net/ipv4/sysctl_net_ipv4.c

   1 /*
   2  * sysctl_net_ipv4.c: sysctl interface to net IPV4 subsystem.
   3  *
   4  * Begun April 1, 1996, Mike Shaver.
   5  * Added /proc/sys/net/ipv4 directory entry (empty =) ). [MS]
   6  */
   7
   8 #include <linux/mm.h>
   9 #include <linux/module.h>
  10 #include <linux/sysctl.h>
  11 #include <linux/igmp.h>
  12 #include <linux/inetdevice.h>
  13 #include <linux/seqlock.h>
  14 #include <linux/init.h>
  15 #include <linux/slab.h>
  16 #include <linux/nsproxy.h>
  17 #include <linux/swap.h>
  18 #include <net/snmp.h>
  19 #include <net/icmp.h>
  20 #include <net/ip.h>
  21 #include <net/route.h>
  22 #include <net/tcp.h>
  23 #include <net/udp.h>
  24 #include <net/cipso_ipv4.h>
  25 #include <net/inet_frag.h>
  26 #include <net/ping.h>
  27 #include <net/tcp_memcontrol.h>
  28
  29 static int zero;
  30 static int one = 1;
  31 static int four = 4;
  32 static int gso_max_segs = GSO_MAX_SEGS;
  33 static int tcp_retr1_max = 255;
  34 static int ip_local_port_range_min[] = { 1, 1 };
  35 static int ip_local_port_range_max[] = { 65535, 65535 };
  36 static int tcp_adv_win_scale_min = -31;
  37 static int tcp_adv_win_scale_max = 31;
  38 static int ip_ttl_min = 1;
  39 static int ip_ttl_max = 255;
  40 static int tcp_syn_retries_min = 1;
  41 static int tcp_syn_retries_max = MAX_TCP_SYNCNT;
  42 static int ip_ping_group_range_min[] = { 0, 0 };
  43 static int ip_ping_group_range_max[] = { GID_T_MAX, GID_T_MAX };
  44
  45 /* Update system visible IP port range */
  46 static void set_local_port_range(int range[2])
  47 {
  48         write_seqlock(&sysctl_local_ports.lock);
  49         sysctl_local_ports.range[0] = range[0];
  50         sysctl_local_ports.range[1] = range[1];
  51         write_sequnlock(&sysctl_local_ports.lock);
  52 }
  53
  54 /* Validate changes from /proc interface. */
  55 static int ipv4_local_port_range(struct ctl_table *table, int write,
  56                                  void __user *buffer,
  57                                  size_t *lenp, loff_t *ppos)
  58 {
  59         int ret;
  60         int range[2];
  61         struct ctl_table tmp = {
  62                 .data = &range,
  63                 .maxlen = sizeof(range),
  64                 .mode = table->mode,
  65                 .extra1 = &ip_local_port_range_min,
  66                 .extra2 = &ip_local_port_range_max,
  67         };
  68
  69         inet_get_local_port_range(range, range + 1);
  70         ret = proc_dointvec_minmax(&tmp, write, buffer, lenp, ppos);
  71
  72         if (write && ret == 0) {
  73                 if (range[1] < range[0])
  74                         ret = -EINVAL;
  75                 else
  76                         set_local_port_range(range);
  77         }
  78
  79         return ret;
  80 }
  81
  82
  83 static void inet_get_ping_group_range_table(struct ctl_table *table, kgid_t *low, kgid_t *high)
  84 {
  85         kgid_t *data = table->data;
  86         unsigned int seq;
  87         do {
  88                 seq = read_seqbegin(&sysctl_local_ports.lock);
  89
  90                 *low = data[0];
  91                 *high = data[1];
  92         } while (read_seqretry(&sysctl_local_ports.lock, seq));
  93 }
  94
  95 /* Update system visible IP port range */
  96 static void set_ping_group_range(struct ctl_table *table, kgid_t low, kgid_t high)
  97 {
  98         kgid_t *data = table->data;
  99         write_seqlock(&sysctl_local_ports.lock);
 100         data[0] = low;
 101         data[1] = high;
 102         write_sequnlock(&sysctl_local_ports.lock);
 103 }
 104
 105 /* Validate changes from /proc interface. */
 106 static int ipv4_ping_group_range(struct ctl_table *table, int write,
 107                                  void __user *buffer,
 108                                  size_t *lenp, loff_t *ppos)
 109 {
 110         struct user_namespace *user_ns = current_user_ns();
 111         int ret;
 112         gid_t urange[2];
 113         kgid_t low, high;
 114         struct ctl_table tmp = {
 115                 .data = &urange,
 116                 .maxlen = sizeof(urange),
 117                 .mode = table->mode,
 118                 .extra1 = &ip_ping_group_range_min,
 119                 .extra2 = &ip_ping_group_range_max,
 120         };
 121
 122         inet_get_ping_group_range_table(table, &low, &high);
 123         urange[0] = from_kgid_munged(user_ns, low);
