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(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         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(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         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 /* Validate changes from /proc interface. */
 139 static int proc_tcp_default_init_rwnd(ctl_table *ctl, int write,
 140                                       void __user *buffer,
 141                                       size_t *lenp, loff_t *ppos)
 142 {
 143         int old_value = *(int *)ctl->data;
 144         int ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
 145         int new_value = *(int *)ctl->data;
 146
 147         if (write && ret == 0 && (new_value < 3 || new_value > 100))
 148                 *(int *)ctl->data = old_value;
 149
 150         return ret;
 151 }
 152
 153 static int proc_tcp_congestion_control(ctl_table *ctl, int write,
 154                                        void __user *buffer, size_t *lenp, loff_t *ppos)
 155 {
 156         char val[TCP_CA_NAME_MAX];
 157         ctl_table tbl = {
 158                 .data = val,
 159                 .maxlen = TCP_CA_NAME_MAX,
 160         };
 161         int ret;
 162
 163         tcp_get_default_congestion_control(val);
 164
 165         ret = proc_dostring(&tbl, write, buffer, lenp, ppos);
 166         if (write && ret == 0)
 167                 ret = tcp_set_default_congestion_control(val);
 168         return ret;
 169 }
 170
 171 static int proc_tcp_available_congestion_control(ctl_table *ctl,
 172                                                  int write,
 173                                                  void __user *buffer, size_t *lenp,
 174                                                  loff_t *ppos)
 175 {
 176         ctl_table tbl = { .maxlen = TCP_CA_BUF_MAX, };
 177         int ret;
 178
 179         tbl.data = kmalloc(tbl.maxlen, GFP_USER);
 180         if (!tbl.data)
 181                 return -ENOMEM;
 182         tcp_get_available_congestion_control(tbl.data, TCP_CA_BUF_MAX);
 183         ret = proc_dostring(&tbl, write, buffer, lenp, ppos);
 184         kfree(tbl.data);
 185         return ret;
 186 }
 187
 188 static int proc_allowed_congestion_control(ctl_table *ctl,
 189                                            int write,
 190                                            void __user *buffer, size_t *lenp,
 191                                            loff_t *ppos)
 192 {
 193         ctl_table tbl = { .maxlen = TCP_CA_BUF_MAX };
 194         int ret;
 195
 196         tbl.data = kmalloc(tbl.maxlen, GFP_USER);
 197         if (!tbl.data)
 198                 return -ENOMEM;
 199
 200         tcp_get_allowed_congestion_control(tbl.data, tbl.maxlen);
 201         ret = proc_dostring(&tbl, write, buffer, lenp, ppos);
 202         if (write && ret == 0)
 203                 ret = tcp_set_allowed_congestion_control(tbl.data);
 204         kfree(tbl.data);
 205         return ret;
 206 }
 207
 208 static int ipv4_tcp_mem(ctl_table *ctl, int write,
 209                            void __user *buffer, size_t *lenp,
 210                            loff_t *ppos)
 211 {
 212         int ret;
 213         unsigned long vec[3];
 214         struct net *net = current->nsproxy->net_ns;
 215 #ifdef CONFIG_MEMCG_KMEM
 216         struct mem_cgroup *memcg;
 217 #endif
 218
 219         ctl_table tmp = {
 220                 .data = &vec,
 221                 .maxlen = sizeof(vec),
 222                 .mode = ctl->mode,
 223         };
 224
 225         if (!write) {
 226                 ctl->data = &net->ipv4.sysctl_tcp_mem;
 227                 return proc_doulongvec_minmax(ctl, write, buffer, lenp, ppos);
 228         }
 229
 230         ret = proc_doulongvec_minmax(&tmp, write, buffer, lenp, ppos);
 231         if (ret)
 232                 return ret;
 233
 234 #ifdef CONFIG_MEMCG_KMEM
 235         rcu_read_lock();
 236         memcg = mem_cgroup_from_task(current);
 237
 238         tcp_prot_mem(memcg, vec[0], 0);
 239         tcp_prot_mem(memcg, vec[1], 1);
 240         tcp_prot_mem(memcg, vec[2], 2);
 241         rcu_read_unlock();
 242 #endif
 243
 244         net->ipv4.sysctl_tcp_mem[0] = vec[0];
 245         net->ipv4.sysctl_tcp_mem[1] = vec[1];
 246         net->ipv4.sysctl_tcp_mem[2] = vec[2];
 247
 248         return 0;
 249 }
 250
