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
  45 /* Update system visible IP port range */
  46 static void set_local_port_range(struct net *net, int range[2])
  47 {
  48         write_seqlock(&net->ipv4.sysctl_local_ports.lock);
  49         net->ipv4.sysctl_local_ports.range[0] = range[0];
  50         net->ipv4.sysctl_local_ports.range[1] = range[1];
  51         write_sequnlock(&net->ipv4.sysctl_local_ports.lock);
  52 }
  53
  54 /* Validate changes from /proc interface. */
  55 static int ipv4_local_port_range(struct ctl_table *table, int write,
  56                                  void __user *buffer,
  57                                  size_t *lenp, loff_t *ppos)
  58 {
  59         struct net *net =
  60                 container_of(table->data, struct net, ipv4.sysctl_local_ports.range);
  61         int ret;
  62         int range[2];
  63         struct ctl_table tmp = {
  64                 .data = &range,
  65                 .maxlen = sizeof(range),
  66                 .mode = table->mode,
  67                 .extra1 = &ip_local_port_range_min,
  68                 .extra2 = &ip_local_port_range_max,
  69         };
  70
  71         inet_get_local_port_range(net, &range[0], &range[1]);
  72
  73         ret = proc_dointvec_minmax(&tmp, write, buffer, lenp, ppos);
  74
  75         if (write && ret == 0) {
  76                 if (range[1] < range[0])
  77                         ret = -EINVAL;
  78                 else
  79                         set_local_port_range(net, range);
  80         }
  81
  82         return ret;
  83 }
  84
  85
  86 static void inet_get_ping_group_range_table(struct ctl_table *table, kgid_t *low, kgid_t *high)
  87 {
  88         kgid_t *data = table->data;
  89         struct net *net =
  90                 container_of(table->data, struct net, ipv4.sysctl_ping_group_range);
  91         unsigned int seq;
  92         do {
  93                 seq = read_seqbegin(&net->ipv4.sysctl_local_ports.lock);
  94
  95                 *low = data[0];
  96                 *high = data[1];
  97         } while (read_seqretry(&net->ipv4.sysctl_local_ports.lock, seq));
  98 }
  99
 100 /* Update system visible IP port range */
 101 static void set_ping_group_range(struct ctl_table *table, kgid_t low, kgid_t high)
 102 {
 103         kgid_t *data = table->data;
 104         struct net *net =
 105                 container_of(table->data, struct net, ipv4.sysctl_ping_group_range);
 106         write_seqlock(&net->ipv4.sysctl_local_ports.lock);
 107         data[0] = low;
 108         data[1] = high;
 109         write_sequnlock(&net->ipv4.sysctl_local_ports.lock);
 110 }
 111
 112 /* Validate changes from /proc interface. */
 113 static int ipv4_ping_group_range(struct ctl_table *table, int write,
 114                                  void __user *buffer,
 115                                  size_t *lenp, loff_t *ppos)
 116 {
 117         struct user_namespace *user_ns = current_user_ns();
 118         int ret;
 119         gid_t urange[2];
 120         kgid_t low, high;
 121         struct ctl_table tmp = {
 122                 .data = &urange,
 123                 .maxlen = sizeof(urange),
 124                 .mode = table->mode,
 125                 .extra1 = &ip_ping_group_range_min,
 126                 .extra2 = &ip_ping_group_range_max,
 127         };
 128
 129         inet_get_ping_group_range_table(table, &low, &high);
 130         urange[0] = from_kgid_munged(user_ns, low);
 131         urange[1] = from_kgid_munged(user_ns, high);
 132         ret = proc_dointvec_minmax(&tmp, write, buffer, lenp, ppos);
 133
 134         if (write && ret == 0) {
 135                 low = make_kgid(user_ns, urange[0]);
 136                 high = make_kgid(user_ns, urange[1]);
 137                 if (!gid_valid(low) || !gid_valid(high) ||
 138                     (urange[1] < urange[0]) || gid_lt(high, low)) {
 139                         low = make_kgid(&init_user_ns, 1);
 140                         high = make_kgid(&init_user_ns, 0);
 141                 }
 142                 set_ping_group_range(table, low, high);
 143         }
 144
 145         return ret;
 146 }
 147
 148 static int proc_tcp_congestion_control(struct ctl_table *ctl, int write,
 149                                        void __user *buffer, size_t *lenp, loff_t *ppos)
 150 {
 151         char val[TCP_CA_NAME_MAX];
 152         struct ctl_table tbl = {
 153                 .data = val,
 154                 .maxlen = TCP_CA_NAME_MAX,
 155         };
 156         int ret;
 157
 158         tcp_get_default_congestion_control(val);
 159
 160         ret = proc_dostring(&tbl, write, buffer, lenp, ppos);
