net/ipv4/sysctl_net_ipv4.c

   1 // SPDX-License-Identifier: GPL-2.0
   2 /*
   3  * sysctl_net_ipv4.c: sysctl interface to net IPV4 subsystem.
   4  *
   5  * Begun April 1, 1996, Mike Shaver.
   6  * Added /proc/sys/net/ipv4 directory entry (empty =) ). [MS]
   7  */
   8
   9 #include <linux/mm.h>
  10 #include <linux/module.h>
  11 #include <linux/sysctl.h>
  12 #include <linux/igmp.h>
  13 #include <linux/inetdevice.h>
  14 #include <linux/seqlock.h>
  15 #include <linux/init.h>
  16 #include <linux/slab.h>
  17 #include <linux/nsproxy.h>
  18 #include <linux/swap.h>
  19 #include <net/snmp.h>
  20 #include <net/icmp.h>
  21 #include <net/ip.h>
  22 #include <net/route.h>
  23 #include <net/tcp.h>
  24 #include <net/udp.h>
  25 #include <net/cipso_ipv4.h>
  26 #include <net/inet_frag.h>
  27 #include <net/ping.h>
  28 #include <net/protocol.h>
  29
  30 static int zero;
  31 static int one = 1;
  32 static int four = 4;
  33 static int thousand = 1000;
  34 static int gso_max_segs = GSO_MAX_SEGS;
  35 static int tcp_retr1_max = 255;
  36 static int ip_local_port_range_min[] = { 1, 1 };
  37 static int ip_local_port_range_max[] = { 65535, 65535 };
  38 static int tcp_adv_win_scale_min = -31;
  39 static int tcp_adv_win_scale_max = 31;
  40 static int ip_privileged_port_min;
  41 static int ip_privileged_port_max = 65535;
  42 static int ip_ttl_min = 1;
  43 static int ip_ttl_max = 255;
  44 static int tcp_syn_retries_min = 1;
  45 static int tcp_syn_retries_max = MAX_TCP_SYNCNT;
  46 static int ip_ping_group_range_min[] = { 0, 0 };
  47 static int ip_ping_group_range_max[] = { GID_T_MAX, GID_T_MAX };
  48
  49 /* obsolete */
  50 static int sysctl_tcp_low_latency __read_mostly;
  51
  52 /* Update system visible IP port range */
  53 static void set_local_port_range(struct net *net, int range[2])
  54 {
  55         bool same_parity = !((range[0] ^ range[1]) & 1);
  56
  57         write_seqlock_bh(&net->ipv4.ip_local_ports.lock);
  58         if (same_parity && !net->ipv4.ip_local_ports.warned) {
  59                 net->ipv4.ip_local_ports.warned = true;
  60                 pr_err_ratelimited("ip_local_port_range: prefer different parity for start/end values.\n");
  61         }
  62         net->ipv4.ip_local_ports.range[0] = range[0];
  63         net->ipv4.ip_local_ports.range[1] = range[1];
  64         write_sequnlock_bh(&net->ipv4.ip_local_ports.lock);
  65 }
  66
  67 /* Validate changes from /proc interface. */
  68 static int ipv4_local_port_range(struct ctl_table *table, int write,
  69                                  void __user *buffer,
  70                                  size_t *lenp, loff_t *ppos)
  71 {
  72         struct net *net =
  73                 container_of(table->data, struct net, ipv4.ip_local_ports.range);
  74         int ret;
  75         int range[2];
  76         struct ctl_table tmp = {
  77                 .data = &range,
  78                 .maxlen = sizeof(range),
  79                 .mode = table->mode,
  80                 .extra1 = &ip_local_port_range_min,
  81                 .extra2 = &ip_local_port_range_max,
  82         };
  83
  84         inet_get_local_port_range(net, &range[0], &range[1]);
  85
  86         ret = proc_dointvec_minmax(&tmp, write, buffer, lenp, ppos);
  87
  88         if (write && ret == 0) {
  89                 /* Ensure that the upper limit is not smaller than the lower,
  90                  * and that the lower does not encroach upon the privileged
  91                  * port limit.
  92                  */
  93                 if ((range[1] < range[0]) ||
  94                     (range[0] < net->ipv4.sysctl_ip_prot_sock))
  95                         ret = -EINVAL;
  96                 else
  97                         set_local_port_range(net, range);
  98         }
  99
 100         return ret;
 101 }
 102
 103 /* Validate changes from /proc interface. */
 104 static int ipv4_privileged_ports(struct ctl_table *table, int write,
 105                                 void __user *buffer, size_t *lenp, loff_t *ppos)
 106 {
 107         struct net *net = container_of(table->data, struct net,
 108             ipv4.sysctl_ip_prot_sock);
 109         int ret;
 110         int pports;
 111         int range[2];
 112         struct ctl_table tmp = {
 113                 .data = &pports,
 114                 .maxlen = sizeof(pports),
 115                 .mode = table->mode,
 116                 .extra1 = &ip_privileged_port_min,
 117                 .extra2 = &ip_privileged_port_max,
 118         };
 119
 120         pports = net->ipv4.sysctl_ip_prot_sock;
 121
 122         ret = proc_dointvec_minmax(&tmp, write, buffer, lenp, ppos);
 123
 124         if (write && ret == 0) {
 125                 inet_get_local_port_range(net, &range[0], &range[1]);
 126                 /* Ensure that the local port range doesn't overlap with the
 127                  * privileged port range.