 251 static int proc_tcp_fastopen_key(ctl_table *ctl, int write, void __user *buffer,
 252                                  size_t *lenp, loff_t *ppos)
 253 {
 254         ctl_table tbl = { .maxlen = (TCP_FASTOPEN_KEY_LENGTH * 2 + 10) };
 255         struct tcp_fastopen_context *ctxt;
 256         int ret;
 257         u32  user_key[4]; /* 16 bytes, matching TCP_FASTOPEN_KEY_LENGTH */
 258
 259         tbl.data = kmalloc(tbl.maxlen, GFP_KERNEL);
 260         if (!tbl.data)
 261                 return -ENOMEM;
 262
 263         rcu_read_lock();
 264         ctxt = rcu_dereference(tcp_fastopen_ctx);
 265         if (ctxt)
 266                 memcpy(user_key, ctxt->key, TCP_FASTOPEN_KEY_LENGTH);
 267         else
 268                 memset(user_key, 0, sizeof(user_key));
 269         rcu_read_unlock();
 270
 271         snprintf(tbl.data, tbl.maxlen, "%08x-%08x-%08x-%08x",
 272                 user_key[0], user_key[1], user_key[2], user_key[3]);
 273         ret = proc_dostring(&tbl, write, buffer, lenp, ppos);
 274
 275         if (write && ret == 0) {
 276                 if (sscanf(tbl.data, "%x-%x-%x-%x", user_key, user_key + 1,
 277                            user_key + 2, user_key + 3) != 4) {
 278                         ret = -EINVAL;
 279                         goto bad_key;
 280                 }
 281                 tcp_fastopen_reset_cipher(user_key, TCP_FASTOPEN_KEY_LENGTH);
 282         }
 283
 284 bad_key:
 285         pr_debug("proc FO key set 0x%x-%x-%x-%x <- 0x%s: %u\n",
 286                user_key[0], user_key[1], user_key[2], user_key[3],
 287                (char *)tbl.data, ret);
 288         kfree(tbl.data);
 289         return ret;
 290 }
 291
 292 static struct ctl_table ipv4_table[] = {
 293         {
 294                 .procname       = "tcp_timestamps",
 295                 .data           = &sysctl_tcp_timestamps,
 296                 .maxlen         = sizeof(int),
 297                 .mode           = 0644,
 298                 .proc_handler   = proc_dointvec
 299         },
 300         {
 301                 .procname       = "tcp_window_scaling",
 302                 .data           = &sysctl_tcp_window_scaling,
 303                 .maxlen         = sizeof(int),
 304                 .mode           = 0644,
 305                 .proc_handler   = proc_dointvec
 306         },
 307         {
 308                 .procname       = "tcp_sack",
 309                 .data           = &sysctl_tcp_sack,
 310                 .maxlen         = sizeof(int),
 311                 .mode           = 0644,
 312                 .proc_handler   = proc_dointvec
 313         },
 314         {
 315                 .procname       = "tcp_retrans_collapse",
 316                 .data           = &sysctl_tcp_retrans_collapse,
 317                 .maxlen         = sizeof(int),
 318                 .mode           = 0644,
 319                 .proc_handler   = proc_dointvec
 320         },
 321         {
 322                 .procname       = "ip_default_ttl",
 323                 .data           = &sysctl_ip_default_ttl,
 324                 .maxlen         = sizeof(int),
 325                 .mode           = 0644,
 326                 .proc_handler   = proc_dointvec_minmax,
 327                 .extra1         = &ip_ttl_min,
 328                 .extra2         = &ip_ttl_max,
 329         },
 330         {
 331                 .procname       = "ip_no_pmtu_disc",
 332                 .data           = &ipv4_config.no_pmtu_disc,
 333                 .maxlen         = sizeof(int),
 334                 .mode           = 0644,
 335                 .proc_handler   = proc_dointvec
 336         },
 337         {
 338                 .procname       = "ip_nonlocal_bind",
 339                 .data           = &sysctl_ip_nonlocal_bind,
 340                 .maxlen         = sizeof(int),
 341                 .mode           = 0644,
 342                 .proc_handler   = proc_dointvec
 343         },
 344         {
 345                 .procname       = "tcp_syn_retries",
 346                 .data           = &sysctl_tcp_syn_retries,
 347                 .maxlen         = sizeof(int),
 348                 .mode           = 0644,
 349                 .proc_handler   = proc_dointvec
 350         },
 351         {
 352                 .procname       = "tcp_synack_retries",
 353                 .data           = &sysctl_tcp_synack_retries,
