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 tcp_retr1_max = 255;
  31 static int ip_local_port_range_min[] = { 1, 1 };
  32 static int ip_local_port_range_max[] = { 65535, 65535 };
  33 static int tcp_adv_win_scale_min = -31;
  34 static int tcp_adv_win_scale_max = 31;
  35 static int ip_ttl_min = 1;
  36 static int ip_ttl_max = 255;
  37 static int ip_ping_group_range_min[] = { 0, 0 };
  38 static int ip_ping_group_range_max[] = { GID_T_MAX, GID_T_MAX };
  39
  40 /* Update system visible IP port range */
  41 static void set_local_port_range(int range[2])
  42 {
  43         write_seqlock(&sysctl_local_ports.lock);
  44         sysctl_local_ports.range[0] = range[0];
  45         sysctl_local_ports.range[1] = range[1];
  46         write_sequnlock(&sysctl_local_ports.lock);
  47 }
  48
  49 /* Validate changes from /proc interface. */
  50 static int ipv4_local_port_range(ctl_table *table, int write,
  51                                  void __user *buffer,
  52                                  size_t *lenp, loff_t *ppos)
  53 {
  54         int ret;
  55         int range[2];
  56         ctl_table tmp = {
  57                 .data = &range,
  58                 .maxlen = sizeof(range),
  59                 .mode = table->mode,
  60                 .extra1 = &ip_local_port_range_min,
  61                 .extra2 = &ip_local_port_range_max,
  62         };
  63
  64         inet_get_local_port_range(range, range + 1);
  65         ret = proc_dointvec_minmax(&tmp, write, buffer, lenp, ppos);
  66
  67         if (write && ret == 0) {
  68                 if (range[1] < range[0])
  69                         ret = -EINVAL;
  70                 else
  71                         set_local_port_range(range);
  72         }
  73
  74         return ret;
  75 }
  76
  77
  78 static void inet_get_ping_group_range_table(struct ctl_table *table, gid_t *low, gid_t *high)
  79 {
  80         gid_t *data = table->data;
  81         unsigned seq;
  82         do {
  83                 seq = read_seqbegin(&sysctl_local_ports.lock);
  84
  85                 *low = data[0];
  86                 *high = data[1];
  87         } while (read_seqretry(&sysctl_local_ports.lock, seq));
  88 }
  89
  90 /* Update system visible IP port range */
  91 static void set_ping_group_range(struct ctl_table *table, gid_t range[2])
  92 {
  93         gid_t *data = table->data;
  94         write_seqlock(&sysctl_local_ports.lock);
  95         data[0] = range[0];
  96         data[1] = range[1];
  97         write_sequnlock(&sysctl_local_ports.lock);
  98 }
  99
 100 /* Validate changes from /proc interface. */
 101 static int ipv4_ping_group_range(ctl_table *table, int write,
 102                                  void __user *buffer,
 103                                  size_t *lenp, loff_t *ppos)
 104 {
 105         int ret;
 106         gid_t range[2];
 107         ctl_table tmp = {
 108                 .data = &range,
 109                 .maxlen = sizeof(range),
 110                 .mode = table->mode,
 111                 .extra1 = &ip_ping_group_range_min,
 112                 .extra2 = &ip_ping_group_range_max,
 113         };
 114
 115         inet_get_ping_group_range_table(table, range, range + 1);
 116         ret = proc_dointvec_minmax(&tmp, write, buffer, lenp, ppos);
 117
 118         if (write && ret == 0)
 119                 set_ping_group_range(table, range);
 120
 121         return ret;
 122 }
 123
 124 static int proc_tcp_congestion_control(ctl_table *ctl, int write,
 125                                        void __user *buffer, size_t *lenp, loff_t *ppos)
 126 {
 127         char val[TCP_CA_NAME_MAX];
 128         ctl_table tbl = {
 129                 .data = val,
 130                 .maxlen = TCP_CA_NAME_MAX,
 131         };
 132         int ret;
 133
 134         tcp_get_default_congestion_control(val);
 135
 136         ret = proc_dostring(&tbl, write, buffer, lenp, ppos);
 137         if (write && ret == 0)
 138                 ret = tcp_set_default_congestion_control(val);
 139         return ret;
 140 }
 141
 142 static int proc_tcp_available_congestion_control(ctl_table *ctl,
 143                                                  int write,
 144                                                  void __user *buffer, size_t *lenp,
 145                                                  loff_t *ppos)
