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 tcp_retr1_max = 255;
  33 static int ip_local_port_range_min[] = { 1, 1 };
  34 static int ip_local_port_range_max[] = { 65535, 65535 };
  35 static int tcp_adv_win_scale_min = -31;
  36 static int tcp_adv_win_scale_max = 31;
  37 static int ip_ttl_min = 1;
  38 static int ip_ttl_max = 255;
  39 static int ip_ping_group_range_min[] = { 0, 0 };
  40 static int ip_ping_group_range_max[] = { GID_T_MAX, GID_T_MAX };
  41
  42 /* Update system visible IP port range */
  43 static void set_local_port_range(int range[2])
  44 {
  45         write_seqlock(&sysctl_local_ports.lock);
  46         sysctl_local_ports.range[0] = range[0];
  47         sysctl_local_ports.range[1] = range[1];
  48         write_sequnlock(&sysctl_local_ports.lock);
  49 }
  50
  51 /* Validate changes from /proc interface. */
  52 static int ipv4_local_port_range(struct ctl_table *table, int write,
  53                                  void __user *buffer,
  54                                  size_t *lenp, loff_t *ppos)
  55 {
  56         int ret;
  57         int range[2];
  58         struct ctl_table tmp = {
  59                 .data = &range,
  60                 .maxlen = sizeof(range),
  61                 .mode = table->mode,
  62                 .extra1 = &ip_local_port_range_min,
  63                 .extra2 = &ip_local_port_range_max,
  64         };
  65
  66         inet_get_local_port_range(range, range + 1);
  67         ret = proc_dointvec_minmax(&tmp, write, buffer, lenp, ppos);
  68
  69         if (write && ret == 0) {
  70                 if (range[1] < range[0])
  71                         ret = -EINVAL;
  72                 else
  73                         set_local_port_range(range);
  74         }
  75
  76         return ret;
  77 }
  78
  79
  80 static void inet_get_ping_group_range_table(struct ctl_table *table, kgid_t *low, kgid_t *high)
  81 {
  82         kgid_t *data = table->data;
  83         unsigned int seq;
  84         do {
  85                 seq = read_seqbegin(&sysctl_local_ports.lock);
  86
  87                 *low = data[0];
  88                 *high = data[1];
  89         } while (read_seqretry(&sysctl_local_ports.lock, seq));
  90 }
  91
  92 /* Update system visible IP port range */
  93 static void set_ping_group_range(struct ctl_table *table, kgid_t low, kgid_t high)
  94 {
  95         kgid_t *data = table->data;
  96         write_seqlock(&sysctl_local_ports.lock);
  97         data[0] = low;
  98         data[1] = high;
  99         write_sequnlock(&sysctl_local_ports.lock);
 100 }
 101
 102 /* Validate changes from /proc interface. */
 103 static int ipv4_ping_group_range(struct ctl_table *table, int write,
 104                                  void __user *buffer,
 105                                  size_t *lenp, loff_t *ppos)
 106 {
 107         struct user_namespace *user_ns = current_user_ns();
 108         int ret;
 109         gid_t urange[2];
 110         kgid_t low, high;
 111         struct ctl_table tmp = {
 112                 .data = &urange,
 113                 .maxlen = sizeof(urange),
 114                 .mode = table->mode,
 115                 .extra1 = &ip_ping_group_range_min,
 116                 .extra2 = &ip_ping_group_range_max,
 117         };
 118
 119         inet_get_ping_group_range_table(table, &low, &high);
 120         urange[0] = from_kgid_munged(user_ns, low);
 121         urange[1] = from_kgid_munged(user_ns, high);
 122         ret = proc_dointvec_minmax(&tmp, write, buffer, lenp, ppos);
 123
 124         if (write && ret == 0) {
 125                 low = make_kgid(user_ns, urange[0]);
 126                 high = make_kgid(user_ns, urange[1]);
 127                 if (!gid_valid(low) || !gid_valid(high) ||
 128                     (urange[1] < urange[0]) || gid_lt(high, low)) {
 129                         low = make_kgid(&init_user_ns, 1);
 130                         high = make_kgid(&init_user_ns, 0);
 131                 }
 132                 set_ping_group_range(table, low, high);
 133         }
 134
 135         return ret;
 136 }
 137
 138 static int proc_tcp_congestion_control(struct ctl_table *ctl, int write,
 139                                        void __user *buffer, size_t *lenp, loff_t *ppos)
 140 {
 141         char val[TCP_CA_NAME_MAX];
 142         struct ctl_table tbl = {
 143                 .data = val,
 144                 .maxlen = TCP_CA_NAME_MAX,
 145         };
 146         int ret;
 147
 148         tcp_get_default_congestion_control(val);
 149
 150         ret = proc_dostring(&tbl, write, buffer, lenp, ppos);
 151         if (write && ret == 0)