 354                 .maxlen         = sizeof(int),
 355                 .mode           = 0644,
 356                 .proc_handler   = proc_dointvec
 357         },
 358         {
 359                 .procname       = "tcp_max_orphans",
 360                 .data           = &sysctl_tcp_max_orphans,
 361                 .maxlen         = sizeof(int),
 362                 .mode           = 0644,
 363                 .proc_handler   = proc_dointvec
 364         },
 365         {
 366                 .procname       = "tcp_max_tw_buckets",
 367                 .data           = &tcp_death_row.sysctl_max_tw_buckets,
 368                 .maxlen         = sizeof(int),
 369                 .mode           = 0644,
 370                 .proc_handler   = proc_dointvec
 371         },
 372         {
 373                 .procname       = "ip_early_demux",
 374                 .data           = &sysctl_ip_early_demux,
 375                 .maxlen         = sizeof(int),
 376                 .mode           = 0644,
 377                 .proc_handler   = proc_dointvec
 378         },
 379         {
 380                 .procname       = "ip_dynaddr",
 381                 .data           = &sysctl_ip_dynaddr,
 382                 .maxlen         = sizeof(int),
 383                 .mode           = 0644,
 384                 .proc_handler   = proc_dointvec
 385         },
 386         {
 387                 .procname       = "tcp_keepalive_time",
 388                 .data           = &sysctl_tcp_keepalive_time,
 389                 .maxlen         = sizeof(int),
 390                 .mode           = 0644,
 391                 .proc_handler   = proc_dointvec_jiffies,
 392         },
 393         {
 394                 .procname       = "tcp_keepalive_probes",
 395                 .data           = &sysctl_tcp_keepalive_probes,
 396                 .maxlen         = sizeof(int),
 397                 .mode           = 0644,
 398                 .proc_handler   = proc_dointvec
 399         },
 400         {
 401                 .procname       = "tcp_keepalive_intvl",
 402                 .data           = &sysctl_tcp_keepalive_intvl,
 403                 .maxlen         = sizeof(int),
 404                 .mode           = 0644,
 405                 .proc_handler   = proc_dointvec_jiffies,
 406         },
 407         {
 408                 .procname       = "tcp_retries1",
 409                 .data           = &sysctl_tcp_retries1,
 410                 .maxlen         = sizeof(int),
 411                 .mode           = 0644,
 412                 .proc_handler   = proc_dointvec_minmax,
 413                 .extra2         = &tcp_retr1_max
 414         },
 415         {
 416                 .procname       = "tcp_retries2",
 417                 .data           = &sysctl_tcp_retries2,
 418                 .maxlen         = sizeof(int),
 419                 .mode           = 0644,
 420                 .proc_handler   = proc_dointvec
 421         },
 422         {
 423                 .procname       = "tcp_fin_timeout",
 424                 .data           = &sysctl_tcp_fin_timeout,
 425                 .maxlen         = sizeof(int),
 426                 .mode           = 0644,
 427                 .proc_handler   = proc_dointvec_jiffies,
 428         },
 429 #ifdef CONFIG_SYN_COOKIES
 430         {
 431                 .procname       = "tcp_syncookies",
 432                 .data           = &sysctl_tcp_syncookies,
 433                 .maxlen         = sizeof(int),
 434                 .mode           = 0644,
 435                 .proc_handler   = proc_dointvec
 436         },
 437 #endif
 438         {
 439                 .procname       = "tcp_fastopen",
 440                 .data           = &sysctl_tcp_fastopen,
 441                 .maxlen         = sizeof(int),
 442                 .mode           = 0644,
 443                 .proc_handler   = proc_dointvec,
 444         },
 445         {
 446                 .procname       = "tcp_fastopen_key",
 447                 .mode           = 0600,
 448                 .maxlen         = ((TCP_FASTOPEN_KEY_LENGTH * 2) + 10),
 449                 .proc_handler   = proc_tcp_fastopen_key,
 450         },
 451         {
 452                 .procname       = "tcp_tw_recycle",
 453                 .data           = &tcp_death_row.sysctl_tw_recycle,
 454                 .maxlen         = sizeof(int),
 455                 .mode           = 0644,
