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 thousand = 1000;
  33 static int gso_max_segs = GSO_MAX_SEGS;
  34 static int tcp_retr1_max = 255;
  35 static int ip_local_port_range_min[] = { 1, 1 };
  36 static int ip_local_port_range_max[] = { 65535, 65535 };
  37 static int tcp_adv_win_scale_min = -31;
  38 static int tcp_adv_win_scale_max = 31;
  39 static int ip_ttl_min = 1;
  40 static int ip_ttl_max = 255;
  41 static int tcp_syn_retries_min = 1;
  42 static int tcp_syn_retries_max = MAX_TCP_SYNCNT;
  43 static int ip_ping_group_range_min[] = { 0, 0 };
  44 static int ip_ping_group_range_max[] = { GID_T_MAX, GID_T_MAX };
  45 static int tcp_delack_seg_min = TCP_DELACK_MIN;
  46 static int tcp_delack_seg_max = 60;
  47 static int tcp_use_userconfig_min;
  48 static int tcp_use_userconfig_max = 1;
  49
  50 /* Update system visible IP port range */
  51 static void set_local_port_range(struct net *net, int range[2])
  52 {
  53         bool same_parity = !((range[0] ^ range[1]) & 1);
  54
  55         write_seqlock_bh(&net->ipv4.ip_local_ports.lock);
  56         if (same_parity && !net->ipv4.ip_local_ports.warned) {
  57                 net->ipv4.ip_local_ports.warned = true;
  58                 pr_err_ratelimited("ip_local_port_range: prefer different parity for start/end values.\n");
  59         }
  60         net->ipv4.ip_local_ports.range[0] = range[0];
  61         net->ipv4.ip_local_ports.range[1] = range[1];
  62         write_sequnlock_bh(&net->ipv4.ip_local_ports.lock);
  63 }
  64
  65 /* Validate changes from /proc interface. */
  66 static int ipv4_local_port_range(struct ctl_table *table, int write,
  67                                  void __user *buffer,
  68                                  size_t *lenp, loff_t *ppos)
  69 {
  70         struct net *net =
  71                 container_of(table->data, struct net, ipv4.ip_local_ports.range);
  72         int ret;
  73         int range[2];
  74         struct ctl_table tmp = {
  75                 .data = &range,
  76                 .maxlen = sizeof(range),
  77                 .mode = table->mode,
  78                 .extra1 = &ip_local_port_range_min,
  79                 .extra2 = &ip_local_port_range_max,
  80         };
  81
  82         inet_get_local_port_range(net, &range[0], &range[1]);
  83
  84         ret = proc_dointvec_minmax(&tmp, write, buffer, lenp, ppos);
  85
  86         if (write && ret == 0) {
  87                 if (range[1] < range[0])
  88                         ret = -EINVAL;
  89                 else
  90                         set_local_port_range(net, range);
  91         }
  92
  93         return ret;
  94 }
  95
  96
  97 static void inet_get_ping_group_range_table(struct ctl_table *table, kgid_t *low, kgid_t *high)
  98 {
  99         kgid_t *data = table->data;
 100         struct net *net =
 101                 container_of(table->data, struct net, ipv4.ping_group_range.range);
 102         unsigned int seq;
 103         do {
 104                 seq = read_seqbegin(&net->ipv4.ping_group_range.lock);
 105
 106                 *low = data[0];
 107                 *high = data[1];
 108         } while (read_seqretry(&net->ipv4.ping_group_range.lock, seq));
 109 }
 110
 111 /* Update system visible IP port range */
 112 static void set_ping_group_range(struct ctl_table *table, kgid_t low, kgid_t high)
 113 {
 114         kgid_t *data = table->data;
 115         struct net *net =
 116                 container_of(table->data, struct net, ipv4.ping_group_range.range);
 117         write_seqlock(&net->ipv4.ping_group_range.lock);
 118         data[0] = low;
 119         data[1] = high;
 120         write_sequnlock(&net->ipv4.ping_group_range.lock);
 121 }
 122
 123 /* Validate changes from /proc interface. */
 124 static int ipv4_ping_group_range(struct ctl_table *table, int write,
 125                                  void __user *buffer,
 126                                  size_t *lenp, loff_t *ppos)
 127 {
 128         struct user_namespace *user_ns = current_user_ns();
 129         int ret;
 130         gid_t urange[2];
 131         kgid_t low, high;
 132         struct ctl_table tmp = {
 133                 .data = &urange,
 134                 .maxlen = sizeof(urange),
 135                 .mode = table->mode,
 136                 .extra1 = &ip_ping_group_range_min,
 137                 .extra2 = &ip_ping_group_range_max,
 138         };
 139
 140         inet_get_ping_group_range_table(table, &low, &high);
 141         urange[0] = from_kgid_munged(user_ns, low);
 142         urange[1] = from_kgid_munged(user_ns, high);
 143         ret = proc_dointvec_minmax(&tmp, write, buffer, lenp, ppos);
 144
 145         if (write && ret == 0) {
 146                 low = make_kgid(user_ns, urange[0]);
 147                 high = make_kgid(user_ns, urange[1]);
 148                 if (!gid_valid(low) || !gid_valid(high))
 149                         return -EINVAL;
 150                 if (urange[1] < urange[0] || gid_lt(high, low)) {
 151                         low = make_kgid(&init_user_ns, 1);
 152                         high = make_kgid(&init_user_ns, 0);
 153                 }
 154                 set_ping_group_range(table, low, high);
 155         }
 156
 157         return ret;
 158 }
 159
 160 /* Validate changes from /proc interface. */
 161 static int proc_tcp_default_init_rwnd(struct ctl_table *ctl, int write,
 162                                       void __user *buffer,
 163                                       size_t *lenp, loff_t *ppos)
 164 {
 165         int old_value = *(int *)ctl->data;