 124         urange[1] = from_kgid_munged(user_ns, high);
 125         ret = proc_dointvec_minmax(&tmp, write, buffer, lenp, ppos);
 126
 127         if (write && ret == 0) {
 128                 low = make_kgid(user_ns, urange[0]);
 129                 high = make_kgid(user_ns, urange[1]);
 130                 if (!gid_valid(low) || !gid_valid(high) ||
 131                     (urange[1] < urange[0]) || gid_lt(high, low)) {
 132                         low = make_kgid(&init_user_ns, 1);
 133                         high = make_kgid(&init_user_ns, 0);
 134                 }
 135                 set_ping_group_range(table, low, high);
 136         }
 137
 138         return ret;
 139 }
 140
 141 static int proc_tcp_congestion_control(struct ctl_table *ctl, int write,
 142                                        void __user *buffer, size_t *lenp, loff_t *ppos)
 143 {
 144         char val[TCP_CA_NAME_MAX];
 145         struct ctl_table tbl = {
 146                 .data = val,
 147                 .maxlen = TCP_CA_NAME_MAX,
 148         };
 149         int ret;
 150
 151         tcp_get_default_congestion_control(val);
 152
 153         ret = proc_dostring(&tbl, write, buffer, lenp, ppos);
 154         if (write && ret == 0)
 155                 ret = tcp_set_default_congestion_control(val);
 156         return ret;
 157 }
 158
 159 static int proc_tcp_available_congestion_control(struct ctl_table *ctl,
 160                                                  int write,
 161                                                  void __user *buffer, size_t *lenp,
 162                                                  loff_t *ppos)
 163 {
 164         struct 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         tcp_get_available_congestion_control(tbl.data, TCP_CA_BUF_MAX);
 171         ret = proc_dostring(&tbl, write, buffer, lenp, ppos);
 172         kfree(tbl.data);
 173         return ret;
 174 }
 175
 176 static int proc_allowed_congestion_control(struct ctl_table *ctl,
 177                                            int write,
 178                                            void __user *buffer, size_t *lenp,
 179                                            loff_t *ppos)
 180 {
 181         struct ctl_table tbl = { .maxlen = TCP_CA_BUF_MAX };
 182         int ret;
 183
 184         tbl.data = kmalloc(tbl.maxlen, GFP_USER);
 185         if (!tbl.data)
 186                 return -ENOMEM;
 187
 188         tcp_get_allowed_congestion_control(tbl.data, tbl.maxlen);
 189         ret = proc_dostring(&tbl, write, buffer, lenp, ppos);
 190         if (write && ret == 0)
 191                 ret = tcp_set_allowed_congestion_control(tbl.data);
 192         kfree(tbl.data);
 193         return ret;
 194 }
 195
 196 static int ipv4_tcp_mem(struct ctl_table *ctl, int write,
 197                            void __user *buffer, size_t *lenp,
 198                            loff_t *ppos)
 199 {
 200         int ret;
 201         unsigned long vec[3];
 202         struct net *net = current->nsproxy->net_ns;
 203 #ifdef CONFIG_MEMCG_KMEM
 204         struct mem_cgroup *memcg;
 205 #endif
 206
 207         struct ctl_table tmp = {
 208                 .data = &vec,
 209                 .maxlen = sizeof(vec),
 210                 .mode = ctl->mode,
 211         };
 212
 213         if (!write) {
 214                 ctl->data = &net->ipv4.sysctl_tcp_mem;
 215                 return proc_doulongvec_minmax(ctl, write, buffer, lenp, ppos);
 216         }
 217
 218         ret = proc_doulongvec_minmax(&tmp, write, buffer, lenp, ppos);
 219         if (ret)
 220                 return ret;
 221
 222 #ifdef CONFIG_MEMCG_KMEM
 223         rcu_read_lock();
 224         memcg = mem_cgroup_from_task(current);
 225
 226         tcp_prot_mem(memcg, vec[0], 0);
 227         tcp_prot_mem(memcg, vec[1], 1);
 228         tcp_prot_mem(memcg, vec[2], 2);
 229         rcu_read_unlock();
 230 #endif
 231
 232         net->ipv4.sysctl_tcp_mem[0] = vec[0];
 233         net->ipv4.sysctl_tcp_mem[1] = vec[1];
 234         net->ipv4.sysctl_tcp_mem[2] = vec[2];
 235
 236         return 0;
 237 }
 238
 239 static int proc_tcp_fastopen_key(struct ctl_table *ctl, int write,
 240                                  void __user *buffer, size_t *lenp,
 241                                  loff_t *ppos)
 242 {
 243         struct ctl_table tbl = { .maxlen = (TCP_FASTOPEN_KEY_LENGTH * 2 + 10) };