 456                 .proc_handler   = proc_dointvec
 457         },
 458         {
 459                 .procname       = "tcp_abort_on_overflow",
 460                 .data           = &sysctl_tcp_abort_on_overflow,
 461                 .maxlen         = sizeof(int),
 462                 .mode           = 0644,
 463                 .proc_handler   = proc_dointvec
 464         },
 465         {
 466                 .procname       = "tcp_stdurg",
 467                 .data           = &sysctl_tcp_stdurg,
 468                 .maxlen         = sizeof(int),
 469                 .mode           = 0644,
 470                 .proc_handler   = proc_dointvec
 471         },
 472         {
 473                 .procname       = "tcp_rfc1337",
 474                 .data           = &sysctl_tcp_rfc1337,
 475                 .maxlen         = sizeof(int),
 476                 .mode           = 0644,
 477                 .proc_handler   = proc_dointvec
 478         },
 479         {
 480                 .procname       = "tcp_max_syn_backlog",
 481                 .data           = &sysctl_max_syn_backlog,
 482                 .maxlen         = sizeof(int),
 483                 .mode           = 0644,
 484                 .proc_handler   = proc_dointvec
 485         },
 486         {
 487                 .procname       = "ip_local_port_range",
 488                 .data           = &sysctl_local_ports.range,
 489                 .maxlen         = sizeof(sysctl_local_ports.range),
 490                 .mode           = 0644,
 491                 .proc_handler   = ipv4_local_port_range,
 492         },
 493         {
 494                 .procname       = "ip_local_reserved_ports",
 495                 .data           = NULL, /* initialized in sysctl_ipv4_init */
 496                 .maxlen         = 65536,
 497                 .mode           = 0644,
 498                 .proc_handler   = proc_do_large_bitmap,
 499         },
 500         {
 501                 .procname       = "igmp_max_memberships",
 502                 .data           = &sysctl_igmp_max_memberships,
 503                 .maxlen         = sizeof(int),
 504                 .mode           = 0644,
 505                 .proc_handler   = proc_dointvec
 506         },
 507         {
 508                 .procname       = "igmp_max_msf",
 509                 .data           = &sysctl_igmp_max_msf,
 510                 .maxlen         = sizeof(int),
 511                 .mode           = 0644,
 512                 .proc_handler   = proc_dointvec
 513         },
 514         {
 515                 .procname       = "inet_peer_threshold",
 516                 .data           = &inet_peer_threshold,
 517                 .maxlen         = sizeof(int),
 518                 .mode           = 0644,
 519                 .proc_handler   = proc_dointvec
 520         },
 521         {
 522                 .procname       = "inet_peer_minttl",
 523                 .data           = &inet_peer_minttl,
 524                 .maxlen         = sizeof(int),
 525                 .mode           = 0644,
 526                 .proc_handler   = proc_dointvec_jiffies,
 527         },
 528         {
 529                 .procname       = "inet_peer_maxttl",
 530                 .data           = &inet_peer_maxttl,
 531                 .maxlen         = sizeof(int),
 532                 .mode           = 0644,
 533                 .proc_handler   = proc_dointvec_jiffies,
 534         },
 535         {
 536                 .procname       = "tcp_orphan_retries",
 537                 .data           = &sysctl_tcp_orphan_retries,
 538                 .maxlen         = sizeof(int),
 539                 .mode           = 0644,
 540                 .proc_handler   = proc_dointvec
 541         },
 542         {
 543                 .procname       = "tcp_fack",
 544                 .data           = &sysctl_tcp_fack,
 545                 .maxlen         = sizeof(int),
 546                 .mode           = 0644,
 547                 .proc_handler   = proc_dointvec
 548         },
 549         {
 550                 .procname       = "tcp_reordering",
 551                 .data           = &sysctl_tcp_reordering,
 552                 .maxlen         = sizeof(int),
 553                 .mode           = 0644,
 554                 .proc_handler   = proc_dointvec
 555         },
 556         {