 128                  */
 129                 if (range[0] < pports)
 130                         ret = -EINVAL;
 131                 else
 132                         net->ipv4.sysctl_ip_prot_sock = pports;
 133         }
 134
 135         return ret;
 136 }
 137
 138 static void inet_get_ping_group_range_table(struct ctl_table *table, kgid_t *low, kgid_t *high)
 139 {
 140         kgid_t *data = table->data;
 141         struct net *net =
 142                 container_of(table->data, struct net, ipv4.ping_group_range.range);
 143         unsigned int seq;
 144         do {
 145                 seq = read_seqbegin(&net->ipv4.ping_group_range.lock);
 146
 147                 *low = data[0];
 148                 *high = data[1];
 149         } while (read_seqretry(&net->ipv4.ping_group_range.lock, seq));
 150 }
 151
 152 /* Update system visible IP port range */
 153 static void set_ping_group_range(struct ctl_table *table, kgid_t low, kgid_t high)
 154 {
 155         kgid_t *data = table->data;
 156         struct net *net =
 157                 container_of(table->data, struct net, ipv4.ping_group_range.range);
 158         write_seqlock(&net->ipv4.ping_group_range.lock);
 159         data[0] = low;
 160         data[1] = high;
 161         write_sequnlock(&net->ipv4.ping_group_range.lock);
 162 }
 163
 164 /* Validate changes from /proc interface. */
 165 static int ipv4_ping_group_range(struct ctl_table *table, int write,
 166                                  void __user *buffer,
 167                                  size_t *lenp, loff_t *ppos)
 168 {
 169         struct user_namespace *user_ns = current_user_ns();
 170         int ret;
 171         gid_t urange[2];
 172         kgid_t low, high;
 173         struct ctl_table tmp = {
 174                 .data = &urange,
 175                 .maxlen = sizeof(urange),
 176                 .mode = table->mode,
 177                 .extra1 = &ip_ping_group_range_min,
 178                 .extra2 = &ip_ping_group_range_max,
 179         };
 180
 181         inet_get_ping_group_range_table(table, &low, &high);
 182         urange[0] = from_kgid_munged(user_ns, low);
 183         urange[1] = from_kgid_munged(user_ns, high);
 184         ret = proc_dointvec_minmax(&tmp, write, buffer, lenp, ppos);
 185
 186         if (write && ret == 0) {
 187                 low = make_kgid(user_ns, urange[0]);
 188                 high = make_kgid(user_ns, urange[1]);
 189                 if (!gid_valid(low) || !gid_valid(high) ||
 190                     (urange[1] < urange[0]) || gid_lt(high, low)) {
 191                         low = make_kgid(&init_user_ns, 1);
 192                         high = make_kgid(&init_user_ns, 0);
 193                 }
 194                 set_ping_group_range(table, low, high);
 195         }
 196
 197         return ret;
 198 }
 199
 200 static int proc_tcp_congestion_control(struct ctl_table *ctl, int write,
 201                                        void __user *buffer, size_t *lenp, loff_t *ppos)
 202 {
 203         char val[TCP_CA_NAME_MAX];
 204         struct ctl_table tbl = {
 205                 .data = val,
 206                 .maxlen = TCP_CA_NAME_MAX,
 207         };
 208         int ret;
 209
 210         tcp_get_default_congestion_control(val);
 211
 212         ret = proc_dostring(&tbl, write, buffer, lenp, ppos);
 213         if (write && ret == 0)
 214                 ret = tcp_set_default_congestion_control(val);
 215         return ret;
 216 }
 217
 218 static int proc_tcp_available_congestion_control(struct ctl_table *ctl,
 219                                                  int write,
 220                                                  void __user *buffer, size_t *lenp,
 221                                                  loff_t *ppos)
 222 {
 223         struct ctl_table tbl = { .maxlen = TCP_CA_BUF_MAX, };
 224         int ret;
 225
 226         tbl.data = kmalloc(tbl.maxlen, GFP_USER);
 227         if (!tbl.data)
 228                 return -ENOMEM;
 229         tcp_get_available_congestion_control(tbl.data, TCP_CA_BUF_MAX);
 230         ret = proc_dostring(&tbl, write, buffer, lenp, ppos);
 231         kfree(tbl.data);
 232         return ret;
 233 }
 234
 235 static int proc_allowed_congestion_control(struct ctl_table *ctl,
 236                                            int write,
 237                                            void __user *buffer, size_t *lenp,
 238                                            loff_t *ppos)
 239 {
 240         struct ctl_table tbl = { .maxlen = TCP_CA_BUF_MAX };
 241         int ret;
 242
 243         tbl.data = kmalloc(tbl.maxlen, GFP_USER);
 244         if (!tbl.data)
 245                 return -ENOMEM;
 246
 247         tcp_get_allowed_congestion_control(tbl.data, tbl.maxlen);
 248         ret = proc_dostring(&tbl, write, buffer, lenp, ppos);
 249         if (write && ret == 0)
 250                 ret = tcp_set_allowed_congestion_control(tbl.data);
 251         kfree(tbl.data);
 252         return ret;
 253 }
 254
 255 static int proc_tcp_fastopen_key(struct ctl_table *ctl, int write,
 256                                  void __user *buffer, size_t *lenp,
 257                                  loff_t *ppos)
 258 {
 259         struct ctl_table tbl = { .maxlen = (TCP_FASTOPEN_KEY_LENGTH * 2 + 10) };
 260         struct tcp_fastopen_context *ctxt;
 261         int ret;
 262         u32  user_key[4]; /* 16 bytes, matching TCP_FASTOPEN_KEY_LENGTH */
 263
 264         tbl.data = kmalloc(tbl.maxlen, GFP_KERNEL);
 265         if (!tbl.data)
 266                 return -ENOMEM;
 267
 268         rcu_read_lock();
 269         ctxt = rcu_dereference(tcp_fastopen_ctx);
 270         if (ctxt)
 271                 memcpy(user_key, ctxt->key, TCP_FASTOPEN_KEY_LENGTH);
 272         else
 273                 memset(user_key, 0, sizeof(user_key));
 274         rcu_read_unlock();
 275
 276         snprintf(tbl.data, tbl.maxlen, "%08x-%08x-%08x-%08x",
 277                 user_key[0], user_key[1], user_key[2], user_key[3]);
 278         ret = proc_dostring(&tbl, write, buffer, lenp, ppos);
 279
 280         if (write && ret == 0) {
 281                 if (sscanf(tbl.data, "%x-%x-%x-%x", user_key, user_key + 1,
 282                            user_key + 2, user_key + 3) != 4) {
