net/ipv4/sysctl_net_ipv4.c

   1 /*
   2  * sysctl_net_ipv4.c: sysctl interface to net IPV4 subsystem.
   3  *
   4  * Begun April 1, 1996, Mike Shaver.
   5  * Added /proc/sys/net/ipv4 directory entry (empty =) ). [MS]
   6  */
   7
   8 #include <linux/mm.h>
   9 #include <linux/module.h>
  10 #include <linux/sysctl.h>
  11 #include <linux/igmp.h>
  12 #include <linux/inetdevice.h>
  13 #include <linux/seqlock.h>
  14 #include <linux/init.h>
  15 #include <linux/slab.h>
  16 #include <linux/nsproxy.h>
  17 #include <linux/swap.h>
  18 #include <net/snmp.h>
  19 #include <net/icmp.h>
  20 #include <net/ip.h>
  21 #include <net/route.h>
  22 #include <net/tcp.h>
  23 #include <net/udp.h>
  24 #include <net/cipso_ipv4.h>
  25 #include <net/inet_frag.h>
  26 #include <net/ping.h>
  27 #include <net/tcp_memcontrol.h>
  28
  29 static int zero;
  30 static int one = 1;
  31 static int four = 4;
  32 static int gso_max_segs = GSO_MAX_SEGS;
  33 static int tcp_retr1_max = 255;
  34 static int ip_local_port_range_min[] = { 1, 1 };
  35 static int ip_local_port_range_max[] = { 65535, 65535 };
  36 static int tcp_adv_win_scale_min = -31;
  37 static int tcp_adv_win_scale_max = 31;
  38 static int ip_ttl_min = 1;
  39 static int ip_ttl_max = 255;
  40 static int tcp_syn_retries_min = 1;
  41 static int tcp_syn_retries_max = MAX_TCP_SYNCNT;
  42 static int ip_ping_group_range_min[] = { 0, 0 };
  43 static int ip_ping_group_range_max[] = { GID_T_MAX, GID_T_MAX };
  44 static int min_sndbuf = SOCK_MIN_SNDBUF;
  45 static int min_rcvbuf = SOCK_MIN_RCVBUF;
  46
  47 /* Update system visible IP port range */
  48 static void set_local_port_range(struct net *net, int range[2])
  49 {
  50         bool same_parity = !((range[0] ^ range[1]) & 1);
  51
  52         write_seqlock(&net->ipv4.ip_local_ports.lock);
  53         if (same_parity && !net->ipv4.ip_local_ports.warned) {
  54                 net->ipv4.ip_local_ports.warned = true;
  55                 pr_err_ratelimited("ip_local_port_range: prefer different parity for start/end values.\n");
  56         }
  57         net->ipv4.ip_local_ports.range[0] = range[0];
  58         net->ipv4.ip_local_ports.range[1] = range[1];
  59         write_sequnlock(&net->ipv4.ip_local_ports.lock);
  60 }
  61
  62 /* Validate changes from /proc interface. */
  63 static int ipv4_local_port_range(struct ctl_table *table, int write,
  64                                  void __user *buffer,
  65                                  size_t *lenp, loff_t *ppos)
  66 {
  67         struct net *net =
  68                 container_of(table->data, struct net, ipv4.ip_local_ports.range);
  69         int ret;
  70         int range[2];
  71         struct ctl_table tmp = {
  72                 .data = &range,
  73                 .maxlen = sizeof(range),
  74                 .mode = table->mode,
  75                 .extra1 = &ip_local_port_range_min,
  76                 .extra2 = &ip_local_port_range_max,
  77         };
  78
  79         inet_get_local_port_range(net, &range[0], &range[1]);
  80
  81         ret = proc_dointvec_minmax(&tmp, write, buffer, lenp, ppos);
  82
  83         if (write && ret == 0) {
  84                 if (range[1] < range[0])
  85                         ret = -EINVAL;
  86                 else
  87                         set_local_port_range(net, range);
  88         }
  89
  90         return ret;
  91 }
  92
  93
  94 static void inet_get_ping_group_range_table(struct ctl_table *table, kgid_t *low, kgid_t *high)
  95 {
  96         kgid_t *data = table->data;
  97         struct net *net =
  98                 container_of(table->data, struct net, ipv4.ping_group_range.range);
  99         unsigned int seq;
 100         do {
 101                 seq = read_seqbegin(&net->ipv4.ip_local_ports.lock);
 102
 103                 *low = data[0];
 104                 *high = data[1];
 105         } while (read_seqretry(&net->ipv4.ip_local_ports.lock, seq));
 106 }
 107
 108 /* Update system visible IP port range */
 109 static void set_ping_group_range(struct ctl_table *table, kgid_t low, kgid_t high)
 110 {
 111         kgid_t *data = table->data;
 112         struct net *net =
 113                 container_of(table->data, struct net, ipv4.ping_group_range.range);
 114         write_seqlock(&net->ipv4.ip_local_ports.lock);
 115         data[0] = low;
 116         data[1] = high;
 117         write_sequnlock(&net->ipv4.ip_local_ports.lock);
 118 }
 119
 120 /* Validate changes from /proc interface. */
 121 static int ipv4_ping_group_range(struct ctl_table *table, int write,
 122                                  void __user *buffer,
 123                                  size_t *lenp, loff_t *ppos)
 124 {
 125         struct user_namespace *user_ns = current_user_ns();
 126         int ret;
 127         gid_t urange[2];
 128         kgid_t low, high;
 129         struct ctl_table tmp = {
 130                 .data = &urange,
 131                 .maxlen = sizeof(urange),
 132                 .mode = table->mode,
 133                 .extra1 = &ip_ping_group_range_min,
 134                 .extra2 = &ip_ping_group_range_max,
 135         };
 136
 137         inet_get_ping_group_range_table(table, &low, &high);