 152                 ret = tcp_set_default_congestion_control(val);
 153         return ret;
 154 }
 155
 156 static int proc_tcp_available_congestion_control(struct ctl_table *ctl,
 157                                                  int write,
 158                                                  void __user *buffer, size_t *lenp,
 159                                                  loff_t *ppos)
 160 {
 161         struct ctl_table tbl = { .maxlen = TCP_CA_BUF_MAX, };
 162         int ret;
 163
 164         tbl.data = kmalloc(tbl.maxlen, GFP_USER);
 165         if (!tbl.data)
 166                 return -ENOMEM;
 167         tcp_get_available_congestion_control(tbl.data, TCP_CA_BUF_MAX);
 168         ret = proc_dostring(&tbl, write, buffer, lenp, ppos);
 169         kfree(tbl.data);
 170         return ret;
 171 }
 172
 173 static int proc_allowed_congestion_control(struct ctl_table *ctl,
 174                                            int write,
 175                                            void __user *buffer, size_t *lenp,
 176                                            loff_t *ppos)
 177 {
 178         struct ctl_table tbl = { .maxlen = TCP_CA_BUF_MAX };
 179         int ret;
 180
 181         tbl.data = kmalloc(tbl.maxlen, GFP_USER);
 182         if (!tbl.data)
 183                 return -ENOMEM;
 184
 185         tcp_get_allowed_congestion_control(tbl.data, tbl.maxlen);
 186         ret = proc_dostring(&tbl, write, buffer, lenp, ppos);
 187         if (write && ret == 0)
 188                 ret = tcp_set_allowed_congestion_control(tbl.data);
 189         kfree(tbl.data);
 190         return ret;
 191 }
 192
 193 static int ipv4_tcp_mem(struct ctl_table *ctl, int write,
 194                            void __user *buffer, size_t *lenp,
 195                            loff_t *ppos)
 196 {
 197         int ret;
 198         unsigned long vec[3];
 199         struct net *net = current->nsproxy->net_ns;
 200 #ifdef CONFIG_MEMCG_KMEM
 201         struct mem_cgroup *memcg;
 202 #endif
 203
 204         struct ctl_table tmp = {
 205                 .data = &vec,
 206                 .maxlen = sizeof(vec),
 207                 .mode = ctl->mode,
 208         };
 209
 210         if (!write) {
 211                 ctl->data = &net->ipv4.sysctl_tcp_mem;
 212                 return proc_doulongvec_minmax(ctl, write, buffer, lenp, ppos);
 213         }
 214
 215         ret = proc_doulongvec_minmax(&tmp, write, buffer, lenp, ppos);
 216         if (ret)
 217                 return ret;
 218
 219 #ifdef CONFIG_MEMCG_KMEM
 220         rcu_read_lock();
 221         memcg = mem_cgroup_from_task(current);
 222
 223         tcp_prot_mem(memcg, vec[0], 0);
 224         tcp_prot_mem(memcg, vec[1], 1);
 225         tcp_prot_mem(memcg, vec[2], 2);
 226         rcu_read_unlock();
 227 #endif
 228
 229         net->ipv4.sysctl_tcp_mem[0] = vec[0];
 230         net->ipv4.sysctl_tcp_mem[1] = vec[1];
 231         net->ipv4.sysctl_tcp_mem[2] = vec[2];
 232
 233         return 0;
 234 }
 235
 236 static int proc_tcp_fastopen_key(struct ctl_table *ctl, int write,
 237                                  void __user *buffer, size_t *lenp,
 238                                  loff_t *ppos)
 239 {
 240         struct ctl_table tbl = { .maxlen = (TCP_FASTOPEN_KEY_LENGTH * 2 + 10) };
 241         struct tcp_fastopen_context *ctxt;
 242         int ret;
 243         u32  user_key[4]; /* 16 bytes, matching TCP_FASTOPEN_KEY_LENGTH */
 244
 245         tbl.data = kmalloc(tbl.maxlen, GFP_KERNEL);
 246         if (!tbl.data)
 247                 return -ENOMEM;
 248
 249         rcu_read_lock();
 250         ctxt = rcu_dereference(tcp_fastopen_ctx);
 251         if (ctxt)
 252                 memcpy(user_key, ctxt->key, TCP_FASTOPEN_KEY_LENGTH);
 253         else
 254                 memset(user_key, 0, sizeof(user_key));
 255         rcu_read_unlock();
 256
 257         snprintf(tbl.data, tbl.maxlen, "%08x-%08x-%08x-%08x",
 258                 user_key[0], user_key[1], user_key[2], user_key[3]);
 259         ret = proc_dostring(&tbl, write, buffer, lenp, ppos);
 260
 261         if (write && ret == 0) {
 262                 if (sscanf(tbl.data, "%x-%x-%x-%x", user_key, user_key + 1,
 263                            user_key + 2, user_key + 3) != 4) {
 264                         ret = -EINVAL;
 265                         goto bad_key;
 266                 }
 267                 tcp_fastopen_reset_cipher(user_key, TCP_FASTOPEN_KEY_LENGTH);
 268         }
 269
 270 bad_key:
 271         pr_debug("proc FO key set 0x%x-%x-%x-%x <- 0x%s: %u\n",
 272                user_key[0], user_key[1], user_key[2], user_key[3],
 273                (char *)tbl.data, ret);