 244         struct tcp_fastopen_context *ctxt;
 245         int ret;
 246         u32  user_key[4]; /* 16 bytes, matching TCP_FASTOPEN_KEY_LENGTH */
 247
 248         tbl.data = kmalloc(tbl.maxlen, GFP_KERNEL);
 249         if (!tbl.data)
 250                 return -ENOMEM;
 251
 252         rcu_read_lock();
 253         ctxt = rcu_dereference(tcp_fastopen_ctx);
 254         if (ctxt)
 255                 memcpy(user_key, ctxt->key, TCP_FASTOPEN_KEY_LENGTH);
 256         else
 257                 memset(user_key, 0, sizeof(user_key));
 258         rcu_read_unlock();
 259
 260         snprintf(tbl.data, tbl.maxlen, "%08x-%08x-%08x-%08x",
 261                 user_key[0], user_key[1], user_key[2], user_key[3]);
 262         ret = proc_dostring(&tbl, write, buffer, lenp, ppos);
 263
 264         if (write && ret == 0) {
 265                 if (sscanf(tbl.data, "%x-%x-%x-%x", user_key, user_key + 1,
 266                            user_key + 2, user_key + 3) != 4) {
 267                         ret = -EINVAL;
 268                         goto bad_key;
 269                 }
 270                 tcp_fastopen_reset_cipher(user_key, TCP_FASTOPEN_KEY_LENGTH);
 271         }
 272
 273 bad_key:
 274         pr_debug("proc FO key set 0x%x-%x-%x-%x <- 0x%s: %u\n",
 275                user_key[0], user_key[1], user_key[2], user_key[3],
 276                (char *)tbl.data, ret);
 277         kfree(tbl.data);
 278         return ret;
 279 }
 280
 281 static struct ctl_table ipv4_table[] = {
 282         {
 283                 .procname       = "tcp_timestamps",
 284                 .data           = &sysctl_tcp_timestamps,
 285                 .maxlen         = sizeof(int),
 286                 .mode           = 0644,
 287                 .proc_handler   = proc_dointvec
 288         },
 289         {
 290                 .procname       = "tcp_window_scaling",
 291                 .data           = &sysctl_tcp_window_scaling,
 292                 .maxlen         = sizeof(int),
 293                 .mode           = 0644,
 294                 .proc_handler   = proc_dointvec
 295         },
 296         {
 297                 .procname       = "tcp_sack",
 298                 .data           = &sysctl_tcp_sack,
 299                 .maxlen         = sizeof(int),
 300                 .mode           = 0644,
 301                 .proc_handler   = proc_dointvec
 302         },
 303         {
 304                 .procname       = "tcp_retrans_collapse",
 305                 .data           = &sysctl_tcp_retrans_collapse,
 306                 .maxlen         = sizeof(int),
 307                 .mode           = 0644,
 308                 .proc_handler   = proc_dointvec
 309         },
 310         {
 311                 .procname       = "ip_default_ttl",
 312                 .data           = &sysctl_ip_default_ttl,
 313                 .maxlen         = sizeof(int),
 314                 .mode           = 0644,
 315                 .proc_handler   = proc_dointvec_minmax,
 316                 .extra1         = &ip_ttl_min,
 317                 .extra2         = &ip_ttl_max,
 318         },
 319         {
 320                 .procname       = "ip_no_pmtu_disc",
 321                 .data           = &ipv4_config.no_pmtu_disc,
 322                 .maxlen         = sizeof(int),
 323                 .mode           = 0644,
 324                 .proc_handler   = proc_dointvec
 325         },
 326         {
 327                 .procname       = "ip_nonlocal_bind",
 328                 .data           = &sysctl_ip_nonlocal_bind,
 329                 .maxlen         = sizeof(int),
 330                 .mode           = 0644,
 331                 .proc_handler   = proc_dointvec
 332         },
 333         {
 334                 .procname       = "tcp_syn_retries",
 335                 .data           = &sysctl_tcp_syn_retries,
 336                 .maxlen         = sizeof(int),
 337                 .mode           = 0644,
 338                 .proc_handler   = proc_dointvec_minmax,
 339                 .extra1         = &tcp_syn_retries_min,
 340                 .extra2         = &tcp_syn_retries_max
 341         },
 342         {
 343                 .procname       = "tcp_synack_retries",
 344                 .data           = &sysctl_tcp_synack_retries,
 345                 .maxlen         = sizeof(int),
 346                 .mode           = 0644,
 347                 .proc_handler   = proc_dointvec