 138         urange[0] = from_kgid_munged(user_ns, low);
 139         urange[1] = from_kgid_munged(user_ns, high);
 140         ret = proc_dointvec_minmax(&tmp, write, buffer, lenp, ppos);
 141
 142         if (write && ret == 0) {
 143                 low = make_kgid(user_ns, urange[0]);
 144                 high = make_kgid(user_ns, urange[1]);
 145                 if (!gid_valid(low) || !gid_valid(high) ||
 146                     (urange[1] < urange[0]) || gid_lt(high, low)) {
 147                         low = make_kgid(&init_user_ns, 1);
 148                         high = make_kgid(&init_user_ns, 0);
 149                 }
 150                 set_ping_group_range(table, low, high);
 151         }
 152
 153         return ret;
 154 }
 155
 156 static int proc_tcp_congestion_control(struct ctl_table *ctl, int write,
 157                                        void __user *buffer, size_t *lenp, loff_t *ppos)
 158 {
 159         char val[TCP_CA_NAME_MAX];
 160         struct ctl_table tbl = {
 161                 .data = val,
 162                 .maxlen = TCP_CA_NAME_MAX,
 163         };
 164         int ret;
 165
 166         tcp_get_default_congestion_control(val);
 167
 168         ret = proc_dostring(&tbl, write, buffer, lenp, ppos);
 169         if (write && ret == 0)
 170                 ret = tcp_set_default_congestion_control(val);
 171         return ret;
 172 }
 173
 174 static int proc_tcp_available_congestion_control(struct ctl_table *ctl,
 175                                                  int write,
 176                                                  void __user *buffer, size_t *lenp,
 177                                                  loff_t *ppos)
 178 {
 179         struct ctl_table tbl = { .maxlen = TCP_CA_BUF_MAX, };
 180         int ret;
 181
 182         tbl.data = kmalloc(tbl.maxlen, GFP_USER);
 183         if (!tbl.data)
 184                 return -ENOMEM;
 185         tcp_get_available_congestion_control(tbl.data, TCP_CA_BUF_MAX);
 186         ret = proc_dostring(&tbl, write, buffer, lenp, ppos);
 187         kfree(tbl.data);
 188         return ret;
 189 }
 190
 191 static int proc_allowed_congestion_control(struct ctl_table *ctl,
 192                                            int write,
 193                                            void __user *buffer, size_t *lenp,
 194                                            loff_t *ppos)
 195 {
 196         struct ctl_table tbl = { .maxlen = TCP_CA_BUF_MAX };
 197         int ret;
 198
 199         tbl.data = kmalloc(tbl.maxlen, GFP_USER);
 200         if (!tbl.data)
 201                 return -ENOMEM;
 202
 203         tcp_get_allowed_congestion_control(tbl.data, tbl.maxlen);
 204         ret = proc_dostring(&tbl, write, buffer, lenp, ppos);
 205         if (write && ret == 0)
 206                 ret = tcp_set_allowed_congestion_control(tbl.data);
 207         kfree(tbl.data);
 208         return ret;
 209 }
 210
 211 static int proc_tcp_fastopen_key(struct ctl_table *ctl, int write,
 212                                  void __user *buffer, size_t *lenp,
 213                                  loff_t *ppos)
 214 {
 215         struct ctl_table tbl = { .maxlen = (TCP_FASTOPEN_KEY_LENGTH * 2 + 10) };
 216         struct tcp_fastopen_context *ctxt;
 217         int ret;
 218         u32  user_key[4]; /* 16 bytes, matching TCP_FASTOPEN_KEY_LENGTH */
 219
 220         tbl.data = kmalloc(tbl.maxlen, GFP_KERNEL);
 221         if (!tbl.data)
 222                 return -ENOMEM;
 223
 224         rcu_read_lock();
 225         ctxt = rcu_dereference(tcp_fastopen_ctx);
 226         if (ctxt)
 227                 memcpy(user_key, ctxt->key, TCP_FASTOPEN_KEY_LENGTH);
 228         else
 229                 memset(user_key, 0, sizeof(user_key));
 230         rcu_read_unlock();
 231
 232         snprintf(tbl.data, tbl.maxlen, "%08x-%08x-%08x-%08x",
 233                 user_key[0], user_key[1], user_key[2], user_key[3]);
 234         ret = proc_dostring(&tbl, write, buffer, lenp, ppos);
 235
 236         if (write && ret == 0) {
 237                 if (sscanf(tbl.data, "%x-%x-%x-%x", user_key, user_key + 1,
 238                            user_key + 2, user_key + 3) != 4) {
 239                         ret = -EINVAL;
 240                         goto bad_key;
 241                 }
 242                 /* Generate a dummy secret but don't publish it. This
 243                  * is needed so we don't regenerate a new key on the
 244                  * first invocation of tcp_fastopen_cookie_gen
 245                  */
 246                 tcp_fastopen_init_key_once(false);
 247                 tcp_fastopen_reset_cipher(user_key, TCP_FASTOPEN_KEY_LENGTH);
 248         }
 249
 250 bad_key:
 251         pr_debug("proc FO key set 0x%x-%x-%x-%x <- 0x%s: %u\n",
 252                user_key[0], user_key[1], user_key[2], user_key[3],
 253                (char *)tbl.data, ret);
 254         kfree(tbl.data);
 255         return ret;
 256 }
 257
 258 static struct ctl_table ipv4_table[] = {
 259         {
 260                 .procname       = "tcp_timestamps",
 261                 .data           = &sysctl_tcp_timestamps,