 274         kfree(tbl.data);
 275         return ret;
 276 }
 277
 278 static struct ctl_table ipv4_table[] = {
 279         {
 280                 .procname       = "tcp_timestamps",
 281                 .data           = &sysctl_tcp_timestamps,
 282                 .maxlen         = sizeof(int),
 283                 .mode           = 0644,
 284                 .proc_handler   = proc_dointvec
 285         },
 286         {
 287                 .procname       = "tcp_window_scaling",
 288                 .data           = &sysctl_tcp_window_scaling,
 289                 .maxlen         = sizeof(int),
 290                 .mode           = 0644,
 291                 .proc_handler   = proc_dointvec
 292         },
 293         {
 294                 .procname       = "tcp_sack",
 295                 .data           = &sysctl_tcp_sack,
 296                 .maxlen         = sizeof(int),
 297                 .mode           = 0644,
 298                 .proc_handler   = proc_dointvec
 299         },
 300         {
 301                 .procname       = "tcp_retrans_collapse",
 302                 .data           = &sysctl_tcp_retrans_collapse,
 303                 .maxlen         = sizeof(int),
 304                 .mode           = 0644,
 305                 .proc_handler   = proc_dointvec
 306         },
 307         {
 308                 .procname       = "ip_default_ttl",
 309                 .data           = &sysctl_ip_default_ttl,
 310                 .maxlen         = sizeof(int),
 311                 .mode           = 0644,
 312                 .proc_handler   = proc_dointvec_minmax,
 313                 .extra1         = &ip_ttl_min,
 314                 .extra2         = &ip_ttl_max,
 315         },
 316         {
 317                 .procname       = "ip_no_pmtu_disc",
 318                 .data           = &ipv4_config.no_pmtu_disc,
 319                 .maxlen         = sizeof(int),
 320                 .mode           = 0644,
 321                 .proc_handler   = proc_dointvec
 322         },
 323         {
 324                 .procname       = "ip_nonlocal_bind",
 325                 .data           = &sysctl_ip_nonlocal_bind,
 326                 .maxlen         = sizeof(int),
 327                 .mode           = 0644,
 328                 .proc_handler   = proc_dointvec
 329         },
 330         {
 331                 .procname       = "tcp_syn_retries",
 332                 .data           = &sysctl_tcp_syn_retries,
 333                 .maxlen         = sizeof(int),
 334                 .mode           = 0644,
 335                 .proc_handler   = proc_dointvec
 336         },
 337         {
 338                 .procname       = "tcp_synack_retries",
 339                 .data           = &sysctl_tcp_synack_retries,
 340                 .maxlen         = sizeof(int),
 341                 .mode           = 0644,
 342                 .proc_handler   = proc_dointvec
 343         },
 344         {
 345                 .procname       = "tcp_max_orphans",
 346                 .data           = &sysctl_tcp_max_orphans,
 347                 .maxlen         = sizeof(int),
 348                 .mode           = 0644,
 349                 .proc_handler   = proc_dointvec
 350         },
 351         {
 352                 .procname       = "tcp_max_tw_buckets",
 353                 .data           = &tcp_death_row.sysctl_max_tw_buckets,
 354                 .maxlen         = sizeof(int),
 355                 .mode           = 0644,
 356                 .proc_handler   = proc_dointvec
 357         },
 358         {
 359                 .procname       = "ip_early_demux",
 360                 .data           = &sysctl_ip_early_demux,
 361                 .maxlen         = sizeof(int),
 362                 .mode           = 0644,
 363                 .proc_handler   = proc_dointvec
 364         },
 365         {
 366                 .procname       = "ip_dynaddr",
 367                 .data           = &sysctl_ip_dynaddr,
 368                 .maxlen         = sizeof(int),
 369                 .mode           = 0644,
 370                 .proc_handler   = proc_dointvec
 371         },
 372         {
 373                 .procname       = "tcp_keepalive_time",
 374                 .data           = &sysctl_tcp_keepalive_time,
 375                 .maxlen         = sizeof(int),
 376                 .mode           = 0644,
 377                 .proc_handler   = proc_dointvec_jiffies,
 378         },
 379         {
 380                 .procname       = "tcp_keepalive_probes",
 381                 .data           = &sysctl_tcp_keepalive_probes,
 382                 .maxlen         = sizeof(int),