 348         },
 349         {
 350                 .procname       = "tcp_max_orphans",
 351                 .data           = &sysctl_tcp_max_orphans,
 352                 .maxlen         = sizeof(int),
 353                 .mode           = 0644,
 354                 .proc_handler   = proc_dointvec
 355         },
 356         {
 357                 .procname       = "tcp_max_tw_buckets",
 358                 .data           = &tcp_death_row.sysctl_max_tw_buckets,
 359                 .maxlen         = sizeof(int),
 360                 .mode           = 0644,
 361                 .proc_handler   = proc_dointvec
 362         },
 363         {
 364                 .procname       = "ip_early_demux",
 365                 .data           = &sysctl_ip_early_demux,
 366                 .maxlen         = sizeof(int),
 367                 .mode           = 0644,
 368                 .proc_handler   = proc_dointvec
 369         },
 370         {
 371                 .procname       = "ip_dynaddr",
 372                 .data           = &sysctl_ip_dynaddr,
 373                 .maxlen         = sizeof(int),
 374                 .mode           = 0644,
 375                 .proc_handler   = proc_dointvec
 376         },
 377         {
 378                 .procname       = "tcp_keepalive_time",
 379                 .data           = &sysctl_tcp_keepalive_time,
 380                 .maxlen         = sizeof(int),
 381                 .mode           = 0644,
 382                 .proc_handler   = proc_dointvec_jiffies,
 383         },
 384         {
 385                 .procname       = "tcp_keepalive_probes",
 386                 .data           = &sysctl_tcp_keepalive_probes,
 387                 .maxlen         = sizeof(int),
 388                 .mode           = 0644,
 389                 .proc_handler   = proc_dointvec
 390         },
 391         {
 392                 .procname       = "tcp_keepalive_intvl",
 393                 .data           = &sysctl_tcp_keepalive_intvl,
 394                 .maxlen         = sizeof(int),
 395                 .mode           = 0644,
 396                 .proc_handler   = proc_dointvec_jiffies,
 397         },
 398         {
 399                 .procname       = "tcp_retries1",
 400                 .data           = &sysctl_tcp_retries1,
 401                 .maxlen         = sizeof(int),
 402                 .mode           = 0644,
 403                 .proc_handler   = proc_dointvec_minmax,
 404                 .extra2         = &tcp_retr1_max
 405         },
 406         {
 407                 .procname       = "tcp_retries2",
 408                 .data           = &sysctl_tcp_retries2,
 409                 .maxlen         = sizeof(int),
 410                 .mode           = 0644,
 411                 .proc_handler   = proc_dointvec
 412         },
 413         {
 414                 .procname       = "tcp_fin_timeout",
 415                 .data           = &sysctl_tcp_fin_timeout,
 416                 .maxlen         = sizeof(int),
 417                 .mode           = 0644,
 418                 .proc_handler   = proc_dointvec_jiffies,
 419         },
 420 #ifdef CONFIG_SYN_COOKIES
 421         {
 422                 .procname       = "tcp_syncookies",
 423                 .data           = &sysctl_tcp_syncookies,
 424                 .maxlen         = sizeof(int),
 425                 .mode           = 0644,
 426                 .proc_handler   = proc_dointvec
 427         },
 428 #endif
 429         {
 430                 .procname       = "tcp_fastopen",
 431                 .data           = &sysctl_tcp_fastopen,
 432                 .maxlen         = sizeof(int),
 433                 .mode           = 0644,
 434                 .proc_handler   = proc_dointvec,
 435         },
 436         {
 437                 .procname       = "tcp_fastopen_key",
 438                 .mode           = 0600,
 439                 .maxlen         = ((TCP_FASTOPEN_KEY_LENGTH * 2) + 10),
 440                 .proc_handler   = proc_tcp_fastopen_key,
 441         },
 442         {
 443                 .procname       = "tcp_tw_recycle",
 444                 .data           = &tcp_death_row.sysctl_tw_recycle,
 445                 .maxlen         = sizeof(int),
 446                 .mode           = 0644,
 447                 .proc_handler   = proc_dointvec
 448         },
 449         {
 450                 .procname       = "tcp_abort_on_overflow",