 262                 .maxlen         = sizeof(int),
 263                 .mode           = 0644,
 264                 .proc_handler   = proc_dointvec
 265         },
 266         {
 267                 .procname       = "tcp_window_scaling",
 268                 .data           = &sysctl_tcp_window_scaling,
 269                 .maxlen         = sizeof(int),
 270                 .mode           = 0644,
 271                 .proc_handler   = proc_dointvec
 272         },
 273         {
 274                 .procname       = "tcp_sack",
 275                 .data           = &sysctl_tcp_sack,
 276                 .maxlen         = sizeof(int),
 277                 .mode           = 0644,
 278                 .proc_handler   = proc_dointvec
 279         },
 280         {
 281                 .procname       = "tcp_retrans_collapse",
 282                 .data           = &sysctl_tcp_retrans_collapse,
 283                 .maxlen         = sizeof(int),
 284                 .mode           = 0644,
 285                 .proc_handler   = proc_dointvec
 286         },
 287         {
 288                 .procname       = "ip_default_ttl",
 289                 .data           = &sysctl_ip_default_ttl,
 290                 .maxlen         = sizeof(int),
 291                 .mode           = 0644,
 292                 .proc_handler   = proc_dointvec_minmax,
 293                 .extra1         = &ip_ttl_min,
 294                 .extra2         = &ip_ttl_max,
 295         },
 296         {
 297                 .procname       = "tcp_syn_retries",
 298                 .data           = &sysctl_tcp_syn_retries,
 299                 .maxlen         = sizeof(int),
 300                 .mode           = 0644,
 301                 .proc_handler   = proc_dointvec_minmax,
 302                 .extra1         = &tcp_syn_retries_min,
 303                 .extra2         = &tcp_syn_retries_max
 304         },
 305         {
 306                 .procname       = "tcp_synack_retries",
 307                 .data           = &sysctl_tcp_synack_retries,
 308                 .maxlen         = sizeof(int),
 309                 .mode           = 0644,
 310                 .proc_handler   = proc_dointvec
 311         },
 312         {
 313                 .procname       = "tcp_max_orphans",
 314                 .data           = &sysctl_tcp_max_orphans,
 315                 .maxlen         = sizeof(int),
 316                 .mode           = 0644,
 317                 .proc_handler   = proc_dointvec
 318         },
 319         {
 320                 .procname       = "tcp_max_tw_buckets",
 321                 .data           = &tcp_death_row.sysctl_max_tw_buckets,
 322                 .maxlen         = sizeof(int),
 323                 .mode           = 0644,
 324                 .proc_handler   = proc_dointvec
 325         },
 326         {
 327                 .procname       = "ip_early_demux",
 328                 .data           = &sysctl_ip_early_demux,
 329                 .maxlen         = sizeof(int),
 330                 .mode           = 0644,
 331                 .proc_handler   = proc_dointvec
 332         },
 333         {
 334                 .procname       = "ip_dynaddr",
 335                 .data           = &sysctl_ip_dynaddr,
 336                 .maxlen         = sizeof(int),
 337                 .mode           = 0644,
 338                 .proc_handler   = proc_dointvec
 339         },
 340         {
 341                 .procname       = "tcp_keepalive_time",
 342                 .data           = &sysctl_tcp_keepalive_time,
 343                 .maxlen         = sizeof(int),
 344                 .mode           = 0644,
 345                 .proc_handler   = proc_dointvec_jiffies,
 346         },
 347         {
 348                 .procname       = "tcp_keepalive_probes",
 349                 .data           = &sysctl_tcp_keepalive_probes,
 350                 .maxlen         = sizeof(int),
 351                 .mode           = 0644,
 352                 .proc_handler   = proc_dointvec
 353         },
 354         {
 355                 .procname       = "tcp_keepalive_intvl",
 356                 .data           = &sysctl_tcp_keepalive_intvl,
 357                 .maxlen         = sizeof(int),
 358                 .mode           = 0644,
 359                 .proc_handler   = proc_dointvec_jiffies,
 360         },
 361         {
 362                 .procname       = "tcp_retries1",
 363                 .data           = &sysctl_tcp_retries1,
 364                 .maxlen         = sizeof(int),
 365                 .mode           = 0644,
 366                 .proc_handler   = proc_dointvec_minmax,
 367                 .extra2         = &tcp_retr1_max
 368         },
 369         {
 370                 .procname       = "tcp_retries2",
 371                 .data           = &sysctl_tcp_retries2,
 372                 .maxlen         = sizeof(int),
 373                 .mode           = 0644,
 374                 .proc_handler   = proc_dointvec
 375         },
 376         {
 377                 .procname       = "tcp_fin_timeout",
 378                 .data           = &sysctl_tcp_fin_timeout,
 379                 .maxlen         = sizeof(int),