 166         int ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
 167         int new_value = *(int *)ctl->data;
 168
 169         if (write && ret == 0 && (new_value < 3 || new_value > 100))
 170                 *(int *)ctl->data = old_value;
 171
 172         return ret;
 173 }
 174
 175 static int proc_tcp_congestion_control(struct ctl_table *ctl, int write,
 176                                        void __user *buffer, size_t *lenp, loff_t *ppos)
 177 {
 178         char val[TCP_CA_NAME_MAX];
 179         struct ctl_table tbl = {
 180                 .data = val,
 181                 .maxlen = TCP_CA_NAME_MAX,
 182         };
 183         int ret;
 184
 185         tcp_get_default_congestion_control(val);
 186
 187         ret = proc_dostring(&tbl, write, buffer, lenp, ppos);
 188         if (write && ret == 0)
 189                 ret = tcp_set_default_congestion_control(val);
 190         return ret;
 191 }
 192
 193 static int proc_tcp_available_congestion_control(struct ctl_table *ctl,
 194                                                  int write,
 195                                                  void __user *buffer, size_t *lenp,
 196                                                  loff_t *ppos)
 197 {
 198         struct ctl_table tbl = { .maxlen = TCP_CA_BUF_MAX, };
 199         int ret;
 200
 201         tbl.data = kmalloc(tbl.maxlen, GFP_USER);
 202         if (!tbl.data)
 203                 return -ENOMEM;
 204         tcp_get_available_congestion_control(tbl.data, TCP_CA_BUF_MAX);
 205         ret = proc_dostring(&tbl, write, buffer, lenp, ppos);
 206         kfree(tbl.data);
 207         return ret;
 208 }
 209
 210 static int proc_allowed_congestion_control(struct ctl_table *ctl,
 211                                            int write,
 212                                            void __user *buffer, size_t *lenp,
 213                                            loff_t *ppos)
 214 {
 215         struct ctl_table tbl = { .maxlen = TCP_CA_BUF_MAX };
 216         int ret;
 217
 218         tbl.data = kmalloc(tbl.maxlen, GFP_USER);
 219         if (!tbl.data)
 220                 return -ENOMEM;
 221
 222         tcp_get_allowed_congestion_control(tbl.data, tbl.maxlen);
 223         ret = proc_dostring(&tbl, write, buffer, lenp, ppos);
 224         if (write && ret == 0)
 225                 ret = tcp_set_allowed_congestion_control(tbl.data);
 226         kfree(tbl.data);
 227         return ret;
 228 }
 229
 230 static int proc_tcp_fastopen_key(struct ctl_table *ctl, int write,
 231                                  void __user *buffer, size_t *lenp,
 232                                  loff_t *ppos)
 233 {
 234         struct ctl_table tbl = { .maxlen = (TCP_FASTOPEN_KEY_LENGTH * 2 + 10) };
 235         struct tcp_fastopen_context *ctxt;
 236         u32  user_key[4]; /* 16 bytes, matching TCP_FASTOPEN_KEY_LENGTH */
 237         __le32 key[4];
 238         int ret, i;
 239
 240         tbl.data = kmalloc(tbl.maxlen, GFP_KERNEL);
 241         if (!tbl.data)
 242                 return -ENOMEM;
 243
 244         rcu_read_lock();
 245         ctxt = rcu_dereference(tcp_fastopen_ctx);
 246         if (ctxt)
 247                 memcpy(key, ctxt->key, TCP_FASTOPEN_KEY_LENGTH);
 248         else
 249                 memset(key, 0, sizeof(key));
 250         rcu_read_unlock();
 251
 252         for (i = 0; i < ARRAY_SIZE(key); i++)
 253                 user_key[i] = le32_to_cpu(key[i]);
 254
 255         snprintf(tbl.data, tbl.maxlen, "%08x-%08x-%08x-%08x",
 256                 user_key[0], user_key[1], user_key[2], user_key[3]);
 257         ret = proc_dostring(&tbl, write, buffer, lenp, ppos);
 258
 259         if (write && ret == 0) {
 260                 if (sscanf(tbl.data, "%x-%x-%x-%x", user_key, user_key + 1,
 261                            user_key + 2, user_key + 3) != 4) {
 262                         ret = -EINVAL;
 263                         goto bad_key;
 264                 }
 265                 /* Generate a dummy secret but don't publish it. This
 266                  * is needed so we don't regenerate a new key on the
 267                  * first invocation of tcp_fastopen_cookie_gen
 268                  */
 269                 tcp_fastopen_init_key_once(false);
 270
 271                 for (i = 0; i < ARRAY_SIZE(user_key); i++)
 272                         key[i] = cpu_to_le32(user_key[i]);
 273
 274                 tcp_fastopen_reset_cipher(key, TCP_FASTOPEN_KEY_LENGTH);
 275         }
 276
 277 bad_key:
 278         pr_debug("proc FO key set 0x%x-%x-%x-%x <- 0x%s: %u\n",
 279                  user_key[0], user_key[1], user_key[2], user_key[3],
 280                (char *)tbl.data, ret);
 281         kfree(tbl.data);
 282         return ret;
 283 }
 284
 285 static struct ctl_table ipv4_table[] = {
 286         {
 287                 .procname       = "tcp_timestamps",
 288                 .data           = &sysctl_tcp_timestamps,
 289                 .maxlen         = sizeof(int),
 290                 .mode           = 0644,
 291                 .proc_handler   = proc_dointvec
 292         },
 293         {
 294                 .procname       = "tcp_window_scaling",
 295                 .data           = &sysctl_tcp_window_scaling,
 296                 .maxlen         = sizeof(int),
 297                 .mode           = 0644,
 298                 .proc_handler   = proc_dointvec
 299         },
 300         {
 301                 .procname       = "tcp_sack",
 302                 .data           = &sysctl_tcp_sack,
 303                 .maxlen         = sizeof(int),
 304                 .mode           = 0644,