 383                 .mode           = 0644,
 384                 .proc_handler   = proc_dointvec
 385         },
 386         {
 387                 .procname       = "tcp_keepalive_intvl",
 388                 .data           = &sysctl_tcp_keepalive_intvl,
 389                 .maxlen         = sizeof(int),
 390                 .mode           = 0644,
 391                 .proc_handler   = proc_dointvec_jiffies,
 392         },
 393         {
 394                 .procname       = "tcp_retries1",
 395                 .data           = &sysctl_tcp_retries1,
 396                 .maxlen         = sizeof(int),
 397                 .mode           = 0644,
 398                 .proc_handler   = proc_dointvec_minmax,
 399                 .extra2         = &tcp_retr1_max
 400         },
 401         {
 402                 .procname       = "tcp_retries2",
 403                 .data           = &sysctl_tcp_retries2,
 404                 .maxlen         = sizeof(int),
 405                 .mode           = 0644,
 406                 .proc_handler   = proc_dointvec
 407         },
 408         {
 409                 .procname       = "tcp_fin_timeout",
 410                 .data           = &sysctl_tcp_fin_timeout,
 411                 .maxlen         = sizeof(int),
 412                 .mode           = 0644,
 413                 .proc_handler   = proc_dointvec_jiffies,
 414         },
 415 #ifdef CONFIG_SYN_COOKIES
 416         {
 417                 .procname       = "tcp_syncookies",
 418                 .data           = &sysctl_tcp_syncookies,
 419                 .maxlen         = sizeof(int),
 420                 .mode           = 0644,
 421                 .proc_handler   = proc_dointvec
 422         },
 423 #endif
 424         {
 425                 .procname       = "tcp_fastopen",
 426                 .data           = &sysctl_tcp_fastopen,
 427                 .maxlen         = sizeof(int),
 428                 .mode           = 0644,
 429                 .proc_handler   = proc_dointvec,
 430         },
 431         {
 432                 .procname       = "tcp_fastopen_key",
 433                 .mode           = 0600,
 434                 .maxlen         = ((TCP_FASTOPEN_KEY_LENGTH * 2) + 10),
 435                 .proc_handler   = proc_tcp_fastopen_key,
 436         },
 437         {
 438                 .procname       = "tcp_tw_recycle",
 439                 .data           = &tcp_death_row.sysctl_tw_recycle,
 440                 .maxlen         = sizeof(int),
 441                 .mode           = 0644,
 442                 .proc_handler   = proc_dointvec
 443         },
 444         {
 445                 .procname       = "tcp_abort_on_overflow",
 446                 .data           = &sysctl_tcp_abort_on_overflow,
 447                 .maxlen         = sizeof(int),
 448                 .mode           = 0644,
 449                 .proc_handler   = proc_dointvec
 450         },
 451         {
 452                 .procname       = "tcp_stdurg",
 453                 .data           = &sysctl_tcp_stdurg,
 454                 .maxlen         = sizeof(int),
 455                 .mode           = 0644,
 456                 .proc_handler   = proc_dointvec
 457         },
 458         {
 459                 .procname       = "tcp_rfc1337",
 460                 .data           = &sysctl_tcp_rfc1337,
 461                 .maxlen         = sizeof(int),
 462                 .mode           = 0644,
 463                 .proc_handler   = proc_dointvec
 464         },
 465         {
 466                 .procname       = "tcp_max_syn_backlog",
 467                 .data           = &sysctl_max_syn_backlog,
 468                 .maxlen         = sizeof(int),
 469                 .mode           = 0644,
 470                 .proc_handler   = proc_dointvec
 471         },
 472         {
 473                 .procname       = "ip_local_port_range",
 474                 .data           = &sysctl_local_ports.range,
 475                 .maxlen         = sizeof(sysctl_local_ports.range),
 476                 .mode           = 0644,
 477                 .proc_handler   = ipv4_local_port_range,
 478         },
 479         {
 480                 .procname       = "ip_local_reserved_ports",
 481                 .data           = NULL, /* initialized in sysctl_ipv4_init */
 482                 .maxlen         = 65536,
 483                 .mode           = 0644,
 484                 .proc_handler   = proc_do_large_bitmap,
 485         },
 486         {
 487                 .procname       = "igmp_max_memberships",
 488                 .data           = &sysctl_igmp_max_memberships,
 489                 .maxlen         = sizeof(int),