 146 {
 147         ctl_table tbl = { .maxlen = TCP_CA_BUF_MAX, };
 148         int ret;
 149
 150         tbl.data = kmalloc(tbl.maxlen, GFP_USER);
 151         if (!tbl.data)
 152                 return -ENOMEM;
 153         tcp_get_available_congestion_control(tbl.data, TCP_CA_BUF_MAX);
 154         ret = proc_dostring(&tbl, write, buffer, lenp, ppos);
 155         kfree(tbl.data);
 156         return ret;
 157 }
 158
 159 static int proc_allowed_congestion_control(ctl_table *ctl,
 160                                            int write,
 161                                            void __user *buffer, size_t *lenp,
 162                                            loff_t *ppos)
 163 {
 164         ctl_table tbl = { .maxlen = TCP_CA_BUF_MAX };
 165         int ret;
 166
 167         tbl.data = kmalloc(tbl.maxlen, GFP_USER);
 168         if (!tbl.data)
 169                 return -ENOMEM;
 170
 171         tcp_get_allowed_congestion_control(tbl.data, tbl.maxlen);
 172         ret = proc_dostring(&tbl, write, buffer, lenp, ppos);
 173         if (write && ret == 0)
 174                 ret = tcp_set_allowed_congestion_control(tbl.data);
 175         kfree(tbl.data);
 176         return ret;
 177 }
 178
 179 static int ipv4_tcp_mem(ctl_table *ctl, int write,
 180                            void __user *buffer, size_t *lenp,
 181                            loff_t *ppos)
 182 {
 183         int ret;
 184         unsigned long vec[3];
 185         struct net *net = current->nsproxy->net_ns;
 186 #ifdef CONFIG_CGROUP_MEM_RES_CTLR_KMEM
 187         struct mem_cgroup *memcg;
 188 #endif
 189
 190         ctl_table tmp = {
 191                 .data = &vec,
 192                 .maxlen = sizeof(vec),
 193                 .mode = ctl->mode,
 194         };
 195
 196         if (!write) {
 197                 ctl->data = &net->ipv4.sysctl_tcp_mem;
 198                 return proc_doulongvec_minmax(ctl, write, buffer, lenp, ppos);
 199         }
 200
 201         ret = proc_doulongvec_minmax(&tmp, write, buffer, lenp, ppos);
 202         if (ret)
 203                 return ret;
 204
 205 #ifdef CONFIG_CGROUP_MEM_RES_CTLR_KMEM
 206         rcu_read_lock();
 207         memcg = mem_cgroup_from_task(current);
 208
 209         tcp_prot_mem(memcg, vec[0], 0);
 210         tcp_prot_mem(memcg, vec[1], 1);
 211         tcp_prot_mem(memcg, vec[2], 2);
 212         rcu_read_unlock();
 213 #endif
 214
 215         net->ipv4.sysctl_tcp_mem[0] = vec[0];
 216         net->ipv4.sysctl_tcp_mem[1] = vec[1];
 217         net->ipv4.sysctl_tcp_mem[2] = vec[2];
 218
 219         return 0;
 220 }
 221
 222 static struct ctl_table ipv4_table[] = {
 223         {
 224                 .procname       = "tcp_timestamps",
 225                 .data           = &sysctl_tcp_timestamps,
 226                 .maxlen         = sizeof(int),
 227                 .mode           = 0644,
 228                 .proc_handler   = proc_dointvec
 229         },
 230         {
 231                 .procname       = "tcp_window_scaling",
 232                 .data           = &sysctl_tcp_window_scaling,
 233                 .maxlen         = sizeof(int),
 234                 .mode           = 0644,
 235                 .proc_handler   = proc_dointvec
 236         },
 237         {
 238                 .procname       = "tcp_sack",
 239                 .data           = &sysctl_tcp_sack,
 240                 .maxlen         = sizeof(int),
 241                 .mode           = 0644,
 242                 .proc_handler   = proc_dointvec
 243         },
 244         {
 245                 .procname       = "tcp_retrans_collapse",
 246                 .data           = &sysctl_tcp_retrans_collapse,
 247                 .maxlen         = sizeof(int),
 248                 .mode           = 0644,
 249                 .proc_handler   = proc_dointvec
 250         },
 251         {
 252                 .procname       = "ip_default_ttl",
 253                 .data           = &sysctl_ip_default_ttl,
 254                 .maxlen         = sizeof(int),
 255                 .mode           = 0644,
 256                 .proc_handler   = proc_dointvec_minmax,
 257                 .extra1         = &ip_ttl_min,
 258                 .extra2         = &ip_ttl_max,
 259         },
 260         {