 380                 .mode           = 0644,
 381                 .proc_handler   = proc_dointvec_jiffies,
 382         },
 383 #ifdef CONFIG_SYN_COOKIES
 384         {
 385                 .procname       = "tcp_syncookies",
 386                 .data           = &sysctl_tcp_syncookies,
 387                 .maxlen         = sizeof(int),
 388                 .mode           = 0644,
 389                 .proc_handler   = proc_dointvec
 390         },
 391 #endif
 392         {
 393                 .procname       = "tcp_fastopen",
 394                 .data           = &sysctl_tcp_fastopen,
 395                 .maxlen         = sizeof(int),
 396                 .mode           = 0644,
 397                 .proc_handler   = proc_dointvec,
 398         },
 399         {
 400                 .procname       = "tcp_fastopen_key",
 401                 .mode           = 0600,
 402                 .maxlen         = ((TCP_FASTOPEN_KEY_LENGTH * 2) + 10),
 403                 .proc_handler   = proc_tcp_fastopen_key,
 404         },
 405         {
 406                 .procname       = "tcp_tw_recycle",
 407                 .data           = &tcp_death_row.sysctl_tw_recycle,
 408                 .maxlen         = sizeof(int),
 409                 .mode           = 0644,
 410                 .proc_handler   = proc_dointvec
 411         },
 412         {
 413                 .procname       = "tcp_abort_on_overflow",
 414                 .data           = &sysctl_tcp_abort_on_overflow,
 415                 .maxlen         = sizeof(int),
 416                 .mode           = 0644,
 417                 .proc_handler   = proc_dointvec
 418         },
 419         {
 420                 .procname       = "tcp_stdurg",
 421                 .data           = &sysctl_tcp_stdurg,
 422                 .maxlen         = sizeof(int),
 423                 .mode           = 0644,
 424                 .proc_handler   = proc_dointvec
 425         },
 426         {
 427                 .procname       = "tcp_rfc1337",
 428                 .data           = &sysctl_tcp_rfc1337,
 429                 .maxlen         = sizeof(int),
 430                 .mode           = 0644,
 431                 .proc_handler   = proc_dointvec
 432         },
 433         {
 434                 .procname       = "tcp_max_syn_backlog",
 435                 .data           = &sysctl_max_syn_backlog,
 436                 .maxlen         = sizeof(int),
 437                 .mode           = 0644,
 438                 .proc_handler   = proc_dointvec
 439         },
 440         {
 441                 .procname       = "igmp_max_memberships",
 442                 .data           = &sysctl_igmp_max_memberships,
 443                 .maxlen         = sizeof(int),
 444                 .mode           = 0644,
 445                 .proc_handler   = proc_dointvec
 446         },
 447         {
 448                 .procname       = "igmp_max_msf",
 449                 .data           = &sysctl_igmp_max_msf,
 450                 .maxlen         = sizeof(int),
 451                 .mode           = 0644,
 452                 .proc_handler   = proc_dointvec
 453         },
 454 #ifdef CONFIG_IP_MULTICAST
 455         {
 456                 .procname       = "igmp_qrv",
 457                 .data           = &sysctl_igmp_qrv,
 458                 .maxlen         = sizeof(int),
 459                 .mode           = 0644,
 460                 .proc_handler   = proc_dointvec_minmax,
 461                 .extra1         = &one
 462         },
 463 #endif
 464         {
 465                 .procname       = "inet_peer_threshold",
 466                 .data           = &inet_peer_threshold,
 467                 .maxlen         = sizeof(int),
 468                 .mode           = 0644,
 469                 .proc_handler   = proc_dointvec
 470         },
 471         {
 472                 .procname       = "inet_peer_minttl",
 473                 .data           = &inet_peer_minttl,
 474                 .maxlen         = sizeof(int),
 475                 .mode           = 0644,
 476                 .proc_handler   = proc_dointvec_jiffies,
 477         },
 478         {
 479                 .procname       = "inet_peer_maxttl",
 480                 .data           = &inet_peer_maxttl,
 481                 .maxlen         = sizeof(int),
 482                 .mode           = 0644,
 483                 .proc_handler   = proc_dointvec_jiffies,
 484         },
 485         {
 486                 .procname       = "tcp_orphan_retries",
 487                 .data           = &sysctl_tcp_orphan_retries,
 488                 .maxlen         = sizeof(int),
 489                 .mode           = 0644,
 490                 .proc_handler   = proc_dointvec
 491         },
 492         {
 493                 .procname       = "tcp_fack",
 494                 .data           = &sysctl_tcp_fack,
 495                 .maxlen         = sizeof(int),
 496                 .mode           = 0644,
 497                 .proc_handler   = proc_dointvec
 498         },
 499         {
 500                 .procname       = "tcp_reordering",
 501                 .data           = &sysctl_tcp_reordering,