 283                         ret = -EINVAL;
 284                         goto bad_key;
 285                 }
 286                 /* Generate a dummy secret but don't publish it. This
 287                  * is needed so we don't regenerate a new key on the
 288                  * first invocation of tcp_fastopen_cookie_gen
 289                  */
 290                 tcp_fastopen_init_key_once(false);
 291                 tcp_fastopen_reset_cipher(user_key, TCP_FASTOPEN_KEY_LENGTH);
 292         }
 293
 294 bad_key:
 295         pr_debug("proc FO key set 0x%x-%x-%x-%x <- 0x%s: %u\n",
 296                user_key[0], user_key[1], user_key[2], user_key[3],
 297                (char *)tbl.data, ret);
 298         kfree(tbl.data);
 299         return ret;
 300 }
 301
 302 static void proc_configure_early_demux(int enabled, int protocol)
 303 {
 304         struct net_protocol *ipprot;
 305 #if IS_ENABLED(CONFIG_IPV6)
 306         struct inet6_protocol *ip6prot;
 307 #endif
 308
 309         rcu_read_lock();
 310
 311         ipprot = rcu_dereference(inet_protos[protocol]);
 312         if (ipprot)
 313                 ipprot->early_demux = enabled ? ipprot->early_demux_handler :
 314                                                 NULL;
 315
 316 #if IS_ENABLED(CONFIG_IPV6)
 317         ip6prot = rcu_dereference(inet6_protos[protocol]);
 318         if (ip6prot)
 319                 ip6prot->early_demux = enabled ? ip6prot->early_demux_handler :
 320                                                  NULL;
 321 #endif
 322         rcu_read_unlock();
 323 }
 324
 325 static int proc_tcp_early_demux(struct ctl_table *table, int write,
 326                                 void __user *buffer, size_t *lenp, loff_t *ppos)
 327 {
 328         int ret = 0;
 329
 330         ret = proc_dointvec(table, write, buffer, lenp, ppos);
 331
 332         if (write && !ret) {
 333                 int enabled = init_net.ipv4.sysctl_tcp_early_demux;
 334
 335                 proc_configure_early_demux(enabled, IPPROTO_TCP);
 336         }
 337
 338         return ret;
 339 }
 340
 341 static int proc_udp_early_demux(struct ctl_table *table, int write,
 342                                 void __user *buffer, size_t *lenp, loff_t *ppos)
 343 {
 344         int ret = 0;
 345
 346         ret = proc_dointvec(table, write, buffer, lenp, ppos);
 347
 348         if (write && !ret) {
 349                 int enabled = init_net.ipv4.sysctl_udp_early_demux;
 350
 351                 proc_configure_early_demux(enabled, IPPROTO_UDP);
 352         }
 353
 354         return ret;
 355 }
 356
 357 static int proc_tfo_blackhole_detect_timeout(struct ctl_table *table,
 358                                              int write,
 359                                              void __user *buffer,
 360                                              size_t *lenp, loff_t *ppos)
 361 {
 362         int ret;
 363
 364         ret = proc_dointvec_minmax(table, write, buffer, lenp, ppos);
 365         if (write && ret == 0)
 366                 tcp_fastopen_active_timeout_reset();
 367
 368         return ret;
 369 }
 370
 371 static int proc_tcp_available_ulp(struct ctl_table *ctl,
 372                                   int write,
 373                                   void __user *buffer, size_t *lenp,
 374                                   loff_t *ppos)
 375 {
 376         struct ctl_table tbl = { .maxlen = TCP_ULP_BUF_MAX, };
 377         int ret;
 378
 379         tbl.data = kmalloc(tbl.maxlen, GFP_USER);
 380         if (!tbl.data)
 381                 return -ENOMEM;
 382         tcp_get_available_ulp(tbl.data, TCP_ULP_BUF_MAX);
 383         ret = proc_dostring(&tbl, write, buffer, lenp, ppos);
 384         kfree(tbl.data);
 385
 386         return ret;
 387 }
 388
 389 static struct ctl_table ipv4_table[] = {
 390         {
 391                 .procname       = "tcp_retrans_collapse",
 392                 .data           = &sysctl_tcp_retrans_collapse,
 393                 .maxlen         = sizeof(int),
 394                 .mode           = 0644,
 395                 .proc_handler   = proc_dointvec
 396         },
 397         {
 398                 .procname       = "tcp_max_orphans",
 399                 .data           = &sysctl_tcp_max_orphans,
 400                 .maxlen         = sizeof(int),
 401                 .mode           = 0644,
 402                 .proc_handler   = proc_dointvec
 403         },
 404         {
 405                 .procname       = "tcp_fastopen",
 406                 .data           = &sysctl_tcp_fastopen,
 407                 .maxlen         = sizeof(int),
 408                 .mode           = 0644,
 409                 .proc_handler   = proc_dointvec,
 410         },
 411         {
 412                 .procname       = "tcp_fastopen_key",
 413                 .mode           = 0600,
 414                 .maxlen         = ((TCP_FASTOPEN_KEY_LENGTH * 2) + 10),
 415                 .proc_handler   = proc_tcp_fastopen_key,
 416         },
 417         {
 418                 .procname       = "tcp_fastopen_blackhole_timeout_sec",
 419                 .data           = &sysctl_tcp_fastopen_blackhole_timeout,
 420                 .maxlen         = sizeof(int),
 421                 .mode           = 0644,
 422                 .proc_handler   = proc_tfo_blackhole_detect_timeout,
 423                 .extra1         = &zero,
 424         },
 425         {
 426                 .procname       = "tcp_abort_on_overflow",
 427                 .data           = &sysctl_tcp_abort_on_overflow,
 428                 .maxlen         = sizeof(int),
 429                 .mode           = 0644,
 430                 .proc_handler   = proc_dointvec
 431         },
 432         {
 433                 .procname       = "tcp_stdurg",
 434                 .data           = &sysctl_tcp_stdurg,
 435                 .maxlen         = sizeof(int),
 436                 .mode           = 0644,
 437                 .proc_handler   = proc_dointvec
 438         },
 439         {
 440                 .procname       = "tcp_rfc1337",
 441                 .data           = &sysctl_tcp_rfc1337,
 442                 .maxlen         = sizeof(int),
 443                 .mode           = 0644,
 444                 .proc_handler   = proc_dointvec