 161         if (write && ret == 0)
 162                 ret = tcp_set_default_congestion_control(val);
 163         return ret;
 164 }
 165
 166 static int proc_tcp_available_congestion_control(struct ctl_table *ctl,
 167                                                  int write,
 168                                                  void __user *buffer, size_t *lenp,
 169                                                  loff_t *ppos)
 170 {
 171         struct ctl_table tbl = { .maxlen = TCP_CA_BUF_MAX, };
 172         int ret;
 173
 174         tbl.data = kmalloc(tbl.maxlen, GFP_USER);
 175         if (!tbl.data)
 176                 return -ENOMEM;
 177         tcp_get_available_congestion_control(tbl.data, TCP_CA_BUF_MAX);
 178         ret = proc_dostring(&tbl, write, buffer, lenp, ppos);
 179         kfree(tbl.data);
 180         return ret;
 181 }
 182
 183 static int proc_allowed_congestion_control(struct ctl_table *ctl,
 184                                            int write,
 185                                            void __user *buffer, size_t *lenp,
 186                                            loff_t *ppos)
 187 {
 188         struct ctl_table tbl = { .maxlen = TCP_CA_BUF_MAX };
 189         int ret;
 190
 191         tbl.data = kmalloc(tbl.maxlen, GFP_USER);
 192         if (!tbl.data)
 193                 return -ENOMEM;
 194
 195         tcp_get_allowed_congestion_control(tbl.data, tbl.maxlen);
 196         ret = proc_dostring(&tbl, write, buffer, lenp, ppos);
 197         if (write && ret == 0)
 198                 ret = tcp_set_allowed_congestion_control(tbl.data);
 199         kfree(tbl.data);
 200         return ret;
 201 }
 202
 203 static int proc_tcp_fastopen_key(struct ctl_table *ctl, int write,
 204                                  void __user *buffer, size_t *lenp,
 205                                  loff_t *ppos)
 206 {
 207         struct ctl_table tbl = { .maxlen = (TCP_FASTOPEN_KEY_LENGTH * 2 + 10) };
 208         struct tcp_fastopen_context *ctxt;
 209         int ret;
 210         u32  user_key[4]; /* 16 bytes, matching TCP_FASTOPEN_KEY_LENGTH */
 211
 212         tbl.data = kmalloc(tbl.maxlen, GFP_KERNEL);
 213         if (!tbl.data)
 214                 return -ENOMEM;
 215
 216         rcu_read_lock();
 217         ctxt = rcu_dereference(tcp_fastopen_ctx);
 218         if (ctxt)
 219                 memcpy(user_key, ctxt->key, TCP_FASTOPEN_KEY_LENGTH);
 220         else
 221                 memset(user_key, 0, sizeof(user_key));
 222         rcu_read_unlock();
 223
 224         snprintf(tbl.data, tbl.maxlen, "%08x-%08x-%08x-%08x",
 225                 user_key[0], user_key[1], user_key[2], user_key[3]);
 226         ret = proc_dostring(&tbl, write, buffer, lenp, ppos);
 227
 228         if (write && ret == 0) {
 229                 if (sscanf(tbl.data, "%x-%x-%x-%x", user_key, user_key + 1,
 230                            user_key + 2, user_key + 3) != 4) {
 231                         ret = -EINVAL;
 232                         goto bad_key;
 233                 }
 234                 /* Generate a dummy secret but don't publish it. This
 235                  * is needed so we don't regenerate a new key on the
 236                  * first invocation of tcp_fastopen_cookie_gen
 237                  */
 238                 tcp_fastopen_init_key_once(false);
 239                 tcp_fastopen_reset_cipher(user_key, TCP_FASTOPEN_KEY_LENGTH);
 240         }
 241
 242 bad_key:
 243         pr_debug("proc FO key set 0x%x-%x-%x-%x <- 0x%s: %u\n",
 244                user_key[0], user_key[1], user_key[2], user_key[3],
 245                (char *)tbl.data, ret);
 246         kfree(tbl.data);
 247         return ret;
 248 }
 249
 250 static struct ctl_table ipv4_table[] = {
 251         {
 252                 .procname       = "tcp_timestamps",
 253                 .data           = &sysctl_tcp_timestamps,
 254                 .maxlen         = sizeof(int),
 255                 .mode           = 0644,
 256                 .proc_handler   = proc_dointvec
 257         },
 258         {
 259                 .procname       = "tcp_window_scaling",
 260                 .data           = &sysctl_tcp_window_scaling,
 261                 .maxlen         = sizeof(int),
 262                 .mode           = 0644,
 263                 .proc_handler   = proc_dointvec
 264         },
 265         {
 266                 .procname       = "tcp_sack",
 267                 .data           = &sysctl_tcp_sack,
 268                 .maxlen         = sizeof(int),
 269                 .mode           = 0644,
 270                 .proc_handler   = proc_dointvec
 271         },
 272         {