 305                 .proc_handler   = proc_dointvec
 306         },
 307         {
 308                 .procname       = "tcp_retrans_collapse",
 309                 .data           = &sysctl_tcp_retrans_collapse,
 310                 .maxlen         = sizeof(int),
 311                 .mode           = 0644,
 312                 .proc_handler   = proc_dointvec
 313         },
 314         {
 315                 .procname       = "ip_default_ttl",
 316                 .data           = &sysctl_ip_default_ttl,
 317                 .maxlen         = sizeof(int),
 318                 .mode           = 0644,
 319                 .proc_handler   = proc_dointvec_minmax,
 320                 .extra1         = &ip_ttl_min,
 321                 .extra2         = &ip_ttl_max,
 322         },
 323         {
 324                 .procname       = "tcp_syn_retries",
 325                 .data           = &sysctl_tcp_syn_retries,
 326                 .maxlen         = sizeof(int),
 327                 .mode           = 0644,
 328                 .proc_handler   = proc_dointvec_minmax,
 329                 .extra1         = &tcp_syn_retries_min,
 330                 .extra2         = &tcp_syn_retries_max
 331         },
 332         {
 333                 .procname       = "tcp_synack_retries",
 334                 .data           = &sysctl_tcp_synack_retries,
 335                 .maxlen         = sizeof(int),
 336                 .mode           = 0644,
 337                 .proc_handler   = proc_dointvec
 338         },
 339         {
 340                 .procname       = "tcp_max_orphans",
 341                 .data           = &sysctl_tcp_max_orphans,
 342                 .maxlen         = sizeof(int),
 343                 .mode           = 0644,
 344                 .proc_handler   = proc_dointvec
 345         },
 346         {
 347                 .procname       = "tcp_max_tw_buckets",
 348                 .data           = &tcp_death_row.sysctl_max_tw_buckets,
 349                 .maxlen         = sizeof(int),
 350                 .mode           = 0644,
 351                 .proc_handler   = proc_dointvec
 352         },
 353         {
 354                 .procname       = "ip_early_demux",
 355                 .data           = &sysctl_ip_early_demux,
 356                 .maxlen         = sizeof(int),
 357                 .mode           = 0644,
 358                 .proc_handler   = proc_dointvec
 359         },
 360         {
 361                 .procname       = "ip_dynaddr",
 362                 .data           = &sysctl_ip_dynaddr,
 363                 .maxlen         = sizeof(int),
 364                 .mode           = 0644,
 365                 .proc_handler   = proc_dointvec
 366         },
 367         {
 368                 .procname       = "tcp_keepalive_time",
 369                 .data           = &sysctl_tcp_keepalive_time,
 370                 .maxlen         = sizeof(int),
 371                 .mode           = 0644,
 372                 .proc_handler   = proc_dointvec_jiffies,
 373         },
 374         {
 375                 .procname       = "tcp_keepalive_probes",
 376                 .data           = &sysctl_tcp_keepalive_probes,
 377                 .maxlen         = sizeof(int),
 378                 .mode           = 0644,
 379                 .proc_handler   = proc_dointvec
 380         },
 381         {
 382                 .procname       = "tcp_keepalive_intvl",
 383                 .data           = &sysctl_tcp_keepalive_intvl,
 384                 .maxlen         = sizeof(int),
 385                 .mode           = 0644,
 386                 .proc_handler   = proc_dointvec_jiffies,
 387         },
 388         {
 389                 .procname       = "tcp_retries1",
 390                 .data           = &sysctl_tcp_retries1,
 391                 .maxlen         = sizeof(int),
 392                 .mode           = 0644,
 393                 .proc_handler   = proc_dointvec_minmax,
 394                 .extra2         = &tcp_retr1_max
 395         },
 396         {
 397                 .procname       = "tcp_retries2",
 398                 .data           = &sysctl_tcp_retries2,
 399                 .maxlen         = sizeof(int),
 400                 .mode           = 0644,
 401                 .proc_handler   = proc_dointvec
 402         },
 403         {
 404                 .procname       = "tcp_fin_timeout",
 405                 .data           = &sysctl_tcp_fin_timeout,
 406                 .maxlen         = sizeof(int),
 407                 .mode           = 0644,
 408                 .proc_handler   = proc_dointvec_jiffies,
 409         },
 410 #ifdef CONFIG_SYN_COOKIES
 411         {
 412                 .procname       = "tcp_syncookies",
 413                 .data           = &sysctl_tcp_syncookies,
 414                 .maxlen         = sizeof(int),
 415                 .mode           = 0644,
 416                 .proc_handler   = proc_dointvec
 417         },
 418 #endif
 419         {
 420                 .procname       = "tcp_fastopen",
 421                 .data           = &sysctl_tcp_fastopen,
 422                 .maxlen         = sizeof(int),
 423                 .mode           = 0644,
 424                 .proc_handler   = proc_dointvec,
 425         },
 426         {
 427                 .procname       = "tcp_fastopen_key",
 428                 .mode           = 0600,
 429                 .maxlen         = ((TCP_FASTOPEN_KEY_LENGTH * 2) + 10),
 430                 .proc_handler   = proc_tcp_fastopen_key,
 431         },
 432         {
 433                 .procname       = "tcp_tw_recycle",