 451                 .data           = &sysctl_tcp_abort_on_overflow,
 452                 .maxlen         = sizeof(int),
 453                 .mode           = 0644,
 454                 .proc_handler   = proc_dointvec
 455         },
 456         {
 457                 .procname       = "tcp_stdurg",
 458                 .data           = &sysctl_tcp_stdurg,
 459                 .maxlen         = sizeof(int),
 460                 .mode           = 0644,
 461                 .proc_handler   = proc_dointvec
 462         },
 463         {
 464                 .procname       = "tcp_rfc1337",
 465                 .data           = &sysctl_tcp_rfc1337,
 466                 .maxlen         = sizeof(int),
 467                 .mode           = 0644,
 468                 .proc_handler   = proc_dointvec
 469         },
 470         {
 471                 .procname       = "tcp_max_syn_backlog",
 472                 .data           = &sysctl_max_syn_backlog,
 473                 .maxlen         = sizeof(int),
 474                 .mode           = 0644,
 475                 .proc_handler   = proc_dointvec
 476         },
 477         {
 478                 .procname       = "ip_local_port_range",
 479                 .data           = &sysctl_local_ports.range,
 480                 .maxlen         = sizeof(sysctl_local_ports.range),
 481                 .mode           = 0644,
 482                 .proc_handler   = ipv4_local_port_range,
 483         },
 484         {
 485                 .procname       = "ip_local_reserved_ports",
 486                 .data           = NULL, /* initialized in sysctl_ipv4_init */
 487                 .maxlen         = 65536,
 488                 .mode           = 0644,
 489                 .proc_handler   = proc_do_large_bitmap,
 490         },
 491         {
 492                 .procname       = "igmp_max_memberships",
 493                 .data           = &sysctl_igmp_max_memberships,
 494                 .maxlen         = sizeof(int),
 495                 .mode           = 0644,
 496                 .proc_handler   = proc_dointvec
 497         },
 498         {
 499                 .procname       = "igmp_max_msf",
 500                 .data           = &sysctl_igmp_max_msf,
 501                 .maxlen         = sizeof(int),
 502                 .mode           = 0644,
 503                 .proc_handler   = proc_dointvec
 504         },
 505         {
 506                 .procname       = "inet_peer_threshold",
 507                 .data           = &inet_peer_threshold,
 508                 .maxlen         = sizeof(int),
 509                 .mode           = 0644,
 510                 .proc_handler   = proc_dointvec
 511         },
 512         {
 513                 .procname       = "inet_peer_minttl",
 514                 .data           = &inet_peer_minttl,
 515                 .maxlen         = sizeof(int),
 516                 .mode           = 0644,
 517                 .proc_handler   = proc_dointvec_jiffies,
 518         },
 519         {
 520                 .procname       = "inet_peer_maxttl",
 521                 .data           = &inet_peer_maxttl,
 522                 .maxlen         = sizeof(int),
 523                 .mode           = 0644,
 524                 .proc_handler   = proc_dointvec_jiffies,
 525         },
 526         {
 527                 .procname       = "tcp_orphan_retries",
 528                 .data           = &sysctl_tcp_orphan_retries,
 529                 .maxlen         = sizeof(int),
 530                 .mode           = 0644,
 531                 .proc_handler   = proc_dointvec
 532         },
 533         {
 534                 .procname       = "tcp_fack",
 535                 .data           = &sysctl_tcp_fack,
 536                 .maxlen         = sizeof(int),
 537                 .mode           = 0644,
 538                 .proc_handler   = proc_dointvec
 539         },
 540         {
 541                 .procname       = "tcp_reordering",
 542                 .data           = &sysctl_tcp_reordering,
 543                 .maxlen         = sizeof(int),
 544                 .mode           = 0644,
 545                 .proc_handler   = proc_dointvec
 546         },
 547         {
 548                 .procname       = "tcp_dsack",
 549                 .data           = &sysctl_tcp_dsack,
 550                 .maxlen         = sizeof(int),