 445         },
 446         {
 447                 .procname       = "inet_peer_threshold",
 448                 .data           = &inet_peer_threshold,
 449                 .maxlen         = sizeof(int),
 450                 .mode           = 0644,
 451                 .proc_handler   = proc_dointvec
 452         },
 453         {
 454                 .procname       = "inet_peer_minttl",
 455                 .data           = &inet_peer_minttl,
 456                 .maxlen         = sizeof(int),
 457                 .mode           = 0644,
 458                 .proc_handler   = proc_dointvec_jiffies,
 459         },
 460         {
 461                 .procname       = "inet_peer_maxttl",
 462                 .data           = &inet_peer_maxttl,
 463                 .maxlen         = sizeof(int),
 464                 .mode           = 0644,
 465                 .proc_handler   = proc_dointvec_jiffies,
 466         },
 467         {
 468                 .procname       = "tcp_fack",
 469                 .data           = &sysctl_tcp_fack,
 470                 .maxlen         = sizeof(int),
 471                 .mode           = 0644,
 472                 .proc_handler   = proc_dointvec
 473         },
 474         {
 475                 .procname       = "tcp_recovery",
 476                 .data           = &sysctl_tcp_recovery,
 477                 .maxlen         = sizeof(int),
 478                 .mode           = 0644,
 479                 .proc_handler   = proc_dointvec,
 480         },
 481         {
 482                 .procname       = "tcp_max_reordering",
 483                 .data           = &sysctl_tcp_max_reordering,
 484                 .maxlen         = sizeof(int),
 485                 .mode           = 0644,
 486                 .proc_handler   = proc_dointvec
 487         },
 488         {
 489                 .procname       = "tcp_dsack",
 490                 .data           = &sysctl_tcp_dsack,
 491                 .maxlen         = sizeof(int),
 492                 .mode           = 0644,
 493                 .proc_handler   = proc_dointvec
 494         },
 495         {
 496                 .procname       = "tcp_mem",
 497                 .maxlen         = sizeof(sysctl_tcp_mem),
 498                 .data           = &sysctl_tcp_mem,
 499                 .mode           = 0644,
 500                 .proc_handler   = proc_doulongvec_minmax,
 501         },
 502         {
 503                 .procname       = "tcp_wmem",
 504                 .data           = &sysctl_tcp_wmem,
 505                 .maxlen         = sizeof(sysctl_tcp_wmem),
 506                 .mode           = 0644,
 507                 .proc_handler   = proc_dointvec_minmax,
 508                 .extra1         = &one,
 509         },
 510         {
 511                 .procname       = "tcp_rmem",
 512                 .data           = &sysctl_tcp_rmem,
 513                 .maxlen         = sizeof(sysctl_tcp_rmem),
 514                 .mode           = 0644,
 515                 .proc_handler   = proc_dointvec_minmax,
 516                 .extra1         = &one,
 517         },
 518         {
 519                 .procname       = "tcp_app_win",
 520                 .data           = &sysctl_tcp_app_win,
 521                 .maxlen         = sizeof(int),
 522                 .mode           = 0644,
 523                 .proc_handler   = proc_dointvec
 524         },
 525         {
 526                 .procname       = "tcp_adv_win_scale",
 527                 .data           = &sysctl_tcp_adv_win_scale,
 528                 .maxlen         = sizeof(int),
 529                 .mode           = 0644,
 530                 .proc_handler   = proc_dointvec_minmax,
 531                 .extra1         = &tcp_adv_win_scale_min,
 532                 .extra2         = &tcp_adv_win_scale_max,
 533         },
 534         {
 535                 .procname       = "tcp_frto",
 536                 .data           = &sysctl_tcp_frto,
 537                 .maxlen         = sizeof(int),
 538                 .mode           = 0644,
 539                 .proc_handler   = proc_dointvec
 540         },
 541         {
 542                 .procname       = "tcp_min_rtt_wlen",
 543                 .data           = &sysctl_tcp_min_rtt_wlen,
 544                 .maxlen         = sizeof(int),
 545                 .mode           = 0644,
 546                 .proc_handler   = proc_dointvec
 547         },
 548         {
 549                 .procname       = "tcp_low_latency",
 550                 .data           = &sysctl_tcp_low_latency,
 551                 .maxlen         = sizeof(int),
 552                 .mode           = 0644,
 553                 .proc_handler   = proc_dointvec
 554         },
 555         {
 556                 .procname       = "tcp_no_metrics_save",
 557                 .data           = &sysctl_tcp_nometrics_save,
 558                 .maxlen         = sizeof(int),
 559                 .mode           = 0644,
 560                 .proc_handler   = proc_dointvec,
 561         },
 562         {
 563                 .procname       = "tcp_moderate_rcvbuf",
 564                 .data           = &sysctl_tcp_moderate_rcvbuf,
 565                 .maxlen         = sizeof(int),
 566                 .mode           = 0644,
 567                 .proc_handler   = proc_dointvec,
 568         },
 569         {
 570                 .procname       = "tcp_tso_win_divisor",
 571                 .data           = &sysctl_tcp_tso_win_divisor,
 572                 .maxlen         = sizeof(int),
 573                 .mode           = 0644,
 574                 .proc_handler   = proc_dointvec,
 575         },
 576         {
 577                 .procname       = "tcp_congestion_control",
 578                 .mode           = 0644,
 579                 .maxlen         = TCP_CA_NAME_MAX,
 580                 .proc_handler   = proc_tcp_congestion_control,
 581         },
 582         {
 583                 .procname       = "tcp_workaround_signed_windows",
 584                 .data           = &sysctl_tcp_workaround_signed_windows,
 585                 .maxlen         = sizeof(int),
 586                 .mode           = 0644,
 587                 .proc_handler   = proc_dointvec
 588         },
 589         {