 502                 .maxlen         = sizeof(int),
 503                 .mode           = 0644,
 504                 .proc_handler   = proc_dointvec
 505         },
 506         {
 507                 .procname       = "tcp_max_reordering",
 508                 .data           = &sysctl_tcp_max_reordering,
 509                 .maxlen         = sizeof(int),
 510                 .mode           = 0644,
 511                 .proc_handler   = proc_dointvec
 512         },
 513         {
 514                 .procname       = "tcp_dsack",
 515                 .data           = &sysctl_tcp_dsack,
 516                 .maxlen         = sizeof(int),
 517                 .mode           = 0644,
 518                 .proc_handler   = proc_dointvec
 519         },
 520         {
 521                 .procname       = "tcp_mem",
 522                 .maxlen         = sizeof(sysctl_tcp_mem),
 523                 .data           = &sysctl_tcp_mem,
 524                 .mode           = 0644,
 525                 .proc_handler   = proc_doulongvec_minmax,
 526         },
 527         {
 528                 .procname       = "tcp_wmem",
 529                 .data           = &sysctl_tcp_wmem,
 530                 .maxlen         = sizeof(sysctl_tcp_wmem),
 531                 .mode           = 0644,
 532                 .proc_handler   = proc_dointvec_minmax,
 533                 .extra1         = &min_sndbuf,
 534         },
 535         {
 536                 .procname       = "tcp_notsent_lowat",
 537                 .data           = &sysctl_tcp_notsent_lowat,
 538                 .maxlen         = sizeof(sysctl_tcp_notsent_lowat),
 539                 .mode           = 0644,
 540                 .proc_handler   = proc_dointvec,
 541         },
 542         {
 543                 .procname       = "tcp_rmem",
 544                 .data           = &sysctl_tcp_rmem,
 545                 .maxlen         = sizeof(sysctl_tcp_rmem),
 546                 .mode           = 0644,
 547                 .proc_handler   = proc_dointvec_minmax,
 548                 .extra1         = &min_rcvbuf,
 549         },
 550         {
 551                 .procname       = "tcp_app_win",
 552                 .data           = &sysctl_tcp_app_win,
 553                 .maxlen         = sizeof(int),
 554                 .mode           = 0644,
 555                 .proc_handler   = proc_dointvec
 556         },
 557         {
 558                 .procname       = "tcp_adv_win_scale",
 559                 .data           = &sysctl_tcp_adv_win_scale,
 560                 .maxlen         = sizeof(int),
 561                 .mode           = 0644,
 562                 .proc_handler   = proc_dointvec_minmax,
 563                 .extra1         = &tcp_adv_win_scale_min,
 564                 .extra2         = &tcp_adv_win_scale_max,
 565         },
 566         {
 567                 .procname       = "tcp_tw_reuse",
 568                 .data           = &sysctl_tcp_tw_reuse,
 569                 .maxlen         = sizeof(int),
 570                 .mode           = 0644,
 571                 .proc_handler   = proc_dointvec
 572         },
 573         {
 574                 .procname       = "tcp_frto",
 575                 .data           = &sysctl_tcp_frto,
 576                 .maxlen         = sizeof(int),
 577                 .mode           = 0644,
 578                 .proc_handler   = proc_dointvec
 579         },
 580         {
 581                 .procname       = "tcp_low_latency",
 582                 .data           = &sysctl_tcp_low_latency,
 583                 .maxlen         = sizeof(int),
 584                 .mode           = 0644,
 585                 .proc_handler   = proc_dointvec
 586         },
 587         {
 588                 .procname       = "tcp_no_metrics_save",
 589                 .data           = &sysctl_tcp_nometrics_save,
 590                 .maxlen         = sizeof(int),
 591                 .mode           = 0644,
 592                 .proc_handler   = proc_dointvec,
 593         },
 594         {
 595                 .procname       = "tcp_moderate_rcvbuf",
 596                 .data           = &sysctl_tcp_moderate_rcvbuf,
 597                 .maxlen         = sizeof(int),
 598                 .mode           = 0644,
 599                 .proc_handler   = proc_dointvec,
 600         },
 601         {
 602                 .procname       = "tcp_tso_win_divisor",
 603                 .data           = &sysctl_tcp_tso_win_divisor,
 604                 .maxlen         = sizeof(int),
 605                 .mode           = 0644,
 606                 .proc_handler   = proc_dointvec,
 607         },
 608         {
 609                 .procname       = "tcp_congestion_control",
 610                 .mode           = 0644,
 611                 .maxlen         = TCP_CA_NAME_MAX,
 612                 .proc_handler   = proc_tcp_congestion_control,
 613         },
 614         {
 615                 .procname       = "tcp_workaround_signed_windows",
 616                 .data           = &sysctl_tcp_workaround_signed_windows,
 617                 .maxlen         = sizeof(int),
 618                 .mode           = 0644,
 619                 .proc_handler   = proc_dointvec