 551                 .mode           = 0644,
 552                 .proc_handler   = proc_dointvec
 553         },
 554         {
 555                 .procname       = "tcp_wmem",
 556                 .data           = &sysctl_tcp_wmem,
 557                 .maxlen         = sizeof(sysctl_tcp_wmem),
 558                 .mode           = 0644,
 559                 .proc_handler   = proc_dointvec_minmax,
 560                 .extra1         = &one,
 561         },
 562         {
 563                 .procname       = "tcp_notsent_lowat",
 564                 .data           = &sysctl_tcp_notsent_lowat,
 565                 .maxlen         = sizeof(sysctl_tcp_notsent_lowat),
 566                 .mode           = 0644,
 567                 .proc_handler   = proc_dointvec,
 568         },
 569         {
 570                 .procname       = "tcp_rmem",
 571                 .data           = &sysctl_tcp_rmem,
 572                 .maxlen         = sizeof(sysctl_tcp_rmem),
 573                 .mode           = 0644,
 574                 .proc_handler   = proc_dointvec_minmax,
 575                 .extra1         = &one,
 576         },
 577         {
 578                 .procname       = "tcp_app_win",
 579                 .data           = &sysctl_tcp_app_win,
 580                 .maxlen         = sizeof(int),
 581                 .mode           = 0644,
 582                 .proc_handler   = proc_dointvec
 583         },
 584         {
 585                 .procname       = "tcp_adv_win_scale",
 586                 .data           = &sysctl_tcp_adv_win_scale,
 587                 .maxlen         = sizeof(int),
 588                 .mode           = 0644,
 589                 .proc_handler   = proc_dointvec_minmax,
 590                 .extra1         = &tcp_adv_win_scale_min,
 591                 .extra2         = &tcp_adv_win_scale_max,
 592         },
 593         {
 594                 .procname       = "tcp_tw_reuse",
 595                 .data           = &sysctl_tcp_tw_reuse,
 596                 .maxlen         = sizeof(int),
 597                 .mode           = 0644,
 598                 .proc_handler   = proc_dointvec
 599         },
 600         {
 601                 .procname       = "tcp_frto",
 602                 .data           = &sysctl_tcp_frto,
 603                 .maxlen         = sizeof(int),
 604                 .mode           = 0644,
 605                 .proc_handler   = proc_dointvec
 606         },
 607         {
 608                 .procname       = "tcp_low_latency",
 609                 .data           = &sysctl_tcp_low_latency,
 610                 .maxlen         = sizeof(int),
 611                 .mode           = 0644,
 612                 .proc_handler   = proc_dointvec
 613         },
 614         {
 615                 .procname       = "tcp_no_metrics_save",
 616                 .data           = &sysctl_tcp_nometrics_save,
 617                 .maxlen         = sizeof(int),
 618                 .mode           = 0644,
 619                 .proc_handler   = proc_dointvec,
 620         },
 621         {
 622                 .procname       = "tcp_moderate_rcvbuf",
 623                 .data           = &sysctl_tcp_moderate_rcvbuf,
 624                 .maxlen         = sizeof(int),
 625                 .mode           = 0644,
 626                 .proc_handler   = proc_dointvec,
 627         },
 628         {
 629                 .procname       = "tcp_tso_win_divisor",
 630                 .data           = &sysctl_tcp_tso_win_divisor,
 631                 .maxlen         = sizeof(int),
 632                 .mode           = 0644,
 633                 .proc_handler   = proc_dointvec,
 634         },
 635         {
 636                 .procname       = "tcp_congestion_control",
 637                 .mode           = 0644,
 638                 .maxlen         = TCP_CA_NAME_MAX,
 639                 .proc_handler   = proc_tcp_congestion_control,
 640         },
 641         {
 642                 .procname       = "tcp_mtu_probing",
 643                 .data           = &sysctl_tcp_mtu_probing,
 644                 .maxlen         = sizeof(int),
 645                 .mode           = 0644,
 646                 .proc_handler   = proc_dointvec,
 647         },
 648         {
 649                 .procname       = "tcp_base_mss",
 650                 .data           = &sysctl_tcp_base_mss,
 651                 .maxlen         = sizeof(int),