 590                 .procname       = "tcp_limit_output_bytes",
 591                 .data           = &sysctl_tcp_limit_output_bytes,
 592                 .maxlen         = sizeof(int),
 593                 .mode           = 0644,
 594                 .proc_handler   = proc_dointvec
 595         },
 596         {
 597                 .procname       = "tcp_challenge_ack_limit",
 598                 .data           = &sysctl_tcp_challenge_ack_limit,
 599                 .maxlen         = sizeof(int),
 600                 .mode           = 0644,
 601                 .proc_handler   = proc_dointvec
 602         },
 603         {
 604                 .procname       = "tcp_slow_start_after_idle",
 605                 .data           = &sysctl_tcp_slow_start_after_idle,
 606                 .maxlen         = sizeof(int),
 607                 .mode           = 0644,
 608                 .proc_handler   = proc_dointvec
 609         },
 610 #ifdef CONFIG_NETLABEL
 611         {
 612                 .procname       = "cipso_cache_enable",
 613                 .data           = &cipso_v4_cache_enabled,
 614                 .maxlen         = sizeof(int),
 615                 .mode           = 0644,
 616                 .proc_handler   = proc_dointvec,
 617         },
 618         {
 619                 .procname       = "cipso_cache_bucket_size",
 620                 .data           = &cipso_v4_cache_bucketsize,
 621                 .maxlen         = sizeof(int),
 622                 .mode           = 0644,
 623                 .proc_handler   = proc_dointvec,
 624         },
 625         {
 626                 .procname       = "cipso_rbm_optfmt",
 627                 .data           = &cipso_v4_rbm_optfmt,
 628                 .maxlen         = sizeof(int),
 629                 .mode           = 0644,
 630                 .proc_handler   = proc_dointvec,
 631         },
 632         {
 633                 .procname       = "cipso_rbm_strictvalid",
 634                 .data           = &cipso_v4_rbm_strictvalid,
 635                 .maxlen         = sizeof(int),
 636                 .mode           = 0644,
 637                 .proc_handler   = proc_dointvec,
 638         },
 639 #endif /* CONFIG_NETLABEL */
 640         {
 641                 .procname       = "tcp_available_congestion_control",
 642                 .maxlen         = TCP_CA_BUF_MAX,
 643                 .mode           = 0444,
 644                 .proc_handler   = proc_tcp_available_congestion_control,
 645         },
 646         {
 647                 .procname       = "tcp_allowed_congestion_control",
 648                 .maxlen         = TCP_CA_BUF_MAX,
 649                 .mode           = 0644,
 650                 .proc_handler   = proc_allowed_congestion_control,
 651         },
 652         {
 653                 .procname       = "tcp_thin_linear_timeouts",
 654                 .data           = &sysctl_tcp_thin_linear_timeouts,
 655                 .maxlen         = sizeof(int),
 656                 .mode           = 0644,
 657                 .proc_handler   = proc_dointvec
 658         },
 659         {
 660                 .procname       = "tcp_early_retrans",
 661                 .data           = &sysctl_tcp_early_retrans,
 662                 .maxlen         = sizeof(int),
 663                 .mode           = 0644,
 664                 .proc_handler   = proc_dointvec_minmax,
 665                 .extra1         = &zero,
 666                 .extra2         = &four,
 667         },
 668         {
 669                 .procname       = "tcp_min_tso_segs",
 670                 .data           = &sysctl_tcp_min_tso_segs,
 671                 .maxlen         = sizeof(int),
 672                 .mode           = 0644,
 673                 .proc_handler   = proc_dointvec_minmax,
 674                 .extra1         = &one,
 675                 .extra2         = &gso_max_segs,
 676         },
 677         {
 678                 .procname       = "tcp_pacing_ss_ratio",
 679                 .data           = &sysctl_tcp_pacing_ss_ratio,
 680                 .maxlen         = sizeof(int),
 681                 .mode           = 0644,
 682                 .proc_handler   = proc_dointvec_minmax,
 683                 .extra1         = &zero,
 684                 .extra2         = &thousand,
 685         },
 686         {
 687                 .procname       = "tcp_pacing_ca_ratio",
 688                 .data           = &sysctl_tcp_pacing_ca_ratio,
 689                 .maxlen         = sizeof(int),
 690                 .mode           = 0644,
 691                 .proc_handler   = proc_dointvec_minmax,
 692                 .extra1         = &zero,
 693                 .extra2         = &thousand,
 694         },
 695         {
 696                 .procname       = "tcp_autocorking",
 697                 .data           = &sysctl_tcp_autocorking,
 698                 .maxlen         = sizeof(int),
 699                 .mode           = 0644,
 700                 .proc_handler   = proc_dointvec_minmax,
 701                 .extra1         = &zero,
 702                 .extra2         = &one,
 703         },
 704         {
 705                 .procname       = "tcp_invalid_ratelimit",
 706                 .data           = &sysctl_tcp_invalid_ratelimit,
 707                 .maxlen         = sizeof(int),
 708                 .mode           = 0644,
 709                 .proc_handler   = proc_dointvec_ms_jiffies,
 710         },
 711         {
 712                 .procname       = "tcp_available_ulp",
 713                 .maxlen         = TCP_ULP_BUF_MAX,
 714                 .mode           = 0444,
 715                 .proc_handler   = proc_tcp_available_ulp,
 716         },
 717         {
 718                 .procname       = "icmp_msgs_per_sec",
 719                 .data           = &sysctl_icmp_msgs_per_sec,
 720                 .maxlen         = sizeof(int),
 721                 .mode           = 0644,
 722                 .proc_handler   = proc_dointvec_minmax,
 723                 .extra1         = &zero,
 724         },
 725         {
 726                 .procname       = "icmp_msgs_burst",
 727                 .data           = &sysctl_icmp_msgs_burst,
 728                 .maxlen         = sizeof(int),