 557                 .procname       = "tcp_dsack",
 558                 .data           = &sysctl_tcp_dsack,
 559                 .maxlen         = sizeof(int),
 560                 .mode           = 0644,
 561                 .proc_handler   = proc_dointvec
 562         },
 563         {
 564                 .procname       = "tcp_wmem",
 565                 .data           = &sysctl_tcp_wmem,
 566                 .maxlen         = sizeof(sysctl_tcp_wmem),
 567                 .mode           = 0644,
 568                 .proc_handler   = proc_dointvec_minmax,
 569                 .extra1         = &one,
 570         },
 571         {
 572                 .procname       = "tcp_rmem",
 573                 .data           = &sysctl_tcp_rmem,
 574                 .maxlen         = sizeof(sysctl_tcp_rmem),
 575                 .mode           = 0644,
 576                 .proc_handler   = proc_dointvec_minmax,
 577                 .extra1         = &one,
 578         },
 579         {
 580                 .procname       = "tcp_app_win",
 581                 .data           = &sysctl_tcp_app_win,
 582                 .maxlen         = sizeof(int),
 583                 .mode           = 0644,
 584                 .proc_handler   = proc_dointvec
 585         },
 586         {
 587                 .procname       = "tcp_adv_win_scale",
 588                 .data           = &sysctl_tcp_adv_win_scale,
 589                 .maxlen         = sizeof(int),
 590                 .mode           = 0644,
 591                 .proc_handler   = proc_dointvec_minmax,
 592                 .extra1         = &tcp_adv_win_scale_min,
 593                 .extra2         = &tcp_adv_win_scale_max,
 594         },
 595         {
 596                 .procname       = "tcp_tw_reuse",
 597                 .data           = &sysctl_tcp_tw_reuse,
 598                 .maxlen         = sizeof(int),
 599                 .mode           = 0644,
 600                 .proc_handler   = proc_dointvec
 601         },
 602         {
 603                 .procname       = "tcp_frto",
 604                 .data           = &sysctl_tcp_frto,
 605                 .maxlen         = sizeof(int),
 606                 .mode           = 0644,
 607                 .proc_handler   = proc_dointvec
 608         },
 609         {
 610                 .procname       = "tcp_low_latency",
 611                 .data           = &sysctl_tcp_low_latency,
 612                 .maxlen         = sizeof(int),
 613                 .mode           = 0644,
 614                 .proc_handler   = proc_dointvec
 615         },
 616         {
 617                 .procname       = "tcp_no_metrics_save",
 618                 .data           = &sysctl_tcp_nometrics_save,
 619                 .maxlen         = sizeof(int),
 620                 .mode           = 0644,
 621                 .proc_handler   = proc_dointvec,
 622         },
 623         {
 624                 .procname       = "tcp_moderate_rcvbuf",
 625                 .data           = &sysctl_tcp_moderate_rcvbuf,
 626                 .maxlen         = sizeof(int),
 627                 .mode           = 0644,
 628                 .proc_handler   = proc_dointvec,
 629         },
 630         {
 631                 .procname       = "tcp_tso_win_divisor",
 632                 .data           = &sysctl_tcp_tso_win_divisor,
 633                 .maxlen         = sizeof(int),
 634                 .mode           = 0644,
 635                 .proc_handler   = proc_dointvec,
 636         },
 637         {
 638                 .procname       = "tcp_congestion_control",
 639                 .mode           = 0644,
 640                 .maxlen         = TCP_CA_NAME_MAX,
 641                 .proc_handler   = proc_tcp_congestion_control,
 642         },
 643         {
 644                 .procname       = "tcp_mtu_probing",
 645                 .data           = &sysctl_tcp_mtu_probing,
 646                 .maxlen         = sizeof(int),
 647                 .mode           = 0644,
 648                 .proc_handler   = proc_dointvec,
 649         },
 650         {
 651                 .procname       = "tcp_base_mss",
 652                 .data           = &sysctl_tcp_base_mss,
 653                 .maxlen         = sizeof(int),
 654                 .mode           = 0644,
 655                 .proc_handler   = proc_dointvec,
 656         },
 657         {