 434                 .data           = &tcp_death_row.sysctl_tw_recycle,
 435                 .maxlen         = sizeof(int),
 436                 .mode           = 0644,
 437                 .proc_handler   = proc_dointvec
 438         },
 439         {
 440                 .procname       = "tcp_abort_on_overflow",
 441                 .data           = &sysctl_tcp_abort_on_overflow,
 442                 .maxlen         = sizeof(int),
 443                 .mode           = 0644,
 444                 .proc_handler   = proc_dointvec
 445         },
 446         {
 447                 .procname       = "tcp_stdurg",
 448                 .data           = &sysctl_tcp_stdurg,
 449                 .maxlen         = sizeof(int),
 450                 .mode           = 0644,
 451                 .proc_handler   = proc_dointvec
 452         },
 453         {
 454                 .procname       = "tcp_rfc1337",
 455                 .data           = &sysctl_tcp_rfc1337,
 456                 .maxlen         = sizeof(int),
 457                 .mode           = 0644,
 458                 .proc_handler   = proc_dointvec
 459         },
 460         {
 461                 .procname       = "tcp_max_syn_backlog",
 462                 .data           = &sysctl_max_syn_backlog,
 463                 .maxlen         = sizeof(int),
 464                 .mode           = 0644,
 465                 .proc_handler   = proc_dointvec
 466         },
 467         {
 468                 .procname       = "igmp_max_memberships",
 469                 .data           = &sysctl_igmp_max_memberships,
 470                 .maxlen         = sizeof(int),
 471                 .mode           = 0644,
 472                 .proc_handler   = proc_dointvec
 473         },
 474         {
 475                 .procname       = "igmp_max_msf",
 476                 .data           = &sysctl_igmp_max_msf,
 477                 .maxlen         = sizeof(int),
 478                 .mode           = 0644,
 479                 .proc_handler   = proc_dointvec
 480         },
 481 #ifdef CONFIG_IP_MULTICAST
 482         {
 483                 .procname       = "igmp_qrv",
 484                 .data           = &sysctl_igmp_qrv,
 485                 .maxlen         = sizeof(int),
 486                 .mode           = 0644,
 487                 .proc_handler   = proc_dointvec_minmax,
 488                 .extra1         = &one
 489         },
 490 #endif
 491         {
 492                 .procname       = "inet_peer_threshold",
 493                 .data           = &inet_peer_threshold,
 494                 .maxlen         = sizeof(int),
 495                 .mode           = 0644,
 496                 .proc_handler   = proc_dointvec
 497         },
 498         {
 499                 .procname       = "inet_peer_minttl",
 500                 .data           = &inet_peer_minttl,
 501                 .maxlen         = sizeof(int),
 502                 .mode           = 0644,
 503                 .proc_handler   = proc_dointvec_jiffies,
 504         },
 505         {
 506                 .procname       = "inet_peer_maxttl",
 507                 .data           = &inet_peer_maxttl,
 508                 .maxlen         = sizeof(int),
 509                 .mode           = 0644,
 510                 .proc_handler   = proc_dointvec_jiffies,
 511         },
 512         {
 513                 .procname       = "tcp_orphan_retries",
 514                 .data           = &sysctl_tcp_orphan_retries,
 515                 .maxlen         = sizeof(int),
 516                 .mode           = 0644,
 517                 .proc_handler   = proc_dointvec
 518         },
 519         {
 520                 .procname       = "tcp_fack",
 521                 .data           = &sysctl_tcp_fack,
 522                 .maxlen         = sizeof(int),
 523                 .mode           = 0644,
 524                 .proc_handler   = proc_dointvec
 525         },
 526         {
 527                 .procname       = "tcp_recovery",
 528                 .data           = &sysctl_tcp_recovery,
 529                 .maxlen         = sizeof(int),
 530                 .mode           = 0644,
 531                 .proc_handler   = proc_dointvec,
 532         },
 533         {
 534                 .procname       = "tcp_reordering",
 535                 .data           = &sysctl_tcp_reordering,
 536                 .maxlen         = sizeof(int),
 537                 .mode           = 0644,
 538                 .proc_handler   = proc_dointvec
 539         },
 540         {
 541                 .procname       = "tcp_max_reordering",
 542                 .data           = &sysctl_tcp_max_reordering,
 543                 .maxlen         = sizeof(int),
 544                 .mode           = 0644,
 545                 .proc_handler   = proc_dointvec
 546         },
 547         {
 548                 .procname       = "tcp_dsack",
 549                 .data           = &sysctl_tcp_dsack,
 550                 .maxlen         = sizeof(int),
 551                 .mode           = 0644,
 552                 .proc_handler   = proc_dointvec
 553         },
 554         {
 555                 .procname       = "tcp_mem",
 556                 .maxlen         = sizeof(sysctl_tcp_mem),
 557                 .data           = &sysctl_tcp_mem,
 558                 .mode           = 0644,
 559                 .proc_handler   = proc_doulongvec_minmax,
 560         },
 561         {
 562                 .procname       = "tcp_wmem",
 563                 .data           = &sysctl_tcp_wmem,
 564                 .maxlen         = sizeof(sysctl_tcp_wmem),