 652                 .mode           = 0644,
 653                 .proc_handler   = proc_dointvec,
 654         },
 655         {
 656                 .procname       = "tcp_workaround_signed_windows",
 657                 .data           = &sysctl_tcp_workaround_signed_windows,
 658                 .maxlen         = sizeof(int),
 659                 .mode           = 0644,
 660                 .proc_handler   = proc_dointvec
 661         },
 662         {
 663                 .procname       = "tcp_limit_output_bytes",
 664                 .data           = &sysctl_tcp_limit_output_bytes,
 665                 .maxlen         = sizeof(int),
 666                 .mode           = 0644,
 667                 .proc_handler   = proc_dointvec
 668         },
 669         {
 670                 .procname       = "tcp_challenge_ack_limit",
 671                 .data           = &sysctl_tcp_challenge_ack_limit,
 672                 .maxlen         = sizeof(int),
 673                 .mode           = 0644,
 674                 .proc_handler   = proc_dointvec
 675         },
 676 #ifdef CONFIG_NET_DMA
 677         {
 678                 .procname       = "tcp_dma_copybreak",
 679                 .data           = &sysctl_tcp_dma_copybreak,
 680                 .maxlen         = sizeof(int),
 681                 .mode           = 0644,
 682                 .proc_handler   = proc_dointvec
 683         },
 684 #endif
 685         {
 686                 .procname       = "tcp_slow_start_after_idle",
 687                 .data           = &sysctl_tcp_slow_start_after_idle,
 688                 .maxlen         = sizeof(int),
 689                 .mode           = 0644,
 690                 .proc_handler   = proc_dointvec
 691         },
 692 #ifdef CONFIG_NETLABEL
 693         {
 694                 .procname       = "cipso_cache_enable",
 695                 .data           = &cipso_v4_cache_enabled,
 696                 .maxlen         = sizeof(int),
 697                 .mode           = 0644,
 698                 .proc_handler   = proc_dointvec,
 699         },
 700         {
 701                 .procname       = "cipso_cache_bucket_size",
 702                 .data           = &cipso_v4_cache_bucketsize,
 703                 .maxlen         = sizeof(int),
 704                 .mode           = 0644,
 705                 .proc_handler   = proc_dointvec,
 706         },
 707         {
 708                 .procname       = "cipso_rbm_optfmt",
 709                 .data           = &cipso_v4_rbm_optfmt,
 710                 .maxlen         = sizeof(int),
 711                 .mode           = 0644,
 712                 .proc_handler   = proc_dointvec,
 713         },
 714         {
 715                 .procname       = "cipso_rbm_strictvalid",
 716                 .data           = &cipso_v4_rbm_strictvalid,
 717                 .maxlen         = sizeof(int),
 718                 .mode           = 0644,
 719                 .proc_handler   = proc_dointvec,
 720         },
 721 #endif /* CONFIG_NETLABEL */
 722         {
 723                 .procname       = "tcp_available_congestion_control",
 724                 .maxlen         = TCP_CA_BUF_MAX,
 725                 .mode           = 0444,
 726                 .proc_handler   = proc_tcp_available_congestion_control,
 727         },
 728         {
 729                 .procname       = "tcp_allowed_congestion_control",
 730                 .maxlen         = TCP_CA_BUF_MAX,
 731                 .mode           = 0644,
 732                 .proc_handler   = proc_allowed_congestion_control,
 733         },
 734         {
 735                 .procname       = "tcp_max_ssthresh",
 736                 .data           = &sysctl_tcp_max_ssthresh,
 737                 .maxlen         = sizeof(int),
 738                 .mode           = 0644,
 739                 .proc_handler   = proc_dointvec,
 740         },
 741         {
 742                 .procname       = "tcp_thin_linear_timeouts",
 743                 .data           = &sysctl_tcp_thin_linear_timeouts,
 744                 .maxlen         = sizeof(int),
 745                 .mode           = 0644,
 746                 .proc_handler   = proc_dointvec
 747         },
 748         {
 749                 .procname       = "tcp_thin_dupack",
 750                 .data           = &sysctl_tcp_thin_dupack,
 751                 .maxlen         = sizeof(int),