 261                 .procname       = "ip_no_pmtu_disc",
 262                 .data           = &ipv4_config.no_pmtu_disc,
 263                 .maxlen         = sizeof(int),
 264                 .mode           = 0644,
 265                 .proc_handler   = proc_dointvec
 266         },
 267         {
 268                 .procname       = "ip_nonlocal_bind",
 269                 .data           = &sysctl_ip_nonlocal_bind,
 270                 .maxlen         = sizeof(int),
 271                 .mode           = 0644,
 272                 .proc_handler   = proc_dointvec
 273         },
 274         {
 275                 .procname       = "tcp_syn_retries",
 276                 .data           = &sysctl_tcp_syn_retries,
 277                 .maxlen         = sizeof(int),
 278                 .mode           = 0644,
 279                 .proc_handler   = proc_dointvec
 280         },
 281         {
 282                 .procname       = "tcp_synack_retries",
 283                 .data           = &sysctl_tcp_synack_retries,
 284                 .maxlen         = sizeof(int),
 285                 .mode           = 0644,
 286                 .proc_handler   = proc_dointvec
 287         },
 288         {
 289                 .procname       = "tcp_max_orphans",
 290                 .data           = &sysctl_tcp_max_orphans,
 291                 .maxlen         = sizeof(int),
 292                 .mode           = 0644,
 293                 .proc_handler   = proc_dointvec
 294         },
 295         {
 296                 .procname       = "tcp_max_tw_buckets",
 297                 .data           = &tcp_death_row.sysctl_max_tw_buckets,
 298                 .maxlen         = sizeof(int),
 299                 .mode           = 0644,
 300                 .proc_handler   = proc_dointvec
 301         },
 302         {
 303                 .procname       = "ip_dynaddr",
 304                 .data           = &sysctl_ip_dynaddr,
 305                 .maxlen         = sizeof(int),
 306                 .mode           = 0644,
 307                 .proc_handler   = proc_dointvec
 308         },
 309         {
 310                 .procname       = "tcp_keepalive_time",
 311                 .data           = &sysctl_tcp_keepalive_time,
 312                 .maxlen         = sizeof(int),
 313                 .mode           = 0644,
 314                 .proc_handler   = proc_dointvec_jiffies,
 315         },
 316         {
 317                 .procname       = "tcp_keepalive_probes",
 318                 .data           = &sysctl_tcp_keepalive_probes,
 319                 .maxlen         = sizeof(int),
 320                 .mode           = 0644,
 321                 .proc_handler   = proc_dointvec
 322         },
 323         {
 324                 .procname       = "tcp_keepalive_intvl",
 325                 .data           = &sysctl_tcp_keepalive_intvl,
 326                 .maxlen         = sizeof(int),
 327                 .mode           = 0644,
 328                 .proc_handler   = proc_dointvec_jiffies,
 329         },
 330         {
 331                 .procname       = "tcp_retries1",
 332                 .data           = &sysctl_tcp_retries1,
 333                 .maxlen         = sizeof(int),
 334                 .mode           = 0644,
 335                 .proc_handler   = proc_dointvec_minmax,
 336                 .extra2         = &tcp_retr1_max
 337         },
 338         {
 339                 .procname       = "tcp_retries2",
 340                 .data           = &sysctl_tcp_retries2,
 341                 .maxlen         = sizeof(int),
 342                 .mode           = 0644,
 343                 .proc_handler   = proc_dointvec
 344         },
 345         {
 346                 .procname       = "tcp_fin_timeout",
 347                 .data           = &sysctl_tcp_fin_timeout,
 348                 .maxlen         = sizeof(int),
 349                 .mode           = 0644,
 350                 .proc_handler   = proc_dointvec_jiffies,
 351         },
 352 #ifdef CONFIG_SYN_COOKIES
 353         {
 354                 .procname       = "tcp_syncookies",
 355                 .data           = &sysctl_tcp_syncookies,
 356                 .maxlen         = sizeof(int),
 357                 .mode           = 0644,
 358                 .proc_handler   = proc_dointvec
 359         },
 360 #endif