 565                 .mode           = 0644,
 566                 .proc_handler   = proc_dointvec_minmax,
 567                 .extra1         = &one,
 568         },
 569         {
 570                 .procname       = "tcp_notsent_lowat",
 571                 .data           = &sysctl_tcp_notsent_lowat,
 572                 .maxlen         = sizeof(sysctl_tcp_notsent_lowat),
 573                 .mode           = 0644,
 574                 .proc_handler   = proc_dointvec,
 575         },
 576         {
 577                 .procname       = "tcp_rmem",
 578                 .data           = &sysctl_tcp_rmem,
 579                 .maxlen         = sizeof(sysctl_tcp_rmem),
 580                 .mode           = 0644,
 581                 .proc_handler   = proc_dointvec_minmax,
 582                 .extra1         = &one,
 583         },
 584         {
 585                 .procname       = "tcp_app_win",
 586                 .data           = &sysctl_tcp_app_win,
 587                 .maxlen         = sizeof(int),
 588                 .mode           = 0644,
 589                 .proc_handler   = proc_dointvec
 590         },
 591         {
 592                 .procname       = "tcp_adv_win_scale",
 593                 .data           = &sysctl_tcp_adv_win_scale,
 594                 .maxlen         = sizeof(int),
 595                 .mode           = 0644,
 596                 .proc_handler   = proc_dointvec_minmax,
 597                 .extra1         = &tcp_adv_win_scale_min,
 598                 .extra2         = &tcp_adv_win_scale_max,
 599         },
 600         {
 601                 .procname       = "tcp_tw_reuse",
 602                 .data           = &sysctl_tcp_tw_reuse,
 603                 .maxlen         = sizeof(int),
 604                 .mode           = 0644,
 605                 .proc_handler   = proc_dointvec
 606         },
 607         {
 608                 .procname       = "tcp_frto",
 609                 .data           = &sysctl_tcp_frto,
 610                 .maxlen         = sizeof(int),
 611                 .mode           = 0644,
 612                 .proc_handler   = proc_dointvec
 613         },
 614         {
 615                 .procname       = "tcp_min_rtt_wlen",
 616                 .data           = &sysctl_tcp_min_rtt_wlen,
 617                 .maxlen         = sizeof(int),
 618                 .mode           = 0644,
 619                 .proc_handler   = proc_dointvec
 620         },
 621         {
 622                 .procname       = "tcp_low_latency",
 623                 .data           = &sysctl_tcp_low_latency,
 624                 .maxlen         = sizeof(int),
 625                 .mode           = 0644,
 626                 .proc_handler   = proc_dointvec
 627         },
 628         {
 629                 .procname       = "tcp_no_metrics_save",
 630                 .data           = &sysctl_tcp_nometrics_save,
 631                 .maxlen         = sizeof(int),
 632                 .mode           = 0644,
 633                 .proc_handler   = proc_dointvec,
 634         },
 635         {
 636                 .procname       = "tcp_moderate_rcvbuf",
 637                 .data           = &sysctl_tcp_moderate_rcvbuf,
 638                 .maxlen         = sizeof(int),
 639                 .mode           = 0644,
 640                 .proc_handler   = proc_dointvec,
 641         },
 642         {
 643                 .procname       = "tcp_tso_win_divisor",
 644                 .data           = &sysctl_tcp_tso_win_divisor,
 645                 .maxlen         = sizeof(int),
 646                 .mode           = 0644,
 647                 .proc_handler   = proc_dointvec,
 648         },
 649         {
 650                 .procname       = "tcp_congestion_control",
 651                 .mode           = 0644,
 652                 .maxlen         = TCP_CA_NAME_MAX,
 653                 .proc_handler   = proc_tcp_congestion_control,
 654         },
 655         {
 656                 .procname       = "tcp_workaround_signed_windows",
 657                 .data           = &sysctl_tcp_workaround_signed_windows,
 658                 .maxlen         = sizeof(int),
 659                 .mode           = 0644,
 660                 .proc_handler   = proc_dointvec
 661         },
 662         {
 663                 .procname       = "tcp_limit_output_bytes",
 664                 .data           = &sysctl_tcp_limit_output_bytes,
 665                 .maxlen         = sizeof(int),
 666                 .mode           = 0644,
 667                 .proc_handler   = proc_dointvec
 668         },
 669         {
 670                 .procname       = "tcp_challenge_ack_limit",
 671                 .data           = &sysctl_tcp_challenge_ack_limit,
 672                 .maxlen         = sizeof(int),
 673                 .mode           = 0644,
 674                 .proc_handler   = proc_dointvec
 675         },
 676         {
 677                 .procname       = "tcp_slow_start_after_idle",
 678                 .data           = &sysctl_tcp_slow_start_after_idle,
 679                 .maxlen         = sizeof(int),
 680                 .mode           = 0644,
 681                 .proc_handler   = proc_dointvec
 682         },
 683 #ifdef CONFIG_NETLABEL
 684         {
 685                 .procname       = "cipso_cache_enable",
 686                 .data           = &cipso_v4_cache_enabled,
 687                 .maxlen         = sizeof(int),
 688                 .mode           = 0644,
 689                 .proc_handler   = proc_dointvec,
 690         },
 691         {
 692                 .procname       = "cipso_cache_bucket_size",