 729                 .mode           = 0644,
 730                 .proc_handler   = proc_dointvec_minmax,
 731                 .extra1         = &zero,
 732         },
 733         {
 734                 .procname       = "udp_mem",
 735                 .data           = &sysctl_udp_mem,
 736                 .maxlen         = sizeof(sysctl_udp_mem),
 737                 .mode           = 0644,
 738                 .proc_handler   = proc_doulongvec_minmax,
 739         },
 740         {
 741                 .procname       = "udp_rmem_min",
 742                 .data           = &sysctl_udp_rmem_min,
 743                 .maxlen         = sizeof(sysctl_udp_rmem_min),
 744                 .mode           = 0644,
 745                 .proc_handler   = proc_dointvec_minmax,
 746                 .extra1         = &one
 747         },
 748         {
 749                 .procname       = "udp_wmem_min",
 750                 .data           = &sysctl_udp_wmem_min,
 751                 .maxlen         = sizeof(sysctl_udp_wmem_min),
 752                 .mode           = 0644,
 753                 .proc_handler   = proc_dointvec_minmax,
 754                 .extra1         = &one
 755         },
 756         { }
 757 };
 758
 759 static struct ctl_table ipv4_net_table[] = {
 760         {
 761                 .procname       = "icmp_echo_ignore_all",
 762                 .data           = &init_net.ipv4.sysctl_icmp_echo_ignore_all,
 763                 .maxlen         = sizeof(int),
 764                 .mode           = 0644,
 765                 .proc_handler   = proc_dointvec
 766         },
 767         {
 768                 .procname       = "icmp_echo_ignore_broadcasts",
 769                 .data           = &init_net.ipv4.sysctl_icmp_echo_ignore_broadcasts,
 770                 .maxlen         = sizeof(int),
 771                 .mode           = 0644,
 772                 .proc_handler   = proc_dointvec
 773         },
 774         {
 775                 .procname       = "icmp_ignore_bogus_error_responses",
 776                 .data           = &init_net.ipv4.sysctl_icmp_ignore_bogus_error_responses,
 777                 .maxlen         = sizeof(int),
 778                 .mode           = 0644,
 779                 .proc_handler   = proc_dointvec
 780         },
 781         {
 782                 .procname       = "icmp_errors_use_inbound_ifaddr",
 783                 .data           = &init_net.ipv4.sysctl_icmp_errors_use_inbound_ifaddr,
 784                 .maxlen         = sizeof(int),
 785                 .mode           = 0644,
 786                 .proc_handler   = proc_dointvec
 787         },
 788         {
 789                 .procname       = "icmp_ratelimit",
 790                 .data           = &init_net.ipv4.sysctl_icmp_ratelimit,
 791                 .maxlen         = sizeof(int),
 792                 .mode           = 0644,
 793                 .proc_handler   = proc_dointvec_ms_jiffies,
 794         },
 795         {
 796                 .procname       = "icmp_ratemask",
 797                 .data           = &init_net.ipv4.sysctl_icmp_ratemask,
 798                 .maxlen         = sizeof(int),
 799                 .mode           = 0644,
 800                 .proc_handler   = proc_dointvec
 801         },
 802         {
 803                 .procname       = "ping_group_range",
 804                 .data           = &init_net.ipv4.ping_group_range.range,
 805                 .maxlen         = sizeof(gid_t)*2,
 806                 .mode           = 0644,
 807                 .proc_handler   = ipv4_ping_group_range,
 808         },
 809         {
 810                 .procname       = "tcp_ecn",
 811                 .data           = &init_net.ipv4.sysctl_tcp_ecn,
 812                 .maxlen         = sizeof(int),
 813                 .mode           = 0644,
 814                 .proc_handler   = proc_dointvec
 815         },
 816         {
 817                 .procname       = "tcp_ecn_fallback",
 818                 .data           = &init_net.ipv4.sysctl_tcp_ecn_fallback,
 819                 .maxlen         = sizeof(int),
 820                 .mode           = 0644,
 821                 .proc_handler   = proc_dointvec
 822         },
 823         {
 824                 .procname       = "ip_dynaddr",
 825                 .data           = &init_net.ipv4.sysctl_ip_dynaddr,
 826                 .maxlen         = sizeof(int),
 827                 .mode           = 0644,
 828                 .proc_handler   = proc_dointvec
 829         },
 830         {
 831                 .procname       = "ip_early_demux",
 832                 .data           = &init_net.ipv4.sysctl_ip_early_demux,
 833                 .maxlen         = sizeof(int),
 834                 .mode           = 0644,
 835                 .proc_handler   = proc_dointvec
 836         },
 837         {
 838                 .procname       = "udp_early_demux",
 839                 .data           = &init_net.ipv4.sysctl_udp_early_demux,
 840                 .maxlen         = sizeof(int),
 841                 .mode           = 0644,
 842                 .proc_handler   = proc_udp_early_demux
 843         },
 844         {
 845                 .procname       = "tcp_early_demux",
 846                 .data           = &init_net.ipv4.sysctl_tcp_early_demux,
 847                 .maxlen         = sizeof(int),
 848                 .mode           = 0644,
 849                 .proc_handler   = proc_tcp_early_demux
 850         },
 851         {
 852                 .procname       = "ip_default_ttl",
 853                 .data           = &init_net.ipv4.sysctl_ip_default_ttl,
 854                 .maxlen         = sizeof(int),
 855                 .mode           = 0644,
 856                 .proc_handler   = proc_dointvec_minmax,
 857                 .extra1         = &ip_ttl_min,
 858                 .extra2         = &ip_ttl_max,
 859         },
 860         {
 861                 .procname       = "ip_local_port_range",
 862                 .maxlen         = sizeof(init_net.ipv4.ip_local_ports.range),
 863                 .data           = &init_net.ipv4.ip_local_ports.range,
 864                 .mode           = 0644,
 865                 .proc_handler   = ipv4_local_port_range,
 866         },
 867         {