 620         },
 621         {
 622                 .procname       = "tcp_limit_output_bytes",
 623                 .data           = &sysctl_tcp_limit_output_bytes,
 624                 .maxlen         = sizeof(int),
 625                 .mode           = 0644,
 626                 .proc_handler   = proc_dointvec
 627         },
 628         {
 629                 .procname       = "tcp_challenge_ack_limit",
 630                 .data           = &sysctl_tcp_challenge_ack_limit,
 631                 .maxlen         = sizeof(int),
 632                 .mode           = 0644,
 633                 .proc_handler   = proc_dointvec
 634         },
 635         {
 636                 .procname       = "tcp_slow_start_after_idle",
 637                 .data           = &sysctl_tcp_slow_start_after_idle,
 638                 .maxlen         = sizeof(int),
 639                 .mode           = 0644,
 640                 .proc_handler   = proc_dointvec
 641         },
 642 #ifdef CONFIG_NETLABEL
 643         {
 644                 .procname       = "cipso_cache_enable",
 645                 .data           = &cipso_v4_cache_enabled,
 646                 .maxlen         = sizeof(int),
 647                 .mode           = 0644,
 648                 .proc_handler   = proc_dointvec,
 649         },
 650         {
 651                 .procname       = "cipso_cache_bucket_size",
 652                 .data           = &cipso_v4_cache_bucketsize,
 653                 .maxlen         = sizeof(int),
 654                 .mode           = 0644,
 655                 .proc_handler   = proc_dointvec,
 656         },
 657         {
 658                 .procname       = "cipso_rbm_optfmt",
 659                 .data           = &cipso_v4_rbm_optfmt,
 660                 .maxlen         = sizeof(int),
 661                 .mode           = 0644,
 662                 .proc_handler   = proc_dointvec,
 663         },
 664         {
 665                 .procname       = "cipso_rbm_strictvalid",
 666                 .data           = &cipso_v4_rbm_strictvalid,
 667                 .maxlen         = sizeof(int),
 668                 .mode           = 0644,
 669                 .proc_handler   = proc_dointvec,
 670         },
 671 #endif /* CONFIG_NETLABEL */
 672         {
 673                 .procname       = "tcp_available_congestion_control",
 674                 .maxlen         = TCP_CA_BUF_MAX,
 675                 .mode           = 0444,
 676                 .proc_handler   = proc_tcp_available_congestion_control,
 677         },
 678         {
 679                 .procname       = "tcp_allowed_congestion_control",
 680                 .maxlen         = TCP_CA_BUF_MAX,
 681                 .mode           = 0644,
 682                 .proc_handler   = proc_allowed_congestion_control,
 683         },
 684         {
 685                 .procname       = "tcp_thin_linear_timeouts",
 686                 .data           = &sysctl_tcp_thin_linear_timeouts,
 687                 .maxlen         = sizeof(int),
 688                 .mode           = 0644,
 689                 .proc_handler   = proc_dointvec
 690         },
 691         {
 692                 .procname       = "tcp_thin_dupack",
 693                 .data           = &sysctl_tcp_thin_dupack,
 694                 .maxlen         = sizeof(int),
 695                 .mode           = 0644,
 696                 .proc_handler   = proc_dointvec
 697         },
 698         {
 699                 .procname       = "tcp_early_retrans",
 700                 .data           = &sysctl_tcp_early_retrans,
 701                 .maxlen         = sizeof(int),
 702                 .mode           = 0644,
 703                 .proc_handler   = proc_dointvec_minmax,
 704                 .extra1         = &zero,
 705                 .extra2         = &four,
 706         },
 707         {
 708                 .procname       = "tcp_min_tso_segs",
 709                 .data           = &sysctl_tcp_min_tso_segs,
 710                 .maxlen         = sizeof(int),
 711                 .mode           = 0644,
 712                 .proc_handler   = proc_dointvec_minmax,
 713                 .extra1         = &one,
 714                 .extra2         = &gso_max_segs,
 715         },
 716         {
 717                 .procname       = "tcp_autocorking",
 718                 .data           = &sysctl_tcp_autocorking,
 719                 .maxlen         = sizeof(int),
 720                 .mode           = 0644,
 721                 .proc_handler   = proc_dointvec_minmax,
 722                 .extra1         = &zero,
 723                 .extra2         = &one,
 724         },
 725         {
 726                 .procname       = "tcp_invalid_ratelimit",
 727                 .data           = &sysctl_tcp_invalid_ratelimit,
 728                 .maxlen         = sizeof(int),
 729                 .mode           = 0644,
 730                 .proc_handler   = proc_dointvec_ms_jiffies,
 731         },
 732         {
 733                 .procname       = "icmp_msgs_per_sec",
 734                 .data           = &sysctl_icmp_msgs_per_sec,
 735                 .maxlen         = sizeof(int),