 658                 .procname       = "tcp_workaround_signed_windows",
 659                 .data           = &sysctl_tcp_workaround_signed_windows,
 660                 .maxlen         = sizeof(int),
 661                 .mode           = 0644,
 662                 .proc_handler   = proc_dointvec
 663         },
 664         {
 665                 .procname       = "tcp_limit_output_bytes",
 666                 .data           = &sysctl_tcp_limit_output_bytes,
 667                 .maxlen         = sizeof(int),
 668                 .mode           = 0644,
 669                 .proc_handler   = proc_dointvec
 670         },
 671         {
 672                 .procname       = "tcp_challenge_ack_limit",
 673                 .data           = &sysctl_tcp_challenge_ack_limit,
 674                 .maxlen         = sizeof(int),
 675                 .mode           = 0644,
 676                 .proc_handler   = proc_dointvec
 677         },
 678 #ifdef CONFIG_NET_DMA
 679         {
 680                 .procname       = "tcp_dma_copybreak",
 681                 .data           = &sysctl_tcp_dma_copybreak,
 682                 .maxlen         = sizeof(int),
 683                 .mode           = 0644,
 684                 .proc_handler   = proc_dointvec
 685         },
 686 #endif
 687         {
 688                 .procname       = "tcp_slow_start_after_idle",
 689                 .data           = &sysctl_tcp_slow_start_after_idle,
 690                 .maxlen         = sizeof(int),
 691                 .mode           = 0644,
 692                 .proc_handler   = proc_dointvec
 693         },
 694 #ifdef CONFIG_NETLABEL
 695         {
 696                 .procname       = "cipso_cache_enable",
 697                 .data           = &cipso_v4_cache_enabled,
 698                 .maxlen         = sizeof(int),
 699                 .mode           = 0644,
 700                 .proc_handler   = proc_dointvec,
 701         },
 702         {
 703                 .procname       = "cipso_cache_bucket_size",
 704                 .data           = &cipso_v4_cache_bucketsize,
 705                 .maxlen         = sizeof(int),
 706                 .mode           = 0644,
 707                 .proc_handler   = proc_dointvec,
 708         },
 709         {
 710                 .procname       = "cipso_rbm_optfmt",
 711                 .data           = &cipso_v4_rbm_optfmt,
 712                 .maxlen         = sizeof(int),
 713                 .mode           = 0644,
 714                 .proc_handler   = proc_dointvec,
 715         },
 716         {
 717                 .procname       = "cipso_rbm_strictvalid",
 718                 .data           = &cipso_v4_rbm_strictvalid,
 719                 .maxlen         = sizeof(int),
 720                 .mode           = 0644,
 721                 .proc_handler   = proc_dointvec,
 722         },
 723 #endif /* CONFIG_NETLABEL */
 724         {
 725                 .procname       = "tcp_available_congestion_control",
 726                 .maxlen         = TCP_CA_BUF_MAX,
 727                 .mode           = 0444,
 728                 .proc_handler   = proc_tcp_available_congestion_control,
 729         },
 730         {
 731                 .procname       = "tcp_allowed_congestion_control",
 732                 .maxlen         = TCP_CA_BUF_MAX,
 733                 .mode           = 0644,
 734                 .proc_handler   = proc_allowed_congestion_control,
 735         },
 736         {
 737                 .procname       = "tcp_max_ssthresh",
 738                 .data           = &sysctl_tcp_max_ssthresh,
 739                 .maxlen         = sizeof(int),
 740                 .mode           = 0644,
 741                 .proc_handler   = proc_dointvec,
 742         },
 743         {
 744                 .procname       = "tcp_thin_linear_timeouts",
 745                 .data           = &sysctl_tcp_thin_linear_timeouts,
 746                 .maxlen         = sizeof(int),
 747                 .mode           = 0644,
 748                 .proc_handler   = proc_dointvec
 749         },
 750         {
 751                 .procname       = "tcp_thin_dupack",
 752                 .data           = &sysctl_tcp_thin_dupack,
 753                 .maxlen         = sizeof(int),
 754                 .mode           = 0644,
 755                 .proc_handler   = proc_dointvec
 756         },
 757         {