 273                 .procname       = "tcp_retrans_collapse",
 274                 .data           = &sysctl_tcp_retrans_collapse,
 275                 .maxlen         = sizeof(int),
 276                 .mode           = 0644,
 277                 .proc_handler   = proc_dointvec
 278         },
 279         {
 280                 .procname       = "ip_default_ttl",
 281                 .data           = &sysctl_ip_default_ttl,
 282                 .maxlen         = sizeof(int),
 283                 .mode           = 0644,
 284                 .proc_handler   = proc_dointvec_minmax,
 285                 .extra1         = &ip_ttl_min,
 286                 .extra2         = &ip_ttl_max,
 287         },
 288         {
 289                 .procname       = "ip_nonlocal_bind",
 290                 .data           = &sysctl_ip_nonlocal_bind,
 291                 .maxlen         = sizeof(int),
 292                 .mode           = 0644,
 293                 .proc_handler   = proc_dointvec
 294         },
 295         {
 296                 .procname       = "tcp_syn_retries",
 297                 .data           = &sysctl_tcp_syn_retries,
 298                 .maxlen         = sizeof(int),
 299                 .mode           = 0644,
 300                 .proc_handler   = proc_dointvec_minmax,
 301                 .extra1         = &tcp_syn_retries_min,
 302                 .extra2         = &tcp_syn_retries_max
 303         },
 304         {
 305                 .procname       = "tcp_synack_retries",
 306                 .data           = &sysctl_tcp_synack_retries,
 307                 .maxlen         = sizeof(int),
 308                 .mode           = 0644,
 309                 .proc_handler   = proc_dointvec
 310         },
 311         {
 312                 .procname       = "tcp_max_orphans",
 313                 .data           = &sysctl_tcp_max_orphans,
 314                 .maxlen         = sizeof(int),
 315                 .mode           = 0644,
 316                 .proc_handler   = proc_dointvec
 317         },
 318         {
 319                 .procname       = "tcp_max_tw_buckets",
 320                 .data           = &tcp_death_row.sysctl_max_tw_buckets,
 321                 .maxlen         = sizeof(int),
 322                 .mode           = 0644,
 323                 .proc_handler   = proc_dointvec
 324         },
 325         {
 326                 .procname       = "ip_early_demux",
 327                 .data           = &sysctl_ip_early_demux,
 328                 .maxlen         = sizeof(int),
 329                 .mode           = 0644,
 330                 .proc_handler   = proc_dointvec
 331         },
 332         {
 333                 .procname       = "ip_dynaddr",
 334                 .data           = &sysctl_ip_dynaddr,
 335                 .maxlen         = sizeof(int),
 336                 .mode           = 0644,
 337                 .proc_handler   = proc_dointvec
 338         },
 339         {
 340                 .procname       = "tcp_keepalive_time",
 341                 .data           = &sysctl_tcp_keepalive_time,
 342                 .maxlen         = sizeof(int),
 343                 .mode           = 0644,
 344                 .proc_handler   = proc_dointvec_jiffies,
 345         },
 346         {
 347                 .procname       = "tcp_keepalive_probes",
 348                 .data           = &sysctl_tcp_keepalive_probes,
 349                 .maxlen         = sizeof(int),
 350                 .mode           = 0644,
 351                 .proc_handler   = proc_dointvec
 352         },
 353         {
 354                 .procname       = "tcp_keepalive_intvl",
 355                 .data           = &sysctl_tcp_keepalive_intvl,
 356                 .maxlen         = sizeof(int),
 357                 .mode           = 0644,
 358                 .proc_handler   = proc_dointvec_jiffies,
 359         },
 360         {
 361                 .procname       = "tcp_retries1",
 362                 .data           = &sysctl_tcp_retries1,
 363                 .maxlen         = sizeof(int),
 364                 .mode           = 0644,
 365                 .proc_handler   = proc_dointvec_minmax,
 366                 .extra2         = &tcp_retr1_max
 367         },
 368         {
 369                 .procname       = "tcp_retries2",
 370                 .data           = &sysctl_tcp_retries2,
 371                 .maxlen         = sizeof(int),
 372                 .mode           = 0644,
 373                 .proc_handler   = proc_dointvec
 374         },
 375         {
 376                 .procname       = "tcp_fin_timeout",
 377                 .data           = &sysctl_tcp_fin_timeout,
