sock,
1069                 .mode           = 0644,
1070                 .proc_handler   = ipv4_privileged_ports,
1071         },
1072 #ifdef CONFIG_NET_L3_MASTER_DEV
1073         {
1074                 .procname       = "udp_l3mdev_accept",
1075                 .data           = &init_net.ipv4.sysctl_udp_l3mdev_accept,
1076                 .maxlen         = sizeof(int),
1077                 .mode           = 0644,
1078                 .proc_handler   = proc_dointvec_minmax,
1079                 .extra1         = &zero,
1080                 .extra2         = &one,
1081         },
1082 #endif
1083         {
1084                 .procname       = "tcp_sack",
1085                 .data           = &init_net.ipv4.sysctl_tcp_sack,
1086                 .maxlen         = sizeof(int),
1087                 .mode           = 0644,
1088                 .proc_handler   = proc_dointvec
1089         },
1090         {
1091                 .procname       = "tcp_window_scaling",
1092                 .data           = &init_net.ipv4.sysctl_tcp_window_scaling,
1093                 .maxlen         = sizeof(int),
1094                 .mode           = 0644,
1095                 .proc_handler   = proc_dointvec
1096         },
1097         {
1098                 .procname       = "tcp_timestamps",
1099                 .data           = &init_net.ipv4.sysctl_tcp_timestamps,
1100                 .maxlen         = sizeof(int),
1101                 .mode           = 0644,
1102                 .proc_handler   = proc_dointvec
1103         },
1104         {
1105                 .procname       = "tcp_early_retrans",
1106                 .data           = &init_net.ipv4.sysctl_tcp_early_retrans,
1107                 .maxlen         = sizeof(int),
1108                 .mode           = 0644,
1109                 .proc_handler   = proc_dointvec_minmax,
1110                 .extra1         = &zero,
1111                 .extra2         = &four,
1112         },
1113         {
1114                 .procname       = "tcp_recovery",
1115                 .data           = &init_net.ipv4.sysctl_tcp_recovery,
1116                 .maxlen         = sizeof(int),
1117                 .mode           = 0644,
1118                 .proc_handler   = proc_dointvec,
1119         },
1120         {
1121                 .procname       = "tcp_thin_linear_timeouts",
1122                 .data           = &init_net.ipv4.sysctl_tcp_thin_linear_timeouts,
1123                 .maxlen         = sizeof(int),
1124                 .mode           = 0644,
1125                 .proc_handler   = proc_dointvec
1126         },
1127         {
1128                 .procname       = "tcp_slow_start_after_idle",
1129                 .data           = &init_net.ipv4.sysctl_tcp_slow_start_after_idle,
1130                 .maxlen         = sizeof(int),
1131                 .mode           = 0644,
1132                 .proc_handler   = proc_dointvec
1133         },
1134         {
1135                 .procname       = "tcp_retrans_collapse",
1136                 .data           = &init_net.ipv4.sysctl_tcp_retrans_collapse,
1137                 .maxlen         = sizeof(int),
1138                 .mode           = 0644,
1139                 .proc_handler   = proc_dointvec
1140         },
1141         {
1142                 .procname       = "tcp_stdurg",
1143                 .data           = &init_net.ipv4.sysctl_tcp_stdurg,
1144                 .maxlen         = sizeof(int),
1145                 .mode           = 0644,
1146                 .proc_handler   = proc_dointvec
1147         },
1148         { }
1149 };
1150
1151 static __net_init int ipv4_sysctl_init_net(struct net *net)
1152 {
1153         struct ctl_table *table;
1154
1155         table = ipv4_net_table;
1156         if (!net_eq(net, &init_net)) {
1157                 int i;
1158
1159                 table = kmemdup(table, sizeof(ipv4_net_table), GFP_KERNEL);
1160                 if (!table)
1161                         goto err_alloc;
1162
1163                 /* Update the variables to point into the current struct net */
1164                 for (i = 0; i < ARRAY_SIZE(ipv4_net_table) - 1; i++)
1165                         table[i].data += (void *)net - (void *)&init_net;
1166         }
1167
1168         net->ipv4.ipv4_hdr = register_net_sysctl(net, "net/ipv4", table);
1169         if (!net->ipv4.ipv4_hdr)
1170                 goto err_reg;
1171
1172         net->ipv4.sysctl_local_reserved_ports = kzalloc(65536 / 8, GFP_KERNEL);
1173         if (!net->ipv4.sysctl_local_reserved_ports)
1174                 goto err_ports;
1175
1176         return 0;
1177
1178 err_ports:
1179         unregister_net_sysctl_table(net->ipv4.ipv4_hdr);
1180 err_reg:
1181         if (!net_eq(net, &init_net))
1182                 kfree(table);
1183 err_alloc:
1184         return -ENOMEM;
1185 }
1186
1187 static __net_exit void ipv4_sysctl_exit_net(struct net *net)
1188 {
1189         struct ctl_table *table;
1190
1191         kfree(net->ipv4.sysctl_local_reserved_ports);
1192         table = net->ipv4.ipv4_hdr->ctl_table_arg;
1193         unregister_net_sysctl_table(net->ipv4.ipv4_hdr);
1194         kfree(table);
1195 }
1196
1197 static __net_initdata struct pernet_operations ipv4_sysctl_ops = {
1198         .init = ipv4_sysctl_init_net,
1199         .exit = ipv4_sysctl_exit_net,
1200 };
1201
1202 static __init int sysctl_ipv4_init(void)
1203 {
1204         struct ctl_table_header *hdr;
1205
1206         hdr = register_net_sysctl(&init_net, "net/ipv4", ipv4_table);
1207         if (!hdr)
1208                 return -ENOMEM;
1209
1210         if (register_pernet_subsys(&ipv4_sysctl_ops)) {
1211                 unregister_net_sysctl_table(hdr);
1212                 return -ENOMEM;
1213         }
1214
1215         return 0;
1216 }
1217
1218 __initcall(sysctl_ipv4_init);