 868                 .procname       = "ip_local_reserved_ports",
 869                 .data           = &init_net.ipv4.sysctl_local_reserved_ports,
 870                 .maxlen         = 65536,
 871                 .mode           = 0644,
 872                 .proc_handler   = proc_do_large_bitmap,
 873         },
 874         {
 875                 .procname       = "ip_no_pmtu_disc",
 876                 .data           = &init_net.ipv4.sysctl_ip_no_pmtu_disc,
 877                 .maxlen         = sizeof(int),
 878                 .mode           = 0644,
 879                 .proc_handler   = proc_dointvec
 880         },
 881         {
 882                 .procname       = "ip_forward_use_pmtu",
 883                 .data           = &init_net.ipv4.sysctl_ip_fwd_use_pmtu,
 884                 .maxlen         = sizeof(int),
 885                 .mode           = 0644,
 886                 .proc_handler   = proc_dointvec,
 887         },
 888         {
 889                 .procname       = "ip_nonlocal_bind",
 890                 .data           = &init_net.ipv4.sysctl_ip_nonlocal_bind,
 891                 .maxlen         = sizeof(int),
 892                 .mode           = 0644,
 893                 .proc_handler   = proc_dointvec
 894         },
 895         {
 896                 .procname       = "fwmark_reflect",
 897                 .data           = &init_net.ipv4.sysctl_fwmark_reflect,
 898                 .maxlen         = sizeof(int),
 899                 .mode           = 0644,
 900                 .proc_handler   = proc_dointvec,
 901         },
 902         {
 903                 .procname       = "tcp_fwmark_accept",
 904                 .data           = &init_net.ipv4.sysctl_tcp_fwmark_accept,
 905                 .maxlen         = sizeof(int),
 906                 .mode           = 0644,
 907                 .proc_handler   = proc_dointvec,
 908         },
 909 #ifdef CONFIG_NET_L3_MASTER_DEV
 910         {
 911                 .procname       = "tcp_l3mdev_accept",
 912                 .data           = &init_net.ipv4.sysctl_tcp_l3mdev_accept,
 913                 .maxlen         = sizeof(int),
 914                 .mode           = 0644,
 915                 .proc_handler   = proc_dointvec_minmax,
 916                 .extra1         = &zero,
 917                 .extra2         = &one,
 918         },
 919 #endif
 920         {
 921                 .procname       = "tcp_mtu_probing",
 922                 .data           = &init_net.ipv4.sysctl_tcp_mtu_probing,
 923                 .maxlen         = sizeof(int),
 924                 .mode           = 0644,
 925                 .proc_handler   = proc_dointvec,
 926         },
 927         {
 928                 .procname       = "tcp_base_mss",
 929                 .data           = &init_net.ipv4.sysctl_tcp_base_mss,
 930                 .maxlen         = sizeof(int),
 931                 .mode           = 0644,
 932                 .proc_handler   = proc_dointvec,
 933         },
 934         {
 935                 .procname       = "tcp_probe_threshold",
 936                 .data           = &init_net.ipv4.sysctl_tcp_probe_threshold,
 937                 .maxlen         = sizeof(int),
 938                 .mode           = 0644,
 939                 .proc_handler   = proc_dointvec,
 940         },
 941         {
 942                 .procname       = "tcp_probe_interval",
 943                 .data           = &init_net.ipv4.sysctl_tcp_probe_interval,
 944                 .maxlen         = sizeof(int),
 945                 .mode           = 0644,
 946                 .proc_handler   = proc_dointvec,
 947         },
 948         {
 949                 .procname       = "igmp_link_local_mcast_reports",
 950                 .data           = &init_net.ipv4.sysctl_igmp_llm_reports,
 951                 .maxlen         = sizeof(int),
 952                 .mode           = 0644,
 953                 .proc_handler   = proc_dointvec
 954         },
 955         {
 956                 .procname       = "igmp_max_memberships",
 957                 .data           = &init_net.ipv4.sysctl_igmp_max_memberships,
 958                 .maxlen         = sizeof(int),
 959                 .mode           = 0644,
 960                 .proc_handler   = proc_dointvec
 961         },
 962         {
 963                 .procname       = "igmp_max_msf",
 964                 .data           = &init_net.ipv4.sysctl_igmp_max_msf,
 965                 .maxlen         = sizeof(int),
 966                 .mode           = 0644,
 967                 .proc_handler   = proc_dointvec
 968         },
 969 #ifdef CONFIG_IP_MULTICAST
 970         {
 971                 .procname       = "igmp_qrv",
 972                 .data           = &init_net.ipv4.sysctl_igmp_qrv,
 973                 .maxlen         = sizeof(int),
 974                 .mode           = 0644,
 975                 .proc_handler   = proc_dointvec_minmax,
 976                 .extra1         = &one
 977         },
 978 #endif
 979         {
 980                 .procname       = "tcp_keepalive_time",
 981                 .data           = &init_net.ipv4.sysctl_tcp_keepalive_time,
 982                 .maxlen         = sizeof(int),
 983                 .mode           = 0644,
 984                 .proc_handler   = proc_dointvec_jiffies,
 985         },
 986         {
 987                 .procname       = "tcp_keepalive_probes",
 988                 .data           = &init_net.ipv4.sysctl_tcp_keepalive_probes,
 989                 .maxlen         = sizeof(int),
 990                 .mode           = 0644,
 991                 .proc_handler   = proc_dointvec
 992         },
 993         {
 994                 .procname       = "tcp_keepalive_intvl",
 995                 .data           = &init_net.ipv4.sysctl_tcp_keepalive_intvl,
 996                 .maxlen         = sizeof(int),
 997                 .mode           = 0644,
 998                 .proc_handler   = proc_dointvec_jiffies,
 999         },
1000         {
1001                 .procname       = "tcp_syn_retries",
1002                 .data           = &init_net.ipv4.sysctl_tcp_syn_retries,
1003                 .maxlen         = sizeof(int),
1004                 .mode           = 0644,
1005                 .proc_handler   = proc_dointvec_minmax,