 490                 .mode           = 0644,
 491                 .proc_handler   = proc_dointvec
 492         },
 493         {
 494                 .procname       = "igmp_max_msf",
 495                 .data           = &sysctl_igmp_max_msf,
 496                 .maxlen         = sizeof(int),
 497                 .mode           = 0644,
 498                 .proc_handler   = proc_dointvec
 499         },
 500         {
 501                 .procname       = "inet_peer_threshold",
 502                 .data           = &inet_peer_threshold,
 503                 .maxlen         = sizeof(int),
 504                 .mode           = 0644,
 505                 .proc_handler   = proc_dointvec
 506         },
 507         {
 508                 .procname       = "inet_peer_minttl",
 509                 .data           = &inet_peer_minttl,
 510                 .maxlen         = sizeof(int),
 511                 .mode           = 0644,
 512                 .proc_handler   = proc_dointvec_jiffies,
 513         },
 514         {
 515                 .procname       = "inet_peer_maxttl",
 516                 .data           = &inet_peer_maxttl,
 517                 .maxlen         = sizeof(int),
 518                 .mode           = 0644,
 519                 .proc_handler   = proc_dointvec_jiffies,
 520         },
 521         {
 522                 .procname       = "tcp_orphan_retries",
 523                 .data           = &sysctl_tcp_orphan_retries,
 524                 .maxlen         = sizeof(int),
 525                 .mode           = 0644,
 526                 .proc_handler   = proc_dointvec
 527         },
 528         {
 529                 .procname       = "tcp_fack",
 530                 .data           = &sysctl_tcp_fack,
 531                 .maxlen         = sizeof(int),
 532                 .mode           = 0644,
 533                 .proc_handler   = proc_dointvec
 534         },
 535         {
 536                 .procname       = "tcp_reordering",
 537                 .data           = &sysctl_tcp_reordering,
 538                 .maxlen         = sizeof(int),
 539                 .mode           = 0644,
 540                 .proc_handler   = proc_dointvec
 541         },
 542         {
 543                 .procname       = "tcp_dsack",
 544                 .data           = &sysctl_tcp_dsack,
 545                 .maxlen         = sizeof(int),
 546                 .mode           = 0644,
 547                 .proc_handler   = proc_dointvec
 548         },
 549         {
 550                 .procname       = "tcp_wmem",
 551                 .data           = &sysctl_tcp_wmem,
 552                 .maxlen         = sizeof(sysctl_tcp_wmem),
 553                 .mode           = 0644,
 554                 .proc_handler   = proc_dointvec_minmax,
 555                 .extra1         = &one,
 556         },
 557         {
 558                 .procname       = "tcp_rmem",
 559                 .data           = &sysctl_tcp_rmem,
 560                 .maxlen         = sizeof(sysctl_tcp_rmem),
 561                 .mode           = 0644,
 562                 .proc_handler   = proc_dointvec_minmax,
 563                 .extra1         = &one,
 564         },
 565         {
 566                 .procname       = "tcp_app_win",
 567                 .data           = &sysctl_tcp_app_win,
 568                 .maxlen         = sizeof(int),
 569                 .mode           = 0644,
 570                 .proc_handler   = proc_dointvec
 571         },
 572         {
 573                 .procname       = "tcp_adv_win_scale",
 574                 .data           = &sysctl_tcp_adv_win_scale,
 575                 .maxlen         = sizeof(int),
 576                 .mode           = 0644,
 577                 .proc_handler   = proc_dointvec_minmax,
 578                 .extra1         = &tcp_adv_win_scale_min,
 579                 .extra2         = &tcp_adv_win_scale_max,
 580         },
 581         {
 582                 .procname       = "tcp_tw_reuse",
 583                 .data           = &sysctl_tcp_tw_reuse,
 584                 .maxlen         = sizeof(int),
 585                 .mode           = 0644,
 586                 .proc_handler   = proc_dointvec
 587         },
 588         {
 589                 .procname       = "tcp_frto",
 590                 .data           = &sysctl_tcp_frto,
 591                 .maxlen         = sizeof(int),
 592                 .mode           = 0644,
 593                 .proc_handler   = proc_dointvec
 594         },
 595         {
 596                 .procname       = "tcp_low_latency",
 597                 .data           = &sysctl_tcp_low_latency,