 378                 .maxlen         = sizeof(int),
 379                 .mode           = 0644,
 380                 .proc_handler   = proc_dointvec_jiffies,
 381         },
 382 #ifdef CONFIG_SYN_COOKIES
 383         {
 384                 .procname       = "tcp_syncookies",
 385                 .data           = &sysctl_tcp_syncookies,
 386                 .maxlen         = sizeof(int),
 387                 .mode           = 0644,
 388                 .proc_handler   = proc_dointvec
 389         },
 390 #endif
 391         {
 392                 .procname       = "tcp_fastopen",
 393                 .data           = &sysctl_tcp_fastopen,
 394                 .maxlen         = sizeof(int),
 395                 .mode           = 0644,
 396                 .proc_handler   = proc_dointvec,
 397         },
 398         {
 399                 .procname       = "tcp_fastopen_key",
 400                 .mode           = 0600,
 401                 .maxlen         = ((TCP_FASTOPEN_KEY_LENGTH * 2) + 10),
 402                 .proc_handler   = proc_tcp_fastopen_key,
 403         },
 404         {
 405                 .procname       = "tcp_tw_recycle",
 406                 .data           = &tcp_death_row.sysctl_tw_recycle,
 407                 .maxlen         = sizeof(int),
 408                 .mode           = 0644,
 409                 .proc_handler   = proc_dointvec
 410         },
 411         {
 412                 .procname       = "tcp_abort_on_overflow",
 413                 .data           = &sysctl_tcp_abort_on_overflow,
 414                 .maxlen         = sizeof(int),
 415                 .mode           = 0644,
 416                 .proc_handler   = proc_dointvec
 417         },
 418         {
 419                 .procname       = "tcp_stdurg",
 420                 .data           = &sysctl_tcp_stdurg,
 421                 .maxlen         = sizeof(int),
 422                 .mode           = 0644,
 423                 .proc_handler   = proc_dointvec
 424         },
 425         {
 426                 .procname       = "tcp_rfc1337",
 427                 .data           = &sysctl_tcp_rfc1337,
 428                 .maxlen         = sizeof(int),
 429                 .mode           = 0644,
 430                 .proc_handler   = proc_dointvec
 431         },
 432         {
 433                 .procname       = "tcp_max_syn_backlog",
 434                 .data           = &sysctl_max_syn_backlog,
 435                 .maxlen         = sizeof(int),
 436                 .mode           = 0644,
 437                 .proc_handler   = proc_dointvec
 438         },
 439         {
 440                 .procname       = "ip_local_reserved_ports",
 441                 .data           = NULL, /* initialized in sysctl_ipv4_init */
 442                 .maxlen         = 65536,
 443                 .mode           = 0644,
 444                 .proc_handler   = proc_do_large_bitmap,
 445         },
 446         {
 447                 .procname       = "igmp_max_memberships",
 448                 .data           = &sysctl_igmp_max_memberships,
 449                 .maxlen         = sizeof(int),
 450                 .mode           = 0644,
 451                 .proc_handler   = proc_dointvec
 452         },
 453         {
 454                 .procname       = "igmp_max_msf",
 455                 .data           = &sysctl_igmp_max_msf,
 456                 .maxlen         = sizeof(int),
 457                 .mode           = 0644,
 458                 .proc_handler   = proc_dointvec
 459         },
 460         {
 461                 .procname       = "inet_peer_threshold",
 462                 .data           = &inet_peer_threshold,
 463                 .maxlen         = sizeof(int),
 464                 .mode           = 0644,
 465                 .proc_handler   = proc_dointvec
 466         },
 467         {
 468                 .procname       = "inet_peer_minttl",
 469                 .data           = &inet_peer_minttl,
 470                 .maxlen         = sizeof(int),
 471                 .mode           = 0644,
 472                 .proc_handler   = proc_dointvec_jiffies,
 473         },
 474         {
 475                 .procname       = "inet_peer_maxttl",
 476                 .data           = &inet_peer_maxttl,
 477                 .maxlen         = sizeof(int),
 478                 .mode           = 0644,
 479                 .proc_handler   = proc_dointvec_jiffies,
 480         },
 481         {
 482                 .procname       = "tcp_orphan_retries",
 483                 .data           = &sysctl_tcp_orphan_retries,
 484                 .maxlen         = sizeof(int),