 361         {
 362                 .procname       = "tcp_tw_recycle",
 363                 .data           = &tcp_death_row.sysctl_tw_recycle,
 364                 .maxlen         = sizeof(int),
 365                 .mode           = 0644,
 366                 .proc_handler   = proc_dointvec
 367         },
 368         {
 369                 .procname       = "tcp_abort_on_overflow",
 370                 .data           = &sysctl_tcp_abort_on_overflow,
 371                 .maxlen         = sizeof(int),
 372                 .mode           = 0644,
 373                 .proc_handler   = proc_dointvec
 374         },
 375         {
 376                 .procname       = "tcp_stdurg",
 377                 .data           = &sysctl_tcp_stdurg,
 378                 .maxlen         = sizeof(int),
 379                 .mode           = 0644,
 380                 .proc_handler   = proc_dointvec
 381         },
 382         {
 383                 .procname       = "tcp_rfc1337",
 384                 .data           = &sysctl_tcp_rfc1337,
 385                 .maxlen         = sizeof(int),
 386                 .mode           = 0644,
 387                 .proc_handler   = proc_dointvec
 388         },
 389         {
 390                 .procname       = "tcp_max_syn_backlog",
 391                 .data           = &sysctl_max_syn_backlog,
 392                 .maxlen         = sizeof(int),
 393                 .mode           = 0644,
 394                 .proc_handler   = proc_dointvec
 395         },
 396         {
 397                 .procname       = "ip_local_port_range",
 398                 .data           = &sysctl_local_ports.range,
 399                 .maxlen         = sizeof(sysctl_local_ports.range),
 400                 .mode           = 0644,
 401                 .proc_handler   = ipv4_local_port_range,
 402         },
 403         {
 404                 .procname       = "ip_local_reserved_ports",
 405                 .data           = NULL, /* initialized in sysctl_ipv4_init */
 406                 .maxlen         = 65536,
 407                 .mode           = 0644,
 408                 .proc_handler   = proc_do_large_bitmap,
 409         },
 410         {
 411                 .procname       = "igmp_max_memberships",
 412                 .data           = &sysctl_igmp_max_memberships,
 413                 .maxlen         = sizeof(int),
 414                 .mode           = 0644,
 415                 .proc_handler   = proc_dointvec
 416         },
 417         {
 418                 .procname       = "igmp_max_msf",
 419                 .data           = &sysctl_igmp_max_msf,
 420                 .maxlen         = sizeof(int),
 421                 .mode           = 0644,
 422                 .proc_handler   = proc_dointvec
 423         },
 424         {
 425                 .procname       = "inet_peer_threshold",
 426                 .data           = &inet_peer_threshold,
 427                 .maxlen         = sizeof(int),
 428                 .mode           = 0644,
 429                 .proc_handler   = proc_dointvec
 430         },
 431         {
 432                 .procname       = "inet_peer_minttl",
 433                 .data           = &inet_peer_minttl,
 434                 .maxlen         = sizeof(int),
 435                 .mode           = 0644,
 436                 .proc_handler   = proc_dointvec_jiffies,
 437         },
 438         {
 439                 .procname       = "inet_peer_maxttl",
 440                 .data           = &inet_peer_maxttl,
 441                 .maxlen         = sizeof(int),
 442                 .mode           = 0644,
 443                 .proc_handler   = proc_dointvec_jiffies,
 444         },
 445         {
 446                 .procname       = "tcp_orphan_retries",
 447                 .data           = &sysctl_tcp_orphan_retries,
 448                 .maxlen         = sizeof(int),
 449                 .mode           = 0644,
 450                 .proc_handler   = proc_dointvec
 451         },
 452         {
 453                 .procname       = "tcp_fack",
 454                 .data           = &sysctl_tcp_fack,
 455                 .maxlen         = sizeof(int),
 456                 .mode           = 0644,
 457                 .proc_handler   = proc_dointvec
 458         },
 459         {
 460                 .procname       = "tcp_reordering",