 598                 .maxlen         = sizeof(int),
 599                 .mode           = 0644,
 600                 .proc_handler   = proc_dointvec
 601         },
 602         {
 603                 .procname       = "tcp_no_metrics_save",
 604                 .data           = &sysctl_tcp_nometrics_save,
 605                 .maxlen         = sizeof(int),
 606                 .mode           = 0644,
 607                 .proc_handler   = proc_dointvec,
 608         },
 609         {
 610                 .procname       = "tcp_moderate_rcvbuf",
 611                 .data           = &sysctl_tcp_moderate_rcvbuf,
 612                 .maxlen         = sizeof(int),
 613                 .mode           = 0644,
 614                 .proc_handler   = proc_dointvec,
 615         },
 616         {
 617                 .procname       = "tcp_tso_win_divisor",
 618                 .data           = &sysctl_tcp_tso_win_divisor,
 619                 .maxlen         = sizeof(int),
 620                 .mode           = 0644,
 621                 .proc_handler   = proc_dointvec,
 622         },
 623         {
 624                 .procname       = "tcp_congestion_control",
 625                 .mode           = 0644,
 626                 .maxlen         = TCP_CA_NAME_MAX,
 627                 .proc_handler   = proc_tcp_congestion_control,
 628         },
 629         {
 630                 .procname       = "tcp_mtu_probing",
 631                 .data           = &sysctl_tcp_mtu_probing,
 632                 .maxlen         = sizeof(int),
 633                 .mode           = 0644,
 634                 .proc_handler   = proc_dointvec,
 635         },
 636         {
 637                 .procname       = "tcp_base_mss",
 638                 .data           = &sysctl_tcp_base_mss,
 639                 .maxlen         = sizeof(int),
 640                 .mode           = 0644,
 641                 .proc_handler   = proc_dointvec,
 642         },
 643         {
 644                 .procname       = "tcp_workaround_signed_windows",
 645                 .data           = &sysctl_tcp_workaround_signed_windows,
 646                 .maxlen         = sizeof(int),
 647                 .mode           = 0644,
 648                 .proc_handler   = proc_dointvec
 649         },
 650         {
 651                 .procname       = "tcp_limit_output_bytes",
 652                 .data           = &sysctl_tcp_limit_output_bytes,
 653                 .maxlen         = sizeof(int),
 654                 .mode           = 0644,
 655                 .proc_handler   = proc_dointvec
 656         },
 657         {
 658                 .procname       = "tcp_challenge_ack_limit",
 659                 .data           = &sysctl_tcp_challenge_ack_limit,
 660                 .maxlen         = sizeof(int),
 661                 .mode           = 0644,
 662                 .proc_handler   = proc_dointvec
 663         },
 664 #ifdef CONFIG_NET_DMA
 665         {
 666                 .procname       = "tcp_dma_copybreak",
 667                 .data           = &sysctl_tcp_dma_copybreak,
 668                 .maxlen         = sizeof(int),
 669                 .mode           = 0644,
 670                 .proc_handler   = proc_dointvec
 671         },
 672 #endif
 673         {
 674                 .procname       = "tcp_slow_start_after_idle",
 675                 .data           = &sysctl_tcp_slow_start_after_idle,
 676                 .maxlen         = sizeof(int),
 677                 .mode           = 0644,
 678                 .proc_handler   = proc_dointvec
 679         },
 680 #ifdef CONFIG_NETLABEL
 681         {
 682                 .procname       = "cipso_cache_enable",
 683                 .data           = &cipso_v4_cache_enabled,
 684                 .maxlen         = sizeof(int),
 685                 .mode           = 0644,
 686                 .proc_handler   = proc_dointvec,
 687         },
 688         {
 689                 .procname       = "cipso_cache_bucket_size",
 690                 .data           = &cipso_v4_cache_bucketsize,
 691                 .maxlen         = sizeof(int),
 692                 .mode           = 0644,
 693                 .proc_handler   = proc_dointvec,
 694         },
 695         {
 696                 .procname       = "cipso_rbm_optfmt",
 697                 .data           = &cipso_v4_rbm_optfmt,
 698                 .maxlen         = sizeof(int),
 699                 .mode           = 0644,
 700                 .proc_handler   = proc_dointvec,
 701         },
 702         {
 703                 .procname       = "cipso_rbm_strictvalid",
 704                 .data           = &cipso_v4_rbm_strictvalid,