 736                 .mode           = 0644,
 737                 .proc_handler   = proc_dointvec_minmax,
 738                 .extra1         = &zero,
 739         },
 740         {
 741                 .procname       = "icmp_msgs_burst",
 742                 .data           = &sysctl_icmp_msgs_burst,
 743                 .maxlen         = sizeof(int),
 744                 .mode           = 0644,
 745                 .proc_handler   = proc_dointvec_minmax,
 746                 .extra1         = &zero,
 747         },
 748         {
 749                 .procname       = "udp_mem",
 750                 .data           = &sysctl_udp_mem,
 751                 .maxlen         = sizeof(sysctl_udp_mem),
 752                 .mode           = 0644,
 753                 .proc_handler   = proc_doulongvec_minmax,
 754         },
 755         {
 756                 .procname       = "udp_rmem_min",
 757                 .data           = &sysctl_udp_rmem_min,
 758                 .maxlen         = sizeof(sysctl_udp_rmem_min),
 759                 .mode           = 0644,
 760                 .proc_handler   = proc_dointvec_minmax,
 761                 .extra1         = &min_rcvbuf,
 762         },
 763         {
 764                 .procname       = "udp_wmem_min",
 765                 .data           = &sysctl_udp_wmem_min,
 766                 .maxlen         = sizeof(sysctl_udp_wmem_min),
 767                 .mode           = 0644,
 768                 .proc_handler   = proc_dointvec_minmax,
 769                 .extra1         = &min_sndbuf,
 770         },
 771         { }
 772 };
 773
 774 static struct ctl_table ipv4_net_table[] = {
 775         {
 776                 .procname       = "icmp_echo_ignore_all",
 777                 .data           = &init_net.ipv4.sysctl_icmp_echo_ignore_all,
 778                 .maxlen         = sizeof(int),
 779                 .mode           = 0644,
 780                 .proc_handler   = proc_dointvec
 781         },
 782         {
 783                 .procname       = "icmp_echo_ignore_broadcasts",
 784                 .data           = &init_net.ipv4.sysctl_icmp_echo_ignore_broadcasts,
 785                 .maxlen         = sizeof(int),
 786                 .mode           = 0644,
 787                 .proc_handler   = proc_dointvec
 788         },
 789         {
 790                 .procname       = "icmp_ignore_bogus_error_responses",
 791                 .data           = &init_net.ipv4.sysctl_icmp_ignore_bogus_error_responses,
 792                 .maxlen         = sizeof(int),
 793                 .mode           = 0644,
 794                 .proc_handler   = proc_dointvec
 795         },
 796         {
 797                 .procname       = "icmp_errors_use_inbound_ifaddr",
 798                 .data           = &init_net.ipv4.sysctl_icmp_errors_use_inbound_ifaddr,
 799                 .maxlen         = sizeof(int),
 800                 .mode           = 0644,
 801                 .proc_handler   = proc_dointvec
 802         },
 803         {
 804                 .procname       = "icmp_ratelimit",
 805                 .data           = &init_net.ipv4.sysctl_icmp_ratelimit,
 806                 .maxlen         = sizeof(int),
 807                 .mode           = 0644,
 808                 .proc_handler   = proc_dointvec_ms_jiffies,
 809         },
 810         {
 811                 .procname       = "icmp_ratemask",
 812                 .data           = &init_net.ipv4.sysctl_icmp_ratemask,
 813                 .maxlen         = sizeof(int),
 814                 .mode           = 0644,
 815                 .proc_handler   = proc_dointvec
 816         },
 817         {
 818                 .procname       = "ping_group_range",
 819                 .data           = &init_net.ipv4.ping_group_range.range,
 820                 .maxlen         = sizeof(gid_t)*2,
 821                 .mode           = 0644,
 822                 .proc_handler   = ipv4_ping_group_range,
 823         },
 824         {
 825                 .procname       = "tcp_ecn",
 826                 .data           = &init_net.ipv4.sysctl_tcp_ecn,
 827                 .maxlen         = sizeof(int),
 828                 .mode           = 0644,
 829                 .proc_handler   = proc_dointvec
 830         },
 831         {
 832                 .procname       = "tcp_ecn_fallback",
 833                 .data           = &init_net.ipv4.sysctl_tcp_ecn_fallback,
 834                 .maxlen         = sizeof(int),
 835                 .mode           = 0644,
 836                 .proc_handler   = proc_dointvec
 837         },
 838         {
 839                 .procname       = "ip_local_port_range",
 840                 .maxlen         = sizeof(init_net.ipv4.ip_local_ports.range),
 841                 .data           = &init_net.ipv4.ip_local_ports.range,
 842                 .mode           = 0644,
 843                 .proc_handler   = ipv4_local_port_range,
 844         },
 845         {
 846                 .procname       = "ip_local_reserved_ports",
 847                 .data           = &init_net.ipv4.sysctl_local_reserved_ports,
 848                 .maxlen         = 65536,
 849                 .mode           = 0644,