 758                 .procname       = "tcp_early_retrans",
 759                 .data           = &sysctl_tcp_early_retrans,
 760                 .maxlen         = sizeof(int),
 761                 .mode           = 0644,
 762                 .proc_handler   = proc_dointvec_minmax,
 763                 .extra1         = &zero,
 764                 .extra2         = &four,
 765         },
 766         {
 767                 .procname       = "tcp_default_init_rwnd",
 768                 .data           = &sysctl_tcp_default_init_rwnd,
 769                 .maxlen         = sizeof(int),
 770                 .mode           = 0644,
 771                 .proc_handler   = proc_tcp_default_init_rwnd
 772         },
 773         {
 774                 .procname       = "udp_mem",
 775                 .data           = &sysctl_udp_mem,
 776                 .maxlen         = sizeof(sysctl_udp_mem),
 777                 .mode           = 0644,
 778                 .proc_handler   = proc_doulongvec_minmax,
 779         },
 780         {
 781                 .procname       = "udp_rmem_min",
 782                 .data           = &sysctl_udp_rmem_min,
 783                 .maxlen         = sizeof(sysctl_udp_rmem_min),
 784                 .mode           = 0644,
 785                 .proc_handler   = proc_dointvec_minmax,
 786                 .extra1         = &one
 787         },
 788         {
 789                 .procname       = "udp_wmem_min",
 790                 .data           = &sysctl_udp_wmem_min,
 791                 .maxlen         = sizeof(sysctl_udp_wmem_min),
 792                 .mode           = 0644,
 793                 .proc_handler   = proc_dointvec_minmax,
 794                 .extra1         = &one
 795         },
 796         { }
 797 };
 798
 799 static struct ctl_table ipv4_net_table[] = {
 800         {
 801                 .procname       = "icmp_echo_ignore_all",
 802                 .data           = &init_net.ipv4.sysctl_icmp_echo_ignore_all,
 803                 .maxlen         = sizeof(int),
 804                 .mode           = 0644,
 805                 .proc_handler   = proc_dointvec
 806         },
 807         {
 808                 .procname       = "icmp_echo_ignore_broadcasts",
 809                 .data           = &init_net.ipv4.sysctl_icmp_echo_ignore_broadcasts,
 810                 .maxlen         = sizeof(int),
 811                 .mode           = 0644,
 812                 .proc_handler   = proc_dointvec
 813         },
 814         {
 815                 .procname       = "icmp_ignore_bogus_error_responses",
 816                 .data           = &init_net.ipv4.sysctl_icmp_ignore_bogus_error_responses,
 817                 .maxlen         = sizeof(int),
 818                 .mode           = 0644,
 819                 .proc_handler   = proc_dointvec
 820         },
 821         {
 822                 .procname       = "icmp_errors_use_inbound_ifaddr",
 823                 .data           = &init_net.ipv4.sysctl_icmp_errors_use_inbound_ifaddr,
 824                 .maxlen         = sizeof(int),
 825                 .mode           = 0644,
 826                 .proc_handler   = proc_dointvec
 827         },
 828         {
 829                 .procname       = "icmp_ratelimit",
 830                 .data           = &init_net.ipv4.sysctl_icmp_ratelimit,
 831                 .maxlen         = sizeof(int),
 832                 .mode           = 0644,
 833                 .proc_handler   = proc_dointvec_ms_jiffies,
 834         },
 835         {
 836                 .procname       = "icmp_ratemask",
 837                 .data           = &init_net.ipv4.sysctl_icmp_ratemask,
 838                 .maxlen         = sizeof(int),
 839                 .mode           = 0644,
 840                 .proc_handler   = proc_dointvec
 841         },
 842         {
 843                 .procname       = "ping_group_range",
 844                 .data           = &init_net.ipv4.sysctl_ping_group_range,
 845                 .maxlen         = sizeof(gid_t)*2,
 846                 .mode           = 0644,
 847                 .proc_handler   = ipv4_ping_group_range,
 848         },
 849         {
 850                 .procname       = "tcp_ecn",
 851                 .data           = &init_net.ipv4.sysctl_tcp_ecn,
 852                 .maxlen         = sizeof(int),
 853                 .mode           = 0644,