 705                 .maxlen         = sizeof(int),
 706                 .mode           = 0644,
 707                 .proc_handler   = proc_dointvec,
 708         },
 709 #endif /* CONFIG_NETLABEL */
 710         {
 711                 .procname       = "tcp_available_congestion_control",
 712                 .maxlen         = TCP_CA_BUF_MAX,
 713                 .mode           = 0444,
 714                 .proc_handler   = proc_tcp_available_congestion_control,
 715         },
 716         {
 717                 .procname       = "tcp_allowed_congestion_control",
 718                 .maxlen         = TCP_CA_BUF_MAX,
 719                 .mode           = 0644,
 720                 .proc_handler   = proc_allowed_congestion_control,
 721         },
 722         {
 723                 .procname       = "tcp_max_ssthresh",
 724                 .data           = &sysctl_tcp_max_ssthresh,
 725                 .maxlen         = sizeof(int),
 726                 .mode           = 0644,
 727                 .proc_handler   = proc_dointvec,
 728         },
 729         {
 730                 .procname       = "tcp_thin_linear_timeouts",
 731                 .data           = &sysctl_tcp_thin_linear_timeouts,
 732                 .maxlen         = sizeof(int),
 733                 .mode           = 0644,
 734                 .proc_handler   = proc_dointvec
 735         },
 736         {
 737                 .procname       = "tcp_thin_dupack",
 738                 .data           = &sysctl_tcp_thin_dupack,
 739                 .maxlen         = sizeof(int),
 740                 .mode           = 0644,
 741                 .proc_handler   = proc_dointvec
 742         },
 743         {
 744                 .procname       = "tcp_early_retrans",
 745                 .data           = &sysctl_tcp_early_retrans,
 746                 .maxlen         = sizeof(int),
 747                 .mode           = 0644,
 748                 .proc_handler   = proc_dointvec_minmax,
 749                 .extra1         = &zero,
 750                 .extra2         = &four,
 751         },
 752         {
 753                 .procname       = "udp_mem",
 754                 .data           = &sysctl_udp_mem,
 755                 .maxlen         = sizeof(sysctl_udp_mem),
 756                 .mode           = 0644,
 757                 .proc_handler   = proc_doulongvec_minmax,
 758         },
 759         {
 760                 .procname       = "udp_rmem_min",
 761                 .data           = &sysctl_udp_rmem_min,
 762                 .maxlen         = sizeof(sysctl_udp_rmem_min),
 763                 .mode           = 0644,
 764                 .proc_handler   = proc_dointvec_minmax,
 765                 .extra1         = &one
 766         },
 767         {
 768                 .procname       = "udp_wmem_min",
 769                 .data           = &sysctl_udp_wmem_min,
 770                 .maxlen         = sizeof(sysctl_udp_wmem_min),
 771                 .mode           = 0644,
 772                 .proc_handler   = proc_dointvec_minmax,
 773                 .extra1         = &one
 774         },
 775         { }
 776 };
 777
 778 static struct ctl_table ipv4_net_table[] = {
 779         {
 780                 .procname       = "icmp_echo_ignore_all",
 781                 .data           = &init_net.ipv4.sysctl_icmp_echo_ignore_all,
 782                 .maxlen         = sizeof(int),
 783                 .mode           = 0644,
 784                 .proc_handler   = proc_dointvec
 785         },
 786         {
 787                 .procname       = "icmp_echo_ignore_broadcasts",
 788                 .data           = &init_net.ipv4.sysctl_icmp_echo_ignore_broadcasts,
 789                 .maxlen         = sizeof(int),
 790                 .mode           = 0644,
 791                 .proc_handler   = proc_dointvec
 792         },
 793         {
 794                 .procname       = "icmp_ignore_bogus_error_responses",
 795                 .data           = &init_net.ipv4.sysctl_icmp_ignore_bogus_error_responses,
 796                 .maxlen         = sizeof(int),
 797                 .mode           = 0644,
 798                 .proc_handler   = proc_dointvec
 799         },
 800         {
 801                 .procname       = "icmp_errors_use_inbound_ifaddr",
 802                 .data           = &init_net.ipv4.sysctl_icmp_errors_use_inbound_ifaddr,
 803                 .maxlen         = sizeof(int),
 804                 .mode           = 0644,
 805                 .proc_handler   = proc_dointvec
 806         },
 807         {
 808                 .procname       = "icmp_ratelimit",
 809                 .data           = &init_net.ipv4.sysctl_icmp_ratelimit,