 461                 .data           = &sysctl_tcp_reordering,
 462                 .maxlen         = sizeof(int),
 463                 .mode           = 0644,
 464                 .proc_handler   = proc_dointvec
 465         },
 466         {
 467                 .procname       = "tcp_ecn",
 468                 .data           = &sysctl_tcp_ecn,
 469                 .maxlen         = sizeof(int),
 470                 .mode           = 0644,
 471                 .proc_handler   = proc_dointvec
 472         },
 473         {
 474                 .procname       = "tcp_dsack",
 475                 .data           = &sysctl_tcp_dsack,
 476                 .maxlen         = sizeof(int),
 477                 .mode           = 0644,
 478                 .proc_handler   = proc_dointvec
 479         },
 480         {
 481                 .procname       = "tcp_wmem",
 482                 .data           = &sysctl_tcp_wmem,
 483                 .maxlen         = sizeof(sysctl_tcp_wmem),
 484                 .mode           = 0644,
 485                 .proc_handler   = proc_dointvec
 486         },
 487         {
 488                 .procname       = "tcp_rmem",
 489                 .data           = &sysctl_tcp_rmem,
 490                 .maxlen         = sizeof(sysctl_tcp_rmem),
 491                 .mode           = 0644,
 492                 .proc_handler   = proc_dointvec
 493         },
 494         {
 495                 .procname       = "tcp_app_win",
 496                 .data           = &sysctl_tcp_app_win,
 497                 .maxlen         = sizeof(int),
 498                 .mode           = 0644,
 499                 .proc_handler   = proc_dointvec
 500         },
 501         {
 502                 .procname       = "tcp_adv_win_scale",
 503                 .data           = &sysctl_tcp_adv_win_scale,
 504                 .maxlen         = sizeof(int),
 505                 .mode           = 0644,
 506                 .proc_handler   = proc_dointvec_minmax,
 507                 .extra1         = &tcp_adv_win_scale_min,
 508                 .extra2         = &tcp_adv_win_scale_max,
 509         },
 510         {
 511                 .procname       = "tcp_tw_reuse",
 512                 .data           = &sysctl_tcp_tw_reuse,
 513                 .maxlen         = sizeof(int),
 514                 .mode           = 0644,
 515                 .proc_handler   = proc_dointvec
 516         },
 517         {
 518                 .procname       = "tcp_frto",
 519                 .data           = &sysctl_tcp_frto,
 520                 .maxlen         = sizeof(int),
 521                 .mode           = 0644,
 522                 .proc_handler   = proc_dointvec
 523         },
 524         {
 525                 .procname       = "tcp_frto_response",
 526                 .data           = &sysctl_tcp_frto_response,
 527                 .maxlen         = sizeof(int),
 528                 .mode           = 0644,
 529                 .proc_handler   = proc_dointvec
 530         },
 531         {
 532                 .procname       = "tcp_low_latency",
 533                 .data           = &sysctl_tcp_low_latency,
 534                 .maxlen         = sizeof(int),
 535                 .mode           = 0644,
 536                 .proc_handler   = proc_dointvec
 537         },
 538         {
 539                 .procname       = "tcp_no_metrics_save",
 540                 .data           = &sysctl_tcp_nometrics_save,
 541                 .maxlen         = sizeof(int),
 542                 .mode           = 0644,
 543                 .proc_handler   = proc_dointvec,
 544         },
 545         {
 546                 .procname       = "tcp_moderate_rcvbuf",
 547                 .data           = &sysctl_tcp_moderate_rcvbuf,
 548                 .maxlen         = sizeof(int),
 549                 .mode           = 0644,
 550                 .proc_handler   = proc_dointvec,
 551         },
 552         {
 553                 .procname       = "tcp_tso_win_divisor",
 554                 .data           = &sysctl_tcp_tso_win_divisor,
 555                 .maxlen         = sizeof(int),
 556                 .mode           = 0644,
 557                 .proc_handler   = proc_dointvec,
 558         },
 559         {
 560                 .procname       = "tcp_congestion_control",