 854                 .proc_handler   = proc_dointvec
 855         },
 856         {
 857                 .procname       = "tcp_mem",
 858                 .maxlen         = sizeof(init_net.ipv4.sysctl_tcp_mem),
 859                 .mode           = 0644,
 860                 .proc_handler   = ipv4_tcp_mem,
 861         },
 862         {
 863                 .procname       = "fwmark_reflect",
 864                 .data           = &init_net.ipv4.sysctl_fwmark_reflect,
 865                 .maxlen         = sizeof(int),
 866                 .mode           = 0644,
 867                 .proc_handler   = proc_dointvec,
 868         },
 869         {
 870                 .procname       = "tcp_fwmark_accept",
 871                 .data           = &init_net.ipv4.sysctl_tcp_fwmark_accept,
 872                 .maxlen         = sizeof(int),
 873                 .mode           = 0644,
 874                 .proc_handler   = proc_dointvec,
 875         },
 876         { }
 877 };
 878
 879 static __net_init int ipv4_sysctl_init_net(struct net *net)
 880 {
 881         struct ctl_table *table;
 882
 883         table = ipv4_net_table;
 884         if (!net_eq(net, &init_net)) {
 885                 table = kmemdup(table, sizeof(ipv4_net_table), GFP_KERNEL);
 886                 if (table == NULL)
 887                         goto err_alloc;
 888
 889                 table[0].data =
 890                         &net->ipv4.sysctl_icmp_echo_ignore_all;
 891                 table[1].data =
 892                         &net->ipv4.sysctl_icmp_echo_ignore_broadcasts;
 893                 table[2].data =
 894                         &net->ipv4.sysctl_icmp_ignore_bogus_error_responses;
 895                 table[3].data =
 896                         &net->ipv4.sysctl_icmp_errors_use_inbound_ifaddr;
 897                 table[4].data =
 898                         &net->ipv4.sysctl_icmp_ratelimit;
 899                 table[5].data =
 900                         &net->ipv4.sysctl_icmp_ratemask;
 901                 table[6].data =
 902                         &net->ipv4.sysctl_ping_group_range;
 903                 table[7].data =
 904                         &net->ipv4.sysctl_tcp_ecn;
 905
 906                 /* Don't export sysctls to unprivileged users */
 907                 if (net->user_ns != &init_user_ns)
 908                         table[0].procname = NULL;
 909         }
 910
 911         /*
 912          * Sane defaults - nobody may create ping sockets.
 913          * Boot scripts should set this to distro-specific group.
 914          */
 915         net->ipv4.sysctl_ping_group_range[0] = make_kgid(&init_user_ns, 1);
 916         net->ipv4.sysctl_ping_group_range[1] = make_kgid(&init_user_ns, 0);
 917
 918         tcp_init_mem(net);
 919
 920         net->ipv4.ipv4_hdr = register_net_sysctl(net, "net/ipv4", table);
 921         if (net->ipv4.ipv4_hdr == NULL)
 922                 goto err_reg;
 923
 924         return 0;
 925
 926 err_reg:
 927         if (!net_eq(net, &init_net))
 928                 kfree(table);
 929 err_alloc:
 930         return -ENOMEM;
 931 }
 932
 933 static __net_exit void ipv4_sysctl_exit_net(struct net *net)
 934 {
 935         struct ctl_table *table;
 936
 937         table = net->ipv4.ipv4_hdr->ctl_table_arg;
 938         unregister_net_sysctl_table(net->ipv4.ipv4_hdr);
 939         kfree(table);
 940 }
 941
 942 static __net_initdata struct pernet_operations ipv4_sysctl_ops = {
 943         .init = ipv4_sysctl_init_net,
 944         .exit = ipv4_sysctl_exit_net,
 945 };
 946
 947 static __init int sysctl_ipv4_init(void)
 948 {
 949         struct ctl_table_header *hdr;
 950         struct ctl_table *i;
 951
 952         for (i = ipv4_table; i->procname; i++) {
 953                 if (strcmp(i->procname, "ip_local_reserved_ports") == 0) {
 954                         i->data = sysctl_local_reserved_ports;
 955                         break;
 956                 }
 957         }
 958         if (!i->procname)
 959                 return -EINVAL;
 960
 961         hdr = register_net_sysctl(&init_net, "net/ipv4", ipv4_table);
 962         if (hdr == NULL)
 963                 return -ENOMEM;
 964
 965         if (register_pernet_subsys(&ipv4_sysctl_ops)) {
 966                 unregister_net_sysctl_table(hdr);
 967                 return -ENOMEM;
 968         }
 969
 970         return 0;
 971 }
 972
 973 __initcall(sysctl_ipv4_init);