 485                 .mode           = 0644,
 486                 .proc_handler   = proc_dointvec
 487         },
 488         {
 489                 .procname       = "tcp_fack",
 490                 .data           = &sysctl_tcp_fack,
 491                 .maxlen         = sizeof(int),
 492                 .mode           = 0644,
 493                 .proc_handler   = proc_dointvec
 494         },
 495         {
 496                 .procname       = "tcp_reordering",
 497                 .data           = &sysctl_tcp_reordering,
 498                 .maxlen         = sizeof(int),
 499                 .mode           = 0644,
 500                 .proc_handler   = proc_dointvec
 501         },
 502         {
 503                 .procname       = "tcp_dsack",
 504                 .data           = &sysctl_tcp_dsack,
 505                 .maxlen         = sizeof(int),
 506                 .mode           = 0644,
 507                 .proc_handler   = proc_dointvec
 508         },
 509         {
 510                 .procname       = "tcp_mem",
 511                 .maxlen         = sizeof(sysctl_tcp_mem),
 512                 .data           = &sysctl_tcp_mem,
 513                 .mode           = 0644,
 514                 .proc_handler   = proc_doulongvec_minmax,
 515         },
 516         {
 517                 .procname       = "tcp_wmem",
 518                 .data           = &sysctl_tcp_wmem,
 519                 .maxlen         = sizeof(sysctl_tcp_wmem),
 520                 .mode           = 0644,
 521                 .proc_handler   = proc_dointvec_minmax,
 522                 .extra1         = &one,
 523         },
 524         {
 525                 .procname       = "tcp_notsent_lowat",
 526                 .data           = &sysctl_tcp_notsent_lowat,
 527                 .maxlen         = sizeof(sysctl_tcp_notsent_lowat),
 528                 .mode           = 0644,
 529                 .proc_handler   = proc_dointvec,
 530         },
 531         {
 532                 .procname       = "tcp_rmem",
 533                 .data           = &sysctl_tcp_rmem,
 534                 .maxlen         = sizeof(sysctl_tcp_rmem),
 535                 .mode           = 0644,
 536                 .proc_handler   = proc_dointvec_minmax,
 537                 .extra1         = &one,
 538         },
 539         {
 540                 .procname       = "tcp_app_win",
 541                 .data           = &sysctl_tcp_app_win,
 542                 .maxlen         = sizeof(int),
 543                 .mode           = 0644,
 544                 .proc_handler   = proc_dointvec
 545         },
 546         {
 547                 .procname       = "tcp_adv_win_scale",
 548                 .data           = &sysctl_tcp_adv_win_scale,
 549                 .maxlen         = sizeof(int),
 550                 .mode           = 0644,
 551                 .proc_handler   = proc_dointvec_minmax,
 552                 .extra1         = &tcp_adv_win_scale_min,
 553                 .extra2         = &tcp_adv_win_scale_max,
 554         },
 555         {
 556                 .procname       = "tcp_tw_reuse",
 557                 .data           = &sysctl_tcp_tw_reuse,
 558                 .maxlen         = sizeof(int),
 559                 .mode           = 0644,
 560                 .proc_handler   = proc_dointvec
 561         },
 562         {
 563                 .procname       = "tcp_frto",
 564                 .data           = &sysctl_tcp_frto,
 565                 .maxlen         = sizeof(int),
 566                 .mode           = 0644,
 567                 .proc_handler   = proc_dointvec
 568         },
 569         {
 570                 .procname       = "tcp_low_latency",
 571                 .data           = &sysctl_tcp_low_latency,
 572                 .maxlen         = sizeof(int),
 573                 .mode           = 0644,
 574                 .proc_handler   = proc_dointvec
 575         },
 576         {
 577                 .procname       = "tcp_no_metrics_save",
 578                 .data           = &sysctl_tcp_nometrics_save,
 579                 .maxlen         = sizeof(int),
 580                 .mode           = 0644,
 581                 .proc_handler   = proc_dointvec,
 582         },
 583         {
 584                 .procname       = "tcp_moderate_rcvbuf",
 585                 .data           = &sysctl_tcp_moderate_rcvbuf,
 586                 .maxlen         = sizeof(int),
 587                 .mode           = 0644,
 588                 .proc_handler   = proc_dointvec,
 589         },
 590         {
 591                 .procname       = "tcp_tso_win_divisor",
 592                 .data           = &sysctl_tcp_tso_win_divisor,