 752                 .mode           = 0644,
 753                 .proc_handler   = proc_dointvec
 754         },
 755         {
 756                 .procname       = "tcp_early_retrans",
 757                 .data           = &sysctl_tcp_early_retrans,
 758                 .maxlen         = sizeof(int),
 759                 .mode           = 0644,
 760                 .proc_handler   = proc_dointvec_minmax,
 761                 .extra1         = &zero,
 762                 .extra2         = &four,
 763         },
 764         {
 765                 .procname       = "tcp_min_tso_segs",
 766                 .data           = &sysctl_tcp_min_tso_segs,
 767                 .maxlen         = sizeof(int),
 768                 .mode           = 0644,
 769                 .proc_handler   = proc_dointvec_minmax,
 770                 .extra1         = &zero,
 771                 .extra2         = &gso_max_segs,
 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 };
 864
 865 static __net_init int ipv4_sysctl_init_net(struct net *net)
 866 {
 867         struct ctl_table *table;
 868
 869         table = ipv4_net_table;
 870         if (!net_eq(net, &init_net)) {
 871                 table = kmemdup(table, sizeof(ipv4_net_table), GFP_KERNEL);
 872                 if (table == NULL)
 873                         goto err_alloc;
 874
 875                 table[0].data =
 876                         &net->ipv4.sysctl_icmp_echo_ignore_all;
 877                 table[1].data =
 878                         &net->ipv4.sysctl_icmp_echo_ignore_broadcasts;
 879                 table[2].data =
 880                         &net->ipv4.sysctl_icmp_ignore_bogus_error_responses;
 881                 table[3].data =
 882                         &net->ipv4.sysctl_icmp_errors_use_inbound_ifaddr;
 883                 table[4].data =
 884                         &net->ipv4.sysctl_icmp_ratelimit;
 885                 table[5].data =
 886                         &net->ipv4.sysctl_icmp_ratemask;
 887                 table[6].data =
 888                         &net->ipv4.sysctl_ping_group_range;
 889                 table[7].data =
 890                         &net->ipv4.sysctl_tcp_ecn;
 891
 892                 /* Don't export sysctls to unprivileged users */
 893                 if (net->user_ns != &init_user_ns)
 894                         table[0].procname = NULL;
 895         }
 896
 897         /*
 898          * Sane defaults - nobody may create ping sockets.
 899          * Boot scripts should set this to distro-specific group.
 900          */
 901         net->ipv4.sysctl_ping_group_range[0] = make_kgid(&init_user_ns, 1);
 902         net->ipv4.sysctl_ping_group_range[1] = make_kgid(&init_user_ns, 0);
 903
 904         tcp_init_mem(net);
 905
 906         net->ipv4.ipv4_hdr = register_net_sysctl(net, "net/ipv4", table);
 907         if (net->ipv4.ipv4_hdr == NULL)
 908                 goto err_reg;
 909
 910         return 0;
 911
 912 err_reg:
 913         if (!net_eq(net, &init_net))
 914                 kfree(table);
 915 err_alloc:
 916         return -ENOMEM;
 917 }
 918
 919 static __net_exit void ipv4_sysctl_exit_net(struct net *net)
 920 {
 921         struct ctl_table *table;
 922
 923         table = net->ipv4.ipv4_hdr->ctl_table_arg;
 924         unregister_net_sysctl_table(net->ipv4.ipv4_hdr);
 925         kfree(table);
 926 }
 927
 928 static __net_initdata struct pernet_operations ipv4_sysctl_ops = {
 929         .init = ipv4_sysctl_init_net,
 930         .exit = ipv4_sysctl_exit_net,
 931 };
 932
 933 static __init int sysctl_ipv4_init(void)
 934 {
 935         struct ctl_table_header *hdr;
 936         struct ctl_table *i;
 937
 938         for (i = ipv4_table; i->procname; i++) {
 939                 if (strcmp(i->procname, "ip_local_reserved_ports") == 0) {
 940                         i->data = sysctl_local_reserved_ports;
 941                         break;
 942                 }
 943         }
 944         if (!i->procname)
 945                 return -EINVAL;
 946
 947         hdr = register_net_sysctl(&init_net, "net/ipv4", ipv4_table);
 948         if (hdr == NULL)
 949                 return -ENOMEM;
 950
 951         if (register_pernet_subsys(&ipv4_sysctl_ops)) {
 952                 unregister_net_sysctl_table(hdr);
 953                 return -ENOMEM;
 954         }
 955
 956         return 0;
 957 }
 958
 959 __initcall(sysctl_ipv4_init);