 561                 .mode           = 0644,
 562                 .maxlen         = TCP_CA_NAME_MAX,
 563                 .proc_handler   = proc_tcp_congestion_control,
 564         },
 565         {
 566                 .procname       = "tcp_abc",
 567                 .data           = &sysctl_tcp_abc,
 568                 .maxlen         = sizeof(int),
 569                 .mode           = 0644,
 570                 .proc_handler   = proc_dointvec,
 571         },
 572         {
 573                 .procname       = "tcp_mtu_probing",
 574                 .data           = &sysctl_tcp_mtu_probing,
 575                 .maxlen         = sizeof(int),
 576                 .mode           = 0644,
 577                 .proc_handler   = proc_dointvec,
 578         },
 579         {
 580                 .procname       = "tcp_base_mss",
 581                 .data           = &sysctl_tcp_base_mss,
 582                 .maxlen         = sizeof(int),
 583                 .mode           = 0644,
 584                 .proc_handler   = proc_dointvec,
 585         },
 586         {
 587                 .procname       = "tcp_workaround_signed_windows",
 588                 .data           = &sysctl_tcp_workaround_signed_windows,
 589                 .maxlen         = sizeof(int),
 590                 .mode           = 0644,
 591                 .proc_handler   = proc_dointvec
 592         },
 593 #ifdef CONFIG_NET_DMA
 594         {
 595                 .procname       = "tcp_dma_copybreak",
 596                 .data           = &sysctl_tcp_dma_copybreak,
 597                 .maxlen         = sizeof(int),
 598                 .mode           = 0644,
 599                 .proc_handler   = proc_dointvec
 600         },
 601 #endif
 602         {
 603                 .procname       = "tcp_slow_start_after_idle",
 604                 .data           = &sysctl_tcp_slow_start_after_idle,
 605                 .maxlen         = sizeof(int),
 606                 .mode           = 0644,
 607                 .proc_handler   = proc_dointvec
 608         },
 609 #ifdef CONFIG_NETLABEL
 610         {
 611                 .procname       = "cipso_cache_enable",
 612                 .data           = &cipso_v4_cache_enabled,
 613                 .maxlen         = sizeof(int),
 614                 .mode           = 0644,
 615                 .proc_handler   = proc_dointvec,
 616         },
 617         {
 618                 .procname       = "cipso_cache_bucket_size",
 619                 .data           = &cipso_v4_cache_bucketsize,
 620                 .maxlen         = sizeof(int),
 621                 .mode           = 0644,
 622                 .proc_handler   = proc_dointvec,
 623         },
 624         {
 625                 .procname       = "cipso_rbm_optfmt",
 626                 .data           = &cipso_v4_rbm_optfmt,
 627                 .maxlen         = sizeof(int),
 628                 .mode           = 0644,
 629                 .proc_handler   = proc_dointvec,
 630         },
 631         {
 632                 .procname       = "cipso_rbm_strictvalid",
 633                 .data           = &cipso_v4_rbm_strictvalid,
 634                 .maxlen         = sizeof(int),
 635                 .mode           = 0644,
 636                 .proc_handler   = proc_dointvec,
 637         },
 638 #endif /* CONFIG_NETLABEL */
 639         {
 640                 .procname       = "tcp_available_congestion_control",
 641                 .maxlen         = TCP_CA_BUF_MAX,
 642                 .mode           = 0444,
 643                 .proc_handler   = proc_tcp_available_congestion_control,
 644         },
 645         {
 646                 .procname       = "tcp_allowed_congestion_control",
 647                 .maxlen         = TCP_CA_BUF_MAX,
 648                 .mode           = 0644,
 649                 .proc_handler   = proc_allowed_congestion_control,
 650         },
 651         {
 652                 .procname       = "tcp_max_ssthresh",
 653                 .data           = &sysctl_tcp_max_ssthresh,
 654                 .maxlen         = sizeof(int),
 655                 .mode           = 0644,
 656                 .proc_handler   = proc_dointvec,
 657         },
 658         {
 659                 .procname       = "tcp_cookie_size",
 660                 .data           = &sysctl_tcp_cookie_size,