 593                 .maxlen         = sizeof(int),
 594                 .mode           = 0644,
 595                 .proc_handler   = proc_dointvec,
 596         },
 597         {
 598                 .procname       = "tcp_congestion_control",
 599                 .mode           = 0644,
 600                 .maxlen         = TCP_CA_NAME_MAX,
 601                 .proc_handler   = proc_tcp_congestion_control,
 602         },
 603         {
 604                 .procname       = "tcp_mtu_probing",
 605                 .data           = &sysctl_tcp_mtu_probing,
 606                 .maxlen         = sizeof(int),
 607                 .mode           = 0644,
 608                 .proc_handler   = proc_dointvec,
 609         },
 610         {
 611                 .procname       = "tcp_base_mss",
 612                 .data           = &sysctl_tcp_base_mss,
 613                 .maxlen         = sizeof(int),
 614                 .mode           = 0644,
 615                 .proc_handler   = proc_dointvec,
 616         },
 617         {
 618                 .procname       = "tcp_workaround_signed_windows",
 619                 .data           = &sysctl_tcp_workaround_signed_windows,
 620                 .maxlen         = sizeof(int),
 621                 .mode           = 0644,
 622                 .proc_handler   = proc_dointvec
 623         },
 624         {
 625                 .procname       = "tcp_limit_output_bytes",
 626                 .data           = &sysctl_tcp_limit_output_bytes,
 627                 .maxlen         = sizeof(int),
 628                 .mode           = 0644,
 629                 .proc_handler   = proc_dointvec
 630         },
 631         {
 632                 .procname       = "tcp_challenge_ack_limit",
 633                 .data           = &sysctl_tcp_challenge_ack_limit,
 634                 .maxlen         = sizeof(int),
 635                 .mode           = 0644,
 636                 .proc_handler   = proc_dointvec
 637         },
 638 #ifdef CONFIG_NET_DMA
 639         {
 640                 .procname       = "tcp_dma_copybreak",
 641                 .data           = &sysctl_tcp_dma_copybreak,
 642                 .maxlen         = sizeof(int),
 643                 .mode           = 0644,
 644                 .proc_handler   = proc_dointvec
 645         },
 646 #endif
 647         {
 648                 .procname       = "tcp_slow_start_after_idle",
 649                 .data           = &sysctl_tcp_slow_start_after_idle,
 650                 .maxlen         = sizeof(int),
 651                 .mode           = 0644,
 652                 .proc_handler   = proc_dointvec
 653         },
 654 #ifdef CONFIG_NETLABEL
 655         {
 656                 .procname       = "cipso_cache_enable",
 657                 .data           = &cipso_v4_cache_enabled,
 658                 .maxlen         = sizeof(int),
 659                 .mode           = 0644,
 660                 .proc_handler   = proc_dointvec,
 661         },
 662         {
 663                 .procname       = "cipso_cache_bucket_size",
 664                 .data           = &cipso_v4_cache_bucketsize,
 665                 .maxlen         = sizeof(int),
 666                 .mode           = 0644,
 667                 .proc_handler   = proc_dointvec,
 668         },
 669         {
 670                 .procname       = "cipso_rbm_optfmt",
 671                 .data           = &cipso_v4_rbm_optfmt,
 672                 .maxlen         = sizeof(int),
 673                 .mode           = 0644,
 674                 .proc_handler   = proc_dointvec,
 675         },
 676         {
 677                 .procname       = "cipso_rbm_strictvalid",
 678                 .data           = &cipso_v4_rbm_strictvalid,
 679                 .maxlen         = sizeof(int),
 680                 .mode           = 0644,
 681                 .proc_handler   = proc_dointvec,
 682         },
 683 #endif /* CONFIG_NETLABEL */
 684         {
 685                 .procname       = "tcp_available_congestion_control",
 686                 .maxlen         = TCP_CA_BUF_MAX,
 687                 .mode           = 0444,
 688                 .proc_handler   = proc_tcp_available_congestion_control,
 689         },
 690         {
 691                 .procname       = "tcp_allowed_congestion_control",
 692                 .maxlen         = TCP_CA_BUF_MAX,
 693                 .mode           = 0644,
 694                 .proc_handler   = proc_allowed_congestion_control,
 695         },
 696         {
 697                 .procname       = "tcp_thin_linear_timeouts",
 698                 .data           = &sysctl_tcp_thin_linear_timeouts,