1006                 .extra1         = &tcp_syn_retries_min,
1007                 .extra2         = &tcp_syn_retries_max
1008         },
1009         {
1010                 .procname       = "tcp_synack_retries",
1011                 .data           = &init_net.ipv4.sysctl_tcp_synack_retries,
1012                 .maxlen         = sizeof(int),
1013                 .mode           = 0644,
1014                 .proc_handler   = proc_dointvec
1015         },
1016 #ifdef CONFIG_SYN_COOKIES
1017         {
1018                 .procname       = "tcp_syncookies",
1019                 .data           = &init_net.ipv4.sysctl_tcp_syncookies,
1020                 .maxlen         = sizeof(int),
1021                 .mode           = 0644,
1022                 .proc_handler   = proc_dointvec
1023         },
1024 #endif
1025         {
1026                 .procname       = "tcp_reordering",
1027                 .data           = &init_net.ipv4.sysctl_tcp_reordering,
1028                 .maxlen         = sizeof(int),
1029                 .mode           = 0644,
1030                 .proc_handler   = proc_dointvec
1031         },
1032         {
1033                 .procname       = "tcp_retries1",
1034                 .data           = &init_net.ipv4.sysctl_tcp_retries1,
1035                 .maxlen         = sizeof(int),
1036                 .mode           = 0644,
1037                 .proc_handler   = proc_dointvec_minmax,
1038                 .extra2         = &tcp_retr1_max
1039         },
1040         {
1041                 .procname       = "tcp_retries2",
1042                 .data           = &init_net.ipv4.sysctl_tcp_retries2,
1043                 .maxlen         = sizeof(int),
1044                 .mode           = 0644,
1045                 .proc_handler   = proc_dointvec
1046         },
1047         {
1048                 .procname       = "tcp_orphan_retries",
1049                 .data           = &init_net.ipv4.sysctl_tcp_orphan_retries,
1050                 .maxlen         = sizeof(int),
1051                 .mode           = 0644,
1052                 .proc_handler   = proc_dointvec
1053         },
1054         {
1055                 .procname       = "tcp_fin_timeout",
1056                 .data           = &init_net.ipv4.sysctl_tcp_fin_timeout,
1057                 .maxlen         = sizeof(int),
1058                 .mode           = 0644,
1059                 .proc_handler   = proc_dointvec_jiffies,
1060         },
1061         {
1062                 .procname       = "tcp_notsent_lowat",
1063                 .data           = &init_net.ipv4.sysctl_tcp_notsent_lowat,
1064                 .maxlen         = sizeof(unsigned int),
1065                 .mode           = 0644,
1066                 .proc_handler   = proc_douintvec,
1067         },
1068         {
1069                 .procname       = "tcp_tw_reuse",
1070                 .data           = &init_net.ipv4.sysctl_tcp_tw_reuse,
1071                 .maxlen         = sizeof(int),
1072                 .mode           = 0644,
1073                 .proc_handler   = proc_dointvec
1074         },
1075         {
1076                 .procname       = "tcp_max_tw_buckets",
1077                 .data           = &init_net.ipv4.tcp_death_row.sysctl_max_tw_buckets,
1078                 .maxlen         = sizeof(int),
1079                 .mode           = 0644,
1080                 .proc_handler   = proc_dointvec
1081         },
1082         {
1083                 .procname       = "tcp_max_syn_backlog",
1084                 .data           = &init_net.ipv4.sysctl_max_syn_backlog,
1085                 .maxlen         = sizeof(int),
1086                 .mode           = 0644,
1087                 .proc_handler   = proc_dointvec
1088         },
1089 #ifdef CONFIG_IP_ROUTE_MULTIPATH
1090         {
1091                 .procname       = "fib_multipath_use_neigh",
1092                 .data           = &init_net.ipv4.sysctl_fib_multipath_use_neigh,
1093                 .maxlen         = sizeof(int),
1094                 .mode           = 0644,
1095                 .proc_handler   = proc_dointvec_minmax,
1096                 .extra1         = &zero,
1097                 .extra2         = &one,
1098         },
1099         {
1100                 .procname       = "fib_multipath_hash_policy",
1101                 .data           = &init_net.ipv4.sysctl_fib_multipath_hash_policy,
1102                 .maxlen         = sizeof(int),
1103                 .mode           = 0644,
1104                 .proc_handler   = proc_dointvec_minmax,
1105                 .extra1         = &zero,
1106                 .extra2         = &one,
1107         },
1108 #endif
1109         {
1110                 .procname       = "ip_unprivileged_port_start",
1111                 .maxlen         = sizeof(int),
1112                 .data           = &init_net.ipv4.sysctl_ip_prot_sock,
1113                 .mode           = 0644,
1114                 .proc_handler   = ipv4_privileged_ports,
1115         },
1116 #ifdef CONFIG_NET_L3_MASTER_DEV
1117         {
1118                 .procname       = "udp_l3mdev_accept",
1119                 .data           = &init_net.ipv4.sysctl_udp_l3mdev_accept,
1120                 .maxlen         = sizeof(int),
1121                 .mode           = 0644,
1122                 .proc_handler   = proc_dointvec_minmax,
1123                 .extra1         = &zero,
1124                 .extra2         = &one,
1125         },
1126 #endif
1127         {
1128                 .procname       = "tcp_sack",
1129                 .data           = &init_net.ipv4.sysctl_tcp_sack,
1130                 .maxlen         = sizeof(int),
1131                 .mode           = 0644,
1132                 .proc_handler   = proc_dointvec
1133         },
1134         {
1135                 .procname       = "tcp_window_scaling",
1136                 .data           = &init_net.ipv4.sysctl_tcp_window_scaling,
1137                 .maxlen         = sizeof(int),
1138                 .mode           = 0644,
1139                 .proc_handler   = proc_dointvec
1140         },
1141         {
1142                 .procname       = "tcp_timestamps",
1143                 .data           = &init_net.ipv4.sysctl_tcp_timestamps,
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);