 810                 .maxlen         = sizeof(int),
 811                 .mode           = 0644,
 812                 .proc_handler   = proc_dointvec_ms_jiffies,
 813         },
 814         {
 815                 .procname       = "icmp_ratemask",
 816                 .data           = &init_net.ipv4.sysctl_icmp_ratemask,
 817                 .maxlen         = sizeof(int),
 818                 .mode           = 0644,
 819                 .proc_handler   = proc_dointvec
 820         },
 821         {
 822                 .procname       = "ping_group_range",
 823                 .data           = &init_net.ipv4.sysctl_ping_group_range,
 824                 .maxlen         = sizeof(gid_t)*2,
 825                 .mode           = 0644,
 826                 .proc_handler   = ipv4_ping_group_range,
 827         },
 828         {
 829                 .procname       = "tcp_ecn",
 830                 .data           = &init_net.ipv4.sysctl_tcp_ecn,
 831                 .maxlen         = sizeof(int),
 832                 .mode           = 0644,
 833                 .proc_handler   = proc_dointvec
 834         },
 835         {
 836                 .procname       = "tcp_mem",
 837                 .maxlen         = sizeof(init_net.ipv4.sysctl_tcp_mem),
 838                 .mode           = 0644,
 839                 .proc_handler   = ipv4_tcp_mem,
 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                 table = kmemdup(table, sizeof(ipv4_net_table), GFP_KERNEL);
 851                 if (table == NULL)
 852                         goto err_alloc;
 853
 854                 table[0].data =
 855                         &net->ipv4.sysctl_icmp_echo_ignore_all;
 856                 table[1].data =
 857                         &net->ipv4.sysctl_icmp_echo_ignore_broadcasts;
 858                 table[2].data =
 859                         &net->ipv4.sysctl_icmp_ignore_bogus_error_responses;
 860                 table[3].data =
 861                         &net->ipv4.sysctl_icmp_errors_use_inbound_ifaddr;
 862                 table[4].data =
 863                         &net->ipv4.sysctl_icmp_ratelimit;
 864                 table[5].data =
 865                         &net->ipv4.sysctl_icmp_ratemask;
 866                 table[6].data =
 867                         &net->ipv4.sysctl_ping_group_range;
 868                 table[7].data =
 869                         &net->ipv4.sysctl_tcp_ecn;
 870
 871                 /* Don't export sysctls to unprivileged users */
 872                 if (net->user_ns != &init_user_ns)
 873                         table[0].procname = NULL;
 874         }
 875
 876         /*
 877          * Sane defaults - nobody may create ping sockets.
 878          * Boot scripts should set this to distro-specific group.
 879          */
 880         net->ipv4.sysctl_ping_group_range[0] = make_kgid(&init_user_ns, 1);
 881         net->ipv4.sysctl_ping_group_range[1] = make_kgid(&init_user_ns, 0);
 882
 883         tcp_init_mem(net);
 884
 885         net->ipv4.ipv4_hdr = register_net_sysctl(net, "net/ipv4", table);
 886         if (net->ipv4.ipv4_hdr == NULL)
 887                 goto err_reg;
 888
 889         return 0;
 890
 891 err_reg:
 892         if (!net_eq(net, &init_net))
 893                 kfree(table);
 894 err_alloc:
 895         return -ENOMEM;
 896 }
 897
 898 static __net_exit void ipv4_sysctl_exit_net(struct net *net)
 899 {
 900         struct ctl_table *table;
 901
 902         table = net->ipv4.ipv4_hdr->ctl_table_arg;
 903         unregister_net_sysctl_table(net->ipv4.ipv4_hdr);
 904         kfree(table);
 905 }
 906
 907 static __net_initdata struct pernet_operations ipv4_sysctl_ops = {
 908         .init = ipv4_sysctl_init_net,
 909         .exit = ipv4_sysctl_exit_net,
 910 };
 911
 912 static __init int sysctl_ipv4_init(void)
 913 {
 914         struct ctl_table_header *hdr;
 915         struct ctl_table *i;
 916
 917         for (i = ipv4_table; i->procname; i++) {
 918                 if (strcmp(i->procname, "ip_local_reserved_ports") == 0) {
 919                         i->data = sysctl_local_reserved_ports;
 920                         break;
 921                 }
 922         }
 923         if (!i->procname)
 924                 return -EINVAL;
 925
 926         hdr = register_net_sysctl(&init_net, "net/ipv4", ipv4_table);
 927         if (hdr == NULL)
 928                 return -ENOMEM;
 929
 930         if (register_pernet_subsys(&ipv4_sysctl_ops)) {
 931                 unregister_net_sysctl_table(hdr);
 932                 return -ENOMEM;
 933         }
 934
 935         return 0;
 936 }
 937
 938 __initcall(sysctl_ipv4_init);