 661                 .maxlen         = sizeof(int),
 662                 .mode           = 0644,
 663                 .proc_handler   = proc_dointvec
 664         },
 665         {
 666                 .procname       = "tcp_thin_linear_timeouts",
 667                 .data           = &sysctl_tcp_thin_linear_timeouts,
 668                 .maxlen         = sizeof(int),
 669                 .mode           = 0644,
 670                 .proc_handler   = proc_dointvec
 671         },
 672         {
 673                 .procname       = "tcp_thin_dupack",
 674                 .data           = &sysctl_tcp_thin_dupack,
 675                 .maxlen         = sizeof(int),
 676                 .mode           = 0644,
 677                 .proc_handler   = proc_dointvec
 678         },
 679         {
 680                 .procname       = "udp_mem",
 681                 .data           = &sysctl_udp_mem,
 682                 .maxlen         = sizeof(sysctl_udp_mem),
 683                 .mode           = 0644,
 684                 .proc_handler   = proc_doulongvec_minmax,
 685         },
 686         {
 687                 .procname       = "udp_rmem_min",
 688                 .data           = &sysctl_udp_rmem_min,
 689                 .maxlen         = sizeof(sysctl_udp_rmem_min),
 690                 .mode           = 0644,
 691                 .proc_handler   = proc_dointvec_minmax,
 692                 .extra1         = &zero
 693         },
 694         {
 695                 .procname       = "udp_wmem_min",
 696                 .data           = &sysctl_udp_wmem_min,
 697                 .maxlen         = sizeof(sysctl_udp_wmem_min),
 698                 .mode           = 0644,
 699                 .proc_handler   = proc_dointvec_minmax,
 700                 .extra1         = &zero
 701         },
 702         { }
 703 };
 704
 705 static struct ctl_table ipv4_net_table[] = {
 706         {
 707                 .procname       = "icmp_echo_ignore_all",
 708                 .data           = &init_net.ipv4.sysctl_icmp_echo_ignore_all,
 709                 .maxlen         = sizeof(int),
 710                 .mode           = 0644,
 711                 .proc_handler   = proc_dointvec
 712         },
 713         {
 714                 .procname       = "icmp_echo_ignore_broadcasts",
 715                 .data           = &init_net.ipv4.sysctl_icmp_echo_ignore_broadcasts,
 716                 .maxlen         = sizeof(int),
 717                 .mode           = 0644,
 718                 .proc_handler   = proc_dointvec
 719         },
 720         {
 721                 .procname       = "icmp_ignore_bogus_error_responses",
 722                 .data           = &init_net.ipv4.sysctl_icmp_ignore_bogus_error_responses,
 723                 .maxlen         = sizeof(int),
 724                 .mode           = 0644,
 725                 .proc_handler   = proc_dointvec
 726         },
 727         {
 728                 .procname       = "icmp_errors_use_inbound_ifaddr",
 729                 .data           = &init_net.ipv4.sysctl_icmp_errors_use_inbound_ifaddr,
 730                 .maxlen         = sizeof(int),
 731                 .mode           = 0644,
 732                 .proc_handler   = proc_dointvec
 733         },
 734         {
 735                 .procname       = "icmp_ratelimit",
 736                 .data           = &init_net.ipv4.sysctl_icmp_ratelimit,
 737                 .maxlen         = sizeof(int),
 738                 .mode           = 0644,
 739                 .proc_handler   = proc_dointvec_ms_jiffies,
 740         },
 741         {
 742                 .procname       = "icmp_ratemask",
 743                 .data           = &init_net.ipv4.sysctl_icmp_ratemask,
 744                 .maxlen         = sizeof(int),
 745                 .mode           = 0644,
 746                 .proc_handler   = proc_dointvec
 747         },
 748         {
 749                 .procname       = "rt_cache_rebuild_count",
 750                 .data           = &init_net.ipv4.sysctl_rt_cache_rebuild_count,
 751                 .maxlen         = sizeof(int),
 752                 .mode           = 0644,
 753                 .proc_handler   = proc_dointvec
 754         },
 755         {
 756                 .procname       = "ping_group_range",