 850                 .proc_handler   = proc_do_large_bitmap,
 851         },
 852         {
 853                 .procname       = "ip_no_pmtu_disc",
 854                 .data           = &init_net.ipv4.sysctl_ip_no_pmtu_disc,
 855                 .maxlen         = sizeof(int),
 856                 .mode           = 0644,
 857                 .proc_handler   = proc_dointvec
 858         },
 859         {
 860                 .procname       = "ip_forward_use_pmtu",
 861                 .data           = &init_net.ipv4.sysctl_ip_fwd_use_pmtu,
 862                 .maxlen         = sizeof(int),
 863                 .mode           = 0644,
 864                 .proc_handler   = proc_dointvec,
 865         },
 866         {
 867                 .procname       = "ip_nonlocal_bind",
 868                 .data           = &init_net.ipv4.sysctl_ip_nonlocal_bind,
 869                 .maxlen         = sizeof(int),
 870                 .mode           = 0644,
 871                 .proc_handler   = proc_dointvec
 872         },
 873         {
 874                 .procname       = "fwmark_reflect",
 875                 .data           = &init_net.ipv4.sysctl_fwmark_reflect,
 876                 .maxlen         = sizeof(int),
 877                 .mode           = 0644,
 878                 .proc_handler   = proc_dointvec,
 879         },
 880         {
 881                 .procname       = "tcp_fwmark_accept",
 882                 .data           = &init_net.ipv4.sysctl_tcp_fwmark_accept,
 883                 .maxlen         = sizeof(int),
 884                 .mode           = 0644,
 885                 .proc_handler   = proc_dointvec,
 886         },
 887         {
 888                 .procname       = "tcp_mtu_probing",
 889                 .data           = &init_net.ipv4.sysctl_tcp_mtu_probing,
 890                 .maxlen         = sizeof(int),
 891                 .mode           = 0644,
 892                 .proc_handler   = proc_dointvec,
 893         },
 894         {
 895                 .procname       = "tcp_base_mss",
 896                 .data           = &init_net.ipv4.sysctl_tcp_base_mss,
 897                 .maxlen         = sizeof(int),
 898                 .mode           = 0644,
 899                 .proc_handler   = proc_dointvec,
 900         },
 901         {
 902                 .procname       = "tcp_probe_threshold",
 903                 .data           = &init_net.ipv4.sysctl_tcp_probe_threshold,
 904                 .maxlen         = sizeof(int),
 905                 .mode           = 0644,
 906                 .proc_handler   = proc_dointvec,
 907         },
 908         {
 909                 .procname       = "tcp_probe_interval",
 910                 .data           = &init_net.ipv4.sysctl_tcp_probe_interval,
 911                 .maxlen         = sizeof(int),
 912                 .mode           = 0644,
 913                 .proc_handler   = proc_dointvec,
 914         },
 915         { }
 916 };
 917
 918 static __net_init int ipv4_sysctl_init_net(struct net *net)
 919 {
 920         struct ctl_table *table;
 921
 922         table = ipv4_net_table;
 923         if (!net_eq(net, &init_net)) {
 924                 int i;
 925
 926                 table = kmemdup(table, sizeof(ipv4_net_table), GFP_KERNEL);
 927                 if (!table)
 928                         goto err_alloc;
 929
 930                 /* Update the variables to point into the current struct net */
 931                 for (i = 0; i < ARRAY_SIZE(ipv4_net_table) - 1; i++)
 932                         table[i].data += (void *)net - (void *)&init_net;
 933         }
 934
 935         net->ipv4.ipv4_hdr = register_net_sysctl(net, "net/ipv4", table);
 936         if (!net->ipv4.ipv4_hdr)
 937                 goto err_reg;
 938
 939         net->ipv4.sysctl_local_reserved_ports = kzalloc(65536 / 8, GFP_KERNEL);
 940         if (!net->ipv4.sysctl_local_reserved_ports)
 941                 goto err_ports;
 942
 943         return 0;
 944
 945 err_ports:
 946         unregister_net_sysctl_table(net->ipv4.ipv4_hdr);
 947 err_reg:
 948         if (!net_eq(net, &init_net))
 949                 kfree(table);
 950 err_alloc:
 951         return -ENOMEM;
 952 }
 953
 954 static __net_exit void ipv4_sysctl_exit_net(struct net *net)
 955 {
 956         struct ctl_table *table;
 957
 958         kfree(net->ipv4.sysctl_local_reserved_ports);
 959         table = net->ipv4.ipv4_hdr->ctl_table_arg;
 960         unregister_net_sysctl_table(net->ipv4.ipv4_hdr);
 961         kfree(table);
 962 }
 963
 964 static __net_initdata struct pernet_operations ipv4_sysctl_ops = {
 965         .init = ipv4_sysctl_init_net,
 966         .exit = ipv4_sysctl_exit_net,
 967 };
 968
 969 static __init int sysctl_ipv4_init(void)
 970 {
 971         struct ctl_table_header *hdr;
 972
 973         hdr = register_net_sysctl(&init_net, "net/ipv4", ipv4_table);
 974         if (!hdr)
 975                 return -ENOMEM;
 976
 977         if (register_pernet_subsys(&ipv4_sysctl_ops)) {
 978                 unregister_net_sysctl_table(hdr);
 979                 return -ENOMEM;
 980         }
 981
 982         return 0;
 983 }
 984
 985 __initcall(sysctl_ipv4_init);