 693                 .data           = &cipso_v4_cache_bucketsize,
 694                 .maxlen         = sizeof(int),
 695                 .mode           = 0644,
 696                 .proc_handler   = proc_dointvec,
 697         },
 698         {
 699                 .procname       = "cipso_rbm_optfmt",
 700                 .data           = &cipso_v4_rbm_optfmt,
 701                 .maxlen         = sizeof(int),
 702                 .mode           = 0644,
 703                 .proc_handler   = proc_dointvec,
 704         },
 705         {
 706                 .procname       = "cipso_rbm_strictvalid",
 707                 .data           = &cipso_v4_rbm_strictvalid,
 708                 .maxlen         = sizeof(int),
 709                 .mode           = 0644,
 710                 .proc_handler   = proc_dointvec,
 711         },
 712 #endif /* CONFIG_NETLABEL */
 713         {
 714                 .procname       = "tcp_available_congestion_control",
 715                 .maxlen         = TCP_CA_BUF_MAX,
 716                 .mode           = 0444,
 717                 .proc_handler   = proc_tcp_available_congestion_control,
 718         },
 719         {
 720                 .procname       = "tcp_allowed_congestion_control",
 721                 .maxlen         = TCP_CA_BUF_MAX,
 722                 .mode           = 0644,
 723                 .proc_handler   = proc_allowed_congestion_control,
 724         },
 725         {
 726                 .procname       = "tcp_thin_linear_timeouts",
 727                 .data           = &sysctl_tcp_thin_linear_timeouts,
 728                 .maxlen         = sizeof(int),
 729                 .mode           = 0644,
 730                 .proc_handler   = proc_dointvec
 731         },
 732         {
 733                 .procname       = "tcp_thin_dupack",
 734                 .data           = &sysctl_tcp_thin_dupack,
 735                 .maxlen         = sizeof(int),
 736                 .mode           = 0644,
 737                 .proc_handler   = proc_dointvec
 738         },
 739         {
 740                 .procname       = "tcp_early_retrans",
 741                 .data           = &sysctl_tcp_early_retrans,
 742                 .maxlen         = sizeof(int),
 743                 .mode           = 0644,
 744                 .proc_handler   = proc_dointvec_minmax,
 745                 .extra1         = &zero,
 746                 .extra2         = &four,
 747         },
 748         {
 749                 .procname       = "tcp_min_tso_segs",
 750                 .data           = &sysctl_tcp_min_tso_segs,
 751                 .maxlen         = sizeof(int),
 752                 .mode           = 0644,
 753                 .proc_handler   = proc_dointvec_minmax,
 754                 .extra1         = &one,
 755                 .extra2         = &gso_max_segs,
 756         },
 757         {
 758                 .procname       = "tcp_pacing_ss_ratio",
 759                 .data           = &sysctl_tcp_pacing_ss_ratio,
 760                 .maxlen         = sizeof(int),
 761                 .mode           = 0644,
 762                 .proc_handler   = proc_dointvec_minmax,
 763                 .extra1         = &zero,
 764                 .extra2         = &thousand,
 765         },
 766         {
 767                 .procname       = "tcp_pacing_ca_ratio",
 768                 .data           = &sysctl_tcp_pacing_ca_ratio,
 769                 .maxlen         = sizeof(int),
 770                 .mode           = 0644,
 771                 .proc_handler   = proc_dointvec_minmax,
 772                 .extra1         = &zero,
 773                 .extra2         = &thousand,
 774         },
 775         {
 776                 .procname       = "tcp_autocorking",
 777                 .data           = &sysctl_tcp_autocorking,
 778                 .maxlen         = sizeof(int),
 779                 .mode           = 0644,
 780                 .proc_handler   = proc_dointvec_minmax,
 781                 .extra1         = &zero,
 782                 .extra2         = &one,
 783         },
 784         {
 785                 .procname       = "tcp_invalid_ratelimit",
 786                 .data           = &sysctl_tcp_invalid_ratelimit,
 787                 .maxlen         = sizeof(int),
 788                 .mode           = 0644,
 789                 .proc_handler   = proc_dointvec_ms_jiffies,
 790         },
 791         {
 792                 .procname       = "tcp_default_init_rwnd",
 793                 .data           = &sysctl_tcp_default_init_rwnd,
 794                 .maxlen         = sizeof(int),
 795                 .mode           = 0644,
 796                 .proc_handler   = proc_tcp_default_init_rwnd
 797         },
 798         {
 799                 .procname       = "icmp_msgs_per_sec",
 800                 .data           = &sysctl_icmp_msgs_per_sec,
 801                 .maxlen         = sizeof(int),
 802                 .mode           = 0644,
 803                 .proc_handler   = proc_dointvec_minmax,
 804                 .extra1         = &zero,
 805         },
 806         {
 807                 .procname       = "icmp_msgs_burst",
 808                 .data           = &sysctl_icmp_msgs_burst,
 809                 .maxlen         = sizeof(int),
 810                 .mode           = 0644,
 811                 .proc_handler   = proc_dointvec_minmax,
 812                 .extra1         = &zero,
 813         },
 814         {
 815                 .procname       = "udp_mem",
 816                 .data           = &sysctl_udp_mem,
 817                 .maxlen         = sizeof(sysctl_udp_mem),