 757                 .data           = &init_net.ipv4.sysctl_ping_group_range,
 758                 .maxlen         = sizeof(init_net.ipv4.sysctl_ping_group_range),
 759                 .mode           = 0644,
 760                 .proc_handler   = ipv4_ping_group_range,
 761         },
 762         {
 763                 .procname       = "tcp_mem",
 764                 .maxlen         = sizeof(init_net.ipv4.sysctl_tcp_mem),
 765                 .mode           = 0644,
 766                 .proc_handler   = ipv4_tcp_mem,
 767         },
 768         { }
 769 };
 770
 771 struct ctl_path net_ipv4_ctl_path[] = {
 772         { .procname = "net", },
 773         { .procname = "ipv4", },
 774         { },
 775 };
 776 EXPORT_SYMBOL_GPL(net_ipv4_ctl_path);
 777
 778 static __net_init int ipv4_sysctl_init_net(struct net *net)
 779 {
 780         struct ctl_table *table;
 781         unsigned long limit;
 782
 783         table = ipv4_net_table;
 784         if (!net_eq(net, &init_net)) {
 785                 table = kmemdup(table, sizeof(ipv4_net_table), GFP_KERNEL);
 786                 if (table == NULL)
 787                         goto err_alloc;
 788
 789                 table[0].data =
 790                         &net->ipv4.sysctl_icmp_echo_ignore_all;
 791                 table[1].data =
 792                         &net->ipv4.sysctl_icmp_echo_ignore_broadcasts;
 793                 table[2].data =
 794                         &net->ipv4.sysctl_icmp_ignore_bogus_error_responses;
 795                 table[3].data =
 796                         &net->ipv4.sysctl_icmp_errors_use_inbound_ifaddr;
 797                 table[4].data =
 798                         &net->ipv4.sysctl_icmp_ratelimit;
 799                 table[5].data =
 800                         &net->ipv4.sysctl_icmp_ratemask;
 801                 table[6].data =
 802                         &net->ipv4.sysctl_rt_cache_rebuild_count;
 803                 table[7].data =
 804                         &net->ipv4.sysctl_ping_group_range;
 805
 806         }
 807
 808         /*
 809          * Sane defaults - nobody may create ping sockets.
 810          * Boot scripts should set this to distro-specific group.
 811          */
 812         net->ipv4.sysctl_ping_group_range[0] = 1;
 813         net->ipv4.sysctl_ping_group_range[1] = 0;
 814
 815         net->ipv4.sysctl_rt_cache_rebuild_count = 4;
 816
 817         limit = nr_free_buffer_pages() / 8;
 818         limit = max(limit, 128UL);
 819         net->ipv4.sysctl_tcp_mem[0] = limit / 4 * 3;
 820         net->ipv4.sysctl_tcp_mem[1] = limit;
 821         net->ipv4.sysctl_tcp_mem[2] = net->ipv4.sysctl_tcp_mem[0] * 2;
 822
 823         net->ipv4.ipv4_hdr = register_net_sysctl_table(net,
 824                         net_ipv4_ctl_path, table);
 825         if (net->ipv4.ipv4_hdr == NULL)
 826                 goto err_reg;
 827
 828         return 0;
 829
 830 err_reg:
 831         if (!net_eq(net, &init_net))
 832                 kfree(table);
 833 err_alloc:
 834         return -ENOMEM;
 835 }
 836
 837 static __net_exit void ipv4_sysctl_exit_net(struct net *net)
 838 {
 839         struct ctl_table *table;
 840
 841         table = net->ipv4.ipv4_hdr->ctl_table_arg;
 842         unregister_net_sysctl_table(net->ipv4.ipv4_hdr);
 843         kfree(table);
 844 }
 845
 846 static __net_initdata struct pernet_operations ipv4_sysctl_ops = {
 847         .init = ipv4_sysctl_init_net,
 848         .exit = ipv4_sysctl_exit_net,
 849 };
 850
 851 static __init int sysctl_ipv4_init(void)
 852 {
 853         struct ctl_table_header *hdr;
 854         struct ctl_table *i;
 855
 856         for (i = ipv4_table; i->procname; i++) {
 857                 if (strcmp(i->procname, "ip_local_reserved_ports") == 0) {
 858                         i->data = sysctl_local_reserved_ports;
 859                         break;
 860                 }
 861         }
 862         if (!i->procname)
 863                 return -EINVAL;
 864
 865         hdr = register_sysctl_paths(net_ipv4_ctl_path, ipv4_table);
 866         if (hdr == NULL)
 867                 return -ENOMEM;
 868
 869         if (register_pernet_subsys(&ipv4_sysctl_ops)) {
 870                 unregister_sysctl_table(hdr);
 871                 return -ENOMEM;
 872         }
 873
 874         return 0;
 875 }
 876
 877 __initcall(sysctl_ipv4_init);