 699                 .maxlen         = sizeof(int),
 700                 .mode           = 0644,
 701                 .proc_handler   = proc_dointvec
 702         },
 703         {
 704                 .procname       = "tcp_thin_dupack",
 705                 .data           = &sysctl_tcp_thin_dupack,
 706                 .maxlen         = sizeof(int),
 707                 .mode           = 0644,
 708                 .proc_handler   = proc_dointvec
 709         },
 710         {
 711                 .procname       = "tcp_early_retrans",
 712                 .data           = &sysctl_tcp_early_retrans,
 713                 .maxlen         = sizeof(int),
 714                 .mode           = 0644,
 715                 .proc_handler   = proc_dointvec_minmax,
 716                 .extra1         = &zero,
 717                 .extra2         = &four,
 718         },
 719         {
 720                 .procname       = "tcp_min_tso_segs",
 721                 .data           = &sysctl_tcp_min_tso_segs,
 722                 .maxlen         = sizeof(int),
 723                 .mode           = 0644,
 724                 .proc_handler   = proc_dointvec_minmax,
 725                 .extra1         = &zero,
 726                 .extra2         = &gso_max_segs,
 727         },
 728         {
 729                 .procname       = "tcp_autocorking",
 730                 .data           = &sysctl_tcp_autocorking,
 731                 .maxlen         = sizeof(int),
 732                 .mode           = 0644,
 733                 .proc_handler   = proc_dointvec_minmax,
 734                 .extra1         = &zero,
 735                 .extra2         = &one,
 736         },
 737         {
 738                 .procname       = "udp_mem",
 739                 .data           = &sysctl_udp_mem,
 740                 .maxlen         = sizeof(sysctl_udp_mem),
 741                 .mode           = 0644,
 742                 .proc_handler   = proc_doulongvec_minmax,
 743         },
 744         {
 745                 .procname       = "udp_rmem_min",
 746                 .data           = &sysctl_udp_rmem_min,
 747                 .maxlen         = sizeof(sysctl_udp_rmem_min),
 748                 .mode           = 0644,
 749                 .proc_handler   = proc_dointvec_minmax,
 750                 .extra1         = &one
 751         },
 752         {
 753                 .procname       = "udp_wmem_min",
 754                 .data           = &sysctl_udp_wmem_min,
 755                 .maxlen         = sizeof(sysctl_udp_wmem_min),
 756                 .mode           = 0644,
 757                 .proc_handler   = proc_dointvec_minmax,
 758                 .extra1         = &one
 759         },
 760         { }
 761 };
 762
 763 static struct ctl_table ipv4_net_table[] = {
 764         {
 765                 .procname       = "icmp_echo_ignore_all",
 766                 .data           = &init_net.ipv4.sysctl_icmp_echo_ignore_all,
 767                 .maxlen         = sizeof(int),
 768                 .mode           = 0644,
 769                 .proc_handler   = proc_dointvec
 770         },
 771         {
 772                 .procname       = "icmp_echo_ignore_broadcasts",
 773                 .data           = &init_net.ipv4.sysctl_icmp_echo_ignore_broadcasts,
 774                 .maxlen         = sizeof(int),
 775                 .mode           = 0644,
 776                 .proc_handler   = proc_dointvec
 777         },
 778         {
 779                 .procname       = "icmp_ignore_bogus_error_responses",
 780                 .data           = &init_net.ipv4.sysctl_icmp_ignore_bogus_error_responses,
 781                 .maxlen         = sizeof(int),
 782                 .mode           = 0644,
 783                 .proc_handler   = proc_dointvec
 784         },
 785         {
 786                 .procname       = "icmp_errors_use_inbound_ifaddr",
 787                 .data           = &init_net.ipv4.sysctl_icmp_errors_use_inbound_ifaddr,
 788                 .maxlen         = sizeof(int),
 789                 .mode           = 0644,
 790                 .proc_handler   = proc_dointvec
 791         },
 792         {
 793                 .procname       = "icmp_ratelimit",
 794                 .data           = &init_net.ipv4.sysctl_icmp_ratelimit,
 795                 .maxlen         = sizeof(int),
 796                 .mode           = 0644,
 797                 .proc_handler   = proc_dointvec_ms_jiffies,
 798         },
 799         {
 800                 .procname       = "icmp_ratemask",
 801                 .data           = &init_net.ipv4.sysctl_icmp_ratemask,
 802                 .maxlen         = sizeof(int),