 818                 .mode           = 0644,
 819                 .proc_handler   = proc_doulongvec_minmax,
 820         },
 821         {
 822                 .procname       = "udp_rmem_min",
 823                 .data           = &sysctl_udp_rmem_min,
 824                 .maxlen         = sizeof(sysctl_udp_rmem_min),
 825                 .mode           = 0644,
 826                 .proc_handler   = proc_dointvec_minmax,
 827                 .extra1         = &one
 828         },
 829         {
 830                 .procname       = "udp_wmem_min",
 831                 .data           = &sysctl_udp_wmem_min,
 832                 .maxlen         = sizeof(sysctl_udp_wmem_min),
 833                 .mode           = 0644,
 834                 .proc_handler   = proc_dointvec_minmax,
 835                 .extra1         = &one
 836         },
 837         {
 838                 .procname       = "tcp_delack_seg",
 839                 .data           = &sysctl_tcp_delack_seg,
 840                 .maxlen         = sizeof(sysctl_tcp_delack_seg),
 841                 .mode           = 0644,
 842                 .proc_handler   = tcp_proc_delayed_ack_control,
 843                 .extra1         = &tcp_delack_seg_min,
 844                 .extra2         = &tcp_delack_seg_max,
 845         },
 846         {
 847                 .procname       = "tcp_use_userconfig",
 848                 .data           = &sysctl_tcp_use_userconfig,
 849                 .maxlen         = sizeof(sysctl_tcp_use_userconfig),
 850                 .mode           = 0644,
 851                 .proc_handler   = tcp_use_userconfig_sysctl_handler,
 852                 .extra1         = &tcp_use_userconfig_min,
 853                 .extra2         = &tcp_use_userconfig_max,
 854         },
 855
 856         { }
 857 };
 858
 859 static struct ctl_table ipv4_net_table[] = {
 860         {
 861                 .procname       = "icmp_echo_ignore_all",
 862                 .data           = &init_net.ipv4.sysctl_icmp_echo_ignore_all,
 863                 .maxlen         = sizeof(int),
 864                 .mode           = 0644,
 865                 .proc_handler   = proc_dointvec
 866         },
 867         {
 868                 .procname       = "icmp_echo_ignore_broadcasts",
 869                 .data           = &init_net.ipv4.sysctl_icmp_echo_ignore_broadcasts,
 870                 .maxlen         = sizeof(int),
 871                 .mode           = 0644,
 872                 .proc_handler   = proc_dointvec
 873         },
 874         {
 875                 .procname       = "icmp_ignore_bogus_error_responses",
 876                 .data           = &init_net.ipv4.sysctl_icmp_ignore_bogus_error_responses,
 877                 .maxlen         = sizeof(int),
 878                 .mode           = 0644,
 879                 .proc_handler   = proc_dointvec
 880         },
 881         {
 882                 .procname       = "icmp_errors_use_inbound_ifaddr",
 883                 .data           = &init_net.ipv4.sysctl_icmp_errors_use_inbound_ifaddr,
 884                 .maxlen         = sizeof(int),
 885                 .mode           = 0644,
 886                 .proc_handler   = proc_dointvec
 887         },
 888         {
 889                 .procname       = "icmp_ratelimit",
 890                 .data           = &init_net.ipv4.sysctl_icmp_ratelimit,
 891                 .maxlen         = sizeof(int),
 892                 .mode           = 0644,
 893                 .proc_handler   = proc_dointvec_ms_jiffies,
 894         },
 895         {
 896                 .procname       = "icmp_ratemask",
 897                 .data           = &init_net.ipv4.sysctl_icmp_ratemask,
 898                 .maxlen         = sizeof(int),
 899                 .mode           = 0644,
 900                 .proc_handler   = proc_dointvec
 901         },
 902         {
 903                 .procname       = "ping_group_range",
 904                 .data           = &init_net.ipv4.ping_group_range.range,
 905                 .maxlen         = sizeof(gid_t)*2,
 906                 .mode           = 0644,
 907                 .proc_handler   = ipv4_ping_group_range,
 908         },
 909         {
 910                 .procname       = "tcp_ecn",
 911                 .data           = &init_net.ipv4.sysctl_tcp_ecn,
 912                 .maxlen         = sizeof(int),
 913                 .mode           = 0644,
 914                 .proc_handler   = proc_dointvec
 915         },
 916         {
 917                 .procname       = "tcp_ecn_fallback",
 918                 .data           = &init_net.ipv4.sysctl_tcp_ecn_fallback,
 919                 .maxlen         = sizeof(int),
 920                 .mode           = 0644,
 921                 .proc_handler   = proc_dointvec
 922         },
 923         {
 924                 .procname       = "ip_local_port_range",
 925                 .maxlen         = sizeof(init_net.ipv4.ip_local_ports.range),
 926                 .data           = &init_net.ipv4.ip_local_ports.range,
 927                 .mode           = 0644,
 928                 .proc_handler   = ipv4_local_port_range,
 929         },
 930         {
 931                 .procname       = "ip_local_reserved_ports",
 932                 .data           = &init_net.ipv4.sysctl_local_reserved_ports,
 933                 .maxlen         = 65536,
 934                 .mode           = 0644,
 935                 .proc_handler   = proc_do_large_bitmap,
 936         },
 937         {
 938                 .procname       = "reserved_port_bind",