 803                 .mode           = 0644,
 804                 .proc_handler   = proc_dointvec
 805         },
 806         {
 807                 .procname       = "ping_group_range",
 808                 .data           = &init_net.ipv4.sysctl_ping_group_range,
 809                 .maxlen         = sizeof(gid_t)*2,
 810                 .mode           = 0644,
 811                 .proc_handler   = ipv4_ping_group_range,
 812         },
 813         {
 814                 .procname       = "tcp_ecn",
 815                 .data           = &init_net.ipv4.sysctl_tcp_ecn,
 816                 .maxlen         = sizeof(int),
 817                 .mode           = 0644,
 818                 .proc_handler   = proc_dointvec
 819         },
 820         {
 821                 .procname       = "ip_local_port_range",
 822                 .maxlen         = sizeof(init_net.ipv4.sysctl_local_ports.range),
 823                 .data           = &init_net.ipv4.sysctl_local_ports.range,
 824                 .mode           = 0644,
 825                 .proc_handler   = ipv4_local_port_range,
 826         },
 827         {
 828                 .procname       = "ip_no_pmtu_disc",
 829                 .data           = &init_net.ipv4.sysctl_ip_no_pmtu_disc,
 830                 .maxlen         = sizeof(int),
 831                 .mode           = 0644,
 832                 .proc_handler   = proc_dointvec
 833         },
 834         {
 835                 .procname       = "ip_forward_use_pmtu",
 836                 .data           = &init_net.ipv4.sysctl_ip_fwd_use_pmtu,
 837                 .maxlen         = sizeof(int),
 838                 .mode           = 0644,
 839                 .proc_handler   = proc_dointvec,
 840         },
 841         { }
 842 };
 843
 844 static __net_init int ipv4_sysctl_init_net(struct net *net)
 845 {
 846         struct ctl_table *table;
 847
 848         table = ipv4_net_table;
 849         if (!net_eq(net, &init_net)) {
 850                 int i;
 851
 852                 table = kmemdup(table, sizeof(ipv4_net_table), GFP_KERNEL);
 853                 if (table == NULL)
 854                         goto err_alloc;
 855
 856                 /* Update the variables to point into the current struct net */
 857                 for (i = 0; i < ARRAY_SIZE(ipv4_net_table) - 1; i++)
 858                         table[i].data += (void *)net - (void *)&init_net;
 859         }
 860
 861         /*
 862          * Sane defaults - nobody may create ping sockets.
 863          * Boot scripts should set this to distro-specific group.
 864          */
 865         net->ipv4.sysctl_ping_group_range[0] = make_kgid(&init_user_ns, 1);
 866         net->ipv4.sysctl_ping_group_range[1] = make_kgid(&init_user_ns, 0);
 867
 868         /*
 869          * Set defaults for local port range
 870          */
 871         seqlock_init(&net->ipv4.sysctl_local_ports.lock);
 872         net->ipv4.sysctl_local_ports.range[0] =  32768;
 873         net->ipv4.sysctl_local_ports.range[1] =  61000;
 874
 875         net->ipv4.ipv4_hdr = register_net_sysctl(net, "net/ipv4", table);
 876         if (net->ipv4.ipv4_hdr == NULL)
 877                 goto err_reg;
 878
 879         return 0;
 880
 881 err_reg:
 882         if (!net_eq(net, &init_net))
 883                 kfree(table);
 884 err_alloc:
 885         return -ENOMEM;
 886 }
 887
 888 static __net_exit void ipv4_sysctl_exit_net(struct net *net)
 889 {
 890         struct ctl_table *table;
 891
 892         table = net->ipv4.ipv4_hdr->ctl_table_arg;
 893         unregister_net_sysctl_table(net->ipv4.ipv4_hdr);
 894         kfree(table);
 895 }
 896
 897 static __net_initdata struct pernet_operations ipv4_sysctl_ops = {
 898         .init = ipv4_sysctl_init_net,
 899         .exit = ipv4_sysctl_exit_net,
 900 };
 901
 902 static __init int sysctl_ipv4_init(void)
 903 {
 904         struct ctl_table_header *hdr;
 905         struct ctl_table *i;
 906
 907         for (i = ipv4_table; i->procname; i++) {
 908                 if (strcmp(i->procname, "ip_local_reserved_ports") == 0) {
 909                         i->data = sysctl_local_reserved_ports;
 910                         break;
 911                 }
 912         }
 913         if (!i->procname)
 914                 return -EINVAL;
 915
 916         hdr = register_net_sysctl(&init_net, "net/ipv4", ipv4_table);
 917         if (hdr == NULL)
 918                 return -ENOMEM;
 919
 920         if (register_pernet_subsys(&ipv4_sysctl_ops)) {
 921                 unregister_net_sysctl_table(hdr);
 922                 return -ENOMEM;
 923         }
 924
 925         return 0;
 926 }
 927
 928 __initcall(sysctl_ipv4_init);