 939                 .data           = &sysctl_reserved_port_bind,
 940                 .maxlen         = sizeof(int),
 941                 .mode           = 0644,
 942                 .proc_handler   = proc_dointvec
 943         },
 944         {
 945                 .procname       = "ip_no_pmtu_disc",
 946                 .data           = &init_net.ipv4.sysctl_ip_no_pmtu_disc,
 947                 .maxlen         = sizeof(int),
 948                 .mode           = 0644,
 949                 .proc_handler   = proc_dointvec
 950         },
 951         {
 952                 .procname       = "ip_forward_use_pmtu",
 953                 .data           = &init_net.ipv4.sysctl_ip_fwd_use_pmtu,
 954                 .maxlen         = sizeof(int),
 955                 .mode           = 0644,
 956                 .proc_handler   = proc_dointvec,
 957         },
 958         {
 959                 .procname       = "ip_nonlocal_bind",
 960                 .data           = &init_net.ipv4.sysctl_ip_nonlocal_bind,
 961                 .maxlen         = sizeof(int),
 962                 .mode           = 0644,
 963                 .proc_handler   = proc_dointvec
 964         },
 965         {
 966                 .procname       = "fwmark_reflect",
 967                 .data           = &init_net.ipv4.sysctl_fwmark_reflect,
 968                 .maxlen         = sizeof(int),
 969                 .mode           = 0644,
 970                 .proc_handler   = proc_dointvec,
 971         },
 972         {
 973                 .procname       = "tcp_fwmark_accept",
 974                 .data           = &init_net.ipv4.sysctl_tcp_fwmark_accept,
 975                 .maxlen         = sizeof(int),
 976                 .mode           = 0644,
 977                 .proc_handler   = proc_dointvec,
 978         },
 979         {
 980                 .procname       = "tcp_mtu_probing",
 981                 .data           = &init_net.ipv4.sysctl_tcp_mtu_probing,
 982                 .maxlen         = sizeof(int),
 983                 .mode           = 0644,
 984                 .proc_handler   = proc_dointvec,
 985         },
 986         {
 987                 .procname       = "tcp_base_mss",
 988                 .data           = &init_net.ipv4.sysctl_tcp_base_mss,
 989                 .maxlen         = sizeof(int),
 990                 .mode           = 0644,
 991                 .proc_handler   = proc_dointvec,
 992         },
 993         {
 994                 .procname       = "tcp_probe_threshold",
 995                 .data           = &init_net.ipv4.sysctl_tcp_probe_threshold,
 996                 .maxlen         = sizeof(int),
 997                 .mode           = 0644,
 998                 .proc_handler   = proc_dointvec,
 999         },
1000         {
1001                 .procname       = "tcp_probe_interval",
1002                 .data           = &init_net.ipv4.sysctl_tcp_probe_interval,
1003                 .maxlen         = sizeof(int),
1004                 .mode           = 0644,
1005                 .proc_handler   = proc_dointvec,
1006         },
1007         {
1008                 .procname       = "igmp_link_local_mcast_reports",
1009                 .data           = &sysctl_igmp_llm_reports,
1010                 .maxlen         = sizeof(int),
1011                 .mode           = 0644,
1012                 .proc_handler   = proc_dointvec
1013         },
1014         { }
1015 };
1016
1017 static __net_init int ipv4_sysctl_init_net(struct net *net)
1018 {
1019         struct ctl_table *table;
1020
1021         table = ipv4_net_table;
1022         if (!net_eq(net, &init_net)) {
1023                 int i;
1024
1025                 table = kmemdup(table, sizeof(ipv4_net_table), GFP_KERNEL);
1026                 if (!table)
1027                         goto err_alloc;
1028
1029                 /* Update the variables to point into the current struct net */
1030                 for (i = 0; i < ARRAY_SIZE(ipv4_net_table) - 1; i++)
1031                         table[i].data += (void *)net - (void *)&init_net;
1032         }
1033
1034         net->ipv4.ipv4_hdr = register_net_sysctl(net, "net/ipv4", table);
1035         if (!net->ipv4.ipv4_hdr)
1036                 goto err_reg;
1037
1038         net->ipv4.sysctl_local_reserved_ports = kzalloc(65536 / 8, GFP_KERNEL);
1039         if (!net->ipv4.sysctl_local_reserved_ports)
1040                 goto err_ports;
1041
1042         return 0;
1043
1044 err_ports:
1045         unregister_net_sysctl_table(net->ipv4.ipv4_hdr);
1046 err_reg:
1047         if (!net_eq(net, &init_net))
1048                 kfree(table);
1049 err_alloc:
1050         return -ENOMEM;
1051 }
1052
1053 static __net_exit void ipv4_sysctl_exit_net(struct net *net)
1054 {
1055         struct ctl_table *table;
1056
1057         kfree(net->ipv4.sysctl_local_reserved_ports);
1058         table = net->ipv4.ipv4_hdr->ctl_table_arg;
1059         unregister_net_sysctl_table(net->ipv4.ipv4_hdr);
1060         kfree(table);
1061 }
1062
1063 static __net_initdata struct pernet_operations ipv4_sysctl_ops = {
1064         .init = ipv4_sysctl_init_net,
1065         .exit = ipv4_sysctl_exit_net,
1066 };
1067
1068 static __init int sysctl_ipv4_init(void)
1069 {
1070         struct ctl_table_header *hdr;
1071
1072         hdr = register_net_sysctl(&init_net, "net/ipv4", ipv4_table);
1073         if (!hdr)
1074                 return -ENOMEM;
1075
1076         if (register_pernet_subsys(&ipv4_sysctl_ops)) {
1077                 unregister_net_sysctl_table(hdr);
1078                 return -ENOMEM;
1079         }
1080
1081         return 0;
1082 }
1083
1084 __initcall(sysctl_ipv4_init);